From: Jianfeng Tan <jianfeng.tan@intel.com>
To: dev@dpdk.org
Cc: yuanhan.liu@linux.intel.com, ferruh.yigit@intel.com,
cunming.liang@intel.com, Jianfeng Tan <jianfeng.tan@intel.com>
Subject: [dpdk-dev] [PATCH v2 6/7] net/virtio_user: enable offloading
Date: Fri, 23 Dec 2016 07:14:25 +0000 [thread overview]
Message-ID: <1482477266-39199-7-git-send-email-jianfeng.tan@intel.com> (raw)
In-Reply-To: <1482477266-39199-1-git-send-email-jianfeng.tan@intel.com>
When used with vhost kernel backend, we can offload at both directions.
- From vhost kernel to virtio_user, the offload is enabled so that
DPDK app can trust the flow is checksum-correct; and if DPDK app
sends it through another port, the checksum needs to be
recalculated or offloaded. It also applies to TSO.
- From virtio_user to vhost_kernel, the offload is enabled so that
kernel can trust the flow is L4-checksum-correct, no need to verify
it; if kernel will consume it, DPDK app should make sure the
l3-checksum is correctly set.
Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
---
drivers/net/virtio/virtio_user/vhost_kernel.c | 61 ++++++++++++++++++++++++++-
1 file changed, 59 insertions(+), 2 deletions(-)
diff --git a/drivers/net/virtio/virtio_user/vhost_kernel.c b/drivers/net/virtio/virtio_user/vhost_kernel.c
index 8984c5c..fb3c454 100644
--- a/drivers/net/virtio/virtio_user/vhost_kernel.c
+++ b/drivers/net/virtio/virtio_user/vhost_kernel.c
@@ -91,6 +91,13 @@ struct vhost_memory_kernel {
#define IFF_ATTACH_QUEUE 0x0200
#define IFF_DETACH_QUEUE 0x0400
+/* Features for GSO (TUNSETOFFLOAD). */
+#define TUN_F_CSUM 0x01 /* You can hand me unchecksummed packets. */
+#define TUN_F_TSO4 0x02 /* I can handle TSO for IPv4 packets */
+#define TUN_F_TSO6 0x04 /* I can handle TSO for IPv6 packets */
+#define TUN_F_TSO_ECN 0x08 /* I can handle TSO with ECN bits. */
+#define TUN_F_UFO 0x10 /* I can handle UFO packets */
+
/* Constants */
#define TUN_DEF_SNDBUF (1ull << 20)
#define PATH_NET_TUN "/dev/net/tun"
@@ -173,6 +180,28 @@ prepare_vhost_memory_kernel(void)
return vm;
}
+/* with below features, vhost kernel does not need to do the checksum and TSO,
+ * these info will be passed to virtio_user through virtio net header.
+ */
+static const uint64_t guest_offloads_mask =
+ (1ULL << VIRTIO_NET_F_GUEST_CSUM) |
+ (1ULL << VIRTIO_NET_F_GUEST_TSO4) |
+ (1ULL << VIRTIO_NET_F_GUEST_TSO6) |
+ (1ULL << VIRTIO_NET_F_GUEST_ECN) |
+ (1ULL << VIRTIO_NET_F_GUEST_UFO);
+
+/* with below features, when flows from virtio_user to vhost kernel
+ * (1) if flows goes up through the kernel networking stack, it does not need
+ * to verify checksum, which can save CPU cycles;
+ * (2) if flows goes through a Linux bridge and outside from an interface
+ * (kernel driver), checksum and TSO will be done by GSO in kernel or even
+ * offloaded into real physical device.
+ */
+static const uint64_t host_offloads_mask =
+ (1ULL << VIRTIO_NET_F_CSUM) |
+ (1ULL << VIRTIO_NET_F_HOST_TSO4) |
+ (1ULL << VIRTIO_NET_F_HOST_TSO6);
+
static int
vhost_kernel_ioctl(struct virtio_user_dev *dev,
enum vhost_user_request req,
@@ -191,10 +220,15 @@ vhost_kernel_ioctl(struct virtio_user_dev *dev,
arg = (void *)vm;
}
- /* Does not work when VIRTIO_F_IOMMU_PLATFORM now, why? */
- if (req_kernel == VHOST_SET_FEATURES)
+ if (req_kernel == VHOST_SET_FEATURES) {
+ /* Does not work when VIRTIO_F_IOMMU_PLATFORM now, why? */
*(uint64_t *)arg &= ~(1ULL << VIRTIO_F_IOMMU_PLATFORM);
+ /* VHOST kernel does not know about below flags */
+ *(uint64_t *)arg &= ~guest_offloads_mask;
+ *(uint64_t *)arg &= ~host_offloads_mask;
+ }
+
for (i = 0; i < VHOST_KERNEL_MAX_QUEUES; ++i) {
if (dev->vhostfds[i] < 0)
continue;
@@ -204,6 +238,15 @@ vhost_kernel_ioctl(struct virtio_user_dev *dev,
break;
}
+ if (!ret && req_kernel == VHOST_GET_FEATURES) {
+ /* with tap as the backend, all these features are supported
+ * but not claimed by vhost-net, so we add them back when
+ * reporting to upper layer.
+ */
+ *((uint64_t *)arg) |= guest_offloads_mask;
+ *((uint64_t *)arg) |= host_offloads_mask;
+ }
+
if (vm)
free(vm);
@@ -271,6 +314,12 @@ vhost_kernel_enable_queue_pair(struct virtio_user_dev *dev,
int hdr_size;
int vhostfd;
int tapfd;
+ unsigned int offload =
+ TUN_F_CSUM |
+ TUN_F_TSO4 |
+ TUN_F_TSO6 |
+ TUN_F_TSO_ECN |
+ TUN_F_UFO;
vhostfd = dev->vhostfds[pair_idx];
@@ -345,6 +394,14 @@ vhost_kernel_enable_queue_pair(struct virtio_user_dev *dev,
goto error;
}
+ /* TODO: before set the offload capabilities, we'd better (1) check
+ * negotiated features to see if necessary to offload; (2) query tap
+ * to see if it supports the offload capabilities.
+ */
+ if (ioctl(tapfd, TUNSETOFFLOAD, offload) != 0)
+ PMD_DRV_LOG(ERR, "TUNSETOFFLOAD ioctl() failed: %s",
+ strerror(errno));
+
if (vhost_kernel_set_backend(vhostfd, tapfd) < 0)
goto error;
--
2.7.4
next prev parent reply other threads:[~2016-12-23 7:13 UTC|newest]
Thread overview: 72+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-12-02 14:31 [dpdk-dev] [PATCH 0/3] virtio_user as an alternative exception path Jianfeng Tan
2016-12-02 14:31 ` [dpdk-dev] [PATCH 1/3] net/virtio_user: add vhost layer Jianfeng Tan
2016-12-08 7:21 ` Yuanhan Liu
2016-12-02 14:31 ` [dpdk-dev] [PATCH 2/3] net/virtio_user: add vhost kernel support Jianfeng Tan
2016-12-02 14:31 ` [dpdk-dev] [PATCH 3/3] net/virtio_user: fix wrongly set features Jianfeng Tan
2016-12-08 8:32 ` Yuanhan Liu
2016-12-02 14:44 ` [dpdk-dev] [PATCH 0/3] virtio_user as an alternative exception path Thomas Monjalon
2016-12-23 7:14 ` [dpdk-dev] [PATCH v2 0/7] " Jianfeng Tan
2016-12-23 7:14 ` [dpdk-dev] [PATCH v2 1/7] net/virtio_user: fix wrongly set features Jianfeng Tan
2016-12-23 7:14 ` [dpdk-dev] [PATCH v2 2/7] net/virtio_user: postpone DRIVER OK notification Jianfeng Tan
[not found] ` <20161226062719.GA19288@yliu-dev.sh.intel.com>
2016-12-26 6:55 ` Tan, Jianfeng
2016-12-26 8:02 ` Yuanhan Liu
2016-12-26 8:26 ` Tan, Jianfeng
2016-12-23 7:14 ` [dpdk-dev] [PATCH v2 3/7] net/virtio_user: move vhost user specific code Jianfeng Tan
2016-12-26 6:28 ` Yuanhan Liu
2016-12-26 6:58 ` Tan, Jianfeng
2016-12-26 7:57 ` Yuanhan Liu
2016-12-29 9:40 ` Tan, Jianfeng
2016-12-23 7:14 ` [dpdk-dev] [PATCH v2 4/7] net/virtio_user: abstract virtio user backend ops Jianfeng Tan
2016-12-26 6:41 ` Yuanhan Liu
2016-12-23 7:14 ` [dpdk-dev] [PATCH v2 5/7] net/virtio_user: add vhost kernel support Jianfeng Tan
2016-12-26 7:44 ` Yuanhan Liu
2017-01-04 7:22 ` Tan, Jianfeng
2017-01-04 7:46 ` Yuanhan Liu
2017-01-09 4:51 ` Jason Wang
2017-01-09 4:39 ` Jason Wang
2017-01-10 6:11 ` Tan, Jianfeng
2017-01-10 8:46 ` Thomas Monjalon
2017-01-10 9:11 ` Tan, Jianfeng
2017-01-11 2:42 ` Jason Wang
2017-01-11 3:13 ` Tan, Jianfeng
2017-01-11 3:23 ` Jason Wang
2017-01-12 9:40 ` Tan, Jianfeng
2017-01-13 2:27 ` Jason Wang
2017-01-11 2:30 ` Tan, Jianfeng
2017-01-11 2:45 ` Jason Wang
2016-12-23 7:14 ` Jianfeng Tan [this message]
2016-12-26 7:53 ` [dpdk-dev] [PATCH v2 6/7] net/virtio_user: enable offloading Yuanhan Liu
2016-12-23 7:14 ` [dpdk-dev] [PATCH v2 7/7] net/virtio_user: enable multiqueue with vhost kernel Jianfeng Tan
2017-01-04 3:59 ` [dpdk-dev] [PATCH v3 0/7] virtio_user as an alternative exception path Jianfeng Tan
2017-01-04 3:59 ` [dpdk-dev] [PATCH v3 1/7] net/virtio_user: fix wrongly set features Jianfeng Tan
2017-01-04 3:59 ` [dpdk-dev] [PATCH v3 2/7] net/virtio_user: fix not properly reset device Jianfeng Tan
2017-01-04 5:46 ` Yuanhan Liu
2017-01-04 3:59 ` [dpdk-dev] [PATCH v3 3/7] net/virtio_user: move vhost user specific code Jianfeng Tan
2017-01-04 6:02 ` Yuanhan Liu
2017-01-04 6:46 ` Tan, Jianfeng
2017-01-04 7:08 ` Yuanhan Liu
2017-01-04 3:59 ` [dpdk-dev] [PATCH v3 4/7] net/virtio_user: abstract virtio user backend ops Jianfeng Tan
2017-01-04 6:11 ` Yuanhan Liu
2017-01-04 3:59 ` [dpdk-dev] [PATCH v3 5/7] net/virtio_user: add vhost kernel support Jianfeng Tan
2017-01-04 6:13 ` Yuanhan Liu
2017-01-04 3:59 ` [dpdk-dev] [PATCH v3 6/7] net/virtio_user: enable offloading Jianfeng Tan
2017-01-04 3:59 ` [dpdk-dev] [PATCH v3 7/7] net/virtio_user: enable multiqueue with vhost kernel Jianfeng Tan
2017-01-09 14:06 ` [dpdk-dev] [PATCH v3 0/7] virtio_user as an alternative exception path Bruce Richardson
2017-01-10 8:46 ` Tan, Jianfeng
2017-01-13 12:18 ` [dpdk-dev] [PATCH v4 0/8] " Jianfeng Tan
2017-01-13 12:18 ` [dpdk-dev] [PATCH v4 1/8] net/virtio_user: fix wrongly get/set features Jianfeng Tan
2017-01-13 12:18 ` [dpdk-dev] [PATCH v4 2/8] net/virtio_user: fix not properly reset device Jianfeng Tan
2017-01-13 12:18 ` [dpdk-dev] [PATCH v4 3/8] net/virtio_user: move vhost user specific code Jianfeng Tan
2017-01-13 12:18 ` [dpdk-dev] [PATCH v4 4/8] net/virtio_user: abstract virtio user backend ops Jianfeng Tan
2017-01-13 12:18 ` [dpdk-dev] [PATCH v4 5/8] net/virtio_user: add vhost kernel support Jianfeng Tan
2017-01-13 12:18 ` [dpdk-dev] [PATCH v4 6/8] net/virtio_user: enable offloading Jianfeng Tan
2017-01-13 12:18 ` [dpdk-dev] [PATCH v4 7/8] net/virtio_user: enable multiqueue with vhost kernel Jianfeng Tan
2017-01-13 12:18 ` [dpdk-dev] [PATCH v4 8/8] doc: add guide to use virtio_user as exceptional path Jianfeng Tan
2017-01-16 6:00 ` Yuanhan Liu
2017-01-16 6:04 ` Yuanhan Liu
2017-01-16 9:44 ` Thomas Monjalon
2017-01-16 9:49 ` Yuanhan Liu
2017-01-16 14:45 ` Thomas Monjalon
2017-01-22 0:46 ` Aws Ismail
2017-01-22 7:16 ` Tan, Jianfeng
2017-01-16 15:05 ` [dpdk-dev] [PATCH v4 0/8] virtio_user as an alternative exception path Yuanhan Liu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1482477266-39199-7-git-send-email-jianfeng.tan@intel.com \
--to=jianfeng.tan@intel.com \
--cc=cunming.liang@intel.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=yuanhan.liu@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).