Adding mlx4_query_values as implementation for ibv_query_values_ex. mlx4_query_values follows the standard extension verb mechanism. This function supports reading the hwclock via mmaping the required space from kernel.
Signed-off-by: Matan Barak <[email protected]> --- src/mlx4.c | 36 ++++++++++++++++++++++++++++++++++++ src/mlx4.h | 3 +++ src/verbs.c | 45 +++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 84 insertions(+) diff --git a/src/mlx4.c b/src/mlx4.c index cc1211f..6d66cf0 100644 --- a/src/mlx4.c +++ b/src/mlx4.c @@ -116,6 +116,28 @@ static struct ibv_context_ops mlx4_ctx_ops = { .detach_mcast = ibv_cmd_detach_mcast }; +static int mlx4_map_internal_clock(struct mlx4_device *dev, + struct ibv_context *ibv_ctx) +{ + struct mlx4_context *context = to_mctx(ibv_ctx); + void *hca_clock_page; + + hca_clock_page = mmap(NULL, dev->page_size, PROT_READ, MAP_SHARED, + ibv_ctx->cmd_fd, dev->page_size * 3); + + if (hca_clock_page == MAP_FAILED) { + fprintf(stderr, PFX + "Warning: Timestamp available,\n" + "but failed to mmap() hca core clock page, errno=%d.\n", + errno); + return -1; + } + + context->hca_core_clock = hca_clock_page + + context->core_clock_offset % dev->page_size; + return 0; +} + static int mlx4_init_context(struct verbs_device *v_device, struct ibv_context *ibv_ctx, int cmd_fd) { @@ -127,6 +149,10 @@ static int mlx4_init_context(struct verbs_device *v_device, __u16 bf_reg_size; struct mlx4_device *dev = to_mdev(&v_device->device); struct verbs_context *verbs_ctx = verbs_get_ctx(ibv_ctx); + struct ibv_query_device_ex_input input_query_device = {.comp_mask = 0}; + struct ibv_device_attr_ex dev_attrs; + uint32_t dev_attrs_comp_mask; + int err; /* memory footprint of mlx4_context and verbs_context share * struct ibv_context. @@ -194,6 +220,12 @@ static int mlx4_init_context(struct verbs_device *v_device, context->bf_buf_size = 0; } + context->hca_core_clock = NULL; + err = _mlx4_query_device_ex(ibv_ctx, &input_query_device, &dev_attrs, + sizeof(dev_attrs), &dev_attrs_comp_mask); + if (!err && dev_attrs_comp_mask & QUERY_DEVICE_RESP_MASK_TIMESTAMP) + mlx4_map_internal_clock(dev, ibv_ctx); + pthread_spin_init(&context->uar_lock, PTHREAD_PROCESS_PRIVATE); ibv_ctx->ops = mlx4_ctx_ops; @@ -210,6 +242,7 @@ static int mlx4_init_context(struct verbs_device *v_device, verbs_set_ctx_op(verbs_ctx, query_device_ex, mlx4_query_device_ex); verbs_set_ctx_op(verbs_ctx, create_cq_ex, mlx4_create_cq_ex); verbs_set_ctx_op(verbs_ctx, poll_cq_ex, mlx4_poll_cq_ex); + verbs_set_ctx_op(verbs_ctx, query_values, mlx4_query_values); return 0; @@ -223,6 +256,9 @@ static void mlx4_uninit_context(struct verbs_device *v_device, munmap(context->uar, to_mdev(&v_device->device)->page_size); if (context->bf_page) munmap(context->bf_page, to_mdev(&v_device->device)->page_size); + if (context->hca_core_clock) + munmap(context->hca_core_clock - context->core_clock_offset, + to_mdev(&v_device->device)->page_size); } diff --git a/src/mlx4.h b/src/mlx4.h index 2465298..8e1935d 100644 --- a/src/mlx4.h +++ b/src/mlx4.h @@ -199,6 +199,7 @@ struct mlx4_context { enum ibv_port_cap_flags caps; } port_query_cache[MLX4_PORTS_NUM]; uint64_t core_clock_offset; + void *hca_core_clock; }; struct mlx4_buf { @@ -403,6 +404,8 @@ int _mlx4_query_device_ex(struct ibv_context *context, int mlx4_query_device_ex(struct ibv_context *context, const struct ibv_query_device_ex_input *input, struct ibv_device_attr_ex *attr, size_t attr_size); +int mlx4_query_values(struct ibv_context *context, + struct ibv_values_ex *values); int mlx4_query_port(struct ibv_context *context, uint8_t port, struct ibv_port_attr *attr); diff --git a/src/verbs.c b/src/verbs.c index a8d6bd7..843ca1e 100644 --- a/src/verbs.c +++ b/src/verbs.c @@ -114,6 +114,51 @@ int mlx4_query_device_ex(struct ibv_context *context, return _mlx4_query_device_ex(context, input, attr, attr_size, NULL); } +#define READL(ptr) (*((uint32_t *)(ptr))) +static int mlx4_read_clock(struct ibv_context *context, uint64_t *cycles) +{ + unsigned int clockhi, clocklo, clockhi1; + int i; + struct mlx4_context *ctx = to_mctx(context); + + if (!ctx->hca_core_clock) + return -EOPNOTSUPP; + + for (i = 0; i < 10; i++) { + clockhi = ntohl(READL(ctx->hca_core_clock)); + clocklo = ntohl(READL(ctx->hca_core_clock + 4)); + clockhi1 = ntohl(READL(ctx->hca_core_clock)); + if (clockhi == clockhi1) + break; + } + + *cycles = (uint64_t)clockhi << 32 | (uint64_t)clocklo; + + return 0; +} + +int mlx4_query_values(struct ibv_context *context, + struct ibv_values_ex *values) +{ + uint32_t comp_mask = 0; + int err = 0; + + if (values->comp_mask & IBV_VALUES_MASK_RAW_CLOCK) { + uint64_t cycles; + + err = mlx4_read_clock(context, &cycles); + if (!err) { + values->raw_clock.tv_sec = 0; + values->raw_clock.tv_nsec = cycles; + comp_mask |= IBV_VALUES_MASK_RAW_CLOCK; + } + } + + values->comp_mask = comp_mask; + + return err; +} + int mlx4_query_port(struct ibv_context *context, uint8_t port, struct ibv_port_attr *attr) { -- 2.1.0 -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
