1
0
mirror of git://sourceware.org/git/lvm2.git synced 2024-12-22 17:35:59 +03:00
lvm2/lib/config/config.c

2008 lines
54 KiB
C
Raw Normal View History

/*
* Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
2011-05-24 17:53:26 +04:00
* Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
*
2004-03-30 23:35:44 +04:00
* This file is part of LVM2.
*
* This copyrighted material is made available to anyone wishing to use,
* modify, copy, or redistribute it subject to the terms and conditions
* of the GNU Lesser General Public License v.2.1.
2004-03-30 23:35:44 +04:00
*
* You should have received a copy of the GNU Lesser General Public License
2004-03-30 23:35:44 +04:00
* along with this program; if not, write to the Free Software Foundation,
* Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
2002-11-18 17:01:16 +03:00
#include "lib.h"
2002-11-18 17:01:16 +03:00
#include "config.h"
#include "crc.h"
#include "device.h"
2004-05-04 22:28:15 +04:00
#include "str_list.h"
#include "toolcontext.h"
#include "lvm-file.h"
#include "memlock.h"
2002-11-18 17:01:16 +03:00
#include <sys/stat.h>
#include <sys/mman.h>
#include <unistd.h>
#include <fcntl.h>
#include <assert.h>
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
#include <ctype.h>
#include <math.h>
#include <float.h>
static const char *_config_source_names[] = {
[CONFIG_UNDEFINED] = "undefined",
[CONFIG_FILE] = "file",
[CONFIG_MERGED_FILES] = "merged files",
[CONFIG_STRING] = "string",
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
[CONFIG_PROFILE_COMMAND] = "command profile",
[CONFIG_PROFILE_METADATA] = "metadata profile",
[CONFIG_FILE_SPECIAL] = "special purpose"
};
struct config_file {
off_t st_size;
char *filename;
int exists;
int keep_open;
struct device *dev;
};
struct config_source {
config_source_t type;
time_t timestamp;
union {
struct config_file *file;
struct config_file *profile;
} source;
struct cft_check_handle *check_handle;
};
/*
* Map each ID to respective definition of the configuration item.
*/
static struct cfg_def_item _cfg_def_items[CFG_COUNT + 1] = {
#define cfg_section(id, name, parent, flags, since_version, comment) {id, parent, name, CFG_TYPE_SECTION, {0}, flags, since_version, comment},
#define cfg(id, name, parent, flags, type, default_value, since_version, comment) {id, parent, name, type, {.v_##type = default_value}, flags, since_version, comment},
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
#define cfg_runtime(id, name, parent, flags, type, since_version, comment) {id, parent, name, type, {.fn_##type = get_default_##id}, flags | CFG_DEFAULT_RUN_TIME, since_version, comment},
#define cfg_array(id, name, parent, flags, types, default_value, since_version, comment) {id, parent, name, CFG_TYPE_ARRAY | types, {.v_CFG_TYPE_STRING = default_value}, flags, since_version, comment},
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
#define cfg_array_runtime(id, name, parent, flags, types, since_version, comment) {id, parent, name, CFG_TYPE_ARRAY | types, {.fn_CFG_TYPE_STRING = get_default_##id}, flags | CFG_DEFAULT_RUN_TIME, since_version, comment},
#include "config_settings.h"
#undef cfg_section
#undef cfg
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
#undef cfg_runtime
#undef cfg_array
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
#undef cfg_array_runtime
};
config_source_t config_get_source_type(struct dm_config_tree *cft)
{
struct config_source *cs = dm_config_get_custom(cft);
return cs ? cs->type : CONFIG_UNDEFINED;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
static inline int _is_profile_based_config_source(config_source_t source)
{
return (source == CONFIG_PROFILE_COMMAND) ||
(source == CONFIG_PROFILE_METADATA);
}
static inline int _is_file_based_config_source(config_source_t source)
{
return (source == CONFIG_FILE) ||
(source == CONFIG_FILE_SPECIAL) ||
_is_profile_based_config_source(source);
}
/*
* public interface
*/
struct dm_config_tree *config_open(config_source_t source,
const char *filename,
int keep_open)
2002-01-07 13:23:52 +03:00
{
struct dm_config_tree *cft = dm_config_create();
struct config_source *cs;
struct config_file *cf;
if (!cft)
return NULL;
if (!(cs = dm_pool_zalloc(cft->mem, sizeof(struct config_source)))) {
log_error("Failed to allocate config source.");
goto fail;
}
2002-01-07 13:23:52 +03:00
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (_is_file_based_config_source(source)) {
if (!(cf = dm_pool_zalloc(cft->mem, sizeof(struct config_file)))) {
log_error("Failed to allocate config file.");
goto fail;
}
cf->keep_open = keep_open;
if (filename &&
!(cf->filename = dm_pool_strdup(cft->mem, filename))) {
log_error("Failed to duplicate filename.");
goto fail;
}
cs->source.file = cf;
}
cs->type = source;
dm_config_set_custom(cft, cs);
return cft;
fail:
dm_config_destroy(cft);
return NULL;
}
/*
* Doesn't populate filename if the file is empty.
*/
int config_file_check(struct dm_config_tree *cft, const char **filename, struct stat *info)
{
struct config_source *cs = dm_config_get_custom(cft);
struct config_file *cf;
struct stat _info;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (!_is_file_based_config_source(cs->type)) {
log_error(INTERNAL_ERROR "config_file_check: expected file, special file or "
"profile config source, found %s config source.",
_config_source_names[cs->type]);
return 0;
}
if (!info)
info = &_info;
cf = cs->source.file;
if (stat(cf->filename, info)) {
log_sys_error("stat", cf->filename);
cf->exists = 0;
return 0;
}
if (!S_ISREG(info->st_mode)) {
log_error("%s is not a regular file", cf->filename);
cf->exists = 0;
return 0;
}
cs->timestamp = info->st_ctime;
cf->exists = 1;
cf->st_size = info->st_size;
if (info->st_size == 0)
log_verbose("%s is empty", cf->filename);
else if (filename)
*filename = cf->filename;
return 1;
}
/*
* Return 1 if config files ought to be reloaded
*/
int config_file_changed(struct dm_config_tree *cft)
{
struct config_source *cs = dm_config_get_custom(cft);
struct config_file *cf;
struct stat info;
if (cs->type != CONFIG_FILE) {
log_error(INTERNAL_ERROR "config_file_changed: expected file config source, "
"found %s config source.", _config_source_names[cs->type]);
return 0;
}
cf = cs->source.file;
if (!cf->filename)
return 0;
if (stat(cf->filename, &info) == -1) {
/* Ignore a deleted config file: still use original data */
if (errno == ENOENT) {
if (!cf->exists)
return 0;
log_very_verbose("Config file %s has disappeared!",
cf->filename);
goto reload;
}
log_sys_error("stat", cf->filename);
log_error("Failed to reload configuration files");
return 0;
}
if (!S_ISREG(info.st_mode)) {
log_error("Configuration file %s is not a regular file",
cf->filename);
goto reload;
}
/* Unchanged? */
if (cs->timestamp == info.st_ctime && cf->st_size == info.st_size)
return 0;
reload:
log_verbose("Detected config file change to %s", cf->filename);
return 1;
}
void config_destroy(struct dm_config_tree *cft)
{
struct config_source *cs;
struct config_file *cf;
if (!cft)
return;
cs = dm_config_get_custom(cft);
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (_is_file_based_config_source(cs->type)) {
cf = cs->source.file;
if (cf && cf->dev)
if (!dev_close(cf->dev))
stack;
}
2002-01-07 13:23:52 +03:00
dm_config_destroy(cft);
}
struct dm_config_tree *config_file_open_and_read(const char *config_file,
config_source_t source,
struct cmd_context *cmd)
{
struct dm_config_tree *cft;
struct stat info;
if (!(cft = config_open(source, config_file, 0))) {
log_error("config_tree allocation failed");
return NULL;
}
/* Is there a config file? */
if (stat(config_file, &info) == -1) {
/* Profile file must be present! */
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (errno == ENOENT && (!_is_profile_based_config_source(source)))
return cft;
log_sys_error("stat", config_file);
goto bad;
}
log_very_verbose("Loading config file: %s", config_file);
if (!config_file_read(cft)) {
log_error("Failed to load config file %s", config_file);
goto bad;
}
return cft;
bad:
config_destroy(cft);
return NULL;
}
struct dm_config_tree *get_config_tree_by_source(struct cmd_context *cmd,
config_source_t source)
{
struct dm_config_tree *cft = cmd->cft;
struct config_source *cs;
while (cft) {
cs = dm_config_get_custom(cft);
if (cs && cs->type == source)
return cft;
cft = cft->cascade;
}
return NULL;
}
/*
* Returns config tree if it was removed.
*/
struct dm_config_tree *remove_config_tree_by_source(struct cmd_context *cmd,
config_source_t source)
{
struct dm_config_tree *previous_cft = NULL;
struct dm_config_tree *cft = cmd->cft;
struct config_source *cs;
while (cft) {
cs = dm_config_get_custom(cft);
if (cs && (cs->type == source)) {
if (previous_cft) {
previous_cft->cascade = cft->cascade;
cmd->cft = previous_cft;
} else
cmd->cft = cft->cascade;
cft->cascade = NULL;
break;
}
previous_cft = cft;
cft = cft->cascade;
}
return cft;
}
struct cft_check_handle *get_config_tree_check_handle(struct cmd_context *cmd,
struct dm_config_tree *cft)
{
struct config_source *cs;
if (!(cs = dm_config_get_custom(cft)))
return NULL;
if (cs->check_handle)
goto out;
/*
* Attach config check handle to all config types but
* CONFIG_FILE_SPECIAL - this one uses its own check
* methods and the cft_check_handle is not applicable here.
*/
if (cs->type != CONFIG_FILE_SPECIAL) {
if (!(cs->check_handle = dm_pool_zalloc(cft->mem, sizeof(*cs->check_handle)))) {
log_error("Failed to allocate configuration check handle.");
return NULL;
}
cs->check_handle->cft = cft;
cs->check_handle->cmd = cmd;
}
out:
return cs->check_handle;
}
int override_config_tree_from_string(struct cmd_context *cmd,
const char *config_settings)
{
struct dm_config_tree *cft_new;
struct config_source *cs = dm_config_get_custom(cmd->cft);
/*
* Follow this sequence:
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
* CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES
*/
if (cs->type == CONFIG_STRING) {
log_error(INTERNAL_ERROR "override_config_tree_from_string: "
"config cascade already contains a string config.");
return 0;
}
if (!(cft_new = dm_config_from_string(config_settings))) {
log_error("Failed to set overridden configuration entries.");
return 0;
}
if (!(cs = dm_pool_zalloc(cft_new->mem, sizeof(struct config_source)))) {
log_error("Failed to allocate config source.");
dm_config_destroy(cft_new);
return 0;
}
cs->type = CONFIG_STRING;
dm_config_set_custom(cft_new, cs);
cmd->cft = dm_config_insert_cascaded_tree(cft_new, cmd->cft);
return 1;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
static int _override_config_tree_from_command_profile(struct cmd_context *cmd,
struct profile *profile)
{
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
struct dm_config_tree *cft = cmd->cft, *cft_previous = NULL;
struct config_source *cs = dm_config_get_custom(cft);
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (cs->type == CONFIG_STRING) {
cft_previous = cft;
cft = cft->cascade;
cs = dm_config_get_custom(cft);
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (cs->type == CONFIG_PROFILE_COMMAND) {
log_error(INTERNAL_ERROR "_override_config_tree_from_command_profile: "
"config cascade already contains a command profile config.");
return 0;
}
if (cft_previous)
dm_config_insert_cascaded_tree(cft_previous, profile->cft);
else
cmd->cft = profile->cft;
dm_config_insert_cascaded_tree(profile->cft, cft);
return 1;
}
static int _override_config_tree_from_metadata_profile(struct cmd_context *cmd,
struct profile *profile)
{
struct dm_config_tree *cft = cmd->cft, *cft_previous = NULL;
struct config_source *cs = dm_config_get_custom(cft);
if (cs->type == CONFIG_STRING) {
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
cft_previous = cft;
cft = cft->cascade;
}
if (cs->type == CONFIG_PROFILE_COMMAND) {
cft_previous = cft;
cft = cft->cascade;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
cs = dm_config_get_custom(cft);
if (cs->type == CONFIG_PROFILE_METADATA) {
log_error(INTERNAL_ERROR "_override_config_tree_from_metadata_profile: "
"config cascade already contains a metadata profile config.");
return 0;
}
if (cft_previous)
dm_config_insert_cascaded_tree(cft_previous, profile->cft);
else
cmd->cft = profile->cft;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
dm_config_insert_cascaded_tree(profile->cft, cft);
return 1;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
int override_config_tree_from_profile(struct cmd_context *cmd,
struct profile *profile)
{
/*
* Follow this sequence:
* CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES
*/
if (!profile->cft && !load_profile(cmd, profile))
return_0;
if (profile->source == CONFIG_PROFILE_COMMAND)
return _override_config_tree_from_command_profile(cmd, profile);
else if (profile->source == CONFIG_PROFILE_METADATA)
return _override_config_tree_from_metadata_profile(cmd, profile);
log_error(INTERNAL_ERROR "override_config_tree_from_profile: incorrect profile source type");
return 0;
}
int config_file_read_fd(struct dm_config_tree *cft, struct device *dev,
off_t offset, size_t size, off_t offset2, size_t size2,
checksum_fn_t checksum_fn, uint32_t checksum)
{
char *fb, *fe;
2002-11-18 17:01:16 +03:00
int r = 0;
int use_mmap = 1;
off_t mmap_offset = 0;
char *buf = NULL;
struct config_source *cs = dm_config_get_custom(cft);
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (!_is_file_based_config_source(cs->type)) {
log_error(INTERNAL_ERROR "config_file_read_fd: expected file, special file "
"or profile config source, found %s config source.",
_config_source_names[cs->type]);
return 0;
}
/* Only use mmap with regular files */
if (!(dev->flags & DEV_REGULAR) || size2)
use_mmap = 0;
if (use_mmap) {
2006-08-17 22:23:44 +04:00
mmap_offset = offset % lvm_getpagesize();
2002-11-18 17:01:16 +03:00
/* memory map the file */
fb = mmap((caddr_t) 0, size + mmap_offset, PROT_READ,
MAP_PRIVATE, dev_fd(dev), offset - mmap_offset);
if (fb == (caddr_t) (-1)) {
log_sys_error("mmap", dev_name(dev));
2002-11-18 17:01:16 +03:00
goto out;
}
fb = fb + mmap_offset;
} else {
if (!(buf = dm_malloc(size + size2))) {
log_error("Failed to allocate circular buffer.");
return 0;
}
if (!dev_read_circular(dev, (uint64_t) offset, size,
(uint64_t) offset2, size2, buf)) {
goto out;
}
fb = buf;
2002-11-18 17:01:16 +03:00
}
if (checksum_fn && checksum !=
(checksum_fn(checksum_fn(INITIAL_CRC, (const uint8_t *)fb, size),
(const uint8_t *)(fb + size), size2))) {
log_error("%s: Checksum error", dev_name(dev));
2002-11-18 17:01:16 +03:00
goto out;
}
fe = fb + size + size2;
if (!dm_config_parse(cft, fb, fe))
2008-01-30 16:19:47 +03:00
goto_out;
2002-11-18 17:01:16 +03:00
r = 1;
out:
if (!use_mmap)
dm_free(buf);
2002-11-18 17:01:16 +03:00
else {
/* unmap the file */
if (munmap(fb - mmap_offset, size + mmap_offset)) {
log_sys_error("munmap", dev_name(dev));
2002-11-18 17:01:16 +03:00
r = 0;
}
}
return r;
}
int config_file_read(struct dm_config_tree *cft)
2002-11-18 17:01:16 +03:00
{
const char *filename = NULL;
struct config_source *cs = dm_config_get_custom(cft);
struct config_file *cf;
2002-11-18 17:01:16 +03:00
struct stat info;
int r;
if (!config_file_check(cft, &filename, &info))
2011-09-02 01:04:14 +04:00
return_0;
2004-05-04 22:28:15 +04:00
/* Nothing to do. E.g. empty file. */
if (!filename)
return 1;
cf = cs->source.file;
if (!cf->dev) {
if (!(cf->dev = dev_create_file(filename, NULL, NULL, 1)))
return_0;
if (!dev_open_readonly_buffered(cf->dev))
return_0;
}
r = config_file_read_fd(cft, cf->dev, 0, (size_t) info.st_size, 0, 0,
(checksum_fn_t) NULL, 0);
2002-11-18 17:01:16 +03:00
if (!cf->keep_open) {
if (!dev_close(cf->dev))
stack;
cf->dev = NULL;
}
2002-11-18 17:01:16 +03:00
return r;
}
time_t config_file_timestamp(struct dm_config_tree *cft)
{
struct config_source *cs = dm_config_get_custom(cft);
return cs->timestamp;
}
#define cfg_def_get_item_p(id) (&_cfg_def_items[id])
#define cfg_def_get_default_value_hint(cmd,item,type,profile) ((item->flags & CFG_DEFAULT_RUN_TIME) ? item->default_value.fn_##type(cmd,profile) : item->default_value.v_##type)
#define cfg_def_get_default_value(cmd,item,type,profile) (item->flags & CFG_DEFAULT_UNDEFINED ? 0 : cfg_def_get_default_value_hint(cmd,item,type,profile))
static int _cfg_def_make_path(char *buf, size_t buf_size, int id, cfg_def_item_t *item, int xlate)
{
int variable = item->flags & CFG_NAME_VARIABLE;
int parent_id = item->parent;
int count, n;
if (id == parent_id) {
buf[0] = '\0';
return 0;
}
count = _cfg_def_make_path(buf, buf_size, parent_id, cfg_def_get_item_p(parent_id), xlate);
if ((n = dm_snprintf(buf + count, buf_size - count, "%s%s%s%s",
count ? "/" : "",
xlate && variable ? "<" : "",
!xlate && variable ? "#" : item->name,
xlate && variable ? ">" : "")) < 0) {
log_error(INTERNAL_ERROR "_cfg_def_make_path: supplied buffer too small for %s/%s",
cfg_def_get_item_p(parent_id)->name, item->name);
buf[0] = '\0';
return 0;
}
return count + n;
}
int config_def_get_path(char *buf, size_t buf_size, int id)
{
return _cfg_def_make_path(buf, buf_size, id, cfg_def_get_item_p(id), 0);
}
static void _get_type_name(char *buf, size_t buf_size, cfg_def_type_t type)
{
(void) dm_snprintf(buf, buf_size, "%s%s%s%s%s%s",
(type & CFG_TYPE_ARRAY) ?
((type & ~CFG_TYPE_ARRAY) ?
" array with values of type:" : " array") : "",
(type & CFG_TYPE_SECTION) ? " section" : "",
(type & CFG_TYPE_BOOL) ? " boolean" : "",
(type & CFG_TYPE_INT) ? " integer" : "",
(type & CFG_TYPE_FLOAT) ? " float" : "",
(type & CFG_TYPE_STRING) ? " string" : "");
}
static void _log_type_error(const char *path, cfg_def_type_t actual,
cfg_def_type_t expected, int suppress_messages)
{
static char actual_type_name[128];
static char expected_type_name[128];
_get_type_name(actual_type_name, sizeof(actual_type_name), actual);
_get_type_name(expected_type_name, sizeof(expected_type_name), expected);
log_warn_suppress(suppress_messages, "Configuration setting \"%s\" has invalid type. "
"Found%s, expected%s.", path,
actual_type_name, expected_type_name);
}
static struct dm_config_value *_get_def_array_values(struct dm_config_tree *cft,
const cfg_def_item_t *def)
{
char *enc_value, *token, *p, *r;
struct dm_config_value *array = NULL, *v = NULL, *oldv = NULL;
if (!def->default_value.v_CFG_TYPE_STRING) {
if (!(array = dm_config_create_value(cft))) {
log_error("Failed to create default empty array for %s.", def->name);
return NULL;
}
array->type = DM_CFG_EMPTY_ARRAY;
return array;
}
if (!(p = token = enc_value = dm_strdup(def->default_value.v_CFG_TYPE_STRING))) {
log_error("_get_def_array_values: dm_strdup failed");
return NULL;
}
/* Proper value always starts with '#'. */
if (token[0] != '#')
goto bad;
while (token) {
/* Move to type identifier. Error on no char. */
token++;
if (!token[0])
goto bad;
/* Move to the actual value and decode any "##" into "#". */
p = token + 1;
while ((p = strchr(p, '#')) && p[1] == '#') {
memmove(p, p + 1, strlen(p));
p++;
}
/* Separate the value out of the whole string. */
if (p)
p[0] = '\0';
if (!(v = dm_config_create_value(cft))) {
log_error("Failed to create default config array value for %s.", def->name);
dm_free(enc_value);
return NULL;
}
if (oldv)
oldv->next = v;
if (!array)
array = v;
switch (toupper(token[0])) {
case 'I':
case 'B':
v->v.i = strtoll(token + 1, &r, 10);
if (*r)
goto bad;
v->type = DM_CFG_INT;
break;
case 'F':
v->v.f = strtod(token + 1, &r);
if (*r)
goto bad;
v->type = DM_CFG_FLOAT;
break;
case 'S':
if (!(r = dm_pool_strdup(cft->mem, token + 1))) {
dm_free(enc_value);
log_error("Failed to duplicate token for default "
"array value of %s.", def->name);
return NULL;
}
v->v.str = r;
v->type = DM_CFG_STRING;
break;
default:
goto bad;
}
oldv = v;
token = p;
}
dm_free(enc_value);
return array;
bad:
log_error(INTERNAL_ERROR "Default array value malformed for \"%s\", "
"value: \"%s\", token: \"%s\".", def->name,
def->default_value.v_CFG_TYPE_STRING, token);
dm_free(enc_value);
return NULL;
}
static int _config_def_check_node_single_value(struct cft_check_handle *handle,
const char *rp, const struct dm_config_value *v,
const cfg_def_item_t *def)
{
/* Check empty array first if present. */
if (v->type == DM_CFG_EMPTY_ARRAY) {
if (!(def->type & CFG_TYPE_ARRAY)) {
_log_type_error(rp, CFG_TYPE_ARRAY, def->type, handle->suppress_messages);
return 0;
}
if (!(def->flags & CFG_ALLOW_EMPTY)) {
log_warn_suppress(handle->suppress_messages,
"Configuration setting \"%s\" invalid. Empty value not allowed.", rp);
return 0;
}
return 1;
}
switch (v->type) {
case DM_CFG_INT:
if (!(def->type & CFG_TYPE_INT) && !(def->type & CFG_TYPE_BOOL)) {
_log_type_error(rp, CFG_TYPE_INT, def->type, handle->suppress_messages);
return 0;
}
break;
case DM_CFG_FLOAT:
if (!(def->type & CFG_TYPE_FLOAT)) {
_log_type_error(rp, CFG_TYPE_FLOAT, def->type, handle-> suppress_messages);
return 0;
}
break;
case DM_CFG_STRING:
if (def->type & CFG_TYPE_BOOL) {
if (!dm_config_value_is_bool(v)) {
log_warn_suppress(handle->suppress_messages,
"Configuration setting \"%s\" invalid. "
"Found string value \"%s\", "
"expected boolean value: 0/1, \"y/n\", "
"\"yes/no\", \"on/off\", "
"\"true/false\".", rp, v->v.str);
return 0;
}
} else if (!(def->type & CFG_TYPE_STRING)) {
_log_type_error(rp, CFG_TYPE_STRING, def->type, handle->suppress_messages);
return 0;
}
break;
default: ;
}
return 1;
}
static int _check_value_differs_from_default(struct cft_check_handle *handle,
const struct dm_config_value *v,
const cfg_def_item_t *def,
struct dm_config_value *v_def)
{
struct dm_config_value *v_def_array, *v_def_iter;
int diff = 0, id;
int64_t i;
float f;
const char *str;
/* if default value is undefined, the value used differs from default */
if (def->flags & CFG_DEFAULT_UNDEFINED) {
diff = 1;
goto out;
}
if (!v_def && (def->type & CFG_TYPE_ARRAY)) {
if (!(v_def_array = v_def_iter = _get_def_array_values(handle->cft, def)))
return_0;
do {
/* iterate over each element of the array and check its value */
if ((v->type != v_def_iter->type) ||
_check_value_differs_from_default(handle, v, def, v_def_iter))
break;
v_def_iter = v_def_iter->next;
v = v->next;
} while (v_def_iter && v);
diff = v || v_def_iter;
dm_pool_free(handle->cft->mem, v_def_array);
} else {
switch (v->type) {
case DM_CFG_INT:
/* int value can be a real int but it can also represent bool */
i = v_def ? v_def->v.i
: def->type & CFG_TYPE_BOOL ?
cfg_def_get_default_value(handle->cmd, def, CFG_TYPE_BOOL, NULL) :
cfg_def_get_default_value(handle->cmd, def, CFG_TYPE_INT, NULL);
diff = i != v->v.i;
break;
case DM_CFG_FLOAT:
f = v_def ? v_def->v.f
: cfg_def_get_default_value(handle->cmd, def, CFG_TYPE_FLOAT, NULL);
diff = fabs(f - v->v.f) < FLT_EPSILON;
break;
case DM_CFG_STRING:
/* string value can be a real string but it can also represent bool */
if (v_def ? v_def->type == DM_CFG_INT : def->type == CFG_TYPE_BOOL) {
i = v_def ? v_def->v.i
: cfg_def_get_default_value(handle->cmd, def, CFG_TYPE_BOOL, NULL);
diff = i != v->v.i;
} else {
str = v_def ? v_def->v.str
: cfg_def_get_default_value(handle->cmd, def, CFG_TYPE_STRING, NULL);
diff = strcmp(str, v->v.str);
}
break;
case DM_CFG_EMPTY_ARRAY:
diff = v_def->type != DM_CFG_EMPTY_ARRAY;
break;
default:
log_error(INTERNAL_ERROR "inconsistent state reached in _check_value_differs_from_default");
return 0;
}
}
out:
if (diff) {
/* mark whole path from bottom to top with CFG_DIFF */
for (id = def->id; id && !(handle->status[id] & CFG_DIFF); id = _cfg_def_items[id].parent)
handle->status[id] |= CFG_DIFF;
}
return diff;
}
static int _config_def_check_node_value(struct cft_check_handle *handle,
const char *rp, const struct dm_config_value *v,
const cfg_def_item_t *def)
{
const struct dm_config_value *v_iter;
if (!v) {
if (def->type != CFG_TYPE_SECTION) {
_log_type_error(rp, CFG_TYPE_SECTION, def->type, handle->suppress_messages);
return 0;
}
return 1;
}
if (v->next) {
if (!(def->type & CFG_TYPE_ARRAY)) {
_log_type_error(rp, CFG_TYPE_ARRAY, def->type, handle->suppress_messages);
return 0;
}
}
v_iter = v;
do {
if (!_config_def_check_node_single_value(handle, rp, v_iter, def))
return 0;
v_iter = v_iter->next;
} while (v_iter);
if (handle->check_diff)
_check_value_differs_from_default(handle, v, def, NULL);
return 1;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
static int _config_def_check_node_is_profilable(struct cft_check_handle *handle,
const char *rp, struct dm_config_node *cn,
const cfg_def_item_t *def)
{
uint16_t flags;
if (!(def->flags & CFG_PROFILABLE)) {
log_warn_suppress(handle->suppress_messages,
"Configuration %s \"%s\" is not customizable by "
"a profile.", cn->v ? "option" : "section", rp);
return 0;
}
flags = def->flags & ~CFG_PROFILABLE;
/*
* Make sure there is no metadata profilable config in the command profile!
*/
if ((handle->source == CONFIG_PROFILE_COMMAND) && (flags & CFG_PROFILABLE_METADATA)) {
log_warn_suppress(handle->suppress_messages,
"Configuration %s \"%s\" is customizable by "
"metadata profile only, not command profile.",
cn->v ? "option" : "section", rp);
return 0;
}
/*
* Make sure there is no command profilable config in the metadata profile!
* (sections do not need to be flagged with CFG_PROFILABLE_METADATA, the
* CFG_PROFILABLE is enough as sections may contain both types inside)
*/
if ((handle->source == CONFIG_PROFILE_METADATA) && cn->v && !(flags & CFG_PROFILABLE_METADATA)) {
log_warn_suppress(handle->suppress_messages,
"Configuration %s \"%s\" is customizable by "
"command profile only, not metadata profile.",
cn->v ? "option" : "section", rp);
return 0;
}
return 1;
}
static int _config_def_check_node(struct cft_check_handle *handle,
const char *vp, char *pvp, char *rp, char *prp,
size_t buf_size, struct dm_config_node *cn)
{
cfg_def_item_t *def;
int sep = vp != pvp; /* don't use '/' separator for top-level node */
if (dm_snprintf(pvp, buf_size, "%s%s", sep ? "/" : "", cn->key) < 0 ||
dm_snprintf(prp, buf_size, "%s%s", sep ? "/" : "", cn->key) < 0) {
log_error("Failed to construct path for configuration node %s.", cn->key);
return 0;
}
if (!(def = (cfg_def_item_t *) dm_hash_lookup(handle->cmd->cft_def_hash, vp))) {
/* If the node is not a section but a setting, fail now. */
if (cn->v) {
log_warn_suppress(handle->suppress_messages,
"Configuration setting \"%s\" unknown.", rp);
config: add support for enhanced config node output There's a possibility to interconnect the dm_config_node with an ID, which in our case is used to reference the configuration definition ID from config_settings.h. So simply interconnecting struct dm_config_node with struct cfg_def_item. This patch also adds support for enhanced config node output besides existing "output line by line". This patch adds a possibility to register a callback that gets called *before* the config node is processed line by line (for example to include any headers on output) and *after* the config node is processed line by line (to include any footers on output). Also, it adds the config node reference itself as the callback arg in addition to have a possibility to extract more information from the config node itself if needed when processing the output callback (e.g. the key name, the id, or whether this is a section or a value etc...). If the config node from lvm.conf/--config tree is recognized and valid, it's always coupled with the config node definition ID from config_settings.h: struct dm_config_node { int id; const char *key; struct dm_config_node *parent, *sib, *child; struct dm_config_value *v; } For example if the dm_config_node *cn holds "devices/dev" configuration, then the cn->id holds "devices_dev_CFG" ID from config_settings.h, -1 if not found in config_settings.h and 0 if matching has not yet been done. To support the enhanced config node output, a new structure has been defined in libdevmapper to register it: struct dm_config_node_out_spec { dm_config_node_out_fn prefix_fn; /* called before processing config node lines */ dm_config_node_out_fn line_fn; /* called for each config node line */ dm_config_node_out_fn suffix_fn; /* called after processing config node lines */ }; Where dm_config_node_out_fn is: typedef int (*dm_config_node_out_fn)(const struct dm_config_node *cn, const char *line, void *baton); (so in comparison to existing callbacks for config node output, it has an extra dm_config_node *cn arg in addition) This patch also adds these functions to libdevmapper: - dm_config_write_node_out - dm_config_write_one_node_out ...which have exactly the same functionality as their counterparts without the "out" suffix. The "*_out" functions adds the extra hooks for enhanced config output (prefix_fn and suffix_fn mentioned above). One can still use the old interface for config node output, this is just an enhancement for those who'd like to modify the output more extensively.
2013-03-05 21:02:13 +04:00
cn->id = -1;
return 0;
}
/* If the node is a section, try if the section name is variable. */
/* Modify virtual path vp in situ and replace the key name with a '#'. */
/* The real path without '#' is still stored in rp variable. */
pvp[sep] = '#', pvp[sep + 1] = '\0';
if (!(def = (cfg_def_item_t *) dm_hash_lookup(handle->cmd->cft_def_hash, vp))) {
log_warn_suppress(handle->suppress_messages,
"Configuration section \"%s\" unknown.", rp);
config: add support for enhanced config node output There's a possibility to interconnect the dm_config_node with an ID, which in our case is used to reference the configuration definition ID from config_settings.h. So simply interconnecting struct dm_config_node with struct cfg_def_item. This patch also adds support for enhanced config node output besides existing "output line by line". This patch adds a possibility to register a callback that gets called *before* the config node is processed line by line (for example to include any headers on output) and *after* the config node is processed line by line (to include any footers on output). Also, it adds the config node reference itself as the callback arg in addition to have a possibility to extract more information from the config node itself if needed when processing the output callback (e.g. the key name, the id, or whether this is a section or a value etc...). If the config node from lvm.conf/--config tree is recognized and valid, it's always coupled with the config node definition ID from config_settings.h: struct dm_config_node { int id; const char *key; struct dm_config_node *parent, *sib, *child; struct dm_config_value *v; } For example if the dm_config_node *cn holds "devices/dev" configuration, then the cn->id holds "devices_dev_CFG" ID from config_settings.h, -1 if not found in config_settings.h and 0 if matching has not yet been done. To support the enhanced config node output, a new structure has been defined in libdevmapper to register it: struct dm_config_node_out_spec { dm_config_node_out_fn prefix_fn; /* called before processing config node lines */ dm_config_node_out_fn line_fn; /* called for each config node line */ dm_config_node_out_fn suffix_fn; /* called after processing config node lines */ }; Where dm_config_node_out_fn is: typedef int (*dm_config_node_out_fn)(const struct dm_config_node *cn, const char *line, void *baton); (so in comparison to existing callbacks for config node output, it has an extra dm_config_node *cn arg in addition) This patch also adds these functions to libdevmapper: - dm_config_write_node_out - dm_config_write_one_node_out ...which have exactly the same functionality as their counterparts without the "out" suffix. The "*_out" functions adds the extra hooks for enhanced config output (prefix_fn and suffix_fn mentioned above). One can still use the old interface for config node output, this is just an enhancement for those who'd like to modify the output more extensively.
2013-03-05 21:02:13 +04:00
cn->id = -1;
return 0;
}
}
handle->status[def->id] |= CFG_USED;
config: add support for enhanced config node output There's a possibility to interconnect the dm_config_node with an ID, which in our case is used to reference the configuration definition ID from config_settings.h. So simply interconnecting struct dm_config_node with struct cfg_def_item. This patch also adds support for enhanced config node output besides existing "output line by line". This patch adds a possibility to register a callback that gets called *before* the config node is processed line by line (for example to include any headers on output) and *after* the config node is processed line by line (to include any footers on output). Also, it adds the config node reference itself as the callback arg in addition to have a possibility to extract more information from the config node itself if needed when processing the output callback (e.g. the key name, the id, or whether this is a section or a value etc...). If the config node from lvm.conf/--config tree is recognized and valid, it's always coupled with the config node definition ID from config_settings.h: struct dm_config_node { int id; const char *key; struct dm_config_node *parent, *sib, *child; struct dm_config_value *v; } For example if the dm_config_node *cn holds "devices/dev" configuration, then the cn->id holds "devices_dev_CFG" ID from config_settings.h, -1 if not found in config_settings.h and 0 if matching has not yet been done. To support the enhanced config node output, a new structure has been defined in libdevmapper to register it: struct dm_config_node_out_spec { dm_config_node_out_fn prefix_fn; /* called before processing config node lines */ dm_config_node_out_fn line_fn; /* called for each config node line */ dm_config_node_out_fn suffix_fn; /* called after processing config node lines */ }; Where dm_config_node_out_fn is: typedef int (*dm_config_node_out_fn)(const struct dm_config_node *cn, const char *line, void *baton); (so in comparison to existing callbacks for config node output, it has an extra dm_config_node *cn arg in addition) This patch also adds these functions to libdevmapper: - dm_config_write_node_out - dm_config_write_one_node_out ...which have exactly the same functionality as their counterparts without the "out" suffix. The "*_out" functions adds the extra hooks for enhanced config output (prefix_fn and suffix_fn mentioned above). One can still use the old interface for config node output, this is just an enhancement for those who'd like to modify the output more extensively.
2013-03-05 21:02:13 +04:00
cn->id = def->id;
if (!_config_def_check_node_value(handle, rp, cn->v, def))
return 0;
/*
* Also check whether this configuration item is allowed
* in certain types of configuration trees as in some
* the use of configuration is restricted, e.g. profiles...
*/
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (_is_profile_based_config_source(handle->source) &&
!_config_def_check_node_is_profilable(handle, rp, cn, def))
return_0;
handle->status[def->id] |= CFG_VALID;
return 1;
}
static int _config_def_check_tree(struct cft_check_handle *handle,
const char *vp, char *pvp, char *rp, char *prp,
size_t buf_size, struct dm_config_node *root)
{
struct dm_config_node *cn;
int valid, r = 1;
size_t len;
for (cn = root->child; cn; cn = cn->sib) {
if ((valid = _config_def_check_node(handle, vp, pvp, rp, prp,
buf_size, cn)) && !cn->v) {
len = strlen(rp);
valid = _config_def_check_tree(handle, vp, pvp + strlen(pvp),
rp, prp + len, buf_size - len, cn);
}
if (!valid)
r = 0;
}
return r;
}
int config_def_check(struct cft_check_handle *handle)
{
cfg_def_item_t *def;
struct dm_config_node *cn;
char vp[CFG_PATH_MAX_LEN], rp[CFG_PATH_MAX_LEN];
size_t rplen;
int id, r = 1;
/*
* vp = virtual path, it might contain substitutes for variable parts
* of the path, used while working with the hash
* rp = real path, the real path of the config element as found in the
* configuration, used for message output
*/
/*
* If the check has already been done and 'skip_if_checked' is set,
* skip the actual check and use last result if available.
* If not available, we must do the check. The global status
* is stored in root node.
*/
if (handle->skip_if_checked && (handle->status[root_CFG_SECTION] & CFG_USED))
return handle->status[root_CFG_SECTION] & CFG_VALID;
/* Nothing to do if checks are disabled and also not forced. */
if (!handle->force_check && !find_config_tree_bool(handle->cmd, config_checks_CFG, NULL))
return 1;
/* Clear 'used' and 'valid' status flags. */
for (id = 0; id < CFG_COUNT; id++)
handle->status[id] &= ~(CFG_USED | CFG_VALID | CFG_DIFF);
/*
* Create a hash of all possible configuration
* sections and settings with full path as a key.
* If section name is variable, use '#' as a substitute.
*/
if (!handle->cmd->cft_def_hash) {
if (!(handle->cmd->cft_def_hash = dm_hash_create(64))) {
log_error("Failed to create configuration definition hash.");
r = 0; goto out;
}
for (id = 1; id < CFG_COUNT; id++) {
def = cfg_def_get_item_p(id);
if (!_cfg_def_make_path(vp, CFG_PATH_MAX_LEN, def->id, def, 0)) {
dm_hash_destroy(handle->cmd->cft_def_hash);
handle->cmd->cft_def_hash = NULL;
r = 0; goto out;
}
if (!dm_hash_insert(handle->cmd->cft_def_hash, vp, def)) {
log_error("Failed to insert configuration to hash.");
r = 0;
goto out;
}
}
}
/*
* Mark this handle as used so next time we know that the check
* has already been done and so we can just reuse the previous
* status instead of running this whole check again.
*/
handle->status[root_CFG_SECTION] |= CFG_USED;
/*
* Allow only sections as top-level elements.
* Iterate top-level sections and dive deeper.
* If any of subsequent checks fails, the whole check fails.
*/
for (cn = handle->cft->root; cn; cn = cn->sib) {
if (!cn->v) {
/* top level node: vp=vp, rp=rp */
if (!_config_def_check_node(handle, vp, vp, rp, rp,
CFG_PATH_MAX_LEN, cn)) {
r = 0; continue;
}
rplen = strlen(rp);
if (!_config_def_check_tree(handle,
vp, vp + strlen(vp),
rp, rp + rplen,
CFG_PATH_MAX_LEN - rplen, cn))
r = 0;
} else {
log_error_suppress(handle->suppress_messages,
"Configuration setting \"%s\" invalid. "
"It's not part of any section.", cn->key);
r = 0;
}
}
out:
if (r)
handle->status[root_CFG_SECTION] |= CFG_VALID;
else
handle->status[root_CFG_SECTION] &= ~CFG_VALID;
return r;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
static int _apply_local_profile(struct cmd_context *cmd, struct profile *profile)
{
if (!profile)
return 0;
/*
* Global metadata profile overrides the local one.
* This simply means the "--metadataprofile" arg
* overrides any profile attached to VG/LV.
*/
if ((profile->source == CONFIG_PROFILE_METADATA) &&
cmd->profile_params->global_metadata_profile)
return 0;
return override_config_tree_from_profile(cmd, profile);
}
const struct dm_config_node *find_config_tree_node(struct cmd_context *cmd, int id, struct profile *profile)
{
cfg_def_item_t *item = cfg_def_get_item_p(id);
char path[CFG_PATH_MAX_LEN];
int profile_applied;
const struct dm_config_node *cn;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
profile_applied = _apply_local_profile(cmd, profile);
_cfg_def_make_path(path, sizeof(path), item->id, item, 0);
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
cn = dm_config_tree_find_node(cmd->cft, path);
if (profile_applied)
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
remove_config_tree_by_source(cmd, profile->source);
return cn;
}
const char *find_config_tree_str(struct cmd_context *cmd, int id, struct profile *profile)
{
cfg_def_item_t *item = cfg_def_get_item_p(id);
char path[CFG_PATH_MAX_LEN];
int profile_applied;
const char *str;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
profile_applied = _apply_local_profile(cmd, profile);
_cfg_def_make_path(path, sizeof(path), item->id, item, 0);
if (item->type != CFG_TYPE_STRING)
log_error(INTERNAL_ERROR "%s cfg tree element not declared as string.", path);
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
str = dm_config_tree_find_str(cmd->cft, path, cfg_def_get_default_value(cmd, item, CFG_TYPE_STRING, profile));
if (profile_applied)
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
remove_config_tree_by_source(cmd, profile->source);
return str;
}
const char *find_config_tree_str_allow_empty(struct cmd_context *cmd, int id, struct profile *profile)
{
cfg_def_item_t *item = cfg_def_get_item_p(id);
char path[CFG_PATH_MAX_LEN];
int profile_applied;
const char *str;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
profile_applied = _apply_local_profile(cmd, profile);
_cfg_def_make_path(path, sizeof(path), item->id, item, 0);
if (item->type != CFG_TYPE_STRING)
log_error(INTERNAL_ERROR "%s cfg tree element not declared as string.", path);
if (!(item->flags & CFG_ALLOW_EMPTY))
log_error(INTERNAL_ERROR "%s cfg tree element not declared to allow empty values.", path);
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
str = dm_config_tree_find_str_allow_empty(cmd->cft, path, cfg_def_get_default_value(cmd, item, CFG_TYPE_STRING, profile));
if (profile_applied)
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
remove_config_tree_by_source(cmd, profile->source);
return str;
}
int find_config_tree_int(struct cmd_context *cmd, int id, struct profile *profile)
{
cfg_def_item_t *item = cfg_def_get_item_p(id);
char path[CFG_PATH_MAX_LEN];
int profile_applied;
int i;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
profile_applied = _apply_local_profile(cmd, profile);
_cfg_def_make_path(path, sizeof(path), item->id, item, 0);
if (item->type != CFG_TYPE_INT)
log_error(INTERNAL_ERROR "%s cfg tree element not declared as integer.", path);
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
i = dm_config_tree_find_int(cmd->cft, path, cfg_def_get_default_value(cmd, item, CFG_TYPE_INT, profile));
if (profile_applied)
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
remove_config_tree_by_source(cmd, profile->source);
return i;
}
int64_t find_config_tree_int64(struct cmd_context *cmd, int id, struct profile *profile)
2011-02-18 17:08:22 +03:00
{
cfg_def_item_t *item = cfg_def_get_item_p(id);
char path[CFG_PATH_MAX_LEN];
int profile_applied;
int i64;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
profile_applied = _apply_local_profile(cmd, profile);
_cfg_def_make_path(path, sizeof(path), item->id, item, 0);
if (item->type != CFG_TYPE_INT)
log_error(INTERNAL_ERROR "%s cfg tree element not declared as integer.", path);
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
i64 = dm_config_tree_find_int64(cmd->cft, path, cfg_def_get_default_value(cmd, item, CFG_TYPE_INT, profile));
if (profile_applied)
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
remove_config_tree_by_source(cmd, profile->source);
return i64;
2011-02-18 17:08:22 +03:00
}
float find_config_tree_float(struct cmd_context *cmd, int id, struct profile *profile)
{
cfg_def_item_t *item = cfg_def_get_item_p(id);
char path[CFG_PATH_MAX_LEN];
int profile_applied;
float f;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
profile_applied = _apply_local_profile(cmd, profile);
_cfg_def_make_path(path, sizeof(path), item->id, item, 0);
if (item->type != CFG_TYPE_FLOAT)
log_error(INTERNAL_ERROR "%s cfg tree element not declared as float.", path);
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
f = dm_config_tree_find_float(cmd->cft, path, cfg_def_get_default_value(cmd, item, CFG_TYPE_FLOAT, profile));
if (profile_applied)
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
remove_config_tree_by_source(cmd, profile->source);
return f;
}
int find_config_tree_bool(struct cmd_context *cmd, int id, struct profile *profile)
{
cfg_def_item_t *item = cfg_def_get_item_p(id);
char path[CFG_PATH_MAX_LEN];
int profile_applied;
int b;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
profile_applied = _apply_local_profile(cmd, profile);
_cfg_def_make_path(path, sizeof(path), item->id, item, 0);
if (item->type != CFG_TYPE_BOOL)
log_error(INTERNAL_ERROR "%s cfg tree element not declared as boolean.", path);
config: add CFG_DEFAULT_RUN_TIME for config options with runtime defaults Previously, we declared a default value as undefined ("NULL") for settings which require runtime context to be set first (e.g. settings for paths that rely on SYSTEM_DIR environment variable or they depend on any other setting in some way). If we want to output default values as they are really used in runtime, we should make it possible to define a default value as function which is evaluated, not just providing a firm constant value as it was before. This patch defines simple prototypes for such functions. Also, there's new helper macros "cfg_runtime" and "cfg_array_runtime" - they provide exactly the same functionality as the original "cfg" and "cfg_array" macros when defining the configuration settings in config_settings.h, but they don't set the constant default value. Instead, they automatically link the configuration setting definition with one of these functions: typedef int (*t_fn_CFG_TYPE_BOOL) (struct cmd_context *cmd, struct profile *profile); typedef int (*t_fn_CFG_TYPE_INT) (struct cmd_context *cmd, struct profile *profile); typedef float (*t_fn_CFG_TYPE_FLOAT) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_STRING) (struct cmd_context *cmd, struct profile *profile); typedef const char* (*t_fn_CFG_TYPE_ARRAY) (struct cmd_context *cmd, struct profile *profile); (The new macros actually set the CFG_DEFAULT_RUNTIME flag properly and set the default value link to the function accordingly). Then such configuration setting requires a function of selected type to be defined. This function has a predefined name: get_default_<id> ...where the <id> is the id of the setting as defined in config_settings.h. For example "backup_archive_dir_CFG" if defined as a setting with default value evaluated in runtime with "cfg_runtime" will automatically have "get_default_backup_archive_dir_CFG" function linked to this setting to get the default value.
2014-03-03 15:34:11 +04:00
b = dm_config_tree_find_bool(cmd->cft, path, cfg_def_get_default_value(cmd, item, CFG_TYPE_BOOL, profile));
if (profile_applied)
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
remove_config_tree_by_source(cmd, profile->source);
return b;
2002-07-11 18:07:43 +04:00
}
2004-05-04 22:28:15 +04:00
/* Insert cn2 after cn1 */
static void _insert_config_node(struct dm_config_node **cn1,
struct dm_config_node *cn2)
2004-05-04 22:28:15 +04:00
{
if (!*cn1) {
*cn1 = cn2;
cn2->sib = NULL;
} else {
cn2->sib = (*cn1)->sib;
(*cn1)->sib = cn2;
}
}
/*
* Merge section cn2 into section cn1 (which has the same name)
* overwriting any existing cn1 nodes with matching names.
*/
static void _merge_section(struct dm_config_node *cn1, struct dm_config_node *cn2,
config_merge_t merge_type)
2004-05-04 22:28:15 +04:00
{
struct dm_config_node *cn, *nextn, *oldn;
struct dm_config_value *cv;
2004-05-04 22:28:15 +04:00
for (cn = cn2->child; cn; cn = nextn) {
nextn = cn->sib;
if (merge_type == CONFIG_MERGE_TYPE_TAGS) {
/* Skip "tags" */
if (!strcmp(cn->key, "tags"))
continue;
}
2004-05-04 22:28:15 +04:00
/* Subsection? */
if (!cn->v)
/* Ignore - we don't have any of these yet */
continue;
/* Not already present? */
if (!(oldn = dm_config_find_node(cn1->child, cn->key))) {
2004-05-04 22:28:15 +04:00
_insert_config_node(&cn1->child, cn);
continue;
}
if (merge_type == CONFIG_MERGE_TYPE_TAGS) {
/* Merge certain value lists */
if ((!strcmp(cn1->key, "activation") &&
!strcmp(cn->key, "volume_list")) ||
(!strcmp(cn1->key, "devices") &&
(!strcmp(cn->key, "filter") || !strcmp(cn->key, "types")))) {
cv = cn->v;
while (cv->next)
cv = cv->next;
cv->next = oldn->v;
}
2004-05-04 22:28:15 +04:00
}
/* Replace values */
oldn->v = cn->v;
}
}
static int _match_host_tags(struct dm_list *tags, const struct dm_config_node *tn)
2004-05-04 22:28:15 +04:00
{
const struct dm_config_value *tv;
2004-05-04 22:28:15 +04:00
const char *str;
for (tv = tn->v; tv; tv = tv->next) {
if (tv->type != DM_CFG_STRING)
2004-05-04 22:28:15 +04:00
continue;
str = tv->v.str;
if (*str == '@')
str++;
if (!*str)
continue;
if (str_list_match_item(tags, str))
return 1;
}
return 0;
}
/* Destructively merge a new config tree into an existing one */
int merge_config_tree(struct cmd_context *cmd, struct dm_config_tree *cft,
struct dm_config_tree *newdata, config_merge_t merge_type)
2004-05-04 22:28:15 +04:00
{
struct dm_config_node *root = cft->root;
struct dm_config_node *cn, *nextn, *oldn, *cn2;
const struct dm_config_node *tn;
struct config_source *cs, *csn;
2004-05-04 22:28:15 +04:00
for (cn = newdata->root; cn; cn = nextn) {
nextn = cn->sib;
if (merge_type == CONFIG_MERGE_TYPE_TAGS) {
/* Ignore tags section */
if (!strcmp(cn->key, "tags"))
2004-05-04 22:28:15 +04:00
continue;
/* If there's a tags node, skip if host tags don't match */
if ((tn = dm_config_find_node(cn->child, "tags"))) {
if (!_match_host_tags(&cmd->tags, tn))
continue;
}
2004-05-04 22:28:15 +04:00
}
if (!(oldn = dm_config_find_node(root, cn->key))) {
2004-05-04 22:28:15 +04:00
_insert_config_node(&cft->root, cn);
if (merge_type == CONFIG_MERGE_TYPE_TAGS) {
/* Remove any "tags" nodes */
for (cn2 = cn->child; cn2; cn2 = cn2->sib) {
if (!strcmp(cn2->key, "tags")) {
cn->child = cn2->sib;
continue;
}
if (cn2->sib && !strcmp(cn2->sib->key, "tags")) {
cn2->sib = cn2->sib->sib;
continue;
}
2004-05-04 22:28:15 +04:00
}
}
continue;
}
_merge_section(oldn, cn, merge_type);
2004-05-04 22:28:15 +04:00
}
/*
* Persistent filter loading is based on timestamp,
* so we need to know the newest timestamp to make right decision
* whether the .cache isn't older then any of configs
*/
cs = dm_config_get_custom(cft);
csn = dm_config_get_custom(newdata);
if (cs && csn && (cs->timestamp < csn->timestamp))
cs->timestamp = csn->timestamp;
2004-05-04 22:28:15 +04:00
return 1;
}
struct out_baton {
FILE *fp;
struct config_def_tree_spec *tree_spec;
struct dm_pool *mem;
};
static int _out_prefix_fn(const struct dm_config_node *cn, const char *line, void *baton)
{
struct out_baton *out = baton;
struct cfg_def_item *cfg_def;
char version[9]; /* 8+1 chars for max version of 7.15.511 */
const char *node_type_name = cn->v ? "option" : "section";
char path[CFG_PATH_MAX_LEN];
if (cn->id < 0)
return 1;
if (!cn->id) {
log_error(INTERNAL_ERROR "Configuration node %s has invalid id.", cn->key);
return 0;
}
if ((out->tree_spec->type == CFG_DEF_TREE_DIFF) &&
(!(out->tree_spec->check_status[cn->id] & CFG_DIFF)))
return 1;
cfg_def = cfg_def_get_item_p(cn->id);
if (out->tree_spec->withcomments) {
_cfg_def_make_path(path, sizeof(path), cfg_def->id, cfg_def, 1);
fprintf(out->fp, "%s# Configuration %s %s.\n", line, node_type_name, path);
if (cfg_def->comment)
fprintf(out->fp, "%s# %s\n", line, cfg_def->comment);
if (cfg_def->flags & CFG_ADVANCED)
fprintf(out->fp, "%s# This configuration %s is advanced.\n", line, node_type_name);
if (cfg_def->flags & CFG_UNSUPPORTED)
fprintf(out->fp, "%s# This configuration %s is not officially supported.\n", line, node_type_name);
if (cfg_def->flags & CFG_NAME_VARIABLE)
fprintf(out->fp, "%s# This configuration %s has variable name.\n", line, node_type_name);
if (cfg_def->flags & CFG_DEFAULT_UNDEFINED)
fprintf(out->fp, "%s# This configuration %s does not have a default value defined.\n", line, node_type_name);
}
if (out->tree_spec->withversions) {
if (dm_snprintf(version, 9, "%u.%u.%u",
(cfg_def->since_version & 0xE000) >> 13,
(cfg_def->since_version & 0x1E00) >> 9,
(cfg_def->since_version & 0x1FF)) == -1) {
log_error("_out_prefix_fn: couldn't create version string");
return 0;
}
fprintf(out->fp, "%s# Since version %s.\n", line, version);
}
return 1;
}
static int _out_line_fn(const struct dm_config_node *cn, const char *line, void *baton)
{
struct out_baton *out = baton;
struct cfg_def_item *cfg_def = cfg_def_get_item_p(cn->id);
if ((out->tree_spec->type == CFG_DEF_TREE_DIFF) &&
(!(out->tree_spec->check_status[cn->id] & CFG_DIFF)))
return 1;
fprintf(out->fp, "%s%s\n", (out->tree_spec->type != CFG_DEF_TREE_CURRENT) &&
(out->tree_spec->type != CFG_DEF_TREE_DIFF) &&
(cfg_def->flags & CFG_DEFAULT_UNDEFINED) ? "#" : "", line);
return 1;
}
static int _out_suffix_fn(const struct dm_config_node *cn, const char *line, void *baton)
{
return 1;
}
int config_write(struct dm_config_tree *cft,
struct config_def_tree_spec *tree_spec,
const char *file, int argc, char **argv)
{
static const struct dm_config_node_out_spec _out_spec = {
.prefix_fn = _out_prefix_fn,
.line_fn = _out_line_fn,
.suffix_fn = _out_suffix_fn
};
const struct dm_config_node *cn;
struct out_baton baton = {
.tree_spec = tree_spec,
.mem = cft->mem
};
int r = 1;
if (!file) {
baton.fp = stdout;
file = "stdout";
} else if (!(baton.fp = fopen(file, "w"))) {
log_sys_error("open", file);
return 0;
}
log_verbose("Dumping configuration to %s", file);
if (!argc) {
if (!dm_config_write_node_out(cft->root, &_out_spec, &baton)) {
log_error("Failure while writing to %s", file);
r = 0;
}
} else while (argc--) {
if ((cn = dm_config_find_node(cft->root, *argv))) {
if (!dm_config_write_one_node_out(cn, &_out_spec, &baton)) {
log_error("Failure while writing to %s", file);
r = 0;
}
} else {
log_error("Configuration node %s not found", *argv);
r = 0;
}
argv++;
}
if (baton.fp && baton.fp != stdout && dm_fclose(baton.fp)) {
stack;
r = 0;
}
return r;
}
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
static struct dm_config_node *_add_def_node(struct dm_config_tree *cft,
struct config_def_tree_spec *spec,
struct dm_config_node *parent,
struct dm_config_node *relay,
cfg_def_item_t *def)
{
struct dm_config_node *cn;
const char *str;
if (!(cn = dm_config_create_node(cft, def->name))) {
log_error("Failed to create default config setting node.");
return NULL;
}
if (!(def->type & CFG_TYPE_SECTION) && (!(cn->v = dm_config_create_value(cft)))) {
log_error("Failed to create default config setting node value.");
return NULL;
}
config: add support for enhanced config node output There's a possibility to interconnect the dm_config_node with an ID, which in our case is used to reference the configuration definition ID from config_settings.h. So simply interconnecting struct dm_config_node with struct cfg_def_item. This patch also adds support for enhanced config node output besides existing "output line by line". This patch adds a possibility to register a callback that gets called *before* the config node is processed line by line (for example to include any headers on output) and *after* the config node is processed line by line (to include any footers on output). Also, it adds the config node reference itself as the callback arg in addition to have a possibility to extract more information from the config node itself if needed when processing the output callback (e.g. the key name, the id, or whether this is a section or a value etc...). If the config node from lvm.conf/--config tree is recognized and valid, it's always coupled with the config node definition ID from config_settings.h: struct dm_config_node { int id; const char *key; struct dm_config_node *parent, *sib, *child; struct dm_config_value *v; } For example if the dm_config_node *cn holds "devices/dev" configuration, then the cn->id holds "devices_dev_CFG" ID from config_settings.h, -1 if not found in config_settings.h and 0 if matching has not yet been done. To support the enhanced config node output, a new structure has been defined in libdevmapper to register it: struct dm_config_node_out_spec { dm_config_node_out_fn prefix_fn; /* called before processing config node lines */ dm_config_node_out_fn line_fn; /* called for each config node line */ dm_config_node_out_fn suffix_fn; /* called after processing config node lines */ }; Where dm_config_node_out_fn is: typedef int (*dm_config_node_out_fn)(const struct dm_config_node *cn, const char *line, void *baton); (so in comparison to existing callbacks for config node output, it has an extra dm_config_node *cn arg in addition) This patch also adds these functions to libdevmapper: - dm_config_write_node_out - dm_config_write_one_node_out ...which have exactly the same functionality as their counterparts without the "out" suffix. The "*_out" functions adds the extra hooks for enhanced config output (prefix_fn and suffix_fn mentioned above). One can still use the old interface for config node output, this is just an enhancement for those who'd like to modify the output more extensively.
2013-03-05 21:02:13 +04:00
cn->id = def->id;
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
if (!(def->type & CFG_TYPE_ARRAY)) {
switch (def->type) {
case CFG_TYPE_SECTION:
cn->v = NULL;
break;
case CFG_TYPE_BOOL:
cn->v->type = DM_CFG_INT;
cn->v->v.i = cfg_def_get_default_value_hint(spec->cmd, def, CFG_TYPE_BOOL, NULL);
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
break;
case CFG_TYPE_INT:
cn->v->type = DM_CFG_INT;
cn->v->v.i = cfg_def_get_default_value_hint(spec->cmd, def, CFG_TYPE_INT, NULL);
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
break;
case CFG_TYPE_FLOAT:
cn->v->type = DM_CFG_FLOAT;
cn->v->v.f = cfg_def_get_default_value_hint(spec->cmd, def, CFG_TYPE_FLOAT, NULL);
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
break;
case CFG_TYPE_STRING:
cn->v->type = DM_CFG_STRING;
if (!(str = cfg_def_get_default_value_hint(spec->cmd, def, CFG_TYPE_STRING, NULL)))
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
str = "";
cn->v->v.str = str;
break;
default:
log_error(INTERNAL_ERROR "_add_def_node: unknown type");
return NULL;
break;
}
} else
cn->v = _get_def_array_values(cft, def);
cn->child = NULL;
if (parent) {
cn->parent = parent;
if (!parent->child)
parent->child = cn;
} else
cn->parent = cn;
if (relay)
relay->sib = cn;
return cn;
}
static int _should_skip_def_node(struct config_def_tree_spec *spec, int section_id, int id)
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
{
cfg_def_item_t *def = cfg_def_get_item_p(id);
uint16_t flags;
if ((def->parent != section_id) ||
(spec->ignoreadvanced && def->flags & CFG_ADVANCED) ||
(spec->ignoreunsupported && def->flags & CFG_UNSUPPORTED))
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
return 1;
switch (spec->type) {
case CFG_DEF_TREE_MISSING:
if (!spec->check_status) {
log_error_once(INTERNAL_ERROR "couldn't determine missing "
"config nodes - unknown status of last config check.");
return 1;
}
if ((spec->check_status[id] & CFG_USED) ||
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
(def->flags & CFG_NAME_VARIABLE) ||
(def->since_version > spec->version))
return 1;
break;
case CFG_DEF_TREE_NEW:
if (def->since_version != spec->version)
return 1;
break;
case CFG_DEF_TREE_PROFILABLE:
case CFG_DEF_TREE_PROFILABLE_CMD:
case CFG_DEF_TREE_PROFILABLE_MDA:
if (!(def->flags & CFG_PROFILABLE) ||
(def->since_version > spec->version))
return 1;
flags = def->flags & ~CFG_PROFILABLE;
if (spec->type == CFG_DEF_TREE_PROFILABLE_CMD) {
if (flags & CFG_PROFILABLE_METADATA)
return 1;
} else if (spec->type == CFG_DEF_TREE_PROFILABLE_MDA) {
if (!(flags & CFG_PROFILABLE_METADATA))
return 1;
}
break;
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
default:
if (def->since_version > spec->version)
return 1;
break;
}
return 0;
}
static struct dm_config_node *_add_def_section_subtree(struct dm_config_tree *cft,
struct config_def_tree_spec *spec,
struct dm_config_node *parent,
struct dm_config_node *relay,
int section_id)
{
struct dm_config_node *cn = NULL, *relay_sub = NULL, *tmp;
cfg_def_item_t *def;
int id;
for (id = 0; id < CFG_COUNT; id++) {
if (_should_skip_def_node(spec, section_id, id))
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
continue;
if (!cn && !(cn = _add_def_node(cft, spec, parent, relay, cfg_def_get_item_p(section_id))))
goto bad;
def = cfg_def_get_item_p(id);
config: use config checks and add support for creating trees from config definition (config_def_create_tree fn) Configuration checking is initiated during config load/processing (_process_config fn) which is part of the command context creation/refresh. This patch also defines 5 types of trees that could be created from the configuration definition (config_settings.h), the cfg_def_tree_t: - CFG_DEF_TREE_CURRENT that denotes a tree of all the configuration nodes that are explicitly defined in lvm.conf/--config - CFG_DEF_TREE_MISSING that denotes a tree of all missing configuration nodes for which default valus are used since they're not explicitly used in lvm.conf/--config - CFG_DEF_TREE_DEFAULT that denotes a tree of all possible configuration nodes with default values assigned, no matter what the actual lvm.conf/--config is - CFG_DEF_TREE_NEW that denotes a tree of all new configuration nodes that appeared in given version - CFG_DEF_TREE_COMPLETE that denotes a tree of the whole configuration tree that is used in LVM2 (a combination of CFG_DEF_TREE_CURRENT + CFG_DEF_TREE_MISSING). This is not implemented yet, it will be added later... The function that creates the definition tree of given type: struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec); Where the "spec" specifies the tree type to be created: struct config_def_tree_spec { cfg_def_tree_t type; /* tree type */ uint16_t version; /* tree at this LVM2 version */ int ignoreadvanced; /* do not include advanced configs */ int ignoreunsupported; /* do not include unsupported configs */ }; This tree can be passed to already existing functions that write the tree on output (like we already do with cmd->cft). There is a new lvm.conf section called "config" with two new options: - config/checks which enables/disables checking (enabled by default) - config/abort_on_errors which enables/disables aborts on any type of mismatch found in the config (disabled by default)
2013-03-05 20:36:10 +04:00
if ((tmp = def->type == CFG_TYPE_SECTION ? _add_def_section_subtree(cft, spec, cn, relay_sub, id)
: _add_def_node(cft, spec, cn, relay_sub, def)))
relay_sub = tmp;
}
return cn;
bad:
log_error("Failed to create default config section node.");
return NULL;
}
struct dm_config_tree *config_def_create_tree(struct config_def_tree_spec *spec)
{
struct dm_config_tree *cft;
struct dm_config_node *root = NULL, *relay = NULL, *tmp;
int id;
if (!(cft = dm_config_create())) {
log_error("Failed to create default config tree.");
return NULL;
}
for (id = root_CFG_SECTION + 1; id < CFG_COUNT; id++) {
if (cfg_def_get_item_p(id)->parent != root_CFG_SECTION)
continue;
if ((tmp = _add_def_section_subtree(cft, spec, root, relay, id))) {
relay = tmp;
if (!root)
root = relay;
}
}
cft->root = root;
return cft;
}
static int _check_profile(struct cmd_context *cmd, struct profile *profile)
{
struct cft_check_handle *handle;
int r;
if (!(handle = dm_pool_zalloc(cmd->libmem, sizeof(*handle)))) {
log_debug("_check_profile: profile check handle allocation failed");
return 0;
}
handle->cmd = cmd;
handle->cft = profile->cft;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
handle->source = profile->source;
/* the check is compulsory - allow only profilable items in a profile config! */
handle->force_check = 1;
/* provide warning messages only if config/checks=1 */
handle->suppress_messages = !find_config_tree_bool(cmd, config_checks_CFG, NULL);
r = config_def_check(handle);
dm_pool_free(cmd->libmem, handle);
return r;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
static int _get_profile_from_list(struct dm_list *list, const char *profile_name,
config_source_t source, struct profile **profile_found)
{
struct profile *profile;
dm_list_iterate_items(profile, list) {
if (!strcmp(profile->name, profile_name)) {
if (profile->source == source) {
*profile_found = profile;
return 1;
}
log_error(INTERNAL_ERROR "Profile %s already added as "
"%s type, but requested type is %s.",
profile_name,
_config_source_names[profile->source],
_config_source_names[source]);
return 0;
}
}
*profile_found = NULL;
return 1;
}
struct profile *add_profile(struct cmd_context *cmd, const char *profile_name, config_source_t source)
{
struct profile *profile;
/* Do some sanity checks first. */
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (!_is_profile_based_config_source(source)) {
log_error(INTERNAL_ERROR "add_profile: incorrect configuration "
"source, expected %s or %s but %s requested",
_config_source_names[CONFIG_PROFILE_COMMAND],
_config_source_names[CONFIG_PROFILE_METADATA],
_config_source_names[source]);
return NULL;
}
if (!profile_name || !*profile_name) {
log_error("Undefined profile name.");
return NULL;
}
if (strchr(profile_name, '/')) {
log_error("%s: bad profile name, it contains '/'.", profile_name);
return NULL;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
/*
* Check if the profile is on the list of profiles to be loaded or if
* not found there, if it's on the list of already loaded profiles.
*/
if (!_get_profile_from_list(&cmd->profile_params->profiles_to_load,
profile_name, source, &profile))
return_NULL;
if (profile)
profile->source = source;
else if (!_get_profile_from_list(&cmd->profile_params->profiles,
profile_name, source, &profile))
return_NULL;
if (profile) {
if (profile->source != source) {
log_error(INTERNAL_ERROR "add_profile: loaded profile "
"has incorrect type, expected %s but %s found",
_config_source_names[source],
_config_source_names[profile->source]);
return NULL;
}
return profile;
}
if (!(profile = dm_pool_zalloc(cmd->libmem, sizeof(*profile)))) {
log_error("profile allocation failed");
return NULL;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
profile->source = source;
profile->name = dm_pool_strdup(cmd->libmem, profile_name);
dm_list_add(&cmd->profile_params->profiles_to_load, &profile->list);
return profile;
}
int load_profile(struct cmd_context *cmd, struct profile *profile) {
static char profile_path[PATH_MAX];
if (critical_section()) {
log_error(INTERNAL_ERROR "trying to load profile %s "
"in critical section.", profile->name);
return 0;
}
if (profile->cft)
return 1;
if (dm_snprintf(profile_path, sizeof(profile_path), "%s/%s.profile",
cmd->profile_params->dir, profile->name) < 0) {
log_error("LVM_SYSTEM_DIR or profile name too long");
return 0;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
if (!(profile->cft = config_file_open_and_read(profile_path, profile->source, cmd)))
return 0;
/*
* *Profile must be valid* otherwise we'd end up with incorrect config!
* If there were config items present that are not supposed to be
* customized by a profile, we could end up with non-deterministic
* behaviour. Therefore, this check is *strictly forced* even if
* config/checks=0. The config/checks=0 will only cause the warning
* messages to be suppressed, but the check itself is always done
* for profiles!
*/
if (!_check_profile(cmd, profile)) {
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
log_error("Ignoring invalid %s %s.",
_config_source_names[profile->source], profile->name);
config_destroy(profile->cft);
profile->cft = NULL;
return 0;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
dm_list_move(&cmd->profile_params->profiles, &profile->list);
return 1;
}
int load_pending_profiles(struct cmd_context *cmd)
{
struct profile *profile, *temp_profile;
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
int r = 1;
dm_list_iterate_items_safe(profile, temp_profile, &cmd->profile_params->profiles_to_load) {
if (!load_profile(cmd, profile))
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
r = 0;
}
config: differentiate command and metadata profiles and consolidate profile handling code - When defining configuration source, the code now uses separate CONFIG_PROFILE_COMMAND and CONFIG_PROFILE_METADATA markers (before, it was just CONFIG_PROFILE that did not make the difference between the two). This helps when checking the configuration if it contains correct set of options which are all in either command-profilable or metadata-profilable group without mixing these groups together - so it's a firm distinction. The "command profile" can't contain "metadata profile" and vice versa! This is strictly checked and if the settings are mixed, such profile is rejected and it's not used. So in the end, the CONFIG_PROFILE_COMMAND set of options and CONFIG_PROFILE_METADATA are mutually exclusive sets. - Marking configuration with one or the other marker will also determine the way these configuration sources are positioned in the configuration cascade which is now: CONFIG_STRING -> CONFIG_PROFILE_COMMAND -> CONFIG_PROFILE_METADATA -> CONFIG_FILE/CONFIG_MERGED_FILES - Marking configuration with one or the other marker will also make it possible to issue a command context refresh (will be probably a part of a future patch) if needed for settings in global profile set. For settings in metadata profile set this is impossible since we can't refresh cmd context in the middle of reading VG/LV metadata and for each VG/LV separately because each VG/LV can have a different metadata profile assinged and it's not possible to change these settings at this level. - When command profile is incorrect, it's rejected *and also* the command exits immediately - the profile *must* be correct for the command that was run with a profile to be executed. Before this patch, when the profile was found incorrect, there was just the warning message and the command continued without profile applied. But it's more correct to exit immediately in this case. - When metadata profile is incorrect, we reject it during command runtime (as we know the profile name from metadata and not early from command line as it is in case of command profiles) and we *do continue* with the command as we're in the middle of operation. Also, the metadata profile is applied directly and on the fly on find_config_tree_* fn call and even if the metadata profile is found incorrect, we still need to return the non-profiled value as found in the other configuration provided or default value. To exit immediately even in this case, we'd need to refactor existing find_config_tree_* fns so they can return error. Currently, these fns return only config values (which end up with default values in the end if the config is not found). - To check the profile validity before use to be sure it's correct, one can use : lvm dumpconfig --commandprofile/--metadataprofile ProfileName --validate (the --commandprofile/--metadataprofile for dumpconfig will come as part of the subsequent patch) - This patch also adds a reference to --commandprofile and --metadataprofile in the cmd help string (which was missing before for the --profile for some commands). We do not mention --profile now as people should use --commandprofile or --metadataprofile directly. However, the --profile is still supported for backward compatibility and it's translated as: --profile == --metadataprofile for lvcreate, vgcreate, lvchange and vgchange (as these commands are able to attach profile to metadata) --profile == --commandprofile for all the other commands (--metadataprofile is not allowed there as it makes no sense) - This patch also contains some cleanups to make the code handling the profiles more readable...
2014-05-20 16:13:10 +04:00
return r;
}
const char *get_default_devices_cache_dir_CFG(struct cmd_context *cmd, struct profile *profile)
{
static char buf[PATH_MAX];
if (dm_snprintf(buf, sizeof(buf), "%s/%s", cmd->system_dir, DEFAULT_CACHE_SUBDIR) < 0) {
log_error("Persistent cache directory name too long.");
return NULL;
}
return dm_pool_strdup(cmd->mem, buf);
}
const char *get_default_devices_cache_CFG(struct cmd_context *cmd, struct profile *profile)
{
const char *cache_dir = NULL, *cache_file_prefix = NULL;
static char buf[PATH_MAX];
/*
* If 'cache_dir' or 'cache_file_prefix' is set, ignore 'cache'.
*/
if (find_config_tree_node(cmd, devices_cache_dir_CFG, profile))
cache_dir = find_config_tree_str(cmd, devices_cache_dir_CFG, profile);
if (find_config_tree_node(cmd, devices_cache_file_prefix_CFG, profile))
cache_file_prefix = find_config_tree_str_allow_empty(cmd, devices_cache_file_prefix_CFG, profile);
if (cache_dir || cache_file_prefix) {
if (dm_snprintf(buf, sizeof(buf),
"%s%s%s/%s.cache",
cache_dir ? "" : cmd->system_dir,
cache_dir ? "" : "/",
cache_dir ? : DEFAULT_CACHE_SUBDIR,
cache_file_prefix ? : DEFAULT_CACHE_FILE_PREFIX) < 0) {
log_error("Persistent cache filename too long.");
return NULL;
}
return dm_pool_strdup(cmd->mem, buf);
}
if (dm_snprintf(buf, sizeof(buf), "%s/%s/%s.cache", cmd->system_dir,
DEFAULT_CACHE_SUBDIR, DEFAULT_CACHE_FILE_PREFIX) < 0) {
log_error("Persistent cache filename too long.");
return NULL;
}
return dm_pool_strdup(cmd->mem, buf);
}
const char *get_default_backup_backup_dir_CFG(struct cmd_context *cmd, struct profile *profile)
{
static char buf[PATH_MAX];
if (dm_snprintf(buf, sizeof(buf), "%s/%s", cmd->system_dir, DEFAULT_BACKUP_SUBDIR) == -1) {
log_error("Couldn't create default backup path '%s/%s'.",
cmd->system_dir, DEFAULT_BACKUP_SUBDIR);
return NULL;
}
return dm_pool_strdup(cmd->mem, buf);
}
const char *get_default_backup_archive_dir_CFG(struct cmd_context *cmd, struct profile *profile)
{
static char buf[PATH_MAX];
if (dm_snprintf (buf, sizeof(buf), "%s/%s", cmd->system_dir, DEFAULT_ARCHIVE_SUBDIR) == -1) {
log_error("Couldn't create default archive path '%s/%s'.",
cmd->system_dir, DEFAULT_ARCHIVE_SUBDIR);
return NULL;
}
return dm_pool_strdup(cmd->mem, buf);
}
const char *get_default_config_profile_dir_CFG(struct cmd_context *cmd, struct profile *profile)
{
static char buf[PATH_MAX];
if (dm_snprintf(buf, sizeof(buf), "%s/%s", cmd->system_dir, DEFAULT_PROFILE_SUBDIR) == -1) {
log_error("Couldn't create default profile path '%s/%s'.",
cmd->system_dir, DEFAULT_PROFILE_SUBDIR);
return NULL;
}
return dm_pool_strdup(cmd->mem, buf);
}
const char *get_default_activation_mirror_image_fault_policy_CFG(struct cmd_context *cmd, struct profile *profile)
{
return find_config_tree_str(cmd, activation_mirror_device_fault_policy_CFG, profile);
}
int get_default_allocation_thin_pool_chunk_size_CFG(struct cmd_context *cmd, struct profile *profile)
{
const char *str;
uint32_t chunk_size;
if (!(str = find_config_tree_str(cmd, allocation_thin_pool_chunk_size_policy_CFG, profile))) {
log_error(INTERNAL_ERROR "Cannot find configuration.");
return 0;
}
if (!strcasecmp(str, "generic"))
chunk_size = DEFAULT_THIN_POOL_CHUNK_SIZE * 2;
else if (!strcasecmp(str, "performance"))
chunk_size = DEFAULT_THIN_POOL_CHUNK_SIZE_PERFORMANCE * 2;
else {
log_error("Thin pool chunk size calculation policy \"%s\" is unrecognised.", str);
return 0;
}
return (int) chunk_size;
}
int get_default_allocation_cache_pool_chunk_size_CFG(struct cmd_context *cmd, struct profile *profile)
{
return DEFAULT_CACHE_POOL_CHUNK_SIZE * 2;
}