Skip to contentSkip to conversation

lib: switch cgroup bit-fields from signed to unsigned int

PATCH In review Li Wang 3 participants 5 messages Jan 18, 2025 – Jan 25, 2025 Library code

lore
Share & tools
lore

Li Wang

author
#0 lore raw

PATCH 1/1 lib: switch cgroup bit-fields from signed to unsigned int

There is a problem in cgroup lib that the declearation int can lead to -1 during the | operation.

Becasue if the field contains uninitialized garbage data, a bit-field declared as int could interpret 0b1 as -1 due to signed arithmetic.

By changing the type to unsigned int, the issue is avoided since unsigned fields cannot represent negative values.

---

Notes:
@Cyril, Petr, I vote to merge this patch before the release. @Guojie, Could you plz repost your memcontrol04 patch based on this change?

 include/tst_test.h |  2 +-
 lib/tst_cgroup.c   | 10 +++++-----
 2 files changed, 6 insertions(+), 6 deletions(-)
include/tst_test.h +1 -1
index 5b3889e64..eb73cd593 100644
@@ -610,7 +610,7 @@ struct tst_fs {
 
 	const char *const *needs_cgroup_ctrls;
 
-	int needs_cgroup_nsdelegate:1;
+	unsigned int needs_cgroup_nsdelegate:1;
 };
 
 /**
lib/tst_cgroup.c +5 -5
index 6055015eb..aa13ac8ec 100644
@@ -44,7 +44,7 @@ struct cgroup_dir {
 	 */
 	int dir_fd;
 
-	int we_created_it:1;
+	unsigned int we_created_it:1;
 };
 
 /* The root of a CGroup hierarchy/tree */
@@ -71,11 +71,11 @@ struct cgroup_root {
 	/* CGroup for current test. Which may have children. */
 	struct cgroup_dir test_dir;
 
-	int nsdelegate:1;
+	unsigned int nsdelegate:1;
 
-	int we_mounted_it:1;
+	unsigned int we_mounted_it:1;
 	/* cpuset is in compatability mode */
-	int no_cpuset_prefix:1;
+	unsigned int no_cpuset_prefix:1;
 };
 
 /* Controller sub-systems */
@@ -138,7 +138,7 @@ struct cgroup_ctrl {
 	/* Runtime; hierarchy the controller is attached to */
 	struct cgroup_root *ctrl_root;
 	/* Runtime; whether we required the controller */
-	int we_require_it:1;
+	unsigned int we_require_it:1;
 };
 
 struct tst_cg_group {
Signature · 5 lines
2.47.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v1 1/1
#1 lore raw

Hi Li,

Quoted 9 lines

There is a problem in cgroup lib that the declearation int can lead to -1 during the | operation.

Becasue if the field contains uninitialized garbage data, a bit-field declared as int could interpret 0b1 as -1 due to signed arithmetic.

By changing the type to unsigned int, the issue is avoided since unsigned fields cannot represent negative values.

Quoted 7 lines

---

Notes:
@Cyril, Petr, I vote to merge this patch before the release. @Guojie, Could you plz repost your memcontrol04 patch based on this change?

+1

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v1 1/1
#2 lore raw
In reply to Li Wang

Hi!

Quoted 6 lines

There is a problem in cgroup lib that the declearation int can lead to -1 during the | operation.

Becasue if the field contains uninitialized garbage data, a bit-field declared as int could interpret 0b1 as -1 due to signed arithmetic.

Uff, so we have a number that can be either -1 or 0. Where did it break? I suppose that things are okay as long as we do if (s->bar | s->bar).

Anyways:

Signature · 5 lines
Cyril Hrubis
[email protected]

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#3 lore raw
Quoted 10 lines · Cyril Hrubis

Hi!

Quoted 6 lines

There is a problem in cgroup lib that the declearation int can lead to -1 during the | operation.

Becasue if the field contains uninitialized garbage data, a bit-field declared as int could interpret 0b1 as -1 due to signed arithmetic.

Uff, so we have a number that can be either -1 or 0. Where did it break?

It doesn't break anything so far, I found that issue while reviewing Guojie's patch-v2, when that memory_recursiveprot declared as int, the test result is always wrong.

However, the root->nsdelegate avoids this problem by negating value twice:

static int cgroup_v2_nsdelegate(void)

{
    return !!roots[0].nsdelegate;
}

But I think we should fix this from the root.

Quoted 12 lines

I suppose that things are okay as long as we do if (s->bar | s->bar).

Anyways:

--
Cyril Hrubis
[email protected]

Signature · 5 lines
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#4 lore raw

Patch merged, thanks!

Quoted 1 line
Signature · 5 lines
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp