Re: [PATCH v2] sbitmap: fix io hung due to race on sbitmap_word::cleared

[Ming Lei <ming.lei@redhat.com>]

On Thu, Jun 06, 2024 at 04:55:17PM +0800, YangYang wrote:
> On 2024/6/4 14:12, Yu Kuai wrote:
> > Hi,
> > 
> > 在 2024/06/04 11:25, Ming Lei 写道:
> > > On Tue, Jun 4, 2024 at 11:12 AM Yang Yang <yang.yang@vivo.com> wrote:
> > > > 
> > > > Configuration for sbq:
> > > >    depth=64, wake_batch=6, shift=6, map_nr=1
> > > > 
> > > > 1. There are 64 requests in progress:
> > > >    map->word = 0xFFFFFFFFFFFFFFFF
> > > > 2. After all the 64 requests complete, and no more requests come:
> > > >    map->word = 0xFFFFFFFFFFFFFFFF, map->cleared = 0xFFFFFFFFFFFFFFFF
> > > > 3. Now two tasks try to allocate requests:
> > > >    T1:                                       T2:
> > > >    __blk_mq_get_tag                          .
> > > >    __sbitmap_queue_get                       .
> > > >    sbitmap_get                               .
> > > >    sbitmap_find_bit                          .
> > > >    sbitmap_find_bit_in_word                  .
> > > >    __sbitmap_get_word  -> nr=-1              __blk_mq_get_tag
> > > >    sbitmap_deferred_clear                    __sbitmap_queue_get
> > > >    /* map->cleared=0xFFFFFFFFFFFFFFFF */     sbitmap_find_bit
> > > >      if (!READ_ONCE(map->cleared))           sbitmap_find_bit_in_word
> > > >        return false;                         __sbitmap_get_word -> nr=-1
> > > >      mask = xchg(&map->cleared, 0)           sbitmap_deferred_clear
> > > >      atomic_long_andnot()                    /* map->cleared=0 */
> > > >                                                if (!(map->cleared))
> > > >                                                  return false;
> > > >                                       /*
> > > >                                        * map->cleared is cleared by T1
> > > >                                        * T2 fail to acquire the tag
> > > >                                        */
> > > > 
> > > > 4. T2 is the sole tag waiter. When T1 puts the tag, T2 cannot be woken
> > > > up due to the wake_batch being set at 6. If no more requests come, T1
> > > > will wait here indefinitely.
> > > > 
> > > > To fix this issue, simply revert commit 661d4f55a794 ("sbitmap:
> > > > remove swap_lock"), which causes this issue.
> > > 
> > > I'd suggest to add the following words in commit log:
> > > 
> > > Check on ->cleared and update on both ->cleared and ->word need to be
> > > done atomically, and using spinlock could be the simplest solution.
> > > 
> > > Otherwise, the patch looks fine for me.
> > 
> > Maybe I'm noob, but I'm confused how can this fix the problem, looks
> > like the race condition doesn't change.
> > 
> > In sbitmap_find_bit_in_word:
> > 
> > 1) __sbitmap_get_word read word;
> > 2) sbitmap_deferred_clear clear cleared;
> > 3) sbitmap_deferred_clear update word;
> > 
> > 2) and 3) are done atomically while 1) can still concurrent with 3):
> > 
> > t1:
> > sbitmap_find_bit_in_word
> >   __sbitmap_get_word
> >   -> read old word, return -1
> >          t2:
> >          sbitmap_find_bit_in_word
> >           __sbitmap_get_word
> >           -> read old word, return -1
> >   sbitmap_deferred_clear
> >   -> clear cleared and update word
> >          sbitmap_deferred_clear
> >          -> cleared is cleared, fail
> > 
> > BYW, I still think it's fine to fix this problem by trying the
> > __sbitmap_get_word() at least one more time if __sbitmap_get_word()
> > failed.
> 
> How about this one:
> 1. Add extra check in sbitmap_find_bit_in_word() referenced from
>    Yu kuai's suggestion.
> 2. Change from atomic_long_andnot to atomic_long_fetch_andnot_release
> 
> ---
>  include/linux/sbitmap.h |  5 +++++
>  lib/sbitmap.c           | 23 ++++++++++++++++++-----
>  2 files changed, 23 insertions(+), 5 deletions(-)
> 
> diff --git a/include/linux/sbitmap.h b/include/linux/sbitmap.h
> index d662cf136021..ec0b0e73c906 100644
> --- a/include/linux/sbitmap.h
> +++ b/include/linux/sbitmap.h
> @@ -36,6 +36,11 @@ struct sbitmap_word {
>      * @cleared: word holding cleared bits
>      */
>     unsigned long cleared ____cacheline_aligned_in_smp;
> +
> +   /**
> +    * @swap_lock: Held while swapping word <-> cleared
> +    */
> +   spinlock_t swap_lock;
>  } ____cacheline_aligned_in_smp;
> 
>  /**
> diff --git a/lib/sbitmap.c b/lib/sbitmap.c
> index 1e453f825c05..63dadf91e40b 100644
> --- a/lib/sbitmap.c
> +++ b/lib/sbitmap.c
> @@ -63,9 +63,13 @@ static inline void update_alloc_hint_after_get(struct sbitmap *sb,
>  static inline bool sbitmap_deferred_clear(struct sbitmap_word *map)
>  {
>     unsigned long mask;
> +   bool ret = false;
> +   unsigned long flags;
> 
> -   if (!READ_ONCE(map->cleared))
> -       return false;
> +   spin_lock_irqsave(&map->swap_lock, flags);
> +
> +   if (!map->cleared)
> +       goto out_unlock;
> 
>     /*
>      * First get a stable cleared mask, setting the old mask to 0.
> @@ -75,9 +79,12 @@ static inline bool sbitmap_deferred_clear(struct sbitmap_word *map)
>     /*
>      * Now clear the masked bits in our free word
>      */
> -   atomic_long_andnot(mask, (atomic_long_t *)&map->word);
> +   atomic_long_fetch_andnot_release(mask, (atomic_long_t *)&map->word);
>     BUILD_BUG_ON(sizeof(atomic_long_t) != sizeof(map->word));
> -   return true;
> +   ret = true;
> +out_unlock:
> +   spin_unlock_irqrestore(&map->swap_lock, flags);
> +   return ret;
>  }
> 
>  int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
> @@ -85,6 +92,7 @@ int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
>               bool alloc_hint)
>  {
>     unsigned int bits_per_word;
> +   int i;
> 
>     if (shift < 0)
>         shift = sbitmap_calculate_shift(depth);
> @@ -116,6 +124,9 @@ int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
>         return -ENOMEM;
>     }
> 
> +   for (i = 0; i < sb->map_nr; i++)
> +       spin_lock_init(&sb->map[i].swap_lock);
> +
>     return 0;
>  }
>  EXPORT_SYMBOL_GPL(sbitmap_init_node);
> @@ -175,11 +186,13 @@ static int sbitmap_find_bit_in_word(struct sbitmap_word *map,
>     int nr;
> 
>     do {
> +       unsigned long cleared = READ_ONCE(map->cleared);
> +

->cleared is stored in standalone cacheline, the above line adds one extra l1
load, so I still prefer to check ->word & ->cleared in sbitmap_deferred_clear().

Thanks, 
Ming