[PATCH] cgroup: Disable preemption for cgrp->freezer.freeze_seq when CONFIG_PREEMPT_RT=y.

[Kuniyuki Iwashima <kuniyu@google.com>]

syzbot reported the splat below. [0]

Commit afa3701c0e45 ("cgroup: cgroup.stat.local time accounting")
introduced cgrp->freezer.freeze_seq.

The writer side is under spin_lock_irq(), but the section is still
preemptible with CONFIG_PREEMPT_RT=y.

Let's wrap the section with preempt_{disable,enable}_nested().

[0]:
WARNING: CPU: 0 PID: 6076 at ./include/linux/seqlock.h:221 __seqprop_assert include/linux/seqlock.h:221 [inline]
WARNING: CPU: 0 PID: 6076 at ./include/linux/seqlock.h:221 cgroup_do_freeze kernel/cgroup/freezer.c:182 [inline]
WARNING: CPU: 0 PID: 6076 at ./include/linux/seqlock.h:221 cgroup_freeze+0x80a/0xf90 kernel/cgroup/freezer.c:309
Modules linked in:
CPU: 0 UID: 0 PID: 6076 Comm: syz.0.17 Not tainted syzkaller #0 PREEMPT_{RT,(full)}
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 08/18/2025
RIP: 0010:__seqprop_assert include/linux/seqlock.h:221 [inline]
RIP: 0010:cgroup_do_freeze kernel/cgroup/freezer.c:182 [inline]
RIP: 0010:cgroup_freeze+0x80a/0xf90 kernel/cgroup/freezer.c:309
Code: 90 e9 9e fb ff ff 44 89 f1 80 e1 07 80 c1 03 38 c1 0f 8c e7 f9 ff ff 4c 89 f7 e8 e1 43 67 00 e9 da f9 ff ff e8 17 68 06 00 90 <0f> 0b 90 e9 10 fc ff ff 44 89 f9 80 e1 07 38 c1 48 8b 0c 24 0f 8c
RSP: 0018:ffffc90003b178e0 EFLAGS: 00010293
RAX: ffffffff81b6c6a9 RBX: 0000000000000000 RCX: ffff88803671bc00
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
RBP: ffffc90003b17a70 R08: 0000000000000000 R09: 0000000000000000
R10: dffffc0000000000 R11: fffffbfff1d6d2a7 R12: dffffc0000000000
R13: 0000000000000000 R14: 0000000000000001 R15: ffff88803623a791
FS:  00005555915ae500(0000) GS:ffff888127017000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000001b33363fff CR3: 0000000023b98000 CR4: 00000000003526f0
Call Trace:
 <TASK>
 cgroup_freeze_write+0x156/0x1c0 kernel/cgroup/cgroup.c:4174
 cgroup_file_write+0x39b/0x740 kernel/cgroup/cgroup.c:4312
 kernfs_fop_write_iter+0x3b0/0x540 fs/kernfs/file.c:352
 new_sync_write fs/read_write.c:593 [inline]
 vfs_write+0x5d5/0xb40 fs/read_write.c:686
 ksys_write+0x14b/0x260 fs/read_write.c:738
 do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
 do_syscall_64+0xfa/0x3b0 arch/x86/entry/syscall_64.c:94
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f0dc3e9eec9
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ffd9b7b6198 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
RAX: ffffffffffffffda RBX: 00007f0dc40f5fa0 RCX: 00007f0dc3e9eec9
RDX: 0000000000000012 RSI: 0000200000000200 RDI: 0000000000000004
RBP: 00007f0dc3f21f91 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007f0dc40f5fa0 R14: 00007f0dc40f5fa0 R15: 0000000000000003
 </TASK>

Fixes: afa3701c0e45 ("cgroup: cgroup.stat.local time accounting")
Reported-by: syzbot+27a2519eb4dad86d0156@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/68de0b21.050a0220.25d7ab.077d.GAE@google.com/
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 kernel/cgroup/freezer.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/cgroup/freezer.c b/kernel/cgroup/freezer.c
index 6c18854bff34..7e779c8a6f89 100644
--- a/kernel/cgroup/freezer.c
+++ b/kernel/cgroup/freezer.c
@@ -179,6 +179,7 @@ static void cgroup_do_freeze(struct cgroup *cgrp, bool freeze, u64 ts_nsec)
 	lockdep_assert_held(&cgroup_mutex);
 
 	spin_lock_irq(&css_set_lock);
+	preempt_disable_nested();
 	write_seqcount_begin(&cgrp->freezer.freeze_seq);
 	if (freeze) {
 		set_bit(CGRP_FREEZE, &cgrp->flags);
@@ -189,6 +190,7 @@ static void cgroup_do_freeze(struct cgroup *cgrp, bool freeze, u64 ts_nsec)
 			cgrp->freezer.freeze_start_nsec);
 	}
 	write_seqcount_end(&cgrp->freezer.freeze_seq);
+	preempt_enable_nested();
 	spin_unlock_irq(&css_set_lock);
 
 	if (freeze)
-- 
2.51.0.618.g983fd99d29-goog