Calling pipe2() with O_NOTIFICATION_PIPE could results in memory leaks
in an error path or CONFIG_WATCH_QUEUE=n. Plug them.
unreferenced object 0xc00000141114a0d8 (size 992):
comm "trinity-c61", pid 1353192, jiffies 4296255779 (age 25989.560s)
hex dump (first 32 bytes):
80 11 00 00 e8 03 00 00 00 00 00 00 00 00 00 00 ................
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
backtrace:
[<00000000abff13d7>] kmem_cache_alloc+0x1b4/0x470
[<000000009502e5d5>] alloc_inode+0xd0/0x130
[<00000000ca1c1a21>] new_inode_pseudo+0x1c/0x80
new_inode_pseudo at fs/inode.c:932
[<000000000c01d1d6>] create_pipe_files+0x48/0x2d0
get_pipe_inode at fs/pipe.c:874
(inlined by) create_pipe_files at fs/pipe.c:914
[<00000000d13ff4c4>] __do_pipe_flags+0x50/0x120
__do_pipe_flags at fs/pipe.c:965
[<0000000003941e42>] do_pipe2+0x3c/0x100
do_pipe2 at fs/pipe.c:1013
[<00000000a006b818>] sys_pipe2+0x1c/0x30
__se_sys_pipe2 at fs/pipe.c:1028
[<00000000a6925b55>] system_call_exception+0xf8/0x1d0
[<000000001c6b0740>] system_call_common+0xe8/0x218
unreferenced object 0xc000001f575ce600 (size 512):
comm "trinity-c61", pid 1353192, jiffies 4296255779 (age 25989.560s)
hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 ad 4e ad de .............N..
ff ff ff ff 00 00 00 00 ff ff ff ff ff ff ff ff ................
backtrace:
[<00000000d74d5e3a>] kmem_cache_alloc_trace+0x1c4/0x2d0
[<0000000061cbc9cb>] alloc_pipe_info+0x88/0x2c0
kmalloc at include/linux/slab.h:554
(inlined by) kzalloc at include/linux/slab.h:666
(inlined by) alloc_pipe_info at fs/pipe.c:793
[<00000000efd6129c>] create_pipe_files+0x6c/0x2d0
get_pipe_inode at fs/pipe.c:883
(inlined by) create_pipe_files at fs/pipe.c:914
[<00000000d13ff4c4>] __do_pipe_flags+0x50/0x120
[<0000000003941e42>] do_pipe2+0x3c/0x100
[<00000000a006b818>] sys_pipe2+0x1c/0x30
[<00000000a6925b55>] system_call_exception+0xf8/0x1d0
[<000000001c6b0740>] system_call_common+0xe8/0x218
unreferenced object 0xc000000d94f20400 (size 1024):
comm "trinity-c61", pid 1353192, jiffies 4296255779 (age 25989.560s)
hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<00000000e60ee00f>] __kmalloc+0x1e4/0x330
[<00000000130e8cc8>] alloc_pipe_info+0x154/0x2c0
kmalloc_array at include/linux/slab.h:594
(inlined by) kcalloc at include/linux/slab.h:605
(inlined by) alloc_pipe_info at fs/pipe.c:810
[<00000000efd6129c>] create_pipe_files+0x6c/0x2d0
[<00000000d13ff4c4>] __do_pipe_flags+0x50/0x120
[<0000000003941e42>] do_pipe2+0x3c/0x100
[<00000000a006b818>] sys_pipe2+0x1c/0x30
[<00000000a6925b55>] system_call_exception+0xf8/0x1d0
[<000000001c6b0740>] system_call_common+0xe8/0x218
Fixes: c73be61cede5 ("pipe: Add general notification queue support")
Signed-off-by: Qian Cai <[email protected]>
---
fs/pipe.c | 11 +++++------
include/linux/watch_queue.h | 6 ++++++
2 files changed, 11 insertions(+), 6 deletions(-)
diff --git a/fs/pipe.c b/fs/pipe.c
index 60dbee457143..f5d74ba1bf8c 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -913,19 +913,18 @@ int create_pipe_files(struct file **res, int flags)
{
struct inode *inode = get_pipe_inode();
struct file *f;
+ int error;
if (!inode)
return -ENFILE;
if (flags & O_NOTIFICATION_PIPE) {
-#ifdef CONFIG_WATCH_QUEUE
- if (watch_queue_init(inode->i_pipe) < 0) {
+ error = watch_queue_init(inode->i_pipe);
+ if (error) {
+ free_pipe_info(inode->i_pipe);
iput(inode);
- return -ENOMEM;
+ return error;
}
-#else
- return -ENOPKG;
-#endif
}
f = alloc_file_pseudo(inode, pipe_mnt, "",
diff --git a/include/linux/watch_queue.h b/include/linux/watch_queue.h
index 5e08db2adc31..c994d1b2cdba 100644
--- a/include/linux/watch_queue.h
+++ b/include/linux/watch_queue.h
@@ -122,6 +122,12 @@ static inline void remove_watch_list(struct watch_list *wlist, u64 id)
*/
#define watch_sizeof(STRUCT) (sizeof(STRUCT) << WATCH_INFO_LENGTH__SHIFT)
+#else
+static inline int watch_queue_init(struct pipe_inode_info *pipe)
+{
+ return -ENOPKG;
+}
+
#endif
#endif /* _LINUX_WATCH_QUEUE_H */
base-commit: 60e720931556fc1034d0981460164dcf02697679
--
2.28.0
On Thu, Oct 01, 2020 at 08:50:55AM -0400, Qian Cai wrote:
> Calling pipe2() with O_NOTIFICATION_PIPE could results in memory leaks
> in an error path or CONFIG_WATCH_QUEUE=n. Plug them.
[snip the copy of bug report]
No objections on the patch itself, but commit message is just about
unreadable. How about something along the lines of the following?
=======================
Calling pipe2() with O_NOTIFICATION_PIPE could results in memory
leaks unless watch_queue_init() is successful.
In case of watch_queue_init() failure in pipe2() we are left
with inode and pipe_inode_info instances that need to be freed. That
failure exit has been introduced in commit c73be61cede5 ("pipe: Add
general notification queue support") and its handling should've been
identical to nearby treatment of alloc_file_pseudo() failures - it
is dealing with the same situation. As it is, the mainline kernel
leaks in that case.
Another problem is that CONFIG_WATCH_QUEUE and !CONFIG_WATCH_QUEUE
cases are treated differently (and the former leaks just pipe_inode_info,
the latter - both pipe_inode_info and inode).
Fixed by providing a dummy wath_queue_init() in !CONFIG_WATCH_QUEUE
case and by having failures of wath_queue_init() handled the same way
we handle alloc_file_pseudo() ones.
Fixes: c73be61cede5 ("pipe: Add general notification queue support")
Signed-off-by: Qian Cai <[email protected]>
=======================
On Thu, 2020-10-01 at 14:16 +0100, Al Viro wrote:
> On Thu, Oct 01, 2020 at 08:50:55AM -0400, Qian Cai wrote:
> > Calling pipe2() with O_NOTIFICATION_PIPE could results in memory leaks
> > in an error path or CONFIG_WATCH_QUEUE=n. Plug them.
>
> [snip the copy of bug report]
>
> No objections on the patch itself, but commit message is just about
> unreadable. How about something along the lines of the following?
>
> =======================
> Calling pipe2() with O_NOTIFICATION_PIPE could results in memory
> leaks unless watch_queue_init() is successful.
>
> In case of watch_queue_init() failure in pipe2() we are left
> with inode and pipe_inode_info instances that need to be freed. That
> failure exit has been introduced in commit c73be61cede5 ("pipe: Add
> general notification queue support") and its handling should've been
> identical to nearby treatment of alloc_file_pseudo() failures - it
> is dealing with the same situation. As it is, the mainline kernel
> leaks in that case.
>
> Another problem is that CONFIG_WATCH_QUEUE and !CONFIG_WATCH_QUEUE
> cases are treated differently (and the former leaks just pipe_inode_info,
> the latter - both pipe_inode_info and inode).
>
> Fixed by providing a dummy wath_queue_init() in !CONFIG_WATCH_QUEUE
> case and by having failures of wath_queue_init() handled the same way
> we handle alloc_file_pseudo() ones.
>
> Fixes: c73be61cede5 ("pipe: Add general notification queue support")
> Signed-off-by: Qian Cai <[email protected]>
> =======================
Thanks Al. This looks very good to me.
On Thu, Oct 01, 2020 at 09:37:02AM -0400, Qian Cai wrote:
> > Fixed by providing a dummy wath_queue_init() in !CONFIG_WATCH_QUEUE
watch_queue_init(), that is
> > case and by having failures of wath_queue_init() handled the same way
ditto
> > we handle alloc_file_pseudo() ones.
> >
> > Fixes: c73be61cede5 ("pipe: Add general notification queue support")
> > Signed-off-by: Qian Cai <[email protected]>
> > =======================
>
> Thanks Al. This looks very good to me.
Applied with that commit message (and typos above corrected) to #fixes
and pushed