A nice writeup of the LKMM (Linux Kernel Memory Model) rules for access
once policies can be found here
https://lwn.net/Articles/799218/#Access-Marking%20Policies .
The locked and unlocked access to eb::blocking_writers should be
annotated accordingly, following this:
Writes:
- locked write must use ONCE, may use plain read
- unlocked write must use ONCE
Reads:
- unlocked read must use ONCE
- locked read may use plain read iff not mixed with unlocked read
- unlocked read then locked must use ONCE
There's one difference on the assembly level, where
btrfs_tree_read_lock_atomic and btrfs_try_tree_read_lock used the cached
value and did not reevaluate it after taking the lock. This could have
missed some opportunities to take the lock in case blocking writers
changed between the calls, but the window is just a few instructions
long. As this is in try-lock, the callers handle that.
Signed-off-by: David Sterba <dsterba@suse.com>
The increment and decrement was inherited from previous version that
used atomics, switched in commit 06297d8cef ("btrfs: switch
extent_buffer blocking_writers from atomic to int"). The only possible
values are 0 and 1 so we can set them directly.
The generated assembly (gcc 9.x) did the direct value assignment in
btrfs_set_lock_blocking_write (asm diff after change in 06297d8cef):
5d: test %eax,%eax
5f: je 62 <btrfs_set_lock_blocking_write+0x22>
61: retq
- 62: lock incl 0x44(%rdi)
- 66: add $0x50,%rdi
- 6a: jmpq 6f <btrfs_set_lock_blocking_write+0x2f>
+ 62: movl $0x1,0x44(%rdi)
+ 69: add $0x50,%rdi
+ 6d: jmpq 72 <btrfs_set_lock_blocking_write+0x32>
The part in btrfs_tree_unlock did a decrement because
BUG_ON(blockers > 1) is probably not a strong hint for the compiler, but
otherwise the output looks safe:
- lock decl 0x44(%rdi)
+ sub $0x1,%eax
+ mov %eax,0x44(%rdi)
Signed-off-by: David Sterba <dsterba@suse.com>
There are two ifs that use eb::blocking_writers. As this is a variable
modified inside and outside of locks, we could minimize number of
accesses to avoid problems with getting different results at different
times.
The access here is locked so this can only race with btrfs_tree_unlock
that sets blocking_writers to 0 without lock and unsets the lock owner.
The first branch is taken only if the same thread already holds the
lock, the second if checks for blocking writers. Here we'd either unlock
and wait, or proceed. Both are valid states of the locking protocol.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
The function belongs to the family of locking functions, so move it
there. The 'noinline' keyword is dropped as it's now an exported
function that does not need it.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The function belongs to the family of locking functions, so move it
there. The 'noinline' keyword is dropped as it's now an exported
function that does not need it.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The function btrfs_assert_tree_locked is used outside of the locking
code so it is exported, however we can make it static inine as it's
fairly trivial.
This is the only locking assertion used in release builds, inlining
improves the text size by 174 bytes and reduces stack consumption in the
callers.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
I've noticed that none of the btrfs_assert_*lock* debugging helpers is
inlined, despite they're short and mostly a value update. Making them
inline shaves 67 from the text size, reduces stack consumption and
perhaps also slightly improves the performance due to avoiding
unnecessary calls.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The file ctree.h serves as a header for everything and has become quite
bloated. Split some helpers that are generic and create a new file that
should be the catch-all for code that's not btrfs-specific.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Those were split out of btrfs_clear_lock_blocking_rw by
aa12c02778 ("btrfs: split btrfs_clear_lock_blocking_rw to read and write helpers")
however at that time this function was unused due to commit
5239834016 ("Btrfs: kill btrfs_clear_path_blocking"). Put the final
nail in the coffin of those 2 functions.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Commit 06297d8cef ("btrfs: switch extent_buffer blocking_writers from
atomic to int") changed the type of blocking_writers but forgot to
adjust relevant code in btrfs_tree_unlock by converting the
smp_mb__after_atomic to smp_mb. This opened up the possibility of a
deadlock due to re-ordering of setting blocking_writers and
checking/waking up the waiter. This particular lockup is explained in a
comment above waitqueue_active() function.
Fix it by converting the memory barrier to a full smp_mb, accounting
for the fact that blocking_writers is a simple integer.
Fixes: 06297d8cef ("btrfs: switch extent_buffer blocking_writers from atomic to int")
Tested-by: Johannes Thumshirn <jthumshirn@suse.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The write_locks is either 0 or 1 and always updated under the lock,
so we don't need the atomic_t semantics.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The spinning_writers is either 0 or 1 and always updated under the lock,
so we don't need the atomic_t semantics.
Signed-off-by: David Sterba <dsterba@suse.com>
The blocking_writers is either 0 or 1 and always updated under the lock,
so we don't need the atomic_t semantics.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Unlike btrfs_tree_lock() and btrfs_tree_read_lock(), the remaining
functions in locking.c will not sleep, thus doesn't make much sense to
record their execution time.
Those events are introduced mainly for user space tool to audit and
detect lock leakage or dead lock.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There are two tree lock events which can sleep:
- btrfs_tree_read_lock()
- btrfs_tree_lock()
Sometimes we may need to look into the concurrency picture of the fs.
For that case, we need the execution time of above two functions and the
owner of @eb.
Here we introduce a trace events for user space tools like bcc, to get
the execution time of above two functions, and get detailed owner info
where eBPF code can't.
All the overhead is hidden behind the trace events, so if events are not
enabled, there is no overhead.
These trace events also output bytenr and generation, allow them to be
pared with unlock events to pin down deadlock.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The member is tracking simple status of the lock, we can use bool for
that and make some room for further space reduction in the structure.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Use the helpers where open coded. On non-debug builds, the warnings will
not trigger and extent_buffer::write_locks become unused and can be
moved to the appropriate section, saving a few bytes.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
The write_locks are a simple counter to track locking balance and used
to assert tree locks. Add helpers to make it conditionally work only in
DEBUG builds. Will be used in followup patches.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Use the helpers where open coded. On non-debug builds, the warnings will
not trigger and extent_buffer::read_locks become unused and can be
moved to the appropriate section, saving a few bytes.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
The read_locks are a simple counter to track locking balance and used to
assert tree locks. Add helpers to make it conditionally work only in
DEBUG builds. Will be used in followup patches.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Use the helpers where open coded. On non-debug builds, the warnings will
not trigger and extent_buffer::spining_readers become unused and can be
moved to the appropriate section, saving a few bytes.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Add helpers for conditional DEBUG build to assert that the extent buffer
spinning_readers constraints are met. Will be used in followup patches.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Use the helpers where open coded. On non-debug builds, the warnings will
not trigger and extent_buffer::spining_writers become unused and can be
moved to the appropriate section, saving a few bytes.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Add helpers for conditional DEBUG build to assert that the extent buffer
spinning_writers constraints are met. Will be used in followup patches.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently, the number of readers and writers is checked and in case
there are any, wait and redo the locks. There's some duplication
before the branches go back to again label, eg. calling wait_event on
blocking_readers twice.
The sequence is transformed
loop:
* wait for readers
* wait for writers
* write_lock
* check readers, unlock and wait for readers, loop
* check writers, unlock and wait for writers, loop
The new sequence is not exactly the same due to the simplification, for
readers it's slightly faster. For the writers, original code does
* wait for writers
* (loop) wait for readers
* wait for writers -- again
while the new goes directly to the reader check. This should behave the
same on a contended lock with multiple writers and readers, but can
reduce number of times we're waiting on something.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
There are many callers that hardcode the desired lock type so we can
avoid the switch and call them directly. Split the current function to
two. There are no remaining users of btrfs_clear_lock_blocking_rw so
it's removed. The call sites will be converted in followup patches.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
There are many callers that hardcode the desired lock type so we can
avoid the switch and call them directly. Split the current function to
two but leave a helper that still takes the variable lock type to make
current code compile. The call sites will be converted in followup
patches.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Use the wrappers and reduce the amount of low-level details about the
waitqueue management.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Remove GPL boilerplate text (long, short, one-line) and keep the rest,
ie. personal, company or original source copyright statements. Add the
SPDX header.
Signed-off-by: David Sterba <dsterba@suse.com>
When performing an unlock on an extent buffer we'd like to order the
decrement of extent_buffer::blocking_writers with waking up any
waiters. In such situations it's sufficient to use smp_mb__after_atomic
rather than the heavy smp_mb. On architectures where atomic operations
are fully ordered (such as x86 or s390) unconditionally executing
a heavyweight smp_mb instruction causes a severe hit to performance
while bringin no improvements in terms of correctness.
The better thing is to use the appropriate smp_mb__after_atomic routine
which will do the correct thing (invoke a full smp_mb or in the case
of ordered atomics insert a compiler barrier). Put another way,
an RMW atomic op + smp_load__after_atomic equals, in terms of
semantics, to a full smp_mb. This ensures that none of the problems
described in the accompanying comment of waitqueue_active occur.
No functional changes.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When a task trying to double lock a extent buffer, there are no
lockdep warning about it because this lock may be in "blocking_lock"
state, and make us hard to debug.
This patch add a WARN_ON() for above condition, it can not report
all deadlock cases(as lock between tasks), but at least helps us
some.
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
The fair reader/writer locks mean that btrfs_clear_path_blocking needs
to strictly follow lock ordering rules even when we already have
blocking locks on a given path.
Before we can clear a blocking lock on the path, we need to make sure
all of the locks have been converted to blocking. This will remove lock
inversions against anyone spinning in write_lock() against the buffers
we're trying to get read locks on. These inversions didn't exist before
the fair read/writer locks, but now we need to be more careful.
We papered over this deadlock in the past by changing
btrfs_try_read_lock() to be a true trylock against both the spinlock and
the blocking lock. This was slower, and not sufficient to fix all the
deadlocks. This patch adds a btrfs_tree_read_lock_atomic(), which
basically means get the spinlock but trylock on the blocking lock.
Signed-off-by: Chris Mason <clm@fb.com>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Reported-by: Patrick Schmid <schmid@phys.ethz.ch>
cc: stable@vger.kernel.org #v3.15+
The Btrfs tree trylock function is poorly named. It always takes
the spinlock and backs off if the blocking lock is held. This
can lead to surprising lockups because people expect it to really be a
trylock.
This commit makes it a pure trylock, both for the spinlock and the
blocking lock. It also reworks the nested lock handling slightly to
avoid taking the read lock while a spinning write lock might be held.
Signed-off-by: Chris Mason <clm@fb.com>
Big patch, but all it does is add statics to functions which
are in fact static, then remove the associated dead-code fallout.
removed functions:
btrfs_iref_to_path()
__btrfs_lookup_delayed_deletion_item()
__btrfs_search_delayed_insertion_item()
__btrfs_search_delayed_deletion_item()
find_eb_for_page()
btrfs_find_block_group()
range_straddles_pages()
extent_range_uptodate()
btrfs_file_extent_length()
btrfs_scrub_cancel_devid()
btrfs_start_transaction_lflush()
btrfs_print_tree() is left because it is used for debugging.
btrfs_start_transaction_lflush() and btrfs_reada_detach() are
left for symmetry.
ulist.c functions are left, another patch will take care of those.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
This does not change the logic of code, but can save us a read_lock.
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
The btrfs locks were unconditionally calling wake_up as the
locks were released. This lead to extra thrashing on the waitqueue,
especially for locks that were dominated by readers.
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
This patch adds the possibilty to read-lock an extent even if it is already
write-locked from the same thread. btrfs_find_all_roots() needs this
capability.
Signed-off-by: Arne Jansen <sensille@gmx.net>
Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
The btrfs metadata btree is the source of significant
lock contention, especially in the root node. This
commit changes our locking to use a reader/writer
lock.
The lock is built on top of rw spinlocks, and it
extends the lock tracking to remember if we have a
read lock or a write lock when we go to blocking. Atomics
count the number of blocking readers or writers at any
given time.
It removes all of the adaptive spinning from the old code
and uses only the spinning/blocking hints inside of btrfs
to decide when it should continue spinning.
In read heavy workloads this is dramatically faster. In write
heavy workloads we're still faster because of less contention
on the root node lock.
We suffer slightly in dbench because we schedule more often
during write locks, but all other benchmarks so far are improved.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Remove static and global declarations and/or definitions. Reduces size
of btrfs.ko by ~3.4kB.
text data bss dec hex filename
402081 7464 200 409745 64091 btrfs.ko.base
398620 7144 200 405964 631cc btrfs.ko.remove-all
Signed-off-by: David Sterba <dsterba@suse.cz>
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>
btrfs_mark_buffer dirty would set dirty bits in the extent_io tree
for the buffers it was dirtying. This may require a kmalloc and it
was not atomic. So, anyone who called btrfs_mark_buffer_dirty had to
set any btree locks they were holding to blocking first.
This commit changes dirty tracking for extent buffers to just use a flag
in the extent buffer. Now that we have one and only one extent buffer
per page, this can be safely done without losing dirty bits along the way.
This also introduces a path->leave_spinning flag that callers of
btrfs_search_slot can use to indicate they will properly deal with a
path returned where all the locks are spinning instead of blocking.
Many of the btree search callers now expect spinning paths,
resulting in better btree concurrency overall.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This reduces contention on the extent buffer spin locks by testing for a
blocking lock before trying to take the spinlock.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
btrfs_tree_locked was being used to make sure a given extent_buffer was
properly locked in a few places. But, it wasn't correct for UP compiled
kernels.
This switches it to using assert_spin_locked instead, and renames it to
btrfs_assert_tree_locked to better reflect how it was really being used.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
David Sterba