2007-06-22 17:13:37

by Doug Thompson

Subject: [PATCH 1/1] fs-sysfs bugfix of sysfs_hash_and_remove dereferencing before checking for NULL

From: Douglas Thompson <[email protected]>

In tracking down one of my bugs in using sysfs, I found the kernel doing a NULL de-reference
in function fs/sysfs/inode.c:sysfs_hash_and_remove(), when I (incorrectly) passed
in a dentry value of NULL.

The check for NULL should occur BEFORE the dereference of 'dir'.
This patch modifies the sequence and does the check FIRST


Signed-off-by: Douglas Thompson <[email protected]>
---

Index: linux-2.6.22-rc4-mm2/fs/sysfs/inode.c
===================================================================
--- linux-2.6.22-rc4-mm2.orig/fs/sysfs/inode.c
+++ linux-2.6.22-rc4-mm2/fs/sysfs/inode.c
@@ -285,7 +285,7 @@ void sysfs_drop_dentry(struct sysfs_dire
int sysfs_hash_and_remove(struct dentry * dir, const char * name)
{
struct sysfs_dirent **pos, *sd;
- struct sysfs_dirent *parent_sd = dir->d_fsdata;
+ struct sysfs_dirent *parent_sd;
int found = 0;

if (!dir)
@@ -295,6 +295,8 @@ int sysfs_hash_and_remove(struct dentry
/* no inode means this hasn't been made visible yet */
return -ENOENT;

+ parent_sd = dir->d_fsdata;
+
mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT);
for (pos = &parent_sd->s_children; *pos; pos = &(*pos)->s_sibling) {
sd = *pos;

2007-06-25 21:57:21

by Greg KH

Subject: Re: [PATCH 1/1] fs-sysfs bugfix of sysfs_hash_and_remove dereferencing before checking for NULL

On Fri, Jun 22, 2007 at 10:13:24AM -0700, Doug Thompson wrote:
> From: Douglas Thompson <[email protected]>
>
> In tracking down one of my bugs in using sysfs, I found the kernel doing a NULL de-reference
> in function fs/sysfs/inode.c:sysfs_hash_and_remove(), when I (incorrectly) passed
> in a dentry value of NULL.
>
> The check for NULL should occur BEFORE the dereference of 'dir'.
> This patch modifies the sequence and does the check FIRST
>
>
> Signed-off-by: Douglas Thompson <[email protected]>
> ---
>
> Index: linux-2.6.22-rc4-mm2/fs/sysfs/inode.c

There's been a lot of churn and change in this area and this patch
doens't apply anymore. Can you try respinning it against the next -mm
release and resending it if it is still needed?

thanks,

greg k-h