[PATCH] fuse: Fix memory leak in fuse_dev_free()

[PATCH] fuse: Fix memory leak in fuse_dev_free()

[Takeshi Misawa]

When ntfs is unmounted, the following leak is
reported by kmemleak.

kmemleak report:

unreferenced object 0xffff880052bf4400 (size 4096):
  comm "mount.ntfs", pid 16530, jiffies 4294861127 (age 3215.836s)
  hex dump (first 32 bytes):
    00 44 bf 52 00 88 ff ff 00 44 bf 52 00 88 ff ff  .D.R.....D.R....
    10 44 bf 52 00 88 ff ff 10 44 bf 52 00 88 ff ff  .D.R.....D.R....
  backtrace:
    [<00000000bf4a2f8d>] fuse_fill_super+0xb22/0x1da0 [fuse]
    [<000000004dde0f0c>] mount_bdev+0x263/0x320
    [<0000000025aebc66>] mount_fs+0x82/0x2bf
    [<0000000042c5a6be>] vfs_kern_mount.part.33+0xbf/0x480
    [<00000000ed10cd5b>] do_mount+0x3de/0x2ad0
    [<00000000d59ff068>] ksys_mount+0xba/0xd0
    [<000000001bda1bcc>] __x64_sys_mount+0xba/0x150
    [<00000000ebe26304>] do_syscall_64+0x151/0x490
    [<00000000d25f2b42>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
    [<000000002e0abd2c>] 0xffffffffffffffff

fuse_dev_alloc() allocate fud->pq.processing.
But this hash table is not freed.

Fix this by freeing fud->pq.processing.

Signed-off-by: Takeshi Misawa <jeliantsurux@gmail.com>
---
Dear Miklos Szeredi

I found kmemleak in fuse and try to create a patch.
Please consider this kmemleak and patch.

Regards.
---
 fs/fuse/inode.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
index 0b94b23b02d4..500aa59dedd2 100644
--- a/fs/fuse/inode.c
+++ b/fs/fuse/inode.c
@@ -1068,6 +1068,7 @@ void fuse_dev_free(struct fuse_dev *fud)
 
 		fuse_conn_put(fc);
 	}
+	kfree(fud->pq.processing);
 	kfree(fud);
 }
 EXPORT_SYMBOL_GPL(fuse_dev_free);
-- 
2.19.1

Re: [PATCH] fuse: Fix memory leak in fuse_dev_free()

[Miklos Szeredi]

On Sun, Dec 9, 2018 at 6:28 AM Takeshi Misawa <jeliantsurux@gmail.com> wrote:
>
> When ntfs is unmounted, the following leak is
> reported by kmemleak.
>
> kmemleak report:
>
> unreferenced object 0xffff880052bf4400 (size 4096):
>   comm "mount.ntfs", pid 16530, jiffies 4294861127 (age 3215.836s)
>   hex dump (first 32 bytes):
>     00 44 bf 52 00 88 ff ff 00 44 bf 52 00 88 ff ff  .D.R.....D.R....
>     10 44 bf 52 00 88 ff ff 10 44 bf 52 00 88 ff ff  .D.R.....D.R....
>   backtrace:
>     [<00000000bf4a2f8d>] fuse_fill_super+0xb22/0x1da0 [fuse]
>     [<000000004dde0f0c>] mount_bdev+0x263/0x320
>     [<0000000025aebc66>] mount_fs+0x82/0x2bf
>     [<0000000042c5a6be>] vfs_kern_mount.part.33+0xbf/0x480
>     [<00000000ed10cd5b>] do_mount+0x3de/0x2ad0
>     [<00000000d59ff068>] ksys_mount+0xba/0xd0
>     [<000000001bda1bcc>] __x64_sys_mount+0xba/0x150
>     [<00000000ebe26304>] do_syscall_64+0x151/0x490
>     [<00000000d25f2b42>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
>     [<000000002e0abd2c>] 0xffffffffffffffff
>
> fuse_dev_alloc() allocate fud->pq.processing.
> But this hash table is not freed.
>
> Fix this by freeing fud->pq.processing.
>
> Signed-off-by: Takeshi Misawa <jeliantsurux@gmail.com>
> ---
> Dear Miklos Szeredi
>
> I found kmemleak in fuse and try to create a patch.
> Please consider this kmemleak and patch.

Thanks, applied.

Miklos


>
> Regards.
> ---
>  fs/fuse/inode.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
> index 0b94b23b02d4..500aa59dedd2 100644
> --- a/fs/fuse/inode.c
> +++ b/fs/fuse/inode.c
> @@ -1068,6 +1068,7 @@ void fuse_dev_free(struct fuse_dev *fud)
>
>                 fuse_conn_put(fc);
>         }
> +       kfree(fud->pq.processing);
>         kfree(fud);
>  }
>  EXPORT_SYMBOL_GPL(fuse_dev_free);
> --
> 2.19.1
>