It might be a kernel disaster if one sysfs entry is freed but
still referenced by sysfs tree.
Recently Dave and Sasha reported one use-after-free problem on
sysfs entry, and the problem has been troubleshooted with help
of debug message added in this patch.
Given sysfs_get_dirent/sysfs_put are exported APIs, even inside
sysfs they are called in many contexts(kobject/attribe add/delete,
inode init/drop, dentry lookup/release, readdir, ...), it is healthful
to check the removed flag before freeing one entry and dump message
if it is freeing without being removed first.
Cc: Dave Jones <da...@redhat.com>
Cc: Sasha Levin <levinsasha...@gmail.com>
Signed-off-by: Ming Lei <ming....@canonical.com>
---
v1:
- use WARN() instead of BUG() as suggested by Dave and Greg
fs/sysfs/dir.c | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/fs/sysfs/dir.c b/fs/sysfs/dir.c
index 1bf016b..e8e0e71 100644
--- a/fs/sysfs/dir.c
+++ b/fs/sysfs/dir.c
@@ -268,6 +268,10 @@ void release_sysfs_dirent(struct sysfs_dirent * sd)
*/
parent_sd = sd->s_parent;
+ WARN(!(sd->s_flags & SYSFS_FLAG_REMOVED),
+ "sysfs: free using entry: %s/%s\n",
+ parent_sd ? parent_sd->s_name : "", sd->s_name);
+
if (sysfs_type(sd) == SYSFS_KOBJ_LINK)
sysfs_put(sd->s_symlink.target_sd);
if (sysfs_type(sd) & SYSFS_COPY_NAME)
@@ -386,7 +390,7 @@ struct sysfs_dirent *sysfs_new_dirent(const char *name,
umode_t mode, int type)
sd->s_name = name;
sd->s_mode = mode;
- sd->s_flags = type;
+ sd->s_flags = type | SYSFS_FLAG_REMOVED;
return sd;
@@ -466,6 +470,9 @@ int __sysfs_add_one(struct sysfs_addrm_cxt *acxt, struct
sysfs_dirent *sd)
ps_iattrs->ia_ctime = ps_iattrs->ia_mtime = CURRENT_TIME;
}
+ /* Mark the entry added into directory tree */
+ sd->s_flags &= ~SYSFS_FLAG_REMOVED;
+
return 0;
}
--
1.7.9.5
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
