Re: [PATCH] mm/page_owner.c: use get_task_comm() to record task command name with the protection of task_lock()

[Qian Cai <quic_qiancai@quicinc.com>]

On Wed, Apr 20, 2022 at 08:28:17PM +0800, Yixuan Cao wrote:
> I noticed that it is advised to access task command name with
> [gs]et_task_comm() in the comment of task_struct->comm,
> which is safer with the protection of task_lock().
> 
> Relative comment in include/linux/sched.h is as follows:
> 
> /*
>  * executable name, excluding path.
>  *
>  * - normally initialized setup_new_exec()
>  * - access it with [gs]et_task_comm()
>  * - lock it with task_lock()
> */
> 
> Signed-off-by: Yixuan Cao <caoyixuan2019@email.szu.edu.cn>
> ---
>  mm/page_owner.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/mm/page_owner.c b/mm/page_owner.c
> index 2743062e92c2..bda8fe2660c0 100644
> --- a/mm/page_owner.c
> +++ b/mm/page_owner.c
> @@ -171,8 +171,7 @@ static inline void __set_page_owner_handle(struct page_ext *page_ext,
>  		page_owner->pid = current->pid;
>  		page_owner->tgid = current->tgid;
>  		page_owner->ts_nsec = local_clock();
> -		strlcpy(page_owner->comm, current->comm,
> -			sizeof(page_owner->comm));
> +		get_task_comm(page_owner->comm, current);

We can't call that thing here.

 WARNING: inconsistent lock state
 5.18.0-rc3-next-20220421-dirty #22 Not tainted
 --------------------------------
 inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage.
 swapper/4/0 [HC0[0]:SC1[1]:HE1:SE0] takes:
 ffff07ff89ad87f8 (&p->alloc_lock){+.?.}-{2:2}, at: __get_task_comm
 {SOFTIRQ-ON-W} state was registered at:
   __lock_acquire
   lock_acquire.part.0
   lock_acquire
   _raw_spin_lock
   __set_task_comm
   kthreadd
   ret_from_fork
 irq event stamp: 50532
 hardirqs last  enabled at (50532):  seqcount_lockdep_reader_access
 hardirqs last disabled at (50531):  seqcount_lockdep_reader_access
 softirqs last  enabled at (50306):  __do_softirq
 softirqs last disabled at (50313):  __irq_exit_rcu

  other info that might help us debug this:
  Possible unsafe locking scenario:

        CPU0
        ----
   lock(&p->alloc_lock);
   <Interrupt>
     lock(&p->alloc_lock);

                *** DEADLOCK ***

 1 lock held by swapper/4/0:
  #0: ffffce91c3d81da0 (rcu_callback){....}-{0:0}, at: rcu_do_batch

               stack backtrace:
 CPU: 4 PID: 0 Comm: swapper/4 Not tainted 5.18.0-rc3-next-20220421-dirty #22
 Call trace:
  dump_backtrace
  show_stack
  dump_stack_lvl
  dump_stack
  print_usage_bug.part.0
  mark_lock_irq
  mark_lock
  mark_usage
  __lock_acquire
  lock_acquire.part.0
  lock_acquire
  _raw_spin_lock
  __get_task_comm
  __get_task_comm at fs/exec.c:1221
  __set_page_owner
  arch___set_bit at include/asm-generic/bitops/non-atomic.h:22
  (inlined by) __set_page_owner_handle at mm/page_owner.c:175
  (inlined by) __set_page_owner at mm/page_owner.c:192
  post_alloc_hook
  get_page_from_freelist
  __alloc_pages
  alloc_pages
  allocate_slab
  new_slab
  ___slab_alloc
  __slab_alloc.constprop.0
  kmem_cache_alloc
  fill_pool
  __debug_object_init
  debug_object_activate
  call_rcu
  put_object
  __delete_object
  kmemleak_free
  slab_free_freelist_hook
  kmem_cache_free
  file_free_rcu
  rcu_do_batch
  rcu_core
  rcu_core_si
  __do_softirq
  __irq_exit_rcu
  irq_exit_rcu
  el1_interrupt
  el1h_64_irq_handler
  el1h_64_irq
  arch_local_irq_enable
  default_idle_call
  cpuidle_idle_call
  do_idle
  cpu_startup_entry
  secondary_start_kernel
  __secondary_switched