This implementation of rw_tryupgrade() behaves slightly differently
from its counterparts on other platforms. It drops the RW_READER lock
and then acquires the RW_WRITER lock leaving a small window where no
lock is held. On other platforms the lock is never released during
the upgrade process. This is necessary under Linux because the kernel
does not provide an upgrade function.
There are currently no callers in the ZFS code where this change in
behavior is a problem. In fact, in most cases the code is already
written such that if the upgrade fails the RW_READER lock is dropped
and the caller blocks waiting to acquire the lock as RW_WRITER.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Matthew Thode <prometheanfire@gentoo.org>
Closes zfsonlinux/zfs#4388
Closes#534
Perf profiling of dd on a zvol revealed that my system spent 3.16% of
its time in random_get_pseudo_bytes(). No SPL consumers need
cryptographic strength entropy, so we can reduce our overhead by
changing the implementation to utilize a fast PRNG.
The Linux kernel did not export a suitable PRNG function until it
exported get_random_int() in Linux 3.10. While we could implement an
autotools check so that we use it when it is available or even try to
access the symbol on older kernels where it is not exported using the
fact that it is exported on newer ones as justification, we can instead
implement our own pseudo-random data generator. For this purpose, I have
written one based on a 128-bit pseudo-random number generator proposed
in a paper by Sebastiano Vigna that itself was based on work by the late
George Marsaglia.
http://vigna.di.unimi.it/ftp/papers/xorshiftplus.pdf
Profiling the same benchmark with an earlier variant of this patch that
used a slightly different generator (roughly same number of
instructions) by the same author showed that time spent in
random_get_pseudo_bytes() dropped to 0.06%. That is a factor of 50
improvement. This particular generator algorithm is also well known to
be fast:
http://xorshift.di.unimi.it/#speed
The benchmark numbers there state that it runs at 1.12ns/64-bits or 7.14
GBps of throughput on an Intel Core i7-4770 in what is presumably a
single-threaded context. Using it in `random_get_pseudo_bytes()` in the
manner I have will probably not reach that level of performance, but it
should be fairly high and many times higher than the Linux
`get_random_bytes()` function that we use now, which runs at 16.3 MB/s
on my Intel Xeon E3-1276v3 processor when measured by using dd on
/dev/urandom.
Also, putting this generator's seed into per-CPU variables allows us to
eliminate overhead from both spin locks and CPU memory barriers, which
is NUMA friendly.
We could have alternatively modified consumers to use something like
`gethrtime() % 3` as suggested by both Matthew Ahrens and Tim Chase, but
that has a few potential problems that this approach avoids:
1. Switching to `gethrtime() % 3` in hot code paths today requires
diverging from illumos-gate and does nothing about potential future
patches from illumos-gate that call our slow `random_get_pseudo_bytes()`
in different hot code paths. Reimplementing `random_get_pseudo_bytes()`
with a per-CPU PRNG avoids both of those things entirely, which means
less work for us in the future.
2. Looking at the code that implements `gethrtime()`, I think it is
unlikely to be faster than this per-CPU PRNG implementation of
`random_get_pseudo_bytes()`. It would be best to go with something fast
now so that there is no point in revisiting this from a performance
perspective.
3. `gethrtime() % 3` can vary in behavior from system to system based on
kernel version, architecture and clock source. In comparison, this
per-CPU PRNG is about ~40 lines of code in `random_get_pseudo_bytes()`
that should behave consistently across all systems regardless of kernel
version, system architecture or machine clock source. It is unlikely
that we would ever need to revisit this per-CPU PRNG while the same
cannot be said for `gethrtime() % 3`.
4. `gethrtime()` uses CPU memory barriers and maybe atomic instructions
depending on the clock source, so replacing `random_get_pseudo_bytes()`
with `gethrtime()` in hot code paths could still require a future person
working on NUMA scalability to reimplement it anyway while this per-CPU
PRNG would not by virtue of using neither CPU memory barriers nor atomic
instructions. Note that I did not check various clock sources for the
presence of atomic instructions. There is simply too much code to read
and given the drawbacks versus this per-cpu PRNG, there is no point in
being certain.
5. I have heard of instances where poor quality pseudo-random numbers
caused problems for HPC code in ways that took more than a year to
identify and were remedied by switching to a higher quality source of
pseudo-random numbers. While filesystems are different than HPC code, I
do not think it is impossible for us to have instances where poor
quality pseudo-random numbers can cause problems. Opting for a well
studied PRNG algorithm that passes tests for statistical randomness over
changing callers to use `gethrtime() % 3` bypasses the need to think
about both whether poor quality pseudo-random numbers can cause problems
and the statistical quality of numbers from `gethrtime() % 3`.
6. `gethrtime()` calls `getrawmonotonic()`, which uses seqlocks. This is
probably not a huge issue, but anyone using kgdb would never be able to
step through a seqlock critical section, which is not a problem either
now or with the per-CPU PRNG:
https://en.wikipedia.org/wiki/Seqlock
The only downside that I can see is that this code's memory requirement
is O(N) where N is NR_CPUS, versus the current code and `gethrtime() %
3`, which are O(1), but that should not be a problem. The seeds will use
64KB of memory at the high end (i.e `NR_CPU == 4096`) and 16 bytes of
memory at the low end (i.e. `NR_CPU == 1`). In either case, we should
only use a few hundred bytes of code for text, especially since
`spl_rand_jump()` should be inlined into `spl_random_init()`, which
should be removed during early boot as part of "Freeing unused kernel
memory". In either case, the memory requirements are minuscule.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tim Chase <tim@chase2k.com>
Closes#372
This patch add a module parameter spl_taskq_kick. When writing non-zero value
to it, it will scan all the taskq, if a taskq contains a task pending for more
than 5 seconds, it will be forced to spawn a new thread. This is use as an
emergency recovery from deadlock, not a general solution.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#529
Previous commit be29e6a updated kobj_read_file() so it no longer
unconditionally passes RLIM64_INFINITY. The vn_rdwr() function
needs to be updated accordingly.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #513
I noticed that the SPL implementation of kobj_read_file is not correct
after comparing it with the userland implementation of kobj_read_file()
in zfsonlinux/zfs#4104.
Note that we no longer pass RLIM64_INFINITY with this, but our vn_rdwr
implementation did not support it anyway, so there is no difference.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#513
To prevent taskq_member holding tq_lock and doing linear search, thus causing
contention. We store the taskq pointer to which the thread belongs in tsd.
This way taskq_member will not need to touch tq_lock, and tsd has per slot
spinlock. So the contention should be reduced greatly.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#500Closes#504Closes#505
If a thread is holding mutex when doing cv_destroy, it might end up waiting a
thread in cv_wait. The waiter would wake up trying to aquire the same mutex
and cause deadlock.
We solve this by move the mutex_enter to the bottom of cv_wait, so that
the waiter will release the cv first, allowing cv_destroy to succeed and have
a chance to free the mutex.
This would create race condition on the cv_mutex. We use xchg to set and check
it to ensure we won't be harmed by the race. This would result in the cv_mutex
debugging becomes best-effort.
Also, the change reveals a race, which was unlikely before, where we call
mutex_destroy while test threads are still holding the mutex. We use
kthread_stop to make sure the threads are exit before mutex_destroy.
Signed-off-by: Chunwei Chen <tuxoko@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tim Chase <tim@chase2k.com>
Issue zfsonlinux/zfs#4166
Issue zfsonlinux/zfs#4106
For earlier versions of the kernel with memalloc_noio_save, it only turns
off __GFP_IO but leaves __GFP_FS untouched during direct reclaim. This
would cause threads to direct reclaim into ZFS and cause deadlock.
Instead, we should stick to using spl_fstrans_mark. Since we would
explicitly turn off both __GFP_IO and __GFP_FS before allocation, it
will work on every version of the kernel.
This impacts kernel versions 3.9-3.17, see upstream kernel commit
torvalds/linux@934f307 for reference.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tim Chase <tim@chase2k.com>
Closes#515
Issue zfsonlinux/zfs#4111
This patch provides 2 new kstats to display task queues:
/proc/spl/taskqs-all - Display all task queues
/proc/spl/taskqs - Display only "active" task queues
A task queue is considered to be "active" if it currently has active
(running) threads or if any of its pending, priority, delay or waitq
lists are not empty.
If the task queue has running threads, displays each thread function's
address (symbolically, if possibly) and its argument.
If the task queue has a non-empty list of pending, priority or delayed
task queue entries (taskq_ent_t), displays each entry's thread function
address and arguemnt.
If the task queue has any waiters, displays each waiting task's pid.
Note: This patch also updates some comments in taskq.h which referred to
"taskq_t" when they should have referred to "taskq_ent_t".
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#491
This patch only addresses the issues identified by the style checker.
It contains no functional changes.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The flags argument in spin_lock_irqsave is modified out side of spin_lock
context. We cannot use a shared variable like tq->tq_lock_flags for them. This
patch removes it and uses local variable for the flags.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#506
When taskq_dispatch() calls taskq_thread_spawn() to create a new thread
for a taskq, linux lockdep warns of possible recursive locking. This is
a false positive.
One such call chain is as follows, when a taskq needs more threads:
taskq_dispatch->taskq_thread_spawn->taskq_dispatch
The initial taskq_dispatch() holds tq_lock on the taskq that needed more
worker threads. The later call into taskq_dispatch() takes
dynamic_taskq->tq_lock. Without subclassing, lockdep believes these
could potentially be the same lock and complains. A similar case occurs
when taskq_dispatch() then calls task_alloc().
This patch uses spin_lock_irqsave_nested() when taking tq_lock, with one
of two new lock subclasses:
subclass taskq
TQ_LOCK_DYNAMIC dynamic_taskq
TQ_LOCK_GENERAL any other
Signed-off-by: Olaf Faaland <faaland1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #480
This reverts commit a430c11f0b. Using
journal_info like this can cause a BUG at kernel fs/jbd2/transaction.c:425!
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #500
The ->journal_info pointer in the task_struct is reserved for use by
filesystems and because the kernel can have multiple file systems on the
same stack due to direct reclaim, each filesystem that touches
->journal_info in a callback function will save the value at the start
of its frame and restore it at the end of its frame. This allows us to
safely use ->journal_info to store a pointer to the taskq's struct in
taskq threads so that ZFS code paths can detect the presence of a taskq.
This could break if the ZFS code were to use taskq_member from the
context of direct reclaim. However, there are no such uses of it in that
manner, so this is safe.
This eliminates an O(N) list traversal under a spinlock with an O(1)
unlocked pointer comparison.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: tuxoko <tuxoko@gmail.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
Closes#500
If a vnode is released asynchronously through areleasef(), it is
possible for the user process to reuse the file descriptor before
areleasef is called. When this happens, getf() will return a stale
reference, any operations in the kernel on that file descriptor will
fail (as it is closed) and the operations meant for that fd will
never occur from userspace's perspective.
We correct this by detecting this condition in getf(), doing a putf
on the old file handle, updating the file descriptor and proceeding
as if everything was fine. When the areleasef() is done, it will
harmlessly decrement the reference counter on the Illumos file handle.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#492
This is needed for architectures that do not have a builtin prefetchw()
Signed-off-by: Dimitri John Ledkov <xnox@ubuntu.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#502
Currently taskq_dispatch() will spawn new task with a condition that the caller
is also a member of the taskq. However, under this condition, it will still
cause deadlock where a task on tq1 is waiting another thread, who is trying to
dispatch a task on tq1. So this patch removes the check.
For example when you do:
zfs send pp/fs0@001 | zfs recv pp/fs0_copy
This will easily deadlock before this patch.
Also, move the seq_task check from taskq_thread_spawn() to taskq_thread()
because it's not used by the caller from taskq_dispatch().
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#496
Linux slab will automatically free empty slab when number of partial slab is
over min_partial, so we don't need to explicitly shrink it. In fact, calling
kmem_cache_shrink from shrinker will cause heavy contention on
kmem_cache_node->list_lock, to the point that it might cause __slab_free to
livelock (see zfsonlinux/zfs#3936)
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes zfsonlinux/zfs#3936
Closes#487
Allocate a kmem cache magazine for every possible CPU which might
be added to the system. This ensures that when one of these CPUs
is enabled it can be safely used immediately.
For many systems the number of online CPUs is identical to the
number of present CPUs so this does imply an increased memory
footprint. In fact, dynamically allocating the array of magazine
pointers instead of using the worst case NR_CPUS can end up
decreasing our memory footprint.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ned Bass <bass6@llnl.gov>
Closes#482
Support grsecurity/PaX kernel configurations where
CONFIG_PAX_USERCOPY_SLABS are enabled. When this kernel option
is enabled slabs which are used to copy between user and kernel
space must be created with SLAB_USERCOPY.
Stock Linux kernels do not have a SLAB_USERCOPY definition so
this causes no change in behavior for non-PAX-enabled kernels.
Verified-by: Wuffleton <null@wuffleton.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2977
Issue #3796
Illumos does not have direct reclaim and code run inside taskq worker
threads is not designed to deal with it. Allowing direct reclaim inside
a worker thread can therefore deadlock. We set PF_MEMALLOC_NOIO through
memalloc_noio_save() to indicate to the kernel's reclaim code that we
are inside a context where memory allocations cannot be allowed to block
on filesystem activity.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue zfsonlinux/zfs#1274
Issue zfsonlinux/zfs#2390
Closes#474
When dynamic taskq is enabled and all threads for a taskq are occupied,
a recursive dispatch can cause a deadlock if calling thread depends on
the recursively-dispatched thread for its return condition.
This patch attempts to create a new thread for recursive dispatch when
none are available.
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#472
This reverts commit 076821e due to a locking issue uncovered in
subsequent testing. An ASSERT is hit due to tq->tq_nspawn being
updated outside the lock. The patch will need to be reworked.
VERIFY3(0 == tq->tq_nspawn) failed (0 == -1)
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #472
When dynamic taskq is enabled and all threads for a taskq are occupied,
a recursive dispatch can cause a deadlock if calling thread depends on
the recursively-dispatched thread for its return condition.
This patch attempts to create a new thread for recursive dispatch when
none are available.
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#472
On Linux the meaning of a processes priority is inverted with respect
to illumos. High values on Linux indicate a _low_ priority while high
value on illumos indicate a _high_ priority.
In order to preserve the logical meaning of the minclsyspri and
maxclsyspri macros when they are used by the illumos wrapper functions
their values have been inverted. This way when changes are merged
from upstream illumos we won't need to remember to invert the macro.
It could also lead to confusion.
Note this change also reverts some of the priorities changes in prior
commit 62aa81a. The rational is as follows:
spl_kmem_cache - High priority may result in blocked memory allocs
spl_system_taskq - May perform I/O for file backed VDEVs
spl_dynamic_taskq - New taskq threads should be spawned promptly
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ned Bass <bass6@llnl.gov>
Issue zfsonlinux/zfs#3607
As described in spl_kmem_cache_destroy() the ->skc_ref count was
added to address the case of a cache reap or grow racing with a
destroy. They are not strictly needed in the alloc/free paths
because consumers of the cache are responsible for not using it
while it's being destroyed.
Removing this code is desirable because there is some evidence that
contention on this atomic negative impacts performance on large-scale
NUMA systems.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tim Chase <tim@chase2k.com>
Issue #463
Add a new defclsyspri macro which can be used to request the default
Linux scheduler priority. Neither the minclsyspri or maxclsyspri map
to the default Linux kernel thread priority. This makes it awkward to
create taskqs which run with the same priority as the rest of the kernel
threads on the system which can lead to performance issues.
All SPL callers which previously used minclsyspri or maxclsyspri have
been changed to use defclsyspri. The vast majority of callers were
part of the test suite which won't have an external impact. The few
places where it could impact performance the change was from maxclsyspri
to defclsyspri. This makes it more likely the process will be scheduled
which may help performance.
To facilitate further performance analysis the spl_taskq_thread_priority
module option has been added. When disabled (0) all newly created kernel
threads will use the default kernel thread priority. When enabled (1)
the specified taskq priority will be used. By default this value is
enabled (1).
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The default kmem debugging (--enable-debug-kmem) can severely impact
performance on large-scale NUMA systems due to the atomic operations
used in the memory accounting. A 32-thread fio test running on a
40-core 80-thread system and performing 100% cached reads with kmem
debugging is:
Enabled:
READ: io=177071MB, aggrb=2951.2MB/s, minb=2951.2MB/s, maxb=2951.2MB/s,
Disabled:
READ: io=271454MB, aggrb=4524.4MB/s, minb=4524.4MB/s, maxb=4524.4MB/s,
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tim Chase <tim@chase2k.com>
Issues #463
Build products from an out of tree build should be written
relative to the build directory. Sources should be referred
to by their locations in the source directory.
This is accomplished by adding the 'src' and 'obj' variables
for the module Makefile.am, using relative paths to reference
source files, and by setting VPATH when source files are not
co-located with the Makefile. This enables the following:
$ mkdir build
$ cd build
$ ../configure
$ make -s
This change also has the advantage of resolving the following
warning which is generated by modern versions of automake.
Makefile.am:00: warning: source file 'xxx' is in a subdirectory,
Makefile.am:00: but option 'subdir-objects' is disabled
Signed-off-by: Turbo Fredriksson <turbo@bayour.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue zfsonlinux/zfs#1082
Add the TASKQ_DYNAMIC flag to the kmem_cache and system taskqs
to reduce the number of idle threads on the system. Additional
threads will be created on demand up to the previous maximum
thread counts. This should have minimal, if any, impact on
performance.
This makes the system taskq consistent with illumos which is
always created as a dynamic taskq with up to 64 threads.
The task limits for the kmem_cache have been increased to avoid
any unnessisary throttling and to keep a larger reserve of
task_t structures on the free list.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tim Chase <tim@chase2k.com>
Closes#458
Setting the TASKQ_DYNAMIC flag will create a taskq with dynamic
semantics. Initially only a single worker thread will be created
to service tasks dispatched to the queue. As additional threads
are needed they will be dynamically spawned up to the max number
specified by 'nthreads'. When the threads are no longer needed,
because the taskq is empty, they will automatically terminate.
Due to the low cost of creating and destroying threads under Linux
by default new threads and spawned and terminated aggressively.
There are two modules options which can be tuned to adjust this
behavior if needed.
* spl_taskq_thread_sequential - The number of sequential tasks,
without interruption, which needed to be handled by a worker
thread before a new worker thread is spawned. Default 4.
* spl_taskq_thread_dynamic - Provides the ability to completely
disable the use of dynamic taskqs on the system. This is provided
for the purposes of debugging and troubleshooting. Default 1
(enabled).
This behavior is fundamentally consistent with the dynamic taskq
implementation found in both illumos and FreeBSD.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tim Chase <tim@chase2k.com>
Closes#458
Commit f752b46e added the cv_wait_interruptible() function to allow
condition variables to be woken by signals. This function and its
timed wait counterpart should have been named cv_wait_sig() to match
the illumos interface which provides the same functionality.
This patch renames the symbol but leaves a #define compatibility
wrapper in place until the ZFS code can be moved to the correct
name.
This patch also makes a small number of cosmetic changes to make
the condvar source and header cstyle clean.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#456
Under Illumos taskq_wait() returns when there are no more tasks
in the queue. This behavior differs from ZoL and FreeBSD where
taskq_wait() returns when all the tasks in the queue at the
beginning of the taskq_wait() call are complete. New tasks
added whilst taskq_wait() is running will be ignored.
This difference in semantics makes it possible that new subtle
issues could be introduced when porting changes from Illumos.
To avoid that possibility the taskq_wait() function is being
updated such that it blocks until the queue in empty.
The previous behavior remains available through the
taskq_wait_outstanding() interface. Note that this function
was previously called taskq_wait_all() but has been renamed
to avoid confusion.
Signed-off-by: Chris Dunlop <chris@onthe.net.au>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#455
This patch only addresses the issues identified by the style checker
in spl-tsd.c. It contains no functional changes.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
To prevent leaking tsd entries, we make tsd_set(key, NULL) remove the tsd
entry for the current thread. This is alright since tsd_get() returns NULL
when the entry doesn't exist.
Signed-off-by: Chunwei Chen <tuxoko@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#443
C type coercion rules require that negative numbers be converted into
positive numbers via wraparound such that a negative -1 becomes a
positive 1. This causes vn_getf to return a file handle when it should
return NULL whenever a positive file descriptor existed with the same
value. We should check for a negative file descriptor and return NULL
instead.
This was caught by ClusterHQ's unit testing.
Reference:
http://stackoverflow.com/questions/50605/signed-to-unsigned-conversion-in-c-is-it-always-safe
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Andriy Gapon <avg@FreeBSD.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#450
When layered on XFS the following warning will be emitted under CentOS7
when entering vfs_fsync() with PF_FSTRANS already set. This is not an
issue for other stock Linux file systems and the warning was removed
for newer kernels. However, to avoid triggering this error PF_FSTRANS
is cleared and then reset in vn_fsync().
WARNING: at fs/xfs/xfs_aops.c:968 xfs_vm_writepage+0x5ab/0x5c0
Call Trace:
[<ffffffff8105dee1>] warn_slowpath_common+0x61/0x80
[<ffffffffa01706fb>] xfs_vm_writepage+0x5ab/0x5c0 [xfs]
[<ffffffff8114b833>] __writepage+0x13/0x50
[<ffffffff8114c341>] write_cache_pages+0x251/0x4d0
[<ffffffff8114c60d>] generic_writepages+0x4d/0x80
[<ffffffffa016fc93>] xfs_vm_writepages+0x43/0x50 [xfs]
[<ffffffff8114d68e>] do_writepages+0x1e/0x40
[<ffffffff81142bd5>] __filemap_fdatawrite_range+0x65/0x80
[<ffffffff81142cea>] filemap_write_and_wait_range+0x2a/0x70
[<ffffffffa017a5b6>] xfs_file_fsync+0x66/0x1f0 [xfs]
[<ffffffff811df54b>] vfs_fsync+0x2b/0x40
[<ffffffffa03a88bd>] vn_fsync+0x2d/0x90 [spl]
[<ffffffffa0520c33>] spa_config_sync+0x503/0x680 [zfs]
[<ffffffffa0520ee4>] spa_config_update+0x134/0x170 [zfs]
[<ffffffffa0520eba>] spa_config_update+0x10a/0x170 [zfs]
[<ffffffffa051c54f>] spa_import+0x5bf/0x7b0 [zfs]
[<ffffffffa055c754>] zfs_ioc_pool_import+0x104/0x150 [zfs]
[<ffffffffa056294f>] zfsdev_ioctl+0x4cf/0x5c0 [zfs]
[<ffffffffa0562480>] ? pool_status_check+0xf0/0xf0 [zfs]
[<ffffffff811c2c85>] do_vfs_ioctl+0x2e5/0x4c0
[<ffffffff811c2f01>] SyS_ioctl+0xa1/0xc0
[<ffffffff815f3219>] system_call_fastpath+0x16/0x1b
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Avoid deadlocks when entering the shrinker from a PF_FSTRANS context.
This patch also reverts commit d0d5dd7 which added MUTEX_FSTRANS. Its
use has been deprecated within ZFS as it was an ineffective mechanism
to eliminate deadlocks. Among other things, it introduced the need for
strict ordering of mutex locking and unlocking in order that the
PF_FSTRANS flag wouldn't set incorrectly.
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#446
In the original implementation of the SPL wrappers were provided
for module initialization and cleanup. This was done to abstract
away any compatibility code which might be needed for the SPL.
As it turned out the only significant compatibility issue was that
the default pwd during module load differed under Illumos and Linux.
Since this is such as minor thing and the wrappers complicate the
code they are being retired.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue zfsonlinux/zfs#2985
Currently, spl_hostid module parameter doesn't do anything, because it will
always be overwritten when calling into hostid_read().
Instead, we should only call into hostid_read() when spl_hostid is not zero,
just as the comment describes.
Signed-off-by: Chunwei Chen <tuxoko@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#427
For performance reasons the reworked kmem code maps vmem_alloc() to
kmalloc_node() for allocations less than spa_kmem_alloc_max. This
allows for more concurrency in the system and less contention of
the virtual address space. Generally, this is a good thing.
However, in the case when the kmalloc_node() fails it makes little
sense to retry it using kmalloc_node() again. It will likely fail
in exactly the same way. A smarter strategy is to abandon this
optimization and retry using spl_vmalloc() which is very likely
to succeed.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ned Bass <bass6@llnl.gov>
Closes#428
The kmem_vasprintf(), kmem_vsprintf(), kobj_open_file(), and vn_openat()
functions should all use the kmem_flags_convert() function to generate
the GFP_* flags. This ensures that they can be safely called in any
context and the correct flags will be used.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#426
The __get_free_pages() function must be used in place of kmalloc()
to ensure the __GFP_COMP is strictly honored. This is due to
kmalloc() being layered on the generic Linux slab caches. It
wasn't until recently that all caches were created using __GFP_COMP.
This means that it is possible for a kmalloc() which passed the
__GFP_COMP flag to be returned a non-compound allocation.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The kmem cache implementation always adds new slabs by dispatching a
task to the spl_kmem_cache taskq to perform the allocation. This is
done because large slabs must be allocated using vmalloc(). It is
possible these allocations will block on IO because the GFP_NOIO flag
is not honored. This can result in a deadlock.
Therefore, a deadlock detection strategy was implemented to deal with
this case. When it is determined, by timeout, that the spl_kmem_cache
thread has deadlocked attempting to add a new slab. Then all callers
attempting to allocate from the cache fall back to using kmalloc()
which does honor all passed flags.
This logic was correct but an optimization in the code allowed for a
deadlock. Because only slabs backed by vmalloc() can deadlock in the
way described above. An optimization was made to only invoke this
deadlock detection code for vmalloc() backed caches. This had the
advantage of making it easy to distinguish these objects when they
were freed.
But this isn't strictly safe. If all the spl_kmem_cache threads end
up deadlocked than we can't grow any of the other caches either. This
can once again result in a deadlock if memory needs to be allocated
from one of these other caches to ensure forward progress.
The fix here is to remove the optimization which limits this fall back
allocation stratagy to vmalloc() backed caches. Doing this means we
may need to take the cache lock in spl_kmem_cache_free() call path.
But this small cost can be mitigated by ignoring objects with virtual
addresses.
For good measure the default number of spl_kmem_cache threads has been
increased from 1 to 4, and made tunable. This alone wouldn't resolve
the original issue since it's still possible for all the threads to be
deadlocked. However, it does help responsiveness by ensuring that a
single deadlocked spl_kmem_cache thread doesn't block allocations from
other caches until the timeout is reached.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
This change is designed to improve the memory utilization of
slabs by more carefully setting their size. The way the code
currently works is problematic for slabs which contain large
objects (>1MB). This is due to slabs being unconditionally
rounded up to a power of two which may result in unused space
at the end of the slab.
The reason the existing code rounds up every slab is because it
assumes it will backed by the buddy allocator. Since the buddy
allocator can only performs power of two allocations this is
desirable because it avoids wasting any space. However, this
logic breaks down if slab is backed by vmalloc() which operates
at a page level granularity. In this case, the optimal thing to
do is calculate the minimum required slab size given certain
constraints (object size, alignment, objects/slab, etc).
Therefore, this patch reworks the spl_slab_size() function so
that it sizes KMC_KMEM slabs differently than KMC_VMEM slabs.
KMC_KMEM slabs are rounded up to the nearest power of two, and
KMC_VMEM slabs are allowed to be the minimum required size.
This change also reduces the default number of objects per slab.
This reduces how much memory a single cache object can pin, which
can result in significant memory saving for highly fragmented
caches. But depending on the workload it may result in slabs
being allocated and freed more frequently. In practice, this
has been shown to be a better default for most workloads.
Also the maximum slab size has been reduced to 4MB on 32-bit
systems. Due to the limited virtual address space it's critical
the we be as frugal as possible. A limit of 4M still lets us
reasonably comfortably allocate a limited number of 1MB objects.
Finally, the kmem:slab_small and kmem:slab_large SPLAT tests
were extended to provide better test coverage of various object
sizes and alignments. Caches are created with random parameters
and their basic functionality is verified by allocating several
slabs worth of objects.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reduce the threshold for detecting a kmem cache deadlock by 10x
from HZ to HZ/10. The reduced value is still several orders of
magnitude large enough to avoid being triggered incorrectly. By
reducing it we allow the system to resolve the issue more quickly.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Many people have noticed that the kmem cache implementation is slow
to release its memory. This patch makes the reclaim behavior more
aggressive by immediately freeing a slab once it is empty. Unused
objects which are cached in the magazines will still prevent a slab
from being freed.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The comment above the Linux 3.16 kernel's clear_bit() states:
/**
* clear_bit - Clears a bit in memory
* @nr: Bit to clear
* @addr: Address to start counting from
*
* clear_bit() is atomic and may not be reordered. However, it does
* not contain a memory barrier, so if it is used for locking purposes,
* you should call smp_mb__before_atomic() and/or smp_mb__after_atomic()
* in order to ensure changes are visible on other processors.
*/
This comment does not make sense in the context of x86 because x86 maps the
operations to barrier(), which is a compiler barrier. However, it does make
sense to me when I consider architectures that reorder around atomic
instructions. In such situations, a processor is allowed to execute the
wake_up_bit() before clear_bit() and we have a race. There are a few
architectures that suffer from this issue.
In such situations, the other processor would wake-up, see the bit is still
taken and go to sleep, while the one responsible for waking it up will
assume that it did its job and continue.
This patch implements a wrapper that maps smp_mb__{before,after}_atomic() to
smp_mb__{before,after}_clear_bit() on older kernels and changes our code to
leverage it in a manner consistent with the mainline kernel.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
