syzbot is reporting that splice()ing from non-empty read side to
already-full write side causes unkillable task, for opipe_prep() is by
error not inverting pipe_full() test.
CPU: 0 PID: 9460 Comm: syz-executor.5 Not tainted 5.6.0-rc3-next-20200228-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
RIP: 0010:rol32 include/linux/bitops.h:105 [inline]
RIP: 0010:iterate_chain_key kernel/locking/lockdep.c:369 [inline]
RIP: 0010:__lock_acquire+0x6a3/0x5270 kernel/locking/lockdep.c:4178
Call Trace:
lock_acquire+0x197/0x420 kernel/locking/lockdep.c:4720
__mutex_lock_common kernel/locking/mutex.c:956 [inline]
__mutex_lock+0x156/0x13c0 kernel/locking/mutex.c:1103
pipe_lock_nested fs/pipe.c:66 [inline]
pipe_double_lock+0x1a0/0x1e0 fs/pipe.c:104
splice_pipe_to_pipe fs/splice.c:1562 [inline]
do_splice+0x35f/0x1520 fs/splice.c:1141
__do_sys_splice fs/splice.c:1447 [inline]
__se_sys_splice fs/splice.c:1427 [inline]
__x64_sys_splice+0x2b5/0x320 fs/splice.c:1427
do_syscall_64+0xf6/0x790 arch/x86/entry/common.c:295
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Reported-by: syzbot+b48daca8639150bc5e73@syzkaller.appspotmail.com
Link: https://syzkaller.appspot.com/bug?id=9386d051e11e09973d5a4cf79af5e8cedf79386d
Fixes: 8cefc107ca54c8b0 ("pipe: Use head and tail pointers for the ring, not cursor and length")
Cc: stable@vger.kernel.org # 5.5+
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
do_splice() is used by io_uring, as will be do_tee(). Move f_mode
checks from sys_{splice,tee}() to do_{splice,tee}(), so they're
enforced for io_uring as well.
Fixes: 7d67af2c0134 ("io_uring: add splice(2) support")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Make do_splice(), so other kernel parts can reuse it Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
This makes the pipe code use separate wait-queues and exclusive waiting
for readers and writers, avoiding a nasty thundering herd problem when
there are lots of readers waiting for data on a pipe (or, less commonly,
lots of writers waiting for a pipe to have space).
While this isn't a common occurrence in the traditional "use a pipe as a
data transport" case, where you typically only have a single reader and
a single writer process, there is one common special case: using a pipe
as a source of "locking tokens" rather than for data communication.
In particular, the GNU make jobserver code ends up using a pipe as a way
to limit parallelism, where each job consumes a token by reading a byte
from the jobserver pipe, and releases the token by writing a byte back
to the pipe.
This pattern is fairly traditional on Unix, and works very well, but
will waste a lot of time waking up a lot of processes when only a single
reader needs to be woken up when a writer releases a new token.
A simplified test-case of just this pipe interaction is to create 64
processes, and then pass a single token around between them (this
test-case also intentionally passes another token that gets ignored to
test the "wake up next" logic too, in case anybody wonders about it):
#include <unistd.h>
int main(int argc, char **argv)
{
int fd[2], counters[2];
pipe(fd);
counters[0] = 0;
counters[1] = -1;
write(fd[1], counters, sizeof(counters));
/* 64 processes */
fork(); fork(); fork(); fork(); fork(); fork();
do {
int i;
read(fd[0], &i, sizeof(i));
if (i < 0)
continue;
counters[0] = i+1;
write(fd[1], counters, (1+(i & 1)) *sizeof(int));
} while (counters[0] < 1000000);
return 0;
}
and in a perfect world, passing that token around should only cause one
context switch per transfer, when the writer of a token causes a
directed wakeup of just a single reader.
But with the "writer wakes all readers" model we traditionally had, on
my test box the above case causes more than an order of magnitude more
scheduling: instead of the expected ~1M context switches, "perf stat"
shows
231,852.37 msec task-clock # 15.857 CPUs utilized
11,250,961 context-switches # 0.049 M/sec
616,304 cpu-migrations # 0.003 M/sec
1,648 page-faults # 0.007 K/sec
1,097,903,998,514 cycles # 4.735 GHz
120,781,778,352 instructions # 0.11 insn per cycle
27,997,056,043 branches # 120.754 M/sec
283,581,233 branch-misses # 1.01% of all branches
14.621273891 seconds time elapsed
0.018243000 seconds user
3.611468000 seconds sys
before this commit.
After this commit, I get
5,229.55 msec task-clock # 3.072 CPUs utilized
1,212,233 context-switches # 0.232 M/sec
103,951 cpu-migrations # 0.020 M/sec
1,328 page-faults # 0.254 K/sec
21,307,456,166 cycles # 4.074 GHz
12,947,819,999 instructions # 0.61 insn per cycle
2,881,985,678 branches # 551.096 M/sec
64,267,015 branch-misses # 2.23% of all branches
1.702148350 seconds time elapsed
0.004868000 seconds user
0.110786000 seconds sys
instead. Much better.
[ Note! This kernel improvement seems to be very good at triggering a
race condition in the make jobserver (in GNU make 4.2.1) for me. It's
a long known bug that was fixed back in June 2017 by GNU make commit
b552b0525198 ("[SV 51159] Use a non-blocking read with pselect to
avoid hangs.").
But there wasn't a new release of GNU make until 4.3 on Jan 19 2020,
so a number of distributions may still have the buggy version. Some
have backported the fix to their 4.2.1 release, though, and even
without the fix it's quite timing-dependent whether the bug actually
is hit. ]
Josh Triplett says:
"I've been hammering on your pipe fix patch (switching to exclusive
wait queues) for a month or so, on several different systems, and I've
run into no issues with it. The patch *substantially* improves
parallel build times on large (~100 CPU) systems, both with parallel
make and with other things that use make's pipe-based jobserver.
All current distributions (including stable and long-term stable
distributions) have versions of GNU make that no longer have the
jobserver bug"
Tested-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This code is ancient, and goes back to when we only had a single page for the pipe buffers. The exact history is hidden in the mists of time (ie "before git", and in fact predates the BK repository too). At that long-ago point in time, it actually helped to try to merge big back-and-forth pipe reads and writes, and not limit pipe reads to the single pipe buffer in length just because that was all we had at a time. However, since then we've expanded the pipe buffers to multiple pages, and this logic really doesn't seem to make sense. And a lot of it is somewhat questionable (ie "hmm, the user asked for a non-blocking read, but we see that there's a writer pending, so let's wait anyway to get the extra data that the writer will have"). But more importantly, it makes the "go to sleep" logic much less obvious, and considering the wakeup issues we've had, I want to make for less of those kinds of things. Cc: David Howells <dhowells@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Similarly to commit 8f868d68d335 ("pipe: Fix missing mask update after
pipe_wait()") this fixes a case where the pipe rewrite ended up caching
the pipe state incorrectly over a pipe lock drop event.
It wasn't quite as obvious, because you needed to splice data from a
pipe to a file, which is a fairly unusual operation, but it's completely
wrong.
Make sure we load the pipe head/tail/size information only after we've
waited for there to be data in the pipe.
While in that file, also make one of the splice helper functions use the
canonical arghument order for pipe_empty(). That's syntactic - pipe
emptiness is just that head and tail are equal, and thus mixing up head
and tail doesn't really matter. It's still wrong, though.
Reported-by: David Sterba <dsterba@suse.cz>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull pipe rework from David Howells:
"This is my set of preparatory patches for building a general
notification queue on top of pipes. It makes a number of significant
changes:
- It removes the nr_exclusive argument from __wake_up_sync_key() as
this is always 1. This prepares for the next step:
- Adds wake_up_interruptible_sync_poll_locked() so that poll can be
woken up from a function that's holding the poll waitqueue
spinlock.
- Change the pipe buffer ring to be managed in terms of unbounded
head and tail indices rather than bounded index and length. This
means that reading the pipe only needs to modify one index, not
two.
- A selection of helper functions are provided to query the state of
the pipe buffer, plus a couple to apply updates to the pipe
indices.
- The pipe ring is allowed to have kernel-reserved slots. This allows
many notification messages to be spliced in by the kernel without
allowing userspace to pin too many pages if it writes to the same
pipe.
- Advance the head and tail indices inside the pipe waitqueue lock
and use wake_up_interruptible_sync_poll_locked() to poke poll
without having to take the lock twice.
- Rearrange pipe_write() to preallocate the buffer it is going to
write into and then drop the spinlock. This allows kernel
notifications to then be added the ring whilst it is filling the
buffer it allocated. The read side is stalled because the pipe
mutex is still held.
- Don't wake up readers on a pipe if there was already data in it
when we added more.
- Don't wake up writers on a pipe if the ring wasn't full before we
removed a buffer"
* tag 'notifications-pipe-prep-20191115' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs:
pipe: Remove sync on wake_ups
pipe: Increase the writer-wakeup threshold to reduce context-switch count
pipe: Check for ring full inside of the spinlock in pipe_write()
pipe: Remove redundant wakeup from pipe_write()
pipe: Rearrange sequence in pipe_write() to preallocate slot
pipe: Conditionalise wakeup in pipe_read()
pipe: Advance tail pointer inside of wait spinlock in pipe_read()
pipe: Allow pipes to have kernel-reserved slots
pipe: Use head and tail pointers for the ring, not cursor and length
Add wake_up_interruptible_sync_poll_locked()
Remove the nr_exclusive argument from __wake_up_sync_key()
pipe: Reduce #inclusion of pipe_fs_i.h
Split pipe->ring_size into two numbers:
(1) pipe->ring_size - indicates the hard size of the pipe ring.
(2) pipe->max_usage - indicates the maximum number of pipe ring slots that
userspace orchestrated events can fill.
This allows for a pipe that is both writable by the general kernel
notification facility and by userspace, allowing plenty of ring space for
notifications to be added whilst preventing userspace from being able to
pin too much unswappable kernel space.
Signed-off-by: David Howells <dhowells@redhat.com>
Convert pipes to use head and tail pointers for the buffer ring rather than
pointer and length as the latter requires two atomic ops to update (or a
combined op) whereas the former only requires one.
(1) The head pointer is the point at which production occurs and points to
the slot in which the next buffer will be placed. This is equivalent
to pipe->curbuf + pipe->nrbufs.
The head pointer belongs to the write-side.
(2) The tail pointer is the point at which consumption occurs. It points
to the next slot to be consumed. This is equivalent to pipe->curbuf.
The tail pointer belongs to the read-side.
(3) head and tail are allowed to run to UINT_MAX and wrap naturally. They
are only masked off when the array is being accessed, e.g.:
pipe->bufs[head & mask]
This means that it is not necessary to have a dead slot in the ring as
head == tail isn't ambiguous.
(4) The ring is empty if "head == tail".
A helper, pipe_empty(), is provided for this.
(5) The occupancy of the ring is "head - tail".
A helper, pipe_occupancy(), is provided for this.
(6) The number of free slots in the ring is "pipe->ring_size - occupancy".
A helper, pipe_space_for_user() is provided to indicate how many slots
userspace may use.
(7) The ring is full if "head - tail >= pipe->ring_size".
A helper, pipe_full(), is provided for this.
Signed-off-by: David Howells <dhowells@redhat.com>
Andreas Grünbacher reports that on the two filesystems that support
iomap directio, it's possible for splice() to return -EAGAIN (instead of
a short splice) if the pipe being written to has less space available in
its pipe buffers than the length supplied by the calling process.
Months ago we fixed splice_direct_to_actor to clamp the length of the
read request to the size of the splice pipe. Do the same to do_splice.
Fixes: 17614445576b6 ("splice: don't read more than available pipe space")
Reported-by: syzbot+3c01db6025f26530cf8d@syzkaller.appspotmail.com
Reported-by: Andreas Grünbacher <andreas.gruenbacher@gmail.com>
Reviewed-by: Andreas Grünbacher <andreas.gruenbacher@gmail.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>