From patchwork Wed Oct 27 06:19:11 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Xuan Zhuo X-Patchwork-Id: 12586467 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 83F28C4332F for ; Wed, 27 Oct 2021 06:19:23 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 71D2C60FBF for ; Wed, 27 Oct 2021 06:19:23 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S238484AbhJ0GVr (ORCPT ); Wed, 27 Oct 2021 02:21:47 -0400 Received: from out30-132.freemail.mail.aliyun.com ([115.124.30.132]:60745 "EHLO out30-132.freemail.mail.aliyun.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S238502AbhJ0GVl (ORCPT ); Wed, 27 Oct 2021 02:21:41 -0400 X-Alimail-AntiSpam: AC=PASS;BC=-1|-1;BR=01201311R601e4;CH=green;DM=||false|;DS=||;FP=0|-1|-1|-1|0|-1|-1|-1;HT=e01e04394;MF=xuanzhuo@linux.alibaba.com;NM=1;PH=DS;RN=6;SR=0;TI=SMTPD_---0Utr8Ryt_1635315554; Received: from localhost(mailfrom:xuanzhuo@linux.alibaba.com fp:SMTPD_---0Utr8Ryt_1635315554) by smtp.aliyun-inc.com(127.0.0.1); Wed, 27 Oct 2021 14:19:14 +0800 From: Xuan Zhuo To: virtualization@lists.linux-foundation.org, netdev@vger.kernel.org Cc: "Michael S. Tsirkin" , Jason Wang , "David S. Miller" , Jakub Kicinski Subject: [PATCH 1/3] virtio: cache indirect desc for split Date: Wed, 27 Oct 2021 14:19:11 +0800 Message-Id: <20211027061913.76276-2-xuanzhuo@linux.alibaba.com> X-Mailer: git-send-email 2.31.0 In-Reply-To: <20211027061913.76276-1-xuanzhuo@linux.alibaba.com> References: <20211027061913.76276-1-xuanzhuo@linux.alibaba.com> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org In the case of using indirect, indirect desc must be allocated and released each time, which increases a lot of cpu overhead. Here, a cache is added for indirect. If the number of indirect desc to be applied for is less than VIRT_QUEUE_CACHE_DESC_NUM, the desc array with the size of VIRT_QUEUE_CACHE_DESC_NUM is fixed and cached for reuse. Signed-off-by: Xuan Zhuo Reported-by: kernel test robot Reported-by: kernel test robot --- drivers/virtio/virtio.c | 6 ++++ drivers/virtio/virtio_ring.c | 63 ++++++++++++++++++++++++++++++------ include/linux/virtio.h | 10 ++++++ 3 files changed, 70 insertions(+), 9 deletions(-) diff --git a/drivers/virtio/virtio.c b/drivers/virtio/virtio.c index 0a5b54034d4b..04bcb74e5b9a 100644 --- a/drivers/virtio/virtio.c +++ b/drivers/virtio/virtio.c @@ -431,6 +431,12 @@ bool is_virtio_device(struct device *dev) } EXPORT_SYMBOL_GPL(is_virtio_device); +void virtio_use_desc_cache(struct virtio_device *dev, bool val) +{ + dev->desc_cache = val; +} +EXPORT_SYMBOL_GPL(virtio_use_desc_cache); + void unregister_virtio_device(struct virtio_device *dev) { int index = dev->index; /* save for after device release */ diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c index dd95dfd85e98..0b9a8544b0e8 100644 --- a/drivers/virtio/virtio_ring.c +++ b/drivers/virtio/virtio_ring.c @@ -117,6 +117,10 @@ struct vring_virtqueue { /* Hint for event idx: already triggered no need to disable. */ bool event_triggered; + /* Is indirect cache used? */ + bool use_desc_cache; + void *desc_cache_chain; + union { /* Available for split ring */ struct { @@ -423,12 +427,47 @@ static unsigned int vring_unmap_one_split(const struct vring_virtqueue *vq, return extra[i].next; } -static struct vring_desc *alloc_indirect_split(struct virtqueue *_vq, +#define VIRT_QUEUE_CACHE_DESC_NUM 4 + +static void desc_cache_chain_free_split(void *chain) +{ + struct vring_desc *desc; + + while (chain) { + desc = chain; + chain = (void *)desc->addr; + kfree(desc); + } +} + +static void desc_cache_put_split(struct vring_virtqueue *vq, + struct vring_desc *desc, int n) +{ + if (vq->use_desc_cache && n <= VIRT_QUEUE_CACHE_DESC_NUM) { + desc->addr = (u64)vq->desc_cache_chain; + vq->desc_cache_chain = desc; + } else { + kfree(desc); + } +} + +static struct vring_desc *alloc_indirect_split(struct vring_virtqueue *vq, unsigned int total_sg, gfp_t gfp) { struct vring_desc *desc; - unsigned int i; + unsigned int i, n; + + if (vq->use_desc_cache && total_sg <= VIRT_QUEUE_CACHE_DESC_NUM) { + if (vq->desc_cache_chain) { + desc = vq->desc_cache_chain; + vq->desc_cache_chain = (void *)desc->addr; + goto got; + } + n = VIRT_QUEUE_CACHE_DESC_NUM; + } else { + n = total_sg; + } /* * We require lowmem mappings for the descriptors because @@ -437,12 +476,13 @@ static struct vring_desc *alloc_indirect_split(struct virtqueue *_vq, */ gfp &= ~__GFP_HIGHMEM; - desc = kmalloc_array(total_sg, sizeof(struct vring_desc), gfp); + desc = kmalloc_array(n, sizeof(struct vring_desc), gfp); if (!desc) return NULL; +got: for (i = 0; i < total_sg; i++) - desc[i].next = cpu_to_virtio16(_vq->vdev, i + 1); + desc[i].next = cpu_to_virtio16(vq->vq.vdev, i + 1); return desc; } @@ -508,7 +548,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq, head = vq->free_head; if (virtqueue_use_indirect(_vq, total_sg)) - desc = alloc_indirect_split(_vq, total_sg, gfp); + desc = alloc_indirect_split(vq, total_sg, gfp); else { desc = NULL; WARN_ON_ONCE(total_sg > vq->split.vring.num && !vq->indirect); @@ -652,7 +692,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq, } if (indirect) - kfree(desc); + desc_cache_put_split(vq, desc, total_sg); END_USE(vq); return -ENOMEM; @@ -717,7 +757,7 @@ static void detach_buf_split(struct vring_virtqueue *vq, unsigned int head, if (vq->indirect) { struct vring_desc *indir_desc = vq->split.desc_state[head].indir_desc; - u32 len; + u32 len, n; /* Free the indirect table, if any, now that it's unmapped. */ if (!indir_desc) @@ -729,10 +769,12 @@ static void detach_buf_split(struct vring_virtqueue *vq, unsigned int head, VRING_DESC_F_INDIRECT)); BUG_ON(len == 0 || len % sizeof(struct vring_desc)); - for (j = 0; j < len / sizeof(struct vring_desc); j++) + n = len / sizeof(struct vring_desc); + + for (j = 0; j < n; j++) vring_unmap_one_split_indirect(vq, &indir_desc[j]); - kfree(indir_desc); + desc_cache_put_split(vq, indir_desc, n); vq->split.desc_state[head].indir_desc = NULL; } else if (ctx) { *ctx = vq->split.desc_state[head].indir_desc; @@ -2199,6 +2241,8 @@ struct virtqueue *__vring_new_virtqueue(unsigned int index, vq->indirect = virtio_has_feature(vdev, VIRTIO_RING_F_INDIRECT_DESC) && !context; vq->event = virtio_has_feature(vdev, VIRTIO_RING_F_EVENT_IDX); + vq->desc_cache_chain = NULL; + vq->use_desc_cache = vdev->desc_cache; if (virtio_has_feature(vdev, VIRTIO_F_ORDER_PLATFORM)) vq->weak_barriers = false; @@ -2329,6 +2373,7 @@ void vring_del_virtqueue(struct virtqueue *_vq) if (!vq->packed_ring) { kfree(vq->split.desc_state); kfree(vq->split.desc_extra); + desc_cache_chain_free_split(vq->desc_cache_chain); } kfree(vq); } diff --git a/include/linux/virtio.h b/include/linux/virtio.h index 41edbc01ffa4..d84b7b8f4070 100644 --- a/include/linux/virtio.h +++ b/include/linux/virtio.h @@ -109,6 +109,7 @@ struct virtio_device { bool failed; bool config_enabled; bool config_change_pending; + bool desc_cache; spinlock_t config_lock; spinlock_t vqs_list_lock; /* Protects VQs list access */ struct device dev; @@ -130,6 +131,15 @@ int register_virtio_device(struct virtio_device *dev); void unregister_virtio_device(struct virtio_device *dev); bool is_virtio_device(struct device *dev); +/** + * virtio_use_desc_cache - virtio ring use desc cache + * + * virtio will cache the allocated indirect desc. + * + * This function must be called before find_vqs. + */ +void virtio_use_desc_cache(struct virtio_device *dev, bool val); + void virtio_break_device(struct virtio_device *dev); void virtio_config_changed(struct virtio_device *dev); From patchwork Wed Oct 27 06:19:12 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Xuan Zhuo X-Patchwork-Id: 12586465 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id A7A30C433F5 for ; Wed, 27 Oct 2021 06:19:22 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 8947860FBF for ; Wed, 27 Oct 2021 06:19:22 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S238478AbhJ0GVq (ORCPT ); Wed, 27 Oct 2021 02:21:46 -0400 Received: from out30-132.freemail.mail.aliyun.com ([115.124.30.132]:40536 "EHLO out30-132.freemail.mail.aliyun.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S238500AbhJ0GVl (ORCPT ); Wed, 27 Oct 2021 02:21:41 -0400 X-Alimail-AntiSpam: AC=PASS;BC=-1|-1;BR=01201311R661e4;CH=green;DM=||false|;DS=||;FP=0|-1|-1|-1|0|-1|-1|-1;HT=e01e04400;MF=xuanzhuo@linux.alibaba.com;NM=1;PH=DS;RN=6;SR=0;TI=SMTPD_---0Utr.kIT_1635315554; Received: from localhost(mailfrom:xuanzhuo@linux.alibaba.com fp:SMTPD_---0Utr.kIT_1635315554) by smtp.aliyun-inc.com(127.0.0.1); Wed, 27 Oct 2021 14:19:15 +0800 From: Xuan Zhuo To: virtualization@lists.linux-foundation.org, netdev@vger.kernel.org Cc: "Michael S. Tsirkin" , Jason Wang , "David S. Miller" , Jakub Kicinski Subject: [PATCH 2/3] virtio: cache indirect desc for packed Date: Wed, 27 Oct 2021 14:19:12 +0800 Message-Id: <20211027061913.76276-3-xuanzhuo@linux.alibaba.com> X-Mailer: git-send-email 2.31.0 In-Reply-To: <20211027061913.76276-1-xuanzhuo@linux.alibaba.com> References: <20211027061913.76276-1-xuanzhuo@linux.alibaba.com> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org In the case of using indirect, indirect desc must be allocated and released each time, which increases a lot of cpu overhead. Here, a cache is added for indirect. If the number of indirect desc to be applied for is less than VIRT_QUEUE_CACHE_DESC_NUM, the desc array with the size of VIRT_QUEUE_CACHE_DESC_NUM is fixed and cached for reuse. Signed-off-by: Xuan Zhuo Reported-by: kernel test robot Reported-by: kernel test robot Reported-by: kernel test robot --- drivers/virtio/virtio_ring.c | 57 +++++++++++++++++++++++++++++++----- 1 file changed, 49 insertions(+), 8 deletions(-) diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c index 0b9a8544b0e8..4fd7bd5bcd70 100644 --- a/drivers/virtio/virtio_ring.c +++ b/drivers/virtio/virtio_ring.c @@ -1074,10 +1074,45 @@ static void vring_unmap_desc_packed(const struct vring_virtqueue *vq, } } -static struct vring_packed_desc *alloc_indirect_packed(unsigned int total_sg, +static void desc_cache_chain_free_packed(void *chain) +{ + struct vring_packed_desc *desc; + + while (chain) { + desc = chain; + chain = (void *)desc->addr; + kfree(desc); + } +} + +static void desc_cache_put_packed(struct vring_virtqueue *vq, + struct vring_packed_desc *desc, int n) +{ + if (vq->use_desc_cache && n <= VIRT_QUEUE_CACHE_DESC_NUM) { + desc->addr = (u64)vq->desc_cache_chain; + vq->desc_cache_chain = desc; + } else { + kfree(desc); + } +} + +static struct vring_packed_desc *alloc_indirect_packed(struct vring_virtqueue *vq, + unsigned int total_sg, gfp_t gfp) { struct vring_packed_desc *desc; + unsigned int n; + + if (vq->use_desc_cache && total_sg <= VIRT_QUEUE_CACHE_DESC_NUM) { + if (vq->desc_cache_chain) { + desc = vq->desc_cache_chain; + vq->desc_cache_chain = (void *)desc->addr; + return desc; + } + n = VIRT_QUEUE_CACHE_DESC_NUM; + } else { + n = total_sg; + } /* * We require lowmem mappings for the descriptors because @@ -1086,7 +1121,7 @@ static struct vring_packed_desc *alloc_indirect_packed(unsigned int total_sg, */ gfp &= ~__GFP_HIGHMEM; - desc = kmalloc_array(total_sg, sizeof(struct vring_packed_desc), gfp); + desc = kmalloc_array(n, sizeof(struct vring_packed_desc), gfp); return desc; } @@ -1106,7 +1141,7 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq, dma_addr_t addr; head = vq->packed.next_avail_idx; - desc = alloc_indirect_packed(total_sg, gfp); + desc = alloc_indirect_packed(vq, total_sg, gfp); if (unlikely(vq->vq.num_free < 1)) { pr_debug("Can't add buf len 1 - avail = 0\n"); @@ -1197,7 +1232,7 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq, for (i = 0; i < err_idx; i++) vring_unmap_desc_packed(vq, &desc[i]); - kfree(desc); + desc_cache_put_packed(vq, desc, total_sg); END_USE(vq); return -ENOMEM; @@ -1422,20 +1457,22 @@ static void detach_buf_packed(struct vring_virtqueue *vq, } if (vq->indirect) { - u32 len; + u32 len, n; /* Free the indirect table, if any, now that it's unmapped. */ desc = state->indir_desc; if (!desc) return; + n = len / sizeof(struct vring_packed_desc); + if (vq->use_dma_api) { len = vq->packed.desc_extra[id].len; - for (i = 0; i < len / sizeof(struct vring_packed_desc); - i++) + for (i = 0; i < n; i++) vring_unmap_desc_packed(vq, &desc[i]); } - kfree(desc); + + desc_cache_put_packed(vq, desc, n); state->indir_desc = NULL; } else if (ctx) { *ctx = state->indir_desc; @@ -1753,6 +1790,8 @@ static struct virtqueue *vring_create_virtqueue_packed( vq->indirect = virtio_has_feature(vdev, VIRTIO_RING_F_INDIRECT_DESC) && !context; vq->event = virtio_has_feature(vdev, VIRTIO_RING_F_EVENT_IDX); + vq->desc_cache_chain = NULL; + vq->use_desc_cache = vdev->desc_cache; if (virtio_has_feature(vdev, VIRTIO_F_ORDER_PLATFORM)) vq->weak_barriers = false; @@ -2374,6 +2413,8 @@ void vring_del_virtqueue(struct virtqueue *_vq) kfree(vq->split.desc_state); kfree(vq->split.desc_extra); desc_cache_chain_free_split(vq->desc_cache_chain); + } else { + desc_cache_chain_free_packed(vq->desc_cache_chain); } kfree(vq); } From patchwork Wed Oct 27 06:19:13 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Xuan Zhuo X-Patchwork-Id: 12586469 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 2A282C433EF for ; Wed, 27 Oct 2021 06:19:24 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 1684A610CA for ; Wed, 27 Oct 2021 06:19:24 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S238492AbhJ0GVs (ORCPT ); Wed, 27 Oct 2021 02:21:48 -0400 Received: from out30-133.freemail.mail.aliyun.com ([115.124.30.133]:37867 "EHLO out30-133.freemail.mail.aliyun.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S238512AbhJ0GVm (ORCPT ); Wed, 27 Oct 2021 02:21:42 -0400 X-Alimail-AntiSpam: AC=PASS;BC=-1|-1;BR=01201311R961e4;CH=green;DM=||false|;DS=||;FP=0|-1|-1|-1|0|-1|-1|-1;HT=e01e04394;MF=xuanzhuo@linux.alibaba.com;NM=1;PH=DS;RN=6;SR=0;TI=SMTPD_---0Utr84H5_1635315555; Received: from localhost(mailfrom:xuanzhuo@linux.alibaba.com fp:SMTPD_---0Utr84H5_1635315555) by smtp.aliyun-inc.com(127.0.0.1); Wed, 27 Oct 2021 14:19:16 +0800 From: Xuan Zhuo To: virtualization@lists.linux-foundation.org, netdev@vger.kernel.org Cc: "Michael S. Tsirkin" , Jason Wang , "David S. Miller" , Jakub Kicinski Subject: [PATCH 3/3] virtio-net: enable virtio indirect cache Date: Wed, 27 Oct 2021 14:19:13 +0800 Message-Id: <20211027061913.76276-4-xuanzhuo@linux.alibaba.com> X-Mailer: git-send-email 2.31.0 In-Reply-To: <20211027061913.76276-1-xuanzhuo@linux.alibaba.com> References: <20211027061913.76276-1-xuanzhuo@linux.alibaba.com> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org If the VIRTIO_RING_F_INDIRECT_DESC negotiation succeeds, and the number of sgs used for sending packets is greater than 1. We must constantly call __kmalloc/kfree to allocate/release desc. In the case of extremely fast package delivery, the overhead cannot be ignored: 27.46% [kernel] [k] virtqueue_add 16.66% [kernel] [k] detach_buf_split 16.51% [kernel] [k] virtnet_xsk_xmit 14.04% [kernel] [k] virtqueue_add_outbuf 5.18% [kernel] [k] __kmalloc 4.08% [kernel] [k] kfree 2.80% [kernel] [k] virtqueue_get_buf_ctx 2.22% [kernel] [k] xsk_tx_peek_desc 2.08% [kernel] [k] memset_erms 0.83% [kernel] [k] virtqueue_kick_prepare 0.76% [kernel] [k] virtnet_xsk_run 0.62% [kernel] [k] __free_old_xmit_ptr 0.60% [kernel] [k] vring_map_one_sg 0.53% [kernel] [k] native_apic_mem_write 0.46% [kernel] [k] sg_next 0.43% [kernel] [k] sg_init_table 0.41% [kernel] [k] kmalloc_slab Compared to not using virtio indirect cache, virtio-net can get a 16% performance improvement when using indirect desc cache. In the test case, the CPU where the package is sent has reached 100%. The following are the PPS in two cases: indirect desc cache | no cache 3074658 | 2685132 3111866 | 2666118 3152527 | 2653632 3125867 | 2669820 3027147 | 2644464 3069211 | 2669777 3038522 | 2675645 3034507 | 2671302 3102257 | 2685504 3083712 | 2692800 3051771 | 2676928 3080684 | 2695040 3147816 | 2720876 3123887 | 2705492 3180963 | 2699520 3191579 | 2676480 3161670 | 2686272 3189768 | 2692588 3174272 | 2686692 3143434 | 2682416 Signed-off-by: Xuan Zhuo --- drivers/net/virtio_net.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index 4ad25a8b0870..0ec29cf90d0a 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -27,9 +27,11 @@ static int napi_weight = NAPI_POLL_WEIGHT; module_param(napi_weight, int, 0444); static bool csum = true, gso = true, napi_tx = true; +static bool virtio_desc_cache = true; module_param(csum, bool, 0444); module_param(gso, bool, 0444); module_param(napi_tx, bool, 0644); +module_param(virtio_desc_cache, bool, 0644); /* FIXME: MTU in config. */ #define GOOD_PACKET_LEN (ETH_HLEN + VLAN_HLEN + ETH_DATA_LEN) @@ -3214,6 +3216,8 @@ static int virtnet_probe(struct virtio_device *vdev) vi->curr_queue_pairs = num_online_cpus(); vi->max_queue_pairs = max_queue_pairs; + virtio_use_desc_cache(vdev, virtio_desc_cache); + /* Allocate/initialize the rx/tx queues, and invoke find_vqs */ err = init_vqs(vi); if (err)