| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20220528011758.7024-5-doshir@vmware.com>
Date: Fri, 27 May 2022 18:17:54 -0700
From: Ronak Doshi <doshir@...are.com>
To: <netdev@...r.kernel.org>
CC: Ronak Doshi <doshir@...are.com>,
VMware PV-Drivers Reviewers <pv-drivers@...are.com>,
"David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>,
Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>,
open list <linux-kernel@...r.kernel.org>
Subject: [PATCH net-next 4/8] vmxnet3: add support for out of order rx completion
Currently, vmxnet3 processes rx completions in-order i.e. no
out of order completion descriptor is expected. With UPT, if
hardware supports LRO, then hardware can report out of order
rx completions. This patch enhances vmxnet3 to add this support.
This supports gets effective only when the corresponding feature
bit is set.
Also, minor enhancements are done for performance.
Signed-off-by: Ronak Doshi <doshir@...are.com>
Acked-by: Guolin Yang <gyang@...are.com>
---
drivers/net/vmxnet3/vmxnet3_drv.c | 70 ++++++++++++++++++++++++++++++++-------
drivers/net/vmxnet3/vmxnet3_int.h | 5 +++
2 files changed, 63 insertions(+), 12 deletions(-)
diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c
index 93f237db463d..94ca3bc1d540 100644
--- a/drivers/net/vmxnet3/vmxnet3_drv.c
+++ b/drivers/net/vmxnet3/vmxnet3_drv.c
@@ -585,6 +585,7 @@ vmxnet3_rq_alloc_rx_buf(struct vmxnet3_rx_queue *rq, u32 ring_idx,
rbi = rbi_base + ring->next2fill;
gd = ring->base + ring->next2fill;
+ rbi->comp_state = VMXNET3_RXD_COMP_PENDING;
if (rbi->buf_type == VMXNET3_RX_BUF_SKB) {
if (rbi->skb == NULL) {
@@ -644,8 +645,10 @@ vmxnet3_rq_alloc_rx_buf(struct vmxnet3_rx_queue *rq, u32 ring_idx,
/* Fill the last buffer but dont mark it ready, or else the
* device will think that the queue is full */
- if (num_allocated == num_to_alloc)
+ if (num_allocated == num_to_alloc) {
+ rbi->comp_state = VMXNET3_RXD_COMP_DONE;
break;
+ }
gd->dword[2] |= cpu_to_le32(ring->gen << VMXNET3_RXD_GEN_SHIFT);
num_allocated++;
@@ -1367,6 +1370,7 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
struct Vmxnet3_RxCompDesc *rcd;
struct vmxnet3_rx_ctx *ctx = &rq->rx_ctx;
u16 segCnt = 0, mss = 0;
+ int comp_offset, fill_offset;
#ifdef __BIG_ENDIAN_BITFIELD
struct Vmxnet3_RxDesc rxCmdDesc;
struct Vmxnet3_RxCompDesc rxComp;
@@ -1639,9 +1643,15 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
rcd_done:
/* device may have skipped some rx descs */
- ring->next2comp = idx;
- num_to_alloc = vmxnet3_cmd_ring_desc_avail(ring);
ring = rq->rx_ring + ring_idx;
+ rbi->comp_state = VMXNET3_RXD_COMP_DONE;
+
+ comp_offset = vmxnet3_cmd_ring_desc_avail(ring);
+ fill_offset = (idx > ring->next2fill ? 0 : ring->size) +
+ idx - ring->next2fill - 1;
+ if (!ring->isOutOfOrder || fill_offset >= comp_offset)
+ ring->next2comp = idx;
+ num_to_alloc = vmxnet3_cmd_ring_desc_avail(ring);
/* Ensure that the writes to rxd->gen bits will be observed
* after all other writes to rxd objects.
@@ -1649,18 +1659,38 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
dma_wmb();
while (num_to_alloc) {
- vmxnet3_getRxDesc(rxd, &ring->base[ring->next2fill].rxd,
- &rxCmdDesc);
- BUG_ON(!rxd->addr);
-
- /* Recv desc is ready to be used by the device */
- rxd->gen = ring->gen;
- vmxnet3_cmd_ring_adv_next2fill(ring);
- num_to_alloc--;
+ rbi = rq->buf_info[ring_idx] + ring->next2fill;
+ if (!(adapter->dev_caps[0] & (1UL << VMXNET3_CAP_OOORX_COMP)))
+ goto refill_buf;
+ if (ring_idx == 0) {
+ /* ring0 Type1 buffers can get skipped; re-fill them */
+ if (rbi->buf_type != VMXNET3_RX_BUF_SKB)
+ goto refill_buf;
+ }
+ if (rbi->comp_state == VMXNET3_RXD_COMP_DONE) {
+refill_buf:
+ vmxnet3_getRxDesc(rxd, &ring->base[ring->next2fill].rxd,
+ &rxCmdDesc);
+ WARN_ON(!rxd->addr);
+
+ /* Recv desc is ready to be used by the device */
+ rxd->gen = ring->gen;
+ vmxnet3_cmd_ring_adv_next2fill(ring);
+ rbi->comp_state = VMXNET3_RXD_COMP_PENDING;
+ num_to_alloc--;
+ } else {
+ /* rx completion hasn't occurred */
+ ring->isOutOfOrder = 1;
+ break;
+ }
+ }
+
+ if (num_to_alloc == 0) {
+ ring->isOutOfOrder = 0;
}
/* if needed, update the register */
- if (unlikely(rq->shared->updateRxProd)) {
+ if (unlikely(rq->shared->updateRxProd) && (ring->next2fill & 0xf) == 0) {
VMXNET3_WRITE_BAR0_REG(adapter,
rxprod_reg[ring_idx] + rq->qid * 8,
ring->next2fill);
@@ -1824,6 +1854,7 @@ vmxnet3_rq_init(struct vmxnet3_rx_queue *rq,
memset(rq->rx_ring[i].base, 0, rq->rx_ring[i].size *
sizeof(struct Vmxnet3_RxDesc));
rq->rx_ring[i].gen = VMXNET3_INIT_GEN;
+ rq->rx_ring[i].isOutOfOrder = 0;
}
if (vmxnet3_rq_alloc_rx_buf(rq, 0, rq->rx_ring[0].size - 1,
adapter) == 0) {
@@ -2014,8 +2045,17 @@ vmxnet3_poll_rx_only(struct napi_struct *napi, int budget)
rxd_done = vmxnet3_rq_rx_complete(rq, adapter, budget);
if (rxd_done < budget) {
+ struct Vmxnet3_RxCompDesc *rcd;
+#ifdef __BIG_ENDIAN_BITFIELD
+ struct Vmxnet3_RxCompDesc rxComp;
+#endif
napi_complete_done(napi, rxd_done);
vmxnet3_enable_intr(adapter, rq->comp_ring.intr_idx);
+ /* after unmasking the interrupt, check if any descriptors were completed */
+ vmxnet3_getRxComp(rcd, &rq->comp_ring.base[rq->comp_ring.next2proc].rcd,
+ &rxComp);
+ if (rcd->gen == rq->comp_ring.gen && napi_reschedule(napi))
+ vmxnet3_disable_intr(adapter, rq->comp_ring.intr_idx);
}
return rxd_done;
}
@@ -3612,6 +3652,12 @@ vmxnet3_probe_device(struct pci_dev *pdev,
adapter->dev_caps[0] = adapter->devcap_supported[0] &
(1UL << VMXNET3_CAP_LARGE_BAR);
}
+ if (!(adapter->ptcap_supported[0] & (1UL << VMXNET3_DCR_ERROR)) &&
+ adapter->ptcap_supported[0] & (1UL << VMXNET3_CAP_OOORX_COMP) &&
+ adapter->devcap_supported[0] & (1UL << VMXNET3_CAP_OOORX_COMP)) {
+ adapter->dev_caps[0] |= adapter->devcap_supported[0] &
+ (1UL << VMXNET3_CAP_OOORX_COMP);
+ }
if (adapter->dev_caps[0])
VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_DCR, adapter->dev_caps[0]);
diff --git a/drivers/net/vmxnet3/vmxnet3_int.h b/drivers/net/vmxnet3/vmxnet3_int.h
index a4f832f0ad5b..5b495ef253e8 100644
--- a/drivers/net/vmxnet3/vmxnet3_int.h
+++ b/drivers/net/vmxnet3/vmxnet3_int.h
@@ -136,6 +136,7 @@ struct vmxnet3_cmd_ring {
u32 next2fill;
u32 next2comp;
u8 gen;
+ u8 isOutOfOrder;
dma_addr_t basePA;
};
@@ -260,9 +261,13 @@ enum vmxnet3_rx_buf_type {
VMXNET3_RX_BUF_PAGE = 2
};
+#define VMXNET3_RXD_COMP_PENDING 0
+#define VMXNET3_RXD_COMP_DONE 1
+
struct vmxnet3_rx_buf_info {
enum vmxnet3_rx_buf_type buf_type;
u16 len;
+ u8 comp_state;
union {
struct sk_buff *skb;
struct page *page;
--
2.11.0
Powered by blists - more mailing lists