]> sourceware.org Git - lvm2.git/log
lvm2.git
12 years agoSupport improperly formated device numbers
Zdenek Kabelac [Tue, 20 Mar 2012 10:47:02 +0000 (10:47 +0000)]
Support improperly formated device numbers

There are kernel drivers (smblk) which set '-1' as their device major number.
This number is listed in /proc/devices then - but the kernel itself is using
just 12 bits - thus device is accessible via 4095 - there is posted patch
for 3.4 to fix this behavior (0 for auto allocation was mean to be used).

However to still allow using such devices with older kernels add some code
to use same behavior - so cut 12 bits from the major number from /proc/devices.

For now use log_warn() - maybe the severity of the message could be lowered
to just verbose level.

12 years agoDo exit if LISTEN_PID environment variable not correct during systemd handover.
Peter Rajnoha [Fri, 16 Mar 2012 21:30:53 +0000 (21:30 +0000)]
Do exit if LISTEN_PID environment variable not correct during systemd handover.

12 years agoAdd comment about slow startup of dmeventd
Zdenek Kabelac [Fri, 16 Mar 2012 19:13:09 +0000 (19:13 +0000)]
Add comment about slow startup of dmeventd

Separat lvm dumpconfig on extra line, so the shell trace output
is not mixed with dump.

12 years agoSkip dmeventd restart test
Zdenek Kabelac [Fri, 16 Mar 2012 19:11:29 +0000 (19:11 +0000)]
Skip dmeventd restart test

If dmevendt is not in /sbin, do not try to run restarting test.
FIXME: add some workaround for testing.

12 years agoRegression fix kernel version check
Zdenek Kabelac [Fri, 16 Mar 2012 19:09:49 +0000 (19:09 +0000)]
Regression fix kernel version check

got changed improperly with large update patch.

12 years agoLonger sleep after clvmd start
Zdenek Kabelac [Fri, 16 Mar 2012 19:08:09 +0000 (19:08 +0000)]
Longer sleep after clvmd start

12 years agoUse pvscan --cache instead of vgscan in lvmetad init script and systemd unit.
Peter Rajnoha [Fri, 16 Mar 2012 15:12:04 +0000 (15:12 +0000)]
Use pvscan --cache instead of vgscan in lvmetad init script and systemd unit.

12 years agoSwitch condition for dmeventd check
Zdenek Kabelac [Fri, 16 Mar 2012 13:44:51 +0000 (13:44 +0000)]
Switch condition for dmeventd check

12 years agoUpdate tests
Zdenek Kabelac [Fri, 16 Mar 2012 13:00:05 +0000 (13:00 +0000)]
Update tests

Indent

Shell improvements - use internal function for checks

Use PVs in ""   (LV and VG cannot have spaces)

Several test very starting 'dmeventd' without annoucing
it via prepade_dmeventd.

Fix some of test actually.

12 years agoUpdate test.sh check.sh utils.sh get.sh
Zdenek Kabelac [Fri, 16 Mar 2012 12:59:43 +0000 (12:59 +0000)]
Update test.sh check.sh utils.sh get.sh

Indent
Better shell usage
Function simplification
More usage of 'get' functions
Don't use valgrind tracing for check and get function (faster)
Update shell debugging (PS4, better stacktrace)
Support paths with spaces
Export SCRIPTNAME for external usage
Watch for dmeventd unexpectedly started during test

12 years agoUpdate aux.sh lvmwrapper
Zdenek Kabelac [Fri, 16 Mar 2012 12:59:02 +0000 (12:59 +0000)]
Update aux.sh lvmwrapper

Indent

Add valgrind support:
 env LVM_TEST_VALGRIND={0123} (the higher level, more commands tested)
 env LVM_TEST_CLVMD=1   runs clvmd within valgrind.
 env VALGRIND  script name executed for each lvm command (def. is valg).

Smarted teardown - should minimize occurence of left dev entries
 (using dmsetup remove -f only as last resort)
 sort removed devices by open count before actual removal

Use "" around string that may contain spaces.

Set log/verbose and activation/retry_deactivation to defined value.

Remove  debug.log after successful lvm command (easier to check output).

12 years agoTest suite skip md_detection
Zdenek Kabelac [Fri, 16 Mar 2012 12:57:28 +0000 (12:57 +0000)]
Test suite skip md_detection

Not normally needed - lets' enable it only for test that
really needs it.

12 years agoAllow to use also special prefixed names for test
Zdenek Kabelac [Fri, 16 Mar 2012 12:56:29 +0000 (12:56 +0000)]
Allow to use also special prefixed names for test

Currently we could not test special prefixes in our test suite.
As teardown will not find such device and basicaly busyloops here,
as at cannot remove such names.

This patch adds possibity to use:

vgcreate  V_$vg1 $dev

Note: you still need to use $PREFIX somewhere in the name.
(And of course, it's really bad idea to use $PREFIX (=LVMTEST)
 for normally used LVs)

The only purpose of this patch is to allow testing cluster with
special vg names that begins with V_ , P_....

12 years agoFix string parsing
Zdenek Kabelac [Fri, 16 Mar 2012 12:53:05 +0000 (12:53 +0000)]
Fix string parsing

Fix propagation of -e option - pass it via internal shell variable.
Fix parsing of /proc/mounts files (don't check for substrings).
 as reported by O.Mangold with suggested patch:
 https://www.redhat.com/archives/linux-lvm/2012-February/msg00030.html
Properly pass arguments with spaces ("$@")
Add validation for YES and EXTOFF variable content.

12 years agoRelax the active-minor test a bit.
Petr Rockai [Fri, 16 Mar 2012 11:36:51 +0000 (11:36 +0000)]
Relax the active-minor test a bit.

12 years agoIt's new.
Petr Rockai [Fri, 16 Mar 2012 10:46:25 +0000 (10:46 +0000)]
It's new.

12 years agoFix a regression in handling --major/--minor arguments to lvcreate & lvchange,
Petr Rockai [Fri, 16 Mar 2012 10:43:52 +0000 (10:43 +0000)]
Fix a regression in handling --major/--minor arguments to lvcreate & lvchange,
by allowing arg_int_value to be used with groupable options.

12 years agoFix name conflicts that prevent down-converting RAID1 when specifying a device
Jonathan Earl Brassow [Thu, 15 Mar 2012 20:00:54 +0000 (20:00 +0000)]
Fix name conflicts that prevent down-converting RAID1 when specifying a device

When down-converting a RAID1 device, it is the last device that is extracted
and removed when the user does not specify a particular device.  However,
when a device is specified (and it is not the last), the device is removed and
the remaining sub-LVs are "shifted down" to fill the hole.  This cause problems
when resuming the LV because if the shifted devices were resumed (and thus
renamed) before the sub-LV being extracted, there would be a name conflict.
The solution is to resume the extracted sub-LVs first so that they can be
properly renamed preventing a possible conflict.

This addresses bug 801967.

12 years agoRemove dmeventd fifos on exit if they are not managed by systemd.
Peter Rajnoha [Thu, 15 Mar 2012 08:45:55 +0000 (08:45 +0000)]
Remove dmeventd fifos on exit if they are not managed by systemd.

12 years agoUpdate configure change
Zdenek Kabelac [Wed, 14 Mar 2012 19:25:04 +0000 (19:25 +0000)]
Update configure change

Always define THIN_CHECK_CMD define - since for now we compile thin code
in more places.

12 years agoJust move declaration
Zdenek Kabelac [Wed, 14 Mar 2012 17:15:22 +0000 (17:15 +0000)]
Just move declaration

12 years agoUse proper "" for absolute paths
Zdenek Kabelac [Wed, 14 Mar 2012 17:14:33 +0000 (17:14 +0000)]
Use proper "" for absolute paths

12 years agoCleanup for conf file
Zdenek Kabelac [Wed, 14 Mar 2012 17:13:45 +0000 (17:13 +0000)]
Cleanup for conf file

12 years agoImprove thin_check option passing
Zdenek Kabelac [Wed, 14 Mar 2012 17:12:05 +0000 (17:12 +0000)]
Improve thin_check option passing

Update a way we handle option passing - so we now support path and options
with space inside.
Fix dm name usage for thin pools with '-' in name.
Use new lvm.conf option thin_check_options to pass in options as string array.

12 years agoAdd --with-thin-check configure option
Zdenek Kabelac [Wed, 14 Mar 2012 17:09:00 +0000 (17:09 +0000)]
Add --with-thin-check configure option

If specified - use given path without test (Path could be empty)
If autodetection is in use - check for command in available PATH.

12 years agoUse SD_ACTIVATION env. var. in systemd units to better detect systemd in use.
Peter Rajnoha [Wed, 14 Mar 2012 15:51:51 +0000 (15:51 +0000)]
Use SD_ACTIVATION env. var. in systemd units to better detect systemd in use.

LISTEN_PID and LISTEN_FDS environment variables are defined only during systemd
"start" action. But we still need to know whether we're activated during
"reload" action as well - we use the reload action to call "dmeventd -R"/"lvmetad -R"
for statefull daemon restart. We can't use normal "restart" as that is simply
composed of "stop" and "start" and we would lose any state the daemon has.

12 years agoWHATS_NEW
Peter Rajnoha [Wed, 14 Mar 2012 12:12:21 +0000 (12:12 +0000)]
WHATS_NEW

12 years agoDo not run a new dmeventd instance on restart if there's no existing one.
Peter Rajnoha [Wed, 14 Mar 2012 11:16:00 +0000 (11:16 +0000)]
Do not run a new dmeventd instance on restart if there's no existing one.

12 years agoFix error message when pvmove LV activation fails with name already in use.
Alasdair Kergon [Tue, 13 Mar 2012 20:21:26 +0000 (20:21 +0000)]
Fix error message when pvmove LV activation fails with name already in use.

12 years agoAdd a toplevel check_lvmetad target.
Petr Rockai [Mon, 12 Mar 2012 15:05:55 +0000 (15:05 +0000)]
Add a toplevel check_lvmetad target.

12 years agoWe need longer window for checking read from fifo.
Zdenek Kabelac [Mon, 12 Mar 2012 14:46:53 +0000 (14:46 +0000)]
We need longer window for checking read from fifo.

Seems like some dmeventd startups are taking more time then expected,
so make the time window longer here.

12 years agoRemoving call of release_vg(NULL)
Zdenek Kabelac [Mon, 12 Mar 2012 14:43:12 +0000 (14:43 +0000)]
Removing call of release_vg(NULL)

Since we are in error path were vg must be always NULL,
skip call of release_vg() like we do in other places.

12 years agoUsing %u modifier to print unsigned values.
Zdenek Kabelac [Mon, 12 Mar 2012 14:41:42 +0000 (14:41 +0000)]
Using %u modifier to print unsigned values.

12 years agoBetter structure layout for device_info
Zdenek Kabelac [Mon, 12 Mar 2012 14:40:41 +0000 (14:40 +0000)]
Better structure layout for device_info

Save some relocation entries and use directly char[].
Since we do not need yes more then 127 partitions per device, use just int8_t.
Move lvm_type_filter_destroy into local static function.

12 years agoSimplify error path code for filter initialization
Zdenek Kabelac [Mon, 12 Mar 2012 14:35:57 +0000 (14:35 +0000)]
Simplify error path code for filter initialization

Use 'int' counter.
Use 'bad' with goto_bad macro.

12 years agoImprove harness code
Zdenek Kabelac [Mon, 12 Mar 2012 14:24:15 +0000 (14:24 +0000)]
Improve harness code

Support timestamping with harness -  using VERBOSE=2
Fix also logging in several situation
(i.e. continue logging multiple test in VERBOSE mode,
do not coredump with empty output).

12 years agoSwitch to normal log_verbose message
Zdenek Kabelac [Mon, 12 Mar 2012 14:18:28 +0000 (14:18 +0000)]
Switch to normal log_verbose message

Here it's not an error case - so do not push this message to stderr.

12 years agoFix error path for create_toolcontext
Zdenek Kabelac [Mon, 12 Mar 2012 14:15:04 +0000 (14:15 +0000)]
Fix error path for create_toolcontext

Never return unfinished toolcontext - since error path is hit on
various stages of initialization we cannot leave it partially uninitialized,
since we would need to spread many more test across the code for config_valid.
Instead return NULL and properly release udev library resources as well.

12 years agoDetect the lvm binary path in lvmetad udev rules.
Peter Rajnoha [Mon, 12 Mar 2012 08:59:55 +0000 (08:59 +0000)]
Detect the lvm binary path in lvmetad udev rules.

We can't use 'DM_SBIN_PATH'. This one is set only for DM devices but not
for all block devices - the pvscan is run on all relevant block devices!

This LVM_SBIN_PATH (as well as DM_SBIN_PATH) detection should be removed
eventually but for upstream solution, we still have to do that as there are
known cases where the binaries are put either in /sbin or /usr/sbin
(some installation systems, initrd systems etc.).

12 years agoRemove some whitespaces.
Milan Broz [Sat, 10 Mar 2012 10:39:28 +0000 (10:39 +0000)]
Remove some whitespaces.
(Test commit.)

12 years agoRemove some whitespaces.
Milan Broz [Sat, 10 Mar 2012 09:32:46 +0000 (09:32 +0000)]
Remove some whitespaces.
(test commit)

12 years agoDo not unlink the socket on daemon shutdown if activation is done via systemd.
Peter Rajnoha [Thu, 8 Mar 2012 14:54:05 +0000 (14:54 +0000)]
Do not unlink the socket on daemon shutdown if activation is done via systemd.

12 years agoAdd 'Restart: on-abort' for lvmetad systemd service.
Peter Rajnoha [Thu, 8 Mar 2012 09:49:09 +0000 (09:49 +0000)]
Add 'Restart: on-abort' for lvmetad systemd service.

Restart lvmetad automatically if it exits due to reception of uncaught signal,
leaving a track in systemd log/syslog.

12 years agoFix warn message and update man page
Zdenek Kabelac [Tue, 6 Mar 2012 09:22:02 +0000 (09:22 +0000)]
Fix warn message and update man page

Fix regression in man page. The chunk size is in kilobyte units on command line
input though in the source code we work with sector size unit
so make it clear in the man page.

Update chunksize for thin pool in man page - it's max value is 1024M == 1G.

Fix warning range message to show proper max value.

12 years agopost-release
Alasdair Kergon [Tue, 6 Mar 2012 04:47:37 +0000 (04:47 +0000)]
post-release

12 years agoupdate reworded string
Alasdair Kergon [Tue, 6 Mar 2012 03:16:05 +0000 (03:16 +0000)]
update reworded string

12 years agoclose ]
Alasdair Kergon [Tue, 6 Mar 2012 03:11:13 +0000 (03:11 +0000)]
close ]

12 years agoChange pvscan --cache syntax
Alasdair Kergon [Tue, 6 Mar 2012 02:54:31 +0000 (02:54 +0000)]
Change pvscan --cache syntax

12 years agopre-release
Alasdair Kergon [Tue, 6 Mar 2012 02:50:40 +0000 (02:50 +0000)]
pre-release

12 years agoAdjust wording
Alasdair Kergon [Tue, 6 Mar 2012 02:42:31 +0000 (02:42 +0000)]
Adjust wording

12 years agore-word warnings to be clearer to user
Alasdair Kergon [Tue, 6 Mar 2012 02:39:25 +0000 (02:39 +0000)]
re-word warnings to be clearer to user

12 years agoSwitch pvscan --cache major:minor to --major --minor.
Alasdair Kergon [Tue, 6 Mar 2012 02:30:49 +0000 (02:30 +0000)]
Switch pvscan --cache major:minor to --major --minor.

12 years agoSome more missing supposedly 64bit operations.
Zdenek Kabelac [Mon, 5 Mar 2012 15:05:24 +0000 (15:05 +0000)]
Some more missing supposedly 64bit operations.

Avoid use 32bit math for extent_size.

12 years agoUse already acquired variable...
Peter Rajnoha [Mon, 5 Mar 2012 14:45:00 +0000 (14:45 +0000)]
Use already acquired variable...

12 years agoFit thin pool metadata into 128MB
Zdenek Kabelac [Mon, 5 Mar 2012 14:19:13 +0000 (14:19 +0000)]
Fit thin pool metadata into 128MB

If the lvcreate may decide some automagical values for a user,
try to keep the pool metadata size into 128MB range for optimal
perfomance (as suggested by Joe).

So if the pool metadata size and chunk_size were not specified,
try to select such values they would fit into 128MB size.

12 years agoImprove warning
Zdenek Kabelac [Mon, 5 Mar 2012 14:15:50 +0000 (14:15 +0000)]
Improve warning

Use thin_dump --repair suggestion in log error message
and use just warning on  deactivation path without repair info
(since node has been deactivated).

Also check whether there is not 16 args for thin_check configured.

12 years agoUse 64bit math
Zdenek Kabelac [Mon, 5 Mar 2012 14:12:57 +0000 (14:12 +0000)]
Use 64bit math

Prevent 32bit overflow and resulting weird error reports when working
with TB sizes..

12 years agoAdd test for name mangling functionality.
Peter Rajnoha [Mon, 5 Mar 2012 12:49:06 +0000 (12:49 +0000)]
Add test for name mangling functionality.

12 years agoCheck for multiple mangled names in auto mangling mode.
Peter Rajnoha [Mon, 5 Mar 2012 12:48:12 +0000 (12:48 +0000)]
Check for multiple mangled names in auto mangling mode.

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.

12 years agoFix dm_task_get_name_unmangled to not unmangle already unmangled name.
Peter Rajnoha [Mon, 5 Mar 2012 12:45:43 +0000 (12:45 +0000)]
Fix dm_task_get_name_unmangled to not unmangle already unmangled name.

In 'auto' and 'hex' mode, these names are already unmangled on ioctl return.
There's no point on trying to do that once again (actually it's a bug!).

12 years agoCheck whether device names are properly mangled on ioctl return.
Peter Rajnoha [Mon, 5 Mar 2012 12:43:03 +0000 (12:43 +0000)]
Check whether device names are properly mangled on ioctl return.

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).

12 years agoClean up internal mangling interface.
Peter Rajnoha [Mon, 5 Mar 2012 12:40:34 +0000 (12:40 +0000)]
Clean up internal mangling interface.

12 years agoValidate udev structures
Zdenek Kabelac [Sun, 4 Mar 2012 17:40:59 +0000 (17:40 +0000)]
Validate udev structures

Avoid using NULL pointers from udev. It seems like some older versions of udev
were improperly returning NULL in some case, so do not silently break here,
and give at least a warning to the user.

12 years agoReturn success for deactivation of thin pool
Zdenek Kabelac [Sun, 4 Mar 2012 17:36:23 +0000 (17:36 +0000)]
Return success for deactivation of thin pool

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.

12 years agoRemove part of FIXME
Zdenek Kabelac [Sun, 4 Mar 2012 16:05:42 +0000 (16:05 +0000)]
Remove part of FIXME

(and reindent a code below)

12 years agoUpdate thin test for thin_check
Zdenek Kabelac [Sun, 4 Mar 2012 16:04:52 +0000 (16:04 +0000)]
Update thin test for thin_check

Test if thin_check is present in system and disable its use, when its missing.

Add testing for poolmetadatasize.

FIXME: Allocation policy for metadata pool might need some relaxing.
(For now it needs to put all block on one PV.)

12 years agoSpeedup test run by a few minutes
Zdenek Kabelac [Sun, 4 Mar 2012 16:02:19 +0000 (16:02 +0000)]
Speedup test run by a few minutes

Reduce disc excercise for some test and focus on LVM testing by
using smaller extent size.

Reduce number of teardown_devs calls and use vg/lvremove instead.

Don't sleep for seconds on pvmove.

FIXME: shell/lvconvert-mirror-basic.sh seems to need more checking.
Test fails for smalled extent size then 512k.

12 years agoAdd FIXMEs
Zdenek Kabelac [Sun, 4 Mar 2012 15:58:31 +0000 (15:58 +0000)]
Add FIXMEs

Processing loop needs some thinking

12 years agoJust make error message more clear
Zdenek Kabelac [Sun, 4 Mar 2012 15:57:27 +0000 (15:57 +0000)]
Just make error message more clear

Make more obvious, the origin LV for snapshot must be active.

12 years agoScan all devices for lvmetad if 'pvscan --cache' used without device list.
Alasdair Kergon [Sat, 3 Mar 2012 18:32:53 +0000 (18:32 +0000)]
Scan all devices for lvmetad if 'pvscan --cache' used without device list.

12 years agoObtain VG list from lvmetad before relying on get_vgs, otherwise PVs in VGs
Alasdair Kergon [Sat, 3 Mar 2012 17:12:21 +0000 (17:12 +0000)]
Obtain VG list from lvmetad before relying on get_vgs, otherwise PVs in VGs
don't appear.

12 years agoReinstate accidentally-deleted line.
Alasdair Kergon [Sat, 3 Mar 2012 17:03:20 +0000 (17:03 +0000)]
Reinstate accidentally-deleted line.

12 years agopost-release
Alasdair Kergon [Sat, 3 Mar 2012 02:08:37 +0000 (02:08 +0000)]
post-release

12 years ago.
Alasdair Kergon [Sat, 3 Mar 2012 01:28:15 +0000 (01:28 +0000)]
.

12 years agopre-release
Alasdair Kergon [Sat, 3 Mar 2012 01:00:49 +0000 (01:00 +0000)]
pre-release

12 years agoCode refactoring
Zdenek Kabelac [Fri, 2 Mar 2012 23:01:10 +0000 (23:01 +0000)]
Code refactoring

Properly test for dm_asprintf result.
Keep unlocking of mutex in the same function and do not spread lock and
unlock over functions.

12 years agoJust space moving
Zdenek Kabelac [Fri, 2 Mar 2012 22:58:23 +0000 (22:58 +0000)]
Just space moving

Don't leave space on EOL.

12 years agoList _thread_registry missed mutex
Zdenek Kabelac [Fri, 2 Mar 2012 22:57:25 +0000 (22:57 +0000)]
List _thread_registry missed mutex

Operation on _thread_registry needs to be covered by mutex.
Cosmetic move a die code after free for valgind short leak list.

12 years agoa fixme
Alasdair Kergon [Fri, 2 Mar 2012 22:44:31 +0000 (22:44 +0000)]
a fixme

12 years agoSupport 16GB for thin pool metadata
Zdenek Kabelac [Fri, 2 Mar 2012 21:53:17 +0000 (21:53 +0000)]
Support 16GB for thin pool metadata

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.

12 years agofix non-lvmetad pvscan macro
Alasdair Kergon [Fri, 2 Mar 2012 21:50:02 +0000 (21:50 +0000)]
fix non-lvmetad pvscan macro

12 years agoAdd support for thin check
Zdenek Kabelac [Fri, 2 Mar 2012 21:49:43 +0000 (21:49 +0000)]
Add support for thin check

Use libdm callback to execute thin_check before activation
thin pool and after deactivation as well.

Supporting thin_check_executable which may pass in extra options for
the tool.

12 years agoPurge remaining trim bits from code
Zdenek Kabelac [Fri, 2 Mar 2012 21:43:26 +0000 (21:43 +0000)]
Purge remaining trim bits from code

12 years agomissing reply frees
Alasdair Kergon [Fri, 2 Mar 2012 21:24:37 +0000 (21:24 +0000)]
missing reply frees

12 years agoPass struct device around internally rather than dev_t.
Alasdair Kergon [Fri, 2 Mar 2012 20:46:36 +0000 (20:46 +0000)]
Pass struct device around internally rather than dev_t.
Add 3rd daemon return state "unknown" for lookups that are carried out
successfully but don't find the item requested.
Avoid issuing error messages when it's expected that a device that's
being looked up in lvmetad might not be there.

12 years agoMerge metadata size checking
Zdenek Kabelac [Fri, 2 Mar 2012 20:18:25 +0000 (20:18 +0000)]
Merge metadata size checking

Move the code for poolmetadatasize operation into one place.
Report override for minimum and maximum size.
Drop _read_thin_params function its error reporting is handled elsewhere.

12 years agoChange pvscan --lvmetad to pvscan --cache.
Alasdair Kergon [Fri, 2 Mar 2012 18:09:46 +0000 (18:09 +0000)]
Change pvscan --lvmetad to pvscan --cache.

12 years agoAdded dm_tree_node_set_callback() for preload and deactivation hooks
Zdenek Kabelac [Fri, 2 Mar 2012 17:31:21 +0000 (17:31 +0000)]
Added dm_tree_node_set_callback() for preload and deactivation hooks

Run users hook after preload for the node is finished,
or after the node has been deactivated.

12 years agoFix estimation of pool metadata device size
Zdenek Kabelac [Fri, 2 Mar 2012 17:25:21 +0000 (17:25 +0000)]
Fix estimation of pool metadata device size

If no size was give the later added minimal size check efectively
disable this code. Also the argument for size now must be kept
in sector_size, so adding division by SECTOR_SIZE (moved into
a const expression)

12 years agoAllow multiple device names with pvscan --lvmetad.
Alasdair Kergon [Fri, 2 Mar 2012 16:58:41 +0000 (16:58 +0000)]
Allow multiple device names with pvscan --lvmetad.
Hold global lock in pvscan --lvmetad.  (This might need refinement.)
Add PV name to "PV gone" messages.
Adjust some log message severities.  (More changes needed.)

12 years agoRemove support for TRIM message
Zdenek Kabelac [Fri, 2 Mar 2012 13:26:08 +0000 (13:26 +0000)]
Remove support for TRIM message

It's been unsupporte for now - and it's not going to be
implemented for thin pool kernel driver - so dropping
appearence of TRIM from libdm and lvm.

12 years agoReinstate error if e.g. daemon comms fails.
Alasdair Kergon [Fri, 2 Mar 2012 02:55:45 +0000 (02:55 +0000)]
Reinstate error if e.g. daemon comms fails.

12 years agoWait for mirror in sync
Zdenek Kabelac [Fri, 2 Mar 2012 00:08:48 +0000 (00:08 +0000)]
Wait for mirror in sync

1s could be too short

12 years agoCheck for error from dup2
Zdenek Kabelac [Thu, 1 Mar 2012 23:20:45 +0000 (23:20 +0000)]
Check for error from dup2

Do not take correct fd as error result.

12 years agoAdd traceback for failpath
Zdenek Kabelac [Thu, 1 Mar 2012 22:55:21 +0000 (22:55 +0000)]
Add traceback for failpath

12 years agoSkip zero length messages
Zdenek Kabelac [Thu, 1 Mar 2012 22:54:17 +0000 (22:54 +0000)]
Skip zero length messages

In case of zero length message, there would be a memory leak on
return path from _do_process_request.

12 years agoCheck allocated pointers
Zdenek Kabelac [Thu, 1 Mar 2012 22:52:59 +0000 (22:52 +0000)]
Check allocated pointers

Test pointers from allocation against NULL.
Error paths should be checked, some of them probably need
some extesions.

12 years agoImprove logging
Zdenek Kabelac [Thu, 1 Mar 2012 22:06:18 +0000 (22:06 +0000)]
Improve logging

Use %m for strerror.
Switch to use 0 for fail return code.

12 years agoCheck pointer before deref in debug message
Zdenek Kabelac [Thu, 1 Mar 2012 21:58:55 +0000 (21:58 +0000)]
Check pointer before deref in debug message

12 years agoExplicitely ignore fail from this label_read
Zdenek Kabelac [Thu, 1 Mar 2012 21:57:43 +0000 (21:57 +0000)]
Explicitely ignore fail from this label_read

This page took 0.069395 seconds and 5 git commands to generate.