[PATCH 6.1.96] gfs2: Fix slab-use-after-free in gfs2_qd_dealloc

[PATCH 6.1.96] gfs2: Fix slab-use-after-free in gfs2_qd_dealloc

[Clayton Casciato]

[ Upstream commit bdcb8aa434c6d36b5c215d02a9ef07551be25a37 ]

In gfs2_put_super(), whether withdrawn or not, the quota should
be cleaned up by gfs2_quota_cleanup().

Otherwise, struct gfs2_sbd will be freed before gfs2_qd_dealloc (rcu
callback) has run for all gfs2_quota_data objects, resulting in
use-after-free.

Also, gfs2_destroy_threads() and gfs2_quota_cleanup() is already called
by gfs2_make_fs_ro(), so in gfs2_put_super(), after calling
gfs2_make_fs_ro(), there is no need to call them again.

The origin of a cherry-pick conflict is the (relevant) code block added in
commit f66af88e3321 ("gfs2: Stop using gfs2_make_fs_ro for withdraw")

There are no references to gfs2_withdrawn() nor gfs2_destroy_threads() in
gfs2_put_super(), so we can simply call gfs2_quota_cleanup() in a new else
block as bdcb8aa434c6 achieves.

Else braces were used for consistency with the if block.

Sponsor: 21SoftWare LLC
Signed-off-by: Clayton Casciato <majortomtosourcecontrol@gmail.com>
---
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index 302d1e43d701..6107cd680176 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -591,6 +591,8 @@ static void gfs2_put_super(struct super_block *sb)
 
 	if (!sb_rdonly(sb)) {
 		gfs2_make_fs_ro(sdp);
+	} else {
+		gfs2_quota_cleanup(sdp);
 	}
 	WARN_ON(gfs2_withdrawing(sdp));

Re: [PATCH 6.1.96] gfs2: Fix slab-use-after-free in gfs2_qd_dealloc

[Greg KH]

On Fri, Jun 28, 2024 at 12:07:52PM -0600, Clayton Casciato wrote:
> [ Upstream commit bdcb8aa434c6d36b5c215d02a9ef07551be25a37 ]
> 
> In gfs2_put_super(), whether withdrawn or not, the quota should
> be cleaned up by gfs2_quota_cleanup().
> 
> Otherwise, struct gfs2_sbd will be freed before gfs2_qd_dealloc (rcu
> callback) has run for all gfs2_quota_data objects, resulting in
> use-after-free.
> 
> Also, gfs2_destroy_threads() and gfs2_quota_cleanup() is already called
> by gfs2_make_fs_ro(), so in gfs2_put_super(), after calling
> gfs2_make_fs_ro(), there is no need to call them again.
> 
> The origin of a cherry-pick conflict is the (relevant) code block added in
> commit f66af88e3321 ("gfs2: Stop using gfs2_make_fs_ro for withdraw")
> 
> There are no references to gfs2_withdrawn() nor gfs2_destroy_threads() in
> gfs2_put_super(), so we can simply call gfs2_quota_cleanup() in a new else
> block as bdcb8aa434c6 achieves.
> 
> Else braces were used for consistency with the if block.
> 
> Sponsor: 21SoftWare LLC

That's not a valid tag for kernel commits, sorry.

> Signed-off-by: Clayton Casciato <majortomtosourcecontrol@gmail.com>

What happened to the original authorship information, and all of the
other signed-off-by that were on the original commit?  YOu can not just
delete them, would you want someone doing that to a patch you
contributed?

as-is, we can't take this, please fix up.

thanks,

greg k-h

Re: [PATCH 6.1.96] gfs2: Fix slab-use-after-free in gfs2_qd_dealloc

[Clayton Casciato]

On 6/29/24 2:10 AM, Greg KH wrote:
> On Fri, Jun 28, 2024 at 12:07:52PM -0600, Clayton Casciato wrote:
>> [ Upstream commit bdcb8aa434c6d36b5c215d02a9ef07551be25a37 ]
>>
>> In gfs2_put_super(), whether withdrawn or not, the quota should
>> be cleaned up by gfs2_quota_cleanup().
>>
>> Otherwise, struct gfs2_sbd will be freed before gfs2_qd_dealloc (rcu
>> callback) has run for all gfs2_quota_data objects, resulting in
>> use-after-free.
>>
>> Also, gfs2_destroy_threads() and gfs2_quota_cleanup() is already called
>> by gfs2_make_fs_ro(), so in gfs2_put_super(), after calling
>> gfs2_make_fs_ro(), there is no need to call them again.
>>
>> The origin of a cherry-pick conflict is the (relevant) code block added in
>> commit f66af88e3321 ("gfs2: Stop using gfs2_make_fs_ro for withdraw")
>>
>> There are no references to gfs2_withdrawn() nor gfs2_destroy_threads() in
>> gfs2_put_super(), so we can simply call gfs2_quota_cleanup() in a new else
>> block as bdcb8aa434c6 achieves.
>>
>> Else braces were used for consistency with the if block.
>>
>> Sponsor: 21SoftWare LLC
> 
> That's not a valid tag for kernel commits, sorry.
> 

The documentation mentions "some people also put extra tags at the end.
They’ll just be ignored for now [...]"

I don't imagine this would be appropriate on a line *after* the sign-off?

>> Signed-off-by: Clayton Casciato <majortomtosourcecontrol@gmail.com>
> 
> What happened to the original authorship information, and all of the
> other signed-off-by that were on the original commit?  YOu can not just
> delete them, would you want someone doing that to a patch you
> contributed?
> 

I didn't understand this and was concerned along the lines of "it is very
impolite to change one submitter’s code and make him endorse your bugs"

> as-is, we can't take this, please fix up.

I've submitted v2

> 
> thanks,
> 
> greg k-h

Thank you for the feedback!

Clayton Casciato