/* * Copyright (c) 2015 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. */ /* * buffer_node.h: VLIB buffer handling node helper macros/inlines * * Copyright (c) 2008 Eliot Dresselhaus * * Permission is hereby granted, free of charge, to any person obtaining * a copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sublicense, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ #ifndef included_vlib_buffer_node_h #define included_vlib_buffer_node_h /** \file vlib buffer/node functions */ /** \brief Finish enqueueing two buffers forward in the graph. Standard dual loop boilerplate element. This is a MACRO, with MULTIPLE SIDE EFFECTS. In the ideal case, <code>next_index == next0 == next1</code>, which means that the speculative enqueue at the top of the dual loop has correctly dealt with both packets. In that case, the macro does nothing at all. @param vm vlib_main_t pointer, varies by thread @param node current node vlib_node_runtime_t pointer @param next_index speculated next index used for both packets @param to_next speculated vector pointer used for both packets @param n_left_to_next number of slots left in speculated vector @param bi0 first buffer index @param bi1 second buffer index @param next0 actual next index to be used for the first packet @param next1 actual next index to be used for the second packet @return @c next_index -- speculative next index to be used for future packets @return @c to_next -- speculative frame to be used for future packets @return @c n_left_to_next -- number of slots left in speculative frame */ #define vlib_validate_buffer_enqueue_x2(vm,node,next_index,to_next,n_left_to_next,bi0,bi1,next0,next1) \ do { \ int enqueue_code = (next0 != next_index) + 2*(next1 != next_index); \ \ if (PREDICT_FALSE (enqueue_code != 0)) \ { \ switch (enqueue_code) \ { \ case 1: \ /* A B A */ \ to_next[-2] = bi1; \ to_next -= 1; \ n_left_to_next += 1; \ vlib_set_next_frame_buffer (vm, node, next0, bi0); \ break; \ \ case 2: \ /* A A B */ \ to_next -= 1; \ n_left_to_next += 1; \ vlib_set_next_frame_buffer (vm, node, next1, bi1); \ break; \ \ case 3: \ /* A B B or A B C */ \ to_next -= 2; \ n_left_to_next += 2; \ vlib_set_next_frame_buffer (vm, node, next0, bi0); \ vlib_set_next_frame_buffer (vm, node, next1, bi1); \ if (next0 == next1) \ { \ vlib_put_next_frame (vm, node, next_index, \ n_left_to_next); \ next_index = next1; \ vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); \ } \ } \ } \ } while (0) /** \brief Finish enqueueing four buffers forward in the graph. Standard quad loop boilerplate element. This is a MACRO, with MULTIPLE SIDE EFFECTS. In the ideal case, <code>next_index == next0 == next1 == next2 == next3</code>, which means that the speculative enqueue at the top of the quad loop has correctly dealt with all four packets. In that case, the macro does nothing at all. @param vm vlib_main_t pointer, varies by thread @param node current node vlib_node_runtime_t pointer @param next_index speculated next index used for both packets @param to_next speculated vector pointer used for both packets @param n_left_to_next number of slots left in speculated vector @param bi0 first buffer index @param bi1 second buffer index @param bi2 third buffer index @param bi3 fourth buffer index @param next0 actual next index to be used for the first packet @param next1 actual next index to be used for the second packet @param next2 actual next index to be used for the third packet @param next3 actual next index to be used for the fourth packet @return @c next_index -- speculative next index to be used for future packets @return @c to_next -- speculative frame to be used for future packets @return @c n_left_to_next -- number of slots left in speculative frame */ #define vlib_validate_buffer_enqueue_x4(vm,node,next_index,to_next,n_left_to_next,bi0,bi1,bi2,bi3,next0,next1,next2,next3) \ do { \ /* After the fact: check the [speculative] enqueue to "next" */ \ u32 fix_speculation = (next_index ^ next0) | (next_index ^ next1) \ | (next_index ^ next2) | (next_index ^ next3); \ if (PREDICT_FALSE(fix_speculation)) \ { \ /* rewind... */ \ to_next -= 4; \ n_left_to_next += 4; \ \ /* If bi0 belongs to "next", send it there */ \ if (next_index == next0) \ { \ to_next[0] = bi0; \ to_next++; \ n_left_to_next --; \ } \ else /* send it where it needs to go */ \ vlib_set_next_frame_buffer (vm, node, next0, bi0); \ \ if (next_index == next1) \ { \ to_next[0] = bi1; \ to_next++; \ n_left_to_next --; \ } \ else \ vlib_set_next_frame_buffer (vm, node, next1, bi1); \ \ if (next_index == next2) \ { \ to_next[0] = bi2; \ to_next++; \ n_left_to_next --; \ } \ else \ vlib_set_next_frame_buffer (vm, node, next2, bi2); \ \ if (next_index == next3) \ { \ to_next[0] = bi3; \ to_next++; \ n_left_to_next --; \ } \ else \ { \ vlib_set_next_frame_buffer (vm, node, next3, bi3); \ \ /* Change speculation: last 2 packets went to the same node*/ \ if (next2 == next3) \ { \ vlib_put_next_frame (vm, node, next_index, n_left_to_next); \ next_index = next3; \ vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); \ } \ } \ } \ } while(0); /** \brief Finish enqueueing one buffer forward in the graph. Standard single loop boilerplate element. This is a MACRO, with MULTIPLE SIDE EFFECTS. In the ideal case, <code>next_index == next0</code>, which means that the speculative enqueue at the top of the single loop has correctly dealt with the packet in hand. In that case, the macro does nothing at all. @param vm vlib_main_t pointer, varies by thread @param node current node vlib_node_runtime_t pointer @param next_index speculated next index used for both packets @param to_next speculated vector pointer used for both packets @param n_left_to_next number of slots left in speculated vector @param bi0 first buffer index @param next0 actual next index to be used for the first packet @return @c next_index -- speculative next index to be used for future packets @return @c to_next -- speculative frame to be used for future packets @return @c n_left_to_next -- number of slots left in speculative frame */ #define vlib_validate_buffer_enqueue_x1(vm,node,next_index,to_next,n_left_to_next,bi0,next0) \ do { \ if (PREDICT_FALSE (next0 != next_index)) \ { \ vlib_put_next_frame (vm, node, next_index, n_left_to_next + 1); \ next_index = next0; \ vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); \ \ to_next[0] = bi0; \ to_next += 1; \ n_left_to_next -= 1; \ } \ } while (0) always_inline uword generic_buffer_node_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame, uword sizeof_trace, void *opaque1, uword opaque2, void (*two_buffers) (vlib_main_t * vm, void *opaque1, uword opaque2, vlib_buffer_t * b0, vlib_buffer_t * b1, u32 * next0, u32 * next1), void (*one_buffer) (vlib_main_t * vm, void *opaque1, uword opaque2, vlib_buffer_t * b0, u32 * next0)) { u32 n_left_from, *from, *to_next; u32 next_index; from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; next_index = node->cached_next_index; if (node->flags & VLIB_NODE_FLAG_TRACE) vlib_trace_frame_buffers_only (vm, node, from, frame->n_vectors, /* stride */ 1, sizeof_trace); while (n_left_from > 0) { u32 n_left_to_next; vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); while (n_left_from >= 4 && n_left_to_next >= 2) { vlib_buffer_t *p0, *p1; u32 pi0, next0; u32 pi1, next1; /* Prefetch next iteration. */ { vlib_buffer_t *p2, *p3; p2 = vlib_get_buffer (vm, from[2]); p3 = vlib_get_buffer (vm, from[3]); vlib_prefetch_buffer_header (p2, LOAD); vlib_prefetch_buffer_header (p3, LOAD); CLIB_PREFETCH (p2->data, 64, LOAD); CLIB_PREFETCH (p3->data, 64, LOAD); } pi0 = to_next[0] = from[0]; pi1 = to_next[1] = from[1]; from += 2; to_next += 2; n_left_from -= 2; n_left_to_next -= 2; p0 = vlib_get_buffer (vm, pi0); p1 = vlib_get_buffer (vm, pi1); two_buffers (vm, opaque1, opaque2, p0, p1, &next0, &next1); vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, n_left_to_next, pi0, pi1, next0, next1); } while (n_left_from > 0 && n_left_to_next > 0) { vlib_buffer_t *p0; u32 pi0, next0; pi0 = from[0]; to_next[0] = pi0; from += 1; to_next += 1; n_left_from -= 1; n_left_to_next -= 1; p0 = vlib_get_buffer (vm, pi0); one_buffer (vm, opaque1, opaque2, p0, &next0); vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, n_left_to_next, pi0, next0); } vlib_put_next_frame (vm, node, next_index, n_left_to_next); } return frame->n_vectors; } static_always_inline void vlib_buffer_enqueue_to_next (vlib_main_t * vm, vlib_node_runtime_t * node, u32 * buffers, u16 * nexts, uword count) { u32 *to_next, n_left_to_next, max; u16 next_index; next_index = nexts[0]; vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); max = clib_min (n_left_to_next, count); while (count) { u32 n_enqueued; if ((nexts[0] != next_index) || n_left_to_next == 0) { vlib_put_next_frame (vm, node, next_index, n_left_to_next); next_index = nexts[0]; vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); max = clib_min (n_left_to_next, count); } #if defined(CLIB_HAVE_VEC512) u16x32 next32 = u16x32_load_unaligned (nexts); next32 = (next32 == u16x32_splat (next32[0])); u64 bitmap = u16x32_msb_mask (next32); n_enqueued = count_trailing_zeros (~bitmap); #elif defined(CLIB_HAVE_VEC256) u16x16 next16 = u16x16_load_unaligned (nexts); next16 = (next16 == u16x16_splat (next16[0])); u64 bitmap = u8x32_msb_mask ((u8x32) next16); n_enqueued = count_trailing_zeros (~bitmap) / 2; #elif defined(CLIB_HAVE_VEC128) && defined(CLIB_HAVE_VEC128_MSB_MASK) u16x8 next8 = u16x8_load_unaligned (nexts); next8 = (next8 == u16x8_splat (next8[0])); u64 bitmap = u8x16_msb_mask ((u8x16) next8); n_enqueued = count_trailing_zeros (~bitmap) / 2; #else u16 x = 0; if (count + 3 < max) { x |= next_index ^ nexts[1]; x |= next_index ^ nexts[2]; x |= next_index ^ nexts[3]; n_enqueued = (x == 0) ? 4 : 1; } else n_enqueued = 1; #endif if (PREDICT_FALSE (n_enqueued > max)) n_enqueued = max; #ifdef CLIB_HAVE_VEC512 if (n_enqueued >= 32) { clib_memcpy_fast (to_next, buffers, 32 * sizeof (u32)); nexts += 32; to_next += 32; buffers += 32; n_left_to_next -= 32; count -= 32; max -= 32; continue; } #endif #ifdef CLIB_HAVE_VEC256 if (n_enqueued >= 16) { clib_memcpy_fast (to_next, buffers, 16 * sizeof (u32)); nexts += 16; to_next += 16; buffers += 16; n_left_to_next -= 16; count -= 16; max -= 16; continue; } #endif #ifdef CLIB_HAVE_VEC128 if (n_enqueued >= 8) { clib_memcpy_fast (to_next, buffers, 8 * sizeof (u32)); nexts += 8; to_next += 8; buffers += 8; n_left_to_next -= 8; count -= 8; max -= 8; continue; } #endif if (n_enqueued >= 4) { clib_memcpy_fast (to_next, buffers, 4 * sizeof (u32)); nexts += 4; to_next += 4; buffers += 4; n_left_to_next -= 4; count -= 4; max -= 4; continue; } /* copy */ to_next[0] = buffers[0]; /* next */ nexts += 1; to_next += 1; buffers += 1; n_left_to_next -= 1; count -= 1; max -= 1; } vlib_put_next_frame (vm, node, next_index, n_left_to_next); } static_always_inline void vlib_buffer_enqueue_to_single_next (vlib_main_t * vm, vlib_node_runtime_t * node, u32 * buffers, u16 next_index, u32 count) { u32 *to_next, n_left_to_next, n_enq; vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); if (PREDICT_TRUE (n_left_to_next >= count)) { clib_memcpy_fast (to_next, buffers, count * sizeof (u32)); n_left_to_next -= count; vlib_put_next_frame (vm, node, next_index, n_left_to_next); return; } n_enq = n_left_to_next; next: clib_memcpy_fast (to_next, buffers, n_enq * sizeof (u32)); n_left_to_next -= n_enq; if (PREDICT_FALSE (count > n_enq)) { count -= n_enq; buffers += n_enq; vlib_put_next_frame (vm, node, next_index, n_left_to_next); vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); n_enq = clib_min (n_left_to_next, count); goto next; } vlib_put_next_frame (vm, node, next_index, n_left_to_next); } static_always_inline u32 vlib_buffer_enqueue_to_thread (vlib_main_t * vm, u32 frame_queue_index, u32 * buffer_indices, u16 * thread_indices, u32 n_packets, int drop_on_congestion) { vlib_thread_main_t *tm = vlib_get_thread_main (); vlib_frame_queue_main_t *fqm; vlib_frame_queue_per_thread_data_t *ptd; u32 n_left = n_packets; u32 drop_list[VLIB_FRAME_SIZE], *dbi = drop_list, n_drop = 0; vlib_frame_queue_elt_t *hf = 0; u32 n_left_to_next_thread = 0, *to_next_thread = 0; u32 next_thread_index, current_thread_index = ~0; int i; fqm = vec_elt_at_index (tm->frame_queue_mains, frame_queue_index); ptd = vec_elt_at_index (fqm->per_thread_data, vm->thread_index); while (n_left) { next_thread_index = thread_indices[0]; if (next_thread_index != current_thread_index) { if (drop_on_congestion && is_vlib_frame_queue_congested (frame_queue_index, next_thread_index, fqm->queue_hi_thresh, ptd->congested_handoff_queue_by_thread_index)) { dbi[0] = buffer_indices[0]; dbi++; n_drop++; goto next; } if (hf) hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_thread; hf = vlib_get_worker_handoff_queue_elt (frame_queue_index, next_thread_index, ptd->handoff_queue_elt_by_thread_index); n_left_to_next_thread = VLIB_FRAME_SIZE - hf->n_vectors; to_next_thread = &hf->buffer_index[hf->n_vectors]; current_thread_index = next_thread_index; } to_next_thread[0] = buffer_indices[0]; to_next_thread++; n_left_to_next_thread--; if (n_left_to_next_thread == 0) { hf->n_vectors = VLIB_FRAME_SIZE; vlib_put_frame_queue_elt (hf); current_thread_index = ~0; ptd->handoff_queue_elt_by_thread_index[next_thread_index] = 0; hf = 0; } /* next */ next: thread_indices += 1; buffer_indices += 1; n_left -= 1; } if (hf) hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_thread; /* Ship frames to the thread nodes */ for (i = 0; i < vec_len (ptd->handoff_queue_elt_by_thread_index); i++) { if (ptd->handoff_queue_elt_by_thread_index[i]) { hf = ptd->handoff_queue_elt_by_thread_index[i]; /* * It works better to let the handoff node * rate-adapt, always ship the handoff queue element. */ if (1 || hf->n_vectors == hf->last_n_vectors) { vlib_put_frame_queue_elt (hf); ptd->handoff_queue_elt_by_thread_index[i] = 0; } else hf->last_n_vectors = hf->n_vectors; } ptd->congested_handoff_queue_by_thread_index[i] = (vlib_frame_queue_t *) (~0); } if (drop_on_congestion && n_drop) vlib_buffer_free (vm, drop_list, n_drop); return n_packets - n_drop; } #endif /* included_vlib_buffer_node_h */ /* * fd.io coding-style-patch-verification: ON * * Local Variables: * eval: (c-set-style "gnu") * End: */