splice: only read in as much information as there is pipe buffer space
PATCH v3 1/1 splice: only read in as much information as there is pipe buffer space
From: Darrick J. Wong <[email protected]>
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.
--- fs/splice.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-)
fs/splice.c +11 -3
@@ -945,12 +945,13 @@ ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,WARN_ON_ONCE (pipe->nrbufs !=0 );while (len) { +unsigned int pipe_pages;size_t read_len;loff_t pos = sd->pos, prev_pos = pos;/* Don't try to read more the pipe has space for. */ - read_len =min_t (size_t, len, - (pipe->buffers - pipe->nrbufs) <<PAGE_SHIFT ); + pipe_pages = pipe->buffers - pipe->nrbufs; + read_len =min (len, (size_t )pipe_pages <<PAGE_SHIFT ); ret =do_splice_to (in, &pos, pipe, read_len, flags);if (unlikely (ret <=0 )) goto out_release;
@@ -1180,8 +1181,15 @@ static long do_splice(struct file *in, loff_t __user *off_in,pipe_lock (opipe); ret =wait_for_space (opipe, flags); -if (!ret) +if (!ret) { +unsigned int pipe_pages; + +/* Don't try to read more the pipe has space for. */ + pipe_pages = opipe->buffers - opipe->nrbufs; + len =min (len, (size_t )pipe_pages <<PAGE_SHIFT ); + ret =do_splice_to (in, &offset, opipe, len, flags); + }pipe_unlock (opipe);if (ret >0 )wakeup_pipe_readers (opipe);
Hi Darrick,
Am Di., 15. Okt. 2019 um 07:33 Uhr schrieb Darrick J. Wong <[email protected]>:
Quoted 7 lines
From: Darrick J. Wong <[email protected]>
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.
Quoted 7 lines
Months ago we fixed
splice_direct_to_actorto clamp the length of the read request to the size of the splice pipe. Do the same todo_splice.Fixes 17614445576b6 ("splice: don't read more than available pipe space")Reported-by [email protected]Reported-by Andreas Grünbacher <[email protected]>Signed-off-by Darrick J. Wong <[email protected]>
I've done some minimal testing on top of 5.4-rc3. This patch fixes the splice issue and also passes the syzbot reproducer. I'll add it to the set of patches I regularly run fstests on now, but we already know fstests doesn't cover splice in the greatest depth possible, so that's unlikely to reveal anything new.
Thanks,
Andreas
Quoted 41 lines
--- fs/splice.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-)fs/splice.c +11
-3@@ -945,12 +945,13 @@ ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,WARN_ON_ONCE (pipe->nrbufs !=0 );while (len) { +unsigned int pipe_pages;size_t read_len;loff_t pos = sd->pos, prev_pos = pos;/* Don't try to read more the pipe has space for. */ - read_len =min_t (size_t, len, - (pipe->buffers - pipe->nrbufs) <<PAGE_SHIFT ); + pipe_pages = pipe->buffers - pipe->nrbufs; + read_len =min (len, (size_t )pipe_pages <<PAGE_SHIFT ); ret =do_splice_to (in, &pos, pipe, read_len, flags);if (unlikely (ret <=0 )) goto out_release;@@ -1180,8 +1181,15 @@ static long do_splice(struct file *in, loff_t __user *off_in,pipe_lock (opipe); ret =wait_for_space (opipe, flags); -if (!ret) +if (!ret) { +unsigned int pipe_pages; + +/* Don't try to read more the pipe has space for. */ + pipe_pages = opipe->buffers - opipe->nrbufs; + len =min (len, (size_t )pipe_pages <<PAGE_SHIFT ); + ret =do_splice_to (in, &offset, opipe, len, flags); + }pipe_unlock (opipe);if (ret >0 )wakeup_pipe_readers (opipe);
Quoted 23 lines · Darrick J. Wong
From: Darrick J. Wong <[email protected]>
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_actorto clamp the length of the read request to the size of the splice pipe. Do the same todo_splice.Fixes 17614445576b6 ("splice: don't read more than available pipe space")Reported-by [email protected]Reported-by Andreas Grünbacher <[email protected]>Signed-off-by Darrick J. Wong <[email protected]>--- fs/splice.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-)fs/splice.c +0
-0
fs/splice.c ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd, · line 948
Quoted 4 lines
fs/splice.c +1
-0@@ -945,12 +945,13 @@ ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,WARN_ON_ONCE (pipe->nrbufs !=0 );while (len) { +unsigned int pipe_pages;
define this as a size_t...
fs/splice.c ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd, · line 954
Quoted 8 lines
fs/splice.c +2
-2size_t read_len;loff_t pos = sd->pos, prev_pos = pos;/* Don't try to read more the pipe has space for. */ - read_len =min_t (size_t, len, - (pipe->buffers - pipe->nrbufs) <<PAGE_SHIFT ); + pipe_pages = pipe->buffers - pipe->nrbufs; + read_len =min (len, (size_t )pipe_pages <<PAGE_SHIFT );
read_len = min_t(size_t, len, pipe_pages << PAGER_SHIFT);
Quoted 4 lines
fs/splice.c +0
-0ret =do_splice_to (in, &pos, pipe, read_len, flags);if (unlikely (ret <=0 )) goto out_release;
fs/splice.c static long do_splice(struct file *in, loff_t __user *off_in, · line 1189
Quoted 10 lines
fs/splice.c +6
-1@@ -1180,8 +1181,15 @@ static long do_splice(struct file *in, loff_t __user *off_in,pipe_lock (opipe); ret =wait_for_space (opipe, flags); -if (!ret) +if (!ret) { +unsigned int pipe_pages; + +/* Don't try to read more the pipe has space for. */ + pipe_pages = opipe->buffers - opipe->nrbufs; + len =min (len, (size_t )pipe_pages <<PAGE_SHIFT );
And same here...
Cheers,
Dave.
Signature · 2 lines
Dave Chinner [email protected]
Quoted 34 lines · Dave Chinner
Quoted 23 lines · Darrick J. Wong
From: Darrick J. Wong <[email protected]>
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_actorto clamp the length of the read request to the size of the splice pipe. Do the same todo_splice.Fixes 17614445576b6 ("splice: don't read more than available pipe space")Reported-by [email protected]Reported-by Andreas Grünbacher <[email protected]>Signed-off-by Darrick J. Wong <[email protected]>--- fs/splice.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-)fs/splice.c +0
-0fs/splice.c ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd, · line 948
Quoted 4 lines
fs/splice.c +1
-0@@ -945,12 +945,13 @@ ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,WARN_ON_ONCE (pipe->nrbufs !=0 );while (len) { +unsigned int pipe_pages;define this as a
size_t...Quoted 3 lines
fs/splice.c +0
-0size_t read_len;loff_t pos = sd->pos, prev_pos = pos;
Quoted 7 lines
fs/splice.c ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd, · line 954
Quoted 5 lines
fs/splice.c +2
-2/* Don't try to read more the pipe has space for. */ - read_len =min_t (size_t, len, - (pipe->buffers - pipe->nrbufs) <<PAGE_SHIFT ); + pipe_pages = pipe->buffers - pipe->nrbufs; + read_len =min (len, (size_t )pipe_pages <<PAGE_SHIFT );read_len = min_t(size_t, len, pipe_pages << PAGER_SHIFT);
If we make pipe_pages have type size_t then we don't need min_t here, right? Since len and read_len are already size_t.
--D
Quoted 24 lines
Quoted 4 lines
fs/splice.c +0
-0ret =do_splice_to (in, &pos, pipe, read_len, flags);if (unlikely (ret <=0 )) goto out_release;fs/splice.c static long do_splice(struct file *in, loff_t __user *off_in, · line 1189
Quoted 10 lines
fs/splice.c +6
-1@@ -1180,8 +1181,15 @@ static long do_splice(struct file *in, loff_t __user *off_in,pipe_lock (opipe); ret =wait_for_space (opipe, flags); -if (!ret) +if (!ret) { +unsigned int pipe_pages; + +/* Don't try to read more the pipe has space for. */ + pipe_pages = opipe->buffers - opipe->nrbufs; + len =min (len, (size_t )pipe_pages <<PAGE_SHIFT );And same here...
Cheers,
Dave.
Signature · 2 lines
Dave Chinner [email protected]
Quoted 7 lines · Darrick J. Wong
From: Darrick J. Wong <[email protected]>
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.
Quoted 5 lines
Months ago we fixed
splice_direct_to_actorto clamp the length of the read request to the size of the splice pipe. Do the same todo_splice.Fixes 17614445576b6 ("splice: don't read more than available pipe space")Reported-by [email protected]
I already invalidated this syzbot report when the previous version of this patch was dropped, as that was what the report appeared to be for. So you don't need this Reported-by line. It's not a big deal, but including it could mislead people into thinking that syzbot found a problem with the commit in the Fixes:
line, rather than a prior version of this patch.
- Eric
Quoted 11 lines · Eric Biggers
Quoted 10 lines · Darrick J. Wong
From: Darrick J. Wong <[email protected]>
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_actorto clamp the length of the read request to the size of the splice pipe. Do the same todo_splice.
Quoted 8 lines
Quoted 2 lines
Fixes 17614445576b6 ("splice: don't read more than available pipe space")Reported-by [email protected]I already invalidated this syzbot report when the previous version of this patch was dropped, as that was what the report appeared to be for. So you don't need this Reported-by line. It's not a big deal, but including it could mislead people into thinking that syzbot found a problem with the commit in the Fixes:
line, rather than a prior version of this patch.
Ok, will drop for v4. Thanks for taking care of the report.
--D
Quoted 1 line
- Eric