Mykyta Iziumtsev(MykytaI) replied on github web page:
platform/linux-generic/pktio/mdev/e1000e.c
line 334
@@ -0,0 +1,605 @@
+/* Copyright (c) 2018, Linaro Limited
+ * All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#include "config.h"
+
+#ifdef ODP_MDEV
+
+#include <linux/types.h>
+#include <protocols/eth.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <string.h>
+#include <sys/mman.h>
+#include <unistd.h>
+
+#include <odp_packet_io_internal.h>
+#include <odp_posix_extensions.h>
+
+#include <odp/api/hints.h>
+#include <odp/api/packet.h>
+#include <odp/api/plat/packet_inlines.h>
+#include <odp/drv/hints.h>
+#include <odp/drv/mmio.h>
+
+#include <pktio/common.h>
+#include <pktio/ethtool.h>
+#include <pktio/mdev.h>
+#include <pktio/sysfs.h>
+#include <pktio/uapi_net_mdev.h>
+
+#define MODULE_NAME "e1000e"
+
+#define E1000E_TX_BUF_SIZE 2048U
+#define E1000E_RX_BUF_SIZE 2048U
+
+/* RX queue definitions */
+#define E1000E_RX_QUEUE_NUM_MAX 1
+
+#define E1000_RDH_OFFSET 0x02810UL
+#define E1000_RDT_OFFSET 0x02818UL
+
+/** RX descriptor */
+typedef struct {
+ odp_u64le_t addr;
+#define E1000E_RXD_STATUS_DONE 0x00000001UL
+#define E1000E_RXD_STATUS_ERR_MASK 0xff000000UL
+ odp_u32le_t status;
+ odp_u16le_t length;
+ odp_u16le_t vlan;
+} e1000e_rx_desc_t;
+
+/** RX queue data */
+typedef struct ODP_ALIGNED_CACHE {
+ e1000e_rx_desc_t *rx_descs; /**< RX queue base */
+ odp_u32le_t *doorbell; /**< RX queue doorbell */
+
+ uint16_t rx_queue_len; /**< Number of RX desc entries */
+
+ uint16_t cidx; /**< Next RX desc to read */
+ odp_u32le_t *pidx; /**< Next RX desc HW is going to write
*/
+
+ mdev_dma_area_t rx_data; /**< RX packet payload area */
+
+ odp_ticketlock_t lock; /**< RX queue lock */
+} e1000e_rx_queue_t;
+
+/* TX queue definitions */
+#define E1000E_TX_QUEUE_NUM_MAX 1
+
+#define E1000_TDH_OFFSET 0x03810UL
+#define E1000_TDT_OFFSET 0x03818UL
+
+typedef struct {
+ odp_u64le_t addr; /* Address of data buffer */
+#define E1000_TXD_CMD_EOP 0x01000000 /* End of Packet */
+#define E1000_TXD_CMD_IFCS 0x02000000 /* Insert FCS (Ethernet CRC) */
+ odp_u32le_t cmd;
+ odp_u32le_t reserved;
+} e1000e_tx_desc_t;
+
+/** TX queue data */
+typedef struct ODP_ALIGNED_CACHE {
+ e1000e_tx_desc_t *tx_descs; /**< TX queue base */
+ odp_u32le_t *doorbell; /**< TX queue doorbell */
+
+ uint16_t tx_queue_len; /**< Number of TX desc entries */
+
+ uint16_t pidx; /**< Next TX desc to write */
+ odp_u32le_t *cidx; /**< Next TX desc HW is going to read */
+
+ mdev_dma_area_t tx_data; /**< TX packet payload area */
+
+ odp_ticketlock_t lock; /**< TX queue lock */
+} e1000e_tx_queue_t;
+
+/** Packet socket using mediated e1000e device */
+typedef struct {
+ /** RX queue hot data */
+ e1000e_rx_queue_t rx_queues[E1000E_RX_QUEUE_NUM_MAX];
+
+ /** TX queue hot data */
+ e1000e_tx_queue_t tx_queues[E1000E_TX_QUEUE_NUM_MAX];
+
+ odp_pool_t pool; /**< pool to alloc packets from */
+
+ odp_bool_t lockless_rx; /**< no locking for RX */
+ odp_bool_t lockless_tx; /**< no locking for TX */
+
+ odp_pktio_capability_t capa; /**< interface capabilities */
+
+ uint8_t *mmio; /**< MMIO region */
+
+ int sockfd; /**< control socket */
+
+ mdev_device_t mdev; /**< Common mdev data */
+} pktio_ops_e1000e_data_t;
+
+static void e1000e_rx_refill(e1000e_rx_queue_t *rxq, uint16_t from,
+ uint16_t num);
+static void e1000e_wait_link_up(pktio_entry_t *pktio_entry);
+static int e1000e_close(pktio_entry_t *pktio_entry);
+
+static int e1000e_mmio_register(pktio_ops_e1000e_data_t *pkt_e1000e,
+ uint64_t offset, uint64_t size)
+{
+ ODP_ASSERT(pkt_e1000e->mmio == NULL);
+
+ pkt_e1000e->mmio = mdev_region_mmap(&pkt_e1000e->mdev, offset, size);
+ if (pkt_e1000e->mmio == MAP_FAILED) {
+ ODP_ERR("Cannot mmap MMIO\n");
+ return -1;
+ }
+
+ ODP_DBG("Register MMIO region: 0x%llx@%016llx\n", size, offset);
+
+ return 0;
+}
+
+static int e1000e_rx_queue_register(pktio_ops_e1000e_data_t *pkt_e1000e,
+ uint64_t offset, uint64_t size)
+{
+ uint16_t rxq_idx = pkt_e1000e->capa.max_input_queues++;
+ e1000e_rx_queue_t *rxq = &pkt_e1000e->rx_queues[rxq_idx];
+ struct ethtool_ringparam ering;
+ int ret;
+
+ ODP_ASSERT(rxq_idx < ARRAY_SIZE(pkt_e1000e->rx_queues));
+
+ odp_ticketlock_init(&rxq->lock);
+
+ ret = ethtool_ringparam_get_fd(pkt_e1000e->sockfd,
+ pkt_e1000e->mdev.if_name, &ering);
+ if (ret) {
+ ODP_ERR("Cannot get queue length\n");
+ return -1;
+ }
+ rxq->rx_queue_len = ering.rx_pending;
+
+ rxq->doorbell =
+ (odp_u32le_t *)(void *)(pkt_e1000e->mmio + E1000_RDT_OFFSET);
+
+ ODP_ASSERT(rxq->rx_queue_len * sizeof(*rxq->rx_descs) <= size);
+
+ rxq->rx_descs = mdev_region_mmap(&pkt_e1000e->mdev, offset, size);
+ if (rxq->rx_descs == MAP_FAILED) {
+ ODP_ERR("Cannot mmap RX queue\n");
+ return -1;
+ }
+
+ rxq->pidx =
+ (odp_u32le_t *)(void *)(pkt_e1000e->mmio + E1000_RDH_OFFSET);
+
+ rxq->rx_data.size = rxq->rx_queue_len * E1000E_RX_BUF_SIZE;
+ ret = mdev_dma_area_alloc(&pkt_e1000e->mdev, &rxq->rx_data);
+ if (ret) {
+ ODP_ERR("Cannot allocate RX queue DMA area\n");
+ return -1;
+ }
+
+ /* Need 1 desc gap to keep tail from touching head */
+ e1000e_rx_refill(rxq, 0, rxq->rx_queue_len - 1);
+
+ ODP_DBG("Register RX queue region: 0x%llx@%016llx\n", size, offset);
+ ODP_DBG(" RX descriptors: %u\n", rxq->rx_queue_len);
+
+ return 0;
+}
+
+static int e1000e_tx_queue_register(pktio_ops_e1000e_data_t *pkt_e1000e,
+ uint64_t offset, uint64_t size)
+{
+ uint16_t txq_idx = pkt_e1000e->capa.max_output_queues++;
+ e1000e_tx_queue_t *txq = &pkt_e1000e->tx_queues[txq_idx];
+ struct ethtool_ringparam ering;
+ int ret;
+
+ ODP_ASSERT(txq_idx < ARRAY_SIZE(pkt_e1000e->tx_queues));
+
+ odp_ticketlock_init(&txq->lock);
+
+ ret = ethtool_ringparam_get_fd(pkt_e1000e->sockfd,
+ pkt_e1000e->mdev.if_name, &ering);
+ if (ret) {
+ ODP_ERR("Cannot get queue length\n");
+ return -1;
+ }
+ txq->tx_queue_len = ering.tx_pending;
+
+ txq->doorbell =
+ (odp_u32le_t *)(void *)(pkt_e1000e->mmio + E1000_TDT_OFFSET);
+
+ ODP_ASSERT(txq->tx_queue_len * sizeof(*txq->tx_descs) <= size);
+
+ txq->tx_descs = mdev_region_mmap(&pkt_e1000e->mdev, offset, size);
+ if (txq->tx_descs == MAP_FAILED) {
+ ODP_ERR("Cannot mmap TX queue\n");
+ return -1;
+ }
+
+ txq->cidx =
+ (odp_u32le_t *)(void *)(pkt_e1000e->mmio + E1000_TDH_OFFSET);
+
+ txq->tx_data.size = txq->tx_queue_len * E1000E_TX_BUF_SIZE;
+ ret = mdev_dma_area_alloc(&pkt_e1000e->mdev, &txq->tx_data);
+ if (ret) {
+ ODP_ERR("Cannot allocate TX queue DMA area\n");
+ return -1;
+ }
+
+ ODP_DBG("Register TX queue region: 0x%llx@%016llx\n", size, offset);
+ ODP_DBG(" TX descriptors: %u\n", txq->tx_queue_len);
+
+ return 0;
+}
+
+static int e1000e_region_info_cb(mdev_device_t *mdev,
+ struct vfio_region_info *region_info)
+{
+ pktio_ops_e1000e_data_t *pkt_e1000e =
+ odp_container_of(mdev, pktio_ops_e1000e_data_t, mdev);
+ mdev_region_class_t class_info;
+
+ if (vfio_get_region_cap_type(region_info, &class_info) < 0) {
+ ODP_ERR("Cannot find class_info in region %u\n",
+ region_info->index);
+ return -1;
+ }
+
+ switch (class_info.type) {
+ case VFIO_NET_MDEV_MMIO:
+ return e1000e_mmio_register(pkt_e1000e,
+ region_info->offset,
+ region_info->size);
+
+ case VFIO_NET_MDEV_RX_RING:
+ return e1000e_rx_queue_register(pkt_e1000e,
+ region_info->offset,
+ region_info->size);
+
+ case VFIO_NET_MDEV_TX_RING:
+ return e1000e_tx_queue_register(pkt_e1000e,
+ region_info->offset,
+ region_info->size);
+
+ default:
+ ODP_ERR("Unexpected region %u (class %u:%u)\n",
+ region_info->index, class_info.type,
+ class_info.subtype);
+ return -1;
+ }
+}
+
+static int e1000e_open(odp_pktio_t id ODP_UNUSED,
+ pktio_entry_t *pktio_entry,
+ const char *resource, odp_pool_t pool)
+{
+ pktio_ops_e1000e_data_t *pkt_e1000e;
+ int ret;
+
+ ODP_ASSERT(pool != ODP_POOL_INVALID);
+
+ if (strncmp(resource, NET_MDEV_PREFIX, strlen(NET_MDEV_PREFIX)))
+ return -1;
+
+ ODP_DBG("%s: probing resource %s\n", MODULE_NAME, resource);
+
+ pkt_e1000e = ODP_OPS_DATA_ALLOC(sizeof(*pkt_e1000e));
+ if (odp_unlikely(pkt_e1000e == NULL)) {
+ ODP_ERR("Failed to allocate pktio_ops_e1000e_data_t struct");
+ return -1;
+ }
+ pktio_entry->s.ops_data = pkt_e1000e;
+
+ memset(pkt_e1000e, 0, sizeof(*pkt_e1000e));
+
+ pkt_e1000e->sockfd = socket(AF_INET, SOCK_DGRAM, 0);
+ if (pkt_e1000e->sockfd == -1) {
+ ODP_ERR("Cannot get device control socket\n");
+ goto out;
+ }
+
+ ret =
+ mdev_device_create(&pkt_e1000e->mdev, MODULE_NAME,
+ resource + strlen(NET_MDEV_PREFIX),
+ e1000e_region_info_cb);
+ if (ret)
+ goto out;
+
+ pkt_e1000e->pool = pool;
+
+ e1000e_wait_link_up(pktio_entry);
+
+ ODP_DBG("%s: open %s is successful\n", MODULE_NAME,
+ pkt_e1000e->mdev.if_name);
+
+ return 0;
+
+out:
+ e1000e_close(pktio_entry);
+ return -1;
+}
+
+static int e1000e_close(pktio_entry_t *pktio_entry)
+{
+ pktio_ops_e1000e_data_t *pkt_e1000e = pktio_entry->s.ops_data;
+
+ ODP_DBG("%s: close %s\n", MODULE_NAME, pkt_e1000e->mdev.if_name);
+
+ mdev_device_destroy(&pkt_e1000e->mdev);
+
+ for (uint16_t i = 0; i < pkt_e1000e->capa.max_input_queues; i++) {
Comment:
Done.
> Mykyta Iziumtsev(MykytaI) wrote:
> Looking at include/config.h.in (where ODP_MDEV comes from) it's not the only
> naming violation. There are at least 7 other pktio (internal) macros which
> don't start with underscore. Please confirm that we really need to add
> underscore for ODP_MDEV and I'll update next patch version with that ...
>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>> @muvarov In this case the variable is being declared at the beginning of
>> scope. The scope in this instance is the block defined by the `for` loop.
>> That citation is saying you shouldn't declare variables in the middle of a
>> scope. While that's still correct C, it can be confusing.
>>> muvarov wrote
>>> ./doc/application-api-guide/api_guide_lines.dox line 162
>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>> @muvarov This idiom is part of C99, so I see no problem with it here. It's
>>>> also why checkpatch doesn't complain.
>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>> We have the various `odpdrv_mmio_uxx_read/write()` APIs. Should we also
>>>>> have `_addr` variants that would similarly enable address loading to
>>>>> handle this sort of awkward casting? E.g.:
>>>>> ```
>>>>> rxq->doorbell = odpdrv_u32le_addr(pkt_e100e->mmio + E1000_RDT_OFFSET);
>>>>> ```
>>>>> Similar comments for other address setups contained here.
>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>> Might be good to document where the various "magic numbers" come from.
>>>>>> Presumably from some hardware manual? Does Intel provide a `.h` file
>>>>>> that we can reference/`#include` for these rather than duplicate them
>>>>>> here?
>>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>>> Agree with @muvarov. `_ODP_MDEV` is fine here. The `ODP` prefix should
>>>>>>> be reserved for things that are part of the external ODP API.
>>>>>>>> muvarov wrote
>>>>>>>> place type out of for(). checkpatch should warn about that.
>>>>>>>>> muvarov wrote
>>>>>>>>> please keep ODP_ and odp_ prefix only for odp api. Internal things
>>>>>>>>> should be named as _ODP_ or without that prefix. This comment not
>>>>>>>>> just for this commit but it has to be corrected.
https://github.com/Linaro/odp/pull/408#discussion_r163485788
updated_at 2018-01-24 08:59:25
