-/*-
- * BSD LICENSE
- *
- * Copyright 2017 6WIND S.A.
- * Copyright 2017 Mellanox.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of 6WIND S.A. nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright 2017 6WIND S.A.
+ * Copyright 2017 Mellanox.
*/
#include <errno.h>
#include <string.h>
#include <unistd.h>
#include <sys/queue.h>
+#include <sys/resource.h>
#include <rte_byteorder.h>
#include <rte_jhash.h>
#include <tap_flow.h>
#include <tap_autoconf.h>
#include <tap_tcmsgs.h>
+#include <tap_rss.h>
#ifndef HAVE_TC_FLOWER
/*
TCA_FLOWER_KEY_VLAN_ETH_TYPE, /* be16 */
};
#endif
+/*
+ * For kernels < 4.2 BPF related enums may not be defined.
+ * Runtime checks will be carried out to gracefully report on TC messages that
+ * are rejected by the kernel. Rejection reasons may be due to:
+ * 1. enum is not defined
+ * 2. enum is defined but kernel is not configured to support BPF system calls,
+ * BPF classifications or BPF actions.
+ */
+#ifndef HAVE_TC_BPF
+enum {
+ TCA_BPF_UNSPEC,
+ TCA_BPF_ACT,
+ TCA_BPF_POLICE,
+ TCA_BPF_CLASSID,
+ TCA_BPF_OPS_LEN,
+ TCA_BPF_OPS,
+};
+#endif
+#ifndef HAVE_TC_BPF_FD
+enum {
+ TCA_BPF_FD = TCA_BPF_OPS + 1,
+ TCA_BPF_NAME,
+};
+#endif
+#ifndef HAVE_TC_ACT_BPF
+#define tc_gen \
+ __u32 index; \
+ __u32 capab; \
+ int action; \
+ int refcnt; \
+ int bindcnt
+
+struct tc_act_bpf {
+ tc_gen;
+};
+
+enum {
+ TCA_ACT_BPF_UNSPEC,
+ TCA_ACT_BPF_TM,
+ TCA_ACT_BPF_PARMS,
+ TCA_ACT_BPF_OPS_LEN,
+ TCA_ACT_BPF_OPS,
+};
+
+#endif
+#ifndef HAVE_TC_ACT_BPF_FD
+enum {
+ TCA_ACT_BPF_FD = TCA_ACT_BPF_OPS + 1,
+ TCA_ACT_BPF_NAME,
+};
+#endif
+
+/* RSS key management */
+enum bpf_rss_key_e {
+ KEY_CMD_GET = 1,
+ KEY_CMD_RELEASE,
+ KEY_CMD_INIT,
+ KEY_CMD_DEINIT,
+};
+
+enum key_status_e {
+ KEY_STAT_UNSPEC,
+ KEY_STAT_USED,
+ KEY_STAT_AVAILABLE,
+};
#define ISOLATE_HANDLE 1
+#define REMOTE_PROMISCUOUS_HANDLE 2
struct rte_flow {
LIST_ENTRY(rte_flow) next; /* Pointer to the next rte_flow structure */
struct rte_flow *remote_flow; /* associated remote flow */
+ int bpf_fd[SEC_MAX]; /* list of bfs fds per ELF section */
+ uint32_t key_idx; /* RSS rule key index into BPF map */
struct nlmsg msg;
};
struct tc_skbedit skbedit;
uint16_t queue;
} skbedit;
+ struct bpf {
+ struct tc_act_bpf bpf;
+ int bpf_fd;
+ const char *annotation;
+ } bpf;
};
};
const struct rte_flow_action actions[],
struct rte_flow_error *error);
+static void
+tap_flow_free(struct pmd_internals *pmd,
+ struct rte_flow *flow);
+
static int
tap_flow_destroy(struct rte_eth_dev *dev,
struct rte_flow *flow,
int set,
struct rte_flow_error *error);
+static int bpf_rss_key(enum bpf_rss_key_e cmd, __u32 *key_idx);
+static int rss_enable(struct pmd_internals *pmd,
+ const struct rte_flow_attr *attr,
+ struct rte_flow_error *error);
+static int rss_add_actions(struct rte_flow *flow, struct pmd_internals *pmd,
+ const struct rte_flow_action_rss *rss,
+ struct rte_flow_error *error);
+
static const struct rte_flow_ops tap_flow_ops = {
.validate = tap_flow_validate,
.create = tap_flow_create,
&adata->skbedit.skbedit);
tap_nlattr_add16(&msg->nh, TCA_SKBEDIT_QUEUE_MAPPING,
adata->skbedit.queue);
+ } else if (strcmp("bpf", adata->id) == 0) {
+ tap_nlattr_add32(&msg->nh, TCA_ACT_BPF_FD, adata->bpf.bpf_fd);
+ tap_nlattr_add(&msg->nh, TCA_ACT_BPF_NAME,
+ strlen(adata->bpf.annotation) + 1,
+ adata->bpf.annotation);
+ tap_nlattr_add(&msg->nh, TCA_ACT_BPF_PARMS,
+ sizeof(adata->bpf.bpf),
+ &adata->bpf.bpf);
} else {
return -1;
}
return -rte_errno;
} else if (flow) {
uint16_t group = attr->group << GROUP_SHIFT;
- uint16_t prio = group | (attr->priority + PRIORITY_OFFSET);
+ uint16_t prio = group | (attr->priority +
+ RSS_PRIORITY_OFFSET + PRIORITY_OFFSET);
flow->msg.t.tcm_info = TC_H_MAKE(prio << 16,
flow->msg.t.tcm_info);
}
TCA_FLOWER_ACT);
}
} else if (actions->type == RTE_FLOW_ACTION_TYPE_RSS) {
- /* Fake RSS support. */
const struct rte_flow_action_rss *rss =
(const struct rte_flow_action_rss *)
actions->conf;
if (action++)
goto exit_action_not_supported;
- if (!rss || rss->num < 1 ||
- (rss->queue[0] > pmd->dev->data->nb_rx_queues - 1))
- goto exit_action_not_supported;
- if (flow) {
- struct action_data adata = {
- .id = "skbedit",
- .skbedit = {
- .skbedit = {
- .action = TC_ACT_PIPE,
- },
- .queue = rss->queue[0],
- },
- };
-
- err = add_actions(flow, 1, &adata,
- TCA_FLOWER_ACT);
+ if (!pmd->rss_enabled) {
+ err = rss_enable(pmd, attr, error);
+ if (err)
+ goto exit_action_not_supported;
}
+ if (flow && rss)
+ err = rss_add_actions(flow, pmd, rss, error);
} else {
goto exit_action_not_supported;
}
flow->msg.t.tcm_handle = handle;
}
+/**
+ * Free the flow opened file descriptors and allocated memory
+ *
+ * @param[in] flow
+ * Pointer to the flow to free
+ *
+ */
+static void
+tap_flow_free(struct pmd_internals *pmd, struct rte_flow *flow)
+{
+ int i;
+
+ if (!flow)
+ return;
+
+ if (pmd->rss_enabled) {
+ /* Close flow BPF file descriptors */
+ for (i = 0; i < SEC_MAX; i++)
+ if (flow->bpf_fd[i] != 0) {
+ close(flow->bpf_fd[i]);
+ flow->bpf_fd[i] = 0;
+ }
+
+ /* Release the map key for this RSS rule */
+ bpf_rss_key(KEY_CMD_RELEASE, &flow->key_idx);
+ flow->key_idx = 0;
+ }
+
+ /* Free flow allocated memory */
+ rte_free(flow);
+}
+
/**
* Create a flow.
*
if (remote_flow)
rte_free(remote_flow);
if (flow)
- rte_free(flow);
+ tap_flow_free(pmd, flow);
return NULL;
}
"couldn't receive kernel ack to our request");
goto end;
}
+
if (remote_flow) {
remote_flow->msg.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
remote_flow->msg.nh.nlmsg_type = RTM_DELTFILTER;
end:
if (remote_flow)
rte_free(remote_flow);
- rte_free(flow);
+ tap_flow_free(pmd, flow);
return ret;
}
* The ISOLATE rule is always present and must have a static handle, as
* the action is changed whether the feature is enabled (DROP) or
* disabled (PASSTHRU).
+ * There is just one REMOTE_PROMISCUOUS rule in all cases. It should
+ * have a static handle such that adding it twice will fail with EEXIST
+ * with any kernel version. Remark: old kernels may falsely accept the
+ * same REMOTE_PROMISCUOUS rules if they had different handles.
*/
if (idx == TAP_ISOLATE)
remote_flow->msg.t.tcm_handle = ISOLATE_HANDLE;
+ else if (idx == TAP_REMOTE_PROMISC)
+ remote_flow->msg.t.tcm_handle = REMOTE_PROMISCUOUS_HANDLE;
else
tap_flow_set_handle(remote_flow);
if (priv_flow_process(pmd, attr, items, actions, NULL,
}
err = tap_nl_recv_ack(pmd->nlsk_fd);
if (err < 0) {
+ /* Silently ignore re-entering remote promiscuous rule */
+ if (errno == EEXIST && idx == TAP_REMOTE_PROMISC)
+ goto success;
RTE_LOG(ERR, PMD,
"Kernel refused TC filter rule creation (%d): %s\n",
errno, strerror(errno));
goto fail;
}
LIST_INSERT_HEAD(&pmd->implicit_flows, remote_flow, next);
+success:
return 0;
fail:
if (remote_flow)
return 0;
}
+#define MAX_RSS_KEYS 256
+#define KEY_IDX_OFFSET (3 * MAX_RSS_KEYS)
+#define SEC_NAME_CLS_Q "cls_q"
+
+const char *sec_name[SEC_MAX] = {
+ [SEC_L3_L4] = "l3_l4",
+};
+
+/**
+ * Enable RSS on tap: create TC rules for queuing.
+ *
+ * @param[in, out] pmd
+ * Pointer to private structure.
+ *
+ * @param[in] attr
+ * Pointer to rte_flow to get flow group
+ *
+ * @param[out] error
+ * Pointer to error reporting if not NULL.
+ *
+ * @return 0 on success, negative value on failure.
+ */
+static int rss_enable(struct pmd_internals *pmd,
+ const struct rte_flow_attr *attr,
+ struct rte_flow_error *error)
+{
+ struct rte_flow *rss_flow = NULL;
+ struct nlmsg *msg = NULL;
+ /* 4096 is the maximum number of instructions for a BPF program */
+ char annotation[64];
+ int i;
+ int err = 0;
+
+ /* unlimit locked memory */
+ struct rlimit memlock_limit = {
+ .rlim_cur = RLIM_INFINITY,
+ .rlim_max = RLIM_INFINITY,
+ };
+ setrlimit(RLIMIT_MEMLOCK, &memlock_limit);
+
+ /* Get a new map key for a new RSS rule */
+ err = bpf_rss_key(KEY_CMD_INIT, NULL);
+ if (err < 0) {
+ rte_flow_error_set(
+ error, EINVAL, RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "Failed to initialize BPF RSS keys");
+
+ return -1;
+ }
+
+ /*
+ * Create BPF RSS MAP
+ */
+ pmd->map_fd = tap_flow_bpf_rss_map_create(sizeof(__u32), /* key size */
+ sizeof(struct rss_key),
+ MAX_RSS_KEYS);
+ if (pmd->map_fd < 0) {
+ RTE_LOG(ERR, PMD,
+ "Failed to create BPF map (%d): %s\n",
+ errno, strerror(errno));
+ rte_flow_error_set(
+ error, ENOTSUP, RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "Kernel too old or not configured "
+ "to support BPF maps");
+
+ return -ENOTSUP;
+ }
+
+ /*
+ * Add a rule per queue to match reclassified packets and direct them to
+ * the correct queue.
+ */
+ for (i = 0; i < pmd->dev->data->nb_rx_queues; i++) {
+ pmd->bpf_fd[i] = tap_flow_bpf_cls_q(i);
+ if (pmd->bpf_fd[i] < 0) {
+ RTE_LOG(ERR, PMD,
+ "Failed to load BPF section %s for queue %d",
+ SEC_NAME_CLS_Q, i);
+ rte_flow_error_set(
+ error, ENOTSUP, RTE_FLOW_ERROR_TYPE_HANDLE,
+ NULL,
+ "Kernel too old or not configured "
+ "to support BPF programs loading");
+
+ return -ENOTSUP;
+ }
+
+ rss_flow = rte_malloc(__func__, sizeof(struct rte_flow), 0);
+ if (!rss_flow) {
+ RTE_LOG(ERR, PMD,
+ "Cannot allocate memory for rte_flow");
+ return -1;
+ }
+ msg = &rss_flow->msg;
+ tc_init_msg(msg, pmd->if_index, RTM_NEWTFILTER, NLM_F_REQUEST |
+ NLM_F_ACK | NLM_F_EXCL | NLM_F_CREATE);
+ msg->t.tcm_info = TC_H_MAKE(0, htons(ETH_P_ALL));
+ tap_flow_set_handle(rss_flow);
+ uint16_t group = attr->group << GROUP_SHIFT;
+ uint16_t prio = group | (i + PRIORITY_OFFSET);
+ msg->t.tcm_info = TC_H_MAKE(prio << 16, msg->t.tcm_info);
+ msg->t.tcm_parent = TC_H_MAKE(MULTIQ_MAJOR_HANDLE, 0);
+
+ tap_nlattr_add(&msg->nh, TCA_KIND, sizeof("bpf"), "bpf");
+ if (tap_nlattr_nested_start(msg, TCA_OPTIONS) < 0)
+ return -1;
+ tap_nlattr_add32(&msg->nh, TCA_BPF_FD, pmd->bpf_fd[i]);
+ snprintf(annotation, sizeof(annotation), "[%s%d]",
+ SEC_NAME_CLS_Q, i);
+ tap_nlattr_add(&msg->nh, TCA_BPF_NAME, strlen(annotation) + 1,
+ annotation);
+ /* Actions */
+ {
+ struct action_data adata = {
+ .id = "skbedit",
+ .skbedit = {
+ .skbedit = {
+ .action = TC_ACT_PIPE,
+ },
+ .queue = i,
+ },
+ };
+ if (add_actions(rss_flow, 1, &adata, TCA_BPF_ACT) < 0)
+ return -1;
+ }
+ tap_nlattr_nested_finish(msg); /* nested TCA_OPTIONS */
+
+ /* Netlink message is now ready to be sent */
+ if (tap_nl_send(pmd->nlsk_fd, &msg->nh) < 0)
+ return -1;
+ err = tap_nl_recv_ack(pmd->nlsk_fd);
+ if (err < 0) {
+ RTE_LOG(ERR, PMD,
+ "Kernel refused TC filter rule creation (%d): %s\n",
+ errno, strerror(errno));
+ return err;
+ }
+ LIST_INSERT_HEAD(&pmd->rss_flows, rss_flow, next);
+ }
+
+ pmd->rss_enabled = 1;
+ return err;
+}
+
+/**
+ * Manage bpf RSS keys repository with operations: init, get, release
+ *
+ * @param[in] cmd
+ * Command on RSS keys: init, get, release
+ *
+ * @param[in, out] key_idx
+ * Pointer to RSS Key index (out for get command, in for release command)
+ *
+ * @return -1 if couldn't get, release or init the RSS keys, 0 otherwise.
+ */
+static int bpf_rss_key(enum bpf_rss_key_e cmd, __u32 *key_idx)
+{
+ __u32 i;
+ int err = 0;
+ static __u32 num_used_keys;
+ static __u32 rss_keys[MAX_RSS_KEYS] = {KEY_STAT_UNSPEC};
+ static __u32 rss_keys_initialized;
+ __u32 key;
+
+ switch (cmd) {
+ case KEY_CMD_GET:
+ if (!rss_keys_initialized) {
+ err = -1;
+ break;
+ }
+
+ if (num_used_keys == RTE_DIM(rss_keys)) {
+ err = -1;
+ break;
+ }
+
+ *key_idx = num_used_keys % RTE_DIM(rss_keys);
+ while (rss_keys[*key_idx] == KEY_STAT_USED)
+ *key_idx = (*key_idx + 1) % RTE_DIM(rss_keys);
+
+ rss_keys[*key_idx] = KEY_STAT_USED;
+
+ /*
+ * Add an offset to key_idx in order to handle a case of
+ * RSS and non RSS flows mixture.
+ * If a non RSS flow is destroyed it has an eBPF map
+ * index 0 (initialized on flow creation) and might
+ * unintentionally remove RSS entry 0 from eBPF map.
+ * To avoid this issue, add an offset to the real index
+ * during a KEY_CMD_GET operation and subtract this offset
+ * during a KEY_CMD_RELEASE operation in order to restore
+ * the real index.
+ */
+ *key_idx += KEY_IDX_OFFSET;
+ num_used_keys++;
+ break;
+
+ case KEY_CMD_RELEASE:
+ if (!rss_keys_initialized)
+ break;
+
+ /*
+ * Subtract offest to restore real key index
+ * If a non RSS flow is falsely trying to release map
+ * entry 0 - the offset subtraction will calculate the real
+ * map index as an out-of-range value and the release operation
+ * will be silently ignored.
+ */
+ key = *key_idx - KEY_IDX_OFFSET;
+ if (key >= RTE_DIM(rss_keys))
+ break;
+
+ if (rss_keys[key] == KEY_STAT_USED) {
+ rss_keys[key] = KEY_STAT_AVAILABLE;
+ num_used_keys--;
+ }
+ break;
+
+ case KEY_CMD_INIT:
+ for (i = 0; i < RTE_DIM(rss_keys); i++)
+ rss_keys[i] = KEY_STAT_AVAILABLE;
+
+ rss_keys_initialized = 1;
+ num_used_keys = 0;
+ break;
+
+ case KEY_CMD_DEINIT:
+ for (i = 0; i < RTE_DIM(rss_keys); i++)
+ rss_keys[i] = KEY_STAT_UNSPEC;
+
+ rss_keys_initialized = 0;
+ num_used_keys = 0;
+ break;
+
+ default:
+ break;
+ }
+
+ return err;
+}
+
+/**
+ * Add RSS hash calculations and queue selection
+ *
+ * @param[in, out] pmd
+ * Pointer to internal structure. Used to set/get RSS map fd
+ *
+ * @param[in] rss
+ * Pointer to RSS flow actions
+ *
+ * @param[out] error
+ * Pointer to error reporting if not NULL.
+ *
+ * @return 0 on success, negative value on failure
+ */
+static int rss_add_actions(struct rte_flow *flow, struct pmd_internals *pmd,
+ const struct rte_flow_action_rss *rss,
+ struct rte_flow_error *error)
+{
+ /* 4096 is the maximum number of instructions for a BPF program */
+ int i;
+ int err;
+ struct rss_key rss_entry = { .hash_fields = 0,
+ .key_size = 0 };
+
+ /* Get a new map key for a new RSS rule */
+ err = bpf_rss_key(KEY_CMD_GET, &flow->key_idx);
+ if (err < 0) {
+ rte_flow_error_set(
+ error, EINVAL, RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "Failed to get BPF RSS key");
+
+ return -1;
+ }
+
+ /* Update RSS map entry with queues */
+ rss_entry.nb_queues = rss->num;
+ for (i = 0; i < rss->num; i++)
+ rss_entry.queues[i] = rss->queue[i];
+ rss_entry.hash_fields =
+ (1 << HASH_FIELD_IPV4_L3_L4) | (1 << HASH_FIELD_IPV6_L3_L4);
+
+ /* Add this RSS entry to map */
+ err = tap_flow_bpf_update_rss_elem(pmd->map_fd,
+ &flow->key_idx, &rss_entry);
+
+ if (err) {
+ RTE_LOG(ERR, PMD,
+ "Failed to update BPF map entry #%u (%d): %s\n",
+ flow->key_idx, errno, strerror(errno));
+ rte_flow_error_set(
+ error, ENOTSUP, RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "Kernel too old or not configured "
+ "to support BPF maps updates");
+
+ return -ENOTSUP;
+ }
+
+
+ /*
+ * Load bpf rules to calculate hash for this key_idx
+ */
+
+ flow->bpf_fd[SEC_L3_L4] =
+ tap_flow_bpf_calc_l3_l4_hash(flow->key_idx, pmd->map_fd);
+ if (flow->bpf_fd[SEC_L3_L4] < 0) {
+ RTE_LOG(ERR, PMD,
+ "Failed to load BPF section %s (%d): %s\n",
+ sec_name[SEC_L3_L4], errno, strerror(errno));
+ rte_flow_error_set(
+ error, ENOTSUP, RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "Kernel too old or not configured "
+ "to support BPF program loading");
+
+ return -ENOTSUP;
+ }
+
+ /* Actions */
+ {
+ struct action_data adata[] = {
+ {
+ .id = "bpf",
+ .bpf = {
+ .bpf_fd = flow->bpf_fd[SEC_L3_L4],
+ .annotation = sec_name[SEC_L3_L4],
+ .bpf = {
+ .action = TC_ACT_PIPE,
+ },
+ },
+ },
+ };
+
+ if (add_actions(flow, RTE_DIM(adata), adata,
+ TCA_FLOWER_ACT) < 0)
+ return -1;
+ }
+
+ return 0;
+}
+
/**
* Manage filter operations.
*