* @round_robin: Allocate bits in strict round-robin order.
*/
bool round_robin;
+
+ /**
+ * @min_shallow_depth: The minimum shallow depth which may be passed to
+ * sbitmap_queue_get_shallow() or __sbitmap_queue_get_shallow().
+ */
+ unsigned int min_shallow_depth;
};
/**
* @shallow_depth: The maximum number of bits to allocate from a single word.
* See sbitmap_get_shallow().
*
+ * If you call this, make sure to call sbitmap_queue_min_shallow_depth() after
+ * initializing @sbq.
+ *
* Return: Non-negative allocated bit number if successful, -1 otherwise.
*/
int __sbitmap_queue_get_shallow(struct sbitmap_queue *sbq,
* @shallow_depth: The maximum number of bits to allocate from a single word.
* See sbitmap_get_shallow().
*
+ * If you call this, make sure to call sbitmap_queue_min_shallow_depth() after
+ * initializing @sbq.
+ *
* Return: Non-negative allocated bit number if successful, -1 otherwise.
*/
static inline int sbitmap_queue_get_shallow(struct sbitmap_queue *sbq,
return nr;
}
+/**
+ * sbitmap_queue_min_shallow_depth() - Inform a &struct sbitmap_queue of the
+ * minimum shallow depth that will be used.
+ * @sbq: Bitmap queue in question.
+ * @min_shallow_depth: The minimum shallow depth that will be passed to
+ * sbitmap_queue_get_shallow() or __sbitmap_queue_get_shallow().
+ *
+ * sbitmap_queue_clear() batches wakeups as an optimization. The batch size
+ * depends on the depth of the bitmap. Since the shallow allocation functions
+ * effectively operate with a different depth, the shallow depth must be taken
+ * into account when calculating the batch size. This function must be called
+ * with the minimum shallow depth that will be used. Failure to do so can result
+ * in missed wakeups.
+ */
+void sbitmap_queue_min_shallow_depth(struct sbitmap_queue *sbq,
+ unsigned int min_shallow_depth);
+
/**
* sbitmap_queue_clear() - Free an allocated bit and wake up waiters on a
* &struct sbitmap_queue.
}
EXPORT_SYMBOL_GPL(sbitmap_bitmap_show);
-static unsigned int sbq_calc_wake_batch(unsigned int depth)
+static unsigned int sbq_calc_wake_batch(struct sbitmap_queue *sbq,
+ unsigned int depth)
{
unsigned int wake_batch;
+ unsigned int shallow_depth;
/*
* For each batch, we wake up one queue. We need to make sure that our
- * batch size is small enough that the full depth of the bitmap is
- * enough to wake up all of the queues.
+ * batch size is small enough that the full depth of the bitmap,
+ * potentially limited by a shallow depth, is enough to wake up all of
+ * the queues.
+ *
+ * Each full word of the bitmap has bits_per_word bits, and there might
+ * be a partial word. There are depth / bits_per_word full words and
+ * depth % bits_per_word bits left over. In bitwise arithmetic:
+ *
+ * bits_per_word = 1 << shift
+ * depth / bits_per_word = depth >> shift
+ * depth % bits_per_word = depth & ((1 << shift) - 1)
+ *
+ * Each word can be limited to sbq->min_shallow_depth bits.
*/
- wake_batch = SBQ_WAKE_BATCH;
- if (wake_batch > depth / SBQ_WAIT_QUEUES)
- wake_batch = max(1U, depth / SBQ_WAIT_QUEUES);
+ shallow_depth = min(1U << sbq->sb.shift, sbq->min_shallow_depth);
+ depth = ((depth >> sbq->sb.shift) * shallow_depth +
+ min(depth & ((1U << sbq->sb.shift) - 1), shallow_depth));
+ wake_batch = clamp_t(unsigned int, depth / SBQ_WAIT_QUEUES, 1,
+ SBQ_WAKE_BATCH);
return wake_batch;
}
*per_cpu_ptr(sbq->alloc_hint, i) = prandom_u32() % depth;
}
- sbq->wake_batch = sbq_calc_wake_batch(depth);
+ sbq->min_shallow_depth = UINT_MAX;
+ sbq->wake_batch = sbq_calc_wake_batch(sbq, depth);
atomic_set(&sbq->wake_index, 0);
sbq->ws = kzalloc_node(SBQ_WAIT_QUEUES * sizeof(*sbq->ws), flags, node);
}
EXPORT_SYMBOL_GPL(sbitmap_queue_init_node);
-void sbitmap_queue_resize(struct sbitmap_queue *sbq, unsigned int depth)
+static void sbitmap_queue_update_wake_batch(struct sbitmap_queue *sbq,
+ unsigned int depth)
{
- unsigned int wake_batch = sbq_calc_wake_batch(depth);
+ unsigned int wake_batch = sbq_calc_wake_batch(sbq, depth);
int i;
if (sbq->wake_batch != wake_batch) {
for (i = 0; i < SBQ_WAIT_QUEUES; i++)
atomic_set(&sbq->ws[i].wait_cnt, 1);
}
+}
+
+void sbitmap_queue_resize(struct sbitmap_queue *sbq, unsigned int depth)
+{
+ sbitmap_queue_update_wake_batch(sbq, depth);
sbitmap_resize(&sbq->sb, depth);
}
EXPORT_SYMBOL_GPL(sbitmap_queue_resize);
}
EXPORT_SYMBOL_GPL(__sbitmap_queue_get_shallow);
+void sbitmap_queue_min_shallow_depth(struct sbitmap_queue *sbq,
+ unsigned int min_shallow_depth)
+{
+ sbq->min_shallow_depth = min_shallow_depth;
+ sbitmap_queue_update_wake_batch(sbq, sbq->sb.depth);
+}
+EXPORT_SYMBOL_GPL(sbitmap_queue_min_shallow_depth);
+
static struct sbq_wait_state *sbq_wake_ptr(struct sbitmap_queue *sbq)
{
int i, wake_index;
seq_puts(m, "}\n");
seq_printf(m, "round_robin=%d\n", sbq->round_robin);
+ seq_printf(m, "min_shallow_depth=%u\n", sbq->min_shallow_depth);
}
EXPORT_SYMBOL_GPL(sbitmap_queue_show);