* [PATCH] vfs: fix vmplice_to_user()
@ 2014-05-27 14:41 Miklos Szeredi
2014-05-27 17:04 ` Linus Torvalds
2014-05-27 17:51 ` Al Viro
0 siblings, 2 replies; 5+ messages in thread
From: Miklos Szeredi @ 2014-05-27 14:41 UTC (permalink / raw)
To: Linus Torvalds, Al Viro; +Cc: linux-kernel, linux-fsdevel
From: Miklos Szeredi <mszeredi@suse.cz>
Commit 6130f5315ee8 "switch vmsplice_to_user() to copy_page_to_iter()" in
v3.15-rc1 broke vmsplice(2).
This patch fixes two bugs:
- count is not initialized to a proper value, which resulted in no data
being copied
- if rw_copy_check_uvector() returns negative then the iov might be leaked.
Tested OK.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
---
fs/splice.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -1537,7 +1537,7 @@ static long vmsplice_to_user(struct file
struct iovec iovstack[UIO_FASTIOV];
struct iovec *iov = iovstack;
struct iov_iter iter;
- ssize_t count = 0;
+ ssize_t count;
pipe = get_pipe_info(file);
if (!pipe)
@@ -1546,8 +1546,9 @@ static long vmsplice_to_user(struct file
ret = rw_copy_check_uvector(READ, uiov, nr_segs,
ARRAY_SIZE(iovstack), iovstack, &iov);
if (ret <= 0)
- return ret;
+ goto out;
+ count = ret;
iov_iter_init(&iter, iov, nr_segs, count, 0);
sd.len = 0;
@@ -1560,6 +1561,7 @@ static long vmsplice_to_user(struct file
ret = __splice_from_pipe(pipe, &sd, pipe_to_user);
pipe_unlock(pipe);
+out:
if (iov != iovstack)
kfree(iov);
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] vfs: fix vmplice_to_user()
2014-05-27 14:41 Miklos Szeredi
@ 2014-05-27 17:04 ` Linus Torvalds
2014-05-27 17:57 ` Al Viro
2014-05-27 17:51 ` Al Viro
1 sibling, 1 reply; 5+ messages in thread
From: Linus Torvalds @ 2014-05-27 17:04 UTC (permalink / raw)
To: Miklos Szeredi; +Cc: Al Viro, Linux Kernel Mailing List, linux-fsdevel
[-- Attachment #1: Type: text/plain, Size: 550 bytes --]
On Tue, May 27, 2014 at 7:41 AM, Miklos Szeredi <miklos@szeredi.hu> wrote:
>
> This patch fixes two bugs:
Mind if I change it to avoid the "goto"? I don't think goto is evil,
but in this case it doesn't seem to buy anything, except to perhaps
make the patch a bit smaller at the expense of making the result a bit
harder to read.
Replacing the
if (ret <= 0)
goto out;
with
if (ret > 0) {
...
}
also allows moving the various variable declarations closer to their use.
So something like the attached..
Linus
[-- Attachment #2: patch.diff --]
[-- Type: text/plain, Size: 1472 bytes --]
fs/splice.c | 30 +++++++++++++++---------------
1 file changed, 15 insertions(+), 15 deletions(-)
diff --git a/fs/splice.c b/fs/splice.c
index 9bc07d2b53cf..6b1115005150 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -1532,12 +1532,9 @@ static long vmsplice_to_user(struct file *file, const struct iovec __user *uiov,
unsigned long nr_segs, unsigned int flags)
{
struct pipe_inode_info *pipe;
- struct splice_desc sd;
- long ret;
struct iovec iovstack[UIO_FASTIOV];
struct iovec *iov = iovstack;
- struct iov_iter iter;
- ssize_t count = 0;
+ long ret;
pipe = get_pipe_info(file);
if (!pipe)
@@ -1545,20 +1542,23 @@ static long vmsplice_to_user(struct file *file, const struct iovec __user *uiov,
ret = rw_copy_check_uvector(READ, uiov, nr_segs,
ARRAY_SIZE(iovstack), iovstack, &iov);
- if (ret <= 0)
- return ret;
+ if (ret > 0) {
+ struct splice_desc sd;
+ struct iov_iter iter;
+ ssize_t count = ret;
- iov_iter_init(&iter, iov, nr_segs, count, 0);
+ iov_iter_init(&iter, iov, nr_segs, count, 0);
- sd.len = 0;
- sd.total_len = count;
- sd.flags = flags;
- sd.u.data = &iter;
- sd.pos = 0;
+ sd.len = 0;
+ sd.total_len = count;
+ sd.flags = flags;
+ sd.u.data = &iter;
+ sd.pos = 0;
- pipe_lock(pipe);
- ret = __splice_from_pipe(pipe, &sd, pipe_to_user);
- pipe_unlock(pipe);
+ pipe_lock(pipe);
+ ret = __splice_from_pipe(pipe, &sd, pipe_to_user);
+ pipe_unlock(pipe);
+ }
if (iov != iovstack)
kfree(iov);
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] vfs: fix vmplice_to_user()
2014-05-27 14:41 Miklos Szeredi
2014-05-27 17:04 ` Linus Torvalds
@ 2014-05-27 17:51 ` Al Viro
1 sibling, 0 replies; 5+ messages in thread
From: Al Viro @ 2014-05-27 17:51 UTC (permalink / raw)
To: Miklos Szeredi; +Cc: Linus Torvalds, linux-kernel, linux-fsdevel
On Tue, May 27, 2014 at 04:41:16PM +0200, Miklos Szeredi wrote:
> From: Miklos Szeredi <mszeredi@suse.cz>
>
> Commit 6130f5315ee8 "switch vmsplice_to_user() to copy_page_to_iter()" in
> v3.15-rc1 broke vmsplice(2).
>
> This patch fixes two bugs:
>
> - count is not initialized to a proper value, which resulted in no data
> being copied
>
> - if rw_copy_check_uvector() returns negative then the iov might be leaked.
>
> Tested OK.
Applied, will push tonight.
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] vfs: fix vmplice_to_user()
2014-05-27 17:04 ` Linus Torvalds
@ 2014-05-27 17:57 ` Al Viro
0 siblings, 0 replies; 5+ messages in thread
From: Al Viro @ 2014-05-27 17:57 UTC (permalink / raw)
To: Linus Torvalds; +Cc: Miklos Szeredi, Linux Kernel Mailing List, linux-fsdevel
On Tue, May 27, 2014 at 10:04:58AM -0700, Linus Torvalds wrote:
> On Tue, May 27, 2014 at 7:41 AM, Miklos Szeredi <miklos@szeredi.hu> wrote:
> >
> > This patch fixes two bugs:
>
> Mind if I change it to avoid the "goto"? I don't think goto is evil,
> but in this case it doesn't seem to buy anything, except to perhaps
> make the patch a bit smaller at the expense of making the result a bit
> harder to read.
>
> Replacing the
>
> if (ret <= 0)
> goto out;
>
> with
>
> if (ret > 0) {
> ...
> }
>
> also allows moving the various variable declarations closer to their use.
>
> So something like the attached..
I'm not sure it's better; I've a pending patch that does combination of
rw_copy_check_uvector() with iov_iter_init(), which means that iov
array will be going to the outer scope anyway...
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] vfs: fix vmplice_to_user()
@ 2014-05-27 19:25 George Spelvin
0 siblings, 0 replies; 5+ messages in thread
From: George Spelvin @ 2014-05-27 19:25 UTC (permalink / raw)
To: viro; +Cc: linux, linux-fsdevel, linux-kernel, miklos, torvalds
You could also get rid of the separate ret/count variables
and use ssize_t everywhere; that's the declared return type
of rw_copy_check_uvector and __splice_from_pipe, after all.
Hunks after the first two are optional extras to make types
consistent in other functions in that file, although I stopped
once I realized how big a job it owuld be to do it all.
Signed-off-by: George Spelvin <linux@horizon.com>
diff --git a/fs/splice.c b/fs/splice.c
index 6b111500..2fce8f44 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -1534,18 +1534,17 @@ static long vmsplice_to_user(struct file *file, const struct iovec __user *uiov,
struct pipe_inode_info *pipe;
struct iovec iovstack[UIO_FASTIOV];
struct iovec *iov = iovstack;
- long ret;
+ ssize_t count;
pipe = get_pipe_info(file);
if (!pipe)
return -EBADF;
- ret = rw_copy_check_uvector(READ, uiov, nr_segs,
+ count = rw_copy_check_uvector(READ, uiov, nr_segs,
ARRAY_SIZE(iovstack), iovstack, &iov);
- if (ret > 0) {
+ if (count > 0) {
struct splice_desc sd;
struct iov_iter iter;
- ssize_t count = ret;
iov_iter_init(&iter, iov, nr_segs, count, 0);
@@ -1556,14 +1555,14 @@ static long vmsplice_to_user(struct file *file, const struct iovec __user *uiov,
sd.pos = 0;
pipe_lock(pipe);
- ret = __splice_from_pipe(pipe, &sd, pipe_to_user);
+ count = __splice_from_pipe(pipe, &sd, pipe_to_user);
pipe_unlock(pipe);
}
if (iov != iovstack)
kfree(iov);
- return ret;
+ return count;
}
/*
@@ -1571,7 +1570,7 @@ static long vmsplice_to_user(struct file *file, const struct iovec __user *uiov,
* as splice-from-memory, where the regular splice is splice-from-file (or
* to file). In both cases the output is a pipe, naturally.
*/
-static long vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
+static ssize_t vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
unsigned long nr_segs, unsigned int flags)
{
struct pipe_inode_info *pipe;
@@ -1585,7 +1584,7 @@ static long vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
.ops = &user_page_pipe_buf_ops,
.spd_release = spd_release_page,
};
- long ret;
+ ssize_t ret;
pipe = get_pipe_info(file);
if (!pipe)
@@ -1626,25 +1625,25 @@ SYSCALL_DEFINE4(vmsplice, int, fd, const struct iovec __user *, iov,
unsigned long, nr_segs, unsigned int, flags)
{
struct fd f;
- long error;
+ ssize_t ret;
if (unlikely(nr_segs > UIO_MAXIOV))
return -EINVAL;
else if (unlikely(!nr_segs))
return 0;
- error = -EBADF;
+ ret = -EBADF;
f = fdget(fd);
if (f.file) {
if (f.file->f_mode & FMODE_WRITE)
- error = vmsplice_to_pipe(f.file, iov, nr_segs, flags);
+ ret = vmsplice_to_pipe(f.file, iov, nr_segs, flags);
else if (f.file->f_mode & FMODE_READ)
- error = vmsplice_to_user(f.file, iov, nr_segs, flags);
+ ret = vmsplice_to_user(f.file, iov, nr_segs, flags);
fdput(f);
}
- return error;
+ return ret;
}
#ifdef CONFIG_COMPAT
^ permalink raw reply related [flat|nested] 5+ messages in thread
end of thread, other threads:[~2014-05-27 19:25 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2014-05-27 19:25 [PATCH] vfs: fix vmplice_to_user() George Spelvin
-- strict thread matches above, loose matches on Subject: below --
2014-05-27 14:41 Miklos Szeredi
2014-05-27 17:04 ` Linus Torvalds
2014-05-27 17:57 ` Al Viro
2014-05-27 17:51 ` Al Viro
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).