When converting a lock, an lkb is in the granted state and also being used
to request a new state. In the case that the conversion was a "try 1cb"
type which has failed, and if the new state was incompatible with the old
state, a callback was being generated to the requesting node. This is
incorrect as callbacks should only be sent to all the other nodes holding
blocking locks. The requesting node should receive the normal (failed)
response to its "try 1cb" conversion request only.
This was discovered while debugging a performance problem on GFS2, however
this fix also speeds up GFS as well. In the GFS2 case the performance gain
is over 10x for cases of write activity to an inode whose glock is cached
on another, idle (wrt that glock) node.
(comment added, dct)
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Tested-by: Abhijith Das <adas@redhat.com>
Signed-off-by: David Teigland <teigland@redhat.com>
Smatch complains because "lkb" is never NULL. Looking at it, the original
code actually adds the new element to the end of the list fine, so we can
just get rid of the if condition. This code is four years old and no one
has complained so it must work.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David Teigland <teigland@redhat.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
The bast mode that appears in the debugfs output should be
useful on both master and process nodes. lkb_highbast is
currently printed, and is only useful on the master node.
lkb_bastmode is only useful on the process node. This
patch sets lkb_bastmode on the master node as well, and
uses that value in the debugfs print.
Signed-off-by: David Teigland <teigland@redhat.com>
When the lock master processes a successful operation (request,
convert, cancel, or unlock), it will process the effects of the
change before sending the reply for the operation. The "effects"
of the operation are:
- blocking callbacks (basts) for any newly granted locks
- waiting or converting locks that can now be granted
The cast is queued on the local node when the reply from the lock
master is received. This means that a lock holder can receive a
bast for a lock mode that is doesn't yet know has been granted.
Signed-off-by: David Teigland <teigland@redhat.com>
When both blocking and completion callbacks are queued for lock,
the dlm would always deliver the completion callback (cast) first.
In some cases the blocking callback (bast) is queued before the
cast, though, and should be delivered first. This patch keeps
track of the order in which they were queued and delivers them
in that order.
This patch also keeps track of the granted mode in the last cast
and eliminates the following bast if the bast mode is compatible
with the preceding cast mode. This happens when a remotely mastered
lock is demoted, e.g. EX->NL, in which case the local node queues
a cast immediately after sending the demote message. In this way
a cast can be queued for a mode, e.g. NL, that makes an in-transit
bast extraneous.
Signed-off-by: David Teigland <teigland@redhat.com>
Replace all GFP_KERNEL and ls_allocation with GFP_NOFS.
ls_allocation would be GFP_KERNEL for userland lockspaces
and GFP_NOFS for file system lockspaces.
It was discovered that any lockspaces on the system can
affect all others by triggering memory reclaim in the
file system which could in turn call back into the dlm
to acquire locks, deadlocking dlm threads that were
shared by all lockspaces, like dlm_recv.
Signed-off-by: David Teigland <teigland@redhat.com>
CC [M] fs/dlm/lock.o
fs/dlm/lock.c: In function ‘find_rsb’:
fs/dlm/lock.c:438: warning: ‘r’ may be used uninitialized in this function
Since r is used on the error path to set r_ret, set it to NULL.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: David Teigland <teigland@redhat.com>
Return immediately from dlm_unlock(CANCEL) if the lock is
granted and not being converted; there's nothing to cancel.
Signed-off-by: David Teigland <teigland@redhat.com>
When a conversion completes successfully and finds that a cancel
of the convert is still in progress (which is now a moot point),
preemptively clear the state associated with outstanding cancel.
That state could cause a subsequent conversion to be ignored.
Also, improve the consistency and content of error and debug
messages in this area.
Signed-off-by: David Teigland <teigland@redhat.com>
The rwlock is almost always used in write mode, so there's no reason
to not use a spinlock instead.
Signed-off-by: David Teigland <teigland@redhat.com>
Record the time the latest blocking callback was queued for
a lock. This will be used for debugging in combination with
lock queue timestamp changes in the previous patch.
Signed-off-by: David Teigland <teigland@redhat.com>
Use ktime instead of jiffies for timestamping lkb's. Also stamp the
time on every lkb whenever it's added to a resource queue, instead of
just stamping locks subject to timeouts. This will allow us to use
timestamps more widely for debugging all locks.
Signed-off-by: David Teigland <teigland@redhat.com>
The lkb bastmode value is set in the context of processing the
lock, and read by the dlm_astd thread. Because it's accessed
in these two separate contexts, the writing/reading ought to
be done under a lock. This is simple to do by setting it and
reading it when the lkb is added to and removed from dlm_astd's
callback list which is properly locked.
Signed-off-by: David Teigland <teigland@redhat.com>
gcc 4.3.0 correctly emits the following warning.
search_rsb_list does not *r_ret if no dlm_rsb is found
and _search_rsb may pass the uninitialized value upstream
on the error path when both calls to search_rsb_list
return non-zero error.
The fix sets *r_ret to NULL on search_rsb_list's not-found path.
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: David Teigland <teigland@redhat.com>
The fix in commit 3650925893 was addressing
the case of a granted PR lock with waiting PR and CW locks. It's a
special case that requires forcing a CW bast. However, that forced CW
bast was incorrectly applying to a second condition where the granted
lock was CW. So, the holder of a CW lock could receive an extraneous CW
bast instead of a PR bast. This fix narrows the original special case to
what was intended.
Signed-off-by: David Teigland <teigland@redhat.com>
When a NOQUEUE request fails, the rsb res_master field is unnecessarily
reset to -1, instead of leaving the valid master setting in place. We
want to save the looked-up master values while the rsb is on the "toss
list" so that another lookup can be avoided if the rsb is soon reused.
The fix is to simply leave res_master value alone.
Signed-off-by: David Teigland <teigland@redhat.com>
We *can* get there from receive_request() and dlm_recover_master_copy()
with namelen too large if incoming request is invalid; BUG() from
DLM_ASSERT() in allocate_rsb() is a bit excessive reaction to that
and in case of dlm_recover_master_copy() we would actually oops before
that while calculating hash of up to 64Kb worth of data - with data
actually being 64 _bytes_ in kmalloc()'ed struct.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Teigland <teigland@redhat.com>
* check that length is large enough to cover the non-variable part of message or
rcom resp. (after checking that it's large enough to cover the header, of
course).
* kill more pointless casts
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Teigland <teigland@redhat.com>
a) don't cast the pointer to dlm_header *, we use it as dlm_message *
anyway.
b) we copy the message into a queue element, then pass the pointer to
copy to dlm_receive_message_saved(); declare it properly to make sure
that we have the right alignment.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Teigland <teigland@redhat.com>
To prevent the master of an rsb from changing rapidly, an unused rsb is kept
on the "toss list" for a period of time to be reused. The toss list was
being cleared completely for each recovery, which is unnecessary. Much of
the benefit of the toss list can be maintained if nodes keep rsb's in their
toss list that they are the master of. These rsb's need to be included
when the resource directory is rebuilt during recovery.
Signed-off-by: David Teigland <teigland@redhat.com>
The invalid lockspace messages are normal and can appear relatively
often. They should be suppressed without debugging enabled.
Signed-off-by: David Teigland <teigland@redhat.com>
In a rare case we may need to repeat a local resource directory lookup
due to a race with removing the rsb and removing the resdir record.
We'll never need to do more than a single additional lookup, though,
so the infinite loop around the lookup can be removed. In addition
to being unnecessary, the infinite loop is dangerous since some other
unknown condition may appear causing the loop to never break.
Signed-off-by: David Teigland <teigland@redhat.com>
Non-forced unlocks should be rejected if the lock is waiting on the
rsb_lookup list for another lock to establish the master node.
Signed-off-by: David Teigland <teigland@redhat.com>
There was some hit and miss validation of messages that has now been
cleaned up and unified. Before processing a message, the new
validate_message() function checks that the lkb is the appropriate type,
process-copy or master-copy, and that the message is from the correct
nodeid for the the given lkb. Other checks and assertions on the
lkb type and nodeid have been removed. The assertions were particularly
bad since they would panic the machine instead of just ignoring the bad
message.
Although other recent patches have made processing old message unlikely,
it still may be possible for an old message to be processed and caught
by these checks.
Signed-off-by: David Teigland <teigland@redhat.com>
Messages from nodes that are no longer members of the lockspace should be
ignored. When nodes are removed from the lockspace, recovery can
sometimes complete quickly enough that messages arrive from a removed node
after recovery has completed. When processed, these messages would often
cause an error message, and could in some cases change some state, causing
problems.
Signed-off-by: David Teigland <teigland@redhat.com>
When a failed request (EBADR or ENOTBLK) is unlocked/canceled instead of
retried, there may be other lkb's waiting on the rsb_lookup list for it
to complete. A call to confirm_master() is needed to move on to the next
waiting lkb since the current one won't be retried.
Signed-off-by: David Teigland <teigland@redhat.com>
When recovery looks at locks waiting for replies, it fails to consider
locks that have already received a reply for their first remote operation,
but not received a reply for secondary, overlapping unlock/cancel. The
appropriate stub reply needs to be called for these waiters.
Appears when we start doing recovery in the presence of a many overlapping
unlock/cancel ops.
Signed-off-by: David Teigland <teigland@redhat.com>
The lkb_ast_type field indicates whether the lkb is on the astqueue list.
When clearing locks for a process, lkb's were being removed from the astqueue
list without clearing the field. If release_lockspace then happened
immediately afterward, it could try to remove the lkb from the list a second
time.
Appears when process calls libdlm dlm_release_lockspace() which first
closes the ls dev triggering clear_proc_locks, and then removes the ls
(a write to control dev) causing release_lockspace().
Signed-off-by: David Teigland <teigland@redhat.com>
The dlm functions in memory.c should use the dlm_ prefix. Also, use
kzalloc/kfree directly for dlm_direntry's, removing the wrapper functions.
Signed-off-by: David Teigland <teigland@redhat.com>
Change log_error() to log_debug() for conditions that can occur in
large number in normal operation.
Signed-off-by: David Teigland <teigland@redhat.com>
This patch adds a proper prototype for some functions in
fs/dlm/dlm_internal.h
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Teigland <teigland@redhat.com>
Introduce a per-lockspace rwsem that's held in read mode by dlm_recv
threads while working in the dlm. This allows dlm_recv activity to be
suspended when the lockspace transitions to, from and between recovery
cycles.
The specific bug prompting this change is one where an in-progress
recovery cycle is aborted by a new recovery cycle. While dlm_recv was
processing a recovery message, the recovery cycle was aborted and
dlm_recoverd began cleaning up. dlm_recv decremented recover_locks_count
on an rsb after dlm_recoverd had reset it to zero. This is fixed by
suspending dlm_recv (taking write lock on the rwsem) before aborting the
current recovery.
The transitions to/from normal and recovery modes are simplified by using
this new ability to block dlm_recv. The switch from normal to recovery
mode means dlm_recv goes from processing locking messages, to saving them
for later, and vice versa. Races are avoided by blocking dlm_recv when
setting the flag that switches between modes.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
If the castaddr passed to the userland API is NULL then don't overwrite the
existing castparam. This allows a different thread to cancel a lock request and
the CANCEL AST gets delivered to the original thread.
bz#306391 (for RHEL4) refers.
Signed-Off-By: Patrick Caulfield <pcaulfie@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Fix a long standing bug where a blocking callback would be missed
when there's a granted lock in PR mode and waiting locks in both
PR and CW modes (and the PR lock was added to the waiting queue
before the CW lock). The logic simply compared the numerical values
of the modes to determine if a blocking callback was required, but in
the one case of PR and CW, the lower valued CW mode blocks the higher
valued PR mode. We just need to add a special check for this PR/CW
case in the tests that decide when a blocking callback is needed.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Add a new flag, DLM_LSFL_FS, to be used when a file system creates a lockspace.
This flag causes the dlm to use GFP_NOFS for allocations instead of GFP_KERNEL.
(This updated version of the patch uses gfp_t for ls_allocation.)
Signed-Off-By: Patrick Caulfield <pcaulfie@redhat.com>
Signed-Off-By: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Add a function that can be used through libdlm by a system daemon to cancel
another process's deadlocked lock. A completion ast with EDEADLK is returned
to the process waiting for the lock.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Various fixes related to the new timeout feature:
- add_timeout() missed setting TIMEWARN flag on lkb's when the
TIMEOUT flag was already set
- clear_proc_locks should remove a dead process's locks from the
timeout list
- the end-of-life calculation for user locks needs to consider that
ETIMEDOUT is equivalent to -DLM_ECANCEL
- make initial default timewarn_cs config value visible in configfs
- change bit position of TIMEOUT_CANCEL flag so it's not copied to
a remote master node
- set timestamp on remote lkb's so a lock dump will display the time
they've been waiting
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
A one liner fix which got missed from the earlier patches.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Fabio Massimo Di Nitto <fabbione@ubuntu.com>
Cc: David Teigland <teigland@redhat.com>
In the rush to get the previous patch set sent, a compilation bug I fixed
shortly before sending somehow got clobbered, probably by a missed quilt
refresh or something.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
When conversion deadlock is detected, cancel the conversion and return
EDEADLK to the application. This is a new default behavior where before
the dlm would allow the deadlock to exist indefinately.
The DLM_LKF_NODLCKWT flag can now be used in a conversion to prevent the
dlm from performing conversion deadlock detection/cancelation on it.
The DLM_LKF_CONVDEADLK flag can continue to be used as before to tell the
dlm to demote the granted mode of the lock being converted if it gets into
a conversion deadlock.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Change the user/kernel device interface used by libdlm:
- Add ability for userspace to check the version of the interface. libdlm
can now adapt to different versions of the kernel interface.
- Increase the size of the flags passed in a lock request so all possible
flags can be used from userspace.
- Add an opaque "xid" value for each lock. This "transaction id" will be
used later to associate locks with each other during deadlock detection.
- Add a "timeout" value for each lock. This is used along with the
DLM_LKF_TIMEOUT flag.
Also, remove a fragment of unused code in device_read().
This patch requires updating libdlm which is backward compatible with
older kernels.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
New features: lock timeouts and time warnings. If the DLM_LKF_TIMEOUT
flag is set, then the request/conversion will be canceled after waiting
the specified number of centiseconds (specified per lock). This feature
is only available for locks requested through libdlm (can be enabled for
kernel dlm users if there's a use for it.)
If the new DLM_LSFL_TIMEWARN flag is set when creating the lockspace, then
a warning message will be sent to userspace (using genetlink) after a
request/conversion has been waiting for a given number of centiseconds
(configurable per node). The time warnings will be used in the future
to do deadlock detection in userspace.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Steven Whitehouse
Dan Carpenter
Tejun Heo
David Teigland
Benny Halevy
Adrian Bunk
Al Viro
Patrick Caulfield