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.
When devices are created - we were not giving meaning error messages
when the failure happened on 'reload' part of creation.
With this patch we are now able to report both name and major:minor.
Enhancment is most visible with 'crypto' devices,
which are using 'secure' memory erase bit.
Adds support for the DM_GET_TARGET_VERSION to dmsetup.
It introduces a new comman "target-version" that will accept list
of targets and print their version.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Analogous to the case of a device with no regions, it is not an
error to attempt to list the stats groups on a device that has no
configured groups: just return success and continue.
Udev debugging is a bit tricky, so to more easily pair cookie ID,
which is the lowest 16 bit - print cookie as hexa number.
This simplify pairing of processed cookies while the 'higher bit flags'
are changed for the same cookie.
Kernels <2.6.27 don't have /sys/dev dir - add code for looking
out device name via longre seach in /sys/block
This makes commands like 'dmsetup dep -o blkdevname' working.
Fix versioning for updated symbols dm_stats_create_region
and dm_stats_create_region.
Only the latest symbol should have global entry.
Since I'm not sure what is currenlty the best option for
old symbols - we added support for easy commenting of them
(so we do not lose information when the symbol appeared
for the first time.)
Note: some old already deleted symbols should have been
restored as comments.
Recent kernel version from kernel commit:
de7180ff908b2bc0342e832dbdaa9a5f1ecaa33a
started to report in cache status line new flag:
no_discard_passdown
Whenever lvm spots unknown status it reports:
Unknown feature in status:
So add reconginzing this feature flag and also report this with
'lvs -o+kernel_discards'
When no_discard_passdown is found in status 'nopassdown' gets reported
for this field (roughly matching what we report for thin-pools).
lvm2 till version 2.02.169 (commit 78d004efa8)
was printing invalid creation_time argument into metadata on 32bit arch.
However with commit ba9820b142 we started
to properly validate all input numbers and thus we refused to accept
invalid metadata with 'garbage' string - but this results in the
situation where metadata produced on older lvm2 on 32 bit architecture
will become unreadable after upgrade.
To fix this case - extend libdm parser in a way, that whenever we
find error integer value, we also check if the parsed value is not for
creation_time node and in this case we let the metadata pass through
with made-up date 2018-05-24 (release date of 2.02.169).
Use the correct loop variable within the loop, instead of reusing the
initial value. Table lines after the first don't get terminated in
the right place.
Signed-off-by: Kurt Garloff <kurt@garloff.de>
Since configure.h is a generated header and it's missing traditional
ifdefs preambule - it can be included & parsed multiple times.
Normally compiler is fine when defines have same value and there is
no warning - yet we don't need to parse this several times
and by adding -include directive we can ensure every file
in the package is rightly compile with configure.h as the
first header file.
Ensure configure.h is always 1st. included header.
Maybe we could eventually introduce gcc -include option, but for now
this better uses dependency tracking.
Also move _REENTRANT and _GNU_SOURCE into configure.h so it
doesn't need to be present in various source files.
This ensures consistent compilation of headers like stdio.h since
it may produce different declaration.
Just for case ensure compiler is not able to optimize
memset() away for resources that are released.
This idea of using memory barrier is taken from openssl.
Other options would be to check for 'explicit_bzero' function.
When preparing ioctl buffer and flatting all parameters,
add table parameters only to ioctl that do process them.
Note: list of ioctl should be kept in sync with kernel code.
DM_DEVICE_CREATE with table is doing several ioctl operations,
however only some of then takes parameters.
Since _create_and_load_v4() reused already existing dm task from
DM_DEVICE_RELOAD it has also kept passing its table parameters
to DM_DEVICE_RESUME ioctl - but this ioctl is supposed to not take
any argument and thus there is no wiping of passed data - and
since kernel returns buffer and shortens dmi->data_size accordingly,
anything past returned data size remained uncleared in zfree()
function.
This has problem if the user used dm_task_secure_data (i.e. cryptsetup),
as in this case binary expact secured data are erased from main memory
after use, but they may have been left in place.
This patch is also closing the possible hole for error path,
which also reuse same dm task structure for DM_DEVICE_REMOVE.
Check that type is always defined, if not make it explicit internal
error (although logged as debug - so catched only with proper lvm.conf
setting).
This ensures later type being NULL can't be dereferenced with coredump.
Since the stats handle is neither bound nor listed before the
attempt to call dm_stats_get_nr_regions(), it will always return
zero: this prevents reporting of any dmstats regions on any
device.
Remove the dm_stats_get_nr_regions() check and instead rely on
the correct return status from dm_stats_populate() which only
returns 0 in the case that there are regions to inspect (and
which logs a specific error for all other cases).
Reported-by: Bryan Gurney <bgurney@redhat.com>
It doesn't make sense to test or warn about the region count until
the stats handle has been listed: at this point it may or may not
contain valid information (but is guaranteed to be correct after
the list).
Commit 3f35146 added a check on the value returned by the
_display_info_cols() function:
1024 if (!_switches[COLS_ARG])
1025 _display_info_long(dmt, &info);
1026 else
1027 r = _display_info_cols(dmt, &info);
1028
1029 return r;
This exposes a bug in the dmstats code in _display_info_cols:
the fact that a device has no regions is explicitly not an error
(and is documented as such in the code), but since the return
code is not changed before leaving the function it is now treated
as an error leading to:
# dmstats list
Command failed.
When no regions exist.
Set the return code to the correct value before returning.
The device-mapper directory now holds a copy of libdm source. At
the moment this code is identical to libdm. Over time code will
migrate out to appropriate places (see doc/refactoring.txt).
The libdm directory still exists, and contains the source for the
libdevmapper shared library, which we will continue to ship (though
not neccessarily update).
All code using libdm should now use the version in device-mapper.
As we start refactoring the code to break dependencies (see doc/refactoring.txt),
I want us to use full paths in the includes (eg, #include "base/data-struct/list.h").
This makes it more obvious when we're breaking abstraction boundaries, eg, including a file in
metadata/ from base/
Recent changes allow some major simplification of the way
lvmcache works and is used. lvmcache_label_scan is now
called in a controlled fashion at the start of commands,
and not via various unpredictable side effects. Remove
various calls to it from other places. lvmcache_label_scan
should not be called from anywhere during a command, because
it produces an incorrect representation of PVs with no MDAs,
and misclassifies them as orphans. This has been a long
standing problem. The invalid flag and rescanning based on
that is no longer used and removed. The 'force' variation is
no longer needed and removed.
btrfs is using fake major:minor device numbers.
try to be smarter and detect used node via DM device name.
This shortens delays, where i.e. lvm2 is asked to deactivate
volume with mounted btrfs as such operation is not retryed
and user is informed about device being in use.
When function dm_stats_populate() returns 0 it's an error and needs
log_error() message - function can't have 'success' returning 0 or
error without reasons.
When doing resume, directly pass location where new updated info
needs to be stored.
_resume_node() ensures the info is ONLY updated when the function
is successful and never changes it on error path.
Update the logic towards more explicit logic.
Preload tree normally does not want to resume, only
in certain cases of extension or new loaded nodes can be
resumed. So introduce new internal variable delay_resume_if_extended
controlable by target.
Patch itself is not changing current existing behaviour,
and rather documents existing problem in more readable way.
lvm2 needs to introduce explicit mechanism how to support more
fain-grained (and safe) logic to i.e. resize thin-pool which
can be sitting on cached raid volume.
Variable props.send_messages has 3 states and was not used properly
here. Activation in this moment does not need to verify thin-pool status
as that has been already checked on preload.
So only if there are some real messages (value 2) call function
for sending them.
Code already has dereferenced UUID before this point,
and its already given we require name & uuid when ading new node
(although uuid could be empty string).
Just like everywhere else - use single if() for major:minor setup
(it basically can't fail as of today anyway)
Always leave funtion with correctly set pointers even on error path.
Replicator never really existed in upstream kernel and its support
got deprecated.
Also its support never got finished so no code is supposed to be
using it anyway.
Libdm symbols are remaining, just the implementation will always
return failure - so any user of:
dm_tree_node_add_replicator_dev_target()
dm_tree_node_add_replicator_target().
will now always recieve error message.
Separate handling of error code from _info_by_dev.
This error can only happeng when we are running out of memory.
In such case there is urgent need to stop any futher proceeding
of command and run to error ASAP.
ATM we want to support delayed resume purely in pvmove case.
So have libdm logic internal to recognize difference beween
pvmove and other targets that do use delayed resume.
This fixes problem introduced with commit aa68b898ff
for mirror-on-mirror or snapshot-on-mirror problem.
TODO: likely added new API call and let libdm user select
delayed nodes explicitely.
Propagate delayed resume at least for preload case in a simple way.
Currently PVMOVE depends on internal logic where 'mirror' with
corelog is 'possible' PVMOVE. In such case resume of 'created'
node is 'delayed'.
This is mostly an ugly internal hack - but for the moment being when we
add propagation for preload - it does work reasonable.
TODO: provide standard API and avoid this internal 'guessing'.
Check first if we need to even link -lrt - since clock functions
are normally emebeded with recent glibc (>=2.17)
Use standard RT_LIBS name.
Avoid duplicate test for realtime clock with lvmlockd
Show better error message when realtime clock support is missing or
disabled.
Link RT_LIBS explicitely with lvmlockd and lvmetad.
Avoid adding -g more then once for debug builds.
Avoid enabling DEBUG_MEM when we build multithreaded tools.
Link executables with -fPIE -pie and --export-dynamic LDFLAGS
Introduce PROGS_FLAGS to add option to pass flags for external libs.
Link lvm2 internally library only when really used.
Link DAEMON_LIBS with daemons.
Pass VALGRIND_CFLAGS internally
Set shell failure mode on couple places.
API for strtod() or strtoul() needs reset of errno, before it's being
called. So add missing resets in missing places and some also some
errno validation for out-of-range numbers.
Dmeventd reuses 'dm_task' struct for some STATUS operation, but due to
missing reinitization of dm_task target list, it has caused misprocesing
of recieved events as the parsed target has been simply added to the
list of existing status and cause multiple actions being called for
single event.
Add function to adjust printing of percent values in better way.
Rounding here is going along following rules:
0% & 100% are always clearly reported with .0 decimal points.
Values slightly above 0% we make sure a nearest bigger
non zero value with given precission is printed
(i.e. 0.01 for %.2f will be shown)
For values closely approaching 100% we again detect and adjust value
that is less then 100 when printed.
(i.e. 99.99 for %.2f will be shown).
For other values we are leaving them with standard rounding mechanism
since we care mainly about corner values 0 & 100 which need to be
printed precisely.
Current existing kernels reports status sometimes in weird form.
Instead of showing what is the exact progress, we need to estimate
this in-sync state from several surrounding states.
Main reason here is to never report 100% sync state for a raid device
which will be undergoing i.e. recovery.
It's not an error to attempt to update regions from an fd that has
been truncated (or otherwise no longer has any allocated extents):
in this case, the call should remove all regions corresponding to
the group, and return an empty region table.
Use log_sys_error rather than log_error if execvp() fails:
/mnt/redhat/xdoio.13752.XIORQ: Created new group with 2 region(s) as group ID 0.
# execvp() failed.
vs:
/var/lib/libvirt/images/rhel7-vm1.qcow2: Created new group with 884 region(s) as group ID 0.
dmfilemapd: execvp failed: No such file or directory
Removing some unused new lines and changing some incorrect "can't
release until this is fixed" comments. Rename license.txt to make
it clear its merely an included file, not itself a licence.
This commit supersedes reverted 1e4462dbfb
to avoid changes to liblvm and the libdm API completely.
The libdevmapper interface compares existing table line retrieved from
the kernel to new table line created to decide if it can suppress a reload.
Any difference between input and output of the table line is taken to be a
change thus causing a table reload.
The dm-raid target started to misorder the raid parameters (e.g. 'raid10_copies')
starting with dm-raid target version 1.9.0 up to (excluding) 1.11.0. This causes
runtime failures (limited to raid10 as of tests) and needs to be reversed to allow
e.g. old lvm2 uspace to run properly.
Check for the aforementioned version range in libdm and adjust creation of the table line
to the respective (mis)ordered sequence inside and correct order outside the range
(as described for the raid target in the kernels Documentation/device-mapper/dm-raid.txt).
This reverts commit 1e4462dbfb
in favour of an enhanced solution avoiding changes in liblvm
completetly by checking the target versions in libdm and emitting
the respective parameter lines.
The libdevmapper interface compares existing table line retrieved from
the kernel to new table line created to decide if it can suppress a reload.
Any difference between input and output of the table line is taken to be a
change thus causing a table reload.
The dm-raid target started to misorder the raid parameters (e.g. 'raid10_copies')
starting with dm-raid target version 1.9.0 up to (excluding) 1.11.0. This causes
runtime failures (limited to raid10 as of tests) and needs to be reversed to allow
e.g. old lvm2 uspace to run properly.
Check for the aforementioned version range and adjust creation of the table line
to the respective (mis)ordered sequence inside and correct order outside the range
(as described for the raid target in the kernels Documentation/device-mapper/dm-raid.txt).
Require that the path argument to dmfilemapd be an absolute path
and document this in tool output, libdevmapper.h and dmfilemapd.8.
The check is also enforced by dm_stats_start_filemapd() to avoid
forking a new process with an invalid path argument.
Older library version was not detecting unknown 'feature' bits
and could let start target without needed option.
New versioned symbol now checks for supported feature bits.
_Base version keeps accepting only previously known features and
mask/ignores unknown bits.
NB: if the older binary passed in 'random' bits, it will not get
metadata2 by chance. New linked binary get new validation function.
Library user is required to not pass 'trash' for unsupported bits,
as such calls will be rejected.
Dm cache target version 1.10 introduces new cache metadata format
(upstream kernel >=4.11).
New format is enable by passing new target feature flag metadata2.
Interace side on libdm uses DM_CACHE_FEATURE_METADATA2.
This feature bit is now also recognized on status
and set in 'feature_flags' field of dm_status_cache structure.
Code also adds check for 'highest' supported feature flag bit.
So it rejects properly any 'unknown' feature bit set by application.
Better code to enforce writethrough caching for cleaner policy.
Only check for cleaner when DM_CACHE_FEATURE_PASSTHROUGH or
DM_CACHE_FEATURE_WRITEBACK is set.
Test for NULL in dm_stats_destroy() and return immediately if
the struct dm_stats pointer is NULL (similar to free(NULL)).
This simplifies cleanup code which otherwise needs to:
out:
if (dms)
dm_stats_destroy(dms);
return;
Older compilers are not able to determine that although group_id
is only assigned in one branch of a conditional, it is never used
used when the other branch is taken:
libdm-stats.c:3319: warning: "group_id" may be used uninitialized in this function
Avoid this by always initialising the variable when it is
declared.
Add a daemon that can be launched to monitor a group of regions
corresponding to the extents of a file, and to update the regions as the
file's allocation changes.
The daemon is intended to be started from a library interface, but can
also be run from the command line:
dmfilemapd <fd> <group_id> <path> <mode> [<foreground>[<log_level>]]
Where fd is a file descriptor open on the mapped file, group_id is the
group identifier of the mapped group and mode is either "inode" or
"path". E.g.:
# dmfilemapd 3 0 vm.img inode 1 3 3<vm.img
...
If foreground is non-zero, the daemon will not fork to run in the
background. If verbose is non-zero, libdm and daemon log messages will
be printed.
It is possible for the group identifier to change when regions are
re-mapped: this occurs when the group leader is deleted (regroup=1 in
dm_stats_update_regions_from_fd()), and another region is created before
the daemon has a chance to recreate the leader region.
The operation is inherently racey since there is currently no way to
atomically move or resize a dm_stats region while retaining its
region_id.
Detect this condition and update the group_id value stored in the
filemap monitor.
A function is also provided in the the stats API to launch the filemap
monitoring daemon:
int dm_stats_start_filemapd(int fd, uint64_t group_id, const char *path,
dm_filemapd_mode_t mode, unsigned foreground,
unsigned verbose);
This carries out the first fork and execs dmfilemapd with the arguments
specified.
A dm_filemapd_mode_t value is specified by the mode argument: either
DM_FILEMAPD_FOLLOW_INODE, or DM_FILEMAPD_FOLLOW_PATH. A helper function,
dm_filemapd_mode_from_string(), is provided to parse a string containing
a valid mode name into the appropriate dm_filemapd_mode_t value.
It's not an error to call dm_stats_group_present() on a handle
that contains no regions.
This causes dmfilemap to log a false backtrace during shutdown
if all regions are removed from the corresponding device:
exiting _filemap_monitor_get_events() with deleted=0, check=0
waiting for FILEMAPD_WAIT
dm message (253:1) [ opencount flush ] @stats_list dmstats [32768] (*1)
<backtrace>
Filemap group removed: exiting.
Change this to only emit a backtrace if the handle is NULL.
Commit 80a6de616a versioned the dm_tree_node_add_raid_target_with_params()
and dm_tree_node_add_raid_target() APIs for compatibility reasons.
There's no user of the latter function, remove it.
Related: rhbz834579
Related: rhbz1191935
Related: rhbz1191978
Commit 27384c52cf lowered the maximum number of devices
back to 64 for compatibility.
Because more members have been added to the API in
'struct dm_tree_node_raid_params *', we have to version
the public libdm RAID API to not break any existing users.
Changes:
- keep the previous 'struct dm_tree_node_raid_params' and
dm_tree_node_add_raid_target_with_params()/dm_tree_node_add_raid_target()
in order to expose the already released public RAID API
- introduce 'struct dm_tree_node_raid_params_v2' and additional functions
dm_tree_node_add_raid_target_with_params_v2()/dm_tree_node_add_raid_target_v2()
to be used by the new lvm2 lib reshape extentions
With this new API, the bitfields for rebuild/writemostly legs in
'struct dm_tree_node_raid_params_v2' can be raised to 256 bits
again (253 legs maximum supported in MD kernel).
Mind that we can limit the maximum usable number via the
DEFAULT_RAID{1}_MAX_IMAGES definition in defaults.h.
Related: rhbz834579
Related: rhbz1191935
Related: rhbz1191978
Commit 64a2fad5d6 raised the maximum number of RAID devices to 64.
Commit e2354ea344 introduced RAID_BITMAP_SIZE as 4 to have
256 bits (4 * 64 bit array members), thus changing the libdm API
unnecessarilly for the time being.
To not change the API, reduce RAID_BITMAP_SIZE to 1.
Remove an unneeded definition of it from libdm-common.h.
If we ever decide to raise past 64, we'll version the API.
Related: rhbz834579
Related: rhbz1191935
Related: rhbz1191978
In order to support striped raid5/6/10 LV reshaping (change
of LV type, stripesize or number of legs), this patch
introduces infrastructure prerequisites to be used
by raid_manip.c extensions in followup patches.
This base is needed for allocation of out-of-place
reshape space required by the MD raid personalities to
avoid writing over data in-place when reading off the
current RAID layout or number of legs and writing out
the new layout or to a different number of legs
(i.e. restripe)
Changes:
- add members reshape_len to 'struct lv_segment' to store
out-of-place reshape length per component rimage
- add member data_copies to struct lv_segment
to support more than 2 raid10 data copies
- make alloc_lv_segment() aware of both reshape_len and data_copies
- adjust all alloc_lv_segment() callers to the new API
- add functions to retrieve the current data offset (needed for
out-of-place reshaping space allocation) and the devices count
from the kernel
- make libdm deptree code aware of reshape_len
- add LV flags for disk add/remove reshaping
- support import/export of the new 'struct lv_segment' members
- enhance lv_extend/_lv_reduce to cope with reshape_len
- add seg_is_*/segtype_is_* macros related to reshaping
- add target version check for reshaping
- grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal
- enhance libdm deptree code to support data_offset (out-of-place reshaping)
and delta_disk (legs add/remove reshaping) target arguments
Related: rhbz834579
Related: rhbz1191935
Related: rhbz1191978
When we preload device with smaller size, we avoid its resume,
so later suspend/resume of full device tree my process all
existing in flight bios.
Also update comment and avoid using confusing opposite meaning.
Add:
- support for segment types raid6_{ls,rs,la,ra}_6
(striped raid with dedicated last Q-Syndrome SubLVs)
- conversion support from raid5_{ls,rs,la,ra} to/from raid6_{ls,rs,la,ra}_6
- setting convenient segtypes on conversions from/to raid4/5/6
- related tests to lvconvert-raid-takeover.sh factoring
out _lvcreate,_lvconvert funxtions
Related: rhbz1366296
Add:
- support for segment type raid6_n_6 (striped raid with dedicated last parity/Q-Syndrome SubLVs)
- conversion support from striped/raid0/raid0_meta/raid4 to/from raid6_n_6
- related tests to lvconvert-raid-takeover.sh
Related: rhbz1366296
Add:
- support for segment type raid5_n (striped raid with dedicated last parity SubLVs)
- conversion support from striped/raid0/raid0_meta/raid4 to/from raid5_n
- related tests to lvconvert-raid-takeover.sh
Related: rhbz1366296
Add a call to update the regions corresponding to a file mapped
group of regions. The regions to be updated must be grouped, to
allow us to correctly identify extents that have been deallocated
since the map was created.
Tables are built of the file extents, and the extents currently
mapped to dmstats regions: if a region no longer has a matching
file extent, it is deleted, and new regions are created for any
file extents without a matching region.
The FIEMAP call returns extents that are currently in-memory (or
journaled) and awaiting allocation in the file system. These have
the FIEMAP_EXTENT_UNKNOWN | FIEMAP_EXTENT_DELALLOC flag bits set
in the fe_flags field - these extents are skipped until they
have a known disk location.
Since it is possile for the 0th extent of the file to have been
deallocated this must also handle the possible deletion and
re-creation of the group leader: if no other region allocation
is taking place the group identifier will not change.
If the group_id passed to _stats_group_id_present is equal to the
special value DM_STATS_GROUP_NOT_PRESENT there is no need to perform
any further tests: return false immediately.
When showing sizes with 'H|human' units we do use standard rounding.
This however is confusing users from time to time,
when the printed number uses some biger units i.e. GiB and there is just
tiny fraction of space missing.
So here is some real-life example with new 'r' unit.
$lvs
LV VG Attr LSize Pool Origin
lvol0 vg -wi-a----- 1.99g
lvol1 vg -wi-a----- <2.00g
lvol2 vg -wi-a----- <2.01g
Meaning is - lvol1 has 'slightly' less then 2.00g - from sign '<' user
can be aware the LV doesn't have full 2.00GiB in size so he
will be less surpriced allocation of 2G volume will not succeed.
$ vgs
VG #PV #LV #SN Attr VSize VFree
vg 2 2 0 wz--n- <6,00g <2,01g
For uses needing 'old' undecorated human unit simply will continue
to use 'H|h' units.
The new R|r may further change when we would recongnize some
other way how to improve readability.
Call _stats_regions_destroy() from dm_stats_list() if dms->regions
is non-NULL. This avoids leaking any pool allocations and ensures
the handle is in a known state: if an error occurs during the list,
dms->regions will be NULL and the handle will appear empty.
The system is likely in some very inconsisten state.
Do not try to make it even more problematic with trying
to invoke tools like thin_check via callback.
The dm_bit_copy() macro uses the source (bs1) bitset size as the
limit for memcpy:
memcpy((bs1) + 1, (bs2) + 1, ((*(bs1) / DM_BITS_PER_INT) + 1)..)
This is safe if the destination bitset is smaller than the source,
or if the two bitsets are of the same size.
With a destination that is larger (e.g. when resizing a bitmap to
add more capacity), the memcpy will overrun the source bitset and
set garbage bits in the destination.
There are nine uses of the macro currently (8 in libdm/regex, and
1 in daemons/cmirrord): in each case the two bitsets are always of
equal size so the behaviour is unchanged.
Fix the macro to use bs2's size to simplify resizing bitsets and
avoid the need for another copy macro.
If FIEMAP returns a single extent after the first call, no extent
boundary is detected and the first extent is not counted by the
normal mechanism.
In this case, increment nr_extents at the same time the extent is
added to the region table, before returning.
It's useful to be able to specify a minimum number of bits for a
new bitmap parsed from a list, for e.g. to allow for expansing a
group without needing to copy/reallocate the bitmap.
Add a backwards compatible symbol for programs linked against old
versions of the library.
It is sometimes convenient to iterate over the set bits in a dm
bitset in reverse order (from the highest set bit toward zero), or
to quickly find the last set bit.
Add dm_bit_get_last() and dm_bit_get_prev(), mirroring the existing
dm_bit_get_first() and dm_bit_get_next().
dm_bit_get_prev() uses __builtin_clz when available to efficiently
test the bitset in reverse.
Split out the loop that iterates over each batch of FIEMAP
extent data from the function that sets up and calls the ioctl
to reduce nesting and simplify local variable use:
_stats_get_extents_for_file()
-> _stats_map_extents()
The _stats_map_extents() function is responsible for detecting
eof and extent boundaries and adding whole, allocated extents
to the file extent table for region creation.
Check that all region_id values specified in a group bitmap are
actually present: although this should not normally happen when
using the dmstats tool, it is possible as a result of manual
changes (or bugs) for a group descriptor to contain one or more
group_id values that do not exist.
Check for this situation when reading group descriptors, warn
the user the user, and clear these bits in the bitmap when
formatting it for output.
If a region has a a DMS_GROUP tag in aux_data where the first
region_id in the bitmap is not the same as the containing region,
dmstats will segfault:
# '2' is never a valid group bitset list for region_id == 0
# dmsetup message vg_hex/root 0 "@stats_set_aux 0 DMS_GROUP=img:2#"
# dmsetup message vg_hex/root 0 "@stats_list"
0: 45383680+16384 16384 dmstats DMS_GROUP=img:2#
1: 46071808+32768 32768 dmstats -
2: 47382528+16384 16384 dmstats -
# dmstats list
Segmentation fault (core dumped)
The crash will occur in some arbitrary dm_stats_get_* property
method - this happens while processing the 1st region_id in the
bitset, because the region is marked as grouped, but there is
no group bitmap present at dms->groups[2]->regions.
Fix this by detecting a mismatch between the expected region_id
and dm_bit_get_first() for the parsed bitset during
_parse_aux_data_group().
Handle files that contain multiple logical extents in a single
physical extent properly:
- In FIEMAP terms a logical extent is a contiguous range of
sectors in the file's address space.
- One or more physically adjacent logical extents comprise a
physical extent: these are the disk areas that will be mapped
to regions.
- An extent boundary occurs when the start sector of extent
n+1 is not equal to (n.start + n.length).
This requires that we accumulate the length values of extents
returned by FIEMAP until a discontinuity is found (since each
struct fiemap_extent returned by FIEMAP only represents a single
logical extent, which may be contiguous with other logical
extents on-disk).
This avoids creating large numbers of regions for physically
adjacent (logical) extents and fixes the earlier behaviour which
would only map the first logical extent of the physical extent,
leaving gaps in the region table for these files.
When mapping regions to a file descriptor, a temporary table of
extent descriptors is built using the dm_pool object building
interface.
Previously this use borrowed the dms->mem region and counter
table pool (since nothing can interleave with the allocation
while the caller is still in dm_stats_create_regions_from_fd()).
This turns out to be problematic for error recovery. When a
region creation operation fails partway through file mapping,
we need to roll back the set of already created regions and
this requires a listed handle: the dm_stats_list() will then
allocate from the same pool as the extents; we either have
to throw away valid list data, or leak the extent table, to
return the handle in a valid state.
Avoid this problem by creating a new, temporary mem pool in
_stats_create_file_regions() to hold the extent data, and
discarding it on exit from the function.
While cleaning up the table of already created regions during a
failed dm_stats_create_regions_from_fd(), list the handle once,
and call _stats_delete_region() directly. This avoids sending a
@stats_list message for each region deleted, reducing runtime
from 6s to 0.7s when cleaning up ~250 out of ~10000 regions:
# time dmstats create --filemap b.img
device-mapper: message ioctl on (253:0) failed: Cannot allocate memory
Failed to create region 246 of 309 at 9388032.
Could not create regions from file /root/b.img
<< pauses here >>
Command failed
real 0m6.267s
user 0m3.770s
sys 0m2.487s
# time dmstats create --filemap b.img
device-mapper: message ioctl on (253:0) failed: Cannot allocate memory
Failed to create region 246 of 309 at 9388032.
Could not create regions from file /root/b.img
Command failed
real 0m0.716s
user 0m0.034s
sys 0m0.581s
Testing the error path requires region creation to start to
fail part way through the operation (in order to have regions
to clean up): the simplest way is to ensure the system is
close to the kernel limit of 1/4 RAM or 1/2 vmalloc space
consumed by dmstats data.
Split dm_stats_delete_region() so that internal callers can manage
the handle state themselves.
dm_stats_delete_region() now just handles checking the state of the
handle, reporting validation errors, and calling dm_stats_list() if
necessary, before calling _stats_delete_region().
The new _stats_delete_region() function performs the actual group
member removal and region deletion, and requires a fully listed
handle to operate.
Callers that repeatedly delete regions can use a single listed
handle for many operations on the same device, avoiding one
message ioctl per region deleted: since @stats_list with many
regions is expensive, this yields large runtime improvements.
If we fail to create a region during dm_stats_create_regions_from_fd(),
we must remove all regions that were created to do this to date. This
needs to loop over the table of region_id values that were populated
by _stats_create_file_regions() before the error.
The code for this failure case in the out_remove branch incorrectly
uses the table index as the region_id:
for (--i; i != DM_STATS_REGION_NOT_PRESENT; i--) {
if (!dm_stats_delete_region(dms, i))
log_error("Could not delete region " FMTu64 ".", i);
}
This causes the cleanup code to delete a completely unrelated set
of regions (since the index here will always be nr_regions..0).
Fix it to pass the actual region_id stored in regions[i] instead.
Fix a silly bug in dm_stats_delete_region() that hugely inflates
runtimes when deleting a large number of regions.
For ~50,000 regions this change reduces the runtime from 98s to
6s on my test systems (a ~93% reduction).
The bug exists because dm_stats_delete_region() applies a truth
test to the return value of dm_stats_get_nr_areas(); this is
never correct usage - it will walk the entire region table and
calculate area counts for each region (which is roughly O(n^2)
in the number of regions, as dm_stats_delete_region() is being
called inside a region walk).
Although the individual area calculation is not that costly,
uselessly running anything 2,500,000,000 times over gets a bit
slow.
A much cheaper test (which is always true if the areas check is
true) is to just test dm_stats_get_nr_regions() or dms->regions;
if either is true it implies at least one area exists.
Old:
Performance counter stats for 'dmstats delete --allregions --alldevices':
98117.791458 task-clock (msec) # 1.000 CPUs utilized
127 context-switches # 0.001 K/sec
3 cpu-migrations # 0.000 K/sec
6,631 page-faults # 0.068 K/sec
307,711,724,562 cycles # 3.136 GHz
544,762,959,577 instructions # 1.77 insn per cycle
84,287,824,115 branches # 859.047 M/sec
2,538,875 branch-misses # 0.00% of all branches
98.119578733 seconds time elapsed
New:
Performance counter stats for 'dmstats delete --allregions --alldevices':
6427.251074 task-clock (msec) # 1.000 CPUs utilized
6 context-switches # 0.001 K/sec
0 cpu-migrations # 0.000 K/sec
6,634 page-faults # 0.001 M/sec
21,613,018,724 cycles # 3.363 GHz
3,794,755,445 instructions # 0.18 insn per cycle
852,974,026 branches # 132.712 M/sec
808,625 branch-misses # 0.09% of all branches
6.428953647 seconds time elapsed
There are two possible errors in _dm_stats_populate_region():
* No region struct in dms->regions[region_id]
* Failure to parse data from @stats_print
These have very different causes: the first occurs where a client
program is populating one region at a time (region_id is a single
region identifier), and has not previously called dm_stats_list()
to dimension the region tables; this is an API usage error.
The second occurs when either we read unparseable data from the
kernel (kernel bug), or where various resource allocations fail.
Separate these two cases out and log separate messages for each
(allocation failures in the path already have their own distinct
message), since the "failed to parse.." message in the un-listed
handle case is confusing and misleading.
Translate log_info() into log_very_verbose() which is macro
supposed to be used by our code.
log_info() is internal macro with eventually some 'symbolic' meaning
in syslogging daemons.
Instead of compiling 2 log call for 2 different logging functions,
and runtime decide which version to use - use only 'newer' function
and when user sets his own OLD dm_log logging translate it runtime
for old arg list set.
The positive part is - we get shorter generated library,
on the negative part this translation means, we always have evaluate
all args and print the message into local on stack buffer, before
we can pass this buffer to the users' logging function with proper
expected parameters (and such function may later decide to discard
logging based on message level so whole printing was unnecessary).
Introduce macros:
log_level(), log_stderr(), log_once(), log_bypass_report()
For easier and more consisten way how to 'decoder' bits
of info from passed 'level'.
This patch fixes potential problem when 'level' of message
might not have always masked right bits.
The dm_stats_delete_region() call removes a region from the bound
device, and, if the region is grouped, from the group leader
group descriptor stored in aux_data.
To do this requires a listed handle: previous versions of the
library do not since no dependencies exist between regions without
grouping.
This leads to strange behaviour when a command built against an old
version of the library is used with one supporting groups. Deleting
a region with dmstats succeeds, but logs errors:
# dmstats list
Name RgID RgSta RgSiz #Areas ArSize ProgID
vg_hex-root 0 0 1.00g 1 1.00g dmstats
vg_hex-root 1 1.00g 1.00g 1 1.00g dmstats
vg_hex-root 2 2.00g 1.00g 1 1.00g dmstats
# dmstats delete --regionid 2 vg_hex/root
Region ID 2 does not exist
Could not delete statistics region.
Command failed
# dmstats list
Name RgID RgSta RgSiz #Areas ArSize ProgID
vg_hex-root 0 0 1.00g 1 1.00g dmstats
vg_hex-root 1 1.00g 1.00g 1 1.00g dmstats
This happens because the call to dm_stats_delete_region() is inside
a dm_stats_walk_*() iterator: upon entry to the call, the iterator
is at its end conditions and about to terminate. Due to the call to
dm_stats_list() inside the function, it returns with an iterator at
the beginning of a walk and performs a further iteration before
exiting. This final loop makes a further attempt to delete the
(already deleted) region, leading to the confusing error messages.
The current dmsetup.c handles DR_STATS and DR_STATS_META reports
separately in _display_info_cols(), meaning that the stats walk
functions are never called for these report types.
Versions before v2.02.159 have a loop using dm_stats_walk_do() and
dm_stats_walk_while(), that executes once for non-stats reports,
and once per region, or area, for DR_STATS/DR_STATS_META reports.
This older behaviour relies on the documented behaviour that the
walk functions will accept a NULL pointer as the struct dm_stats*
argument.
This was broken by commit f1f2df7b: the NULL test on dms and
dms->regions were incorrectly moved from the dm_stats_walk_end()
wrapper to the internal '_stats_walk_end()' helper.
Since the pointer is dereferenced in between these points, using
an older dmsetup with current libdm results in a segfault when
running a non-stats report:
# dmsetup info -c vg00/lvol0
Segmentation fault (core dumped)
Restore the NULL checks to the wrapper function as intended.
The dm_report_group_output_and_pop_all calls dm_report_output and
dm_report_group_pop for all the items that are currently in report
group. This is just a shortcut that makes it easier to output and
pop group's content so the group handle can be reused again without
a need to initialize and configure it again.
The functionality of dm_report_group_output_and_pop_all is the
same as dm_report_destroy but without destroying the report group
handle.
This patch moves printing of starting '{' character for JSON output up
untili it's known there's any further output following - either the
content or ending '}' character.
Also, remove unnecessary switch for different report group types and
calling individual functions to handle dm_report_group_create as that
code is shared for all existing types at the moment.
Calling dm_report_destroy_rows makes it possible to destroy any report
content we have but at the same time it doesn't destroy the report
handle itself, thus it's possible to reuse that handle again for new
report content.
Functionally, this is the same as calling dm_report_output with the
report handle but omitting the output iself. This functionality may
be useful if we, for whatever reason, need to discard the report
content and start a fresh new one but with the same report configuration
and initialization and thus we can just reuse the existing handle.
Support aggregate group and region histograms by allocating a new
histogram from the pool and populating it with a sum of the histogram
data for the areas contained in the region or group.
To avoid repeatedly summing the same histogram data, cache the pointer
in the group and regions structs for subsequent access. The aggregate
histograms are allocated from the same pool as the area histograms in
the corresponding handle and will be discarded at each list or populate
operation.
Add a call to create dmstats regions that correspond to the extents
present in a file descriptor open on a file in a local file system.
The file must reside on a file system type that correctly supports
physical extent location data in the FIEMAP ioctl.
Regions are optionally placed into a group with a user-defined alias.
File systems that do not support physical offsets in FIEMAP (btrfs
currently) are detected via fstatfs() - although attempting to map
a --filemap group on btrfs will fail anyway with the generic error
"Not on a device-mapper device" this is confusing; the file system
mount is on a device-mapper device, but btrfs' volume layer masks
this in the returned st_dev field since the returned logical file
extents may span multiple physical devices.
The function _stats_remove_region_id_from_group() incorecctly set
the group_id to DM_STATS_GROUP_NOT_PRESENT _before_ the call to
_stats_group_destroy(). This will cause the destroy function to
return immediately without doing anything:
339 static void _stats_group_destroy(struct dm_stats_group *group)
340 {
341 if (!_stats_group_present(group))
342 return;
Invalidating the ID in _stats_region_region_id_from_group() is
redundant anyway; it is rightly done as the last operation in
_stats_group_destroy (and it is not possible for anything to see
the old value between the two calls).
Remove the change to group_id to ensure that the alias and bitset
resources are correctly freed.
The call to dm_stats_walk_start() before the do statement makes
dm_stats_walk_do() behave inconsistently depending on context;
wrap them in an additional do { } while (0) so that the macro
always expands to a valid statement.
If after extracting stats arguments and group tags nothing remains
of aux_data but '-' set the region->aux_data field to the empty
string to match behaviour for non-grouped regions.
Although not harmful do not allow a group containing regions with
histograms since it is not currently possible to present histogram
data aggregated for the group.
Although a non-zero value for the number of ticks spent doing IO
should imply a non-zero number of IOs in the interval test for
this explicitly to avoid a divide-by-zero in the event of bad
counter data.
It's possible for interval_ns to be zero if the interval is not
set or the clock is misconfigured. Test for this before using the
value as the divisor in the utilisation calculation.
Walk flags are ULL constants; cast the result to a uint64_t before
logging with a FMTx64 format specifier to avoid a compiler warning:
warning: format ‘%lx’ expects argument of type ‘long unsigned int’,
but argument 5 has type ‘long long unsigned int’
Make it clear in libdevmapper.h, and in function argument names, that
libdm-stats uses the aux_data field internally and that any values set
for user_data are appended to the library values before being stored
with a region, and similarly, that internal data fields will be stripped
prior to returning any previously stored user_data.
Add support do dm_stats_walk*() to walk over the set of
available groups using the cursor embedded in the dm_stats
handle, and to obtain the type of the object at the current
stats cursor location. A set of flags is introduced to
control which objects are visited:
DM_STATS_WALK_AREA
DM_STATS_WALK_REGION
DM_STATS_WALK_GROUP
DM_STATS_WALK_ALL
A final flag suppresses visits to regions that contain only a
single area - since the aggregate of such a region is idential
to the area it contains this allows these duplicates to be
filtered out:
DM_STATS_WALK_SKIP_SINGLE_AREA
If flags are not initialised before beginning a walk the default
set matches the behaviour of previous versions of the library.
Also accept group identifiers as immediate arguments to the
counter, metric, and property functions by adding control
flags to the region and area identifiers passed in.
Region and area properties are mapped to their equivalents for
the group (for example: group size is reported as the sum of
all regions contained in the group). Counter and metric values
are aggregated for the region or group.
Introduce constants for the buffer sizes that libdm-stats uses:
one for messages sent to the kernel, one for rows of response data
returned, and a pair for the "start+len" range and histogram bounds
strings.
Add a grouping facility to the libdm-stats library that allows the
user to bind several regions together as a group. Groups may be
used to aggregate data from several regions for reporting, or to
select and sort among large sets of regions.
A textual descriptor ("group tag") is associated with each group
and is stored in the first group member's aux_data field. The
tag contains the group member list and an optional alias for the
group, allowing the user to assign meaningful names to groups of
regions.
These descriptors are parsed in @stats_list message responses and
populate the resulting region and area tables with the group
structure.
Groups with overlapping regions are permitted but since this will
result in some events being counted more than once a warning is
printed in this case.
Nested and overlapping groups are not currently supported and
attempting to create these configurations results in error.
