Audit_alloc_mark() assign pathname to audit_mark->path, on error path from fsnotify_add_inode_mark(), fsnotify_put_mark will free memory of audit_mark->path, but the caller of audit_alloc_mark will free the pathname again, so there will be double free problem. Fix this by resetting audit_mark->path to NULL pointer on error path from fsnotify_add_inode_mark(). Fixes: 7b1293234084d ("fsnotify: Add group pointer in fsnotify_init_mark()") Signed-off-by: Gaosheng Cui <cuigaosheng1(a)huawei.com&gt;

--- kernel/audit_fsnotify.c | 1 + 1 file changed, 1 insertion(+) diff --git a/kernel/audit_fsnotify.c b/kernel/audit_fsnotify.c index 6432a37ac1c9..c565fbf66ac8 100644 --- a/kernel/audit_fsnotify.c +++ b/kernel/audit_fsnotify.c @@ -102,6 +102,7 @@ struct audit_fsnotify_mark *audit_alloc_mark(struct audit_krule *krule, char *pa ret = fsnotify_add_inode_mark(&audit_mark->mark, inode, 0); if (ret < 0) { + audit_mark->path = NULL; fsnotify_put_mark(&audit_mark->mark); audit_mark = ERR_PTR(ret); } -- 2.25.1

On Mon, Aug 22, 2022 at 4:50 AM Jan Kara <jack(a)suse.cz&gt; wrote: > On Mon 22-08-22 10:29:05, Gaosheng Cui wrote: > > Audit_alloc_mark() assign pathname to audit_mark->path, on error path > > from fsnotify_add_inode_mark(), fsnotify_put_mark will free memory > > of audit_mark->path, but the caller of audit_alloc_mark will free > > the pathname again, so there will be double free problem. > > > > Fix this by resetting audit_mark->path to NULL pointer on error path > > from fsnotify_add_inode_mark(). > > > > Fixes: 7b1293234084d ("fsnotify: Add group pointer in fsnotify_init_mark()") > > Signed-off-by: Gaosheng Cui <cuigaosheng1(a)huawei.com&gt; > > Good spotting! The patch looks good to me. Feel free to add: > > Reviewed-by: Jan Kara <jack(a)suse.cz&gt; > > > --- > > kernel/audit_fsnotify.c | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/kernel/audit_fsnotify.c b/kernel/audit_fsnotify.c > > index 6432a37ac1c9..c565fbf66ac8 100644 > > --- a/kernel/audit_fsnotify.c > > +++ b/kernel/audit_fsnotify.c > > @@ -102,6 +102,7 @@ struct audit_fsnotify_mark *audit_alloc_mark(struct audit_krule *krule, char *pa > > > > ret = fsnotify_add_inode_mark(&audit_mark->mark, inode, 0); > > if (ret < 0) { > > + audit_mark->path = NULL; > > fsnotify_put_mark(&audit_mark->mark); As I'm tracing the code path from audit through fsnotify, and back into audit, I'm wondering if we still have a problem. When fsnotify_add_inode_mark() fails it will end up freeing not just audit_mark->path, but audit_mark itself via audit_fsnotify_mark_free() (via a call into fsnotify_put_mark()), yes?

On Mon, Aug 22, 2022 at 11:20 AM Jan Kara <jack(a)suse.cz&gt; wrote: > On Mon 22-08-22 10:34:15, Paul Moore wrote: > > On Mon, Aug 22, 2022 at 4:50 AM Jan Kara <jack(a)suse.cz&gt; wrote: > > > On Mon 22-08-22 10:29:05, Gaosheng Cui wrote: > > > > Audit_alloc_mark() assign pathname to audit_mark->path, on error path > > > > from fsnotify_add_inode_mark(), fsnotify_put_mark will free memory > > > > of audit_mark->path, but the caller of audit_alloc_mark will free > > > > the pathname again, so there will be double free problem. > > > > > > > > Fix this by resetting audit_mark->path to NULL pointer on error path > > > > from fsnotify_add_inode_mark(). > > > > > > > > Fixes: 7b1293234084d ("fsnotify: Add group pointer in fsnotify_init_mark()") > > > > Signed-off-by: Gaosheng Cui <cuigaosheng1(a)huawei.com&gt; > > > > > > Good spotting! The patch looks good to me. Feel free to add: > > > > > > Reviewed-by: Jan Kara <jack(a)suse.cz&gt; > > > > > > > --- > > > > kernel/audit_fsnotify.c | 1 + > > > > 1 file changed, 1 insertion(+) > > > > > > > > diff --git a/kernel/audit_fsnotify.c b/kernel/audit_fsnotify.c > > > > index 6432a37ac1c9..c565fbf66ac8 100644 > > > > --- a/kernel/audit_fsnotify.c > > > > +++ b/kernel/audit_fsnotify.c > > > > @@ -102,6 +102,7 @@ struct audit_fsnotify_mark *audit_alloc_mark(struct audit_krule *krule, char *pa > > > > > > > > ret = fsnotify_add_inode_mark(&audit_mark->mark, inode, 0); > > > > if (ret < 0) { > > > > + audit_mark->path = NULL; > > > > fsnotify_put_mark(&audit_mark->mark); > > > > As I'm tracing the code path from audit through fsnotify, and back > > into audit, I'm wondering if we still have a problem. When > > fsnotify_add_inode_mark() fails it will end up freeing not just > > audit_mark->path, but audit_mark itself via audit_fsnotify_mark_free() > > (via a call into fsnotify_put_mark()), yes? > > I don't think so. fsnotify_add_mark_locked() will call fsnotify_put_mark() > but that is just a counter part to fsnotify_get_mark() a few lines above. > The caller of fsnotify_add_inode_mark() still holds its own mark reference > which prevents mark from being freed. Okay, that sounds reasonable, but I'm still looking for a code path that only frees audit_mark:path and not the audit_mark itself. What am I not seeing?

On Mon, Aug 22, 2022 at 1:59 PM Jan Kara <jack(a)suse.cz&gt; wrote: > On Mon 22-08-22 13:39:23, Paul Moore wrote: >> On Mon, Aug 22, 2022 at 11:20 AM Jan Kara <jack(a)suse.cz&gt; wrote: >>> On Mon 22-08-22 10:34:15, Paul Moore wrote: >>>> On Mon, Aug 22, 2022 at 4:50 AM Jan Kara <jack(a)suse.cz&gt; wrote: >>>>> On Mon 22-08-22 10:29:05, Gaosheng Cui wrote: >>>>>> Audit_alloc_mark() assign pathname to audit_mark->path, on error path >>>>>> from fsnotify_add_inode_mark(), fsnotify_put_mark will free memory >>>>>> of audit_mark->path, but the caller of audit_alloc_mark will free >>>>>> the pathname again, so there will be double free problem. >>>>>> >>>>>> Fix this by resetting audit_mark->path to NULL pointer on error path >>>>>> from fsnotify_add_inode_mark(). >>>>>> >>>>>> Fixes: 7b1293234084d ("fsnotify: Add group pointer in fsnotify_init_mark()") >>>>>> Signed-off-by: Gaosheng Cui <cuigaosheng1(a)huawei.com&gt; >>>>> Good spotting! The patch looks good to me. Feel free to add: >>>>> >>>>> Reviewed-by: Jan Kara <jack(a)suse.cz&gt; >>>>> >>>>>> --- >>>>>> kernel/audit_fsnotify.c | 1 + >>>>>> 1 file changed, 1 insertion(+) >>>>>> >>>>>> diff --git a/kernel/audit_fsnotify.c b/kernel/audit_fsnotify.c >>>>>> index 6432a37ac1c9..c565fbf66ac8 100644 >>>>>> --- a/kernel/audit_fsnotify.c >>>>>> +++ b/kernel/audit_fsnotify.c >>>>>> @@ -102,6 +102,7 @@ struct audit_fsnotify_mark *audit_alloc_mark(struct audit_krule *krule, char *pa >>>>>> >>>>>> ret = fsnotify_add_inode_mark(&audit_mark->mark, inode, 0); >>>>>> if (ret < 0) { >>>>>> + audit_mark->path = NULL; >>>>>> fsnotify_put_mark(&audit_mark->mark); >>>> As I'm tracing the code path from audit through fsnotify, and back >>>> into audit, I'm wondering if we still have a problem. When >>>> fsnotify_add_inode_mark() fails it will end up freeing not just >>>> audit_mark->path, but audit_mark itself via audit_fsnotify_mark_free() >>>> (via a call into fsnotify_put_mark()), yes? >>> I don't think so. fsnotify_add_mark_locked() will call fsnotify_put_mark() >>> but that is just a counter part to fsnotify_get_mark() a few lines above. >>> The caller of fsnotify_add_inode_mark() still holds its own mark reference >>> which prevents mark from being freed. >> Okay, that sounds reasonable, but I'm still looking for a code path >> that only frees audit_mark:path and not the audit_mark itself. What >> am I not seeing? > The callers of audit_alloc_mark() call kfree(path) if audit_alloc_mark() > returns error (which is a sensible thing because in some cases path indeed > needs freeing). Of course! Thanks for that, I think I got a bit of tunnel vision on this for some reason. I'll merge this into audit/stable-6.0 now and once testing is complete I'll send it up to Linus. Thanks everyone!