vmxnet3 emulation has recently added several new features including
support for uniform passthrough(UPT). To make UPT work vmxnet3 has
to be enhanced as per the new specification. This patch series
extends the vmxnet3 driver to leverage these new features.
Compatibility is maintained using existing vmxnet3 versioning mechanism as
follows:
- new features added to vmxnet3 emulation are associated with new vmxnet3
version viz. vmxnet3 version 7.
- emulation advertises all the versions it supports to the driver.
- during initialization, vmxnet3 driver picks the highest version number
supported by both the emulation and the driver and configures emulation
to run at that version.
In particular, following changes are introduced:
Patch 1:
This patch introduces utility macros for vmxnet3 version 7 comparison
and updates Copyright information.
Patch 2:
This patch adds new capability registers to fine control enablement of
individual features based on emulation and passthrough.
Patch 3:
This patch adds support for large passthrough BAR register.
Patch 4:
This patch adds support for out of order rx completion processing.
Patch 5:
This patch introduces new command to set ring buffer sizes to pass this
information to the hardware.
Patch 6:
For better performance, hardware has a requirement to limit number of TSO
descriptors. This patch adds that support.
Patch 7:
With vmxnet3 version 7, new descriptor fields are used to indicate
encapsulation offload.
Patch 8:
With all vmxnet3 version 7 changes incorporated in the vmxnet3 driver,
with this patch, the driver can configure emulation to run at vmxnet3
version 7.
Changes in v2->v3:
- use correct byte ordering for ringBufSize
Changes in v2:
- use local rss_fields variable for the rss capability checks in patch 2
Ronak Doshi (8):
vmxnet3: prepare for version 7 changes
vmxnet3: add support for capability registers
vmxnet3: add support for large passthrough BAR register
vmxnet3: add support for out of order rx completion
vmxnet3: add command to set ring buffer sizes
vmxnet3: limit number of TXDs used for TSO packet
vmxnet3: use ext1 field to indicate encapsulated packet
vmxnet3: update to version 7
drivers/net/vmxnet3/Makefile | 2 +-
drivers/net/vmxnet3/upt1_defs.h | 2 +-
drivers/net/vmxnet3/vmxnet3_defs.h | 80 ++++++++--
drivers/net/vmxnet3/vmxnet3_drv.c | 291 ++++++++++++++++++++++++++++++----
drivers/net/vmxnet3/vmxnet3_ethtool.c | 116 ++++++++++++--
drivers/net/vmxnet3/vmxnet3_int.h | 24 ++-
6 files changed, 457 insertions(+), 58 deletions(-)
--
2.11.0
vmxnet3 is currently at version 6 and this patch initiates the
preparation to accommodate changes for upto version 7. Introduced
utility macros for vmxnet3 version 7 comparison and update Copyright
information.
Signed-off-by: Ronak Doshi <[email protected]>
Acked-by: Guolin Yang <[email protected]>
---
drivers/net/vmxnet3/Makefile | 2 +-
drivers/net/vmxnet3/upt1_defs.h | 2 +-
drivers/net/vmxnet3/vmxnet3_defs.h | 2 +-
drivers/net/vmxnet3/vmxnet3_drv.c | 2 +-
drivers/net/vmxnet3/vmxnet3_ethtool.c | 2 +-
drivers/net/vmxnet3/vmxnet3_int.h | 5 ++++-
6 files changed, 9 insertions(+), 6 deletions(-)
diff --git a/drivers/net/vmxnet3/Makefile b/drivers/net/vmxnet3/Makefile
index 7a38925f4165..a666a88ac1ff 100644
--- a/drivers/net/vmxnet3/Makefile
+++ b/drivers/net/vmxnet3/Makefile
@@ -2,7 +2,7 @@
#
# Linux driver for VMware's vmxnet3 ethernet NIC.
#
-# Copyright (C) 2007-2021, VMware, Inc. All Rights Reserved.
+# Copyright (C) 2007-2022, VMware, Inc. All Rights Reserved.
#
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
diff --git a/drivers/net/vmxnet3/upt1_defs.h b/drivers/net/vmxnet3/upt1_defs.h
index f9f3a23d1698..41c0660a0c54 100644
--- a/drivers/net/vmxnet3/upt1_defs.h
+++ b/drivers/net/vmxnet3/upt1_defs.h
@@ -1,7 +1,7 @@
/*
* Linux driver for VMware's vmxnet3 ethernet NIC.
*
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
diff --git a/drivers/net/vmxnet3/vmxnet3_defs.h b/drivers/net/vmxnet3/vmxnet3_defs.h
index 74d4e8bc4abc..9f91ebb10137 100644
--- a/drivers/net/vmxnet3/vmxnet3_defs.h
+++ b/drivers/net/vmxnet3/vmxnet3_defs.h
@@ -1,7 +1,7 @@
/*
* Linux driver for VMware's vmxnet3 ethernet NIC.
*
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c
index 93e8d119d45f..6fc6a2a26161 100644
--- a/drivers/net/vmxnet3/vmxnet3_drv.c
+++ b/drivers/net/vmxnet3/vmxnet3_drv.c
@@ -1,7 +1,7 @@
/*
* Linux driver for VMware's vmxnet3 ethernet NIC.
*
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
diff --git a/drivers/net/vmxnet3/vmxnet3_ethtool.c b/drivers/net/vmxnet3/vmxnet3_ethtool.c
index 3172d46c0335..e41e76757c5b 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethtool.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethtool.c
@@ -1,7 +1,7 @@
/*
* Linux driver for VMware's vmxnet3 ethernet NIC.
*
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
diff --git a/drivers/net/vmxnet3/vmxnet3_int.h b/drivers/net/vmxnet3/vmxnet3_int.h
index 7027ff483fa5..5251c3439d6a 100644
--- a/drivers/net/vmxnet3/vmxnet3_int.h
+++ b/drivers/net/vmxnet3/vmxnet3_int.h
@@ -1,7 +1,7 @@
/*
* Linux driver for VMware's vmxnet3 ethernet NIC.
*
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
@@ -81,6 +81,7 @@
#define VMXNET3_RSS
#endif
+#define VMXNET3_REV_7 6 /* Vmxnet3 Rev. 7 */
#define VMXNET3_REV_6 5 /* Vmxnet3 Rev. 6 */
#define VMXNET3_REV_5 4 /* Vmxnet3 Rev. 5 */
#define VMXNET3_REV_4 3 /* Vmxnet3 Rev. 4 */
@@ -431,6 +432,8 @@ struct vmxnet3_adapter {
(adapter->version >= VMXNET3_REV_5 + 1)
#define VMXNET3_VERSION_GE_6(adapter) \
(adapter->version >= VMXNET3_REV_6 + 1)
+#define VMXNET3_VERSION_GE_7(adapter) \
+ (adapter->version >= VMXNET3_REV_7 + 1)
/* must be a multiple of VMXNET3_RING_SIZE_ALIGN */
#define VMXNET3_DEF_TX_RING_SIZE 512
--
2.11.0
This patch adds a new command to set ring buffer sizes. This is
required to pass the buffer size information to passthrough devices.
For performance reasons, with version7 and later, ring1 will contain
only mtu size buffers (bound to 3K). Packets > 3K will use both ring1
and ring2.
Also, ring sizes are round down to power of 2 and ring2 default
size is increased to 512.
Signed-off-by: Ronak Doshi <[email protected]>
Acked-by: Guolin Yang <[email protected]>
---
drivers/net/vmxnet3/vmxnet3_defs.h | 11 +++++++
drivers/net/vmxnet3/vmxnet3_drv.c | 57 +++++++++++++++++++++++++++--------
drivers/net/vmxnet3/vmxnet3_ethtool.c | 7 +++++
drivers/net/vmxnet3/vmxnet3_int.h | 3 +-
4 files changed, 65 insertions(+), 13 deletions(-)
diff --git a/drivers/net/vmxnet3/vmxnet3_defs.h b/drivers/net/vmxnet3/vmxnet3_defs.h
index 8d626611ab2d..415e4c9993ef 100644
--- a/drivers/net/vmxnet3/vmxnet3_defs.h
+++ b/drivers/net/vmxnet3/vmxnet3_defs.h
@@ -99,6 +99,9 @@ enum {
VMXNET3_CMD_SET_COALESCE,
VMXNET3_CMD_REGISTER_MEMREGS,
VMXNET3_CMD_SET_RSS_FIELDS,
+ VMXNET3_CMD_RESERVED4,
+ VMXNET3_CMD_RESERVED5,
+ VMXNET3_CMD_SET_RING_BUFFER_SIZE,
VMXNET3_CMD_FIRST_GET = 0xF00D0000,
VMXNET3_CMD_GET_QUEUE_STATUS = VMXNET3_CMD_FIRST_GET,
@@ -743,6 +746,13 @@ enum Vmxnet3_RSSField {
VMXNET3_RSS_FIELDS_ESPIP6 = 0x0020,
};
+struct Vmxnet3_RingBufferSize {
+ __le16 ring1BufSizeType0;
+ __le16 ring1BufSizeType1;
+ __le16 ring2BufSizeType1;
+ __le16 pad;
+};
+
/* If the command data <= 16 bytes, use the shared memory directly.
* otherwise, use variable length configuration descriptor.
*/
@@ -750,6 +760,7 @@ union Vmxnet3_CmdInfo {
struct Vmxnet3_VariableLenConfDesc varConf;
struct Vmxnet3_SetPolling setPolling;
enum Vmxnet3_RSSField setRssFields;
+ struct Vmxnet3_RingBufferSize ringBufSize;
__le64 data[2];
};
diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c
index 94ca3bc1d540..c2f99c01a130 100644
--- a/drivers/net/vmxnet3/vmxnet3_drv.c
+++ b/drivers/net/vmxnet3/vmxnet3_drv.c
@@ -2681,6 +2681,23 @@ vmxnet3_setup_driver_shared(struct vmxnet3_adapter *adapter)
}
static void
+vmxnet3_init_bufsize(struct vmxnet3_adapter *adapter)
+{
+ struct Vmxnet3_DriverShared *shared = adapter->shared;
+ union Vmxnet3_CmdInfo *cmdInfo = &shared->cu.cmdInfo;
+ unsigned long flags;
+
+ if (!VMXNET3_VERSION_GE_7(adapter))
+ return;
+
+ cmdInfo->ringBufSize = adapter->ringBufSize;
+ spin_lock_irqsave(&adapter->cmd_lock, flags);
+ VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_CMD,
+ VMXNET3_CMD_SET_RING_BUFFER_SIZE);
+ spin_unlock_irqrestore(&adapter->cmd_lock, flags);
+}
+
+static void
vmxnet3_init_coalesce(struct vmxnet3_adapter *adapter)
{
struct Vmxnet3_DriverShared *shared = adapter->shared;
@@ -2818,6 +2835,7 @@ vmxnet3_activate_dev(struct vmxnet3_adapter *adapter)
goto activate_err;
}
+ vmxnet3_init_bufsize(adapter);
vmxnet3_init_coalesce(adapter);
vmxnet3_init_rssfields(adapter);
@@ -2991,19 +3009,29 @@ static void
vmxnet3_adjust_rx_ring_size(struct vmxnet3_adapter *adapter)
{
size_t sz, i, ring0_size, ring1_size, comp_size;
- if (adapter->netdev->mtu <= VMXNET3_MAX_SKB_BUF_SIZE -
- VMXNET3_MAX_ETH_HDR_SIZE) {
- adapter->skb_buf_size = adapter->netdev->mtu +
- VMXNET3_MAX_ETH_HDR_SIZE;
- if (adapter->skb_buf_size < VMXNET3_MIN_T0_BUF_SIZE)
- adapter->skb_buf_size = VMXNET3_MIN_T0_BUF_SIZE;
-
- adapter->rx_buf_per_pkt = 1;
+ /* With version7 ring1 will have only T0 buffers */
+ if (!VMXNET3_VERSION_GE_7(adapter)) {
+ if (adapter->netdev->mtu <= VMXNET3_MAX_SKB_BUF_SIZE -
+ VMXNET3_MAX_ETH_HDR_SIZE) {
+ adapter->skb_buf_size = adapter->netdev->mtu +
+ VMXNET3_MAX_ETH_HDR_SIZE;
+ if (adapter->skb_buf_size < VMXNET3_MIN_T0_BUF_SIZE)
+ adapter->skb_buf_size = VMXNET3_MIN_T0_BUF_SIZE;
+
+ adapter->rx_buf_per_pkt = 1;
+ } else {
+ adapter->skb_buf_size = VMXNET3_MAX_SKB_BUF_SIZE;
+ sz = adapter->netdev->mtu - VMXNET3_MAX_SKB_BUF_SIZE +
+ VMXNET3_MAX_ETH_HDR_SIZE;
+ adapter->rx_buf_per_pkt = 1 + (sz + PAGE_SIZE - 1) / PAGE_SIZE;
+ }
} else {
- adapter->skb_buf_size = VMXNET3_MAX_SKB_BUF_SIZE;
- sz = adapter->netdev->mtu - VMXNET3_MAX_SKB_BUF_SIZE +
- VMXNET3_MAX_ETH_HDR_SIZE;
- adapter->rx_buf_per_pkt = 1 + (sz + PAGE_SIZE - 1) / PAGE_SIZE;
+ adapter->skb_buf_size = min((int)adapter->netdev->mtu + VMXNET3_MAX_ETH_HDR_SIZE,
+ VMXNET3_MAX_SKB_BUF_SIZE);
+ adapter->rx_buf_per_pkt = 1;
+ adapter->ringBufSize.ring1BufSizeType0 = cpu_to_le16(adapter->skb_buf_size);
+ adapter->ringBufSize.ring1BufSizeType1 = 0;
+ adapter->ringBufSize.ring2BufSizeType1 = cpu_to_le16(PAGE_SIZE);
}
/*
@@ -3019,6 +3047,11 @@ vmxnet3_adjust_rx_ring_size(struct vmxnet3_adapter *adapter)
ring1_size = (ring1_size + sz - 1) / sz * sz;
ring1_size = min_t(u32, ring1_size, VMXNET3_RX_RING2_MAX_SIZE /
sz * sz);
+ /* For v7 and later, keep ring size power of 2 for UPT */
+ if (VMXNET3_VERSION_GE_7(adapter)) {
+ ring0_size = rounddown_pow_of_two(ring0_size);
+ ring1_size = rounddown_pow_of_two(ring1_size);
+ }
comp_size = ring0_size + ring1_size;
for (i = 0; i < adapter->num_rx_queues; i++) {
diff --git a/drivers/net/vmxnet3/vmxnet3_ethtool.c b/drivers/net/vmxnet3/vmxnet3_ethtool.c
index 397b268f7bc5..ce3993282c0f 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethtool.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethtool.c
@@ -718,6 +718,13 @@ vmxnet3_set_ringparam(struct net_device *netdev,
new_rx_ring2_size = min_t(u32, new_rx_ring2_size,
VMXNET3_RX_RING2_MAX_SIZE);
+ /* For v7 and later, keep ring size power of 2 for UPT */
+ if (VMXNET3_VERSION_GE_7(adapter)) {
+ new_tx_ring_size = rounddown_pow_of_two(new_tx_ring_size);
+ new_rx_ring_size = rounddown_pow_of_two(new_rx_ring_size);
+ new_rx_ring2_size = rounddown_pow_of_two(new_rx_ring2_size);
+ }
+
/* rx data ring buffer size has to be a multiple of
* VMXNET3_RXDATA_DESC_SIZE_ALIGN
*/
diff --git a/drivers/net/vmxnet3/vmxnet3_int.h b/drivers/net/vmxnet3/vmxnet3_int.h
index 5b495ef253e8..cb87731f5f1c 100644
--- a/drivers/net/vmxnet3/vmxnet3_int.h
+++ b/drivers/net/vmxnet3/vmxnet3_int.h
@@ -408,6 +408,7 @@ struct vmxnet3_adapter {
dma_addr_t pm_conf_pa;
dma_addr_t rss_conf_pa;
bool queuesExtEnabled;
+ struct Vmxnet3_RingBufferSize ringBufSize;
u32 devcap_supported[8];
u32 ptcap_supported[8];
u32 dev_caps[8];
@@ -449,7 +450,7 @@ struct vmxnet3_adapter {
/* must be a multiple of VMXNET3_RING_SIZE_ALIGN */
#define VMXNET3_DEF_TX_RING_SIZE 512
#define VMXNET3_DEF_RX_RING_SIZE 1024
-#define VMXNET3_DEF_RX_RING2_SIZE 256
+#define VMXNET3_DEF_RX_RING2_SIZE 512
#define VMXNET3_DEF_RXDATA_DESC_SIZE 128
--
2.11.0
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 <[email protected]>
Acked-by: Guolin Yang <[email protected]>
---
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
Till vmxnet3 version 6, om field of transmit descriptor was used
to indicate encapsulated offload packet and msscof was used to
indirectly indicate TSO/CSO. From version 7 and later, ext1 field
will be used to indicate whether packet is encapsulated or not and
om fields will continue to indicate if the packet is TSO or CSO.
Signed-off-by: Ronak Doshi <[email protected]>
Acked-by: Guolin Yang <[email protected]>
---
drivers/net/vmxnet3/vmxnet3_defs.h | 14 ++++++++------
drivers/net/vmxnet3/vmxnet3_drv.c | 18 +++++++++++++++---
2 files changed, 23 insertions(+), 9 deletions(-)
diff --git a/drivers/net/vmxnet3/vmxnet3_defs.h b/drivers/net/vmxnet3/vmxnet3_defs.h
index cb9dc72f2b3d..41d6767283a6 100644
--- a/drivers/net/vmxnet3/vmxnet3_defs.h
+++ b/drivers/net/vmxnet3/vmxnet3_defs.h
@@ -148,17 +148,17 @@ struct Vmxnet3_TxDesc {
#ifdef __BIG_ENDIAN_BITFIELD
u32 msscof:14; /* MSS, checksum offset, flags */
- u32 ext1:1;
+ u32 ext1:1; /* set to 1 to indicate inner csum/tso, vmxnet3 v7 */
u32 dtype:1; /* descriptor type */
- u32 oco:1;
+ u32 oco:1; /* Outer csum offload */
u32 gen:1; /* generation bit */
u32 len:14;
#else
u32 len:14;
u32 gen:1; /* generation bit */
- u32 oco:1;
+ u32 oco:1; /* Outer csum offload */
u32 dtype:1; /* descriptor type */
- u32 ext1:1;
+ u32 ext1:1; /* set to 1 to indicate inner csum/tso, vmxnet3 v7 */
u32 msscof:14; /* MSS, checksum offset, flags */
#endif /* __BIG_ENDIAN_BITFIELD */
@@ -262,11 +262,13 @@ struct Vmxnet3_RxCompDesc {
u32 rqID:10; /* rx queue/ring ID */
u32 sop:1; /* Start of Packet */
u32 eop:1; /* End of Packet */
- u32 ext1:2;
+ u32 ext1:2; /* bit 0: indicating v4/v6/.. is for inner header */
+ /* bit 1: indicating rssType is based on inner header */
u32 rxdIdx:12; /* Index of the RxDesc */
#else
u32 rxdIdx:12; /* Index of the RxDesc */
- u32 ext1:2;
+ u32 ext1:2; /* bit 0: indicating v4/v6/.. is for inner header */
+ /* bit 1: indicating rssType is based on inner header */
u32 eop:1; /* End of Packet */
u32 sop:1; /* Start of Packet */
u32 rqID:10; /* rx queue/ring ID */
diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c
index 2ee3a2e39f10..5c42b4a008e8 100644
--- a/drivers/net/vmxnet3/vmxnet3_drv.c
+++ b/drivers/net/vmxnet3/vmxnet3_drv.c
@@ -1161,7 +1161,12 @@ vmxnet3_tq_xmit(struct sk_buff *skb, struct vmxnet3_tx_queue *tq,
if (ctx.mss) {
if (VMXNET3_VERSION_GE_4(adapter) && skb->encapsulation) {
gdesc->txd.hlen = ctx.l4_offset + ctx.l4_hdr_size;
- gdesc->txd.om = VMXNET3_OM_ENCAP;
+ if (VMXNET3_VERSION_GE_7(adapter)) {
+ gdesc->txd.om = VMXNET3_OM_TSO;
+ gdesc->txd.ext1 = 1;
+ } else {
+ gdesc->txd.om = VMXNET3_OM_ENCAP;
+ }
gdesc->txd.msscof = ctx.mss;
if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM)
@@ -1178,8 +1183,15 @@ vmxnet3_tq_xmit(struct sk_buff *skb, struct vmxnet3_tx_queue *tq,
skb->encapsulation) {
gdesc->txd.hlen = ctx.l4_offset +
ctx.l4_hdr_size;
- gdesc->txd.om = VMXNET3_OM_ENCAP;
- gdesc->txd.msscof = 0; /* Reserved */
+ if (VMXNET3_VERSION_GE_7(adapter)) {
+ gdesc->txd.om = VMXNET3_OM_CSUM;
+ gdesc->txd.msscof = ctx.l4_offset +
+ skb->csum_offset;
+ gdesc->txd.ext1 = 1;
+ } else {
+ gdesc->txd.om = VMXNET3_OM_ENCAP;
+ gdesc->txd.msscof = 0; /* Reserved */
+ }
} else {
gdesc->txd.hlen = ctx.l4_offset;
gdesc->txd.om = VMXNET3_OM_CSUM;
--
2.11.0
For vmxnet3 to work in UPT mode, the BAR sizes have been increased.
The PT page has been extended to 2 pages and also includes OOB pages
as a part of PT BAR. This patch enhances vmxnet3 to use appropriate
BAR offsets based on the capability registered. To use new offsets,
VMXNET3_CAP_LARGE_BAR needs to be set by the device. If it is not set
then the device will use legacy PT page layout.
Signed-off-by: Ronak Doshi <[email protected]>
Acked-by: Guolin Yang <[email protected]>
---
drivers/net/vmxnet3/vmxnet3_defs.h | 14 ++++++++++++--
drivers/net/vmxnet3/vmxnet3_drv.c | 25 ++++++++++++++++++++-----
drivers/net/vmxnet3/vmxnet3_ethtool.c | 6 +++---
drivers/net/vmxnet3/vmxnet3_int.h | 3 +++
4 files changed, 38 insertions(+), 10 deletions(-)
diff --git a/drivers/net/vmxnet3/vmxnet3_defs.h b/drivers/net/vmxnet3/vmxnet3_defs.h
index 0157155ff677..8d626611ab2d 100644
--- a/drivers/net/vmxnet3/vmxnet3_defs.h
+++ b/drivers/net/vmxnet3/vmxnet3_defs.h
@@ -57,8 +57,18 @@ enum {
VMXNET3_REG_RXPROD2 = 0xA00 /* Rx Producer Index for ring 2 */
};
-#define VMXNET3_PT_REG_SIZE 4096 /* BAR 0 */
-#define VMXNET3_VD_REG_SIZE 4096 /* BAR 1 */
+/* For Large PT BAR, the following offset to DB register */
+enum {
+ VMXNET3_REG_LB_TXPROD = 0x1000, /* Tx Producer Index */
+ VMXNET3_REG_LB_RXPROD = 0x1400, /* Rx Producer Index for ring 1 */
+ VMXNET3_REG_LB_RXPROD2 = 0x1800, /* Rx Producer Index for ring 2 */
+};
+
+#define VMXNET3_PT_REG_SIZE 4096 /* BAR 0 */
+#define VMXNET3_LARGE_PT_REG_SIZE 8192 /* large PT pages */
+#define VMXNET3_VD_REG_SIZE 4096 /* BAR 1 */
+#define VMXNET3_LARGE_BAR0_REG_SIZE (4096 * 4096) /* LARGE BAR 0 */
+#define VMXNET3_OOB_REG_SIZE (4094 * 4096) /* OOB pages */
#define VMXNET3_REG_ALIGN 8 /* All registers are 8-byte aligned. */
#define VMXNET3_REG_ALIGN_MASK 0x7
diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c
index edc4f23d4965..93f237db463d 100644
--- a/drivers/net/vmxnet3/vmxnet3_drv.c
+++ b/drivers/net/vmxnet3/vmxnet3_drv.c
@@ -1207,7 +1207,7 @@ vmxnet3_tq_xmit(struct sk_buff *skb, struct vmxnet3_tx_queue *tq,
if (tx_num_deferred >= le32_to_cpu(tq->shared->txThreshold)) {
tq->shared->txNumDeferred = 0;
VMXNET3_WRITE_BAR0_REG(adapter,
- VMXNET3_REG_TXPROD + tq->qid * 8,
+ adapter->tx_prod_offset + tq->qid * 8,
tq->tx_ring.next2fill);
}
@@ -1359,8 +1359,8 @@ static int
vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
struct vmxnet3_adapter *adapter, int quota)
{
- static const u32 rxprod_reg[2] = {
- VMXNET3_REG_RXPROD, VMXNET3_REG_RXPROD2
+ u32 rxprod_reg[2] = {
+ adapter->rx_prod_offset, adapter->rx_prod2_offset
};
u32 num_pkts = 0;
bool skip_page_frags = false;
@@ -2783,9 +2783,9 @@ vmxnet3_activate_dev(struct vmxnet3_adapter *adapter)
for (i = 0; i < adapter->num_rx_queues; i++) {
VMXNET3_WRITE_BAR0_REG(adapter,
- VMXNET3_REG_RXPROD + i * VMXNET3_REG_ALIGN,
+ adapter->rx_prod_offset + i * VMXNET3_REG_ALIGN,
adapter->rx_queue[i].rx_ring[0].next2fill);
- VMXNET3_WRITE_BAR0_REG(adapter, (VMXNET3_REG_RXPROD2 +
+ VMXNET3_WRITE_BAR0_REG(adapter, (adapter->rx_prod2_offset +
(i * VMXNET3_REG_ALIGN)),
adapter->rx_queue[i].rx_ring[1].next2fill);
}
@@ -3608,6 +3608,10 @@ vmxnet3_probe_device(struct pci_dev *pdev,
if (VMXNET3_VERSION_GE_7(adapter)) {
adapter->devcap_supported[0] = VMXNET3_READ_BAR1_REG(adapter, VMXNET3_REG_DCR);
adapter->ptcap_supported[0] = VMXNET3_READ_BAR1_REG(adapter, VMXNET3_REG_PTCR);
+ if (adapter->devcap_supported[0] & (1UL << VMXNET3_CAP_LARGE_BAR)) {
+ adapter->dev_caps[0] = adapter->devcap_supported[0] &
+ (1UL << VMXNET3_CAP_LARGE_BAR);
+ }
if (adapter->dev_caps[0])
VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_DCR, adapter->dev_caps[0]);
@@ -3617,6 +3621,17 @@ vmxnet3_probe_device(struct pci_dev *pdev,
spin_unlock_irqrestore(&adapter->cmd_lock, flags);
}
+ if (VMXNET3_VERSION_GE_7(adapter) &&
+ adapter->dev_caps[0] & (1UL << VMXNET3_CAP_LARGE_BAR)) {
+ adapter->tx_prod_offset = VMXNET3_REG_LB_TXPROD;
+ adapter->rx_prod_offset = VMXNET3_REG_LB_RXPROD;
+ adapter->rx_prod2_offset = VMXNET3_REG_LB_RXPROD2;
+ } else {
+ adapter->tx_prod_offset = VMXNET3_REG_TXPROD;
+ adapter->rx_prod_offset = VMXNET3_REG_RXPROD;
+ adapter->rx_prod2_offset = VMXNET3_REG_RXPROD2;
+ }
+
if (VMXNET3_VERSION_GE_6(adapter)) {
spin_lock_irqsave(&adapter->cmd_lock, flags);
VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_CMD,
diff --git a/drivers/net/vmxnet3/vmxnet3_ethtool.c b/drivers/net/vmxnet3/vmxnet3_ethtool.c
index 458f2da1ebab..397b268f7bc5 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethtool.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethtool.c
@@ -520,7 +520,7 @@ vmxnet3_get_regs(struct net_device *netdev, struct ethtool_regs *regs, void *p)
for (i = 0; i < adapter->num_tx_queues; i++) {
struct vmxnet3_tx_queue *tq = &adapter->tx_queue[i];
- buf[j++] = VMXNET3_READ_BAR0_REG(adapter, VMXNET3_REG_TXPROD +
+ buf[j++] = VMXNET3_READ_BAR0_REG(adapter, adapter->tx_prod_offset +
i * VMXNET3_REG_ALIGN);
buf[j++] = VMXNET3_GET_ADDR_LO(tq->tx_ring.basePA);
@@ -548,9 +548,9 @@ vmxnet3_get_regs(struct net_device *netdev, struct ethtool_regs *regs, void *p)
for (i = 0; i < adapter->num_rx_queues; i++) {
struct vmxnet3_rx_queue *rq = &adapter->rx_queue[i];
- buf[j++] = VMXNET3_READ_BAR0_REG(adapter, VMXNET3_REG_RXPROD +
+ buf[j++] = VMXNET3_READ_BAR0_REG(adapter, adapter->rx_prod_offset +
i * VMXNET3_REG_ALIGN);
- buf[j++] = VMXNET3_READ_BAR0_REG(adapter, VMXNET3_REG_RXPROD2 +
+ buf[j++] = VMXNET3_READ_BAR0_REG(adapter, adapter->rx_prod2_offset +
i * VMXNET3_REG_ALIGN);
buf[j++] = VMXNET3_GET_ADDR_LO(rq->rx_ring[0].basePA);
diff --git a/drivers/net/vmxnet3/vmxnet3_int.h b/drivers/net/vmxnet3/vmxnet3_int.h
index a7c8f80702c2..a4f832f0ad5b 100644
--- a/drivers/net/vmxnet3/vmxnet3_int.h
+++ b/drivers/net/vmxnet3/vmxnet3_int.h
@@ -406,6 +406,9 @@ struct vmxnet3_adapter {
u32 devcap_supported[8];
u32 ptcap_supported[8];
u32 dev_caps[8];
+ u16 tx_prod_offset;
+ u16 rx_prod_offset;
+ u16 rx_prod2_offset;
};
#define VMXNET3_WRITE_BAR0_REG(adapter, reg, val) \
--
2.11.0
Hello:
This series was applied to netdev/net-next.git (master)
by Paolo Abeni <[email protected]>:
On Tue, 7 Jun 2022 20:23:45 -0700 you wrote:
> vmxnet3 emulation has recently added several new features including
> support for uniform passthrough(UPT). To make UPT work vmxnet3 has
> to be enhanced as per the new specification. This patch series
> extends the vmxnet3 driver to leverage these new features.
>
> Compatibility is maintained using existing vmxnet3 versioning mechanism as
> follows:
> - new features added to vmxnet3 emulation are associated with new vmxnet3
> version viz. vmxnet3 version 7.
> - emulation advertises all the versions it supports to the driver.
> - during initialization, vmxnet3 driver picks the highest version number
> supported by both the emulation and the driver and configures emulation
> to run at that version.
>
> [...]
Here is the summary with links:
- [v3,net-next,1/8] vmxnet3: prepare for version 7 changes
https://git.kernel.org/netdev/net-next/c/55f0395fcace
- [v3,net-next,2/8] vmxnet3: add support for capability registers
https://git.kernel.org/netdev/net-next/c/6f91f4ba046e
- [v3,net-next,3/8] vmxnet3: add support for large passthrough BAR register
https://git.kernel.org/netdev/net-next/c/543fb6740541
- [v3,net-next,4/8] vmxnet3: add support for out of order rx completion
https://git.kernel.org/netdev/net-next/c/2c5a5748105a
- [v3,net-next,5/8] vmxnet3: add command to set ring buffer sizes
(no matching commit)
- [v3,net-next,6/8] vmxnet3: limit number of TXDs used for TSO packet
https://git.kernel.org/netdev/net-next/c/d2857b99a74b
- [v3,net-next,7/8] vmxnet3: use ext1 field to indicate encapsulated packet
https://git.kernel.org/netdev/net-next/c/60cafa0395c2
- [v3,net-next,8/8] vmxnet3: update to version 7
https://git.kernel.org/netdev/net-next/c/acc38e041bd3
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html