Rename (__)memcg_kmem_(un)charge() into (__)memcg_kmem_(un)charge_page()
to better reflect what they are actually doing:
1) call __memcg_kmem_(un)charge_memcg() to actually charge or uncharge
the current memcg
2) set or clear the PageKmemcg flag
Signed-off-by: Roman Gushchin <guro@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Link: http://lkml.kernel.org/r/20200109202659.752357-4-guro@fb.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Andrei Vagin reported that commit 0ddad21d3e ("pipe: use exclusive
waits when reading or writing") broke one of the CRIU tests. He even
has a trivial reproducer:
#include <unistd.h>
#include <sys/types.h>
#include <sys/wait.h>
int main()
{
int p[2];
pid_t p1, p2;
int status;
if (pipe(p) == -1)
return 1;
p1 = fork();
if (p1 == 0) {
close(p[1]);
read(p[0], &status, sizeof(status));
return 0;
}
p2 = fork();
if (p2 == 0) {
close(p[1]);
read(p[0], &status, sizeof(status));
return 0;
}
sleep(1);
close(p[1]);
wait(&status);
wait(&status);
return 0;
}
and the problem - once he points it out - is obvious. We use these nice
exclusive waits, but when the last writer goes away, it then needs to
wake up _every_ reader (and conversely, the last reader disappearing
needs to wake every writer, of course).
In fact, when going through this, we had several small oddities around
how to wake things. We did in fact wake every reader when we changed
the size of the pipe buffers. But that's entirely pointless, since that
just acts as a possible source of new space - no new data to read.
And when we change the size of the buffer, we don't need to wake all
writers even when we add space - that case acts just as if somebody made
space by reading, and any writer that finds itself not filling it up
entirely will wake the next one.
On the other hand, on the exit path, we tried to limit the wakeups with
the proper poll keys etc, which is entirely pointless, because at that
point we obviously need to wake up everybody. So don't do that: just
wake up everybody - but only do that if the counts changed to zero.
So fix those non-IO wakeups to be more proper: space change doesn't add
any new data, but it might make room for writers, so it wakes up a
writer. And the actual changes to reader/writer counts should wake up
everybody, since everybody is affected (ie readers will all see EOF if
the writers have gone away, and writers will all get EPIPE if all
readers have gone away).
Fixes: 0ddad21d3e ("pipe: use exclusive waits when reading or writing")
Reported-and-tested-by: Andrei Vagin <avagin@gmail.com>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
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>
LTP pipeio_1 test is hanging with v5.5-rc2-385-gb8e382a185eb,
with read side observing empty pipe and sleeping and write
side running out of space and then sleeping as well. In this
scenario there are 5 writers and 1 reader.
Problem is that after pipe_write() reacquires pipe lock, it
re-checks for empty pipe with potentially stale 'head' and
doesn't wake up read side anymore. pipe->tail can advance
beyond 'head', because there are multiple writers.
Use pipe->head for empty pipe check after reacquiring lock
to observe current state.
Testing: With patch, LTP pipeio_1 ran successfully in loop for 1 hour.
Without patch it hanged within a minute.
Fixes: 1b6b26ae70 ("pipe: fix and clarify pipe write wakeup logic")
Reported-by: Rachel Sibley <rasibley@redhat.com>
Signed-off-by: Jan Stancek <jstancek@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There's no need to separately check for signals while inside the locked
region, since we're going to do "wait_event_interruptible()" right
afterwards anyway, and the error handling is much simpler there.
The check for whether we had already read anything was also redundant,
since we no longer do the odd merging of reads when there are pending
writers.
But perhaps more importantly, this adds commentary about why we still
need to wake up possible writers even though we didn't read any data,
and why we can skip all the finishing touches now if we get a signal (or
had a signal pending) while waiting for more data.
[ This is a split-out cleanup from my "make pipe IO use exclusive wait
queues" thing, which I can't apply because it triggers a nasty bug in
the GNU make jobserver - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
pipe_wait() may be simple, but since it relies on the pipe lock, it
means that we have to do the wakeup while holding the lock. That's
unfortunate, because the very first thing the waked entity will want to
do is to get the pipe lock for itself.
So get rid of the pipe_wait() usage by simply releasing the pipe lock,
doing the wakeup (if required) and then using wait_event_interruptible()
to wait on the right condition instead.
wait_event_interruptible() handles races on its own by comparing the
wakeup condition before and after adding itself to the wait queue, so
you can use an optimistic unlocked condition for it.
Cc: David Howells <dhowells@redhat.com>
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>
This is the read side version of the previous commit: it simplifies the
logic to only wake up waiting writers when necessary, and makes sure to
use a synchronous wakeup. This time not so much for GNU make jobserver
reasons (that pipe never fills up), but simply to get the writer going
quickly again.
A bit less verbose commentary this time, if only because I assume that
the write side commentary isn't going to be ignored if you touch this
code.
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The pipe rework ends up having been extra painful, partly becaused of
actual bugs with ordering and caching of the pipe state, but also
because of subtle performance issues.
In particular, the pipe rework caused the kernel build to inexplicably
slow down.
The reason turns out to be that the GNU make jobserver (which limits the
parallelism of the build) uses a pipe to implement a "token" system: a
parallel submake will read a character from the pipe to get the job
token before starting a new job, and will write a character back to the
pipe when it is done. The overall job limit is thus easily controlled
by just writing the appropriate number of initial token characters into
the pipe.
But to work well, that really means that the old behavior of write
wakeups being synchronous (WF_SYNC) is very important - when the pipe
writer wakes up a reader, we want the reader to actually get scheduled
immediately. Otherwise you lose the parallelism of the build.
The pipe rework lost that synchronous wakeup on write, and we had
clearly all forgotten the reasons and rules for it.
This rewrites the pipe write wakeup logic to do the required Wsync
wakeups, but also clarifies the logic and avoids extraneous wakeups.
It also ends up addign a number of comments about what oit does and why,
so that we hopefully don't end up forgetting about this next time we
change this code.
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The kernel wait queues have a basic rule to them: you add yourself to
the wait-queue first, and then you check the things that you're going to
wait on. That avoids the races with the event you're waiting for.
The same goes for poll/select logic: the "poll_wait()" goes first, and
then you check the things you're polling for.
Of course, if you use locking, the ordering doesn't matter since the
lock will serialize with anything that changes the state you're looking
at. That's not the case here, though.
So move the poll_wait() first in pipe_poll(), before you start looking
at the pipe state.
Fixes: 8cefc107ca ("pipe: Use head and tail pointers for the ring, not cursor and length")
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Merge two fixes for the pipe rework from David Howells:
"Here are a couple of patches to fix bugs syzbot found in the pipe
changes:
- An assertion check will sometimes trip when polling a pipe because
the ring size and indices used are approximate and may be being
changed simultaneously.
An equivalent approximate calculation was done previously, but
without the assertion check, so I've just dropped the check. To
make it accurate, the pipe mutex would need to be taken or the spin
lock could be used - but usage of the spinlock would need to be
rolled out into splice, iov_iter and other places for that.
- The index mask and the max_usage values cannot be cached across
pipe_wait() as F_SETPIPE_SZ could have been called during the wait.
This can cause pipe_write() to break"
* pipe-rework:
pipe: Fix missing mask update after pipe_wait()
pipe: Remove assertion from pipe_poll()
Fix pipe_write() to not cache the ring index mask and max_usage as their
values are invalidated by calling pipe_wait() because the latter
function drops the pipe lock, thereby allowing F_SETPIPE_SZ change them.
Without this, pipe_write() may subsequently miscalculate the array
indices and pipe fullness, leading to an oops like the following:
BUG: KASAN: slab-out-of-bounds in pipe_write+0xc25/0xe10 fs/pipe.c:481
Write of size 8 at addr ffff8880771167a8 by task syz-executor.3/7987
...
CPU: 1 PID: 7987 Comm: syz-executor.3 Not tainted 5.4.0-rc2-syzkaller #0
...
Call Trace:
pipe_write+0xc25/0xe10 fs/pipe.c:481
call_write_iter include/linux/fs.h:1895 [inline]
new_sync_write+0x3fd/0x7e0 fs/read_write.c:483
__vfs_write+0x94/0x110 fs/read_write.c:496
vfs_write+0x18a/0x520 fs/read_write.c:558
ksys_write+0x105/0x220 fs/read_write.c:611
__do_sys_write fs/read_write.c:623 [inline]
__se_sys_write fs/read_write.c:620 [inline]
__x64_sys_write+0x6e/0xb0 fs/read_write.c:620
do_syscall_64+0xca/0x5d0 arch/x86/entry/common.c:290
entry_SYSCALL_64_after_hwframe+0x49/0xbe
This is not a problem for pipe_read() as the mask is recalculated on
each pass of the loop, after pipe_wait() has been called.
Fixes: 8cefc107ca ("pipe: Use head and tail pointers for the ring, not cursor and length")
Reported-by: syzbot+838eb0878ffd51f27c41@syzkaller.appspotmail.com
Signed-off-by: David Howells <dhowells@redhat.com>
Cc: Eric Biggers <ebiggers@kernel.org>
[ Changed it to use a temporary variable 'mask' to avoid long lines -Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
An assertion check was added to pipe_poll() to make sure that the ring
occupancy isn't seen to overflow the ring size. However, since no locks
are held when the three values are read, it is possible for F_SETPIPE_SZ
to intervene and muck up the calculation, thereby causing the oops.
Fix this by simply removing the assertion and accepting that the
calculation might be approximate.
Note that the previous code also had a similar issue, though there was
no assertion check, since the occupancy counter and the ring size were
not read with a lock held, so it's possible that the poll check might
have malfunctioned then too.
Also wake up all the waiters so that they can reissue their checks if
there was a competing read or write.
Fixes: 8cefc107ca ("pipe: Use head and tail pointers for the ring, not cursor and length")
Reported-by: syzbot+d37abaade33a934f16f2@syzkaller.appspotmail.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Eric Biggers <ebiggers@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
-----BEGIN PGP SIGNATURE-----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=FKC7
-----END PGP SIGNATURE-----
Merge tag 'notifications-pipe-prep-20191115' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs
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
In commit 3975b097e5 ("convert stream-like files -> stream_open, even
if they use noop_llseek") Kirill used a coccinelle script to change
"nonseekable_open()" to "stream_open()", which changed the trivial cases
of stream-like file descriptors to the new model with FMODE_STREAM.
However, the two big cases - sockets and pipes - don't actually have
that trivial pattern at all, and were thus never converted to
FMODE_STREAM even though it makes lots of sense to do so.
That's particularly true when looking forward to the next change:
getting rid of FMODE_ATOMIC_POS entirely, and just using FMODE_STREAM to
decide whether f_pos updates are needed or not. And if they are, we'll
always do them atomically.
This came up because KCSAN (correctly) noted that the non-locked f_pos
updates are data races: they are clearly benign for the case where we
don't care, but it would be good to just not have that issue exist at
all.
Note that the reason we used FMODE_ATOMIC_POS originally is that only
doing it for the minimal required case is "safer" in that it's possible
that the f_pos locking can cause unnecessary serialization across the
whole write() call. And in the worst case, that kind of serialization
can cause deadlock issues: think writers that need readers to empty the
state using the same file descriptor.
[ Note that the locking is per-file descriptor - because it protects
"f_pos", which is obviously per-file descriptor - so it only affects
cases where you literally use the same file descriptor to both read
and write.
So a regular pipe that has separate reading and writing file
descriptors doesn't really have this situation even though it's the
obvious case of "reader empties what a bit writer concurrently fills"
But we want to make pipes as being stream-line anyway, because we
don't want the unnecessary overhead of locking, and because a named
pipe can be (ab-)used by reading and writing to the same file
descriptor. ]
There are likely a lot of other cases that might want FMODE_STREAM, and
looking for ".llseek = no_llseek" users and other cases that don't have
an lseek file operation at all and making them use "stream_open()" might
be a good idea. But pipes and sockets are likely to be the two main
cases.
Cc: Kirill Smelkov <kirr@nexedi.com>
Cc: Eic Dumazet <edumazet@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: Marco Elver <elver@google.com>
Cc: Andrea Parri <parri.andrea@gmail.com>
Cc: Paul McKenney <paulmck@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Increase the threshold at which the reader sends a wake event to the
writers in the queue such that the queue must be half empty before the wake
is issued rather than the wake being issued when just a single slot
available.
This reduces the number of context switches in the tests significantly,
without altering the amount of work achieved. With my pipe-bench program,
there's a 20% reduction versus an unpatched kernel.
Suggested-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: David Howells <dhowells@redhat.com>
Make pipe_write() check to see if the ring has become full between it
taking the pipe mutex, checking the ring status and then taking the
spinlock.
This can happen if a notification is written into the pipe as that happens
without the pipe mutex.
Signed-off-by: David Howells <dhowells@redhat.com>
Rearrange the sequence in pipe_write() so that the allocation of the new
buffer, the allocation of a ring slot and the attachment to the ring is
done under the pipe wait spinlock and then the lock is dropped and the
buffer can be filled.
The data copy needs to be done with the spinlock unheld and irqs enabled,
so the lock needs to be dropped first. However, the reader can't progress
as we're holding pipe->mutex.
We also need to drop the lock as that would impact others looking at the
pipe waitqueue, such as poll(), the consumer and a future kernel message
writer.
We just abandon the preallocated slot if we get a copy error. Future
writes may continue it and a future read will eventually recycle it.
Signed-off-by: David Howells <dhowells@redhat.com>
Only do a wakeup in pipe_read() if we made space in a completely full
buffer. The producer shouldn't be waiting on pipe->wait otherwise.
Signed-off-by: David Howells <dhowells@redhat.com>
Advance the pipe ring tail pointer inside of wait spinlock in pipe_read()
so that the pipe can be written into with kernel notifications from
contexts where pipe->mutex cannot be taken.
Signed-off-by: David Howells <dhowells@redhat.com>
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>
Convert the pipe filesystem to the new internal mount API as the old
one will be obsoleted and removed. This allows greater flexibility in
communication of mount parameters between userspace, the VFS and the
filesystem.
See Documentation/filesystems/mount_api.txt for more information.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Once upon a time we used to set ->d_name of e.g. pipefs root
so that d_path() on pipes would work. These days it's
completely pointless - dentries of pipes are not even connected
to pipefs root. However, mount_pseudo() had set the root
dentry name (passed as the second argument) and callers
kept inventing names to pass to it. Including those that
didn't *have* any non-root dentries to start with...
All of that had been pointless for about 8 years now; it's
time to get rid of that cargo-culting...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Merge page ref overflow branch.
Jann Horn reported that he can overflow the page ref count with
sufficient memory (and a filesystem that is intentionally extremely
slow).
Admittedly it's not exactly easy. To have more than four billion
references to a page requires a minimum of 32GB of kernel memory just
for the pointers to the pages, much less any metadata to keep track of
those pointers. Jann needed a total of 140GB of memory and a specially
crafted filesystem that leaves all reads pending (in order to not ever
free the page references and just keep adding more).
Still, we have a fairly straightforward way to limit the two obvious
user-controllable sources of page references: direct-IO like page
references gotten through get_user_pages(), and the splice pipe page
duplication. So let's just do that.
* branch page-refs:
fs: prevent page refcount overflow in pipe_buf_get
mm: prevent get_user_pages() from overflowing page refcount
mm: add 'try_get_page()' helper function
mm: make page ref count overflow check tighter and more explicit
Change pipe_buf_get() to return a bool indicating whether it succeeded
in raising the refcount of the page (if the thing in the pipe is a page).
This removes another mechanism for overflowing the page refcount. All
callers converted to handle a failure.
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Matthew Wilcox <willy@infradead.org>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull misc vfs updates from Al Viro:
"Assorted fixes (really no common topic here)"
* 'work.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
vfs: Make __vfs_write() static
vfs: fix preadv64v2 and pwritev64v2 compat syscalls with offset == -1
pipe: stop using ->can_merge
splice: don't merge into linked buffers
fs: move generic stat response attr handling to vfs_getattr_nosec
orangefs: don't reinitialize result_mask in ->getattr
fs/devpts: always delete dcache dentry-s in dput()
Move the memcg_kmem_enabled() checks into memcg kmem charge/uncharge
functions, so, the users don't have to explicitly check that condition.
This is purely code cleanup patch without any functional change. Only
the order of checks in memcg_charge_slab() can potentially be changed
but the functionally it will be same. This should not matter as
memcg_charge_slab() is not in the hot path.
Link: http://lkml.kernel.org/r/20190103161203.162375-1-shakeelb@google.com
Signed-off-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Roman Gushchin <guro@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Al Viro pointed out that since there is only one pipe buffer type to which
new data can be appended, it isn't necessary to have a ->can_merge field in
struct pipe_buf_operations, we can just check for a magic type.
Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Before this patch, it was possible for two pipes to affect each other after
data had been transferred between them with tee():
============
$ cat tee_test.c
int main(void) {
int pipe_a[2];
if (pipe(pipe_a)) err(1, "pipe");
int pipe_b[2];
if (pipe(pipe_b)) err(1, "pipe");
if (write(pipe_a[1], "abcd", 4) != 4) err(1, "write");
if (tee(pipe_a[0], pipe_b[1], 2, 0) != 2) err(1, "tee");
if (write(pipe_b[1], "xx", 2) != 2) err(1, "write");
char buf[5];
if (read(pipe_a[0], buf, 4) != 4) err(1, "read");
buf[4] = 0;
printf("got back: '%s'\n", buf);
}
$ gcc -o tee_test tee_test.c
$ ./tee_test
got back: 'abxx'
$
============
As suggested by Al Viro, fix it by creating a separate type for
non-mergeable pipe buffers, then changing the types of buffers in
splice_pipe_to_pipe() and link_pipe().
Cc: <stable@vger.kernel.org>
Fixes: 7c77f0b3f9 ("splice: implement pipe to pipe splicing")
Fixes: 70524490ee ("[PATCH] splice: add support for sys_tee()")
Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull vfs open-related updates from Al Viro:
- "do we need fput() or put_filp()" rules are gone - it's always fput()
now. We keep track of that state where it belongs - in ->f_mode.
- int *opened mess killed - in finish_open(), in ->atomic_open()
instances and in fs/namei.c code around do_last()/lookup_open()/atomic_open().
- alloc_file() wrappers with saner calling conventions are introduced
(alloc_file_clone() and alloc_file_pseudo()); callers converted, with
much simplification.
- while we are at it, saner calling conventions for path_init() and
link_path_walk(), simplifying things inside fs/namei.c (both on
open-related paths and elsewhere).
* 'work.open3' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (40 commits)
few more cleanups of link_path_walk() callers
allow link_path_walk() to take ERR_PTR()
make path_init() unconditionally paired with terminate_walk()
document alloc_file() changes
make alloc_file() static
do_shmat(): grab shp->shm_file earlier, switch to alloc_file_clone()
new helper: alloc_file_clone()
create_pipe_files(): switch the first allocation to alloc_file_pseudo()
anon_inode_getfile(): switch to alloc_file_pseudo()
hugetlb_file_setup(): switch to alloc_file_pseudo()
ocxlflash_getfile(): switch to alloc_file_pseudo()
cxl_getfile(): switch to alloc_file_pseudo()
... and switch shmem_file_setup() to alloc_file_pseudo()
__shmem_file_setup(): reorder allocations
new wrapper: alloc_file_pseudo()
kill FILE_{CREATED,OPENED}
switch atomic_open() and lookup_open() to returning 0 in all success cases
document ->atomic_open() changes
->atomic_open(): return 0 in all success cases
get rid of 'opened' in path_openat() and the helpers downstream
...
alloc_file_clone(old_file, mode, ops): create a new struct file with
->f_path equal to that of old_file. pipe converted.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... so that it could set both ->f_flags and ->f_mode, without callers
having to set ->f_flags manually.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... just use put_pipe_info() to get the pipe->files down to 1 and let
fput()-called pipe_release() do freeing.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The poll() changes were not well thought out, and completely
unexplained. They also caused a huge performance regression, because
"->poll()" was no longer a trivial file operation that just called down
to the underlying file operations, but instead did at least two indirect
calls.
Indirect calls are sadly slow now with the Spectre mitigation, but the
performance problem could at least be largely mitigated by changing the
"->get_poll_head()" operation to just have a per-file-descriptor pointer
to the poll head instead. That gets rid of one of the new indirections.
But that doesn't fix the new complexity that is completely unwarranted
for the regular case. The (undocumented) reason for the poll() changes
was some alleged AIO poll race fixing, but we don't make the common case
slower and more complex for some uncommon special case, so this all
really needs way more explanations and most likely a fundamental
redesign.
[ This revert is a revert of about 30 different commits, not reverted
individually because that would just be unnecessarily messy - Linus ]
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Using this helper removes an in-kernel call to the sys_pipe2() syscall.
This patch is part of a series which removes in-kernel calls to syscalls.
On this basis, the syscall entry path can be streamlined. For details, see
http://lkml.kernel.org/r/20180325162527.GA17492@light.dominikbrodowski.net
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
This is the mindless scripted replacement of kernel use of POLL*
variables as described by Al, done by this script:
for V in IN OUT PRI ERR RDNORM RDBAND WRNORM WRBAND HUP RDHUP NVAL MSG; do
L=`git grep -l -w POLL$V | grep -v '^t' | grep -v /um/ | grep -v '^sa' | grep -v '/poll.h$'|grep -v '^D'`
for f in $L; do sed -i "-es/^\([^\"]*\)\(\<POLL$V\>\)/\\1E\\2/" $f; done
done
with de-mangling cleanups yet to come.
NOTE! On almost all architectures, the EPOLL* constants have the same
values as the POLL* constants do. But they keyword here is "almost".
For various bad reasons they aren't the same, and epoll() doesn't
actually work quite correctly in some cases due to this on Sparc et al.
The next patch from Al will sort out the final differences, and we
should be all done.
Scripted-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The pipe buffer limits are accessed without any locking, and may be
changed at any time by the sysctl handlers. In theory this could cause
problems for expressions like the following:
pipe_user_pages_hard && user_bufs > pipe_user_pages_hard
... since the assembly code might reference the 'pipe_user_pages_hard'
memory location multiple times, and if the admin removes the limit by
setting it to 0, there is a very brief window where processes could
incorrectly observe the limit to be exceeded.
Fix this by loading the limits with READ_ONCE() prior to use.
Link: http://lkml.kernel.org/r/20180111052902.14409-8-ebiggers3@gmail.com
Signed-off-by: Eric Biggers <ebiggers@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Willy Tarreau <w@1wt.eu>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: "Luis R . Rodriguez" <mcgrof@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
round_pipe_size() calculates the number of pages the requested size
corresponds to, then rounds the page count up to the next power of 2.
However, it also rounds everything < PAGE_SIZE up to PAGE_SIZE.
Therefore, there's no need to actually translate the size into a page
count; we just need to round the size up to the next power of 2.
We do need to verify the size isn't greater than (1 << 31), since on
32-bit systems roundup_pow_of_two() would be undefined in that case. But
that can just be combined with the UINT_MAX check which we need anyway
now.
Finally, update pipe_set_size() to not redundantly check the return value
of round_pipe_size() for the "invalid size" case twice.
Link: http://lkml.kernel.org/r/20180111052902.14409-7-ebiggers3@gmail.com
Signed-off-by: Eric Biggers <ebiggers@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: "Luis R . Rodriguez" <mcgrof@kernel.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Willy Tarreau <w@1wt.eu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A pipe's size is represented as an 'unsigned int'. As expected, writing a
value greater than UINT_MAX to /proc/sys/fs/pipe-max-size fails with
EINVAL. However, the F_SETPIPE_SZ fcntl silently truncates such values to
32 bits, rather than failing with EINVAL as expected. (It *does* fail
with EINVAL for values above (1 << 31) but <= UINT_MAX.)
Fix this by moving the check against UINT_MAX into round_pipe_size() which
is called in both cases.
Link: http://lkml.kernel.org/r/20180111052902.14409-6-ebiggers3@gmail.com
Signed-off-by: Eric Biggers <ebiggers@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: "Luis R . Rodriguez" <mcgrof@kernel.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Willy Tarreau <w@1wt.eu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
With pipe-user-pages-hard set to 'N', users were actually only allowed up
to 'N - 1' buffers; and likewise for pipe-user-pages-soft.
Fix this to allow up to 'N' buffers, as would be expected.
Link: http://lkml.kernel.org/r/20180111052902.14409-5-ebiggers3@gmail.com
Fixes: b0b91d18e2 ("pipe: fix limit checking in pipe_set_size()")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Acked-by: Willy Tarreau <w@1wt.eu>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: "Luis R . Rodriguez" <mcgrof@kernel.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
pipe-user-pages-hard and pipe-user-pages-soft are only supposed to apply
to unprivileged users, as documented in both Documentation/sysctl/fs.txt
and the pipe(7) man page.
However, the capabilities are actually only checked when increasing a
pipe's size using F_SETPIPE_SZ, not when creating a new pipe. Therefore,
if pipe-user-pages-hard has been set, the root user can run into it and be
unable to create pipes. Similarly, if pipe-user-pages-soft has been set,
the root user can run into it and have their pipes limited to 1 page each.
Fix this by allowing the privileged override in both cases.
Link: http://lkml.kernel.org/r/20180111052902.14409-4-ebiggers3@gmail.com
Fixes: 759c01142a ("pipe: limit the per-user amount of pages allocated in pipes")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: "Luis R . Rodriguez" <mcgrof@kernel.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Willy Tarreau <w@1wt.eu>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
pipe_proc_fn() is no longer needed, as it only calls through to
proc_dopipe_max_size(). Just put proc_dopipe_max_size() in the ctl_table
entry directly, and remove the unneeded EXPORT_SYMBOL() and the ENOSYS
stub for it.
(The reason the ENOSYS stub isn't needed is that the pipe-max-size
ctl_table entry is located directly in 'kern_table' rather than being
registered separately. Therefore, the entry is already only defined when
the kernel is built with sysctl support.)
Link: http://lkml.kernel.org/r/20180111052902.14409-3-ebiggers3@gmail.com
Signed-off-by: Eric Biggers <ebiggers@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: "Luis R . Rodriguez" <mcgrof@kernel.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Willy Tarreau <w@1wt.eu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Patch series "pipe: buffer limits fixes and cleanups", v2.
This series simplifies the sysctl handler for pipe-max-size and fixes
another set of bugs related to the pipe buffer limits:
- The root user wasn't allowed to exceed the limits when creating new
pipes.
- There was an off-by-one error when checking the limits, so a limit of
N was actually treated as N - 1.
- F_SETPIPE_SZ accepted values over UINT_MAX.
- Reading the pipe buffer limits could be racy.
This patch (of 7):
Before validating the given value against pipe_min_size,
do_proc_dopipe_max_size_conv() calls round_pipe_size(), which rounds the
value up to pipe_min_size. Therefore, the second check against
pipe_min_size is redundant. Remove it.
Link: http://lkml.kernel.org/r/20180111052902.14409-2-ebiggers3@gmail.com
Signed-off-by: Eric Biggers <ebiggers@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: "Luis R . Rodriguez" <mcgrof@kernel.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Willy Tarreau <w@1wt.eu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
pipe_max_size is assigned directly via procfs sysctl:
static struct ctl_table fs_table[] = {
...
{
.procname = "pipe-max-size",
.data = &pipe_max_size,
.maxlen = sizeof(int),
.mode = 0644,
.proc_handler = &pipe_proc_fn,
.extra1 = &pipe_min_size,
},
...
int pipe_proc_fn(struct ctl_table *table, int write, void __user *buf,
size_t *lenp, loff_t *ppos)
{
...
ret = proc_dointvec_minmax(table, write, buf, lenp, ppos)
...
and then later rounded in-place a few statements later:
...
pipe_max_size = round_pipe_size(pipe_max_size);
...
This leaves a window of time between initial assignment and rounding
that may be visible to other threads. (For example, one thread sets a
non-rounded value to pipe_max_size while another reads its value.)
Similar reads of pipe_max_size are potentially racy:
pipe.c :: alloc_pipe_info()
pipe.c :: pipe_set_size()
Add a new proc_dopipe_max_size() that consolidates reading the new value
from the user buffer, verifying bounds, and calling round_pipe_size()
with a single assignment to pipe_max_size.
Link: http://lkml.kernel.org/r/1507658689-11669-4-git-send-email-joe.lawrence@redhat.com
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Reported-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Roman Gushchin
Linus Torvalds
Jan Stancek
David Howells
Al Viro
Matthew Wilcox
Shakeel Butt
Jann Horn
Christoph Hellwig
Dominik Brodowski
Eric Biggers
Joe Lawrence