[v3] net/virtio: fix Rx scatter offload
Checks
Commit Message
From: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
Report Rx scatter offload capability depending on VIRTIO_NET_F_MRG_RXBUF.
If Rx scatter is not requested, ensure that provided Rx buffers on
each Rx queue are big enough to fit Rx packets up to configured MTU.
Fixes: ce17eddefc20 ("ethdev: introduce Rx queue offloads API")
Cc: stable@dpdk.org
Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
v3:
- fix segfault on MTU set if an Rx queue is not setup
v2:
- do not overwrite Rx offloads when Rx scatter is added
drivers/net/virtio/virtio.h | 2 +
drivers/net/virtio/virtio_ethdev.c | 65 ++++++++++++++++++++++++++++++
drivers/net/virtio/virtio_ethdev.h | 5 +++
drivers/net/virtio/virtio_rxtx.c | 10 +++++
4 files changed, 82 insertions(+)
Comments
On 7/20/21 9:54 AM, Andrew Rybchenko wrote:
> From: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
>
> Report Rx scatter offload capability depending on VIRTIO_NET_F_MRG_RXBUF.
>
> If Rx scatter is not requested, ensure that provided Rx buffers on
> each Rx queue are big enough to fit Rx packets up to configured MTU.
>
> Fixes: ce17eddefc20 ("ethdev: introduce Rx queue offloads API")
> Cc: stable@dpdk.org
>
> Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
> Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
> ---
> v3:
> - fix segfault on MTU set if an Rx queue is not setup
>
> v2:
> - do not overwrite Rx offloads when Rx scatter is added
>
> drivers/net/virtio/virtio.h | 2 +
> drivers/net/virtio/virtio_ethdev.c | 65 ++++++++++++++++++++++++++++++
> drivers/net/virtio/virtio_ethdev.h | 5 +++
> drivers/net/virtio/virtio_rxtx.c | 10 +++++
> 4 files changed, 82 insertions(+)
>
Thanks for the fix.
I see my R-by is already there, but I confirm this is good to me.
Maxime
> -----Original Message-----
> From: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
> Sent: Tuesday, July 20, 2021 3:55 PM
> To: Maxime Coquelin <maxime.coquelin@redhat.com>; Xia, Chenbo
> <chenbo.xia@intel.com>; Andrew Rybchenko <arybchenko@solarflare.com>; Ananyev,
> Konstantin <konstantin.ananyev@intel.com>; Shahaf Shuler <shahafs@mellanox.com>
> Cc: dev@dpdk.org; Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>; stable@dpdk.org
> Subject: [PATCH v3] net/virtio: fix Rx scatter offload
>
> From: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
>
> Report Rx scatter offload capability depending on VIRTIO_NET_F_MRG_RXBUF.
>
> If Rx scatter is not requested, ensure that provided Rx buffers on
> each Rx queue are big enough to fit Rx packets up to configured MTU.
>
> Fixes: ce17eddefc20 ("ethdev: introduce Rx queue offloads API")
> Cc: stable@dpdk.org
>
> Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
> Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
> ---
> --
> 2.30.2
Applied to next-virtio/main. Thanks
On 7/20/21 7:19 PM, Maxime Coquelin wrote:
>
>
> On 7/20/21 9:54 AM, Andrew Rybchenko wrote:
>> From: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
>>
>> Report Rx scatter offload capability depending on VIRTIO_NET_F_MRG_RXBUF.
>>
>> If Rx scatter is not requested, ensure that provided Rx buffers on
>> each Rx queue are big enough to fit Rx packets up to configured MTU.
>>
>> Fixes: ce17eddefc20 ("ethdev: introduce Rx queue offloads API")
>> Cc: stable@dpdk.org
>>
>> Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
>> Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
>> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
>> ---
>> v3:
>> - fix segfault on MTU set if an Rx queue is not setup
>>
>> v2:
>> - do not overwrite Rx offloads when Rx scatter is added
>>
>> drivers/net/virtio/virtio.h | 2 +
>> drivers/net/virtio/virtio_ethdev.c | 65 ++++++++++++++++++++++++++++++
>> drivers/net/virtio/virtio_ethdev.h | 5 +++
>> drivers/net/virtio/virtio_rxtx.c | 10 +++++
>> 4 files changed, 82 insertions(+)
>>
>
> Thanks for the fix.
> I see my R-by is already there, but I confirm this is good to me.
It was inherited from v1, since changes from v1 to v3 are really minor
fixes.
On 7/21/21 11:29 AM, Andrew Rybchenko wrote:
> On 7/20/21 7:19 PM, Maxime Coquelin wrote:
>>
>>
>> On 7/20/21 9:54 AM, Andrew Rybchenko wrote:
>>> From: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
>>>
>>> Report Rx scatter offload capability depending on
>>> VIRTIO_NET_F_MRG_RXBUF.
>>>
>>> If Rx scatter is not requested, ensure that provided Rx buffers on
>>> each Rx queue are big enough to fit Rx packets up to configured MTU.
>>>
>>> Fixes: ce17eddefc20 ("ethdev: introduce Rx queue offloads API")
>>> Cc: stable@dpdk.org
>>>
>>> Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
>>> Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
>>> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
>>> ---
>>> v3:
>>> - fix segfault on MTU set if an Rx queue is not setup
>>>
>>> v2:
>>> - do not overwrite Rx offloads when Rx scatter is added
>>>
>>> drivers/net/virtio/virtio.h | 2 +
>>> drivers/net/virtio/virtio_ethdev.c | 65 ++++++++++++++++++++++++++++++
>>> drivers/net/virtio/virtio_ethdev.h | 5 +++
>>> drivers/net/virtio/virtio_rxtx.c | 10 +++++
>>> 4 files changed, 82 insertions(+)
>>>
>>
>> Thanks for the fix.
>> I see my R-by is already there, but I confirm this is good to me.
>
> It was inherited from v1, since changes from v1 to v3 are really minor
> fixes.
>
Yes, no problem. I was just to let Chenbo know, so that he can add it to
the PR.
Maxime
@@ -167,6 +167,7 @@ struct virtio_hw {
uint8_t started;
uint8_t weak_barriers;
uint8_t vlan_strip;
+ bool rx_ol_scatter;
uint8_t has_tx_offload;
uint8_t has_rx_offload;
uint8_t use_vec_rx;
@@ -180,6 +181,7 @@ struct virtio_hw {
uint8_t duplex;
uint8_t intr_lsc;
uint16_t max_mtu;
+ size_t max_rx_pkt_len;
/*
* App management thread and virtio interrupt handler thread
* both can change device state, this lock is meant to avoid
@@ -841,6 +841,56 @@ virtio_dev_allmulticast_disable(struct rte_eth_dev *dev)
return 0;
}
+uint16_t
+virtio_rx_mem_pool_buf_size(struct rte_mempool *mp)
+{
+ return rte_pktmbuf_data_room_size(mp) - RTE_PKTMBUF_HEADROOM;
+}
+
+bool
+virtio_rx_check_scatter(uint16_t max_rx_pkt_len, uint16_t rx_buf_size,
+ bool rx_scatter_enabled, const char **error)
+{
+ if (!rx_scatter_enabled && max_rx_pkt_len > rx_buf_size) {
+ *error = "Rx scatter is disabled and RxQ mbuf pool object size is too small";
+ return false;
+ }
+
+ return true;
+}
+
+static bool
+virtio_check_scatter_on_all_rx_queues(struct rte_eth_dev *dev,
+ uint16_t frame_size)
+{
+ struct virtio_hw *hw = dev->data->dev_private;
+ struct virtnet_rx *rxvq;
+ struct virtqueue *vq;
+ unsigned int qidx;
+ uint16_t buf_size;
+ const char *error;
+
+ if (hw->vqs == NULL)
+ return true;
+
+ for (qidx = 0; (vq = hw->vqs[2 * qidx + VTNET_SQ_RQ_QUEUE_IDX]) != NULL;
+ qidx++) {
+ rxvq = &vq->rxq;
+ if (rxvq->mpool == NULL)
+ continue;
+ buf_size = virtio_rx_mem_pool_buf_size(rxvq->mpool);
+
+ if (!virtio_rx_check_scatter(frame_size, buf_size,
+ hw->rx_ol_scatter, &error)) {
+ PMD_INIT_LOG(ERR, "MTU check for RxQ %u failed: %s",
+ qidx, error);
+ return false;
+ }
+ }
+
+ return true;
+}
+
#define VLAN_TAG_LEN 4 /* 802.3ac tag (not DMA'd) */
static int
virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
@@ -858,6 +908,15 @@ virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
RTE_ETHER_MIN_MTU, max_frame_size - ether_hdr_len);
return -EINVAL;
}
+
+ if (!virtio_check_scatter_on_all_rx_queues(dev, frame_size)) {
+ PMD_INIT_LOG(ERR, "MTU vs Rx scatter and Rx buffers check failed");
+ return -EINVAL;
+ }
+
+ hw->max_rx_pkt_len = frame_size;
+ dev->data->dev_conf.rxmode.max_rx_pkt_len = hw->max_rx_pkt_len;
+
return 0;
}
@@ -2042,6 +2101,8 @@ virtio_dev_configure(struct rte_eth_dev *dev)
if (rxmode->max_rx_pkt_len > hw->max_mtu + ether_hdr_len)
req_features &= ~(1ULL << VIRTIO_NET_F_MTU);
+ hw->max_rx_pkt_len = rxmode->max_rx_pkt_len;
+
if (rx_offloads & (DEV_RX_OFFLOAD_UDP_CKSUM |
DEV_RX_OFFLOAD_TCP_CKSUM))
req_features |= (1ULL << VIRTIO_NET_F_GUEST_CSUM);
@@ -2090,6 +2151,8 @@ virtio_dev_configure(struct rte_eth_dev *dev)
if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
hw->vlan_strip = 1;
+ hw->rx_ol_scatter = (rx_offloads & DEV_RX_OFFLOAD_SCATTER);
+
if ((rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER) &&
!virtio_with_feature(hw, VIRTIO_NET_F_CTRL_VLAN)) {
PMD_DRV_LOG(ERR,
@@ -2445,6 +2508,8 @@ virtio_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
host_features = VIRTIO_OPS(hw)->get_features(hw);
dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP;
dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_JUMBO_FRAME;
+ if (host_features & (1ULL << VIRTIO_NET_F_MRG_RXBUF))
+ dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_SCATTER;
if (host_features & (1ULL << VIRTIO_NET_F_GUEST_CSUM)) {
dev_info->rx_offload_capa |=
DEV_RX_OFFLOAD_TCP_CKSUM |
@@ -120,4 +120,9 @@ int virtio_dev_close(struct rte_eth_dev *dev);
int virtio_inject_pkts(struct rte_eth_dev *dev, struct rte_mbuf **tx_pkts,
int nb_pkts);
+bool virtio_rx_check_scatter(uint16_t max_rx_pkt_len, uint16_t rx_buf_size,
+ bool rx_scatter_enabled, const char **error);
+
+uint16_t virtio_rx_mem_pool_buf_size(struct rte_mempool *mp);
+
#endif /* _VIRTIO_ETHDEV_H_ */
@@ -634,6 +634,8 @@ virtio_dev_rx_queue_setup(struct rte_eth_dev *dev,
struct virtqueue *vq = hw->vqs[vq_idx];
struct virtnet_rx *rxvq;
uint16_t rx_free_thresh;
+ uint16_t buf_size;
+ const char *error;
PMD_INIT_FUNC_TRACE();
@@ -642,6 +644,14 @@ virtio_dev_rx_queue_setup(struct rte_eth_dev *dev,
return -EINVAL;
}
+ buf_size = virtio_rx_mem_pool_buf_size(mp);
+ if (!virtio_rx_check_scatter(hw->max_rx_pkt_len, buf_size,
+ hw->rx_ol_scatter, &error)) {
+ PMD_INIT_LOG(ERR, "RxQ %u Rx scatter check failed: %s",
+ queue_idx, error);
+ return -EINVAL;
+ }
+
rx_free_thresh = rx_conf->rx_free_thresh;
if (rx_free_thresh == 0)
rx_free_thresh =