Benjamin Marzinski [Thu, 14 May 2009 04:38:28 +0000 (23:38 -0500)]
multipath-tools: Improvement to max_fds
Setting max_fds to unlimited doesn't actually work. In the kernel, there is a
fixed limit to the maximum number of open fds a process can have. If you try
to set max_fds to greater than this, it fails. This patch replaces the special
value "unlimited" with a new special value, "max". If you set max_fds to "max",
multipath will use the actual system limit, which it looks up from
/proc/sys/fs/nr_open.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Benjamin Marzinski [Thu, 30 Apr 2009 16:31:37 +0000 (11:31 -0500)]
Add uid, gid, and mode config attributes
This adds the ability to set uid, gid, and mode for the multipath device
nodes. Also, kpartx created device nodes will inherit the uid, gid, and
mode of their parent device. These attributes can be set in either the
defaults or the multipaths section of /etc/multipath.conf
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Mike Snitzer [Wed, 29 Apr 2009 19:26:32 +0000 (15:26 -0400)]
multipathd: restrict /var/run/multipathd.sock permissions further
Use a more restrictive umask for /var/run/multipathd.sock
Group and Other do not need to access the socket.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Mike Snitzer [Wed, 29 Apr 2009 19:26:16 +0000 (15:26 -0400)]
fix small issues in cli_handlers
- properly check cli_list_wildcards()'s MALLOC returned pointer
- add missing newline to "blacklisted" reply
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Mike Snitzer [Wed, 29 Apr 2009 19:25:42 +0000 (15:25 -0400)]
cleanup various MALLOC/REALLOC callers
- alloc_hwe and alloc_mpe should've been used
- MALLOC and REALLOC returned pointer must be checked before use
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Mike Snitzer [Wed, 29 Apr 2009 19:25:09 +0000 (15:25 -0400)]
do not allow relative path names to be added to the pathvec
CVE-2009-0115 taught us that such paths should not be tolerated
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Mike Snitzer [Wed, 29 Apr 2009 19:24:24 +0000 (15:24 -0400)]
avoid NULL pointers in discovery code
Both sysfs_pathinfo() and path_offline() could hit NULL pointers when
trying to dereference the newly acquired parent sysfs_device. By being
more defensive a crash I experienced in path_offline() is avoided.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Mike Snitzer [Wed, 29 Apr 2009 19:23:28 +0000 (15:23 -0400)]
multipathd: add error checking to client's process_req
If multipathd (server) crashes fail any multipathd (client)
communication gracefully. This patch adds error checking that prevents
calls to recv_packet() if the send_packet() call failed.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Mike Snitzer [Tue, 28 Apr 2009 21:50:21 +0000 (23:50 +0200)]
[lib] fix double frees
A careless merge from Hannes' repo introduced double frees.
Christophe Varoqui [Wed, 22 Apr 2009 00:09:39 +0000 (02:09 +0200)]
[lib] prio_arg leaked here through cherry-picks
main patch is not yet merged. Remove for now.
Hannes Reinecke [Wed, 4 Mar 2009 12:37:43 +0000 (13:37 +0100)]
Strip trailing blanks from wwid
We should be stripping any trailing blanks from the wwid to catch
invalid wwid with just blanks.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 25 Feb 2009 15:09:15 +0000 (16:09 +0100)]
Only pass block events to multipath
We can easily preselect the events from a udev rule; this takes some
load off the multipath daemon itself.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Tue, 24 Feb 2009 13:13:56 +0000 (14:13 +0100)]
Fixup multibus zero-path handling
The 'multibus' pgpolicy function currently always adds a pathgroup,
even when no paths are present. This confuses setup_map() which
doesn't expect this behaviour.
References: 476330
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Thu, 19 Feb 2009 15:19:45 +0000 (16:19 +0100)]
Return 'ghost' state when port is in standby
Currently, issuing a TUR command on a path in standby
mode, it returns a Unit Attention status. This is
reported as a Failed path by the SLES11 tur checker.
Instead, this has to be reported as Ghost path based
on the additional sense code and sense code qualifier values.
References: 475816
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Mon, 9 Feb 2009 12:59:57 +0000 (13:59 +0100)]
Use noflush for kpartx
kpartx has to use 'noflush' to be able to complete for an all-paths
down scenario. Otherwise no uevent might be sent and multipath
will stall.
References: 473352
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Fri, 6 Feb 2009 13:57:12 +0000 (14:57 +0100)]
leastpending IO loadbalancing is not displayed properly
While using least pending IO lb policy for the LUN, the
dm_least_pending module is getting loaded and devices
are getting created properly. But multipath
display command does not display the paths for this LUN.
This is due to some explicit parsing for 'round-robin' in
dmparser.c, which has to be updated to accomodate other
load-balancing policies.
References: 444199
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 4 Feb 2009 10:34:15 +0000 (11:34 +0100)]
Retry tur on driver errors
Sending SG_IO commands might return driver errors, too. And
some of these errors should cause the tur to be retried.
References: bnc#456961
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 28 Jan 2009 08:24:10 +0000 (09:24 +0100)]
Plug memory leaks
Running the internal memory checker revealed quite some memory
leaks.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 28 Jan 2009 08:19:36 +0000 (09:19 +0100)]
Correct definition of dbg_malloc()
Any malloc function should return 'void *' to avoid type
mismatch errors.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Mon, 26 Jan 2009 12:01:33 +0000 (13:01 +0100)]
Error checking for VECTOR_XXX defines
Christophe doesn't believe in error checking. I, however, have
been tripping over them all the time.
References: bnc#469269
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Fri, 23 Jan 2009 12:04:15 +0000 (13:04 +0100)]
Fixup whitespace issues
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 21 Jan 2009 13:11:38 +0000 (14:11 +0100)]
Fix 'remove' uevent handling
A remove event might be handled after the failed devices have already
been purged from the multipath structure, so a failure here is not
an error.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 21 Jan 2009 12:26:51 +0000 (13:26 +0100)]
Update misleading debug messages for directio checker
io_getevents returns the number of events received, so '1'
is actual a success. And calling 'strerror(errno)' unconditionally
here will lead to false errors as the errno value won't be
updated then.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Thu, 15 Jan 2009 11:03:34 +0000 (12:03 +0100)]
Add new HP arrays to internal hardware table
This patch is to include entries for 2 new HP arrays in the hwtable.
References: bnc#442133
Signed-off-by: Vijayakumar Balasubramanian <vijayakumar@hp.com>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 7 Jan 2009 10:05:45 +0000 (11:05 +0100)]
Add more debugging output
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 7 Jan 2009 09:59:46 +0000 (10:59 +0100)]
Shuffle to the call to memset in logarea_init()
We should call memset only after we know the area has been
allocated.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 7 Jan 2009 09:52:25 +0000 (10:52 +0100)]
Fix sysfs_attr_get_value()
sysfs_attr_get_value() should return NULL if the attribute was not
found or found to be empty. And we should increase the attribute
value size to avoid overflows.
And overflows should be truncated, not ignored.
References: bnc#456747
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Fri, 19 Dec 2008 08:43:23 +0000 (09:43 +0100)]
Using condlog() for sysfs debugging output
The sysfs code still used 'printf' for debugging. Changed that
to using condlog().
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Thu, 18 Dec 2008 15:20:09 +0000 (16:20 +0100)]
Add missing unlock() on exit path
We forgot to unlock the paths on exit.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Thu, 18 Dec 2008 09:03:15 +0000 (10:03 +0100)]
Remove kdev_t.h include
Pointless and giving the users the wrong impression.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Tue, 9 Dec 2008 11:26:49 +0000 (12:26 +0100)]
Set 'max_fds' parameter for multipath
max_fds should be enabled for multipath, too.
References: 457443
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 3 Dec 2008 09:38:04 +0000 (10:38 +0100)]
Split off IBM ESS entries
For some reason the regular expression doesn't work properly.
So split off the two entries.
References: bnc#439763
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Mon, 24 Nov 2008 07:31:28 +0000 (08:31 +0100)]
Check vector_foreach_slot for NULL argument
Most crashes are due to vector_foreach_slot not checking for NULL
argument. Do that now, and add some more checks for not accessing
a NULL pointer indirection.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Fri, 21 Nov 2008 13:20:34 +0000 (14:20 +0100)]
multipathd crashes in update_multipath()
For an all-paths-down scenario we can have an mpp with no
paths attached. So check for it.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Thu, 20 Nov 2008 12:33:10 +0000 (13:33 +0100)]
Allow zero paths for device-mapper strings
There is no reason why we shouldn't allow zero paths in the device
mapper output.
References: 435688
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Thu, 20 Nov 2008 11:33:39 +0000 (12:33 +0100)]
Search for correct hwe in update_multipath()
When updating the multipath structure we only should update the
hardware entry if it doesn't exist anymore. And we should take
care of not selecting invalid entries in the paths vector as
the hardware entry for this is invalid, too.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 19 Nov 2008 10:04:40 +0000 (11:04 +0100)]
Valgrind fixes for checkers
Declare an array does not zero it out. Hence we might read some
random garbage here.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Tue, 18 Nov 2008 10:33:09 +0000 (11:33 +0100)]
More valgrind fixes
valgrind found some memory which wasn't freed on exit.
Do it now.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Fri, 14 Nov 2008 09:02:56 +0000 (10:02 +0100)]
libmultipath: zero out lines in print.c
static declaration of line within a function doesn't place them in
the bss segment, hence they are not zeroed automatically.
Gives funny effects occasionally.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Tue, 11 Nov 2008 12:44:09 +0000 (13:44 +0100)]
Update hardware table for new HP arrays
References: 442133
Signed-off-by: Vijayakumar Balasubramanian <vijaykumar@hp.com>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Tue, 4 Nov 2008 08:40:26 +0000 (09:40 +0100)]
Check for empty mpvecs in mpvec_garbage_collector()
During shutdown the mpvec pointer can indeed be empty, so we should
check it first before trying to access it.
References: 437245
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Thu, 30 Oct 2008 12:46:15 +0000 (13:46 +0100)]
Do not check for valid mp context in get_state()
When called from 'multipath -ll' get_state() does not
have a valid context. But the checkers should run nevertheless
as we need the state here but are not interested in any long-term
issues.
References: 433659
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 29 Oct 2008 15:58:52 +0000 (16:58 +0100)]
Fixup %n to handle '!' special case
The kernel name may contain an '!', which should be translated
into a '/'. So do it here.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Tue, 21 Oct 2008 08:22:16 +0000 (10:22 +0200)]
Increase buffer size in find_loop_by_file()
find_loop_by_file() uses an internal buffer of 20 chars, which is compared
to a buffer of the size of 64 chars. Not good.
References: 436428
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Tue, 14 Oct 2008 06:38:29 +0000 (08:38 +0200)]
kpartx -l does not remove it's loop device
When doing a kpartx -l it does not remove it's loop device when done.
The appended patch to multipath-tools could fix this.
References: 417266
Signed-off-by: Philipp Zimmer <pzimmer@novell.com>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Tue, 14 Oct 2008 06:21:02 +0000 (08:21 +0200)]
Read the verbosity level from multipath.conf configuration file
With this patch, multipathd can read the verbosity level from
the multipath.conf configuration file.
For example:
verbosity 5
References: 388284
Signed-off-by: Ritesh Raj Sarraf <rsarraf@netapp.com>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Thu, 9 Oct 2008 10:53:38 +0000 (12:53 +0200)]
fopen returns NULL on failure, not -1
References: 432598
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Fri, 26 Sep 2008 13:31:07 +0000 (15:31 +0200)]
multipath: Check return value for get_inq()
Check the return value of get_inq() to correctly handle devices
with unconnected LUNs.
References: 419123
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 23 Jul 2008 10:11:08 +0000 (12:11 +0200)]
Further whitespace cleanup
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 23 Jul 2008 10:04:40 +0000 (12:04 +0200)]
Increase RDAC priority values
A priority value of '0' will cause a path to never be selected,
so we should make sure to return always a non-zero value for
available paths.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Hannes Reinecke [Wed, 23 Jul 2008 09:20:53 +0000 (11:20 +0200)]
Valgrind fixes
Valgrind found some issues. And clear up whitespaces while we're at it.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Christophe Varoqui [Sun, 19 Apr 2009 22:44:38 +0000 (00:44 +0200)]
[lib] set PASSCRED netlink socket option
Christophe Varoqui [Sun, 19 Apr 2009 22:16:58 +0000 (00:16 +0200)]
[lib] check udev and netlink messages sender uid
Ignore if not 0.
This change was suggested by Kay Sievers. It required switching from
recv() to recvmsg().
Mike Snitzer [Fri, 17 Apr 2009 22:07:58 +0000 (18:07 -0400)]
multipath-tools: install libmultipath.so in $syslibdir
Allows libmultipath.so to be installed in the proper lib dir
(e.g. /lib64). $libdir is already used for /lib/multipath
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Mike Snitzer [Fri, 17 Apr 2009 22:07:59 +0000 (18:07 -0400)]
multipath-tools: update multipathd.init.redhat
Various small improvements to Red Hat's multipathd initscript.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Mike Snitzer [Fri, 17 Apr 2009 22:07:57 +0000 (18:07 -0400)]
multipath-tools: update scsi_id arguments
Switch over to using --whitelisted and --device
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Christophe Varoqui [Sun, 19 Apr 2009 21:44:28 +0000 (23:44 +0200)]
[lib] define _GNU_SOURCE in preparation for using recvmsg()
And deal with consequences :
1/ rename util::basename() to util::basenamecpy() to resolve
namespace conflict GNU version
2/ don't redefine _GNU_SOURCE if already set in regex.c to avoid
a build warning
Christophe Varoqui [Wed, 15 Apr 2009 19:47:01 +0000 (21:47 +0200)]
[lib] more selective mask to catch uevent from netlink
Kay Sievers reports 0x01 is the only group used by the kernel.
As udev will start using other groups, pollution will result from
our lack of filtering.
Christophe Varoqui [Fri, 3 Apr 2009 22:33:25 +0000 (00:33 +0200)]
[lib] bump version to 0.4.9 in preparation for release
Christophe Varoqui [Fri, 3 Apr 2009 22:29:49 +0000 (00:29 +0200)]
[multipathd] actually propose restorequeuing et disablequeuing in CLI
this bit was missing from the forward-port
Benjamin Marzinski [Fri, 3 Apr 2009 22:09:55 +0000 (00:09 +0200)]
[multipathd] Add options to multipathd to turn off queueing
Even when the last path of a multipath device is deleted, it can't be
removed until all the queued IO is flushed. For devices that have
no_path_retry set to queue, this doesn't automatically happen.
This patch adds a "flush_on_last_del" config file option, that causes the
multipath device to automatically turn off queueing when the last path is
deleted. It also adds the "disablequeueing" and "restorequeueing"
multipathd cli commands.
Benjamin Marzinski [Fri, 13 Mar 2009 20:55:14 +0000 (15:55 -0500)]
set pthread stack size to at least PTHREAD_STACK_MIN
Setting the stacksize too small just causes
pthread_attr_setstacksize() to fail, leaving you with the default stack
size. On some architectures, the default stacksize is large, like 10Mb.
Since you start one waiter thread per multipath device, every 100
devices eats up 1Gb of memory.
The other problem is that when I actually read the pthread_attr_init man
page (it can fail. who knew?), I saw that it can fail with ENOMEM. Also,
that it had a function to free it, and that the result of reinitializing
an attr that hadn't been freed was undefined. Clearly, this function
wasn't intended to be called over and over without ever freeing the
attr, which is how we've been using it in multipathd. So, in the spirit
of writing code to the interface, instead of to how it appears to be
currently implemented, how about this.
Benjamin Marzinski [Thu, 12 Mar 2009 18:38:11 +0000 (13:38 -0500)]
remove deleted path from pathvec
When the last path in a multipath map was removed, the path wasn't getting
deleted from the pathvec before it was getting freed.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Benjamin Marzinski [Fri, 3 Apr 2009 04:19:36 +0000 (23:19 -0500)]
Clean up multipath linking
This is a patch to fix up the linking. It does two things. First, it makes
libmultipath.so install to /lib/ just like a normal shared library, so you
don't have to use -rpath to link to it. Second, and more importantly,
it moves the libaio linking into libcheckdirectio.so, where it belongs. Since
libcheckdirectio.so is a dynamic shared object, multipath and multipathd don't
know what functions they need to link in from libaio. This fixes the directio
lockup for me.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Benjamin Marzinski [Fri, 3 Apr 2009 04:44:19 +0000 (23:44 -0500)]
Miscellaneous multipath bugfixes
This is mostly a cleanup of some bugs that recently got introduced. In
ACT_RESIZE we were trying to create a read-only device before we tried
to create a read/write one (I also added the ability to fail back to
read-only in ACT_RELOAD). There were some printouts that I assume were
for debugging, and some duplicate code. And I switched it so that
dm_simplecmd_flush did flushing, and dm_simplecmd_noflush didn't.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Christophe Varoqui [Thu, 2 Apr 2009 21:23:12 +0000 (23:23 +0200)]
[multipathd] minor logging fix for the resize handler
add missing format param
Christophe Varoqui [Thu, 2 Apr 2009 21:17:29 +0000 (23:17 +0200)]
[lib] use explicit names dm_simplecmd_{flush,noflush}
easier than the magic value used in static dm_simple_cmd to drive
the flush/noflush behaviour
Konrad Rzeszutek [Thu, 2 Apr 2009 21:04:17 +0000 (23:04 +0200)]
Support resizing of multipath maps.
This is patch that initially showed up on dm-devel mailing list:
http://www.linux-archive.org/device-mapper-development/162594-multipath-tools-libmultipath-configure-c-libmu.html
which was posted on dm-devel mailing list, but never ported
over to work with the git version. This forward-port by me
works.
Benjamin Marzinski [Thu, 2 Apr 2009 20:36:41 +0000 (22:36 +0200)]
[multipathd] signal deadlock
If multipathd is run with -v3, both the SIGHUP, and the SIGUSR1 signal handlers
will log a message. If a multipathd thread receives one of these signals while
it has a log lock held, it deadlocks itself. Also, the SIGHUP handler will grab
the vecs lock, so if any thread receives a SIGHUP while holding the vecs lock,
it deadlocks itself. This commit blocks the appropriate signals to guard
against this.
Hannes Reinecke [Wed, 1 Apr 2009 20:31:01 +0000 (22:31 +0200)]
[multipathd] /var/run/multipathd.sock is world-writable
Due to an stray 'umask()' the socket file is in fact world-writable,
allowing for an easy exploit.
References: 458598
Hannes Reinecke [Fri, 27 Mar 2009 22:50:18 +0000 (23:50 +0100)]
Stack overflow in uev_trigger()
uev_discard uses sscanf to write a 10 byte string into an array,
but I forgot to take the trailing NULL byte into account.
Konrad Rzeszutek [Tue, 24 Mar 2009 18:14:37 +0000 (19:14 +0100)]
Don't display the state of the newly added map during addition in the daemon
and don't switch groups.
A previous commit mass-changed #ifdef DAEMON to check for 'mpp->waiter'.
Unfortunatly when the 'domap' function is called with ACT_CREATE in the daemon,
the mpp->waiter is not set, hence the multipath client mode logic is choosen.
Fixing this triggers another issues which is that newly added path via
ACT_CREATE won't have their waitevent thread created as the caller checks
mpp->action (which changed to ACT_NOTHING) and won't start the thread.
Konrad Rzeszutek [Tue, 24 Mar 2009 18:04:31 +0000 (19:04 +0100)]
Free prio_name and checker_name now that we dynamically allocate them.
In earlier releases the prio_name and checker_name would point
to statically allocated values. Now that we strdup them, we need
to free them during shutdown.
Konrad Rzeszutek [Tue, 24 Mar 2009 18:01:05 +0000 (19:01 +0100)]
Unload prio and checkers libraries during shutdown.
Our statup sequence is 'load_config', 'init_checkers', and 'init_prio'.
Both init_* functions reset the list of prio and checkers, which is
unfortunate as in load_config, depending on the multipat.conf, would
load prio and checker libraries. This results in double-loading of
the libraries and a memory leak.
Konrad Rzeszutek [Tue, 24 Mar 2009 17:49:14 +0000 (18:49 +0100)]
Race-condition fix with free_waiter threads during shutdown.
When we shutdown, the main process locks the mutex, causing
all of the free_waiter functions to pile up on their lock.
Once we unlock in the main process, all of the free_waiters
start working. However the next instruction in the main proces
is to destroy the mutex. The short window is all the free_waiter
threads have to do their cleanup before they attempt to unlock
the mutex - which might have been de-allocated (and set to NULL).
End result can be a seg-fault.
This fix adds a ref-count to the mutex so that during shutdown
we spin and wait until all of the free_waiter functions
have completed and the ref-count is set to zero.
Konrad Rzeszutek [Tue, 24 Mar 2009 17:22:06 +0000 (18:22 +0100)]
[checkers] update doc for path checkers and use MALLOC instead of malloc
The internal memory debugger throws a fit if the path checkers are
using the malloc instead of the 'MALLOC' macro.
Konrad Rzeszutek [Tue, 24 Mar 2009 17:16:48 +0000 (18:16 +0100)]
[checkers] Allocate only once the EMC-clariion global checker's stage area
The multi-path global context on every path checker init would be
set. In scenarios where you have four paths per multipath, that context
would be set four times instead of once causing a small memory leak
(4 bytes per block disk).
Babu Moger [Sat, 7 Mar 2009 14:49:10 +0000 (15:49 +0100)]
[lib] add LSI hardware settings to internal defaults
This patch adds the hwtable for LSI rdac generic storage. Also sets the feature pg_init_retries to maximum(50).
Benjamin Marzinski [Sun, 18 Jan 2009 22:44:51 +0000 (23:44 +0100)]
[lib] plug fd leak in callout error path and and error messages
multipathd was not releasing fds when a fork failed in execute_program.
Also, the error messages from execute_program were not very
helpful. This patch fixes both
Christophe Varoqui [Wed, 17 Dec 2008 23:13:30 +0000 (00:13 +0100)]
[lib] dm_flush_map on a non-multipath devmap do nothing and returns 0
So that "multipath -F" can succeed when non-multipath maps are active
on the system.
Christophe Varoqui [Wed, 17 Dec 2008 23:01:37 +0000 (00:01 +0100)]
[lib] don't add up errors in dm_flush_maps()
to avoid return code overflow, as pointed by "Bryn M. Reeves",
Redhat.
Christophe Varoqui [Wed, 17 Dec 2008 22:39:47 +0000 (23:39 +0100)]
[multipathd] log the uevent discarding as debug information
Konrad Rzeszutek [Wed, 17 Dec 2008 22:24:40 +0000 (23:24 +0100)]
[lib] Memory leak when deleting and re-adding device-mapper entries.
If we remove a map (with configuration details specified in the multipath.conf file)
and later re-add it, the pointer to the config is over-written without first
de-allocating the values.
Konrad Rzeszutek [Wed, 17 Dec 2008 22:23:09 +0000 (23:23 +0100)]
[multipathd] Memory leak in ev_add_map.
When using dm_mapname it makes a strdup of the returned value. We use
the dm_mapname return value (alias) in our function but neglected
to free it at the exit points.
Konrad Rzeszutek [Wed, 17 Dec 2008 22:22:01 +0000 (23:22 +0100)]
[lib] The internal debug memory leak printout has output jumbled.
When using the internal memory leak the function that call realloc get a
confused output. The fix puts the list of operation in the right
order.
Konrad Rzeszutek [Wed, 17 Dec 2008 22:20:44 +0000 (23:20 +0100)]
[lib] Need to use FREE macro instead of free.
When using the internal memory leak checker it is imperative that
all free use the FREE macro. Otherwise accounting errors show up.
Konrad Rzeszutek [Wed, 17 Dec 2008 22:19:19 +0000 (23:19 +0100)]
[lib] Substantial memory leak when large amount of LUNs are present
dm_get_name() was creating a vector, populating it with multipaths,
then only deleting the vector itself, which leaked all of the multipaths.
If the number of LUNs (ie. multipaths) was large then the leak was equally large.
Signed-off-by: Konrad Rzeszutek <konrad@virtualiron.com>
Benjamin Marzinski [Thu, 11 Dec 2008 22:12:40 +0000 (16:12 -0600)]
Fix user_friendly_names for WWIDs with spaces
The user_friendly_names code didn't handle WWIDs with spaces correctly. This
fixes it by taking everything from the start of the WWID to the end of the line
as the WWID. The only downside is that if someone hand edits their bindings
file, they may leave stray whitespace at the end of the line, and cause it to
not match. This could ne solved by quoting the WWIDs and escaping any quotes
that they might have, but I don't think that people hand edit their bindings
files enough to make it a problem.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Benjamin Marzinski [Thu, 11 Dec 2008 22:11:40 +0000 (16:11 -0600)]
Fix broken major:minor device handling
devt2devname was not actually setting devname, so multipath was not able to
do operate on devices using major:minor
This patch sets devname, and also checks for an unlikely buffer overrun.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Benjamin Marzinski [Thu, 11 Dec 2008 22:10:27 +0000 (16:10 -0600)]
Fix for parsing multipath.conf files without spaces before brackets
There's a bug in bug in the multipath.conf code that keeps multipath from
correctly parsing config files where there is no space between a section name
and the opening bracket. For instance
devices {
device {
...
}
}
works but
devices {
device{
...
}
}
doesn't. This patch makes sure that brackets are the recognized as seperate from
the token that they follow, unless they are part of a quoted string.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Benjamin Marzinski [Thu, 11 Dec 2008 22:07:25 +0000 (16:07 -0600)]
config file cleanup and a defaults file
This patch cleans up multipath.conf.annotated and multipath.conf.synthetic adds
a multipath.conf.defaults file that lists the builtin default configs.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Benjamin Marzinski [Thu, 11 Dec 2008 22:06:07 +0000 (16:06 -0600)]
Fix for kpartx on sparc64
This typedef change is necessary to make kpartx work on sparc64 machines.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Benjamin Marzinski [Thu, 11 Dec 2008 22:03:28 +0000 (16:03 -0600)]
Fix for uevent devpath handling
This is necessary to make uevents work on fedora, since devpath appears as
something like:
'/devices/pci0000:00/0000:00:0a.0/0000:06:00.0/host11/rport-11:0-1/target11:0:1/11:0:1:0/block/sdi'
It simply strips off the everything up to the /block.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Christophe Varoqui [Sun, 30 Nov 2008 23:35:42 +0000 (00:35 +0100)]
[doc] path_checker can't be set in a multipath {} config block
All docs wrongly stated that as valid. It is not, mainly
because path checker selection happens before assigning the
path to a multipath.
Thanks to Matthew Kent for pointing this out.
Christophe Varoqui [Sun, 30 Nov 2008 22:56:46 +0000 (23:56 +0100)]
[lib] add '%c' wildcard to display path checker type in cli
$ sudo multipathd show paths format "%d %i %c"
dev hcil checker
sdb 5:0:0:0 readsector0
sdc 5:0:0:1 readsector0
Matthew Kent [Sun, 30 Nov 2008 22:55:10 +0000 (23:55 +0100)]
[build] rpm specfile fix
catch up with recent file movement.
Christophe Varoqui [Mon, 24 Nov 2008 22:14:59 +0000 (23:14 +0100)]
[multipath] set "multipath -f|-F" return code correctly
Thanks to Diedrich Ehlerding <diedrich.ehlerding@fujitsu-siemens.com>
for reporting this.
Christophe Varoqui [Thu, 20 Nov 2008 22:16:01 +0000 (23:16 +0100)]
strip ANSI color markers in non-interactive mode
Tested with piped and redirected stdout, using either "multipath -l"
or "multipathd show topo". Seems there are no more blatant regression.
Let me know if there are, because I'm tempted to go a little further
with those ANSI codes by highlighting degrated/failure situation.
Christophe Varoqui [Wed, 19 Nov 2008 23:41:45 +0000 (00:41 +0100)]
[multipathd] easier multipathd interaction method
Previously using the CLI commands from the shell was done using :
multipathd -k"show paths format \"%w %d %i %s %t %T %D\""
Now you can also use :
multipathd -- show paths format "%w %d %i %s %t %T %D"
or
multipathd show paths format "%w %d %i %s %t %T %D"
Christophe Varoqui [Wed, 19 Nov 2008 22:26:59 +0000 (23:26 +0100)]
[lib] fix quote doubling when using "show conf"