Brian Behlendorf [Thu, 15 Jan 2015 23:11:45 +0000 (15:11 -0800)]
Use __get_free_pages() for emergency objects
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>
Brian Behlendorf [Mon, 15 Dec 2014 22:06:18 +0000 (14:06 -0800)]
Refine slab cache sizing
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>
Brian Behlendorf [Tue, 16 Dec 2014 00:02:48 +0000 (16:02 -0800)]
Reduce kmem cache deadlock threshold
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>
The spl-module-parameters(5) was not kept up to date. Refresh
the man page so that it lists all the possible module options,
describes what the do, and justify why the default values are
set they way the are.
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>
Richard Yao [Thu, 4 Dec 2014 23:47:51 +0000 (18:47 -0500)]
Enforce architecture-specific barriers around clear_bit()
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>
Richard Yao [Sun, 13 Jul 2014 18:45:20 +0000 (14:45 -0400)]
Add hooks for disabling direct reclaim
The port of XFS to Linux introduced a thread-specific PF_FSTRANS bit
that is used to mark contexts which are processing transactions. When
set, allocations in this context can dip into kernel memory reserves
to avoid deadlocks during writeback. Linux 3.9 provided the additional
PF_MEMALLOC_NOIO for disabling __GFP_IO in page allocations, which XFS
began using in 3.15.
This patch implements hooks for marking transactions via PF_FSTRANS.
When an allocation is performed in the context of PF_FSTRANS, any
KM_SLEEP allocation is transparently converted to a GFP_NOIO allocation.
Additionally, when using a Linux 3.9 or newer kernel, it will set
PF_MEMALLOC_NOIO to prevent direct reclaim from entering pageout() on
on any KM_PUSHPAGE or KM_NOSLEEP allocation. This effectively allows
the spl_vmalloc() helper function to be used safely in a thread which
is responsible for IO.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
1. It replaces the preprocessor kmem flag to gfp flag mapping with
proper translation logic. This eliminates the potential for
surprises that were previously possible where kmem flags were
mapped to gfp flags.
2. It maps vmem_alloc() allocations to kmem_alloc() for allocations
sized less than or equal to the newly-added spl_kmem_alloc_max
parameter. This ensures that small allocations will not contend
on a single global lock, large allocations can still be handled,
and potentially limited virtual address space will not be squandered.
This behavior is entirely different than under Illumos due to
different memory management strategies employed by the respective
kernels. However, this functionally provides the semantics required.
3. The --disable-debug-kmem, --enable-debug-kmem (default), and
--enable-debug-kmem-tracking allocators have been unified in to
a single spl_kmem_alloc_impl() allocation function. This was
done to simplify the code and make it more maintainable.
4. Improve portability by exposing an implementation of the memory
allocations functions that can be safely used in the same way
they are used on Illumos. Specifically, callers may safely
use KM_SLEEP in contexts which perform filesystem IO. This
allows us to eliminate an entire class of Linux specific changes
which were previously required to avoid deadlocking the system.
This change will be largely transparent to existing callers but there
are a few caveats:
1. Because the headers were refactored and extraneous includes removed
callers may find they need to explicitly add additional #includes.
In particular, kmem_cache.h must now be explicitly includes to
access the SPL's kmem cache implementation. This behavior is
different from Illumos but it was done to avoid always masking
the Linux slab functions when kmem.h is included.
2. Callers, like Lustre, which made assumptions about the definitions
of KM_SLEEP, KM_NOSLEEP, and KM_PUSHPAGE will need to be updated.
Other callers such as ZFS which did not will not require changes.
3. KM_PUSHPAGE is no longer overloaded to imply GFP_NOIO. It retains
its original meaning of allowing allocations to access reserved
memory. KM_PUSHPAGE callers can be converted back to KM_SLEEP.
4. The KM_NODEBUG flags has been retired and the default warning
threshold increased to 32k.
5. The kmem_virt() functions has been removed. For callers which
need to distinguish between a physical and virtual address use
is_vmalloc_addr().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Address all cstyle issues in the kmem, vmem, and kmem_cache source
and headers. This will done to make it easier to review subsequent
changes which will rework the kmem/vmem implementation.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
This change introduces no functional changes to the memory management
interfaces. It only restructures the existing codes by separating the
kmem, vmem, and kmem cache implementations in the separate source and
header files.
Splitting this functionality in to separate files required the addition
of spl_vmem_{init,fini}() and spl_kmem_cache_{initi,fini}() functions.
Additionally, several minor changes to the #include's were required to
accommodate the removal of extraneous header from kmem.h.
But again, while large this patch introduces no functional changes.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Tim Chase [Sun, 11 Jan 2015 16:37:43 +0000 (10:37 -0600)]
Use current_kernel_time() in the time compatibility wrappers
Since the Linux kernel's utimens family of functions uses
current_kernel_time(), we need to do the same in the context of ZFS
or else there can be discrepencies in timestamps (they go backward)
if userland code does:
Brian Behlendorf [Thu, 20 Nov 2014 01:01:20 +0000 (20:01 -0500)]
Fix debug object on stack warning
When running the SPLAT tests on a kernel with CONFIG_DEBUG_OBJECTS=y
enabled the following warning is generated.
ODEBUG: object is on stack, but not annotated
WARNING: at lib/debugobjects.c:300 __debug_object_init+0x221/0x480()
This is caused by the test cases placing a debug object on the stack
rather than the heap. This isn't harmful since they are small objects
but to make CONFIG_DEBUG_OBJECTS=y happy the objects have been relocated
to the heap. This impacted taskq tests 1, 3, and 7.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #424
Chunwei Chen [Fri, 19 Dec 2014 03:31:59 +0000 (11:31 +0800)]
mutex: force serialization on mutex_exit() to fix races
It is known that mutexes in Linux are not safe when using them to
synchronize the freeing of object in which the mutex is embedded:
http://lwn.net/Articles/575477/
The known places in ZFS which are suspected to suffer from the race
condition are zio->io_lock and dbuf->db_mtx.
* zio uses zio->io_lock and zio->io_cv to synchronize freeing
between zio_wait() and zio_done().
* dbuf uses dbuf->db_mtx to protect reference counting.
This patch fixes this kind of race by forcing serialization on
mutex_exit() with a spin lock, making the mutex safe by sacrificing
a bit of performance and memory overhead.
This issue most commonly manifests itself as a deadlock in the zio
pipeline caused by a process spinning on the damaged mutex. Similar
deadlocks have been reported for the dbuf->db_mtx mutex. And it can
also cause a NULL dereference or bad paging request under the right
circumstances.
This issue any many like it are linked off the zfsonlinux/zfs#2523
issue. Specifically this fix resolves at least the following
outstanding issues:
Signed-off-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Richard Yao <ryao@gentoo.org>
Closes #421
Ned Bass [Fri, 14 Nov 2014 18:18:05 +0000 (10:18 -0800)]
Remove compat includes from sys/types.h
Don't include the compatibility code in linux/*_compat.h in the public
header sys/types.h. This causes problems when an external code base
includes the ZFS headers and has its own conflicting compatibility code.
Lustre, in particular, defined SHRINK_STOP for compatibility with
pre-3.12 kernels in a way that conflicted with the SPL's definition.
Because Lustre ZFS OSD includes ZFS headers it fails to build due to a
'"SHRINK_STOP" redefined' compiler warning. To avoid such conflicts
only include the compat headers from .c files or private headers.
Also, for consistency, include sys/*.h before linux/*.h then sort by
header name.
Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #411
When the SPL was originally written Linux tracepoints were still
in their infancy. Therefore, an entire debugging subsystem was
added to facilite tracing which served us well for many years.
Now that Linux tracepoints have matured they provide all the
functionality of the previous tracing subsystem. Rather than
maintain parallel functionality it makes sense to fully adopt
tracepoints. Therefore, this patch retires the legacy debugging
infrastructure.
See zfsonlinux/zfs@bc9f413 for the tracepoint changes.
Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #408
As long as we can fit a minimum of one object/slab there's no reason
to prevent the creation of the cache. This effectively pushes the
maximum object size up to 32MB. The splat cache tests were extended
accordingly to verify this functionality.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Marcel Wysocki [Fri, 31 Oct 2014 23:26:29 +0000 (16:26 -0700)]
Add config/compile to config/.gitignore
This file may be added by automake and therefore should be added
to config/.gitignore. For the full list of possible auxiliary
programs see the full automake documentation.
Richard Yao [Tue, 17 Jun 2014 23:37:02 +0000 (19:37 -0400)]
kmem_cache: Call constructor/destructor on each alloc/free
This has a few benefits. First, it fixes a regression that "Rework
generic memory allocation interfaces" appears to have triggered in
splat's slab_reap and slab_age tests. Second, it makes porting code from
Illumos to ZFSOnLinux easier. Third, it has the side effect of making
reclaim from slab caches that specify reclaim functions an order of
magnitude faster. The splat slab_reap test usually took 30 to 40
seconds. With this change, it takes 3 to 4.
Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #369
Tim Chase [Thu, 2 Oct 2014 12:40:05 +0000 (07:40 -0500)]
Linux 3.12 compat: shrinker semantics
The new shrinker API as of Linux 3.12 modifies "struct shrinker" by
replacing the @shrink callback with the pair of @count_objects and
@scan_objects. It also requires the return value of @count_objects to
return the number of objects actually freed whereas the previous @shrink
callback returned the number of remaining freeable objects.
This patch adds support for the new @scan_objects return value semantics
and updates the splat shrinker test case appropriately.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com>
Closes #403
Brian Behlendorf [Sun, 19 Oct 2014 21:42:10 +0000 (14:42 -0700)]
Merge branch 'cleanup'
Over the years the SPL code bases has accumulated compatibly code
to allow it to build against a wide range of Linux kernels. In
general this is desirable because it makes the code flexible.
However, once support for these old kernels is no longer needed
and is no longer being actively tested it should be removed. This
helps keep the code simple and understandable.
The spl-0.6.x releases have supported kernels all the way back to
2.6.26. This patch stack moves that cut off up to 2.6.32 and newer
kernels. This ensures we still support all the major enterprise
distributions which are largely locked in to 2.6.32 based kernels.
And at the same time we can shed a large amount of compatibility
code which simplifies maintenance and new development.
Signed-off-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #395
The vfs_fsync() function has been available since Linux 2.6.29.
There is no longer a need to maintain this compatibility code.
However, the HAVE_2ARGS_VFS_FSYNC check was left in place
since that change occured after 2.6.32.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
As of Linux 2.6.32 the proc handlers where updated to expect only
five arguments. Therefore there is no longer a need to maintain
this compatibility code and this infrastructure can be simplified.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The groups_search() function was never exported by a mainline kernel
therefore we drop this compatibility code and always provide our own
implementation.
Additionally, the cred_t structure has been available since 2.6.29
so there is no longer a need to maintain compatibility code.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
This function has never been exported by any mainline and was only
briefly available under RHEL5. Therefore this check is being removed
and the code update to always use the wrapper function.
The next step will be to eliminate all this code. If ZFS were updated
not to assume that it's pwd was / there would be no need for this.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
After the removable of get_vmalloc_info(), the unused global memory
variables, and the optional dcache/icache shrinkers there is no
longer a need for the kallsyms compatibility code. This allows
us to eliminate another brittle area of the code by removing the
kernel upcall this functionality depended on for older kernels.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
This is optional functionality which may or may not be useful to
ZFS when using older kernels. It is never a hard requirement.
Therefore this functionality is being removed from the SPL and
a simpler slimmed down version will be added to ZFS.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Platforms such as Illumos and FreeBSD have historically provided
global variables which summerize the memory state of a system.
Linux on the otherhand doesn't expose any of this information
to kernel modules and uses entirely different mechanisms for
memory management.
In order to simplify the original ZFS port to Linux these global
variables were emulated by the SPL for the benefit of ZFS. As ZoL
has matured over the years it has moved steadily away from these
interfaces and now no longer depends on them at all.
Therefore, this patch completely removes the global variables
availrmem, minfree, desfree, lotsfree, needfree, swapfs_minfree,
and swapfs_reserve. This greatly simplifies the memory management
code and eliminates a common area of confusion.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The get_vmalloc_info() function was used to back the vmem_size()
function. This was always problematic and resulted in brittle
code because the kernel never provided a clean interface for
modules.
However, it turns out that the only caller of this function in
ZFS uses it to determine the total virtual address space size.
This can be determined easily without get_vmalloc_info() so
vmem_size() has been updated to take this approach which allows
us to shed the get_vmalloc_info() dependency.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The inode structure has used i_mutex as its internal locking
primitive since 2.6.16. The compatibility code to check for
the previous semaphore primitive has been removed. However,
the wrapper function itself is being kept because it's entirely
possible this primitive will change again to allow finer grained
locking.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The atomic64_xchg() and atomic64_cmpxchg() functions have been
available since Linux 2.6.24. There is no longer a need to
maintain this compatibility code.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Tue, 30 Sep 2014 22:48:25 +0000 (18:48 -0400)]
Simplify the time compatibility wrappers
Many of the time functions had grown overly complex in order to
handle kernel compatibility issues. However, as of Linux 2.6.26
all the required functionality is available. This allows us to
retire numerous configure checks and greatly simplify the time
compatibility wrappers.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Tue, 30 Sep 2014 22:07:07 +0000 (18:07 -0400)]
Map highbit64() to fls64()
The fls64() function has been available since Linux 2.6.16 and
it should be used to implemented highbit64(). This allows us
to provide an optimized implementation and simplify the code.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Tue, 30 Sep 2014 22:15:22 +0000 (18:15 -0400)]
Remove CTL_UNNUMBERED sysctl interface
Support for the CTL_UNNUMBERED sysctl interface was removed in
Linux 2.6.19. There is no longer any reason to maintain this
compatibility code. There also issue any reason to keep around
the CTL_NAME macro and helpers so they have been retired.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Fri, 17 Oct 2014 22:10:19 +0000 (15:10 -0700)]
Remove utsname() wrapper
There is no longer a need to wrap this because utsname() is provided
by the kernel and can be called directly. This will require a small
change in the ZFS code because utsname is expected to be a global
structure and not a function.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Tue, 30 Sep 2014 18:51:53 +0000 (14:51 -0400)]
Remove sysctl_vfs_cache_pressure assumption
The generic SPL cache shrinkers make the assumption that the
caches only contain VFS cache data and therefore should be scaled
based on vfs_cache_pressure. This is not strictly true and it
should not be assumed.
Removing this tuning should not have any impact on the stock
behavior because vfs_cache_pressure=100 by default. This means
that no scaling will take place.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Mon, 29 Sep 2014 20:48:35 +0000 (16:48 -0400)]
Remove adaptive mutex implementation
Since the Linux 2.6.29 kernel all mutexes have been adaptive mutexs.
There is no longer any point in keeping this code so it is being
removed to simplify the code.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Mon, 29 Sep 2014 23:42:15 +0000 (19:42 -0400)]
Update code to use misc_register()/misc_deregister()
When the SPL was originally written it was designed to use the
device_create() and device_destroy() functions. Unfortunately,
these functions changed considerably over the years making them
difficult to rely on.
As it turns out a better choice would have been to use the
misc_register()/misc_deregister() functions. This interface
for registering character devices has remained stable, is simple,
and provides everything we need.
Therefore the code has been reworked to use this interface. The
higher level ZFS code has always depended on these same interfaces
so this is also as a step towards minimizing our kernel dependencies.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Tom Prince [Thu, 9 Oct 2014 17:08:00 +0000 (14:08 -0300)]
Install header during post-build rather than post-install.
New versions of dkms clean up the build directory after installing.
It appears that this was always intended, but had rm -rf "/path/to/build/*"
(note the quotes), which prevented it from working.
Also, the build step is already installing stuff into the directory where
these files go, so installing our stuff there as part of build rather than
install makes sense.
Signed-off-by: Tom Prince <tom.prince@clusterhq.com> Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #399
Brian Behlendorf [Tue, 30 Sep 2014 21:47:10 +0000 (17:47 -0400)]
Fix bug in SPLAT taskq:front
While running SPLAT on a kernel with CONFIG_DEBUG_ATOMIC_SLEEP
enabled the taskq:front was flagged as a test which might sleep
which in an unsafe context. Specifically, the splat_vprint()
function which internally takes a mutex was being called under
a spin lock. Moving the log function outside the spin lock
cleanly solves this issue.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The smp_mb__{before,after}_clear_bit functions have been renamed
smp_mb__{before,after}_atomic. Rather than adding a compatibility
function to handle this the code has been updated to use smp_wmb().
This has the advantage of being a stable functionally equivalent
interface. On many architectures smp_mb__after_clear_bit() expands
to smp_wmb(). Others might be able to do something slightly more
efficient but this will be safe and correct on all of them.
Signed-off-by: Turbo Fredriksson <turbo@bayour.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #386
Richard Yao [Sun, 29 Jun 2014 05:00:02 +0000 (01:00 -0400)]
Cleanup vn_rename() and vn_remove()
zfsonlinux/spl#bcb15891ab394e11615eee08bba1fd85ac32e158 implemented
Linux 3.6+ support by adding duplicate vn_rename and vn_remove
functions. The new ones were cleaner, but the duplicate functions made
the codebase less maintainable. This adds some compatibility shims that
allow us to retire the older vn_rename and vn_remove in favor of the new
ones on old kernels. The result is a net 143 line reduction in lines of
code and a cleaner codebase.
Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #370
Ned Bass [Sat, 9 Aug 2014 00:41:22 +0000 (17:41 -0700)]
Linux 3.17 compat: remove wait_on_bit action function
Linux kernel 3.17 removes the action function argument from
wait_on_bit(). Add autoconf test and compatibility macro to support
the new interface.
The former "wait_on_bit" interface required an 'action' function to
be provided which does the actual waiting. There were over 20 such
functions in the kernel, many of them identical, though most cases
can be satisfied by one of just two functions: one which uses
io_schedule() and one which just uses schedule(). This API change
was made to consolidate all of those redundant wait functions.
For small objects the Linux slab allocator should be used to make the most
efficient use of the memory. However, large objects are not supported by
the Linux slab and therefore the SPL implementation is preferred. A cutoff
of 16K was determined to be optimal for architectures using 4K pages.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: DHE <git@dehacked.net>
Issue #356
Closes #379
Reinstate the correct default behavior of returning the number of objects
in the cache for reclaim. This behavior was disabled in recent releases
to do occasional reports of spinning in shrink_slabs(). Those issues have
been resolved and can no longer can be reproduced. See commit 376dc35.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: DHE <git@dehacked.net>
Issue #358
Closes #379
Tim Chase [Sun, 27 Apr 2014 01:56:03 +0000 (20:56 -0500)]
Add functions and macros as used upstream.
Added highbit64() and howmany() which are used in recent upstream
code. Both highbit() and highbit64() should at some point be
re-factored to use the optimized fls() and fls64() functions.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Prakash Surya <surya1@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com>
Closes #363
Brian Behlendorf [Wed, 16 Jul 2014 21:00:57 +0000 (14:00 -0700)]
Rate limit debugging stack traces
There have been issues in the past where excessive debug logging
to the console has resulted in significant performance impacts.
In the vast majority of these cases only a few stack traces are
required to diagnose the issue. Therefore, stack traces dumped to
the console will now we limited to 5 every 60s.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Prakash Surya <surya1@llnl.gov>
Closes #374
Tim Chase [Wed, 18 Jun 2014 20:22:50 +0000 (15:22 -0500)]
Safer debugging and assertion macros.
Spl's debugging and assertion macros macro used the typical do/while(0)
form for if/else friendliness, however, this limits their use in contexts
where a do loop is not valid; such as within another multi-statement
style macro.
The following macros have been converted to not use do/while(0):
PANIC, ASSERT, ASSERTF, VERIFY, VERIFY3_IMPL
PANIC has been converted to a wrapper around the new spl_PANIC() function.
The other macros have been converted to use the "&&" operator for the
branch-predicition conditional and also to use spl_PANIC().
The __ASSERT() macro was not touched. It is only used by the debugging
infrastructure and that code, including this macro, will be retired when
the tracepoint patches are merged.
Signed-off-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #367
Brian Behlendorf [Thu, 29 May 2014 21:06:32 +0000 (14:06 -0700)]
Fix DKMS package upgrade and packager
Running 'yum upgrade spl-dkms' package could appear to work properly
and still leave you with no spl modules installed. This will occur
when only the spl release, and not the version, are incremented.
This may be the case for a fast moving spl-testing repository.
During the upgrade process DKMS will realize that spl-x.y.z is already
installed and remove it. DKMS then correctly builds the new modules
for spl-x.y.z. However, as a final step when the old spl-x.y.z-r is
removed the %preun script runs and removes the newly build modules.
To handle this case the %preun script has been updated to only run
when the installed version exactly matches the full spec file version.
This change also updated ChangeLog section based on the DKMS
reference spec file.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Thu, 29 May 2014 21:55:40 +0000 (14:55 -0700)]
Restrict release number to META version
When creating packages in a git repository the release number
can be automatically set by 'git describe'. This normally works
well but if your repository has newer tags which match the form
NAME-VERSION* the release may be incorrectly calculated. To
prevent this the match patten has been restricted to NAME-VERSION.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Fri, 16 May 2014 16:29:21 +0000 (09:29 -0700)]
Add spl_kmem_cache_reclaim module option
The correct behavior for all registered shrinkers is to return the
number of objects in their cache. In theory this allows the Linux
VM to balance memory reclaim across all registered caches.
In commit b9b3715 this behavior was disabled in favor of returning
-1 which notifies the VM that no additional objects are available
for reclaim. This was done as a workaround to resolve thrashing
in shrink_slabs() which could occur when memory was low and numerous
core where in reclaim. Unfortunately, this has been observed to
increase the likelihood of OOM events when SPL slab consumers are
responsible for consuming the majority of memory.
Therefore, this patch makes this behavior tunable. Setting the
spl_kmem_cache_reclaim module option to 0x1 will result in the
shrinker only being called once. This is the default behavior.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Prakash Surya <surya1@llnl.gov>
Closes #358
For small objects the Linux slab allocator has several advantages
over its counterpart in the SPL. These include:
1) It is more memory-efficient and packs objects more tightly.
2) It is continually tuned to maximize performance.
Therefore it makes sense to layer the SPLs slab allocator on top
of the Linux slab allocator. This allows us to leverage the
advantages above while preserving the Illumos semantics we depend
on. However, there are some things we need to be careful of:
1) The Linux slab allocator was never designed to work well with
large objects. Because the SPL slab must still handle this use
case a cut off limit was added to transition from Linux slab
backed objects to kmem or vmem backed slabs.
spl_kmem_cache_slab_limit - Objects less than or equal to this
size in bytes will be backed by the Linux slab. By default
this value is zero which disables the Linux slab functionality.
Reasonable values for this cut off limit are in the range of
4096-16386 bytes.
spl_kmem_cache_kmem_limit - Objects less than or equal to this
size in bytes will be backed by a kmem slab. Objects over this
size will be vmem backed instead. This value defaults to
1/8 a page, or 512 bytes on an x86_64 architecture.
2) Be aware that using the Linux slab may inadvertently introduce
new deadlocks. Care has been taken previously to ensure that
all allocations which occur in the write path use GFP_NOIO.
However, there may be internal allocations performed in the
Linux slab which do not honor these flags. If this is the case
a deadlock may occur.
The path forward is definitely to start relying on the Linux slab.
But for that to happen we need to start building confidence that
there aren't any unexpected surprises lurking for us. And ideally
need to move completely away from using the SPLs slab for large
memory allocations. This patch is a first step.
NOTES:
1) The KMC_NOMAGAZINE flag was leveraged to support the Linux slab
backed caches but it is not supported for kmem/vmem backed caches.
2) Regardless of the spl_kmem_cache_*_limit settings a cache may
be explicitly set to a given type by passed the KMC_KMEM,
KMC_VMEM, or KMC_SLAB flags during cache creation.
3) The constructors, destructors, and reclaim callbacks are all
functional and will be called regardless of the cache type.
4) KMC_SLAB caches will not appear in /proc/spl/kmem/slab due to
the issues involved in presenting correct object accounting.
Instead they will appear in /proc/slabinfo under the same names.
5) Several kmem SPLAT tests needed to be fixed because they relied
incorrectly on internal kmem slab accounting. With the updated
test cases all the SPLAT tests pass as expected.
6) An autoconf test was added to ensure that the __GFP_COMP flag
was correctly added to the default flags used when allocating
a slab. This is required to ensure all pages in higher order
slabs are properly refcounted, see ae16ed9.
7) When using the SLUB allocator there is no need to attempt to
set the __GFP_COMP flag. This has been the default behavior
for the SLUB since Linux 2.6.25.
8) When using the SLUB it may be desirable to set the slub_nomerge
kernel parameter to prevent caches from being merged.
Original-patch-by: DHE <git@dehacked.net> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Prakash Surya <surya1@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Signed-off-by: DHE <git@dehacked.net> Signed-off-by: Chunwei Chen <tuxoko@gmail.com>
Closes #356
Andrey Vesnovaty [Wed, 28 Aug 2013 02:09:25 +0000 (05:09 +0300)]
Evenly distribute the taskq threads across available CPUs
The problem is described in commit aeeb4e0c0ae75b99ebbaa3056f0afc8e12949532.
However, instead of disabling the binding to CPU altogether we just keep the
last CPU index across calls to taskq_create() and thus achieve even
distribution of the taskq threads across all available CPUs.
The implementation based on assumption that task queues initialization
performed in serial manner.
When using __get_free_pages to get high order memory, only the first page's
_count will set to 1, other's will be 0. When an internal page get passed into
rbd, it will eventully go into tcp_sendpage. There, it will be called with
get_page and put_page, and get freed erroneously when _count jump back to 0.
The solution to this problem is to use compound page. All pages in a
high order compound page share a single _count. So get_page and put_page in
tcp_sendpage will not cause _count jump to 0.
Signed-off-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #251
Using the ARM reference simulation (fast model foundation v8) I
cross compiled spl and zfs, to confirm it works on ARMv8 (64 bit
arm architecture, called aarch64 in Linux).
As it is based on previous ARM porting, the resulting patch is
disappointingly small, there was very little to do. The code fixes
the compile issues and has light testing done.
Signed-off-by: Jorgen Lundman <lundman@lundman.net> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #351
By default maximal number of objects in slab can't exceed (16*2 - 1) and slab
size can't exceed 32M.
Today's high end servers having couple hundreds of RAM available for ARC may
run into a trouble with virtual memory because of the restriction mentioned
above.
Problem:
Reasons for very high number of virtual memory allocations:
* Real slab size very small relative to the size of the entire RAM
* Slabs allocated on virtual memory and fill entire ARC
The result is very high number of allocated virtual memory ranges (hundreds of
ranges). When virtual memory subsystem manages high number of ranges its
performance become so poor that it freezes from time to time.
Solution:
Number of objects per slab should be increased taking into account maximal
slab size which can also be increased if needed.
Signed-off-by: Andrey Vesnovaty <andrey.vesnovaty@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #337
Chunwei Chen [Tue, 25 Feb 2014 09:16:55 +0000 (17:16 +0800)]
Add ddi_time_after and friends
When comparing times gotten from ddi_get_lbolt, we have to take account of
wrap around of jiffies. Therefore, we cannot use 't1 < t2'. Instead we should
use 't1 - t2 < 0'.
This patch add ddi_time_after and friends to address this issue. They have
strict type restriction, clock_t for vanilla and int64_t for 64 version, to
prevent type conversion from screwing things.
Signed-off-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #335
Richard Yao [Tue, 12 Mar 2013 01:16:36 +0000 (21:16 -0400)]
Simplify hostid logic
There is plenty of compatibility code for a hw_hostid
that isn't used by anything. At the same time, there are apparently
issues with the current hostid logic. coredumb in #zfsonlinux on
freenode reported that Fedora 17 changes its hostid on every boot, which
required force importing his pool. A suggestion by wca was to adopt
FreeBSD's behavior, where it treats hostid as zero if /etc/hostid does
not exist
Adopting FreeBSD's behavior permits us to eliminate plenty of code,
including a userland helper that invokes the system's hostid as a
fallback.
Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #224
Tim Chase [Fri, 11 Apr 2014 13:55:10 +0000 (08:55 -0500)]
Call kthread_create() correctly with fixed arguments.
The kernel's kthread_create() function is defined as "..." and there is
no va_list variant at the moment. The task name is pre-formatted into
a local buffer and passed to kthread_create() with fixed arguments.
Signed-off-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #347
Tim Chase [Wed, 26 Mar 2014 13:29:24 +0000 (08:29 -0500)]
Support post-3.13 kthread_create() semantics.
Provide spl_kthread_create() as a wrapper to the kernel's kthread_create()
to provide pre-3.13 semantics. Re-try if the call is interrupted or if it
would have returned -ENOMEM. Otherwise return NULL.
Signed-off-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #339
Due to certain assumptions made in the the cred:groupmember test it
could result in false positives when run on specific distributions.
This was solely a bug in the test case and not in the groupmember()
function which the test case was validating.
To prevent future false positives the test case has been rewritten
to be both more rigerous and to make fewer assumptions about the
system.
Minor style cleanup was done to cr_groups_search() and groupmember()
functions.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
By setting __GFP_NORETRY the kernel memory reclaim logic was allowed to
abort early and dump a falled allocation stack to the console. Since
this was done in a tight loop to fill memory it could result in a large
number of stacks being dumped to the console. This in turn slowed down
the test sufficiently so it exceeded the time limit and failed.
To resolve this issue the __GFP_NORETRY flag is being removed. This is
how it should have been called originally to ensure we're simulating
the behavior of most callers which will use the GFP_KERNEL flag.
In addition, the reclaim granularity of 1000 objects was far to coarse
for this to be a realistic test. For kmem:slab_reclaim there might
only be a few thousand objects total in the cache. Therefore, the
SPLAT_KMEM_OBJ_RECLAIM constant for these tests was lowered. This
will cause the reclaim callback to run more frequently which makes
for a better test case.
The frequency of the cache reaping in kmem:slab_reap was increased
to accommodate the reduced number of objects released during the
reclaim.
These changes only impact the test cases and were done to remove
false positives caused by the test case itself.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf [Fri, 17 Jan 2014 19:40:59 +0000 (11:40 -0800)]
Remove incorrect use of EXTRA_DIST for man pages
Setting the 'dist_' prefix is the correct way to instruct Automake
to include these files in the distribution. The EXTRA_DIST variable
is reserved for files which are not covered by the automatic rules.
When this code was written it appears to have been assumed that
every taskq would have a large number of threads. In this case
it would make sense to attempt to evenly bind the threads over
all available CPUs. However, it failed to consider that creating
taskqs with a small number of threads will cause the CPUs with
lower ids become over-subscribed.
For this reason the kthread_bind() call is being removed and
we're leaving the kernel to schedule these threads as it sees fit.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #325
Related to issue #257 which added Linux 3.10 compatibility. For
ARM and Sparc architectures we must explicitly include the
<linux/vmalloc.h> header to ensure the vmalloc_info structure
is always defined when available.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #257
Closes #291
Use the standard Linux MODULE_VERSION macro to expose the installed
spl and splat module versions. This will also automatically add a
checksum of the .c files and headers in "srcversion". See:
Richard Yao [Mon, 25 Nov 2013 16:22:33 +0000 (11:22 -0500)]
Linux 3.13 compat: Pass NULL for new delegated inode argument
This check was originally added for SLES10, a093c6a, to check for
a 'struct vfsmount *' argument which they added. However, since
SLES10 is based on a 2.6.16 kernel which is no longer supported
this functionality was dropped. The checks were refactored to
support Linux 3.13 without concern for historical versions.
Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #312
Richard Yao [Thu, 8 Aug 2013 08:30:55 +0000 (04:30 -0400)]
Linux 3.13 compat: Remove unused flags variable from __cv_init()
GCC 4.8.1 complained about an unused flags variable when building
against Linux 2.6.26.8:
/var/tmp/portage/sys-kernel/spl-9999/work/spl-9999/module/spl/../../module/spl/spl-condvar.c:
In function ‘__cv_init’:
/var/tmp/portage/sys-kernel/spl-9999/work/spl-9999/module/spl/../../module/spl/spl-condvar.c:39:6:
error: variable ‘flags’ set but not used
[-Werror=unused-but-set-variable]
int flags = KM_SLEEP;
^
cc1: all warnings being treated as errors
Additionally, the superfluous code uses a preempt_count variable that is
no longer available on Linux 3.13. Deleting the unnecessary code fixes a
Linux 3.13 compatibility issue.
Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #312
This is a first draft of a spl-module-parameters(5) man page. I have
just extracted the parameter name and its description with modinfo,
then checked the source what type it is and its default value.
This will need more work, preferably someone that actually know these
values and what to use them for. Similar to zfsonlinux/zfs#1856, but
for the spl.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes zfsonlinux/zfs#1856
Cyril Plisko [Thu, 14 Nov 2013 21:58:35 +0000 (13:58 -0800)]
Tighten spl dependency on spl-kmod
Make spl depend on the same version of spl-kmod, rather than on same or
better. When yum repository contains a number of versions the dependency
resolution breaks on trying to install non-latest version.
Signed-off-by: Cyril Plisko <cyril.plisko@mountall.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes zfsonlinux/zfs#1677
Richard Yao [Tue, 5 Nov 2013 16:35:54 +0000 (11:35 -0500)]
Linux 3.12 compat: New shrinker API
torvalds/linux@24f7c6 introduced a new shrinker API while
torvalds/linux@a0b021 dropped support for the old shrinker API.
This patch adds support for the new shrinker API by wrapping
the old one with the new one.
This change also reorganizes the autotools checks on the shrinker
API such that the configure script will fail early if an unknown
API is encountered in the future.
Support for the set_shrinker() API which was used by Linux 2.6.22
and older has been dropped. As a general rule compatibility is
only maintained back to Linux 2.6.26.
Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes zfsonlinux/zfs#1732
Closes zfsonlinux/zfs#1822
Closes #293
Closes #307
Ned Bass [Fri, 1 Nov 2013 20:37:58 +0000 (13:37 -0700)]
Emulate illumos interface cv_timedwait_hires()
Needed for Illumos #3582. This interface is supposed to support
a variable-resolution timeout with nanosecond granularity. This
implementation rounds up to microsecond resolution, as nanosecond-
precision timing is rarely needed for real-world performance
tuning and may incur unnecessary busy-waiting. usleep_range() is
used if available, otherwise udelay() or msleep() are used
depending on the length of the delay interval.
Add flags from sys/callo.h as these are used to control the behavior of
cv_timedwait_hires(). Specifically,
CALLOUT_FLAG_ABSOLUTE
Normally, the expiration passed to the timeout API functions is
an expiration interval. If this flag is specified, then it is
interpreted as the expiration time itself.
CALLOUT_FLAG_ROUNDUP
Roundup the expiration time to the next resolution boundary. If this
flag is not specified, the expiration time is rounded down.
projects / spl / log