> -----Original Message-----
> From: Hemant Agrawal <hemant.agra...@nxp.com>
> Sent: Monday, April 12, 2021 10:17 PM
>
> This patch add support for connecting with modem and creating the ipc
> channel as queues with modem for the exchange of data.
>
> Signed-off-by: Nipun Gupta <nipun.gu...@nxp.com>
> Signed-off-by: Hemant Agrawal <hemant.agra...@nxp.com>
> ---
> drivers/baseband/la12xx/bbdev_la12xx.c | 564 ++++++++++++++++++++-
> drivers/baseband/la12xx/bbdev_la12xx.h | 5 +-
> drivers/baseband/la12xx/bbdev_la12xx_ipc.h | 180 +++++++
> 3 files changed, 746 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/baseband/la12xx/bbdev_la12xx.c
> b/drivers/baseband/la12xx/bbdev_la12xx.c
> index 7e9be74bb4..0a68686205 100644
> --- a/drivers/baseband/la12xx/bbdev_la12xx.c
> +++ b/drivers/baseband/la12xx/bbdev_la12xx.c
> @@ -3,6 +3,11 @@
> */
>
> #include <string.h>
> +#include <unistd.h>
> +#include <fcntl.h>
> +#include <sys/ioctl.h>
> +#include <sys/mman.h>
> +#include <dirent.h>
>
> #include <rte_common.h>
> #include <rte_bus_vdev.h>
> @@ -31,11 +36,561 @@ struct bbdev_la12xx_params {
> #define BBDEV_LA12XX_VDEV_MODEM_ID_ARG "modem"
> #define LA12XX_MAX_MODEM 4
>
> +#define LA12XX_MAX_CORES 4
> +#define BBDEV_LA12XX_LDPC_ENC_CORE 0
> +#define BBDEV_LA12XX_LDPC_DEC_CORE 1
> +
> static const char * const bbdev_la12xx_valid_params[] = {
> BBDEV_LA12XX_MAX_NB_QUEUES_ARG,
> BBDEV_LA12XX_VDEV_MODEM_ID_ARG,
> };
>
> +static const struct rte_bbdev_op_cap bbdev_capabilities[] = {
> + {
> + .type = RTE_BBDEV_OP_LDPC_ENC,
> + .cap.ldpc_enc = {
> + .capability_flags =
> + RTE_BBDEV_LDPC_CRC_24A_ATTACH
> |
> +
> RTE_BBDEV_LDPC_CRC_24B_ATTACH,
I believe you also support RTE_BBDEV_LDPC_RATE_MATCH based on future commits.
Still in that very commit I dont believe support anything yet
> + .num_buffers_src =
> +
> RTE_BBDEV_LDPC_MAX_CODE_BLOCKS,
> + .num_buffers_dst =
> +
> RTE_BBDEV_LDPC_MAX_CODE_BLOCKS,
> + }
> + },
> + {
> + .type = RTE_BBDEV_OP_LDPC_DEC,
> + .cap.ldpc_dec = {
> + .capability_flags =
> + RTE_BBDEV_LDPC_CRC_TYPE_24A_CHECK |
> +
> RTE_BBDEV_LDPC_CRC_TYPE_24B_CHECK |
> +
> RTE_BBDEV_LDPC_CRC_TYPE_24B_DROP,
Similar comment, on that very commit I dont believe you support anything yet.
Future commits and documentation suggests you support more than these.
> + .num_buffers_src =
> +
> RTE_BBDEV_LDPC_MAX_CODE_BLOCKS,
> + .num_buffers_hard_out =
> +
> RTE_BBDEV_LDPC_MAX_CODE_BLOCKS,
> + }
> + },
> + RTE_BBDEV_END_OF_CAPABILITIES_LIST()
> +};
> +
> +static struct rte_bbdev_queue_conf default_queue_conf = {
> + .queue_size = MAX_CHANNEL_DEPTH,
> +};
> +
> +/* Get device info */
> +static void
> +la12xx_info_get(struct rte_bbdev *dev,
> + struct rte_bbdev_driver_info *dev_info) {
> + PMD_INIT_FUNC_TRACE();
> +
> + dev_info->driver_name = RTE_STR(DRIVER_NAME);
> + dev_info->max_num_queues = LA12XX_MAX_QUEUES;
> + dev_info->queue_size_lim = MAX_CHANNEL_DEPTH;
> + dev_info->hardware_accelerated = true;
> + dev_info->max_dl_queue_priority = 0;
> + dev_info->max_ul_queue_priority = 0;
> + dev_info->default_queue_conf = default_queue_conf;
> + dev_info->capabilities = bbdev_capabilities;
> + dev_info->cpu_flag_reqs = NULL;
> + dev_info->min_alignment = 64;
> +
> + BBDEV_LA12XX_PMD_DEBUG("got device info from %u", dev->data-
> >dev_id);
> +}
> +
> +/* Release queue */
> +static int
> +la12xx_queue_release(struct rte_bbdev *dev, uint16_t q_id) {
> + RTE_SET_USED(dev);
> + RTE_SET_USED(q_id);
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + /* TODO: Implement */
I believe this should be implemented in that commit. Not added later in the
serie from what I could see.
> +
> + return 0;
> +}
> +
> +#define HUGEPG_OFFSET(A) \
> + ((uint64_t) ((unsigned long) (A) \
> + - ((uint64_t)ipc_priv->hugepg_start.host_vaddr)))
> +
> +static int ipc_queue_configure(uint32_t channel_id,
> + ipc_t instance, struct bbdev_la12xx_q_priv *q_priv) {
> + ipc_userspace_t *ipc_priv = (ipc_userspace_t *)instance;
> + ipc_instance_t *ipc_instance = ipc_priv->instance;
> + ipc_ch_t *ch;
> + void *vaddr;
> + uint32_t i = 0;
> + uint32_t msg_size = sizeof(struct bbdev_ipc_enqueue_op);
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + BBDEV_LA12XX_PMD_DEBUG("%x %p", ipc_instance->initialized,
> + ipc_priv->instance);
> + ch = &(ipc_instance->ch_list[channel_id]);
> +
> + BBDEV_LA12XX_PMD_DEBUG("channel: %u, depth: %u, msg size:
> %u",
> + channel_id, q_priv->queue_size, msg_size);
> +
> + /* Start init of channel */
> + ch->md.ring_size = rte_cpu_to_be_32(q_priv->queue_size);
> + ch->md.pi = 0;
> + ch->md.ci = 0;
> + ch->md.msg_size = msg_size;
> + for (i = 0; i < q_priv->queue_size; i++) {
> + vaddr = rte_malloc(NULL, msg_size, RTE_CACHE_LINE_SIZE);
> + if (!vaddr)
> + return IPC_HOST_BUF_ALLOC_FAIL;
> + /* Only offset now */
> + ch->bd[i].modem_ptr =
> + rte_cpu_to_be_32(HUGEPG_OFFSET(vaddr));
> + ch->bd[i].host_virt_l = lower_32_bits(vaddr);
> + ch->bd[i].host_virt_h = upper_32_bits(vaddr);
> + q_priv->msg_ch_vaddr[i] = vaddr;
> + /* Not sure use of this len may be for CRC*/
What is that comment refering to? Is that a TODO?
> + ch->bd[i].len = 0;
> + }
> + ch->host_ipc_params =
> + rte_cpu_to_be_32(HUGEPG_OFFSET(q_priv->host_params));
> + ch->bl_initialized = 1;
> +
> + BBDEV_LA12XX_PMD_DEBUG("Channel configured");
> + return IPC_SUCCESS;
> +}
> +
> +static int
> +la12xx_e200_queue_setup(struct rte_bbdev *dev,
> + struct bbdev_la12xx_q_priv *q_priv)
> +{
> + struct bbdev_la12xx_private *priv = dev->data->dev_private;
> + ipc_userspace_t *ipc_priv = priv->ipc_priv;
> + struct gul_hif *mhif;
> + ipc_metadata_t *ipc_md;
> + ipc_ch_t *ch;
> + int instance_id = 0, i;
> + int ret;
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + switch (q_priv->op_type) {
> + case RTE_BBDEV_OP_LDPC_ENC:
> + q_priv->la12xx_core_id = BBDEV_LA12XX_LDPC_ENC_CORE;
> + break;
> + case RTE_BBDEV_OP_LDPC_DEC:
> + q_priv->la12xx_core_id = BBDEV_LA12XX_LDPC_DEC_CORE;
> + break;
> + default:
> + BBDEV_LA12XX_PMD_ERR("Unsupported op type\n");
> + return -1;
> + }
> +
> + mhif = (struct gul_hif *)ipc_priv->mhif_start.host_vaddr;
> + /* offset is from start of PEB */
> + ipc_md = (ipc_metadata_t *)((size_t)ipc_priv->peb_start.host_vaddr
> +
> + mhif->ipc_regs.ipc_mdata_offset);
> + ch = &ipc_md->instance_list[instance_id].ch_list[q_priv->q_id];
> +
> + if (q_priv->q_id < priv->num_valid_queues) {
> + ipc_br_md_t *md, *host_md;
> + ipc_ch_t *host_rx_ch;
> +
> + host_rx_ch =
> + &ipc_md->instance_list[instance_id].ch_list[q_priv-
> >q_id
> + + HOST_RX_QUEUEID_OFFSET];
> + md = &(ch->md);
> + host_md = &(host_rx_ch->md);
> +
> + q_priv->feca_blk_id = rte_cpu_to_be_32(ch->feca_blk_id);
> + q_priv->feca_blk_id_be32 = ch->feca_blk_id;
> + q_priv->host_pi = rte_be_to_cpu_32(host_md->pi);
> + q_priv->host_ci = rte_be_to_cpu_32(md->ci);
> + q_priv->host_params = (host_ipc_params_t *)
> + (rte_be_to_cpu_32(ch->host_ipc_params) +
> + ((size_t)ipc_priv->hugepg_start.host_vaddr));
> +
> + for (i = 0; i < q_priv->queue_size; i++) {
> + uint32_t h, l;
> +
> + h = host_rx_ch->bd[i].host_virt_h;
> + l = host_rx_ch->bd[i].host_virt_l;
> + q_priv->msg_ch_vaddr[i] = (void *)join_32_bits(h, l);
> + }
> +
> + BBDEV_LA12XX_PMD_WARN(
> + "Queue [%d] already configured, not configuring
> again",
> + q_priv->q_id);
> + return 0;
> + }
> +
> + BBDEV_LA12XX_PMD_DEBUG("setting up queue %d", q_priv->q_id);
> +
> + q_priv->host_params = rte_zmalloc(NULL,
> sizeof(host_ipc_params_t),
> + RTE_CACHE_LINE_SIZE);
> + ch->host_ipc_params =
> + rte_cpu_to_be_32(HUGEPG_OFFSET(q_priv->host_params));
> +
> + /* Call ipc_configure_channel */
> + ret = ipc_queue_configure((q_priv->q_id +
> HOST_RX_QUEUEID_OFFSET),
> + ipc_priv, q_priv);
> + if (ret) {
> + BBDEV_LA12XX_PMD_ERR("Unable to setup queue (%d)
> (err=%d)",
> + q_priv->q_id, ret);
> + return ret;
> + }
> +
> + /* Set queue properties for LA12xx device */
> + switch (q_priv->op_type) {
> + case RTE_BBDEV_OP_LDPC_ENC:
> + if (priv->num_ldpc_enc_queues >=
> MAX_LDPC_ENC_FECA_QUEUES) {
> + BBDEV_LA12XX_PMD_ERR(
> + "num_ldpc_enc_queues reached max
> value");
> + return -1;
> + }
> + ch->la12xx_core_id =
> +
> rte_cpu_to_be_32(BBDEV_LA12XX_LDPC_ENC_CORE);
> + ch->feca_blk_id = rte_cpu_to_be_32(priv-
> >num_ldpc_enc_queues++);
> + break;
> + case RTE_BBDEV_OP_LDPC_DEC:
> + if (priv->num_ldpc_dec_queues >=
> MAX_LDPC_DEC_FECA_QUEUES) {
> + BBDEV_LA12XX_PMD_ERR(
> + "num_ldpc_dec_queues reached max
> value");
> + return -1;
> + }
> + ch->la12xx_core_id =
> +
> rte_cpu_to_be_32(BBDEV_LA12XX_LDPC_DEC_CORE);
> + ch->feca_blk_id = rte_cpu_to_be_32(priv-
> >num_ldpc_dec_queues++);
> + break;
> + default:
> + BBDEV_LA12XX_PMD_ERR("Not supported op type\n");
> + return -1;
> + }
> + ch->op_type = rte_cpu_to_be_32(q_priv->op_type);
> + ch->depth = rte_cpu_to_be_32(q_priv->queue_size);
> +
> + /* Store queue config here */
> + q_priv->feca_blk_id = rte_cpu_to_be_32(ch->feca_blk_id);
> + q_priv->feca_blk_id_be32 = ch->feca_blk_id;
> +
> + return 0;
> +}
> +
> +/* Setup a queue */
> +static int
> +la12xx_queue_setup(struct rte_bbdev *dev, uint16_t q_id,
> + const struct rte_bbdev_queue_conf *queue_conf) {
> + struct bbdev_la12xx_private *priv = dev->data->dev_private;
> + struct rte_bbdev_queue_data *q_data;
> + struct bbdev_la12xx_q_priv *q_priv;
> + int ret;
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + /* Move to setup_queues callback */
> + q_data = &dev->data->queues[q_id];
> + q_data->queue_private = rte_zmalloc(NULL,
> + sizeof(struct bbdev_la12xx_q_priv), 0);
> + if (!q_data->queue_private) {
> + BBDEV_LA12XX_PMD_ERR("Memory allocation failed for
> qpriv");
> + return -ENOMEM;
> + }
> + q_priv = q_data->queue_private;
> + q_priv->q_id = q_id;
> + q_priv->bbdev_priv = dev->data->dev_private;
> + q_priv->queue_size = queue_conf->queue_size;
> + q_priv->op_type = queue_conf->op_type;
> +
> + ret = la12xx_e200_queue_setup(dev, q_priv);
> + if (ret) {
> + BBDEV_LA12XX_PMD_ERR("e200_queue_setup failed for
> qid: %d",
> + q_id);
> + return ret;
> + }
> +
> + /* Store queue config here */
> + priv->num_valid_queues++;
> +
> + return 0;
> +}
> +
> +static int
> +la12xx_start(struct rte_bbdev *dev)
> +{
> + struct bbdev_la12xx_private *priv = dev->data->dev_private;
> + ipc_userspace_t *ipc_priv = priv->ipc_priv;
> + int ready = 0;
> + struct gul_hif *hif_start;
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + hif_start = (struct gul_hif *)ipc_priv->mhif_start.host_vaddr;
> +
> + /* Set Host Read bit */
> + SET_HIF_HOST_RDY(hif_start, HIF_HOST_READY_IPC_APP);
> +
> + /* Now wait for modem ready bit */
> + while (!ready)
> + ready = CHK_HIF_MOD_RDY(hif_start,
> HIF_MOD_READY_IPC_APP);
> +
> + return 0;
> +}
> +
> +static const struct rte_bbdev_ops pmd_ops = {
> + .info_get = la12xx_info_get,
> + .queue_setup = la12xx_queue_setup,
> + .queue_release = la12xx_queue_release,
> + .start = la12xx_start
> +};
> +static struct hugepage_info *
> +get_hugepage_info(void)
> +{
> + struct hugepage_info *hp_info;
> + struct rte_memseg *mseg;
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + /* TODO - find a better way */
> + hp_info = rte_malloc(NULL, sizeof(struct hugepage_info), 0);
> + if (!hp_info) {
> + BBDEV_LA12XX_PMD_ERR("Unable to allocate on local
> heap");
> + return NULL;
> + }
> +
> + mseg = rte_mem_virt2memseg(hp_info, NULL);
> + hp_info->vaddr = mseg->addr;
> + hp_info->paddr = rte_mem_virt2phy(mseg->addr);
> + hp_info->len = mseg->len;
> +
> + return hp_info;
> +}
I was not sure what this is trying to do. Is that assuming a single huge page
being used and keep track of it for future DMA activity? Is that valid?
> +
> +static int open_ipc_dev(int modem_id)
> +{
> + char dev_initials[16], dev_path[PATH_MAX];
> + struct dirent *entry;
> + int dev_ipc = 0;
> + DIR *dir;
> +
> + dir = opendir("/dev/");
> + if (!dir) {
> + BBDEV_LA12XX_PMD_ERR("Unable to open /dev/");
> + return -1;
> + }
> +
> + sprintf(dev_initials, "gulipcgul%d", modem_id);
> +
> + while ((entry = readdir(dir)) != NULL) {
> + if (!strncmp(dev_initials, entry->d_name,
> + sizeof(dev_initials) - 1))
> + break;
> + }
> +
> + if (!entry) {
> + BBDEV_LA12XX_PMD_ERR("Error: No gulipcgul%d device",
> modem_id);
> + return -1;
> + }
> +
> + sprintf(dev_path, "/dev/%s", entry->d_name);
> + dev_ipc = open(dev_path, O_RDWR);
> + if (dev_ipc < 0) {
> + BBDEV_LA12XX_PMD_ERR("Error: Cannot open %s",
> dev_path);
> + return -errno;
> + }
> +
> + return dev_ipc;
> +}
> +
> +static int
> +setup_la12xx_dev(struct rte_bbdev *dev) {
> + struct bbdev_la12xx_private *priv = dev->data->dev_private;
> + ipc_userspace_t *ipc_priv = priv->ipc_priv;
> + struct hugepage_info *hp = NULL;
> + ipc_channel_us_t *ipc_priv_ch = NULL;
> + int dev_ipc = 0, dev_mem = 0, i;
> + ipc_metadata_t *ipc_md;
> + struct gul_hif *mhif;
> + uint32_t phy_align = 0;
> + int ret;
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + if (!ipc_priv) {
> + /* TODO - get a better way */
> + /* Get the hugepage info against it */
> + hp = get_hugepage_info();
> + if (!hp) {
> + BBDEV_LA12XX_PMD_ERR("Unable to get hugepage
> info");
> + ret = -ENOMEM;
> + goto err;
> + }
> +
> + BBDEV_LA12XX_PMD_DEBUG("0x%" PRIx64 " %p %x",
> + hp->paddr, hp->vaddr, (int)hp->len);
> +
> + ipc_priv = rte_zmalloc(0, sizeof(ipc_userspace_t), 0);
> + if (ipc_priv == NULL) {
> + BBDEV_LA12XX_PMD_ERR(
> + "Unable to allocate memory for ipc priv");
> + ret = -ENOMEM;
> + goto err;
> + }
> +
> + for (i = 0; i < IPC_MAX_CHANNEL_COUNT; i++) {
> + ipc_priv_ch = rte_zmalloc(0,
> + sizeof(ipc_channel_us_t), 0);
> + if (ipc_priv_ch == NULL) {
> + BBDEV_LA12XX_PMD_ERR(
> + "Unable to allocate memory for
> channels");
> + ret = -ENOMEM;
> + }
> + ipc_priv->channels[i] = ipc_priv_ch;
> + }
> +
> + dev_mem = open("/dev/mem", O_RDWR);
> + if (dev_mem < 0) {
> + BBDEV_LA12XX_PMD_ERR("Error: Cannot open
> /dev/mem");
> + ret = -errno;
> + goto err;
> + }
> +
> + ipc_priv->instance_id = 0;
> + ipc_priv->dev_mem = dev_mem;
> +
> + ipc_priv->sys_map.hugepg_start.host_phys = hp->paddr;
> + ipc_priv->sys_map.hugepg_start.size = hp->len;
> +
> + ipc_priv->hugepg_start.host_phys = hp->paddr;
> + ipc_priv->hugepg_start.host_vaddr = hp->vaddr;
> + ipc_priv->hugepg_start.size = hp->len;
> +
> + rte_free(hp);
> + }
> +
> + dev_ipc = open_ipc_dev(priv->modem_id);
> + if (dev_ipc < 0) {
> + BBDEV_LA12XX_PMD_ERR("Error: open_ipc_dev failed");
> + goto err;
> + }
> + ipc_priv->dev_ipc = dev_ipc;
> +
> + /* Send IOCTL to get system map */
> + /* Send IOCTL to put hugepg_start map */
> + ret = ioctl(ipc_priv->dev_ipc, IOCTL_GUL_IPC_GET_SYS_MAP,
> + &ipc_priv->sys_map);
> + if (ret) {
> + BBDEV_LA12XX_PMD_ERR(
> + "IOCTL_GUL_IPC_GET_SYS_MAP ioctl failed");
> + goto err;
> + }
> +
> + phy_align = (ipc_priv->sys_map.mhif_start.host_phys % 0x1000);
I would replace the 0x1000 magic number used for alignment
> + ipc_priv->mhif_start.host_vaddr =
> + mmap(0, ipc_priv->sys_map.mhif_start.size + phy_align,
> + (PROT_READ | PROT_WRITE), MAP_SHARED, ipc_priv-
> >dev_mem,
> + (ipc_priv->sys_map.mhif_start.host_phys - phy_align));
> + if (ipc_priv->mhif_start.host_vaddr == MAP_FAILED) {
> + BBDEV_LA12XX_PMD_ERR("MAP failed:");
> + ret = -errno;
> + goto err;
> + }
> +
> + ipc_priv->mhif_start.host_vaddr = (void *)((size_t)
> + (ipc_priv->mhif_start.host_vaddr) + phy_align);
> +
> + phy_align = (ipc_priv->sys_map.peb_start.host_phys % 0x1000);
> + ipc_priv->peb_start.host_vaddr =
> + mmap(0, ipc_priv->sys_map.peb_start.size + phy_align,
> + (PROT_READ | PROT_WRITE), MAP_SHARED, ipc_priv-
> >dev_mem,
> + (ipc_priv->sys_map.peb_start.host_phys - phy_align));
> + if (ipc_priv->peb_start.host_vaddr == MAP_FAILED) {
> + BBDEV_LA12XX_PMD_ERR("MAP failed:");
> + ret = -errno;
> + goto err;
> + }
> +
> + ipc_priv->peb_start.host_vaddr = (void *)((size_t)
> + (ipc_priv->peb_start.host_vaddr) + phy_align);
> +
> + phy_align = (ipc_priv->sys_map.modem_ccsrbar.host_phys %
> 0x1000);
> + ipc_priv->modem_ccsrbar.host_vaddr =
> + mmap(0, ipc_priv->sys_map.modem_ccsrbar.size +
> phy_align,
> + (PROT_READ | PROT_WRITE), MAP_SHARED, ipc_priv-
> >dev_mem,
> + (ipc_priv->sys_map.modem_ccsrbar.host_phys -
> phy_align));
> + if (ipc_priv->modem_ccsrbar.host_vaddr == MAP_FAILED) {
> + BBDEV_LA12XX_PMD_ERR("MAP failed:");
> + ret = -errno;
> + goto err;
> + }
> +
> + ipc_priv->modem_ccsrbar.host_vaddr = (void *)((size_t)
> + (ipc_priv->modem_ccsrbar.host_vaddr) + phy_align);
> +
> + ipc_priv->hugepg_start.modem_phys =
> + ipc_priv->sys_map.hugepg_start.modem_phys;
> +
> + ipc_priv->mhif_start.host_phys =
> + ipc_priv->sys_map.mhif_start.host_phys;
> + ipc_priv->mhif_start.size = ipc_priv->sys_map.mhif_start.size;
> + ipc_priv->peb_start.host_phys = ipc_priv-
> >sys_map.peb_start.host_phys;
> + ipc_priv->peb_start.size = ipc_priv->sys_map.peb_start.size;
> +
> + BBDEV_LA12XX_PMD_INFO("peb 0x%" PRIx64 " %p %x",
> + ipc_priv->peb_start.host_phys,
> + ipc_priv->peb_start.host_vaddr,
> + ipc_priv->peb_start.size);
> + BBDEV_LA12XX_PMD_INFO("hugepg 0x%" PRIx64 " %p %x",
> + ipc_priv->hugepg_start.host_phys,
> + ipc_priv->hugepg_start.host_vaddr,
> + ipc_priv->hugepg_start.size);
> + BBDEV_LA12XX_PMD_INFO("mhif 0x%" PRIx64 " %p %x",
> + ipc_priv->mhif_start.host_phys,
> + ipc_priv->mhif_start.host_vaddr,
> + ipc_priv->mhif_start.size);
> + mhif = (struct gul_hif *)ipc_priv->mhif_start.host_vaddr;
> +
> + /* offset is from start of PEB */
> + ipc_md = (ipc_metadata_t *)((size_t)ipc_priv->peb_start.host_vaddr
> +
> + mhif->ipc_regs.ipc_mdata_offset);
> +
> + if (sizeof(ipc_metadata_t) != mhif->ipc_regs.ipc_mdata_size) {
> + BBDEV_LA12XX_PMD_ERR(
> + "\n ipc_metadata_t=%d mhif-
> >ipc_regs.ipc_mdata_size=%x",
> + (int)sizeof(ipc_metadata_t),
> + mhif->ipc_regs.ipc_mdata_size);
> + BBDEV_LA12XX_PMD_ERR(
> + "--> mhif->ipc_regs.ipc_mdata_offset= %x",
> + mhif->ipc_regs.ipc_mdata_offset);
> + BBDEV_LA12XX_PMD_ERR(
> + "gul_hif size=%d", (int)sizeof(struct gul_hif));
> + return IPC_MD_SZ_MISS_MATCH;
> + }
> +
> + ipc_priv->instance = (ipc_instance_t *)
> + (&ipc_md->instance_list[ipc_priv->instance_id]);
> +
> + BBDEV_LA12XX_PMD_DEBUG("finish host init");
> +
> + priv->ipc_priv = ipc_priv;
> +
> + return 0;
> +
> +err:
> + rte_free(hp);
> + rte_free(ipc_priv);
> + rte_free(ipc_priv_ch);
> + if (dev_mem)
> + close(dev_mem);
> + if (dev_ipc)
> + close(dev_ipc);
> +
> + return ret;
> +}
> +
> static inline int
> parse_u16_arg(const char *key, const char *value, void *extra_args) { @@ -
> 123,6 +678,7 @@ la12xx_bbdev_create(struct rte_vdev_device *vdev,
> struct rte_bbdev *bbdev;
> const char *name = rte_vdev_device_name(vdev);
> struct bbdev_la12xx_private *priv;
> + int ret;
>
> PMD_INIT_FUNC_TRACE();
>
> @@ -152,7 +708,13 @@ la12xx_bbdev_create(struct rte_vdev_device
> *vdev,
>
> BBDEV_LA12XX_PMD_INFO("Setting Up %s: DevId=%d,
> ModemId=%d",
> name, bbdev->data->dev_id, priv-
> >modem_id);
> - bbdev->dev_ops = NULL;
> + ret = setup_la12xx_dev(bbdev);
> + if (ret) {
> + BBDEV_LA12XX_PMD_ERR("IPC Setup failed for %s", name);
> + rte_free(bbdev->data->dev_private);
> + return ret;
> + }
> + bbdev->dev_ops = &pmd_ops;
> bbdev->device = &vdev->device;
> bbdev->data->socket_id = 0;
> bbdev->intr_handle = NULL;
> diff --git a/drivers/baseband/la12xx/bbdev_la12xx.h
> b/drivers/baseband/la12xx/bbdev_la12xx.h
> index 5228502331..c94f08e059 100644
> --- a/drivers/baseband/la12xx/bbdev_la12xx.h
> +++ b/drivers/baseband/la12xx/bbdev_la12xx.h
> @@ -14,7 +14,7 @@
> #define MAX_CHANNEL_DEPTH 16
> /* private data structure */
> struct bbdev_la12xx_private {
> - void *ipc_priv;
> + ipc_userspace_t *ipc_priv;
> uint8_t num_valid_queues;
> uint8_t max_nb_queues;
> uint8_t num_ldpc_enc_queues;
> @@ -52,5 +52,6 @@ struct bbdev_la12xx_q_priv {
>
> #define lower_32_bits(x) ((uint32_t)((uint64_t)x)) #define upper_32_bits(x)
> ((uint32_t)(((uint64_t)(x) >> 16) >> 16))
> -
> +#define join_32_bits(upper, lower) \
> + ((size_t)(((uint64_t)(upper) << 32) | (uint32_t)(lower)))
> #endif
> diff --git a/drivers/baseband/la12xx/bbdev_la12xx_ipc.h
> b/drivers/baseband/la12xx/bbdev_la12xx_ipc.h
> index 9aa5562981..9d5789f726 100644
> --- a/drivers/baseband/la12xx/bbdev_la12xx_ipc.h
> +++ b/drivers/baseband/la12xx/bbdev_la12xx_ipc.h
> @@ -4,9 +4,175 @@
> #ifndef __BBDEV_LA12XX_IPC_H__
> #define __BBDEV_LA12XX_IPC_H__
>
> +#define LA12XX_MAX_QUEUES 20
> +
> +/** No. of max channel per instance */
> +#define IPC_MAX_CHANNEL_COUNT (64)
> +
> /** No. of max channel per instance */
> #define IPC_MAX_DEPTH (16)
>
> +/** No. of max IPC instance per modem */
> +#define IPC_MAX_INSTANCE_COUNT (1)
> +
> +#define HOST_RX_QUEUEID_OFFSET LA12XX_MAX_QUEUES
> +
> +#define MAX_MEM_POOL_COUNT 8
> +
> +/** Error codes */
> +#define IPC_SUCCESS (0) /** IPC operation success */
> +#define IPC_INPUT_INVALID (-1) /** Invalid input to API */
> +#define IPC_CH_INVALID (-2) /** Channel no is invalid */
> +#define IPC_INSTANCE_INVALID (-3) /** Instance no is invalid */
> +#define IPC_MEM_INVALID (-4) /** Insufficient memory */
> +#define IPC_CH_FULL (-5) /** Channel is full */
> +#define IPC_CH_EMPTY (-6) /** Channel is empty */
> +#define IPC_BL_EMPTY (-7) /** Free buffer list is empty */
> +#define IPC_BL_FULL (-8) /** Free buffer list is full */
> +#define IPC_HOST_BUF_ALLOC_FAIL (-9) /** DPDK malloc fail */
> +#define IPC_MD_SZ_MISS_MATCH (-10) /** META DATA sz in mhif miss
> matched*/
> +#define IPC_MALLOC_FAIL (-11) /** system malloc fail */
> +#define IPC_IOCTL_FAIL (-12) /** IOCTL call failed */
> +#define IPC_MMAP_FAIL (-14) /** MMAP fail */
> +#define IPC_OPEN_FAIL (-15) /** OPEN fail */
> +#define IPC_EVENTFD_FAIL (-16) /** eventfd initialization failed */
> +#define IPC_NOT_IMPLEMENTED (-17) /** IPC feature is not
> implemented yet*/
> +
> +#define SET_HIF_HOST_RDY(hif, RDY_MASK) (hif->host_ready |=
> RDY_MASK)
> +#define CHK_HIF_MOD_RDY(hif, RDY_MASK) (hif->mod_ready &
> RDY_MASK)
> +
> +/* Host Ready bits */
> +#define HIF_HOST_READY_HOST_REGIONS (1 << 0)
> +#define HIF_HOST_READY_IPC_LIB (1 << 12)
> +#define HIF_HOST_READY_IPC_APP (1 << 13)
> +#define HIF_HOST_READY_FECA (1 << 14)
> +
> +/* Modem Ready bits */
> +#define HIF_MOD_READY_IPC_LIB (1 << 5)
> +#define HIF_MOD_READY_IPC_APP (1 << 6)
> +#define HIF_MOD_READY_FECA (1 << 7)
> +
> +typedef void *ipc_t;
> +
> +struct ipc_msg {
> + int chid;
> + void *addr;
> + uint32_t len;
> + uint8_t flags;
> +};
> +
> +typedef struct {
> + uint64_t host_phys;
> + uint32_t modem_phys;
> + void *host_vaddr;
> + uint32_t size;
> +} mem_range_t;
> +
> +#define GUL_IPC_MAGIC 'R'
> +
> +#define IOCTL_GUL_IPC_GET_SYS_MAP _IOW(GUL_IPC_MAGIC, 1, struct
> ipc_msg
> +*) #define IOCTL_GUL_IPC_CHANNEL_REGISTER _IOWR(GUL_IPC_MAGIC, 4,
> +struct ipc_msg *) #define IOCTL_GUL_IPC_CHANNEL_DEREGISTER \
> + _IOWR(GUL_IPC_MAGIC, 5, struct ipc_msg *) #define
> +IOCTL_GUL_IPC_CHANNEL_RAISE_INTERRUPT _IOW(GUL_IPC_MAGIC, 6, int
> *)
> +
> +/** buffer ring common metadata */
> +typedef struct ipc_bd_ring_md {
> + volatile uint32_t pi; /**< Producer index and flag (MSB)
> + * which flip for each Ring wrapping
> + */
> + volatile uint32_t ci; /**< Consumer index and flag (MSB)
> + * which flip for each Ring wrapping
> + */
> + uint32_t ring_size; /**< depth (Used to roll-over pi/ci) */
> + uint32_t msg_size; /**< Size of the each buffer */
> +} __rte_packed ipc_br_md_t;
> +
> +/** IPC buffer descriptor */
> +typedef struct ipc_buffer_desc {
> + union {
> + uint64_t host_virt; /**< msg's host virtual address */
> + struct {
> + uint32_t host_virt_l;
> + uint32_t host_virt_h;
> + };
> + };
> + uint64_t host_phy; /**< msg's host physical address */
No intention to use rte_iova_t here?
> + uint32_t modem_ptr; /**< msg's modem physical address */
> + uint32_t len; /**< msg len */
> + uint64_t crc; /**< crc */
> +} __rte_packed ipc_bd_t;
> +
> +typedef struct ipc_channel {
> + uint32_t ch_id; /**< Channel id */
> + uint32_t bl_initialized;/**< Set when buffer list is initialized */
> + ipc_br_md_t md;
> + ipc_bd_t bd[IPC_MAX_DEPTH];
> + uint32_t op_type;/* BBDEV operation supported on this channel */
> + uint32_t depth; /* Channel depth */
> + uint32_t feca_blk_id;/* FECA Transport Block ID for processing */
> + uint32_t la12xx_core_id;/* LA12xx core ID to scheduled work for it*/
> + uint32_t host_ipc_params;/* Address for host IPC parameters */ }
> +__rte_packed ipc_ch_t;
> +
> +typedef struct ipc_instance {
> + uint32_t initialized; /**< Set in ipc_init */
> + ipc_ch_t ch_list[IPC_MAX_CHANNEL_COUNT];
> + /**< Channel descriptors in this instance */ } __rte_packed
> +ipc_instance_t;
> +
> +typedef struct ipc_metadata {
> + ipc_instance_t instance_list[IPC_MAX_INSTANCE_COUNT];
> +} __rte_packed ipc_metadata_t;
> +
> +typedef struct ipc_channel_us_priv {
> + int32_t eventfd;
> + uint32_t channel_id;
> + /* In flight packets status for buffer list. */
> + uint8_t bufs_inflight[IPC_MAX_DEPTH];
> +} ipc_channel_us_t;
> +
> +typedef struct {
> + uint64_t host_phys;
> + uint32_t modem_phys;
> + uint32_t size;
> +} mem_strt_addr_t;
> +
> +typedef struct {
> + mem_strt_addr_t modem_ccsrbar;
> + mem_strt_addr_t peb_start; /* PEB meta data */
> + mem_strt_addr_t mhif_start; /* MHIF meta daat */
> + mem_strt_addr_t hugepg_start; /* Modem to access hugepage */ }
> +sys_map_t;
> +
> +typedef struct ipc_priv_t {
> + int instance_id;
> + int dev_ipc;
> + int dev_mem;
> + struct rte_mempool *rtemempool[MAX_MEM_POOL_COUNT];
> + sys_map_t sys_map;
> + mem_range_t modem_ccsrbar;
> + mem_range_t peb_start;
> + mem_range_t mhif_start;
> + mem_range_t hugepg_start;
> + ipc_channel_us_t *channels[IPC_MAX_CHANNEL_COUNT];
> + ipc_instance_t *instance;
> +} ipc_userspace_t;
> +
> +/** Structure specifying enqueue operation (enqueue at LA1224) */
> +struct bbdev_ipc_enqueue_op {
> + /** Status of operation that was performed */
> + int32_t status;
> + /** CRC Status of SD operation that was performed */
> + int32_t crc_stat_addr;
> + /** HARQ Output buffer memory length for Shared Decode.
> + * Filled by LA12xx.
> + */
What is this? Is HARQ supported or not? Is this memory accessible.
> + uint32_t out_len;
> + /** Reserved (for 8 byte alignment) */
> + uint32_t rsvd;
> +};
> +
> /* This shared memory would be on the host side which have copy of some
> * of the parameters which are also part of Shared BD ring. Read access
> * of these parameters from the host side would not be over PCI.
> @@ -17,4 +183,18 @@ typedef struct host_ipc_params {
> volatile uint32_t modem_ptr[IPC_MAX_DEPTH]; } __rte_packed
> host_ipc_params_t;
>
> +struct hif_ipc_regs {
> + uint32_t ipc_mdata_offset;
> + uint32_t ipc_mdata_size;
> +} __rte_packed;
> +
> +struct gul_hif {
> + uint32_t ver;
> + uint32_t hif_ver;
> + uint32_t status;
> + volatile uint32_t host_ready;
> + volatile uint32_t mod_ready;
> + struct hif_ipc_regs ipc_regs;
> +} __rte_packed;
> +
> #endif
> --
> 2.17.1
