There's a failure case where an application gets a cqe entry, but
the kernel can then overwrite it before the application is done
reading it. This can happen since the io_uring_{get,wait}_completion()
interface both returns a CQE pointer AND increments the ring index.
If the kernel reuses this entry before the applications is done reading
it, the contents may be corrupted.
Remove the CQ head increment from the CQE retrieval, and put it into
a separate helper, io_uring_cqe_seen(). The application must call this
helper when it got a new CQE entry through one of the above calls, and
it's now done reading it.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Also changes the fsync prep helper to require passing in the actual
flag, not just a boolean for fsync vs fdatasync.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This cache hint has been removed from the io_uring series, as there's
some hesitation to leak this information to userspace. Remove it from
the header file and the man page. We can always re-introduce it later,
if we get some variant of this included.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
A previous fix that ensured we pass back the right error messed
up the normal return, which is number of entries submitted.
This makes the poll test cases fail.
Fixes: 8260029608 ("queue: ensure io_uring_submit() returns the right error")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We weren't passing back -errno for the system call failure.
This meant any error got turned into EPERM as far as the
caller was concerned.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This really wants to be a "will we over-fill the ring?" kind of
check, but the sqe_head/sqe_tail should not be that far apart. If they
are, that's a bug elsewhere. So just kill the check.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
With a recent change to the kernel side, we now fully guarantee that
once io_uring_enter(2) returns that X entries have been submitted, it's
completely safe to reuse tohse entries. This used to not always be the
case, if an SQE had to be punted to async context for submission.
This makes for a more reliable and nicer interface.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Applications should not need to care about this, we can pass it in
ourselves. Once the libc support is there, we won't expose this
parameter either.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We would randomly fail depending on what garbage was passed in
the two sigset related fields.
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Just setup one ring in the child process, and poll for when the
read side of the pipe is readable. From the parent, write something
to the pipe.
If nothing happens within 1 second, fail the test.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
- Add a timeout, if we don't finish in one second we are definitely
hung because the poll remove command didn't work.
- Don't assume completion ordering
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Let's have the various helpers be in usefully named functions, no
need to bundle them all into the same one.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We don't need any of the information in there in the caller, and
this makes it harder to abuse as we don't require the caller to
have memset() the struct first.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
- Don't use polled IO for the io_uring-test, that'll fail if the
fs/device doesn't suppor it
- copy read-in was broken for io_uring-cp
Signed-off-by: Jens Axboe <axboe@kernel.dk>
- Fixed buffers are now available through io_uring_register()
- Various thread/wq options are now dead and automatic instead
- sqe->index is now sqe->buf_index
- Fixed buffers require flag, not separate opcode
Signed-off-by: Jens Axboe <axboe@kernel.dk>
- io_uring_sqe added a data field that's passed back at completion
- io_uring_sqe added an index field, for fixed buffer locations
- io_uring_setup(2) system call added a 'nr_iovecs' field
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This is cleaner than having the app juggle an SQ and CQ ring,
just wrap them in struct io_uring and have the API always take
that. This means the app doesn't need to worry about the different
types of rings, and that we only need to pass in one argument
for setup/teardown.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This now exposes two helpers:
io_uring_get_completion()
Return a completion, if we have one (or more) available in
the ring
io_uring_wait_completion()
Return a completion, waiting for it if necessary
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe