Re: [PATCH] cgroup/cpuset: fix panic caused by partcmd_update

[Waiman Long <longman@redhat.com>]

On 7/29/24 22:55, chenridong wrote:
>
>
> On 2024/7/30 10:34, Waiman Long wrote:
>> On 7/29/24 21:53, Chen Ridong wrote:
>>> We find a bug as below:
>>> BUG: unable to handle page fault for address: 00000003
>>> PGD 0 P4D 0
>>> Oops: 0000 [#1] PREEMPT SMP NOPTI
>>> CPU: 3 PID: 358 Comm: bash Tainted: G        W I 6.6.0-10893-g60d6
>>> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.15.0-1 
>>> 04/4
>>> RIP: 0010:partition_sched_domains_locked+0x483/0x600
>>> Code: 01 48 85 d2 74 0d 48 83 05 29 3f f8 03 01 f3 48 0f bc c2 89 c0 
>>> 48 9
>>> RSP: 0018:ffffc90000fdbc58 EFLAGS: 00000202
>>> RAX: 0000000100000003 RBX: ffff888100b3dfa0 RCX: 0000000000000000
>>> RDX: 0000000000000000 RSI: 0000000000000000 RDI: 000000000002fe80
>>> RBP: ffff888100b3dfb0 R08: 0000000000000001 R09: 0000000000000000
>>> R10: ffffc90000fdbcb0 R11: 0000000000000004 R12: 0000000000000002
>>> R13: ffff888100a92b48 R14: 0000000000000000 R15: 0000000000000000
>>> FS:  00007f44a5425740(0000) GS:ffff888237d80000(0000) 
>>> knlGS:0000000000000
>>> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>> CR2: 0000000100030973 CR3: 000000010722c000 CR4: 00000000000006e0
>>> Call Trace:
>>>   <TASK>
>>>   ? show_regs+0x8c/0xa0
>>>   ? __die_body+0x23/0xa0
>>>   ? __die+0x3a/0x50
>>>   ? page_fault_oops+0x1d2/0x5c0
>>>   ? partition_sched_domains_locked+0x483/0x600
>>>   ? search_module_extables+0x2a/0xb0
>>>   ? search_exception_tables+0x67/0x90
>>>   ? kernelmode_fixup_or_oops+0x144/0x1b0
>>>   ? __bad_area_nosemaphore+0x211/0x360
>>>   ? up_read+0x3b/0x50
>>>   ? bad_area_nosemaphore+0x1a/0x30
>>>   ? exc_page_fault+0x890/0xd90
>>>   ? __lock_acquire.constprop.0+0x24f/0x8d0
>>>   ? __lock_acquire.constprop.0+0x24f/0x8d0
>>>   ? asm_exc_page_fault+0x26/0x30
>>>   ? partition_sched_domains_locked+0x483/0x600
>>>   ? partition_sched_domains_locked+0xf0/0x600
>>>   rebuild_sched_domains_locked+0x806/0xdc0
>>>   update_partition_sd_lb+0x118/0x130
>>>   cpuset_write_resmask+0xffc/0x1420
>>>   cgroup_file_write+0xb2/0x290
>>>   kernfs_fop_write_iter+0x194/0x290
>>>   new_sync_write+0xeb/0x160
>>>   vfs_write+0x16f/0x1d0
>>>   ksys_write+0x81/0x180
>>>   __x64_sys_write+0x21/0x30
>>>   x64_sys_call+0x2f25/0x4630
>>>   do_syscall_64+0x44/0xb0
>>>   entry_SYSCALL_64_after_hwframe+0x78/0xe2
>>> RIP: 0033:0x7f44a553c887
>>>
>>> It can be reproduced with cammands:
>>> cd /sys/fs/cgroup/
>>> mkdir test
>>> cd test/
>>> echo +cpuset > ../cgroup.subtree_control
>>> echo root > cpuset.cpus.partition
>>> echo 0-3 > cpuset.cpus // 3 is nproc
>> What do you mean by "3 is nproc"? Are there only 3 CPUs in the 
>> system? What are the value of /sys/fs/cgroup/cpuset.cpu*?
> Yes, I tested it with qemu, only 3 cpus are available.
> # cat /sys/fs/cgroup/cpuset.cpus.effective
> 0-3
> This case is taking all cpus away from root, test should fail to be a 
> valid root, it should not rebuild scheduling domains.
I see. So there are 4 CPUs in the systems. So nproc should be 4. That is 
why I got confused when you said nproc is 3. I think you should clarify 
this in your patch.
>
>>>
>>> This issue is caused by the incorrect rebuilding of scheduling domains.
>>> In this scenario, test/cpuset.cpus.partition should be an invalid root
>>> and should not trigger the rebuilding of scheduling domains. When 
>>> calling
>>> update_parent_effective_cpumask with partcmd_update, if newmask is not
>>> null, it should recheck newmask whether there are cpus is available
>>> for parect/cs that has tasks.
>>>
>>> Fixes: 0c7f293efc87 ("cgroup/cpuset: Add 
>>> cpuset.cpus.exclusive.effective for v2")
>>> Signed-off-by: Chen Ridong <chenridong@huawei.com>
>>> ---
>>>   kernel/cgroup/cpuset.c | 2 ++
>>>   1 file changed, 2 insertions(+)
>>>
>>> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
>>> index 40ec4abaf440..a9b6d56eeffa 100644
>>> --- a/kernel/cgroup/cpuset.c
>>> +++ b/kernel/cgroup/cpuset.c
>>> @@ -1991,6 +1991,8 @@ static int 
>>> update_parent_effective_cpumask(struct cpuset *cs, int cmd,
>>>               part_error = PERR_CPUSEMPTY;
>>>               goto write_error;
>>>           }
>>> +        /* Check newmask again, whether cpus are available for 
>>> parent/cs */
>>> +        nocpu |= tasks_nocpu_error(parent, cs, newmask);
>>>           /*
>>>            * partcmd_update with newmask:
>>
>> The code change looks reasonable to me. However, I would like to know 
>> more about the reproduction steps.

I am OK with this patch other than missing some information in your 
reproduction step.

Cheers,
Longman