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

From: YangYang
Date: Thu May 30 2024 - 06:56:19 EST


On 2024/5/30 9:49, Yu Kuai wrote:
Hi,

在 2024/05/27 12:26, Yang Yang 写道:
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
                                       */

This looks correct.


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.

Fix this issue by adding a new flag swap_inprogress to indicate whether
the swap is ongoing.

Fixes: 661d4f55a794 ("sbitmap: remove swap_lock")
Signed-off-by: Yang Yang <yang.yang@xxxxxxxx>
---
  include/linux/sbitmap.h |  5 +++++
  lib/sbitmap.c           | 22 ++++++++++++++++++++--
  2 files changed, 25 insertions(+), 2 deletions(-)

diff --git a/include/linux/sbitmap.h b/include/linux/sbitmap.h
index d662cf136021..b88a9e4997ab 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_inprogress: set to 1 when swapping word <-> cleared
+     */
+    atomic_t swap_inprogress;
  } ____cacheline_aligned_in_smp;
  /**
diff --git a/lib/sbitmap.c b/lib/sbitmap.c
index 1e453f825c05..d4bb258fe8b0 100644
--- a/lib/sbitmap.c
+++ b/lib/sbitmap.c
@@ -62,10 +62,19 @@ static inline void update_alloc_hint_after_get(struct sbitmap *sb,
   */
  static inline bool sbitmap_deferred_clear(struct sbitmap_word *map)
  {
-    unsigned long mask;
+    unsigned long mask, flags;
+    int zero = 0;
-    if (!READ_ONCE(map->cleared))
+    if (!READ_ONCE(map->cleared)) {
+        if (atomic_read(&map->swap_inprogress))
+            goto out_wait;
          return false;
+    }
+
+    if (!atomic_try_cmpxchg(&map->swap_inprogress, &zero, 1))
+        goto out_wait;
+
+    local_irq_save(flags);
      /*
       * First get a stable cleared mask, setting the old mask to 0.
@@ -77,6 +86,15 @@ static inline bool sbitmap_deferred_clear(struct sbitmap_word *map)
       */
      atomic_long_andnot(mask, (atomic_long_t *)&map->word);
      BUILD_BUG_ON(sizeof(atomic_long_t) != sizeof(map->word));
+
+    atomic_set(&map->swap_inprogress, 0);
+    smp_mb__after_atomic();
+    local_irq_restore(flags);
+    return true;
+
+out_wait:
+    while (atomic_read(&map->swap_inprogress))
+        ;

However, I really don't like this change.

How about following patch? I think it can fix the prblem as well.

The following patch cannot fix the problem.

Thanks.


diff --git a/lib/sbitmap.c b/lib/sbitmap.c
index 1e453f825c05..d1b9bd61d296 100644
--- a/lib/sbitmap.c
+++ b/lib/sbitmap.c
@@ -175,12 +175,14 @@ static int sbitmap_find_bit_in_word(struct sbitmap_word *map,
        int nr;

        do {
+               bool clear = READ_ONCE(map->cleared) != 0;
+
                nr = __sbitmap_get_word(&map->word, depth,
                                        alloc_hint, wrap);
-               if (nr != -1)
-                       break;
-               if (!sbitmap_deferred_clear(map))
+               if (nr != -1 || !clear)
                        break;
+
+               sbitmap_deferred_clear(map);
        } while (1);

        return nr;

Thanks,
Kuai
      return true;
  }