Some block devices, like dm, bubble back -EAGAIN through the completion
handler. We check for this in io_read(), but don't honor it for when
we have copied the iov. Return -EAGAIN for this case before retrying,
to force punt to io-wq.

Fixes: bcf5a063 ("io_uring: support true async buffered reads, if file provides it")
Reported-by: Zorro Lang <zlang@redhat.com>
Tested-by: Zorro Lang <zlang@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

If we already have mapped the necessary data for retry, then don't set
it up again. It's a pointless operation, and we leak the iovec if it's
a large (non-stack) vec.

Fixes: b63534c4 ("io_uring: re-issue block requests that failed because of resources")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

This isn't safe, and isn't needed either. We are guaranteed that any
work we queue is on a live task (and will be run), or it goes to
our backup io-wq threads if the task is exiting.
Signed-off-by: Jens Axboe <axboe@kernel.dk>

If task_work ends up being marked for cancelation, we go through a
cancelation helper instead of the queue path. In converting task_work to
always hold a ctx reference, this path was missed. Make sure that
io_req_task_cancel() puts the reference that is being held against the
ctx.

Fixes: 6d816e08 ("io_uring: hold 'ctx' reference around task_work queue + execute")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Always grab work environment for deferred links. The assumption that we
will be running it always from the task in question is false, as exiting
tasks may mean that we're deferring this one to a thread helper. And at
that point it's too late to grab the work environment.

Fixes: debb85f4 ("io_uring: factor out grab_env() from defer_prep()")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

While looking for ->files in ->defer_list, consider that requests there
may actually be links.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

While trying to cancel requests with ->files, it also should look for
requests in ->defer_list, otherwise it might end up hanging a thread.

Cancel all requests in ->defer_list up to the last request there with
matching ->files, that's needed to follow drain ordering semantics.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

If we exceed UIO_FASTIOV, we don't handle the transition correctly
between an allocated vec for requests that are queued with IOSQE_ASYNC.
Store the iovec appropriately and re-set it in the iter iov in case
it changed.

Fixes: ff6165b2 ("io_uring: retain iov_iter state over io_read/io_write calls")
Reported-by: Nick Hill <nick@nickhill.org>
Tested-by: Norman Maurer <norman.maurer@googlemail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Actually two things that need fixing up here:

- The io_rw_reissue() -EAGAIN retry is explicit to block devices and
  regular files, so don't ever attempt to do that on other types of
  files.

- If we hit -EAGAIN on a nonblock marked file, don't arm poll handler for
  it. It should just complete with -EAGAIN.

Cc: stable@vger.kernel.org
Reported-by: Norman Maurer <norman.maurer@googlemail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

While io_sqe_file_register() failed in __io_sqe_files_update(),
table->files[i] still point to the original file which may freed
soon, and that will trigger use-after-free problems.

Cc: stable@vger.kernel.org
Fixes: f3bd9dae ("io_uring: fix memleak in __io_sqe_files_update()")
Signed-off-by: Jiufei Xue <jiufei.xue@linux.alibaba.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Index here is already the position of the file in fixed_file_table, we
should not use io_file_from_index() again to get it. Otherwise, the
wrong file which still in use may be released unexpectedly.

Cc: stable@vger.kernel.org # v5.6
Fixes: 05f3fb3c ("io_uring: avoid ring quiesce for fixed file set unregister and update")
Signed-off-by: Jiufei Xue <jiufei.xue@linux.alibaba.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

These events happen inline from submission, so there's no need to
bounce them through the original task. Just set them up for retry
and issue retry directly instead of going over task_work.
Signed-off-by: Jens Axboe <axboe@kernel.dk>

This normally isn't hit, as polling is mostly done on NVMe with deep
queue depths. But if we do run into request starvation, we need to
ensure that retries are properly serialized.
Reported-by: Andres Freund <andres@anarazel.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Make sure we clear req->result, which was set to -EAGAIN for retry
purposes, when moving it to the reissue list. Otherwise we can end up
retrying a request more than once, which leads to weird results in
the io-wq handling (and other spots).

Cc: stable@vger.kernel.org
Reported-by: Andres Freund <andres@anarazel.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

The man page for io_uring generally claims were consistent with what
preadv2 and pwritev2 accept, but turns out there's a slight discrepancy
in how offset == -1 is handled for pipes/streams. preadv doesn't allow
it, but preadv2 does. This currently causes io_uring to return -EINVAL
if that is attempted, but we should allow that as documented.

This change makes us consistent with preadv2/pwritev2 for just passing
in a NULL ppos for streams if the offset is -1.

Cc: stable@vger.kernel.org # v5.7+
Reported-by: Benedikt Ames <wisp3rwind@posteo.eu>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

We need to call kiocb_done() for any ret < 0 to ensure that we always
get the proper -ERESTARTSYS (and friends) transformation done.

At some point this should be tied into general error handling, so we
can get rid of the various (mostly network) related commands that check
and perform this substitution.
Signed-off-by: Jens Axboe <axboe@kernel.dk>

There's no point in using the poll handler if we can't do a nonblocking
IO attempt of the operation, since we'll need to go async anyway. In
fact this is actively harmful, as reading from eg pipes won't return 0
to indicate EOF.

Cc: stable@vger.kernel.org # v5.7+
Reported-by: Benedikt Ames <wisp3rwind@posteo.eu>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

We do the initial accounting of locked_vm and pinned_vm before we have
setup ctx->sqo_mm, which means we can end up having not accounted the
memory at setup time, but still decrement it when we exit. This causes
an imbalance in the accounting.

Setup ctx->sqo_mm earlier in io_uring_create(), before we do the first
accounting of mm->{locked,pinned}_vm. This also unifies the state
grabbing for the ctx, and eliminates a failure case in
io_sq_offload_start().

Fixes: f74441e6 ("io_uring: account locked memory before potential error case")
Reported-by: Robert M. Muncrief <rmuncrief@humanavance.com>
Reported-by: Niklas Schnelle <schnelle@linux.ibm.com>
Tested-by: Niklas Schnelle <schnelle@linux.ibm.com>
Tested-by: Robert M. Muncrief <rmuncrief@humanavance.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Some consumers of the iov_iter will return an error, but still have
bytes consumed in the iterator. This is an issue for -EAGAIN, since we
rely on a sane iov_iter state across retries.

Fix this by ensuring that we revert consumed bytes, if any, if the file
operations have consumed any bytes from iterator. This is similar to what
generic_file_read_iter() does, and is always safe as we have the previous
bytes count handy already.

Fixes: ff6165b2 ("io_uring: retain iov_iter state over io_read/io_write calls")
Reported-by: Dmitry Shulyak <yashulyak@gmail.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Don't forget to update wqe->hash_tail after cancelling a pending work
item, if it was hashed.

Cc: stable@vger.kernel.org # 5.7+
Reported-by: Dmitry Shulyak <yashulyak@gmail.com>
Fixes: 86f3cd1b ("io-wq: handle hashed writes in chains")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

If an application is doing reads on signalfd, and we arm the poll handler
because there's no data available, then the wakeup can recurse on the
tasks sighand->siglock as the signal delivery from task_work_add() will
use TWA_SIGNAL and that attempts to lock it again.

We can detect the signalfd case pretty easily by comparing the poll->head
wait_queue_head_t with the target task signalfd wait queue. Just use
normal task wakeup for this case.

Cc: stable@vger.kernel.org # v5.7+
Signed-off-by: Jens Axboe <axboe@kernel.dk>

If io_import_iovec() returns an error, return iovec is undefined and
must not be used, so don't set it to NULL when failing.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

kfree() handles NULL pointers well, but io_{read,write}() checks it
because of performance reasons. Leave a comment there for those who are
tempted to patch it.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Setting and clearing REQ_F_OVERFLOW in io_uring_cancel_files() and
io_cqring_overflow_flush() are racy, because they might be called
asynchronously.

REQ_F_OVERFLOW flag in only needed for files cancellation, so if it can
be guaranteed that requests _currently_ marked inflight can't be
overflown, the problem will be solved with removing the flag
altogether.

That's how the patch works, it removes inflight status of a request
in io_cqring_fill_event() whenever it should be thrown into CQ-overflow
list. That's Ok to do, because no opcode specific handling can be done
after io_cqring_fill_event(), the same assumption as with "struct
io_completion" patches.
And it already have a good place for such cleanups, which is
io_clean_op(). A nice side effect of this is removing this inflight
check from the hot path.

note on synchronisation: now __io_cqring_fill_event() may be taking two
spinlocks simultaneously, completion_lock and inflight_lock. It's fine,
because we never do that in reverse order, and CQ-overflow of inflight
requests shouldn't happen often.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

We currently use system_wq, which is unbounded in terms of number of
workers. This means that if we're exiting tons of rings at the same
time, then we'll briefly spawn tons of event kworkers just for a very
short blocking time as the rings exit.

Use system_unbound_wq instead, which has a sane cap on the concurrency
level.
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_rw_prep_async() goes through a dance of clearing req->io, calling
the iovec import, then re-setting req->io. Provide an internal helper
that does the right thing without needing state tweaked to get there.

This enables further cleanups in io_read, io_write, and
io_resubmit_prep(), but that's left for another time.
Signed-off-by: Jens Axboe <axboe@kernel.dk>

The 5.9 merge moved this function io_uring, which means that we don't
need to retain the generic nature of it. Clean up this part by removing
redundant checks, and just inlining the small remainder in
io_rw_should_retry().

No functional changes in this patch.
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Commit f254ac04 ("io_uring: enable lookup of links holding inflight files")
only handled 2 out of the three head link cases we have, we also need to
lookup and cancel work that is blocked in io-wq if that work has a link
that's holding a reference to the files structure.

Put the "cancel head links that hold this request pending" logic into
io_attempt_cancel(), which will to through the motions of finding and
canceling head links that hold the current inflight files stable request
pending.

Cc: stable@vger.kernel.org
Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Pull io_uring fixes from Jens Axboe:
 "A few differerent things in here.

  Seems like syzbot got some more io_uring bits wired up, and we got a
  handful of reports and the associated fixes are in here.

  General fixes too, and a lot of them marked for stable.

  Lastly, a bit of fallout from the async buffered reads, where we now
  more easily trigger short reads. Some applications don't really like
  that, so the io_read() code now handles short reads internally, and
  got a cleanup along the way so that it's now easier to read (and
  documented). We're now passing tests that failed before"

* tag 'io_uring-5.9-2020-08-15' of git://git.kernel.dk/linux-block:
  io_uring: short circuit -EAGAIN for blocking read attempt
  io_uring: sanitize double poll handling
  io_uring: internally retry short reads
  io_uring: retain iov_iter state over io_read/io_write calls
  task_work: only grab task signal lock when needed
  io_uring: enable lookup of links holding inflight files
  io_uring: fail poll arm on queue proc failure
  io_uring: hold 'ctx' reference around task_work queue + execute
  fs: RWF_NOWAIT should imply IOCB_NOIO
  io_uring: defer file table grabbing request cleanup for locked requests
  io_uring: add missing REQ_F_COMP_LOCKED for nested requests
  io_uring: fix recursive completion locking on oveflow flush
  io_uring: use TWA_SIGNAL for task_work uncondtionally
  io_uring: account locked memory before potential error case
  io_uring: set ctx sq/cq entry count earlier
  io_uring: Fix NULL pointer dereference in loop_rw_iter()
  io_uring: add comments on how the async buffered read retry works
  io_uring: io_async_buf_func() need not test page bit

Commit 1355c31e ("asm-generic: pgalloc: provide generic pmd_alloc_one()
and pmd_free_one()") converted parisc to use generic version of
pmd_alloc_one() but it missed the fact that parisc uses order-1 pages for
PMD.

Restore the original version of pmd_alloc_one() for parisc, just use
GFP_PGTABLE_KERNEL that implies __GFP_ZERO instead of GFP_KERNEL and
memset.

Fixes: 1355c31e ("asm-generic: pgalloc: provide generic pmd_alloc_one() and pmd_free_one()")
Reported-by: Meelis Roos <mroos@linux.ee>
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Tested-by: Meelis Roos <mroos@linux.ee>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Link: https://lkml.kernel.org/r/9f2b5ebd-e4a4-0fa1-6cd3-4b9f6892d1ad@linux.eeSigned-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull block fixes from Jens Axboe:
 "A few fixes on the block side of things:

   - Discard granularity fix (Coly)

   - rnbd cleanups (Guoqing)

   - md error handling fix (Dan)

   - md sysfs fix (Junxiao)

   - Fix flush request accounting, which caused an IO slowdown for some
     configurations (Ming)

   - Properly propagate loop flag for partition scanning (Lennart)"

* tag 'block-5.9-2020-08-14' of git://git.kernel.dk/linux-block:
  block: fix double account of flush request's driver tag
  loop: unset GENHD_FL_NO_PART_SCAN on LOOP_CONFIGURE
  rnbd: no need to set bi_end_io in rnbd_bio_map_kern
  rnbd: remove rnbd_dev_submit_io
  md-cluster: Fix potential error pointer dereference in resize_bitmaps()
  block: check queue's limits.discard_granularity in __blkdev_issue_discard()
  md: get sysfs entry after redundancy attr group create

Pull RISC-V fix from Palmer Dabbelt:
 "I collected a single fix during the merge window: we managed to break
  the early trap setup on !MMU, this fixes it"

* tag 'riscv-for-linus-5.9-mw1' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: Setup exception vector for nommu platform

Pull arch/sh updates from Rich Felker:
 "Cleanup, SECCOMP_FILTER support, message printing fixes, and other
  changes to arch/sh"

* tag 'sh-for-5.9' of git://git.libc.org/linux-sh: (34 commits)
  sh: landisk: Add missing initialization of sh_io_port_base
  sh: bring syscall_set_return_value in line with other architectures
  sh: Add SECCOMP_FILTER
  sh: Rearrange blocks in entry-common.S
  sh: switch to copy_thread_tls()
  sh: use the generic dma coherent remap allocator
  sh: don't allow non-coherent DMA for NOMMU
  dma-mapping: consolidate the NO_DMA definition in kernel/dma/Kconfig
  sh: unexport register_trapped_io and match_trapped_io_handler
  sh: don't include <asm/io_trapped.h> in <asm/io.h>
  sh: move the ioremap implementation out of line
  sh: move ioremap_fixed details out of <asm/io.h>
  sh: remove __KERNEL__ ifdefs from non-UAPI headers
  sh: sort the selects for SUPERH alphabetically
  sh: remove -Werror from Makefiles
  sh: Replace HTTP links with HTTPS ones
  arch/sh/configs: remove obsolete CONFIG_SOC_CAMERA*
  sh: stacktrace: Remove stacktrace_ops.stack()
  sh: machvec: Modernize printing of kernel messages
  sh: pci: Modernize printing of kernel messages
  ...

One case was missed in the short IO retry handling, and that's hitting
-EAGAIN on a blocking attempt read (eg from io-wq context). This is a
problem on sockets that are marked as non-blocking when created, they
don't carry any REQ_F_NOWAIT information to help us terminate them
instead of perpetually retrying.

Fixes: 227c0c96 ("io_uring: internally retry short reads")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

There's a bit of confusion on the matching pairs of poll vs double poll,
depending on if the request is a pure poll (IORING_OP_POLL_ADD) or
poll driven retry.

Add io_poll_get_double() that returns the double poll waitqueue, if any,
and io_poll_get_single() that returns the original poll waitqueue. With
that, remove the argument to io_poll_remove_double().

Finally ensure that wait->private is cleared once the double poll handler
has run, so that remove knows it's already been seen.

Cc: stable@vger.kernel.org # v5.8
Reported-by: syzbot+7f617d4a9369028b8a2c@syzkaller.appspotmail.com
Fixes: 18bceab1 ("io_uring: allow POLL_ADD with double poll_wait() users")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Pull more perf tools updates from Arnaldo Carvalho de Melo:
 "Fixes:
   - Fixes for 'perf bench numa'.

   - Always memset source before memcpy in 'perf bench mem'.

   - Quote CC and CXX for their arguments to fix build in environments
     using those variables to pass more than just the compiler names.

   - Fix module symbol processing, addressing regression detected via
     "perf test".

   - Allow multiple probes in record+script_probe_vfs_getname.sh 'perf
     test' entry.

  Improvements:
   - Add script to autogenerate socket family name id->string table from
     copy of kernel header, used so far in 'perf trace'.

   - 'perf ftrace' improvements to provide similar options for this
     utility so that one can go from 'perf record', 'perf trace', etc to
     'perf ftrace' just by changing the name of the subcommand.

   - Prefer new "sched:sched_waking" trace event when it exists in 'perf
     sched' post processing.

   - Update POWER9 metrics to utilize other metrics.

   - Fall back to querying debuginfod if debuginfo not found locally.

  Miscellaneous:
   - Sync various kvm headers with kernel sources"

* tag 'perf-tools-2020-08-14' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux: (40 commits)
  perf ftrace: Make option description initials all capital letters
  perf build-ids: Fall back to debuginfod query if debuginfo not found
  perf bench numa: Remove dead code in parse_nodes_opt()
  perf stat: Update POWER9 metrics to utilize other metrics
  perf ftrace: Add change log
  perf: ftrace: Add set_tracing_options() to set all trace options
  perf ftrace: Add option --tid to filter by thread id
  perf ftrace: Add option -D/--delay to delay tracing
  perf: ftrace: Allow set graph depth by '--graph-opts'
  perf ftrace: Add support for trace option tracing_thresh
  perf ftrace: Add option 'verbose' to show more info for graph tracer
  perf ftrace: Add support for tracing option 'irq-info'
  perf ftrace: Add support for trace option funcgraph-irqs
  perf ftrace: Add support for trace option sleep-time
  perf ftrace: Add support for tracing option 'func_stack_trace'
  perf tools: Add general function to parse sublevel options
  perf ftrace: Add option '--inherit' to trace children processes
  perf ftrace: Show trace column header
  perf ftrace: Add option '-m/--buffer-size' to set per-cpu buffer size
  perf ftrace: Factor out function write_tracing_file_int()
  ...

Pull x86 fixes from Ingo Molnar:
 "Misc fixes and small updates all around the place:

   - Fix mitigation state sysfs output

   - Fix an FPU xstate/sxave code assumption bug triggered by
     Architectural LBR support

   - Fix Lightning Mountain SoC TSC frequency enumeration bug

   - Fix kexec debug output

   - Fix kexec memory range assumption bug

   - Fix a boundary condition in the crash kernel code

   - Optimize porgatory.ro generation a bit

   - Enable ACRN guests to use X2APIC mode

   - Reduce a __text_poke() IRQs-off critical section for the benefit of
     PREEMPT_RT"

* tag 'x86-urgent-2020-08-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/alternatives: Acquire pte lock with interrupts enabled
  x86/bugs/multihit: Fix mitigation reporting when VMX is not in use
  x86/fpu/xstate: Fix an xstate size check warning with architectural LBRs
  x86/purgatory: Don't generate debug info for purgatory.ro
  x86/tsr: Fix tsc frequency enumeration bug on Lightning Mountain SoC
  kexec_file: Correctly output debugging information for the PT_LOAD ELF header
  kexec: Improve & fix crash_exclude_mem_range() to handle overlapping ranges
  x86/crash: Correct the address boundary of function parameters
  x86/acrn: Remove redundant chars from ACRN signature
  x86/acrn: Allow ACRN guest to use X2APIC mode

Pull scheduler fixes from Ingo Molnar:
 "Two fixes: fix a new tracepoint's output value, and fix the formatting
  of show-state syslog printouts"

* tag 'sched-urgent-2020-08-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/debug: Fix the alignment of the show-state debug output
  sched: Fix use of count for nr_running tracepoint

Pull perf fixes from Ingo Molnar:
 "Misc fixes, an expansion of perf syscall access to CAP_PERFMON
  privileged tools, plus a RAPL HW-enablement for Intel SPR platforms"

* tag 'perf-urgent-2020-08-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/rapl: Add support for Intel SPR platform
  perf/x86/rapl: Support multiple RAPL unit quirks
  perf/x86/rapl: Fix missing psys sysfs attributes
  hw_breakpoint: Remove unused __register_perf_hw_breakpoint() declaration
  kprobes: Remove show_registers() function prototype
  perf/core: Take over CAP_SYS_PTRACE creds to CAP_PERFMON capability