IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
For dm_is_dm_major to determine whether the major number given as
an argument belongs to a DM device, libdm code needs to know what
the actual DM major is to do the comparison.
It may happen that the dm-mod module is not loaded during this
call and so for the completness let's try our best before we start
giving various errors - we can still make use of dm-mod autoloading,
though only since kernels 2.6.36 where this feature was introduced.
Commit 94786a3bbf introduced
another bug - since sscanf needs extra 1 byte for \0.
Since there is no easy way to do a macro evaluation for (PATH_MAX-1)
and string concatation of this number to get resulting (%4095s) - let's
go with easiest path and restore extra byte for 0.
Other option would be to prepare sscanf parsing string in runtime.
But lets resolve it when we look at PATH_MAX handling later...
Add extra safety detection for thin pool transaction id
and query pool status after confirmed message.
In case there is a missmatch, immeditelly abort further
processing.
Avoid playing with +1.
PATH_MAX code needs probably more thinking anyway, since
there is no MAX path in Linux - user may easily create path
with 64kB chars - so 4kB buffer is surelly not enough for
such dirs.
Note:
http://insanecoding.blogspot.cz/2007/11/pathmax-simply-isnt.html
This patch adds a new flag --deferred to dmsetup remove. If this flag is
specified and the device is open, it is scheduled to be deleted on
close.
struct dm_info is extended.
The existing dm_task_get_info() is converted into a wrapper around the
new version dm_task_get_info_with_deferred_remove() so existing binaries
can still use the old smaller structure.
Recompiled code will pick up the new larger structure.
From: Mikulas Patocka <mpatocka@redhat.com>
Using "[ ]" operator together with "&&" (or ",") inside causes the
string list to be matched if and only if all the items given match
the value reported and the number of items also match. This is
strict list matching and the original behaviour we already have.
In contrast to that, the new "{ }" operator together with "&&" inside
causes the string list to be matched if and only if all the items given
match the value reported but the number of items don't need to match.
So we can provide a subset in selection criteria and if the subset
is found, it matches.
For example:
$ lvs -o name,tags
LV LV Tags
lvol0 a
lvol1 a,b
lvol2 b,c,x
lvol3 a,b,y
$ lvs -o name,tags -S 'tags=[a,b]'
LV LV Tags
lvol1 a,b
$ lvs -o name,tags -S 'tags={a,b}'
LV LV Tags
lvol1 a,b
lvol3 a,b,y
So in the example above the a,b is subset of a,b,y and therefore
it also matches.
Clearly, when using "||" (or "#") inside, the { } and [ ] is the
same:
$ lvs -o name,tags -S 'tags=[a#b]'
LV LV Tags
lvol0 a
lvol1 a,b
lvol2 b,c,x
lvol3 a,b,y
$ lvs -o name,tags -S 'tags={a#b}'
LV LV Tags
lvol0 a
lvol1 a,b
lvol2 b,c,x
lvol3 a,b,y
Also in addition to the above feature, fix list with single value
matching when using [ ]:
Before this patch:
$ lvs -o name,tags -S 'tags=[a]'
LV LV Tags
lvol0 a
lvol1 a,b
lvol3 a,b,y
With this patch applied:
$ lvs -o name,tags -S 'tags=[a]'
LV LV Tags
lvol0 a
In case neither [] or {} is used, assume {} (the behaviour is not
changed here):
$ lvs -o name,tags -S 'tags=a'
LV LV Tags
lvol0 a
lvol1 a,b
lvol3 a,b,y
So in new terms 'tags=a' is equal to 'tags={a}'.
2.02.106 added suffixes to some LV uuids in the kernel.
If any of these LVs is activated with 2.02.105 or earlier,
and then a later version is used, the LVs appear invisible and
activation commands fail.
The code now has to check the kernel for both old and new uuids.
Change the help heading from 'Common Fields' to 'Special Fields' for
the fields: selected, help, ?
Remove the code that does 'all' processing with these special fields as
each of them changes the behaviour of the command in an undesirable way.
'lvs -o all,selected' was of course just printing help.
(via internal expansion to 'lv_all,common_all')
and if we ignored the help fields, then '-o common_all' would still
pull in 'selected' and change the way rows were output.
In contrast to per-type reserved values that are applied for all fields
of that type, per-field reserved values are only applied for concrete
field only.
Also add 'struct dm_report_field_reserved_value' to libdm for per-field
reserved value definition. This is defined by field number (an index
in the 'fields' array which is given for the dm_report_init_with_selection
function during report initialization) and the value to use for any
of the specified reserved names.
A field where it has no meaning to do any type of comparison is the
implicit "help" or "?" field. The error given was a bit cryptic
before this patch, the FLD_UNCOMPARABLE flag makes it easier to identify
this situation anywhere in the code and provide much better error message.
This flag can be applied to other fields that may appear in the future -
mostly usable for implicit fields as they always have special purpose
(so we're not exporting it in libdevmapper for now - usual reporting
fields don't need this).
Before this patch:
$ vgs -S help=1
dm_report_object: no data assigned to field help
dm_report_object: no data assigned to field help
(...which is true actually, but let's provide something better...)
With this patch applied:
$vgs -S help=1
Selection field is uncomparable: help.
Selection syntax error at 'help=1'.
$vgs -S '(name=vg && help=1) || vg_size > 1g'
Selection field is uncomparable: help.
Selection syntax error at 'help=1) || vg_size > 1g'.
It's better to have implicit fields at the very end of the output
so users can see them without scrolling back if the list of fields
is long (the "help" is also an implicit field now so it should be
easily visible).
We have "help" and "?" defined as implicit fields now. As such, we
don't need to export these names in libdevmapper (as it was introduced
by commit 7c86131233 within this release).
If anyone uses these field names by mistake, the libdevmapper code can
error out correctly if it detects that the set of explicit field names
(the ones supplied by "fields" arg in dm_report_init/dm_report_init_with_selection)
contains any of the implicit field names (the ones defined internally
by libdevmapper itself).
Making "help" and "?" implicit also simplifies code since the
dm_report_init caller (lvm/dmsetup) doesn't need to check on
dm_report_init return whether "help" or "?" was hit while parsing
fields/sort keys in libdevmapper.
The libdevmapper now sets internal "RH_ALREADY_REPORTED" flag
after it reports the "help" or "?" implicit field. Then libdevmapper
itself checks for this flag in dm_report_object and if found,
the actual reporting is skipped (because the "help" implicit field
was reported instead of the actual report).
Fix gcc warnings:
libdm-report.c:1952:5: warning: "end_op_flag_hit" may be used uninitialized in this function [-Wmaybe-uninitialized]
libdm-report.c:2232:28: warning: "custom" may be used uninitialized in this function [-Wmaybe-uninitialized]
And snap_percent is not 0% in dm < 1.10.0 so
don't test comparison with 0% here.
Implicit fields are fields that are registered with the report
and reported internally by libdevmapper itself (compared to explicit
fields that are registered by the layer above libdevmapper - e.g. LVM,
dmsetup...).
The "selected" field is the implicit field (for now the only one)
that reports the result of the selection. Since the selection itself
is the property of the libdevmapper, the upper layer using dm_report_init
can't register this field itself and it must be done directly at
libdevmapper layer.
The "selected" field is internally registered as part of the "common"
report type with id 0x80000000 (the last bit in uin32_t) which is then
reserved (the explicit report types are then checked if they do not
contain this id and if yes, we error out).
This way, the "selected" field is recognized by all libdevmapper users
that initialize the reporting with "dm_report_init_with_selection".
If reporting is initialized with the classical "dm_report_init",
there's no functional change (so the "selected" field is not defined
and it's not recognized).
Make dm_report_init_with_selection to accept an argument with an
array of reserved values where each element contains a triple:
{dm report field type, reserved value, array of strings representing this value}
When the selection is parsed, we always check whether a string
representation of some reserved value is not hit and if it is,
we use the reserved value assigned for this string instead of
trying to parse it as a value of certain field type.
This makes it possible to define selections like:
... --select lv_major=undefined (or -1 or unknown or undef or whatever string representations are registered for this reserved value in the future)
... --select lv_read_ahead=auto
... --select vg_mda_copies=unmanaged
With this, each time the field value of certain type is hit
and when we compare it with the selection, we use the proper
value for comparison.
For now, register these reserved values that are used at the moment
(also more descriptive names are used for the values):
const uint64_t _reserved_number_undef_64 = UINT64_MAX;
const uint64_t _reserved_number_unmanaged_64 = UINT64_MAX - 1;
const uint64_t _reserved_size_auto_64 = UINT64_MAX;
{
{DM_REPORT_FIELD_TYPE_NUMBER, _reserved_number_undef_64, {"-1", "undefined", "undef", "unknown", NULL}},
{DM_REPORT_FIELD_TYPE_NUMBER, _reserved_number_unmanaged_64, {"unmanaged", NULL}},
{DM_REPORT_FIELD_TYPE_SIZE, _reserved_size_auto_64, {"auto", NULL}},
NULL
}
Same reserved value of different field types do not collide.
All arrays are null-terminated.
The list of reserved values is automatically displayed within
selection help output:
Selection operands
------------------
...
Reserved values
---------------
-1, undefined, undef, unknown - Reserved value for undefined numeric value. [number]
unmanaged - Reserved value for unmanaged number of metadata copies in VG. [number]
auto - Reserved value for size that is automatically calculated. [size]
Selection operators
-------------------
...
When the field list is displayed as help for constructing selection
criteria, show also the field value type. This is useful for users
to know what set of operators are allowed for the type - the subsequent
"Selection operands" section in the help output summarize all known
types that can be used in selection.
The "<lvm command> -S/--select help" shows help (including list of fields to match against):
...field list here including the field type name...
Selection operands
------------------
field - Reporting field.
number - Non-negative integer value.
size - Floating point value with units specified.
string - Characters quoted by ' or " or unquoted.
string list - Strings enclosed by [ ] and elements delimited by either
"all items must match" or "at least one item must match" operator.
regular expression - Characters quoted by ' or " or unquoted.
Selection operators
-------------------
Comparison operators:
=~ - Matching regular expression.
!~ - Not matching regular expression.
= - Equal to.
!= - Not equal to.
>= - Greater than or equal to.
> - Greater than
<= - Less than or equal to.
< - Less than.
Logical and grouping operators:
&& - All fields must match
, - All fields must match
|| - At least one field must match
# - At least one field must match
! - Logical negation
( - Left parenthesis
) - Right parenthesis
[ - List start
] - List end
Selection list items are enclosed in '[' and ']' (if there's only
one item, the '[' and ']' can be omitted). Each element of the list
is a string (either quoted or unquoted, like the usual string operand
used in selection) and each element is delimited either by conjunction
(meaining "match all") or disjunction operator (meaning "match any").
For example, if "," is the conjuction operator and "/" is the
disjunction operator then:
lv_tags=[a,b,c]
...will match all fields where tags contain *all* a, b and c.
lv_tags=[a/b/c]
...will match all fields where tags contain *any* of a, b, or c.
Mixing operators within the list is not supported:
lv_tags=[a,b/c]
...will give an error.
The order in which items are defined in the selection do not matter.
This patch enhances the selection parsing functionality to recognize
such lists.
The {pv,vg,lv,seg}_tags and lv_modules fields are reported as string
lists using the new dm_report_field_string_list - so we just pass
the list to the fn that takes care of reporting and item sorting itself.
Add a separate dm_report_field_string_list fn to libdevmapper to
support reporting string lists. Before, the code used libdevmappers's
dm_report_field_string fn which required formatting the list to a
single string. This functionality is now moved to libdevmapper
and the code that needs to report the string list just needs
to pass the list itself and libdevmapper will take care of this.
This also enhances code reuse.
The dm_report_field_string_list also accepts an argument to define
custom delimiter to use. If not defined, a default "," (comma) is
used as item delimiter in the string list reported.
The dm_report_field_string_list automatically sorts the items in
the list before formatting it to a final string. It also encodes
the position and length within the final string where each element
can be found. This can be used to support checking against each
list item reported since since when formatted as a single string
for the actual report, we would lose this information otherwise
(we don't want to copy each item, the position and length within
the final string is enough for us to get the original items back).
When such lists are checked against the selection tree, we can check
each item individually this way and we can support operators like
"match any" and "match all".
The list of strings is used quite frequently and we'd like to reuse
this simple structure for report selection support too. Make it part
of libdevmapper for general reuse throughout the code.
This also simplifies the LVM code a bit since we don't need to
include and manage lvm-types.h anymore (the string list was the
only structure defined there).
This is rebased and edited version of the original design and
patch proposed by Jun'ichi Nomura:
http://www.redhat.com/archives/dm-devel/2007-April/msg00025.html
The dm_report_init_with_selection is the same as dm_report_init
but it contains an additional argument to set the selection
in the form of a string that contains field names to check against and
selection operators. The selection string is parsend and a selection
tree is composed for use in the checks against individual fields when
the report is processed. The parsed selection tree is stored in dm_report
structure as "selection_root".
This is rebased and edited version of the original design and
patch proposed by Jun'ichi Nomura:
http://www.redhat.com/archives/dm-devel/2007-April/msg00025.html
Add support for parsing numbers, strings (quoted or unquoted), regexes
and operators amogst these operands in selection condition supplied.
This is rebased and edited version of the original design and
patch proposed by Jun'ichi Nomura:
http://www.redhat.com/archives/dm-devel/2007-April/msg00025.html
This patch defines operators and structures that will be used
to store the report selection against which the actual values
reported will be checked.
Selection operators
-------------------
Comparison operators:
=~ - Matching regular expression.
!~ - Not matching regular expression.
= - Equal to.
!= - Not equal to.
>= - Greater than or equal to.
> - Greater than
<= - Less than or equal to.
< - Less than.
Logical and grouping operators:
&& - All fields must match
, - All fields must match
|| - At least one field must match
# - At least one field must match
! - Logical negation
( - Left parenthesis
) - Right parenthesis
This makes it easier to check against the fields (following patches for
report selection) and check whether size units are allowed or not
with the field value.
As part of better error handling, remove DM devices that have been
sucessfully created but failed to load a table. This can happen
when pvmove'ing in a cluster and the cluster mirror daemon is not
running on a remote node - the mapping table failing to load as a
result. In this case, any revert would work on other nodes running
cmirrord because the DM devices on those nodes did succeed in loading.
However, because no table was able to load on the non-cmirrord nodes,
there is no table present that points to what needs to be reverted.
This causes the empty DM device to remain on the system without being
present in any LVM representation.
This patch should only be considered a partial fix to the overall
problem. This is because only the device which failed to load a
table is removed. Any LVs that may have been loaded as requirements
to the DM device that failed to load may be left in place. Complete
clean-up will require tracking those devices which have been created
as dependencies and removing them along with the device that failed
to load a table.
Share DM_REPORT_FIELD_RESERVED_NAME_{HELP,HELP_ALT} between libdm and
any libdm user to handle reserved field names, in this case the virtual
field name to show help instead of failing on unrecognized field.
The libdm user also needs to check the field name so it can fire
proper code in this case (cleanup, exit etc.).
If there ever would be a second call to dm_lib_init()
and envvar would be improperly set, some last set value
would be used while it should reset to default mangling mode.
When the node enters dtree with implicit dependency, it
automatically has udev flags from parent node
and could not be changed later when the node has been
entered again via i.e lvm's preload tracking.
Resolve this by tracking whether the node has been
created by implicit dependency tracking or has been
entered explicitely. Implicit node could be later
upgraded by an explicit _add_dev() with proper udev_flags.
For implicit devices add special udev flags to avoid
any scan and udev rule processing if we resume such device.
Patch allows easier removing of orphan nodes.
We need to use "--verifyudev" for dmsetup mangle command used in
the name-mangling test since without the --verifyudev, we'd end up
with the failed rename.
Also, add direct check for the dev nodes - node with old name must
be gone and node with new name must be present. Before, we checked
just the output of the command.
One bug popped up here when renaming with udev and libdevmapper
fallback checking the udev when target mangle mode is "none"
(fixme added in the libdevmapper's node rename code).
Reuse _node_send_messages for just checking
for valid transaction_id with preload.
This allows earlier detection of incosistent thin pool.
Code does the same thing, except for sending messages.
Improve testing of transation_id to not allow other difference
then either kernel TID is equal or is lower by oned and there
are queued messages for transaction.
Mark messages as submitted if the transaction_id is already matching.
Do not try to deactivate node on failure here and leave it on
proper error path of the caller.
Deactivation of top level node has to happen,
before traversing subtree.
Swap list logic and rather append new nodes to the head
and then use normal iteration.
(in-release update)
Avoid introducing libdm structure allocated in library user.
Use direct call with all currently supported args.
When new arg is added, new function will cover it.
I am reverting the commit below - removing the new 'dm_config_get_int'
function and simply calling 'dm_config_get_uint32' while casting the
'int *' pointer parameter.
Commit being reverted:
commit 94377dfd5e
Author: Jonathan Brassow <jbrassow@redhat.com>
Date: Mon Jan 27 05:26:19 2014 -0600
Misc: New function for reading lvm config file fields
Introduce 'dm_config_get_int', which will be used by the upcoming
cachepool segment type.
This patch defines a structure for holding all of the device-mapper
cache target's status information. The associated function provides
an easy way for higher levels (LVM) to consume the information.
This patch finishes the device-mapper interface for the cache and
cachepool segment types (i.e. the cache target).
This patch adds the cache segment type - the second of two necessary
to create cache logical volumes. This segment type references the
cachepool (the small fast device) and the origin (the large slow device);
linking them to create the cache device. The cache device is the
hierarchical device-mapper device that the user ulitmately makes use
of.
The cache segment sources the information necessary to construct the
device-mapper cache target from the origin and cachepool segments to
which it links.
This patch adds the new cachepool segment type - the first of two
necessary to eventually create 'cache' logical volumes. In addition
to the new segment type, updates to makefiles, configure files, the
lv_segment struct, and some necessary libdevmapper flags.
The cachepool is the LV and corresponding segment type that will hold
all information pertinent to the cache itself - it's size, cachemode,
cache policy, core arguments (like migration_threshold), etc.
Revert activated volumes if callback fails.
This is currently used only for thin_check failure support.
When thin_check detects failure in thin metadata device, it deactivate
volumes in reversed order that have been preloaded for thin pool activation.
After this change lvm command will not leave active pool subvolumes
in dm table.
Pass dnode pointer instead of rather unknown child pointer.
The pointer is currently unused and passing child pointer
is quite undefined, while dnode has at least some usability.
Add internal error warning when string value is used
as sort value for numerical field.
Using log_warn since the function itself does not return error,
so we do not confuse log_error() checker.
On modern systems udev manages nodes in /dev/mapper directory.
It creates, deletes and renames the nodes according to the
state of the kernel driver.
When the dmsetup is compiled without udev support (--enable-udev_sync)
and runs on the system with running udevd it tries to manage nodes in
/dev/mapper too, so it can race with udev.
dmsetup checks if the node was created/deleted/renamed with the stat
syscall, and skips the operation if it was. However, if udev
creates/deletes/renames the node after the stat syscall and before the
mknod/unlink/rename syscall, dmsetup reports an error.
Since in the system everything happened as expected, skip reporting
error for such case.
These races can be easily provoked by inserting sleep at appropriate
places.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
This file may be included by other programs, so it should be compliant
with the C standard.
* use __linux__ instead of linux - __linux__ is always defined, linux is
not defined when gcc runs in standard-compliant mode (with -std=c89 or
-std=c99) because the C standard doesn't allow polluting namespace
with arbitrary defines.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Just like we have symbolic names assigned to general DM udev flags
(DM_UDEV_* flags), we have the same for any subsystem flags now
(DM_SUBSYSTEM_UDEV_FLAG*), making it easier to use.
When images and their associated metadata are removed from a RAID1 LV,
the remaining sub-LVs are "shifted" down to fill the gaps. For
example, if there is a 3-way mirror:
[0][1][2]
and we remove device#0, the devices will be shifted down
[1][2]
and renamed.
[0][1]
This can create a problem for resume_lv (specifically,
dm_tree_activate_children) during the renaming process though. This
is because it will attempt to rename the higher indexed sub-LVs first
and find that it cannot because there are currently other sub-LVs with
that name. The solution is to check for a conflicting name before
attempting to rename. If a conflict is found and that conflicting
sub-LV is also in the process of renaming, we can defer the current
rename until the conflicting sub-LV has renamed and cleared the
conflict.
Now that resume_lv can handle these types of rename conflicts, we can
remove the workaround in RAID that was attempting to resume a RAID1
LV from the bottom-up in order to force a proper rename in assending
order before attempting a resume on the top-level LV. This "hack"
only worked for single machine use-cases of LVM. Clearing this up
paves the way for exclusive activation of RAID LVs in a cluster.
Do not allow passing '' names to kernel.
This test was missing also in kernel, so it has allowed
to create device with '' name. This then confused dmsetup tool,
since such name is unexpected and unsupported. To remove
such name from table, user has to use -j -m to specify which device
should be removed.
This patch fixes the posibility to run this operation:
dmsetup rename existingdev ''
after this operation commands like 'dmsetup table' are failing.
This patch prohibits to use such name.
Recent kernels allow messages to respond with a string.
Add dm_task_get_message_response() to libdevmapper to perform some
basic sanity checks and return this.
Have 'dmsetup message' display any response.
DM statistics will make extensive use of this.
(From Mikulas.)
libdm-common.c:883:42: warning: pointer/integer type mismatch in conditional expression
define log_sys_error(x, y) log_err("%s%s%s failed: %s", y, *y ? ": " : "", x, strerror(errno))
So the "y" which was 'path ? : "SELinux context reset"' from
previous commit did not quite fit the other "? :" in the log_sys_macro.
- null_fd resource leak on error path in _reopen_fd_null fn
- dead code in verify_message in clvmd code
- dead code in _init_filter_components in toolcontext code
- null dereference in dm_prepare_selinux_context on error path if
setfscreatecon fails while resetting SELinux context
Support tests with abort when libdm encounters internal
error - i.e. for dmsetup tool.
Code execution will be aborted when
env var DM_ABORT_ON_INTERNAL_ERRORS is set to 1
When resuming a node needed by a higher layer of the tree,
if the resume fails, only remove it if the node did not
originally have a live table.
Ref. 97f8454ecc
Clear send_messages flag when they have been delivered successfully.
There is no need to validate it for all other activations of the same
node in the dm_tree.
Also add extra debug message which shows the reason for skipping
sending of messages because the transaction_id has already the matching
value.
Show 'at' pointer address with pool name.
It's useful for debugging to be able to locate pointer address in the
debug trace log. It's only available when compiled with extra debug
compilation flag DEBUG_POOL in make.tmpl.
This patch adds the ability to set the minimum and maximum I/O rate for
sync operations in RAID LVs. The options are available for 'lvcreate' and
'lvchange' and are as follows:
--minrecoveryrate <Rate> [bBsSkKmMgG]
--maxrecoveryrate <Rate> [bBsSkKmMgG]
The rate is specified in size/sec/device. If a suffix is not given,
kiB/sec/device is assumed. Setting the rate to 0 removes the preference.
This patch may not be fully correct. It tries to solve
the imbalanced suspend counter.
The problem starts when some LV is created and fails in resume path.
(i.e. resuming to large PV (enforced) over small loop devices)
This fails in _resume_node() after dm_task_run(). And while
existing device with empty table is left in inactive table,
further calls are reporting this device is in suspend state.
When later the lvm2 tries to rollback created device and deactivate it,
it will end with internal error, when we try to decrement
never incremented suspend counter.
As an 'easy fix' for now update suspend counter only for live nodes.
TODO: explore better fix.
Since we use get_status also in dmeventd, which may use one pool
for a single device, in case it would be repeatedly returning error,
it may not be freeing the pool and would cause slow but steady growth.
To stay safe in the error path release any allocated memory.
To detect mounted device, use also /proc/self/mountinfo
as so far the check was only able to detect ext4 mounted filesystem.
TODO:
Once proper testing for this feature is added, it may appear,
mountinfo check is enough and covers all cases and sysfs check
could be removed.
'lvchange' is used to alter a RAID 1 logical volume's write-mostly and
write-behind characteristics. The '--writemostly' parameter takes a
PV as an argument with an optional trailing character to specify whether
to set ('y'), unset ('n'), or toggle ('t') the value. If no trailing
character is given, it will set the flag.
Synopsis:
lvchange [--writemostly <PV>:{t|y|n}] [--writebehind <count>] vg/lv
Example:
lvchange --writemostly /dev/sdb1:y --writebehind 512 vg/raid1_lv
The last character in the 'lv_attr' field is used to show whether a device
has the WriteMostly flag set. It is signified with a 'w'. If the device
has failed, the 'p'artial flag has priority.
Example ("nosync" raid1 with mismatch_cnt and writemostly):
[~]# lvs -a --segment vg
LV VG Attr #Str Type SSize
raid1 vg Rwi---r-m 2 raid1 500.00m
[raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m
[raid1_rimage_1] vg Iwi---r-w 1 linear 500.00m
[raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m
[raid1_rmeta_1] vg ewi---r-- 1 linear 4.00m
Example (raid1 with mismatch_cnt, writemostly - but failed drive):
[~]# lvs -a --segment vg
LV VG Attr #Str Type SSize
raid1 vg rwi---r-p 2 raid1 500.00m
[raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m
[raid1_rimage_1] vg Iwi---r-p 1 linear 500.00m
[raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m
[raid1_rmeta_1] vg ewi---r-p 1 linear 4.00m
A new reportable field has been added for writebehind as well. If
write-behind has not been set or the LV is not RAID1, the field will
be blank.
Example (writebehind is set):
[~]# lvs -a -o name,attr,writebehind vg
LV Attr WBehind
lv rwi-a-r-- 512
[lv_rimage_0] iwi-aor-w
[lv_rimage_1] iwi-aor--
[lv_rmeta_0] ewi-aor--
[lv_rmeta_1] ewi-aor--
Example (writebehind is not set):
[~]# lvs -a -o name,attr,writebehind vg
LV Attr WBehind
lv rwi-a-r--
[lv_rimage_0] iwi-aor-w
[lv_rimage_1] iwi-aor--
[lv_rmeta_0] ewi-aor--
[lv_rmeta_1] ewi-aor--
Revert commit 31c24dd9f2. This commit
was used to force a RAID device-mapper table to be loaded into the
kernel despite the fact that it was identical to the one already
loaded. The effect allowed a RAID array with a transiently failed
device to refresh and reintegrate the failed device. This operation
is better done in the kernel on a 'resume'. Since,
'lvchange --refresh' already performs a suspend/resume cycle, the
above commit is not needed once the kernel change is made. Reverting
the commit removes an unnecessary (at least for now) change to the
device-mapper interface.
I've updated the dm_status_raid structure and dm_get_status_raid()
function to make it handle the new kernel status fields that will
be coming in dm-raid v1.5.0. It is backwards compatible with the
old status line - initializing the new fields to '0'. The new
structure is also more amenable to future changes. It includes a
'reserved' field that is currently initialized to zero but could
be used to hold flags describing new features. It also now uses
pointers for the character strings instead of attempting to allocate
their space along with the structure (causing the size of the
structure to be variable). This allows future fields to be appended.
The new fields that are available are:
- sync_action : shows what the sync thread in the kernel is doing
(idle, frozen, resync, recover, check, repair, or
reshape)
- mismatch_count: shows the number of discrepancies which were
found or repaired by a "check" or "repair"
process, respectively.
Previous commit included changes to WHATSNEW, but the code changes
were missing. Here is the description from the previous commit:
commit bbc6378b73
Author: Jonathan Brassow <jbrassow@redhat.com>
Date: Thu Feb 21 11:31:36 2013 -0600
RAID: Make 'lvchange --refresh' restore transiently failed RAID PVs
A new function (dm_tree_node_force_identical_table_reload) was added to
avoid the suppression of identical table reloads. This allows RAID LVs
to reload the on-disk superblock information that contains which devices
have failed and the bitmaps. If the failed device has returned, this has
the effect of restoring the device and initiating recovery. Without this
patch, the user had to completely deactivate their RAID LV and re-activate
it in order to restore the failed device. Now they simply need to
suspend and resume (which is done by 'lvchange --refresh').
The identical table suppression is only avoided if the LV is not PARTAIL
(i.e. all of it's devices can be seen and read by LVM) and the kernel
status of the array contains failed devices. In other words, the function
will only be called in the case where we may have success in restoring
a failed device in the array.
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.
Export this functionality from libdevmapper just for
convenience and general use when reading boolean values
which could be defined either in a numeric way with 0/1
or by using strings with "true"/"false", "yes"/"no",
"on"/"off", "y"/"n".
When a section was empty in a configuration tree (no children - this is
allowed) and we were looking for a config node inside that section, the
_find_config_node function incorrectly returned the section itself if
the node inside that section was not found.
For example the configuration below:
The config:
abc {
}
And a function call to get the "def" node inside "abc" section:
_find_config_node(..., "abc/def")
...returned the "abc" node instead of NULL ("def" not found).
This in turn caused segfaults in the code using lookups in such
a configuration tree as we (correctly) expected that the node
returned was always the one we were looking for or NULL if not
found. But if incorrect node was returned instead, we processed
that as if this was the node we were looking for and so we
processed its value as well. But sections don't have values => segfault.
On glibc, those are erroneously (namespace pollution) pulled in via
other headers. this doesn't work with conformant libcs (musl libc in
this case), we simply need to include all needed headers.
Signed-Off-By: John Spencer <maillist-lvm@barfooze.de>
This patch fixes problem reported here:
https://www.redhat.com/archives/dm-devel/2013-January/msg00311.html
Fixing it by separating function for duplicating string token.
---
When /etc/lvm/lvm.conf is truncated at the first '"' of a line, all LVM
utilities crash with a segfault.
The segfault only seems to occur if the last character is the first '"'
(double quote) of a line. If you truncate it at any other point, lvm
detects the error and report parse error
lvm.conf ends like this.
$hexdump -C lvm.conf
....
69 72 20 3d 20 22 2f 64 65 76 22 0a 0a 0a 20 20 |ir = "/dev"... |
20 20 23 20 41 6e 20 61 72 72 61 79 20 6f 66 20 | # An array of |
64 69 72 65 63 74 6f 72 69 65 73 20 74 68 61 74 |directories that|
20 63 6f 6e 74 61 69 6e 20 74 68 65 20 64 65 76 | contain the dev|
69 63 65 20 6e 6f 64 65 73 20 79 6f 75 20 77 69 |ice nodes you wi|
73 68 0a 20 20 20 20 23 20 74 6f 20 75 73 65 20 |sh. # to use |
77 69 74 68 20 4c 56 4d 32 2e 0a 20 20 20 20 73 |with LVM2.. s|
63 61 6e 20 3d 20 5b 20 22 2f 78 22 2c 0a 20 20 |can = [ "/x",. |
20 20 20 20 20 20 20 20 20 20 20 22 | "|
...
Reported-by: dongmao zhang <dmzhang suse com>
Similar to the way thin* accesses its kernel status, we add a method
for RAID to grab the various values in its status output without the
higher levels (LVM) having to understand how to parse the output.
Added functions include:
- lib/activate/dev_manager.c:dev_manager_raid_status()
Pulls the status line from the kernel
- libdm/libdm-deptree.c:dm_get_status_raid()
Parses status line and puts components into dm_status_raid struct
- lib/activate/activate.c:lv_raid_dev_health()
Accesses dm_status_raid to deliver raid dev_health string
The new structure and functions can provide a more unified way to access
status information. ('lv_raid_percent' could switch to using these
functions, for example.)
Add log/debug_classes to lvm.conf to allow debug messages to be
classified and filtered at runtime.
The dm_errno field is only used by log_error(), so I've redefined it
for log_debug() messages to hold the message class.
By default, all existing messages appear, but we can add categories that
generate high volumes of data, such as logging all traffic to/from
lvmetad.
If the resume of preloaded node fails, do not leave such
node in the table - since it may not be easy to detach such
node later when the node is i.e. internal.
i.e. failing activation of the thin pool with mismatching
chunk size may leave -tpool device in the table, which
could have been then removed only by dmsetup command.
$ export DM_DISABLE_UDEV=1
$ dmsetup create test --table "0 1 zero"
Udev is running and DM_DISABLE_UDEV environment variable is set. Bypassing udev, device-mapper library will manage device nodes in device directory.
$ lvchange -ay vg/lvol0
Udev is running and DM_DISABLE_UDEV environment variable is set. Bypassing udev, LVM will manage logical volume symlinks in device directory.
Udev is running and DM_DISABLE_UDEV environment variable is set. Bypassing udev, LVM will obtain device list by scanning device directory.
Udev is running and DM_DISABLE_UDEV environment variable is set. Bypassing udev, device-mapper library will manage device nodes in device directory.
Setting this environment variable will cause a full fallback
to old direct node and symlink management in libdevmapper and lvm2.
It means:
- disabling udev synchronization
(--noudevsync in dmsetup and --noudevsync + activation/udev_sync=0
lvm2 config)
- disabling dm and any subsystem related udev rules
(--noudevrules in dmsetup and activation/udev_rules=0 lvm2 config)
- management of nodes/symlinks under /dev directly by libdevmapper/lvm2
(--verifyudev in dmsetup and activation/verify_udev_operations=1
lvm2 config)
- not obtaining any device list from udev database
(devices/obtain_device_list_from_udev=0 lvm2 config)
Note: we could set all of these before - there's no functional change!
However the DM_DISABLE_UDEV environment variable is a nice shortcut
to make it easier for libdevmapper users so that one can switch off all
of the udev management off at one go directly on the command line,
without a need to modify any source or add any extra switches.
cookie_set variable found in the struct dm_task should be always
set to 1 after dm_task_set_cookie_call, even if udev_sync is disabled
as the cookie itself carries synchronization informations *as well as*
extra flags to control other aspects of udev support.
For example, one could disable the synchronization itself, but still
direct the libdm code to disable library fallback via
DM_UDEV_DISABLE_LIBRARY_FALLBACK flag. These extra flags still need
to be carried out!
A concrete example:
$ dmsetup create test --table "0 1 zero" --noudevsync
This disables synchronization with udev. As the --verifyudev option is
not used, we don't want to do any corrections. In other words, we
need DM_UDEV_DISABLE_LIBRARY_FALLBACK flag to be used. However,
with --noudevsync this was not the case - the flag was ignored!
This patch fixes the case when noudevsync is used but there are still
some extra flags passed within the cookie flag part. The synchronization
part of the cookie stays zero (which is ok as dm_udev_wait call on such a
cookie is simply a NOOP).
Use log_warn to print non-fatal warning messages.
Use of log_error would confuse checker for testing
whether proper error has been reported for some real error.
On each ioctl return, the device UUID is decoded from \xNN format.
If the UUID of the device being *removed* is malformed (e.g. it
hasn't been corrected before), just remove it without any error
as the UUID is not needed anymore - the device is gone anyway.
Otherwise a misleading error message would be issued just after
the removal:
# dmsetup remove test
The UUID "a b" should be mangled but it contains blacklisted characters.
Command failed
Just like we already have existing mangling support for
device-mapper names, we need exactly the same for device-mapper
UUIDs as their character whitelist is wider than what udev supports.
In case udev is used to create entries in /dev based on UUIDs
and these UUIDs contain characters not supported by udev,
we'll end up with incorrect /dev content for such devices.
So we need to mangle them to a form that is supported by udev.
The mangling used for UUIDs follows the mangling used for names
(that is already supported and used throughout). That means,
setting the name mangling mode via dm_set_name_mangling_mode
affects mangling used for UUIDs in exactly the same manner.
It would be useless to add a new and separate
dm_set_uuid_mangling_mode fn, we'll reuse existing interface.
(un)mangle_name -> (un)mangle_string
check_multiple_mangled_name_allowed -> check_multiple_mangled_string_allowed
Just for clarity as the same functions will be reused to (un)mangle dm UUIDs.
Patch clears the flag if thin pool is stacked over mirror.
Since thin pool could be used to stack device over mirrors,
it needs resume properly i.e. mirrors with corelog which are otherwise
unconditionally skipped (for pvmove functionality).
If we were defining a section (which is a node without a value) and
the value was created automatically on dm_config_create_node call,
we were wasting resources as the next step after creating the config
node itself was assigning NULL for the node's value.
The dm_config_node_create + dm_config_create_value sequence should be
used instead for settings and dm_config_node_create alone for sections.
The majority of the code already used the correct sequence. Though
with dm_config_node_create fn creating the value as well, the pool
memory was being trashed this way.
This patch removes the node value initialization on dm_config_create_node
fn call and keeps it for the direct dm_config_create_value fn call.
This patch adds support for RAID10. It is not the default at this
stage. The user needs to specify '--type raid10' if they would like
RAID10 instead of stacked mirror over stripe.
Adding couple INTERNAL_ERROR reports for unwanted parameters:
Ensure the 'top' metadata node cannot be NULL for lvmetad.
Make obvious vginfo2 cannot be NULL.
Report internal error if handler and vg is undefined.
Check for handle in poll_vg().
Ensure seg is not NULL in dev_manager_transient().
Report missing read_ahead for _lv_read_ahead_single().
Check for report handler in dm_report_object().
Check missing VG in _vgreduce_single().
A regression introduced in 2.02.89 (11e520256b)
caused the lvm dumpconfig <node> to print out
the node as well as its subsequent siblings.
The information about "only_one" mode got lost.
Before this patch (just an example node):
# lvm dumpconfig global/use_lvmetad
use_lvmetad=1
thin_check_executable="/usr/sbin/thin_check"
thin_check_options="-q"
(...all nodes to the end of the section)
With this patch applied:
# lvm dumpconfig global/use_lvmetad
use_lvmetad=1
With latest changes in the udev, some deprecated functions were removed
from libudev amongst which there was the "udev_get_dev_path" function
we used to compare a device directory used in udev and directore set in
libdevmapper. The "/dev" is hardcoded in udev now (udev version >= 183).
Amongst other changes and from packager's point of view, it's also
important to note that the libudev development library ("libudev-devel")
could now be a part of the systemd development library ("systemd-devel")
because of the udev + systemd merge.
Auto mode can't deal with multiple mangled names. We can do that while working
in hex mode, but in auto mode, this would lead to device name ambiguity.
Be more strict when unmangling names on ioctl return - require the name to be
properly mangled in 'auto' and 'hex' mode. There really should not be any
blacklisted character since the names should be renamed already (by means of
renaming it directly or running 'dmsetup mangle' for automatic rename).
if the thin_check fail on thin pool - still return successful deactivation,
since lvremove would currently fail.
TODO: find some way to not run check with lvremove.
Add some hack math to allow 16GB devices to be passed as thinpool metadata.
Since kernel has put in limit to not allow which are just bigger then
some predefined constant in kernel but not matching 16GB so any device bigger
is rejected.
FIXME: Current code still might need more tweaks to be more generic.
Since lvm seems to call driver_version(NULL, 0) this would lead
to crash. Though the combination of the code is probably very hard to hit.
If the user doesn't supply version buffer, just skip printing to buffer.
pvcreate gives
WARNING: Ignoring unsupported value for metadata/pvmetadataignore.
It was warning if there is no config file entry instead of only if the node
exists but is empty.
Should be faster then strncpy - since we could avoid clearing 4KB pages
with each strncpy(...,PATH_MAX).
Also it's easy to check whether string fit - and eventually avoid
to continue working we incomplete string.
If we have good enough glibc to return number of needed chars, do not
loop try to reach good size, but use this size directly for allocation,
saving also last strdup.
Since now we start with 16 bytes - skip buffer realloc for shorter string.
Device-mapper in kernel uses '\' as escape character so it's better
to double it to avoid any confusion when using existing device names
with '\' in the table specification.
For example:
dmsetup create x --table "0 8 linear /dev/mapper/a\x20b 0"
should pass just fine now without a need to explicitly escape the '\' char
like this:
dmsetup create x --table "0 8 linear /dev/mapper/a\\x20b 0"
If dm_task_get_name or dm_task_get_names gets called, these will return
unmangled form of the names so the name mangling stays totally transparent
to any libdevmapper user (unless DM_STRING_MANGLING_NONE is used in which
case the name is not touched and it is is returned as it is in kernel).
For example:
dmsetup create "a b" - will create a\x20b device in kernel and so udev will
create /dev/mapper/a\x20b
dm_task_get_name/names will still return "a b"
In AUTO mode, the libdevmapper user can still query the device by using
the mangled ("a\x20b") or unmangled form of the name when calling dm_task_set_name.
If mangled name is provided, it's detected and the name is kept as it is.
If unmangled name is provided, it will be mangled. IOW in AUTO mode it's
totally transparent and it should not require any changes in the code
using libdevmapper.
However, any libdevmapper user must be aware of the fact that the mangled form
of the name appears in /dev/mapper (udev just can't deal with those blacklisted
characters).
dm_task_get_name_mangled will always return mangled form of the name while
the dm_task_get_name_unmangled will always return unmangled form of the name
irrespective of the global setting (dm_set/get_name_mangling_mode).
This is handy in situations where we need to detect whether the name is already
mangled or not. Also display functions make use of it.
Use the DEV_NAME macro to use the mangled form of the name if present,
use normal name otherwise (we store both forms - mangled and unmangled in
struct dm_task). Mangled form should be always preferred over unmangled
with the exception of the situations where we divide one task into several
others (like "create and load") - we need to avoid mangling the name twice
(because of multiple dm_task_set_name calls)!
If dm_task_set_name/newname is called, the name provided will be
automatically translated to correct encoded form with the hex enconding
so any character not on udev whitelist will be mangled with \xNN
format where NN is hex value of the character used.
By default, the name mangling mode used is the one set during
configure with the '--with-default-name-mangling' option.
This option configures the default name mangling mode used, one of:
AUTO, NONE and HEX.
The name mangling is primarily used to support udev character whitelist
(0-9, A-Z, a-z, #*-.:=@_) so any character that is not on udev whitelist
will get translated into an encoded form \xNN where NN is the hex value
of the character.
It was not possible to pass down the DM_[FORCE|NO]SYNC flags to
'dm_tree_node_add_raid_target'. This meant that converting to 'raid1' from
'mirror' would cause a full resync. (It also meant that '--nosync' was
ineffective when creating a 'raid1' LV.)
I've taken the 'reserved' parameter in 'dm_tree_node_add_raid_target' and
used it for the "flags" parameter. Now it is possible to pass the sync
flags and any other flags that may come up.
In case of zero bytes would be read from sysfs, it would store '\0' on
temp_buf[-1] address.
Simplify some buffer length calculation and use strcpy if we've just
checked string fits in give buffer.
Replace jump label error: with bad: commonly used in libdm.
Replace asserts with test for failing memory allocation.
Add at least stack traces.
Index counter starts from 1 (0 reserved for error), so replacing fingerprint.
Since the function dm_get_next_target() returns NULL as 'next' pointer
so it's not a 'real' error - set 0 to all parameters when NULL is
returned because of missing head.
i.e. one of use case::
do {
next = dm_get_next_target(dmt, next, &start, &length,
&target_type, ¶ms);
size += length;
} while (next);
Using PRELOAD part would lead to problems when the problem
would happen before vg_write and vg_commit.
Also this change is necessary for snapshot creation sequence.
This is accomplished by reading associated sysfs information. For a dm device,
this is /sys/dev/block/major:minor/dm/name (supported in kernel version >= 2.6.29,
for older kernels, the behaviour is the same as for non-dm devices).
For a non-dm device, this is a readlink on /sys/dev/block/major:minor, e.g.
/sys/dev/block/253:0 --> ../../devices/virtual/block/dm-0.
The last component of the path is a proper kernel name (block device name).
One can request to read only kernel names by setting the 'prefer_kernel_name'
argument if needed.
LVM- prefix.
Try harder not to leave stray empty devices around (locally or remotely) when
reverting changes after failures while there are inactive tables.
If we know major:minor number of device (which is known after resume) we will
try to use sysfs to set/get read ahead parameters of device.
This avoid potential problem of blocking commands like 'dmsetup info' awaiting
for device being usable for open/close - i.e. overfilled thin pool may block
such command.
Add dm_get_status_thin_pool and dm_get_status_thin functions to
parse 'params' argument which is received via dm_get_next_target.
Returns filed structure allocated from given mempool.
RAID is not like traditional LVM mirroring. LVM mirroring required failed
devices to be removed or the logical volume would simply hang. RAID arrays can
keep on running with failed devices. In fact, for RAID types other than RAID1,
removing a device would mean substituting an error target or converting to a
lower level RAID (e.g. RAID6 -> RAID5, or RAID4/5 to RAID0). Therefore, rather
than removing a failed device unconditionally and potentially allocating a
replacement, RAID allows the user to "replace" a device with a new one. This
approach is a 1-step solution vs the current 2-step solution.
example> lvconvert --replace <dev_to_remove> vg/lv [possible_replacement_PVs]
'--replace' can be specified more than once.
example> lvconvert --replace /dev/sdb1 --replace /dev/sdc1 vg/lv
Avoid creation of target type name when it's longer then
DM_MAX_TYPE_NAME (noticed by static analyzer where the
sp.target_type might be missing '\0' at the end.)
Before patch:
$> dmsetup create long
0 1000 looooooooooooooooooooooooooong
^D
device-mapper: reload ioctl failed: Invalid argument
After patch:
$> dmsetup create xxx
0 1000 looooooooooooooooooooooooooong
Target type name looooooooooooooooooooooooooong is too long.
Command failed
Remove DM_THIN_ERROR_DEVICE_ID from API.
Remove API warning.
Drop code that was using DM_THIN_ERROR_DEVICE_ID (already commented)
Remove debug message which slipped in through some previous commit.
A little code shuffling and adding support for
DM_THIN_ERROR_DEVICE_ID which might be eventually be used
for activation of thin which is going to be deleted.
For now we do not need it lvm.
Add a new node flag send_messages that is used to simplify
test when to call _node_send_messages().
Add call to _node_send_messages when pool is deeper in the tree.
There should be no need for retry for our internal devices - it would be hinding
our own bug in the tree processing.
Update error messages to show also also device name.
No WHATS_NEW - in release fix.
When DEBUG_MEM is used, the memory is trashed with extra pattern before real
free() is called, and as this memory was marked as non accessible when used with
valgrind, make it again usable.
Certain errno codes could be expected in some situations thus
add experimental support for them.
When expected errno is set after ioctl error - function skips error
printing and exits succefully.
Currently only useful for thin pool messages.
Version 2 of the userspace log protocol accepts return information during the
DM_ULOG_CTR exchange. The return information contains the name of the log
device that is being used (if there is one). The kernel can then register the
device via 'dm_get_device'. Amoung other things, this allows for userspace to
assemble a correct dependency tree of devices - critical for LVM handling of
suspend/resume calls.
Also, update dm-log-userspace.h to match the kernel header associated with
this protocol change. (Includes a version inc.)
The upstream kernel version that this file mirrors has changed, here is the
commit message:
commit 86a54a4802df10d23ccd655e2083e812fe990243
Author: Jonathan Brassow <jbrassow@redhat.com>
Date: Thu Jan 13 19:59:52 2011 +0000
dm log userspace: add version number to comms
This patch adds a 'version' field to the 'dm_ulog_request'
structure.
The 'version' field is taken from a portion of the unused
'padding' field in the 'dm_ulog_request' structure. This
was done to avoid changing the size of the structure and
possibly disrupting backwards compatibility.
The version number will help notify user-space daemons
when a change has been made to the kernel/userspace
log API.
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
The current code does not always assign proper udev flags to sub-LVs (e.g.
mirror images and log LVs). This shows up especially during a splitmirror
operation in which an image is split off from a mirror to form a new LV.
A mirror with a disk log is actually composed of 4 different LVs: the 2
mirror images, the log, and the top-level LV that "glues" them all together.
When a 2-way mirror is split into two linear LVs, two of those LVs must be
removed. The segments of the image which is not split off to form the new
LV are transferred to the top-level LV. This is done so that the original
LV can maintain its major/minor, UUID, and name. The sub-lv from which the
segments were transferred gets an error segment as a transitory process
before it is eventually removed. (Note that if the error target was not put
in place, a resume_lv would result in two LVs pointing to the same segment!
If the machine crashes before the eventual removal of the sub-LV, the result
would be a residual LV with the same mapping as the original (now linear) LV.)
So, the two LVs that need to be removed are now the log device and the sub-LV
with the error segment. If udev_flags are not properly set, a resume will
cause the error LV to come up and be scanned by udev. This causes I/O errors.
Additionally, when udev scans sub-LVs (or former sub-LVs), it can cause races
when we are trying to remove those LVs. This is especially bad during failure
conditions.
When the mirror is suspended, the top-level along with its sub-LVs are
suspended. The changes (now 2 linear devices and the yet-to-be-removed log
and error LV) are committed. When the resume takes place on the original
LV, there are no longer links to the other sub-lvs through the LVM metadata.
The links are implicitly handled by querying the kernel for a list of
dependencies. This is done in the '_add_dev' function (which is recursively
called for each dependency found) - called through the following chain:
_add_dev
dm_tree_add_dev_with_udev_flags
<*** DM / LVM divide ***>
_add_dev_to_dtree
_add_lv_to_dtree
_create_partial_dtree
_tree_action
dev_manager_activate
_lv_activate_lv
_lv_resume
lv_resume_if_active
When udev flags are calculated by '_get_udev_flags', it is done by referencing
the 'logical_volume' structure. Those flags are then passed down into
'dm_tree_add_dev_with_udev_flags', which in turn passes them to '_add_dev'.
Unfortunately, when '_add_dev' is finding the dependencies, it has no way to
calculate their proper udev_flags. This is because it is below the DM/LVM
divide - it doesn't have access to the logical_volume structure. In fact,
'_add_dev' simply reuses the udev_flags given for the initial device! This
virtually guarentees the udev_flags are wrong for all the dependencies unless
they are reset by some other mechanism. The current code provides no such
mechanism. Even if '_add_new_lv_to_dtree' were called on the sub-devices -
which it isn't - entries already in the tree are simply passed over, failing
to reset any udev_flags. The solution must retain its implicit nature of
discovering dependencies and be able to go back over the dependencies found
to properly set the udev_flags.
My solution simply calls a new function before leaving '_add_new_lv_to_dtree'
that iterates over the dtree nodes to properly reset the udev_flags of any
children. It is important that this function occur after the '_add_dev' has
done its job of querying the kernel for a list of dependencies. It is this
list of children that we use to look up their respective LVs and properly
calculate the udev_flags.
This solution has worked for single machine, cluster, and cluster w/ exclusive
activation.
Make limits for thin data_block_size and device_id part of public API.
FIXME: read them possible from some kernel header file in the future ?
But we may need to support different values for different versions ?
