From e7cde313e1df14ea118386d678c3e47219090a0d Mon Sep 17 00:00:00 2001 From: Mohsin Kazmi Date: Tue, 26 Jun 2018 17:20:11 +0200 Subject: vhost: Add multi-versioning support It also refactors the vhost code which was in one big file vhost-user.c. Receive side code is in vhost_user_input.c and Transmit side code is in vhost_user_output.c Change-Id: I1b539b5008685889723e228265786a2a3e9f3a78 Signed-off-by: Mohsin Kazmi --- src/vnet/devices/virtio/vhost_user_input.c | 678 +++++++++++++++++++++++++++++ 1 file changed, 678 insertions(+) create mode 100644 src/vnet/devices/virtio/vhost_user_input.c (limited to 'src/vnet/devices/virtio/vhost_user_input.c') diff --git a/src/vnet/devices/virtio/vhost_user_input.c b/src/vnet/devices/virtio/vhost_user_input.c new file mode 100644 index 00000000000..3db44731b5e --- /dev/null +++ b/src/vnet/devices/virtio/vhost_user_input.c @@ -0,0 +1,678 @@ +/* + *------------------------------------------------------------------ + * vhost-user-input + * + * Copyright (c) 2014-2018 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + *------------------------------------------------------------------ + */ + +#include /* for open */ +#include +#include +#include +#include +#include +#include /* for iovec */ +#include +#include + +#include +#include + +#include +#include + +#include + +#include +#include +#include + +#include +#include + +/* + * When an RX queue is down but active, received packets + * must be discarded. This value controls up to how many + * packets will be discarded during each round. + */ +#define VHOST_USER_DOWN_DISCARD_COUNT 256 + +/* + * When the number of available buffers gets under this threshold, + * RX node will start discarding packets. + */ +#define VHOST_USER_RX_BUFFER_STARVATION 32 + +/* + * On the receive side, the host should free descriptors as soon + * as possible in order to avoid TX drop in the VM. + * This value controls the number of copy operations that are stacked + * before copy is done for all and descriptors are given back to + * the guest. + * The value 64 was obtained by testing (48 and 128 were not as good). + */ +#define VHOST_USER_RX_COPY_THRESHOLD 64 + +vlib_node_registration_t vhost_user_input_node; + +#define foreach_vhost_user_input_func_error \ + _(NO_ERROR, "no error") \ + _(NO_BUFFER, "no available buffer") \ + _(MMAP_FAIL, "mmap failure") \ + _(INDIRECT_OVERFLOW, "indirect descriptor overflows table") \ + _(UNDERSIZED_FRAME, "undersized ethernet frame received (< 14 bytes)") \ + _(FULL_RX_QUEUE, "full rx queue (possible driver tx drop)") + +typedef enum +{ +#define _(f,s) VHOST_USER_INPUT_FUNC_ERROR_##f, + foreach_vhost_user_input_func_error +#undef _ + VHOST_USER_INPUT_FUNC_N_ERROR, +} vhost_user_input_func_error_t; + +static __clib_unused char *vhost_user_input_func_error_strings[] = { +#define _(n,s) s, + foreach_vhost_user_input_func_error +#undef _ +}; + +static_always_inline void +vhost_user_rx_trace (vhost_trace_t * t, + vhost_user_intf_t * vui, u16 qid, + vlib_buffer_t * b, vhost_user_vring_t * txvq) +{ + vhost_user_main_t *vum = &vhost_user_main; + u32 last_avail_idx = txvq->last_avail_idx; + u32 desc_current = txvq->avail->ring[last_avail_idx & txvq->qsz_mask]; + vring_desc_t *hdr_desc = 0; + virtio_net_hdr_mrg_rxbuf_t *hdr; + u32 hint = 0; + + memset (t, 0, sizeof (*t)); + t->device_index = vui - vum->vhost_user_interfaces; + t->qid = qid; + + hdr_desc = &txvq->desc[desc_current]; + if (txvq->desc[desc_current].flags & VIRTQ_DESC_F_INDIRECT) + { + t->virtio_ring_flags |= 1 << VIRTIO_TRACE_F_INDIRECT; + /* Header is the first here */ + hdr_desc = map_guest_mem (vui, txvq->desc[desc_current].addr, &hint); + } + if (txvq->desc[desc_current].flags & VIRTQ_DESC_F_NEXT) + { + t->virtio_ring_flags |= 1 << VIRTIO_TRACE_F_SIMPLE_CHAINED; + } + if (!(txvq->desc[desc_current].flags & VIRTQ_DESC_F_NEXT) && + !(txvq->desc[desc_current].flags & VIRTQ_DESC_F_INDIRECT)) + { + t->virtio_ring_flags |= 1 << VIRTIO_TRACE_F_SINGLE_DESC; + } + + t->first_desc_len = hdr_desc ? hdr_desc->len : 0; + + if (!hdr_desc || !(hdr = map_guest_mem (vui, hdr_desc->addr, &hint))) + { + t->virtio_ring_flags |= 1 << VIRTIO_TRACE_F_MAP_ERROR; + } + else + { + u32 len = vui->virtio_net_hdr_sz; + memcpy (&t->hdr, hdr, len > hdr_desc->len ? hdr_desc->len : len); + } +} + +static_always_inline u32 +vhost_user_input_copy (vhost_user_intf_t * vui, vhost_copy_t * cpy, + u16 copy_len, u32 * map_hint) +{ + void *src0, *src1, *src2, *src3; + if (PREDICT_TRUE (copy_len >= 4)) + { + if (PREDICT_FALSE (!(src2 = map_guest_mem (vui, cpy[0].src, map_hint)))) + return 1; + if (PREDICT_FALSE (!(src3 = map_guest_mem (vui, cpy[1].src, map_hint)))) + return 1; + + while (PREDICT_TRUE (copy_len >= 4)) + { + src0 = src2; + src1 = src3; + + if (PREDICT_FALSE + (!(src2 = map_guest_mem (vui, cpy[2].src, map_hint)))) + return 1; + if (PREDICT_FALSE + (!(src3 = map_guest_mem (vui, cpy[3].src, map_hint)))) + return 1; + + CLIB_PREFETCH (src2, 64, LOAD); + CLIB_PREFETCH (src3, 64, LOAD); + + clib_memcpy ((void *) cpy[0].dst, src0, cpy[0].len); + clib_memcpy ((void *) cpy[1].dst, src1, cpy[1].len); + copy_len -= 2; + cpy += 2; + } + } + while (copy_len) + { + if (PREDICT_FALSE (!(src0 = map_guest_mem (vui, cpy->src, map_hint)))) + return 1; + clib_memcpy ((void *) cpy->dst, src0, cpy->len); + copy_len -= 1; + cpy += 1; + } + return 0; +} + +/** + * Try to discard packets from the tx ring (VPP RX path). + * Returns the number of discarded packets. + */ +static_always_inline u32 +vhost_user_rx_discard_packet (vlib_main_t * vm, + vhost_user_intf_t * vui, + vhost_user_vring_t * txvq, u32 discard_max) +{ + /* + * On the RX side, each packet corresponds to one descriptor + * (it is the same whether it is a shallow descriptor, chained, or indirect). + * Therefore, discarding a packet is like discarding a descriptor. + */ + u32 discarded_packets = 0; + u32 avail_idx = txvq->avail->idx; + while (discarded_packets != discard_max) + { + if (avail_idx == txvq->last_avail_idx) + goto out; + + u16 desc_chain_head = + txvq->avail->ring[txvq->last_avail_idx & txvq->qsz_mask]; + txvq->last_avail_idx++; + txvq->used->ring[txvq->last_used_idx & txvq->qsz_mask].id = + desc_chain_head; + txvq->used->ring[txvq->last_used_idx & txvq->qsz_mask].len = 0; + vhost_user_log_dirty_ring (vui, txvq, + ring[txvq->last_used_idx & txvq->qsz_mask]); + txvq->last_used_idx++; + discarded_packets++; + } + +out: + CLIB_MEMORY_BARRIER (); + txvq->used->idx = txvq->last_used_idx; + vhost_user_log_dirty_ring (vui, txvq, idx); + return discarded_packets; +} + +/* + * In case of overflow, we need to rewind the array of allocated buffers. + */ +static __clib_unused void +vhost_user_input_rewind_buffers (vlib_main_t * vm, + vhost_cpu_t * cpu, vlib_buffer_t * b_head) +{ + u32 bi_current = cpu->rx_buffers[cpu->rx_buffers_len]; + vlib_buffer_t *b_current = vlib_get_buffer (vm, bi_current); + b_current->current_length = 0; + b_current->flags = 0; + while (b_current != b_head) + { + cpu->rx_buffers_len++; + bi_current = cpu->rx_buffers[cpu->rx_buffers_len]; + b_current = vlib_get_buffer (vm, bi_current); + b_current->current_length = 0; + b_current->flags = 0; + } + cpu->rx_buffers_len++; +} + +static __clib_unused u32 +vhost_user_if_input (vlib_main_t * vm, + vhost_user_main_t * vum, + vhost_user_intf_t * vui, + u16 qid, vlib_node_runtime_t * node, + vnet_hw_interface_rx_mode mode) +{ + vhost_user_vring_t *txvq = &vui->vrings[VHOST_VRING_IDX_TX (qid)]; + u16 n_rx_packets = 0; + u32 n_rx_bytes = 0; + u16 n_left; + u32 n_left_to_next, *to_next; + u32 next_index = VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT; + u32 n_trace = vlib_get_trace_count (vm, node); + u32 map_hint = 0; + u16 thread_index = vlib_get_thread_index (); + u16 copy_len = 0; + + { + /* do we have pending interrupts ? */ + vhost_user_vring_t *rxvq = &vui->vrings[VHOST_VRING_IDX_RX (qid)]; + f64 now = vlib_time_now (vm); + + if ((txvq->n_since_last_int) && (txvq->int_deadline < now)) + vhost_user_send_call (vm, txvq); + + if ((rxvq->n_since_last_int) && (rxvq->int_deadline < now)) + vhost_user_send_call (vm, rxvq); + } + + /* + * For adaptive mode, it is optimized to reduce interrupts. + * If the scheduler switches the input node to polling due + * to burst of traffic, we tell the driver no interrupt. + * When the traffic subsides, the scheduler switches the node back to + * interrupt mode. We must tell the driver we want interrupt. + */ + if (PREDICT_FALSE (mode == VNET_HW_INTERFACE_RX_MODE_ADAPTIVE)) + { + if ((node->flags & + VLIB_NODE_FLAG_SWITCH_FROM_POLLING_TO_INTERRUPT_MODE) || + !(node->flags & + VLIB_NODE_FLAG_SWITCH_FROM_INTERRUPT_TO_POLLING_MODE)) + /* Tell driver we want notification */ + txvq->used->flags = 0; + else + /* Tell driver we don't want notification */ + txvq->used->flags = VRING_USED_F_NO_NOTIFY; + } + + if (PREDICT_FALSE (txvq->avail->flags & 0xFFFE)) + return 0; + + n_left = (u16) (txvq->avail->idx - txvq->last_avail_idx); + + /* nothing to do */ + if (PREDICT_FALSE (n_left == 0)) + return 0; + + if (PREDICT_FALSE (!vui->admin_up || !(txvq->enabled))) + { + /* + * Discard input packet if interface is admin down or vring is not + * enabled. + * "For example, for a networking device, in the disabled state + * client must not supply any new RX packets, but must process + * and discard any TX packets." + */ + vhost_user_rx_discard_packet (vm, vui, txvq, + VHOST_USER_DOWN_DISCARD_COUNT); + return 0; + } + + if (PREDICT_FALSE (n_left == (txvq->qsz_mask + 1))) + { + /* + * Informational error logging when VPP is not + * receiving packets fast enough. + */ + vlib_error_count (vm, node->node_index, + VHOST_USER_INPUT_FUNC_ERROR_FULL_RX_QUEUE, 1); + } + + if (n_left > VLIB_FRAME_SIZE) + n_left = VLIB_FRAME_SIZE; + + /* + * For small packets (<2kB), we will not need more than one vlib buffer + * per packet. In case packets are bigger, we will just yeld at some point + * in the loop and come back later. This is not an issue as for big packet, + * processing cost really comes from the memory copy. + * The assumption is that big packets will fit in 40 buffers. + */ + if (PREDICT_FALSE (vum->cpus[thread_index].rx_buffers_len < n_left + 1 || + vum->cpus[thread_index].rx_buffers_len < 40)) + { + 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[thread_index].rx_buffers + + curr_len, + VHOST_USER_RX_BUFFERS_N - curr_len, + VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); + + if (PREDICT_FALSE + (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[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, + vlib_get_thread_index (), + vui->sw_if_index, flush); + + vlib_error_count (vm, vhost_user_input_node.index, + VHOST_USER_INPUT_FUNC_ERROR_NO_BUFFER, flush); + } + } + + while (n_left > 0) + { + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left > 0 && n_left_to_next > 0) + { + vlib_buffer_t *b_head, *b_current; + u32 bi_current; + u16 desc_current; + u32 desc_data_offset; + vring_desc_t *desc_table = txvq->desc; + + 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 + * check for the next buffer prefetch. + */ + n_left = 0; + break; + } + + desc_current = + txvq->avail->ring[txvq->last_avail_idx & txvq->qsz_mask]; + 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[thread_index].rx_buffers) + [vum->cpus[thread_index]. + rx_buffers_len - 1], LOAD); + + /* Just preset the used descriptor id and length for later */ + txvq->used->ring[txvq->last_used_idx & txvq->qsz_mask].id = + desc_current; + txvq->used->ring[txvq->last_used_idx & txvq->qsz_mask].len = 0; + vhost_user_log_dirty_ring (vui, txvq, + ring[txvq->last_used_idx & + txvq->qsz_mask]); + + /* The buffer should already be initialized */ + b_head->total_length_not_including_first_buffer = 0; + b_head->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID; + + if (PREDICT_FALSE (n_trace)) + { + //TODO: next_index is not exactly known at that point + vlib_trace_buffer (vm, node, next_index, b_head, + /* follow_chain */ 0); + vhost_trace_t *t0 = + vlib_add_trace (vm, node, b_head, sizeof (t0[0])); + vhost_user_rx_trace (t0, vui, qid, b_head, txvq); + n_trace--; + vlib_set_trace_count (vm, node, n_trace); + } + + /* This depends on the setup but is very consistent + * So I think the CPU branch predictor will make a pretty good job + * at optimizing the decision. */ + if (txvq->desc[desc_current].flags & VIRTQ_DESC_F_INDIRECT) + { + desc_table = map_guest_mem (vui, txvq->desc[desc_current].addr, + &map_hint); + desc_current = 0; + if (PREDICT_FALSE (desc_table == 0)) + { + vlib_error_count (vm, node->node_index, + VHOST_USER_INPUT_FUNC_ERROR_MMAP_FAIL, 1); + goto out; + } + } + + if (PREDICT_TRUE (vui->is_any_layout) || + (!(desc_table[desc_current].flags & VIRTQ_DESC_F_NEXT))) + { + /* ANYLAYOUT or single buffer */ + desc_data_offset = vui->virtio_net_hdr_sz; + } + else + { + /* CSR case without ANYLAYOUT, skip 1st buffer */ + desc_data_offset = desc_table[desc_current].len; + } + + while (1) + { + /* Get more input if necessary. Or end of packet. */ + if (desc_data_offset == desc_table[desc_current].len) + { + if (PREDICT_FALSE (desc_table[desc_current].flags & + VIRTQ_DESC_F_NEXT)) + { + desc_current = desc_table[desc_current].next; + desc_data_offset = 0; + } + else + { + goto out; + } + } + + /* Get more output if necessary. Or end of packet. */ + if (PREDICT_FALSE + (b_current->current_length == VLIB_BUFFER_DATA_SIZE)) + { + if (PREDICT_FALSE + (vum->cpus[thread_index].rx_buffers_len == 0)) + { + /* Cancel speculation */ + to_next--; + n_left_to_next++; + + /* + * Checking if there are some left buffers. + * If not, just rewind the used buffers and stop. + * Note: Scheduled copies are not cancelled. This is + * not an issue as they would still be valid. Useless, + * but valid. + */ + vhost_user_input_rewind_buffers (vm, + &vum->cpus + [thread_index], + b_head); + n_left = 0; + goto stop; + } + + /* Get next output */ + vum->cpus[thread_index].rx_buffers_len--; + u32 bi_next = + (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; + b_current = vlib_get_buffer (vm, bi_current); + } + + /* Prepare a copy order executed later for the data */ + 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; + cpy->len = VLIB_BUFFER_DATA_SIZE - b_current->current_length; + cpy->len = (cpy->len > desc_data_l) ? desc_data_l : cpy->len; + cpy->dst = (uword) (vlib_buffer_get_current (b_current) + + b_current->current_length); + cpy->src = desc_table[desc_current].addr + desc_data_offset; + + desc_data_offset += cpy->len; + + b_current->current_length += cpy->len; + b_head->total_length_not_including_first_buffer += cpy->len; + } + + out: + CLIB_PREFETCH (&n_left, sizeof (n_left), LOAD); + + n_rx_bytes += b_head->total_length_not_including_first_buffer; + n_rx_packets++; + + b_head->total_length_not_including_first_buffer -= + b_head->current_length; + + /* consume the descriptor and return it as used */ + txvq->last_avail_idx++; + txvq->last_used_idx++; + + VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b_head); + + vnet_buffer (b_head)->sw_if_index[VLIB_RX] = vui->sw_if_index; + vnet_buffer (b_head)->sw_if_index[VLIB_TX] = (u32) ~ 0; + b_head->error = 0; + + { + u32 next0 = VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT; + + /* redirect if feature path enabled */ + vnet_feature_start_device_input_x1 (vui->sw_if_index, &next0, + b_head); + + u32 bi = to_next[-1]; //Cannot use to_next[-1] in the macro + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi, next0); + } + + n_left--; + + /* + * Although separating memory copies from virtio ring parsing + * is beneficial, we can offer to perform the copies from time + * to time in order to free some space in the ring. + */ + if (PREDICT_FALSE (copy_len >= VHOST_USER_RX_COPY_THRESHOLD)) + { + if (PREDICT_FALSE + (vhost_user_input_copy (vui, vum->cpus[thread_index].copy, + copy_len, &map_hint))) + { + vlib_error_count (vm, node->node_index, + VHOST_USER_INPUT_FUNC_ERROR_MMAP_FAIL, 1); + } + copy_len = 0; + + /* give buffers back to driver */ + CLIB_MEMORY_BARRIER (); + txvq->used->idx = txvq->last_used_idx; + vhost_user_log_dirty_ring (vui, txvq, idx); + } + } + stop: + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + /* Do the memory copies */ + if (PREDICT_FALSE + (vhost_user_input_copy (vui, vum->cpus[thread_index].copy, + copy_len, &map_hint))) + { + vlib_error_count (vm, node->node_index, + VHOST_USER_INPUT_FUNC_ERROR_MMAP_FAIL, 1); + } + + /* give buffers back to driver */ + CLIB_MEMORY_BARRIER (); + txvq->used->idx = txvq->last_used_idx; + vhost_user_log_dirty_ring (vui, txvq, idx); + + /* interrupt (call) handling */ + if ((txvq->callfd_idx != ~0) && + !(txvq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT)) + { + txvq->n_since_last_int += n_rx_packets; + + if (txvq->n_since_last_int > vum->coalesce_frames) + vhost_user_send_call (vm, txvq); + } + + /* increase rx counters */ + vlib_increment_combined_counter + (vnet_main.interface_main.combined_sw_if_counters + + VNET_INTERFACE_COUNTER_RX, + vlib_get_thread_index (), vui->sw_if_index, n_rx_packets, n_rx_bytes); + + vnet_device_increment_rx_packets (thread_index, n_rx_packets); + + return n_rx_packets; +} + +VLIB_NODE_FN (vhost_user_input_node) (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + vhost_user_main_t *vum = &vhost_user_main; + uword n_rx_packets = 0; + vhost_user_intf_t *vui; + vnet_device_input_runtime_t *rt = + (vnet_device_input_runtime_t *) node->runtime_data; + vnet_device_and_queue_t *dq; + + vec_foreach (dq, rt->devices_and_queues) + { + if (clib_smp_swap (&dq->interrupt_pending, 0) || + (node->state == VLIB_NODE_STATE_POLLING)) + { + vui = + pool_elt_at_index (vum->vhost_user_interfaces, dq->dev_instance); + n_rx_packets = vhost_user_if_input (vm, vum, vui, dq->queue_id, node, + dq->mode); + } + } + + return n_rx_packets; +} + +#ifndef CLIB_MARCH_VARIANT +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (vhost_user_input_node) = { + .type = VLIB_NODE_TYPE_INPUT, + .name = "vhost-user-input", + .sibling_of = "device-input", + + /* Will be enabled if/when hardware is detected. */ + .state = VLIB_NODE_STATE_DISABLED, + + .format_buffer = format_ethernet_header_with_length, + .format_trace = format_vhost_trace, + + .n_errors = VHOST_USER_INPUT_FUNC_N_ERROR, + .error_strings = vhost_user_input_func_error_strings, +}; +/* *INDENT-ON* */ +#endif + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ -- cgit 1.2.3-korg