On 2023/9/22 07:31, Wenjia Zhang wrote:
<...>
+static int smc_lo_register_dmb(struct smcd_dev *smcd, struct smcd_dmb *dmb,
+ void *client_priv)
+{
+ struct smc_lo_dmb_node *dmb_node, *tmp_node;
+ struct smc_lo_dev *ldev = smcd->priv;
+ int sba_idx, rc;
+
+ /* check space for new dmb */
+ for_each_clear_bit(sba_idx, ldev->sba_idx_mask, SMC_LODEV_MAX_DMBS) {
+ if (!test_and_set_bit(sba_idx, ldev->sba_idx_mask))
+ break;
+ }
+ if (sba_idx == SMC_LODEV_MAX_DMBS)
+ return -ENOSPC;
+
+ dmb_node = kzalloc(sizeof(*dmb_node), GFP_KERNEL);
+ if (!dmb_node) {
+ rc = -ENOMEM;
+ goto err_bit;
+ }
+
+ dmb_node->sba_idx = sba_idx;
+ dmb_node->cpu_addr = kzalloc(dmb->dmb_len, GFP_KERNEL |
+ __GFP_NOWARN | __GFP_NORETRY |
+ __GFP_NOMEMALLOC);
kzalloc()/kmalloc() allocates physically contigueous memory. Are you sure it is suitable for allocating the dmb?
Yes, physically contigueous memory is little expensive here. I initially wanted to see the best performance.
I tried using vzalloc here, and the performance dropped a bit (2%~8%) compared to kzalloc. I think it is acceptable.
- ipc-benchmark
kzalloc vzalloc
Message
rate (msg/s) 152076 145753(-4.16%)
- sockperf
kzalloc vzalloc
Bandwidth(MBps) 8491.638 8002.380(-5.76%)
Latency(us) 3.222 3.508(+8.88%)
- nginx/wrk
kzalloc vzalloc
Requests/s 272519.36 256490.94(-5.88%)
- redis-benchmark
kzalloc vzalloc
GET(Requests/s) 123304.56 120084.05(-2.61%)
SET(Requests/s) 122062.87 118800.12(-2.67%)
+ if (!dmb_node->cpu_addr) {
+ rc = -ENOMEM;
+ goto err_node;
+ }
+ dmb_node->len = dmb->dmb_len;
+ dmb_node->dma_addr = (dma_addr_t)dmb_node->cpu_addr;
+
+again:
+ /* add new dmb into hash table */
+ get_random_bytes(&dmb_node->token, sizeof(dmb_node->token));
+ write_lock(&ldev->dmb_ht_lock);
+ hash_for_each_possible(ldev->dmb_ht, tmp_node, list, dmb_node->token) {
+ if (tmp_node->token == dmb_node->token) {
+ write_unlock(&ldev->dmb_ht_lock);
+ goto again;
+ }
+ }
+ hash_add(ldev->dmb_ht, &dmb_node->list, dmb_node->token);
+ write_unlock(&ldev->dmb_ht_lock);
+
+ dmb->sba_idx = dmb_node->sba_idx;
+ dmb->dmb_tok = dmb_node->token;
+ dmb->cpu_addr = dmb_node->cpu_addr;
+ dmb->dma_addr = dmb_node->dma_addr;
+ dmb->dmb_len = dmb_node->len;
+
+ return 0;
+
+err_node:
+ kfree(dmb_node);
+err_bit:
+ clear_bit(sba_idx, ldev->sba_idx_mask);
+ return rc;
+}
+
+static int smc_lo_unregister_dmb(struct smcd_dev *smcd, struct smcd_dmb *dmb)
+{
+ struct smc_lo_dmb_node *dmb_node = NULL, *tmp_node;
+ struct smc_lo_dev *ldev = smcd->priv;
+
+ /* remove dmb from hash table */
+ write_lock(&ldev->dmb_ht_lock);
+ hash_for_each_possible(ldev->dmb_ht, tmp_node, list, dmb->dmb_tok) {
+ if (tmp_node->token == dmb->dmb_tok) {
+ dmb_node = tmp_node;
+ break;
+ }
+ }
+ if (!dmb_node) {
+ write_unlock(&ldev->dmb_ht_lock);
+ return -EINVAL;
+ }
+ hash_del(&dmb_node->list);
+ write_unlock(&ldev->dmb_ht_lock);
+
+ clear_bit(dmb_node->sba_idx, ldev->sba_idx_mask);
+ kfree(dmb_node->cpu_addr);
+ kfree(dmb_node);
+
+ return 0;
+}
+
