diff options
author | Damjan Marion <damarion@cisco.com> | 2017-04-05 19:18:20 +0200 |
---|---|---|
committer | Dave Barach <openvpp@barachs.net> | 2017-04-06 11:31:39 +0000 |
commit | 586afd762bfa149f5ca167bd5fd5a0cd59ce94fe (patch) | |
tree | 808b57c61e0fe1a181871bb1ad94398c5ba42671 /src/vnet/devices/virtio | |
parent | bc799c92d761a2d45105aa6a1685b3663687d2a4 (diff) |
Use thread local storage for thread index
This patch deprecates stack-based thread identification,
Also removes requirement that thread stacks are adjacent.
Finally, possibly annoying for some folks, it renames
all occurences of cpu_index and cpu_number with thread
index. Using word "cpu" is misleading here as thread can
be migrated ti different CPU, and also it is not related
to linux cpu index.
Change-Id: I68cdaf661e701d2336fc953dcb9978d10a70f7c1
Signed-off-by: Damjan Marion <damarion@cisco.com>
Diffstat (limited to 'src/vnet/devices/virtio')
-rw-r--r-- | src/vnet/devices/virtio/vhost-user.c | 127 |
1 files changed, 65 insertions, 62 deletions
diff --git a/src/vnet/devices/virtio/vhost-user.c b/src/vnet/devices/virtio/vhost-user.c index 00807dc0398..5e720f6510a 100644 --- a/src/vnet/devices/virtio/vhost-user.c +++ b/src/vnet/devices/virtio/vhost-user.c @@ -331,7 +331,7 @@ vhost_user_tx_thread_placement (vhost_user_intf_t * vui) { //Let's try to assign one queue to each thread u32 qid = 0; - u32 cpu_index = 0; + u32 thread_index = 0; vui->use_tx_spinlock = 0; while (1) { @@ -341,20 +341,21 @@ vhost_user_tx_thread_placement (vhost_user_intf_t * vui) if (!rxvq->started || !rxvq->enabled) continue; - vui->per_cpu_tx_qid[cpu_index] = qid; - cpu_index++; - if (cpu_index == vlib_get_thread_main ()->n_vlib_mains) + vui->per_cpu_tx_qid[thread_index] = qid; + thread_index++; + if (thread_index == vlib_get_thread_main ()->n_vlib_mains) return; } //We need to loop, meaning the spinlock has to be used vui->use_tx_spinlock = 1; - if (cpu_index == 0) + if (thread_index == 0) { //Could not find a single valid one - for (cpu_index = 0; - cpu_index < vlib_get_thread_main ()->n_vlib_mains; cpu_index++) + for (thread_index = 0; + thread_index < vlib_get_thread_main ()->n_vlib_mains; + thread_index++) { - vui->per_cpu_tx_qid[cpu_index] = 0; + vui->per_cpu_tx_qid[thread_index] = 0; } return; } @@ -368,7 +369,7 @@ vhost_user_rx_thread_placement () vhost_user_intf_t *vui; vhost_cpu_t *vhc; u32 *workers = 0; - u32 cpu_index; + u32 thread_index; vlib_main_t *vm; //Let's list all workers cpu indexes @@ -400,9 +401,9 @@ vhost_user_rx_thread_placement () continue; i %= vec_len (vui_workers); - cpu_index = vui_workers[i]; + thread_index = vui_workers[i]; i++; - vhc = &vum->cpus[cpu_index]; + vhc = &vum->cpus[thread_index]; iaq.qid = qid; iaq.vhost_iface_index = vui - vum->vhost_user_interfaces; @@ -429,14 +430,14 @@ vhost_user_rx_thread_placement () vhc->operation_mode = mode; } - for (cpu_index = vum->input_cpu_first_index; - cpu_index < vum->input_cpu_first_index + vum->input_cpu_count; - cpu_index++) + for (thread_index = vum->input_cpu_first_index; + thread_index < vum->input_cpu_first_index + vum->input_cpu_count; + thread_index++) { vlib_node_state_t state = VLIB_NODE_STATE_POLLING; - vhc = &vum->cpus[cpu_index]; - vm = vlib_mains ? vlib_mains[cpu_index] : &vlib_global_main; + vhc = &vum->cpus[thread_index]; + vm = vlib_mains ? vlib_mains[thread_index] : &vlib_global_main; switch (vhc->operation_mode) { case VHOST_USER_INTERRUPT_MODE: @@ -532,7 +533,7 @@ vhost_user_set_interrupt_pending (vhost_user_intf_t * vui, u32 ifq) { vhost_user_main_t *vum = &vhost_user_main; vhost_cpu_t *vhc; - u32 cpu_index; + u32 thread_index; vhost_iface_and_queue_t *vhiq; vlib_main_t *vm; u32 ifq2; @@ -553,8 +554,8 @@ vhost_user_set_interrupt_pending (vhost_user_intf_t * vui, u32 ifq) if ((vhiq->vhost_iface_index == (ifq >> 8)) && (VHOST_VRING_IDX_TX (vhiq->qid) == (ifq & 0xff))) { - cpu_index = vhc - vum->cpus; - vm = vlib_mains ? vlib_mains[cpu_index] : &vlib_global_main; + thread_index = vhc - vum->cpus; + vm = vlib_mains ? vlib_mains[thread_index] : &vlib_global_main; /* * Convert RX virtqueue number in the lower byte to vring * queue index for the input node process. Top bytes contain @@ -1592,7 +1593,7 @@ vhost_user_if_input (vlib_main_t * vm, u32 n_trace = vlib_get_trace_count (vm, node); u16 qsz_mask; u32 map_hint = 0; - u16 cpu_index = os_get_cpu_number (); + u16 thread_index = vlib_get_thread_index (); u16 copy_len = 0; { @@ -1651,32 +1652,32 @@ vhost_user_if_input (vlib_main_t * vm, * in the loop and come back later. This is not an issue as for big packet, * processing cost really comes from the memory copy. */ - if (PREDICT_FALSE (vum->cpus[cpu_index].rx_buffers_len < n_left + 1)) + if (PREDICT_FALSE (vum->cpus[thread_index].rx_buffers_len < n_left + 1)) { - u32 curr_len = vum->cpus[cpu_index].rx_buffers_len; - vum->cpus[cpu_index].rx_buffers_len += + u32 curr_len = vum->cpus[thread_index].rx_buffers_len; + vum->cpus[thread_index].rx_buffers_len += vlib_buffer_alloc_from_free_list (vm, - vum->cpus[cpu_index].rx_buffers + + vum->cpus[thread_index].rx_buffers + curr_len, VHOST_USER_RX_BUFFERS_N - curr_len, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); if (PREDICT_FALSE - (vum->cpus[cpu_index].rx_buffers_len < + (vum->cpus[thread_index].rx_buffers_len < VHOST_USER_RX_BUFFER_STARVATION)) { /* In case of buffer starvation, discard some packets from the queue * and log the event. * We keep doing best effort for the remaining packets. */ - u32 flush = (n_left + 1 > vum->cpus[cpu_index].rx_buffers_len) ? - n_left + 1 - vum->cpus[cpu_index].rx_buffers_len : 1; + u32 flush = (n_left + 1 > vum->cpus[thread_index].rx_buffers_len) ? + n_left + 1 - vum->cpus[thread_index].rx_buffers_len : 1; flush = vhost_user_rx_discard_packet (vm, vui, txvq, flush); n_left -= flush; vlib_increment_simple_counter (vnet_main. interface_main.sw_if_counters + VNET_INTERFACE_COUNTER_DROP, - os_get_cpu_number (), + vlib_get_thread_index (), vui->sw_if_index, flush); vlib_error_count (vm, vhost_user_input_node.index, @@ -1696,7 +1697,7 @@ vhost_user_if_input (vlib_main_t * vm, u32 desc_data_offset; vring_desc_t *desc_table = txvq->desc; - if (PREDICT_FALSE (vum->cpus[cpu_index].rx_buffers_len <= 1)) + if (PREDICT_FALSE (vum->cpus[thread_index].rx_buffers_len <= 1)) { /* Not enough rx_buffers * Note: We yeld on 1 so we don't need to do an additional @@ -1707,17 +1708,18 @@ vhost_user_if_input (vlib_main_t * vm, } desc_current = txvq->avail->ring[txvq->last_avail_idx & qsz_mask]; - vum->cpus[cpu_index].rx_buffers_len--; - bi_current = (vum->cpus[cpu_index].rx_buffers) - [vum->cpus[cpu_index].rx_buffers_len]; + vum->cpus[thread_index].rx_buffers_len--; + bi_current = (vum->cpus[thread_index].rx_buffers) + [vum->cpus[thread_index].rx_buffers_len]; b_head = b_current = vlib_get_buffer (vm, bi_current); to_next[0] = bi_current; //We do that now so we can forget about bi_current to_next++; n_left_to_next--; vlib_prefetch_buffer_with_index (vm, - (vum->cpus[cpu_index].rx_buffers) - [vum->cpus[cpu_index]. + (vum-> + cpus[thread_index].rx_buffers) + [vum->cpus[thread_index]. rx_buffers_len - 1], LOAD); /* Just preset the used descriptor id and length for later */ @@ -1791,7 +1793,7 @@ vhost_user_if_input (vlib_main_t * vm, (b_current->current_length == VLIB_BUFFER_DATA_SIZE)) { if (PREDICT_FALSE - (vum->cpus[cpu_index].rx_buffers_len == 0)) + (vum->cpus[thread_index].rx_buffers_len == 0)) { /* Cancel speculation */ to_next--; @@ -1805,17 +1807,18 @@ vhost_user_if_input (vlib_main_t * vm, * but valid. */ vhost_user_input_rewind_buffers (vm, - &vum->cpus[cpu_index], + &vum->cpus + [thread_index], b_head); n_left = 0; goto stop; } /* Get next output */ - vum->cpus[cpu_index].rx_buffers_len--; + vum->cpus[thread_index].rx_buffers_len--; u32 bi_next = - (vum->cpus[cpu_index].rx_buffers)[vum->cpus - [cpu_index].rx_buffers_len]; + (vum->cpus[thread_index].rx_buffers)[vum->cpus + [thread_index].rx_buffers_len]; b_current->next_buffer = bi_next; b_current->flags |= VLIB_BUFFER_NEXT_PRESENT; bi_current = bi_next; @@ -1823,7 +1826,7 @@ vhost_user_if_input (vlib_main_t * vm, } /* Prepare a copy order executed later for the data */ - vhost_copy_t *cpy = &vum->cpus[cpu_index].copy[copy_len]; + vhost_copy_t *cpy = &vum->cpus[thread_index].copy[copy_len]; copy_len++; u32 desc_data_l = desc_table[desc_current].len - desc_data_offset; @@ -1880,7 +1883,7 @@ vhost_user_if_input (vlib_main_t * vm, if (PREDICT_FALSE (copy_len >= VHOST_USER_RX_COPY_THRESHOLD)) { if (PREDICT_FALSE - (vhost_user_input_copy (vui, vum->cpus[cpu_index].copy, + (vhost_user_input_copy (vui, vum->cpus[thread_index].copy, copy_len, &map_hint))) { clib_warning @@ -1905,7 +1908,7 @@ vhost_user_if_input (vlib_main_t * vm, /* Do the memory copies */ if (PREDICT_FALSE - (vhost_user_input_copy (vui, vum->cpus[cpu_index].copy, + (vhost_user_input_copy (vui, vum->cpus[thread_index].copy, copy_len, &map_hint))) { clib_warning ("Memory mapping error on interface hw_if_index=%d " @@ -1933,9 +1936,9 @@ vhost_user_if_input (vlib_main_t * vm, vlib_increment_combined_counter (vnet_main.interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), vui->sw_if_index, n_rx_packets, n_rx_bytes); + vlib_get_thread_index (), vui->sw_if_index, n_rx_packets, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, n_rx_packets); + vnet_device_increment_rx_packets (thread_index, n_rx_packets); return n_rx_packets; } @@ -1946,15 +1949,15 @@ vhost_user_input (vlib_main_t * vm, { vhost_user_main_t *vum = &vhost_user_main; uword n_rx_packets = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); vhost_iface_and_queue_t *vhiq; vhost_user_intf_t *vui; vhost_cpu_t *vhc; - vhc = &vum->cpus[cpu_index]; + vhc = &vum->cpus[thread_index]; if (PREDICT_TRUE (vhc->operation_mode == VHOST_USER_POLLING_MODE)) { - vec_foreach (vhiq, vum->cpus[cpu_index].rx_queues) + vec_foreach (vhiq, vum->cpus[thread_index].rx_queues) { vui = &vum->vhost_user_interfaces[vhiq->vhost_iface_index]; n_rx_packets += vhost_user_if_input (vm, vum, vui, vhiq->qid, node); @@ -2096,7 +2099,7 @@ vhost_user_tx (vlib_main_t * vm, vhost_user_vring_t *rxvq; u16 qsz_mask; u8 error; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 map_hint = 0; u8 retry = 8; u16 copy_len; @@ -2116,7 +2119,7 @@ vhost_user_tx (vlib_main_t * vm, qid = VHOST_VRING_IDX_RX (*vec_elt_at_index - (vui->per_cpu_tx_qid, os_get_cpu_number ())); + (vui->per_cpu_tx_qid, vlib_get_thread_index ())); rxvq = &vui->vrings[qid]; if (PREDICT_FALSE (vui->use_tx_spinlock)) vhost_user_vring_lock (vui, qid); @@ -2143,10 +2146,10 @@ retry: if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { - vum->cpus[cpu_index].current_trace = + vum->cpus[thread_index].current_trace = vlib_add_trace (vm, node, b0, - sizeof (*vum->cpus[cpu_index].current_trace)); - vhost_user_tx_trace (vum->cpus[cpu_index].current_trace, + sizeof (*vum->cpus[thread_index].current_trace)); + vhost_user_tx_trace (vum->cpus[thread_index].current_trace, vui, qid / 2, b0, rxvq); } @@ -2188,14 +2191,14 @@ retry: { // Get a header from the header array virtio_net_hdr_mrg_rxbuf_t *hdr = - &vum->cpus[cpu_index].tx_headers[tx_headers_len]; + &vum->cpus[thread_index].tx_headers[tx_headers_len]; tx_headers_len++; hdr->hdr.flags = 0; hdr->hdr.gso_type = 0; hdr->num_buffers = 1; //This is local, no need to check // Prepare a copy order executed later for the header - vhost_copy_t *cpy = &vum->cpus[cpu_index].copy[copy_len]; + vhost_copy_t *cpy = &vum->cpus[thread_index].copy[copy_len]; copy_len++; cpy->len = vui->virtio_net_hdr_sz; cpy->dst = buffer_map_addr; @@ -2220,7 +2223,7 @@ retry: else if (vui->virtio_net_hdr_sz == 12) //MRG is available { virtio_net_hdr_mrg_rxbuf_t *hdr = - &vum->cpus[cpu_index].tx_headers[tx_headers_len - 1]; + &vum->cpus[thread_index].tx_headers[tx_headers_len - 1]; //Move from available to used buffer rxvq->used->ring[rxvq->last_used_idx & qsz_mask].id = @@ -2282,7 +2285,7 @@ retry: } { - vhost_copy_t *cpy = &vum->cpus[cpu_index].copy[copy_len]; + vhost_copy_t *cpy = &vum->cpus[thread_index].copy[copy_len]; copy_len++; cpy->len = bytes_left; cpy->len = (cpy->len > buffer_len) ? buffer_len : cpy->len; @@ -2325,8 +2328,8 @@ retry: if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { - vum->cpus[cpu_index].current_trace->hdr = - vum->cpus[cpu_index].tx_headers[tx_headers_len - 1]; + vum->cpus[thread_index].current_trace->hdr = + vum->cpus[thread_index].tx_headers[tx_headers_len - 1]; } n_left--; //At the end for error counting when 'goto done' is invoked @@ -2336,7 +2339,7 @@ retry: done: //Do the memory copies if (PREDICT_FALSE - (vhost_user_tx_copy (vui, vum->cpus[cpu_index].copy, + (vhost_user_tx_copy (vui, vum->cpus[thread_index].copy, copy_len, &map_hint))) { clib_warning ("Memory mapping error on interface hw_if_index=%d " @@ -2386,7 +2389,7 @@ done3: vlib_increment_simple_counter (vnet_main.interface_main.sw_if_counters + VNET_INTERFACE_COUNTER_DROP, - os_get_cpu_number (), vui->sw_if_index, n_left); + vlib_get_thread_index (), vui->sw_if_index, n_left); } vlib_buffer_free (vm, vlib_frame_args (frame), frame->n_vectors); @@ -2773,11 +2776,11 @@ vhost_user_send_interrupt_process (vlib_main_t * vm, case ~0: vec_foreach (vhc, vum->cpus) { - u32 cpu_index = vhc - vum->cpus; + u32 thread_index = vhc - vum->cpus; f64 next_timeout; next_timeout = timeout; - vec_foreach (vhiq, vum->cpus[cpu_index].rx_queues) + vec_foreach (vhiq, vum->cpus[thread_index].rx_queues) { vui = &vum->vhost_user_interfaces[vhiq->vhost_iface_index]; vhost_user_vring_t *rxvq = |