This ended up causing some noise in places such as rxrpc running in softirq.
The warning is misleading in this case as the mutex trylock and unlock
operations are done within the same context; and therefore we need not
worry about the PI-boosting issues that comes along with no single-owner
lock guarantees.
While we don't want to support this in mutexes, there is no way out of
this yet; so lets get rid of the WARNs for now, as it is only fair to
code that has historically relied on non-preemptible softirq guarantees.
In addition, changing the lock type is also unviable: exclusive rwsems
have the same issue (just not the WARN_ON) and counting semaphores
would introduce a performance hit as mutexes are a lot more optimized.
This reverts:
a0855d24fc: ("locking/mutex: Complain upon mutex API misuse in IRQ contexts")
Fixes: a0855d24fc: ("locking/mutex: Complain upon mutex API misuse in IRQ contexts")
Signed-off-by: Davidlohr Bueso <[email protected]>
Tested-by: David Howells <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
At a slight footprint cost (24 vs 32 bytes), mutexes are more optimal
than semaphores; it's also a nicer interface for mutual exclusion,
which is why they are encouraged over binary semaphores, when possible.
Replace the hyperv_mmio_lock, its semantics implies traditional lock
ownership; that is, the lock owner is the same for both lock/unlock
operations. Therefore it is safe to convert.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Sasha Levin <[email protected]>
With some considerations, the custom pat_rbtree implementation can be
simplified to use most of the generic interval_tree machinery:
- The tree inorder traversal can slightly differ when there are key
('start') collisions in the tree due to one going left and another right.
This, however, only affects the output of debugfs' pat_memtype_list file.
- Generic interval trees are now fully closed [a, b], for which we need
to adjust the last endpoint (ie: end - 1).
- Erasing logic must remain untouched as well.
- In order for the types to remain u64, the 'memtype_interval' calls are
introduced, as opposed to simply using struct interval_tree.
In addition, the PAT tree might potentially also benefit by the fast overlap
detection for the insertion case when looking up the first overlapping node
in the tree.
No change in behavior is intended.
Finally, I've tested this on various servers, via sanity warnings, running
side by side with the current version and so far see no differences in the
returned pointer node when doing memtype_rb_lowest_match() lookups.
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Thomas Gleixner <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
At a slight footprint cost (24 vs 32 bytes), mutexes are more optimal
than semaphores; it's also a nicer interface for mutual exclusion,
which is why they are encouraged over binary semaphores, when possible.
For both i_link_lock and i_ext_lock (and hence i_hash_lock which I
annotated for the hash lock mapping hackery for lockdep), their semantics
imply traditional lock ownership; that is, the lock owner is the same for
both lock/unlock operations and does not run in irq context. Therefore
it is safe to convert.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: David Sterba <[email protected]>
At a slight footprint cost (24 vs 32 bytes), mutexes are more optimal
than semaphores; it's also a nicer interface for mutual exclusion,
which is why they are encouraged over binary semaphores, when possible.
There is also lockdep support.
For both f_sem and b_sem, their semantics imply traditional lock
ownership; that is, the lock owner is the same for both lock/unlock
operations and not under irq contexts (ie for trylock/unlock scenarios).
Therefore it is safe to convert.
Signed-off-by: Davidlohr Bueso <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>
At a slight footprint cost (24 vs 32 bytes), mutexes are more optimal
than semaphores; it's also a nicer interface for mutual exclusion,
which is why they are encouraged over binary semaphores, when possible.
For both v_sem and z_sem, their semantics imply traditional lock
ownership; that is, the lock owner is the same for both lock/unlock
operations. Therefore it is safe to convert.
Signed-off-by: Davidlohr Bueso <[email protected]>
Acked-by: Valdis Kletnieks <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>
Add warning checks if mutex_trylock() or mutex_unlock() are used in
IRQ contexts, under CONFIG_DEBUG_MUTEXES=y.
While the mutex rules and semantics are explicitly documented, this allows
to expose any abusers and robustifies the whole thing.
While trylock and unlock are non-blocking, calling from IRQ context
is still forbidden (lock must be within the same context as unlock).
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Currently rwsems is the only locking primitive that lacks this
debug feature. Add it under CONFIG_DEBUG_RWSEMS and do the magic
checking in the locking fastpath (trylock) operation such that
we cover all cases. The unlocking part is pretty straightforward.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Acked-by: Waiman Long <[email protected]>
Cc: [email protected]
Cc: Davidlohr Bueso <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
Simplify the timerqueue code by using cached rbtrees and rely on the tree
leftmost node semantics to get the timer with earliest expiration time.
This is a drop in conversion, and therefore semantics remain untouched.
The runtime overhead of cached rbtrees is be pretty much the same as the
current head->next method, noting that when removing the leftmost node,
a common operation for the timerqueue, the rb_next(leftmost) is O(1) as
well, so the next timer will either be the right node or its parent.
Therefore no extra pointer chasing. Finally, the size of the struct
timerqueue_head remains the same.
Passes several hours of rcutorture.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Link: https://lkml.kernel.org/r/20190724152323.bojciei3muvfxalm@linux-r8p5
Our msg priorities became an rbtree as of d6629859b3 ("ipc/mqueue:
improve performance of send/recv"). However, consuming a msg in
msg_get() remains logarithmic (still being better than the case before
of course). By applying well known techniques to cache pointers we can
have the node with the highest priority in O(1), which is specially nice
for the rt cases. Furthermore, some callers can call msg_get() in a
loop.
A new msg_tree_erase() helper is also added to encapsulate the tree
removal and node_cache game. Passes ltp mq testcases.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Manfred Spraul <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This is a lot more appropriate than PI_LIST, which in the kernel one
would assume that it has to do with priority-inheritance; which is not
-- furthermore futexes make use of plists so this can be even more
confusing, albeit the debug nature of the config option.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Andrew Morton <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The current check does not take into account the previous value of
pinned_vm; thus it is quite bogus as is. Fix this by checking the
new value after the (optimistic) atomic inc.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>
Holding mmap_sem exclusively for a gup() is an overkill. Lets
share the lock and replace the gup call for gup_longterm(), as
it is better suited for the lifetime of the pinning.
Fixes: c0c77d8fb7 ("xsk: add user memory registration support sockopt")
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: David S. Miller <[email protected]>
Cc: Bjorn Topel <[email protected]>
Cc: Magnus Karlsson <[email protected]>
CC: [email protected]
Acked-by: Björn Töpel <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
The current comment for the barrier that guarantees that waiter increment
is always before taking the hb spinlock (barrier (A)) needs to be fixed as
it is misplaced.
This is obviously referring to hb_waiters_inc, which is a full barrier.
Reported-by: Peter Zijlstra <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
ib_umem_get() uses gup_longterm() and relies on the lock to stabilze the
vma_list, so we cannot really get rid of mmap_sem altogether, but now that
the counter is atomic, we can get of some complexity that mmap_sem brings
with only pinned_vm.
Reviewed-by: Ira Weiny <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>
usnic_uiom_get_pages() uses gup_longterm() so we cannot really get rid of
mmap_sem altogether in the driver, but we can get rid of some complexity
that mmap_sem brings with only pinned_vm. We can get rid of the wq
altogether as we no longer need to defer work to unpin pages as the
counter is now atomic. We also share the lock.
Acked-by: Parvi Kaustubhi <[email protected]>
Reviewed-by: Ira Weiny <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>
This driver already uses gup_fast() and thus we can just drop the mmap_sem
protection around the pinned_vm counter. Note that the window between when
hfi1_can_pin_pages() is called and the actual counter is incremented
remains the same as mmap_sem was _only_ used for when ->pinned_vm was
touched.
Reviewed-by: Ira Weiny <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>
The driver uses mmap_sem for both pinned_vm accounting and
get_user_pages(). Because rdma drivers might want to use gup_longterm() in
the future we still need some sort of mmap_sem serialization (as opposed
to removing it entirely by using gup_fast()). Now that pinned_vm is atomic
the writer lock can therefore be converted to reader.
This also fixes a bug that __qib_get_user_pages was not taking into
account the current value of pinned_vm.
Reviewed-by: Ira Weiny <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>
The driver uses mmap_sem for both pinned_vm accounting and
get_user_pages(). By using gup_fast() and letting the mm handle the lock
if needed, we can no longer rely on the semaphore and simplify the whole
thing.
Reviewed-by: Ira Weiny <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>
Taking a sleeping lock to _only_ increment a variable is quite the
overkill, and pretty much all users do this. Furthermore, some drivers
(ie: infiniband and scif) that need pinned semantics can go to quite
some trouble to actually delay via workqueue (un)accounting for pinned
pages when not possible to acquire it.
By making the counter atomic we no longer need to hold the mmap_sem and
can simply some code around it for pinned_vm users. The counter is 64-bit
such that we need not worry about overflows such as rdma user input
controlled from userspace.
Reviewed-by: Ira Weiny <[email protected]>
Reviewed-by: Christoph Lameter <[email protected]>
Reviewed-by: Daniel Jordan <[email protected]>
Reviewed-by: Jan Kara <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>
Some users, specifically futexes and rwsems, required fixes
that allowed the callers to be safe when wakeups occur before
they are expected by wake_up_q(). Such scenarios also play
games and rely on reference counting, and until now were
pivoting on wake_q doing it. With the wake_q_add() call being
moved down, this can no longer be the case. As such we end up
with a a double task refcounting overhead; and these callers
care enough about this (being rather core-ish).
This patch introduces a wake_q_add_safe() call that serves
for callers that have already done refcounting and therefore the
task is 'safe' from wake_q point of view (int that it requires
reference throughout the entire queue/>wakeup cycle). In the one
case it has internal reference counting, in the other case it
consumes the reference counting.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Waiman Long <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Xie Yongji <[email protected]>
Cc: Yongji Xie <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: https://lkml.kernel.org/r/20181218195352.7orq3upiwfdbrdne@linux-r8p5
Signed-off-by: Ingo Molnar <[email protected]>
At the cost of an extra pointer, we can avoid the O(logN) cost of
finding the first element in the tree (smallest node), which is
something heavily required for histograms. Specifically, the following
are converted to rb_root_cached, and users accordingly:
hist::entries_in_array
hist::entries_in
hist::entries
hist::entries_collapsed
hist_entry::hroot_in
hist_entry::hroot_out
Signed-off-by: Davidlohr Bueso <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
[ Added some missing conversions to rb_first_cached() ]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
At the cost of an extra pointer, we can avoid the O(logN) cost of
finding the first element in the tree (smallest node), which is
something required for any of the strlist or intlist traversals
(XXX_for_each_entry()). There are a number of users in perf of these
(particularly strlists), including probes, and buildid.
Signed-off-by: Davidlohr Bueso <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
At the cost of an extra pointer, we can avoid the O(logN) cost of
finding the first element in the tree (smallest node), which is
something required for nearly every in/srcline callchain node deletion
(in/srcline__tree_delete()).
Signed-off-by: Davidlohr Bueso <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
At the cost of an extra pointer, we can avoid the O(logN) cost of
finding the first element in the tree (smallest node), which is
something required for nearly every operation dealing with
machine->guests and threads->entries.
The conversion is straightforward, however, it's worth noticing that the
rb_erase_init() calls have been replaced by rb_erase_cached() which has
no _init() flavor, however, the node is explicitly cleared next anyway,
which was redundant until now.
Signed-off-by: Davidlohr Bueso <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
There is no reason why we rearm the waitiqueue upon every fetch_events
retry (for when events are found yet send_events() fails). If nothing
else, this saves four lock operations per retry, and furthermore reduces
the scope of the lock even further.
[[email protected]: restore code to original position, fix and reflow comment]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Jason Baron <[email protected]>
Cc: Al Viro <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This patch aims at reducing ep wq.lock hold times in epoll_wait(2). For
the blocking case, there is no need to constantly take and drop the
spinlock, which is only needed to manipulate the waitqueue.
The call to ep_events_available() is now lockless, and only exposed to
benign races. Here, if false positive (returns available events and
does not see another thread deleting an epi from the list) we call into
send_events and then the list's state is correctly seen. Otoh, if a
false negative and we don't see a list_add_tail(), for example, from irq
callback, then it is rechecked again before blocking, which will see the
correct state.
In order for more accuracy to see concurrent list_del_init(), use the
list_empty_careful() variant -- of course, this won't be safe against
insertions from wakeup.
For the overflow list we obviously need to prevent load/store tearing as
we don't want to see partial values while the ready list is disabled.
[[email protected]: forgotten fixlets]
Link: http://lkml.kernel.org/r/20181109155258.jxcr4t2pnz6zqct3@linux-r8p5
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Suggested-by: Jason Baron <[email protected]>
Cc: Al Viro <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The ep->ovflist is a secondary ready-list to temporarily store events
that might occur when doing sproc without holding the ep->wq.lock. This
accounts for every time we check for ready events and also send events
back to userspace; both callbacks, particularly the latter because of
copy_to_user, can account for a non-trivial time.
As such, the unlikely() check to see if the pointer is being used, seems
both misleading and sub-optimal. In fact, we go to an awful lot of
trouble to sync both lists, and populating the ovflist is far from an
uncommon scenario.
For example, profiling a concurrent epoll_wait(2) benchmark, with
CONFIG_PROFILE_ANNOTATED_BRANCHES shows that for a two threads a 33%
incorrect rate was seen; and when incrementally increasing the number of
epoll instances (which is used, for example for multiple queuing load
balancing models), up to a 90% incorrect rate was seen.
Similarly, by deleting the prediction, 3% throughput boost was seen
across incremental threads.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Andrew Morton <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Jason Baron <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The current logic is a bit convoluted. Lets simplify this with a
standard list_for_each_entry_safe() loop instead and just break out
after maxevents is reached.
While at it, remove an unnecessary indentation level in the loop when
there are in fact ready events.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Andrew Morton <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Jason Baron <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Patch series "epoll: some miscellaneous optimizations".
The following are some incremental optimizations on some of the epoll
core. Each patch has the details, but together, the series is seen to
shave off measurable cycles on a number of systems and workloads.
For example, on a 40-core IB, a pipetest as well as parallel
epoll_wait() benchmark show around a 20-30% increase in raw operations
per second when the box is fully occupied (incremental thread counts),
and up to 15% performance improvement with lower counts.
Passes ltp epoll related testcases.
This patch(of 6):
All callers pass the EP_MAX_NESTS constant already, so lets simplify
this a tad and get rid of the redundant parameter for nested eventpolls.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Andrew Morton <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Jason Baron <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This is already done for us internally by the signal machinery.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Wolfram Sang <[email protected]>
This program benchmarks concurrent epoll_wait(2) for file descriptors
that are monitored with with EPOLLIN along various semantics, by a
single epoll instance. Such conditions can be found when using
single/combined or multiple queuing when load balancing.
Each thread has a number of private, nonblocking file descriptors,
referred to as fdmap. A writer thread will constantly be writing to the
fdmaps of all threads, minimizing each threads's chances of epoll_wait
not finding any ready read events and blocking as this is not what we
want to stress. Full details in the start of the C file.
Committer testing:
# perf bench
Usage:
perf bench [<common options>] <collection> <benchmark> [<options>]
# List of all available benchmark collections:
sched: Scheduler and IPC benchmarks
mem: Memory access benchmarks
numa: NUMA scheduling and MM benchmarks
futex: Futex stressing benchmarks
epoll: Epoll stressing benchmarks
all: All benchmarks
# perf bench epoll
# List of available benchmarks for collection 'epoll':
wait: Benchmark epoll concurrent epoll_waits
all: Run all futex benchmarks
# perf bench epoll wait
# Running 'epoll/wait' benchmark:
Run summary [PID 19295]: 3 threads monitoring on 64 file-descriptors for 8 secs.
[thread 0] fdmap: 0xdaa650 ... 0xdaa74c [ 328241 ops/sec ]
[thread 1] fdmap: 0xdaa900 ... 0xdaa9fc [ 351695 ops/sec ]
[thread 2] fdmap: 0xdaabb0 ... 0xdaacac [ 381423 ops/sec ]
Averaged 353786 operations/sec (+- 4.35%), total secs = 8
#
Committer notes:
Fix the build on debian:experimental-x-mips, debian:experimental-x-mipsel
and others:
CC /tmp/build/perf/bench/epoll-wait.o
bench/epoll-wait.c: In function 'writerfn':
bench/epoll-wait.c:399:12: error: format '%ld' expects argument of type 'long int', but argument 2 has type 'size_t' {aka 'unsigned int'} [-Werror=format=]
printinfo("exiting writer-thread (total full-loops: %ld)\n", iter);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~
bench/epoll-wait.c:86:31: note: in definition of macro 'printinfo'
do { if (__verbose) { printf(fmt, ## arg); fflush(stdout); } } while (0)
^~~
cc1: all warnings being treated as errors
Signed-off-by: Davidlohr Bueso <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Davidlohr Bueso <[email protected]>
Cc: Jason Baron <[email protected]> <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Link: http://lkml.kernel.org/r/20181106182349.thdkpvshkna5vd7o@linux-r8p5>
[ Applied above fixup as per Davidlohr's request ]
[ Use inttypes.h to print rlim_t fields, fixing the build on Alpine Linux / musl libc ]
[ Check if eventfd() is available, i.e. if HAVE_EVENTFD is defined ]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
In sysvipc we have an ids->tables_initialized regarding the rhashtable,
introduced in 0cfb6aee70 ("ipc: optimize semget/shmget/msgget for lots
of keys")
It's there, specifically, to prevent nil pointer dereferences, from using
an uninitialized api. Considering how rhashtable_init() can fail
(probably due to ENOMEM, if anything), this made the overall ipc
initialization capable of failure as well. That alone is ugly, but fine,
however I've spotted a few issues regarding the semantics of
tables_initialized (however unlikely they may be):
- There is inconsistency in what we return to userspace: ipc_addid()
returns ENOSPC which is certainly _wrong_, while ipc_obtain_object_idr()
returns EINVAL.
- After we started using rhashtables, ipc_findkey() can return nil upon
!tables_initialized, but the caller expects nil for when the ipc
structure isn't found, and can therefore call into ipcget() callbacks.
Now that rhashtable initialization cannot fail, we can properly get rid of
the hack altogether.
[[email protected]: commit id extended to 12 digits]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Manfred Spraul <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Herbert Xu <[email protected]>
Cc: Kees Cook <[email protected]>
Cc: Michael Kerrisk <[email protected]>
Cc: Michal Hocko <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
rhashtable_init() may fail due to -ENOMEM, thus making the entire api
unusable. This patch removes this scenario, however unlikely. In order
to guarantee memory allocation, this patch always ends up doing
GFP_KERNEL|__GFP_NOFAIL for both the tbl as well as
alloc_bucket_spinlocks().
Upon the first table allocation failure, we shrink the size to the
smallest value that makes sense and retry with __GFP_NOFAIL semantics.
With the defaults, this means that from 64 buckets, we retry with only 4.
Any later issues regarding performance due to collisions or larger table
resizing (when more memory becomes available) is the least of our
problems.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Manfred Spraul <[email protected]>
Acked-by: Herbert Xu <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Kees Cook <[email protected]>
Cc: Michael Kerrisk <[email protected]>
Cc: Michal Hocko <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
As of ce91f6ee5b ("mm: kvmalloc does not fallback to vmalloc for
incompatible gfp flags") we can simplify the caller and trust kvzalloc()
to just do the right thing. For the case of the GFP_ATOMIC context, we
can drop the __GFP_NORETRY flag for obvious reasons, and for the
__GFP_NOWARN case, however, it is changed such that the caller passes the
flag instead of making bucket_table_alloc() handle it.
This slightly changes the gfp flags passed on to nested_table_alloc() as
it will now also use GFP_ATOMIC | __GFP_NOWARN. However, I consider this
a positive consequence as for the same reasons we want nowarn semantics in
bucket_table_alloc().
[[email protected]: commit id extended to 12 digits, line wraps updated]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Manfred Spraul <[email protected]>
Acked-by: Michal Hocko <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Herbert Xu <[email protected]>
Cc: Kees Cook <[email protected]>
Cc: Michael Kerrisk <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
ipc/util.c contains multiple functions to get the ipc object pointer given
an id number.
There are two sets of function: One set verifies the sequence counter part
of the id number, other functions do not check the sequence counter.
The standard for function names in ipc/util.c is
- ..._check() functions verify the sequence counter
- ..._idr() functions do not verify the sequence counter
ipc_lock() is an exception: It does not verify the sequence counter value,
but this is not obvious from the function name.
Furthermore, shm.c is the only user of this helper. Thus, we can simply
move the logic into shm_lock() and get rid of the function altogether.
[[email protected]: most of changelog]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Manfred Spraul <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Herbert Xu <[email protected]>
Cc: Kees Cook <[email protected]>
Cc: Michael Kerrisk <[email protected]>
Cc: Michal Hocko <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Similar to other calls, ep_poll() is not called with interrupts disabled,
and we can therefore avoid the irq save/restore dance and just disable
local irqs. In fact, the call should never be called in irq context at
all, considering that the only path is
epoll_wait(2) -> do_epoll_wait() -> ep_poll().
When running on a 2 socket 40-core (ht) IvyBridge a common pipe based
epoll_wait(2) microbenchmark, the following performance improvements are
seen:
# threads vanilla dirty
1 1805587 2106412
2 1854064 2090762
4 1805484 2017436
8 1751222 1974475
16 1725299 1962104
32 1378463 1571233
64 787368 900784
Which is a pretty constantly near 15%.
Also add a lockdep check such that we detect any mischief before
deadlocking.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Andrew Morton <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Jason Baron <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Both functions are similar to the context of ep_modify(), called via
epoll_ctl(2). Just like ep_modify(), saving and restoring interrupts is
an overkill in these calls as it will never be called with irqs disabled.
While ep_remove() can be called directly from EPOLL_CTL_DEL, it can also
be called when releasing the file, but this also complies with the above.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Jason Baron <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Patch series "fs/epoll: loosen irq safety when possible".
Both patches replace saving+restoring interrupts when taking the ep->lock
(now the waitqueue lock), with just disabling local irqs. This shows
immediate performance benefits in patch 1 for an epoll workload running on
Xen. The main concern we need to have with this sort of changes in epoll
is the ep_poll_callback() which is passed to the wait queue wakeup and is
done very often under irq context, this patch does not touch this call.
Patches have been tested pretty heavily with the customer workload,
microbenchmarks, ltp testcases and two high level workloads that use epoll
under the hood: nginx and libevent benchmarks.
This patch (of 2):
Saving and restoring interrupts in ep_scan_ready_list() is an
overkill as it is never called with interrupts disabled. Loosen
this to simply disabling local irqs such that archs where managing
irqs is expensive or virtual environments. This patch yields
some throughput improvements on a workload that is epoll intensive
running on a single Xen DomU.
1 Job 7500 --> 8800 enq/s (+17%)
2 Jobs 14000 --> 15200 enq/s (+8%)
3 Jobs 20500 --> 22300 enq/s (+8%)
4 Jobs 25000 --> 28000 enq/s (+8-12)%
On bare metal:
For a 2-socket 40-core (ht) IvyBridge on a few workloads, unfortunately I
don't have a xen environment and the results for Xen I do have (which
numbers are in patch 1) I don't have the actual workload, so cannot
compare them directly.
1) Different configurations were used for a epoll_wait (pipes io)
microbench (http://linux-scalability.org/epoll/epoll-test.c) and shows
around a 7-10% improvement in overall total number of times the
epoll_wait() loops when using both regular and nested epolls, so very
raw numbers, but measurable nonetheless.
# threads vanilla dirty
1 1677717 1805587
2 1660510 1854064
4 1610184 1805484
8 1577696 1751222
16 1568837 1725299
32 1291532 1378463
64 752584 787368
Note that stddev is pretty small.
2) Another pipe test, which shows no real measurable improvement.
(http://www.xmailserver.org/linux-patches/pipetest.c)
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Jason Baron <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
In order for load/store tearing prevention to work, _all_ accesses to
the variable in question need to be done around READ and WRITE_ONCE()
macros. Ensure everyone does so for q->status variable for
semtimedop().
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Manfred Spraul <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
rhashtable_init() currently does not take into account the user-passed
min_size parameter unless param->nelem_hint is set as well. As such,
the default size (number of buckets) will always be HASH_DEFAULT_SIZE
even if the smallest allowed size is larger than that. Remediate this
by unconditionally calling into rounded_hashtable_size() and handling
things accordingly.
Signed-off-by: Davidlohr Bueso <[email protected]>
Acked-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Both smatch and coverity are reporting potential issues with spectre
variant 1 with the 'semnum' index within the sma->sems array, ie:
ipc/sem.c:388 sem_lock() warn: potential spectre issue 'sma->sems'
ipc/sem.c:641 perform_atomic_semop_slow() warn: potential spectre issue 'sma->sems'
ipc/sem.c:721 perform_atomic_semop() warn: potential spectre issue 'sma->sems'
Avoid any possible speculation by using array_index_nospec() thus
ensuring the semnum value is bounded to [0, sma->sem_nsems). With the
exception of sem_lock() all of these are slowpaths.
Link: http://lkml.kernel.org/r/20180423171131.njs4rfm2yzyeg6do@linux-n805
Signed-off-by: Davidlohr Bueso <[email protected]>
Reported-by: Dan Carpenter <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: "Gustavo A. R. Silva" <[email protected]>
Cc: Manfred Spraul <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Patch series "ipc/shm: shmat() fixes around nil-page".
These patches fix two issues reported[1] a while back by Joe and Andrea
around how shmat(2) behaves with nil-page.
The first reverts a commit that it was incorrectly thought that mapping
nil-page (address=0) was a no no with MAP_FIXED. This is not the case,
with the exception of SHM_REMAP; which is address in the second patch.
I chose two patches because it is easier to backport and it explicitly
reverts bogus behaviour. Both patches ought to be in -stable and ltp
testcases need updated (the added testcase around the cve can be
modified to just test for SHM_RND|SHM_REMAP).
[1] lkml.kernel.org/r/20180430172152.nfa564pvgpk3ut7p@linux-n805
This patch (of 2):
Commit 95e91b831f ("ipc/shm: Fix shmat mmap nil-page protection")
worked on the idea that we should not be mapping as root addr=0 and
MAP_FIXED. However, it was reported that this scenario is in fact
valid, thus making the patch both bogus and breaks userspace as well.
For example X11's libint10.so relies on shmat(1, SHM_RND) for lowmem
initialization[1].
[1] https://cgit.freedesktop.org/xorg/xserver/tree/hw/xfree86/os-support/linux/int10/linux.c#n347
Link: http://lkml.kernel.org/r/[email protected]
Fixes: 95e91b831f ("ipc/shm: Fix shmat mmap nil-page protection")
Signed-off-by: Davidlohr Bueso <[email protected]>
Reported-by: Joe Lawrence <[email protected]>
Reported-by: Andrea Arcangeli <[email protected]>
Cc: Manfred Spraul <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
No changes in refcount semantics -- key init is false; replace
static_key_enable with static_branch_enable
static_key_slow_inc|dec with static_branch_inc|dec
static_key_false with static_branch_unlikely
Added a '_key' suffix to udp and udpv6 encap_needed, for better
self documentation.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
No changes in refcount semantics -- key init is false; replace
static_key_slow_inc|dec with static_branch_inc|dec
static_key_false with static_branch_unlikely
Added a '_key' suffix to generic_xdp_needed, for better self
documentation.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
No changes in refcount semantics -- key init is false; replace
static_key_slow_inc|dec with static_branch_inc|dec
static_key_false with static_branch_unlikely
Added a '_key' suffix to netstamp_needed, for better self
documentation.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
No changes in semantics -- key init is false; replace
static_key_slow_inc|dec with static_branch_inc|dec
static_key_false with static_branch_unlikely
Added a '_key' suffix to both ingress_needed and egress_needed,
for better self documentation.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
No changes in refcount semantics -- key init is false; replace
static_key_slow_inc|dec with static_branch_inc|dec
static_key_false with static_branch_unlikely
Added a '_key' suffix to memalloc_socks, for better self
documentation.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
No changes in refcount semantics -- key init is false; replace
static_key_slow_inc|dec with static_branch_inc|dec
static_key_false with static_branch_unlikely
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
There is a permission discrepancy when consulting msq ipc object
metadata between /proc/sysvipc/msg (0444) and the MSG_STAT shmctl
command. The later does permission checks for the object vs S_IRUGO.
As such there can be cases where EACCESS is returned via syscall but the
info is displayed anyways in the procfs files.
While this might have security implications via info leaking (albeit no
writing to the msq metadata), this behavior goes way back and showing
all the objects regardless of the permissions was most likely an
overlook - so we are stuck with it. Furthermore, modifying either the
syscall or the procfs file can cause userspace programs to break (ie
ipcs). Some applications require getting the procfs info (without root
privileges) and can be rather slow in comparison with a syscall -- up to
500x in some reported cases for shm.
This patch introduces a new MSG_STAT_ANY command such that the msq ipc
object permissions are ignored, and only audited instead. In addition,
I've left the lsm security hook checks in place, as if some policy can
block the call, then the user has no other choice than just parsing the
procfs file.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Reported-by: Robert Kettler <[email protected]>
Cc: Eric W. Biederman <[email protected]>
Cc: Kees Cook <[email protected]>
Cc: Manfred Spraul <[email protected]>
Cc: Michael Kerrisk <[email protected]>
Cc: Michal Hocko <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
There is a permission discrepancy when consulting shm ipc object
metadata between /proc/sysvipc/sem (0444) and the SEM_STAT semctl
command. The later does permission checks for the object vs S_IRUGO.
As such there can be cases where EACCESS is returned via syscall but the
info is displayed anyways in the procfs files.
While this might have security implications via info leaking (albeit no
writing to the sma metadata), this behavior goes way back and showing
all the objects regardless of the permissions was most likely an
overlook - so we are stuck with it. Furthermore, modifying either the
syscall or the procfs file can cause userspace programs to break (ie
ipcs). Some applications require getting the procfs info (without root
privileges) and can be rather slow in comparison with a syscall -- up to
500x in some reported cases for shm.
This patch introduces a new SEM_STAT_ANY command such that the sem ipc
object permissions are ignored, and only audited instead. In addition,
I've left the lsm security hook checks in place, as if some policy can
block the call, then the user has no other choice than just parsing the
procfs file.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Reported-by: Robert Kettler <[email protected]>
Cc: Eric W. Biederman <[email protected]>
Cc: Kees Cook <[email protected]>
Cc: Manfred Spraul <[email protected]>
Cc: Michael Kerrisk <[email protected]>
Cc: Michal Hocko <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Patch series "sysvipc: introduce STAT_ANY commands", v2.
The following patches adds the discussed (see [1]) new command for shm
as well as for sems and msq as they are subject to the same
discrepancies for ipc object permission checks between the syscall and
via procfs. These new commands are justified in that (1) we are stuck
with this semantics as changing syscall and procfs can break userland;
and (2) some users can benefit from performance (for large amounts of
shm segments, for example) from not having to parse the procfs
interface.
Once merged, I will submit the necesary manpage updates. But I'm thinking
something like:
: diff --git a/man2/shmctl.2 b/man2/shmctl.2
: index 7bb503999941..bb00bbe21a57 100644
: --- a/man2/shmctl.2
: +++ b/man2/shmctl.2
: @@ -41,6 +41,7 @@
: .\" 2005-04-25, mtk -- noted aberrant Linux behavior w.r.t. new
: .\" attaches to a segment that has already been marked for deletion.
: .\" 2005-08-02, mtk: Added IPC_INFO, SHM_INFO, SHM_STAT descriptions.
: +.\" 2018-02-13, dbueso: Added SHM_STAT_ANY description.
: .\"
: .TH SHMCTL 2 2017-09-15 "Linux" "Linux Programmer's Manual"
: .SH NAME
: @@ -242,6 +243,18 @@ However, the
: argument is not a segment identifier, but instead an index into
: the kernel's internal array that maintains information about
: all shared memory segments on the system.
: +.TP
: +.BR SHM_STAT_ANY " (Linux-specific)"
: +Return a
: +.I shmid_ds
: +structure as for
: +.BR SHM_STAT .
: +However, the
: +.I shm_perm.mode
: +is not checked for read access for
: +.IR shmid ,
: +resembing the behaviour of
: +/proc/sysvipc/shm.
: .PP
: The caller can prevent or allow swapping of a shared
: memory segment with the following \fIcmd\fP values:
: @@ -287,7 +300,7 @@ operation returns the index of the highest used entry in the
: kernel's internal array recording information about all
: shared memory segments.
: (This information can be used with repeated
: -.B SHM_STAT
: +.B SHM_STAT/SHM_STAT_ANY
: operations to obtain information about all shared memory segments
: on the system.)
: A successful
: @@ -328,7 +341,7 @@ isn't accessible.
: \fIshmid\fP is not a valid identifier, or \fIcmd\fP
: is not a valid command.
: Or: for a
: -.B SHM_STAT
: +.B SHM_STAT/SHM_STAT_ANY
: operation, the index value specified in
: .I shmid
: referred to an array slot that is currently unused.
This patch (of 3):
There is a permission discrepancy when consulting shm ipc object metadata
between /proc/sysvipc/shm (0444) and the SHM_STAT shmctl command. The
later does permission checks for the object vs S_IRUGO. As such there can
be cases where EACCESS is returned via syscall but the info is displayed
anyways in the procfs files.
While this might have security implications via info leaking (albeit no
writing to the shm metadata), this behavior goes way back and showing all
the objects regardless of the permissions was most likely an overlook - so
we are stuck with it. Furthermore, modifying either the syscall or the
procfs file can cause userspace programs to break (ie ipcs). Some
applications require getting the procfs info (without root privileges) and
can be rather slow in comparison with a syscall -- up to 500x in some
reported cases.
This patch introduces a new SHM_STAT_ANY command such that the shm ipc
object permissions are ignored, and only audited instead. In addition,
I've left the lsm security hook checks in place, as if some policy can
block the call, then the user has no other choice than just parsing the
procfs file.
[1] https://lkml.org/lkml/2017/12/19/220
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Acked-by: Michal Hocko <[email protected]>
Cc: Michael Kerrisk <[email protected]>
Cc: Manfred Spraul <[email protected]>
Cc: Eric W. Biederman <[email protected]>
Cc: Kees Cook <[email protected]>
Cc: Robert Kettler <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
While running rt-tests' pi_stress program I got the following splat:
rq->clock_update_flags < RQCF_ACT_SKIP
WARNING: CPU: 27 PID: 0 at kernel/sched/sched.h:960 assert_clock_updated.isra.38.part.39+0x13/0x20
[...]
<IRQ>
enqueue_top_rt_rq+0xf4/0x150
? cpufreq_dbs_governor_start+0x170/0x170
sched_rt_rq_enqueue+0x65/0x80
sched_rt_period_timer+0x156/0x360
? sched_rt_rq_enqueue+0x80/0x80
__hrtimer_run_queues+0xfa/0x260
hrtimer_interrupt+0xcb/0x220
smp_apic_timer_interrupt+0x62/0x120
apic_timer_interrupt+0xf/0x20
</IRQ>
[...]
do_idle+0x183/0x1e0
cpu_startup_entry+0x5f/0x70
start_secondary+0x192/0x1d0
secondary_startup_64+0xa5/0xb0
We can get rid of it be the "traditional" means of adding an
update_rq_clock() call after acquiring the rq->lock in
do_sched_rt_period_timer().
The case for the RT task throttling (which this workload also hits)
can be ignored in that the skip_update call is actually bogus and
quite the contrary (the request bits are removed/reverted).
By setting RQCF_UPDATED we really don't care if the skip is happening
or not and will therefore make the assert_clock_updated() check happy.
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Matt Fleming <[email protected]>
Acked-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
No changes in refcount semantics -- key init is false; replace
static_key_slow_inc|dec with static_branch_inc|dec
static_key_false with static_branch_unlikely
Added a '_key' suffix to i2c_trace_msg, for better self
documentation.
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: David Howells <[email protected]>
Signed-off-by: Wolfram Sang <[email protected]>
No changes in semantics -- key init is true; replace
static_key_slow_dec with static_branch_disable
static_key_true with static_branch_likely
The first is because we never actually do any couterpart incs,
thus there is really no reference counting semantics going on.
Use the more proper static_branch_disable() construct.
Also added a '_key' suffix to supports_deactivate, for better
self documentation.
Cc: Thomas Gleixner <[email protected]>
Cc: Jason Cooper <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Marc Zyngier <[email protected]>
At the point of sysfs callback, the call to gup is
done without mmap_sem (or any lock for that matter).
This is racy. As such, use the get_user_pages_fast()
alternative and safely avoid taking the lock, if possible.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Tony Luck <[email protected]>
get_user_pages() must be called with mmap_sem held, currently
it is not. In fact it is called under the user db_table->mutex.
To fix this we can convert gup to use the fast alternative,
and safely avoid taking mmap_sem, if possible. Furthermore
this is safe wrt to the mutex as other callers that take the
lock (unmap and alloc_db) are not called under mmap_sem
(hence possible deadlock).
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>
Things can explode for locktorture if the user does combinations
of nwriters_stress=0 nreaders_stress=0. Fix this by not assuming
we always want to torture writer threads.
Reported-by: Jeremy Linton <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Reviewed-by: Jeremy Linton <[email protected]>
Tested-by: Jeremy Linton <[email protected]>
We should account for nreader threads, not writers in this
callback. Could even trigger a div by 0 if the user explicitly
disables writers.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
It was reported that the whole futex bench breaks when dealing with
non-contiguously numbered cpus.
$ echo 0 | sudo tee /sys/devices/system/cpu/cpu3/online
$ ./perf bench futex all
perf: pthread_create: Operation not permitted
Run summary [PID 14934]: 7 threads, each ....
James had implemented an approach with cpumaps that use an in house
flavor. Instead of re-inventing the wheel, I've redone the patch such
that we use the perf's util/cpumap.c interface instead.
Applies to all futex benchmarks.
Suggested-by: Arnaldo Carvalho de Melo <[email protected]>
Originally-from: James Yang <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Davidlohr Bueso <[email protected]>
Cc: Kim Phillips <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
For a custom microbenchmark on a 3.30GHz Xeon SandyBridge, which calls
IPC_STAT over and over, it was calculated that, on avg the cost of
ipc_get_maxid() for increasing amounts of keys was:
10 keys: ~900 cycles
100 keys: ~15000 cycles
1000 keys: ~150000 cycles
10000 keys: ~2100000 cycles
This is unsurprising as maxid is currently O(n).
By having the max_id available in O(1) we save all those cycles for each
semctl(_STAT) command, the idr_find can be expensive -- which some real
(customer) workloads actually poll on.
Note that this used to be the case, until commit 7ca7e564e0 ("ipc:
store ipcs into IDRs"). The cost is the extra idr_find when doing
RMIDs, but we simply go backwards, and should not take too many
iterations to find the new value.
[[email protected]: coding-style fixes]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Manfred Spraul <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Patch series "sysvipc: ipc-key management improvements".
Here are a few improvements I spotted while eyeballing Guillaume's
rhashtable implementation for ipc keys. The first and fourth patches
are the interesting ones, the middle two are trivial.
This patch (of 4):
The next_id object-allocation functionality was introduced in commit
03f5956680 ("ipc: add sysctl to specify desired next object id").
Given that these new entries are _only_ exported under the
CONFIG_CHECKPOINT_RESTORE option, there is no point for the common case
to even know about ->next_id. As such rewrite ipc_buildid() such that
it can do away with the field as well as unnecessary branches when
adding a new identifier. The end result also better differentiates both
cases, so the code ends up being cleaner; albeit the small duplications
regarding the default case.
[[email protected]: coding-style fixes]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Manfred Spraul <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Fengguang reported soft lockups while running the rbtree and interval
tree test modules. The logic for these tests all occur in init phase,
and we currently are pounding with the default values for number of
nodes and number of iterations of each test. Reduce the latter by two
orders of magnitude. This does not influence the value of the tests in
that one thousand times by default is enough to get the picture.
Link: http://lkml.kernel.org/r/20171109161715.xai2dtwqw2frhkcm@linux-n805
Signed-off-by: Davidlohr Bueso <[email protected]>
Reported-by: Fengguang Wu <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Andrea brought to my attention that the L->{L,S} guarantees are
completely bogus for this case. I was looking at the diagram, from the
offending commit, when that _is_ the race, we had the load reordered
already.
What we need is at least S->L semantics, thus simply use
wq_has_sleeper() to serialize the call for good.
Link: http://lkml.kernel.org/r/[email protected]
Fixes: 46acef048a (mm,compaction: serialize waitqueue_active() checks)
Signed-off-by: Davidlohr Bueso <[email protected]>
Reported-by: Andrea Parri <[email protected]>
Cc: Vlastimil Babka <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
For example, the following could occur, making us miss a wakeup:
CPU0 CPU1
kvm_vcpu_block kvm_mips_comparecount_func
[L] swait_active(&vcpu->wq)
[S] prepare_to_swait(&vcpu->wq)
[L] if (!kvm_vcpu_has_pending_timer(vcpu))
schedule() [S] queue_timer_int(vcpu)
Ensure that the swait_active() check is not hoisted over the interrupt.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>
Particularly because kvmppc_fast_vcpu_kick_hv() is a callback,
ensure that we properly serialize wq active checks in order to
avoid potentially missing a wakeup due to racing with the waiter
side.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>
Davidlohr Bueso