diff options
Diffstat (limited to 'drivers/net/ethernet/netronome/nfp/bpf')
| -rw-r--r-- | drivers/net/ethernet/netronome/nfp/bpf/cmsg.c | 453 | ||||
| -rw-r--r-- | drivers/net/ethernet/netronome/nfp/bpf/fw.h | 157 | ||||
| -rw-r--r-- | drivers/net/ethernet/netronome/nfp/bpf/jit.c | 988 | ||||
| -rw-r--r-- | drivers/net/ethernet/netronome/nfp/bpf/main.c | 248 | ||||
| -rw-r--r-- | drivers/net/ethernet/netronome/nfp/bpf/main.h | 207 | ||||
| -rw-r--r-- | drivers/net/ethernet/netronome/nfp/bpf/offload.c | 198 | ||||
| -rw-r--r-- | drivers/net/ethernet/netronome/nfp/bpf/verifier.c | 159 |
7 files changed, 2189 insertions, 221 deletions
diff --git a/drivers/net/ethernet/netronome/nfp/bpf/cmsg.c b/drivers/net/ethernet/netronome/nfp/bpf/cmsg.c new file mode 100644 index 000000000000..80d3aa0fc9d3 --- /dev/null +++ b/drivers/net/ethernet/netronome/nfp/bpf/cmsg.c @@ -0,0 +1,453 @@ +/* + * Copyright (C) 2017 Netronome Systems, Inc. + * + * This software is dual licensed under the GNU General License Version 2, + * June 1991 as shown in the file COPYING in the top-level directory of this + * source tree or the BSD 2-Clause License provided below. You have the + * option to license this software under the complete terms of either license. + * + * The BSD 2-Clause License: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * 1. Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * 2. 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. + * + * 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. + */ + +#include <linux/bpf.h> +#include <linux/bitops.h> +#include <linux/bug.h> +#include <linux/jiffies.h> +#include <linux/skbuff.h> +#include <linux/wait.h> + +#include "../nfp_app.h" +#include "../nfp_net.h" +#include "fw.h" +#include "main.h" + +#define cmsg_warn(bpf, msg...) nn_dp_warn(&(bpf)->app->ctrl->dp, msg) + +#define NFP_BPF_TAG_ALLOC_SPAN (U16_MAX / 4) + +static bool nfp_bpf_all_tags_busy(struct nfp_app_bpf *bpf) +{ + u16 used_tags; + + used_tags = bpf->tag_alloc_next - bpf->tag_alloc_last; + + return used_tags > NFP_BPF_TAG_ALLOC_SPAN; +} + +static int nfp_bpf_alloc_tag(struct nfp_app_bpf *bpf) +{ + /* All FW communication for BPF is request-reply. To make sure we + * don't reuse the message ID too early after timeout - limit the + * number of requests in flight. + */ + if (nfp_bpf_all_tags_busy(bpf)) { + cmsg_warn(bpf, "all FW request contexts busy!\n"); + return -EAGAIN; + } + + WARN_ON(__test_and_set_bit(bpf->tag_alloc_next, bpf->tag_allocator)); + return bpf->tag_alloc_next++; +} + +static void nfp_bpf_free_tag(struct nfp_app_bpf *bpf, u16 tag) +{ + WARN_ON(!__test_and_clear_bit(tag, bpf->tag_allocator)); + + while (!test_bit(bpf->tag_alloc_last, bpf->tag_allocator) && + bpf->tag_alloc_last != bpf->tag_alloc_next) + bpf->tag_alloc_last++; +} + +static struct sk_buff * +nfp_bpf_cmsg_alloc(struct nfp_app_bpf *bpf, unsigned int size) +{ + struct sk_buff *skb; + + skb = nfp_app_ctrl_msg_alloc(bpf->app, size, GFP_KERNEL); + skb_put(skb, size); + + return skb; +} + +static struct sk_buff * +nfp_bpf_cmsg_map_req_alloc(struct nfp_app_bpf *bpf, unsigned int n) +{ + unsigned int size; + + size = sizeof(struct cmsg_req_map_op); + size += sizeof(struct cmsg_key_value_pair) * n; + + return nfp_bpf_cmsg_alloc(bpf, size); +} + +static unsigned int nfp_bpf_cmsg_get_tag(struct sk_buff *skb) +{ + struct cmsg_hdr *hdr; + + hdr = (struct cmsg_hdr *)skb->data; + + return be16_to_cpu(hdr->tag); +} + +static struct sk_buff *__nfp_bpf_reply(struct nfp_app_bpf *bpf, u16 tag) +{ + unsigned int msg_tag; + struct sk_buff *skb; + + skb_queue_walk(&bpf->cmsg_replies, skb) { + msg_tag = nfp_bpf_cmsg_get_tag(skb); + if (msg_tag == tag) { + nfp_bpf_free_tag(bpf, tag); + __skb_unlink(skb, &bpf->cmsg_replies); + return skb; + } + } + + return NULL; +} + +static struct sk_buff *nfp_bpf_reply(struct nfp_app_bpf *bpf, u16 tag) +{ + struct sk_buff *skb; + + nfp_ctrl_lock(bpf->app->ctrl); + skb = __nfp_bpf_reply(bpf, tag); + nfp_ctrl_unlock(bpf->app->ctrl); + + return skb; +} + +static struct sk_buff *nfp_bpf_reply_drop_tag(struct nfp_app_bpf *bpf, u16 tag) +{ + struct sk_buff *skb; + + nfp_ctrl_lock(bpf->app->ctrl); + skb = __nfp_bpf_reply(bpf, tag); + if (!skb) + nfp_bpf_free_tag(bpf, tag); + nfp_ctrl_unlock(bpf->app->ctrl); + + return skb; +} + +static struct sk_buff * +nfp_bpf_cmsg_wait_reply(struct nfp_app_bpf *bpf, enum nfp_bpf_cmsg_type type, + int tag) +{ + struct sk_buff *skb; + int i, err; + + for (i = 0; i < 50; i++) { + udelay(4); + skb = nfp_bpf_reply(bpf, tag); + if (skb) + return skb; + } + + err = wait_event_interruptible_timeout(bpf->cmsg_wq, + skb = nfp_bpf_reply(bpf, tag), + msecs_to_jiffies(5000)); + /* We didn't get a response - try last time and atomically drop + * the tag even if no response is matched. + */ + if (!skb) + skb = nfp_bpf_reply_drop_tag(bpf, tag); + if (err < 0) { + cmsg_warn(bpf, "%s waiting for response to 0x%02x: %d\n", + err == ERESTARTSYS ? "interrupted" : "error", + type, err); + return ERR_PTR(err); + } + if (!skb) { + cmsg_warn(bpf, "timeout waiting for response to 0x%02x\n", + type); + return ERR_PTR(-ETIMEDOUT); + } + + return skb; +} + +static struct sk_buff * +nfp_bpf_cmsg_communicate(struct nfp_app_bpf *bpf, struct sk_buff *skb, + enum nfp_bpf_cmsg_type type, unsigned int reply_size) +{ + struct cmsg_hdr *hdr; + int tag; + + nfp_ctrl_lock(bpf->app->ctrl); + tag = nfp_bpf_alloc_tag(bpf); + if (tag < 0) { + nfp_ctrl_unlock(bpf->app->ctrl); + dev_kfree_skb_any(skb); + return ERR_PTR(tag); + } + + hdr = (void *)skb->data; + hdr->ver = CMSG_MAP_ABI_VERSION; + hdr->type = type; + hdr->tag = cpu_to_be16(tag); + + __nfp_app_ctrl_tx(bpf->app, skb); + + nfp_ctrl_unlock(bpf->app->ctrl); + + skb = nfp_bpf_cmsg_wait_reply(bpf, type, tag); + if (IS_ERR(skb)) + return skb; + + hdr = (struct cmsg_hdr *)skb->data; + /* 0 reply_size means caller will do the validation */ + if (reply_size && skb->len != reply_size) { + cmsg_warn(bpf, "cmsg drop - wrong size %d != %d!\n", + skb->len, reply_size); + goto err_free; + } + if (hdr->type != __CMSG_REPLY(type)) { + cmsg_warn(bpf, "cmsg drop - wrong type 0x%02x != 0x%02lx!\n", + hdr->type, __CMSG_REPLY(type)); + goto err_free; + } + + return skb; +err_free: + dev_kfree_skb_any(skb); + return ERR_PTR(-EIO); +} + +static int +nfp_bpf_ctrl_rc_to_errno(struct nfp_app_bpf *bpf, + struct cmsg_reply_map_simple *reply) +{ + static const int res_table[] = { + [CMSG_RC_SUCCESS] = 0, + [CMSG_RC_ERR_MAP_FD] = -EBADFD, + [CMSG_RC_ERR_MAP_NOENT] = -ENOENT, + [CMSG_RC_ERR_MAP_ERR] = -EINVAL, + [CMSG_RC_ERR_MAP_PARSE] = -EIO, + [CMSG_RC_ERR_MAP_EXIST] = -EEXIST, + [CMSG_RC_ERR_MAP_NOMEM] = -ENOMEM, + [CMSG_RC_ERR_MAP_E2BIG] = -E2BIG, + }; + u32 rc; + + rc = be32_to_cpu(reply->rc); + if (rc >= ARRAY_SIZE(res_table)) { + cmsg_warn(bpf, "FW responded with invalid status: %u\n", rc); + return -EIO; + } + + return res_table[rc]; +} + +long long int +nfp_bpf_ctrl_alloc_map(struct nfp_app_bpf *bpf, struct bpf_map *map) +{ + struct cmsg_reply_map_alloc_tbl *reply; + struct cmsg_req_map_alloc_tbl *req; + struct sk_buff *skb; + u32 tid; + int err; + + skb = nfp_bpf_cmsg_alloc(bpf, sizeof(*req)); + if (!skb) + return -ENOMEM; + + req = (void *)skb->data; + req->key_size = cpu_to_be32(map->key_size); + req->value_size = cpu_to_be32(map->value_size); + req->max_entries = cpu_to_be32(map->max_entries); + req->map_type = cpu_to_be32(map->map_type); + req->map_flags = 0; + + skb = nfp_bpf_cmsg_communicate(bpf, skb, CMSG_TYPE_MAP_ALLOC, + sizeof(*reply)); + if (IS_ERR(skb)) + return PTR_ERR(skb); + + reply = (void *)skb->data; + err = nfp_bpf_ctrl_rc_to_errno(bpf, &reply->reply_hdr); + if (err) + goto err_free; + + tid = be32_to_cpu(reply->tid); + dev_consume_skb_any(skb); + + return tid; +err_free: + dev_kfree_skb_any(skb); + return err; +} + +void nfp_bpf_ctrl_free_map(struct nfp_app_bpf *bpf, struct nfp_bpf_map *nfp_map) +{ + struct cmsg_reply_map_free_tbl *reply; + struct cmsg_req_map_free_tbl *req; + struct sk_buff *skb; + int err; + + skb = nfp_bpf_cmsg_alloc(bpf, sizeof(*req)); + if (!skb) { + cmsg_warn(bpf, "leaking map - failed to allocate msg\n"); + return; + } + + req = (void *)skb->data; + req->tid = cpu_to_be32(nfp_map->tid); + + skb = nfp_bpf_cmsg_communicate(bpf, skb, CMSG_TYPE_MAP_FREE, + sizeof(*reply)); + if (IS_ERR(skb)) { + cmsg_warn(bpf, "leaking map - I/O error\n"); + return; + } + + reply = (void *)skb->data; + err = nfp_bpf_ctrl_rc_to_errno(bpf, &reply->reply_hdr); + if (err) + cmsg_warn(bpf, "leaking map - FW responded with: %d\n", err); + + dev_consume_skb_any(skb); +} + +static int +nfp_bpf_ctrl_entry_op(struct bpf_offloaded_map *offmap, + enum nfp_bpf_cmsg_type op, + u8 *key, u8 *value, u64 flags, u8 *out_key, u8 *out_value) +{ + struct nfp_bpf_map *nfp_map = offmap->dev_priv; + struct nfp_app_bpf *bpf = nfp_map->bpf; + struct bpf_map *map = &offmap->map; + struct cmsg_reply_map_op *reply; + struct cmsg_req_map_op *req; + struct sk_buff *skb; + int err; + + /* FW messages have no space for more than 32 bits of flags */ + if (flags >> 32) + return -EOPNOTSUPP; + + skb = nfp_bpf_cmsg_map_req_alloc(bpf, 1); + if (!skb) + return -ENOMEM; + + req = (void *)skb->data; + req->tid = cpu_to_be32(nfp_map->tid); + req->count = cpu_to_be32(1); + req->flags = cpu_to_be32(flags); + + /* Copy inputs */ + if (key) + memcpy(&req->elem[0].key, key, map->key_size); + if (value) + memcpy(&req->elem[0].value, value, map->value_size); + + skb = nfp_bpf_cmsg_communicate(bpf, skb, op, + sizeof(*reply) + sizeof(*reply->elem)); + if (IS_ERR(skb)) + return PTR_ERR(skb); + + reply = (void *)skb->data; + err = nfp_bpf_ctrl_rc_to_errno(bpf, &reply->reply_hdr); + if (err) + goto err_free; + + /* Copy outputs */ + if (out_key) + memcpy(out_key, &reply->elem[0].key, map->key_size); + if (out_value) + memcpy(out_value, &reply->elem[0].value, map->value_size); + + dev_consume_skb_any(skb); + + return 0; +err_free: + dev_kfree_skb_any(skb); + return err; +} + +int nfp_bpf_ctrl_update_entry(struct bpf_offloaded_map *offmap, + void *key, void *value, u64 flags) +{ + return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_UPDATE, + key, value, flags, NULL, NULL); +} + +int nfp_bpf_ctrl_del_entry(struct bpf_offloaded_map *offmap, void *key) +{ + return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_DELETE, + key, NULL, 0, NULL, NULL); +} + +int nfp_bpf_ctrl_lookup_entry(struct bpf_offloaded_map *offmap, + void *key, void *value) +{ + return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_LOOKUP, + key, NULL, 0, NULL, value); +} + +int nfp_bpf_ctrl_getfirst_entry(struct bpf_offloaded_map *offmap, + void *next_key) +{ + return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_GETFIRST, + NULL, NULL, 0, next_key, NULL); +} + +int nfp_bpf_ctrl_getnext_entry(struct bpf_offloaded_map *offmap, + void *key, void *next_key) +{ + return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_GETNEXT, + key, NULL, 0, next_key, NULL); +} + +void nfp_bpf_ctrl_msg_rx(struct nfp_app *app, struct sk_buff *skb) +{ + struct nfp_app_bpf *bpf = app->priv; + unsigned int tag; + + if (unlikely(skb->len < sizeof(struct cmsg_reply_map_simple))) { + cmsg_warn(bpf, "cmsg drop - too short %d!\n", skb->len); + goto err_free; + } + + nfp_ctrl_lock(bpf->app->ctrl); + + tag = nfp_bpf_cmsg_get_tag(skb); + if (unlikely(!test_bit(tag, bpf->tag_allocator))) { + cmsg_warn(bpf, "cmsg drop - no one is waiting for tag %u!\n", + tag); + goto err_unlock; + } + + __skb_queue_tail(&bpf->cmsg_replies, skb); + wake_up_interruptible_all(&bpf->cmsg_wq); + + nfp_ctrl_unlock(bpf->app->ctrl); + + return; +err_unlock: + nfp_ctrl_unlock(bpf->app->ctrl); +err_free: + dev_kfree_skb_any(skb); +} diff --git a/drivers/net/ethernet/netronome/nfp/bpf/fw.h b/drivers/net/ethernet/netronome/nfp/bpf/fw.h new file mode 100644 index 000000000000..cfcc7bcb2c67 --- /dev/null +++ b/drivers/net/ethernet/netronome/nfp/bpf/fw.h @@ -0,0 +1,157 @@ +/* + * Copyright (C) 2017 Netronome Systems, Inc. + * + * This software is dual licensed under the GNU General License Version 2, + * June 1991 as shown in the file COPYING in the top-level directory of this + * source tree or the BSD 2-Clause License provided below. You have the + * option to license this software under the complete terms of either license. + * + * The BSD 2-Clause License: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * 1. Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * 2. 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. + * + * 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 NFP_BPF_FW_H +#define NFP_BPF_FW_H 1 + +#include <linux/bitops.h> +#include <linux/types.h> + +enum bpf_cap_tlv_type { + NFP_BPF_CAP_TYPE_FUNC = 1, + NFP_BPF_CAP_TYPE_ADJUST_HEAD = 2, + NFP_BPF_CAP_TYPE_MAPS = 3, +}; + +struct nfp_bpf_cap_tlv_func { + __le32 func_id; + __le32 func_addr; +}; + +struct nfp_bpf_cap_tlv_adjust_head { + __le32 flags; + __le32 off_min; + __le32 off_max; + __le32 guaranteed_sub; + __le32 guaranteed_add; +}; + +#define NFP_BPF_ADJUST_HEAD_NO_META BIT(0) + +struct nfp_bpf_cap_tlv_maps { + __le32 types; + __le32 max_maps; + __le32 max_elems; + __le32 max_key_sz; + __le32 max_val_sz; + __le32 max_elem_sz; +}; + +/* + * Types defined for map related control messages + */ +#define CMSG_MAP_ABI_VERSION 1 + +enum nfp_bpf_cmsg_type { + CMSG_TYPE_MAP_ALLOC = 1, + CMSG_TYPE_MAP_FREE = 2, + CMSG_TYPE_MAP_LOOKUP = 3, + CMSG_TYPE_MAP_UPDATE = 4, + CMSG_TYPE_MAP_DELETE = 5, + CMSG_TYPE_MAP_GETNEXT = 6, + CMSG_TYPE_MAP_GETFIRST = 7, + __CMSG_TYPE_MAP_MAX, +}; + +#define CMSG_TYPE_MAP_REPLY_BIT 7 +#define __CMSG_REPLY(req) (BIT(CMSG_TYPE_MAP_REPLY_BIT) | (req)) + +#define CMSG_MAP_KEY_LW 16 +#define CMSG_MAP_VALUE_LW 16 + +enum nfp_bpf_cmsg_status { + CMSG_RC_SUCCESS = 0, + CMSG_RC_ERR_MAP_FD = 1, + CMSG_RC_ERR_MAP_NOENT = 2, + CMSG_RC_ERR_MAP_ERR = 3, + CMSG_RC_ERR_MAP_PARSE = 4, + CMSG_RC_ERR_MAP_EXIST = 5, + CMSG_RC_ERR_MAP_NOMEM = 6, + CMSG_RC_ERR_MAP_E2BIG = 7, +}; + +struct cmsg_hdr { + u8 type; + u8 ver; + __be16 tag; +}; + +struct cmsg_reply_map_simple { + struct cmsg_hdr hdr; + __be32 rc; +}; + +struct cmsg_req_map_alloc_tbl { + struct cmsg_hdr hdr; + __be32 key_size; /* in bytes */ + __be32 value_size; /* in bytes */ + __be32 max_entries; + __be32 map_type; + __be32 map_flags; /* reserved */ +}; + +struct cmsg_reply_map_alloc_tbl { + struct cmsg_reply_map_simple reply_hdr; + __be32 tid; +}; + +struct cmsg_req_map_free_tbl { + struct cmsg_hdr hdr; + __be32 tid; +}; + +struct cmsg_reply_map_free_tbl { + struct cmsg_reply_map_simple reply_hdr; + __be32 count; +}; + +struct cmsg_key_value_pair { + __be32 key[CMSG_MAP_KEY_LW]; + __be32 value[CMSG_MAP_VALUE_LW]; +}; + +struct cmsg_req_map_op { + struct cmsg_hdr hdr; + __be32 tid; + __be32 count; + __be32 flags; + struct cmsg_key_value_pair elem[0]; +}; + +struct cmsg_reply_map_op { + struct cmsg_reply_map_simple reply_hdr; + __be32 count; + __be32 resv; + struct cmsg_key_value_pair elem[0]; +}; +#endif diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c index 995e95410b11..56451edf01c2 100644 --- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c +++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2016 Netronome Systems, Inc. + * Copyright (C) 2016-2017 Netronome Systems, Inc. * * This software is dual licensed under the GNU General License Version 2, * June 1991 as shown in the file COPYING in the top-level directory of this @@ -33,6 +33,7 @@ #define pr_fmt(fmt) "NFP net bpf: " fmt +#include <linux/bug.h> #include <linux/kernel.h> #include <linux/bpf.h> #include <linux/filter.h> @@ -66,12 +67,6 @@ next2 = nfp_meta_next(next)) static bool -nfp_meta_has_next(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) -{ - return meta->l.next != &nfp_prog->insns; -} - -static bool nfp_meta_has_prev(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) { return meta->l.prev != &nfp_prog->insns; @@ -90,19 +85,25 @@ static void nfp_prog_push(struct nfp_prog *nfp_prog, u64 insn) static unsigned int nfp_prog_current_offset(struct nfp_prog *nfp_prog) { - return nfp_prog->start_off + nfp_prog->prog_len; + return nfp_prog->prog_len; } -static unsigned int -nfp_prog_offset_to_index(struct nfp_prog *nfp_prog, unsigned int offset) +static bool +nfp_prog_confirm_current_offset(struct nfp_prog *nfp_prog, unsigned int off) { - return offset - nfp_prog->start_off; + /* If there is a recorded error we may have dropped instructions; + * that doesn't have to be due to translator bug, and the translation + * will fail anyway, so just return OK. + */ + if (nfp_prog->error) + return true; + return !WARN_ON_ONCE(nfp_prog_current_offset(nfp_prog) != off); } /* --- Emitters --- */ static void __emit_cmd(struct nfp_prog *nfp_prog, enum cmd_tgt_map op, - u8 mode, u8 xfer, u8 areg, u8 breg, u8 size, bool sync) + u8 mode, u8 xfer, u8 areg, u8 breg, u8 size, bool sync, bool indir) { enum cmd_ctx_swap ctx; u64 insn; @@ -120,14 +121,15 @@ __emit_cmd(struct nfp_prog *nfp_prog, enum cmd_tgt_map op, FIELD_PREP(OP_CMD_CNT, size) | FIELD_PREP(OP_CMD_SIG, sync) | FIELD_PREP(OP_CMD_TGT_CMD, cmd_tgt_act[op].tgt_cmd) | + FIELD_PREP(OP_CMD_INDIR, indir) | FIELD_PREP(OP_CMD_MODE, mode); nfp_prog_push(nfp_prog, insn); } static void -emit_cmd(struct nfp_prog *nfp_prog, enum cmd_tgt_map op, - u8 mode, u8 xfer, swreg lreg, swreg rreg, u8 size, bool sync) +emit_cmd_any(struct nfp_prog *nfp_prog, enum cmd_tgt_map op, u8 mode, u8 xfer, + swreg lreg, swreg rreg, u8 size, bool sync, bool indir) { struct nfp_insn_re_regs reg; int err; @@ -148,7 +150,22 @@ emit_cmd(struct nfp_prog *nfp_prog, enum cmd_tgt_map op, return; } - __emit_cmd(nfp_prog, op, mode, xfer, reg.areg, reg.breg, size, sync); + __emit_cmd(nfp_prog, op, mode, xfer, reg.areg, reg.breg, size, sync, + indir); +} + +static void +emit_cmd(struct nfp_prog *nfp_prog, enum cmd_tgt_map op, u8 mode, u8 xfer, + swreg lreg, swreg rreg, u8 size, bool sync) +{ + emit_cmd_any(nfp_prog, op, mode, xfer, lreg, rreg, size, sync, false); +} + +static void +emit_cmd_indir(struct nfp_prog *nfp_prog, enum cmd_tgt_map op, u8 mode, u8 xfer, + swreg lreg, swreg rreg, u8 size, bool sync) +{ + emit_cmd_any(nfp_prog, op, mode, xfer, lreg, rreg, size, sync, true); } static void @@ -172,22 +189,28 @@ __emit_br(struct nfp_prog *nfp_prog, enum br_mask mask, enum br_ev_pip ev_pip, nfp_prog_push(nfp_prog, insn); } -static void emit_br_def(struct nfp_prog *nfp_prog, u16 addr, u8 defer) +static void +emit_br_relo(struct nfp_prog *nfp_prog, enum br_mask mask, u16 addr, u8 defer, + enum nfp_relo_type relo) { - if (defer > 2) { + if (mask == BR_UNC && defer > 2) { pr_err("BUG: branch defer out of bounds %d\n", defer); nfp_prog->error = -EFAULT; return; } - __emit_br(nfp_prog, BR_UNC, BR_EV_PIP_UNCOND, BR_CSS_NONE, addr, defer); + + __emit_br(nfp_prog, mask, + mask != BR_UNC ? BR_EV_PIP_COND : BR_EV_PIP_UNCOND, + BR_CSS_NONE, addr, defer); + + nfp_prog->prog[nfp_prog->prog_len - 1] |= + FIELD_PREP(OP_RELO_TYPE, relo); } static void emit_br(struct nfp_prog *nfp_prog, enum br_mask mask, u16 addr, u8 defer) { - __emit_br(nfp_prog, mask, - mask != BR_UNC ? BR_EV_PIP_COND : BR_EV_PIP_UNCOND, - BR_CSS_NONE, addr, defer); + emit_br_relo(nfp_prog, mask, addr, defer, RELO_BR_REL); } static void @@ -230,9 +253,11 @@ emit_immed(struct nfp_prog *nfp_prog, swreg dst, u16 imm, return; } - __emit_immed(nfp_prog, reg.areg, reg.breg, imm >> 8, width, - invert, shift, reg.wr_both, - reg.dst_lmextn, reg.src_lmextn); + /* Use reg.dst when destination is No-Dest. */ + __emit_immed(nfp_prog, + swreg_type(dst) == NN_REG_NONE ? reg.dst : reg.areg, + reg.breg, imm >> 8, width, invert, shift, + reg.wr_both, reg.dst_lmextn, reg.src_lmextn); } static void @@ -458,6 +483,21 @@ static void wrp_immed(struct nfp_prog *nfp_prog, swreg dst, u32 imm) } } +static void +wrp_immed_relo(struct nfp_prog *nfp_prog, swreg dst, u32 imm, + enum nfp_relo_type relo) +{ + if (imm > 0xffff) { + pr_err("relocation of a large immediate!\n"); + nfp_prog->error = -EFAULT; + return; + } + emit_immed(nfp_prog, dst, imm, IMMED_WIDTH_ALL, false, IMMED_SHIFT_0B); + + nfp_prog->prog[nfp_prog->prog_len - 1] |= + FIELD_PREP(OP_RELO_TYPE, relo); +} + /* ur_load_imm_any() - encode immediate or use tmp register (unrestricted) * If the @imm is small enough encode it directly in operand and return * otherwise load @imm to a spare register and return its encoding. @@ -490,24 +530,179 @@ static void wrp_nops(struct nfp_prog *nfp_prog, unsigned int count) emit_nop(nfp_prog); } +static void wrp_mov(struct nfp_prog *nfp_prog, swreg dst, swreg src) +{ + emit_alu(nfp_prog, dst, reg_none(), ALU_OP_NONE, src); +} + +static void wrp_reg_mov(struct nfp_prog *nfp_prog, u16 dst, u16 src) +{ + wrp_mov(nfp_prog, reg_both(dst), reg_b(src)); +} + +/* wrp_reg_subpart() - load @field_len bytes from @offset of @src, write the + * result to @dst from low end. + */ static void -wrp_br_special(struct nfp_prog *nfp_prog, enum br_mask mask, - enum br_special special) +wrp_reg_subpart(struct nfp_prog *nfp_prog, swreg dst, swreg src, u8 field_len, + u8 offset) { - emit_br(nfp_prog, mask, 0, 0); + enum shf_sc sc = offset ? SHF_SC_R_SHF : SHF_SC_NONE; + u8 mask = (1 << field_len) - 1; - nfp_prog->prog[nfp_prog->prog_len - 1] |= - FIELD_PREP(OP_BR_SPECIAL, special); + emit_ld_field_any(nfp_prog, dst, mask, src, sc, offset * 8, true); } -static void wrp_mov(struct nfp_prog *nfp_prog, swreg dst, swreg src) +static void +addr40_offset(struct nfp_prog *nfp_prog, u8 src_gpr, swreg offset, + swreg *rega, swreg *regb) { - emit_alu(nfp_prog, dst, reg_none(), ALU_OP_NONE, src); + if (offset == reg_imm(0)) { + *rega = reg_a(src_gpr); + *regb = reg_b(src_gpr + 1); + return; + } + + emit_alu(nfp_prog, imm_a(nfp_prog), reg_a(src_gpr), ALU_OP_ADD, offset); + emit_alu(nfp_prog, imm_b(nfp_prog), reg_b(src_gpr + 1), ALU_OP_ADD_C, + reg_imm(0)); + *rega = imm_a(nfp_prog); + *regb = imm_b(nfp_prog); } -static void wrp_reg_mov(struct nfp_prog *nfp_prog, u16 dst, u16 src) +/* NFP has Command Push Pull bus which supports bluk memory operations. */ +static int nfp_cpp_memcpy(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) { - wrp_mov(nfp_prog, reg_both(dst), reg_b(src)); + bool descending_seq = meta->ldst_gather_len < 0; + s16 len = abs(meta->ldst_gather_len); + swreg src_base, off; + bool src_40bit_addr; + unsigned int i; + u8 xfer_num; + + off = re_load_imm_any(nfp_prog, meta->insn.off, imm_b(nfp_prog)); + src_40bit_addr = meta->ptr.type == PTR_TO_MAP_VALUE; + src_base = reg_a(meta->insn.src_reg * 2); + xfer_num = round_up(len, 4) / 4; + + if (src_40bit_addr) + addr40_offset(nfp_prog, meta->insn.src_reg, off, &src_base, + &off); + + /* Setup PREV_ALU fields to override memory read length. */ + if (len > 32) + wrp_immed(nfp_prog, reg_none(), + CMD_OVE_LEN | FIELD_PREP(CMD_OV_LEN, xfer_num - 1)); + + /* Memory read from source addr into transfer-in registers. */ + emit_cmd_any(nfp_prog, CMD_TGT_READ32_SWAP, + src_40bit_addr ? CMD_MODE_40b_BA : CMD_MODE_32b, 0, + src_base, off, xfer_num - 1, true, len > 32); + + /* Move from transfer-in to transfer-out. */ + for (i = 0; i < xfer_num; i++) + wrp_mov(nfp_prog, reg_xfer(i), reg_xfer(i)); + + off = re_load_imm_any(nfp_prog, meta->paired_st->off, imm_b(nfp_prog)); + + if (len <= 8) { + /* Use single direct_ref write8. */ + emit_cmd(nfp_prog, CMD_TGT_WRITE8_SWAP, CMD_MODE_32b, 0, + reg_a(meta->paired_st->dst_reg * 2), off, len - 1, + true); + } else if (len <= 32 && IS_ALIGNED(len, 4)) { + /* Use single direct_ref write32. */ + emit_cmd(nfp_prog, CMD_TGT_WRITE32_SWAP, CMD_MODE_32b, 0, + reg_a(meta->paired_st->dst_reg * 2), off, xfer_num - 1, + true); + } else if (len <= 32) { + /* Use single indirect_ref write8. */ + wrp_immed(nfp_prog, reg_none(), + CMD_OVE_LEN | FIELD_PREP(CMD_OV_LEN, len - 1)); + emit_cmd_indir(nfp_prog, CMD_TGT_WRITE8_SWAP, CMD_MODE_32b, 0, + reg_a(meta->paired_st->dst_reg * 2), off, + len - 1, true); + } else if (IS_ALIGNED(len, 4)) { + /* Use single indirect_ref write32. */ + wrp_immed(nfp_prog, reg_none(), + CMD_OVE_LEN | FIELD_PREP(CMD_OV_LEN, xfer_num - 1)); + emit_cmd_indir(nfp_prog, CMD_TGT_WRITE32_SWAP, CMD_MODE_32b, 0, + reg_a(meta->paired_st->dst_reg * 2), off, + xfer_num - 1, true); + } else if (len <= 40) { + /* Use one direct_ref write32 to write the first 32-bytes, then + * another direct_ref write8 to write the remaining bytes. + */ + emit_cmd(nfp_prog, CMD_TGT_WRITE32_SWAP, CMD_MODE_32b, 0, + reg_a(meta->paired_st->dst_reg * 2), off, 7, + true); + + off = re_load_imm_any(nfp_prog, meta->paired_st->off + 32, + imm_b(nfp_prog)); + emit_cmd(nfp_prog, CMD_TGT_WRITE8_SWAP, CMD_MODE_32b, 8, + reg_a(meta->paired_st->dst_reg * 2), off, len - 33, + true); + } else { + /* Use one indirect_ref write32 to write 4-bytes aligned length, + * then another direct_ref write8 to write the remaining bytes. + */ + u8 new_off; + + wrp_immed(nfp_prog, reg_none(), + CMD_OVE_LEN | FIELD_PREP(CMD_OV_LEN, xfer_num - 2)); + emit_cmd_indir(nfp_prog, CMD_TGT_WRITE32_SWAP, CMD_MODE_32b, 0, + reg_a(meta->paired_st->dst_reg * 2), off, + xfer_num - 2, true); + new_off = meta->paired_st->off + (xfer_num - 1) * 4; + off = re_load_imm_any(nfp_prog, new_off, imm_b(nfp_prog)); + emit_cmd(nfp_prog, CMD_TGT_WRITE8_SWAP, CMD_MODE_32b, + xfer_num - 1, reg_a(meta->paired_st->dst_reg * 2), off, + (len & 0x3) - 1, true); + } + + /* TODO: The following extra load is to make sure data flow be identical + * before and after we do memory copy optimization. + * + * The load destination register is not guaranteed to be dead, so we + * need to make sure it is loaded with the value the same as before + * this transformation. + * + * These extra loads could be removed once we have accurate register + * usage information. + */ + if (descending_seq) + xfer_num = 0; + else if (BPF_SIZE(meta->insn.code) != BPF_DW) + xfer_num = xfer_num - 1; + else + xfer_num = xfer_num - 2; + + switch (BPF_SIZE(meta->insn.code)) { + case BPF_B: + wrp_reg_subpart(nfp_prog, reg_both(meta->insn.dst_reg * 2), + reg_xfer(xfer_num), 1, + IS_ALIGNED(len, 4) ? 3 : (len & 3) - 1); + break; + case BPF_H: + wrp_reg_subpart(nfp_prog, reg_both(meta->insn.dst_reg * 2), + reg_xfer(xfer_num), 2, (len & 3) ^ 2); + break; + case BPF_W: + wrp_mov(nfp_prog, reg_both(meta->insn.dst_reg * 2), + reg_xfer(0)); + break; + case BPF_DW: + wrp_mov(nfp_prog, reg_both(meta->insn.dst_reg * 2), + reg_xfer(xfer_num)); + wrp_mov(nfp_prog, reg_both(meta->insn.dst_reg * 2 + 1), + reg_xfer(xfer_num + 1)); + break; + } + + if (BPF_SIZE(meta->insn.code) != BPF_DW) + wrp_immed(nfp_prog, reg_both(meta->insn.dst_reg * 2 + 1), 0); + + return 0; } static int @@ -540,20 +735,20 @@ data_ld(struct nfp_prog *nfp_prog, swreg offset, u8 dst_gpr, int size) } static int -data_ld_host_order(struct nfp_prog *nfp_prog, u8 src_gpr, swreg offset, - u8 dst_gpr, int size) +data_ld_host_order(struct nfp_prog *nfp_prog, u8 dst_gpr, + swreg lreg, swreg rreg, int size, enum cmd_mode mode) { unsigned int i; u8 mask, sz; - /* We load the value from the address indicated in @offset and then + /* We load the value from the address indicated in rreg + lreg and then * mask out the data we don't need. Note: this is little endian! */ sz = max(size, 4); mask = size < 4 ? GENMASK(size - 1, 0) : 0; - emit_cmd(nfp_prog, CMD_TGT_READ32_SWAP, CMD_MODE_32b, 0, - reg_a(src_gpr), offset, sz / 4 - 1, true); + emit_cmd(nfp_prog, CMD_TGT_READ32_SWAP, mode, 0, + lreg, rreg, sz / 4 - 1, true); i = 0; if (mask) @@ -570,6 +765,26 @@ data_ld_host_order(struct nfp_prog *nfp_prog, u8 src_gpr, swreg offset, } static int +data_ld_host_order_addr32(struct nfp_prog *nfp_prog, u8 src_gpr, swreg offset, + u8 dst_gpr, u8 size) +{ + return data_ld_host_order(nfp_prog, dst_gpr, reg_a(src_gpr), offset, + size, CMD_MODE_32b); +} + +static int +data_ld_host_order_addr40(struct nfp_prog *nfp_prog, u8 src_gpr, swreg offset, + u8 dst_gpr, u8 size) +{ + swreg rega, regb; + + addr40_offset(nfp_prog, src_gpr, offset, ®a, ®b); + + return data_ld_host_order(nfp_prog, dst_gpr, rega, regb, + size, CMD_MODE_40b_BA); +} + +static int construct_data_ind_ld(struct nfp_prog *nfp_prog, u16 offset, u16 src, u8 size) { swreg tmp_reg; @@ -583,7 +798,7 @@ construct_data_ind_ld(struct nfp_prog *nfp_prog, u16 offset, u16 src, u8 size) imm_a(nfp_prog), ALU_OP_ADD, reg_imm(size)); emit_alu(nfp_prog, reg_none(), plen_reg(nfp_prog), ALU_OP_SUB, imm_a(nfp_prog)); - wrp_br_special(nfp_prog, BR_BLO, OP_BR_GO_ABORT); + emit_br_relo(nfp_prog, BR_BLO, BR_OFF_RELO, 0, RELO_BR_GO_ABORT); /* Load data */ return data_ld(nfp_prog, imm_b(nfp_prog), 0, size); @@ -596,7 +811,7 @@ static int construct_data_ld(struct nfp_prog *nfp_prog, u16 offset, u8 size) /* Check packet length */ tmp_reg = ur_load_imm_any(nfp_prog, offset + size, imm_a(nfp_prog)); emit_alu(nfp_prog, reg_none(), plen_reg(nfp_prog), ALU_OP_SUB, tmp_reg); - wrp_br_special(nfp_prog, BR_BLO, OP_BR_GO_ABORT); + emit_br_relo(nfp_prog, BR_BLO, BR_OFF_RELO, 0, RELO_BR_GO_ABORT); /* Load data */ tmp_reg = re_load_imm_any(nfp_prog, offset, imm_b(nfp_prog)); @@ -975,9 +1190,6 @@ wrp_test_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, { const struct bpf_insn *insn = &meta->insn; - if (insn->off < 0) /* TODO */ - return -EOPNOTSUPP; - wrp_test_reg_one(nfp_prog, insn->dst_reg * 2, alu_op, insn->src_reg * 2, br_mask, insn->off); wrp_test_reg_one(nfp_prog, insn->dst_reg * 2 + 1, alu_op, @@ -995,9 +1207,6 @@ wrp_cmp_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, u8 reg = insn->dst_reg * 2; swreg tmp_reg; - if (insn->off < 0) /* TODO */ - return -EOPNOTSUPP; - tmp_reg = ur_load_imm_any(nfp_prog, imm & ~0U, imm_b(nfp_prog)); if (!swap) emit_alu(nfp_prog, reg_none(), reg_a(reg), ALU_OP_SUB, tmp_reg); @@ -1027,9 +1236,6 @@ wrp_cmp_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, areg = insn->dst_reg * 2; breg = insn->src_reg * 2; - if (insn->off < 0) /* TODO */ - return -EOPNOTSUPP; - if (swap) { areg ^= breg; breg ^= areg; @@ -1052,6 +1258,136 @@ static void wrp_end32(struct nfp_prog *nfp_prog, swreg reg_in, u8 gpr_out) SHF_SC_R_ROT, 16); } +static int adjust_head(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + swreg tmp = imm_a(nfp_prog), tmp_len = imm_b(nfp_prog); + struct nfp_bpf_cap_adjust_head *adjust_head; + u32 ret_einval, end; + + adjust_head = &nfp_prog->bpf->adjust_head; + + /* Optimized version - 5 vs 14 cycles */ + if (nfp_prog->adjust_head_location != UINT_MAX) { + if (WARN_ON_ONCE(nfp_prog->adjust_head_location != meta->n)) + return -EINVAL; + + emit_alu(nfp_prog, pptr_reg(nfp_prog), + reg_a(2 * 2), ALU_OP_ADD, pptr_reg(nfp_prog)); + emit_alu(nfp_prog, plen_reg(nfp_prog), + plen_reg(nfp_prog), ALU_OP_SUB, reg_a(2 * 2)); + emit_alu(nfp_prog, pv_len(nfp_prog), + pv_len(nfp_prog), ALU_OP_SUB, reg_a(2 * 2)); + + wrp_immed(nfp_prog, reg_both(0), 0); + wrp_immed(nfp_prog, reg_both(1), 0); + + /* TODO: when adjust head is guaranteed to succeed we can + * also eliminate the following if (r0 == 0) branch. + */ + + return 0; + } + + ret_einval = nfp_prog_current_offset(nfp_prog) + 14; + end = ret_einval + 2; + + /* We need to use a temp because offset is just a part of the pkt ptr */ + emit_alu(nfp_prog, tmp, + reg_a(2 * 2), ALU_OP_ADD_2B, pptr_reg(nfp_prog)); + + /* Validate result will fit within FW datapath constraints */ + emit_alu(nfp_prog, reg_none(), + tmp, ALU_OP_SUB, reg_imm(adjust_head->off_min)); + emit_br(nfp_prog, BR_BLO, ret_einval, 0); + emit_alu(nfp_prog, reg_none(), + reg_imm(adjust_head->off_max), ALU_OP_SUB, tmp); + emit_br(nfp_prog, BR_BLO, ret_einval, 0); + + /* Validate the length is at least ETH_HLEN */ + emit_alu(nfp_prog, tmp_len, + plen_reg(nfp_prog), ALU_OP_SUB, reg_a(2 * 2)); + emit_alu(nfp_prog, reg_none(), + tmp_len, ALU_OP_SUB, reg_imm(ETH_HLEN)); + emit_br(nfp_prog, BR_BMI, ret_einval, 0); + + /* Load the ret code */ + wrp_immed(nfp_prog, reg_both(0), 0); + wrp_immed(nfp_prog, reg_both(1), 0); + + /* Modify the packet metadata */ + emit_ld_field(nfp_prog, pptr_reg(nfp_prog), 0x3, tmp, SHF_SC_NONE, 0); + + /* Skip over the -EINVAL ret code (defer 2) */ + emit_br(nfp_prog, BR_UNC, end, 2); + + emit_alu(nfp_prog, plen_reg(nfp_prog), + plen_reg(nfp_prog), ALU_OP_SUB, reg_a(2 * 2)); + emit_alu(nfp_prog, pv_len(nfp_prog), + pv_len(nfp_prog), ALU_OP_SUB, reg_a(2 * 2)); + + /* return -EINVAL target */ + if (!nfp_prog_confirm_current_offset(nfp_prog, ret_einval)) + return -EINVAL; + + wrp_immed(nfp_prog, reg_both(0), -22); + wrp_immed(nfp_prog, reg_both(1), ~0); + + if (!nfp_prog_confirm_current_offset(nfp_prog, end)) + return -EINVAL; + + return 0; +} + +static int +map_lookup_stack(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + struct bpf_offloaded_map *offmap; + struct nfp_bpf_map *nfp_map; + bool load_lm_ptr; + u32 ret_tgt; + s64 lm_off; + swreg tid; + + offmap = (struct bpf_offloaded_map *)meta->arg1.map_ptr; + nfp_map = offmap->dev_priv; + + /* We only have to reload LM0 if the key is not at start of stack */ + lm_off = nfp_prog->stack_depth; + lm_off += meta->arg2.var_off.value + meta->arg2.off; + load_lm_ptr = meta->arg2_var_off || lm_off; + + /* Set LM0 to start of key */ + if (load_lm_ptr) + emit_csr_wr(nfp_prog, reg_b(2 * 2), NFP_CSR_ACT_LM_ADDR0); + + /* Load map ID into a register, it should actually fit as an immediate + * but in case it doesn't deal with it here, not in the delay slots. + */ + tid = ur_load_imm_any(nfp_prog, nfp_map->tid, imm_a(nfp_prog)); + + emit_br_relo(nfp_prog, BR_UNC, BR_OFF_RELO + BPF_FUNC_map_lookup_elem, + 2, RELO_BR_HELPER); + ret_tgt = nfp_prog_current_offset(nfp_prog) + 2; + + /* Load map ID into A0 */ + wrp_mov(nfp_prog, reg_a(0), tid); + + /* Load the return address into B0 */ + wrp_immed_relo(nfp_prog, reg_b(0), ret_tgt, RELO_IMMED_REL); + + if (!nfp_prog_confirm_current_offset(nfp_prog, ret_tgt)) + return -EINVAL; + + /* Reset the LM0 pointer */ + if (!load_lm_ptr) + return 0; + + emit_csr_wr(nfp_prog, stack_reg(nfp_prog), NFP_CSR_ACT_LM_ADDR0); + wrp_nops(nfp_prog, 3); + + return 0; +} + /* --- Callbacks --- */ static int mov_reg64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) { @@ -1486,14 +1822,29 @@ mem_ldx_data(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, tmp_reg = re_load_imm_any(nfp_prog, meta->insn.off, imm_b(nfp_prog)); - return data_ld_host_order(nfp_prog, meta->insn.src_reg * 2, tmp_reg, - meta->insn.dst_reg * 2, size); + return data_ld_host_order_addr32(nfp_prog, meta->insn.src_reg * 2, + tmp_reg, meta->insn.dst_reg * 2, size); +} + +static int +mem_ldx_emem(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, + unsigned int size) +{ + swreg tmp_reg; + + tmp_reg = re_load_imm_any(nfp_prog, meta->insn.off, imm_b(nfp_prog)); + + return data_ld_host_order_addr40(nfp_prog, meta->insn.src_reg * 2, + tmp_reg, meta->insn.dst_reg * 2, size); } static int mem_ldx(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, unsigned int size) { + if (meta->ldst_gather_len) + return nfp_cpp_memcpy(nfp_prog, meta); + if (meta->ptr.type == PTR_TO_CTX) { if (nfp_prog->type == BPF_PROG_TYPE_XDP) return mem_ldx_xdp(nfp_prog, meta, size); @@ -1508,6 +1859,9 @@ mem_ldx(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, return mem_ldx_stack(nfp_prog, meta, size, meta->ptr.off + meta->ptr.var_off.value); + if (meta->ptr.type == PTR_TO_MAP_VALUE) + return mem_ldx_emem(nfp_prog, meta, size); + return -EOPNOTSUPP; } @@ -1630,8 +1984,6 @@ static int mem_stx8(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) static int jump(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) { - if (meta->insn.off < 0) /* TODO */ - return -EOPNOTSUPP; emit_br(nfp_prog, BR_UNC, meta->insn.off, 0); return 0; @@ -1646,9 +1998,6 @@ static int jeq_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) or1 = reg_a(insn->dst_reg * 2); or2 = reg_b(insn->dst_reg * 2 + 1); - if (insn->off < 0) /* TODO */ - return -EOPNOTSUPP; - if (imm & ~0U) { tmp_reg = ur_load_imm_any(nfp_prog, imm & ~0U, imm_b(nfp_prog)); emit_alu(nfp_prog, imm_a(nfp_prog), @@ -1689,15 +2038,32 @@ static int jle_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) return wrp_cmp_imm(nfp_prog, meta, BR_BHS, true); } +static int jsgt_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + return wrp_cmp_imm(nfp_prog, meta, BR_BLT, true); +} + +static int jsge_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + return wrp_cmp_imm(nfp_prog, meta, BR_BGE, false); +} + +static int jslt_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + return wrp_cmp_imm(nfp_prog, meta, BR_BLT, false); +} + +static int jsle_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + return wrp_cmp_imm(nfp_prog, meta, BR_BGE, true); +} + static int jset_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) { const struct bpf_insn *insn = &meta->insn; u64 imm = insn->imm; /* sign extend */ swreg tmp_reg; - if (insn->off < 0) /* TODO */ - return -EOPNOTSUPP; - if (!imm) { meta->skip = true; return 0; @@ -1726,9 +2092,6 @@ static int jne_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) u64 imm = insn->imm; /* sign extend */ swreg tmp_reg; - if (insn->off < 0) /* TODO */ - return -EOPNOTSUPP; - if (!imm) { emit_alu(nfp_prog, reg_none(), reg_a(insn->dst_reg * 2), ALU_OP_OR, reg_b(insn->dst_reg * 2 + 1)); @@ -1753,9 +2116,6 @@ static int jeq_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) { const struct bpf_insn *insn = &meta->insn; - if (insn->off < 0) /* TODO */ - return -EOPNOTSUPP; - emit_alu(nfp_prog, imm_a(nfp_prog), reg_a(insn->dst_reg * 2), ALU_OP_XOR, reg_b(insn->src_reg * 2)); emit_alu(nfp_prog, imm_b(nfp_prog), reg_a(insn->dst_reg * 2 + 1), @@ -1787,6 +2147,26 @@ static int jle_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) return wrp_cmp_reg(nfp_prog, meta, BR_BHS, true); } +static int jsgt_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + return wrp_cmp_reg(nfp_prog, meta, BR_BLT, true); +} + +static int jsge_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + return wrp_cmp_reg(nfp_prog, meta, BR_BGE, false); +} + +static int jslt_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + return wrp_cmp_reg(nfp_prog, meta, BR_BLT, false); +} + +static int jsle_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + return wrp_cmp_reg(nfp_prog, meta, BR_BGE, true); +} + static int jset_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) { return wrp_test_reg(nfp_prog, meta, ALU_OP_AND, BR_BNE); @@ -1797,9 +2177,22 @@ static int jne_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) return wrp_test_reg(nfp_prog, meta, ALU_OP_XOR, BR_BNE); } +static int call(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) +{ + switch (meta->insn.imm) { + case BPF_FUNC_xdp_adjust_head: + return adjust_head(nfp_prog, meta); + case BPF_FUNC_map_lookup_elem: + return map_lookup_stack(nfp_prog, meta); + default: + WARN_ONCE(1, "verifier allowed unsupported function\n"); + return -EOPNOTSUPP; + } +} + static int goto_out(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) { - wrp_br_special(nfp_prog, BR_UNC, OP_BR_GO_OUT); + emit_br_relo(nfp_prog, BR_UNC, BR_OFF_RELO, 0, RELO_BR_GO_OUT); return 0; } @@ -1860,6 +2253,10 @@ static const instr_cb_t instr_cb[256] = { [BPF_JMP | BPF_JGE | BPF_K] = jge_imm, [BPF_JMP | BPF_JLT | BPF_K] = jlt_imm, [BPF_JMP | BPF_JLE | BPF_K] = jle_imm, + [BPF_JMP | BPF_JSGT | BPF_K] = jsgt_imm, + [BPF_JMP | BPF_JSGE | BPF_K] = jsge_imm, + [BPF_JMP | BPF_JSLT | BPF_K] = jslt_imm, + [BPF_JMP | BPF_JSLE | BPF_K] = jsle_imm, [BPF_JMP | BPF_JSET | BPF_K] = jset_imm, [BPF_JMP | BPF_JNE | BPF_K] = jne_imm, [BPF_JMP | BPF_JEQ | BPF_X] = jeq_reg, @@ -1867,99 +2264,64 @@ static const instr_cb_t instr_cb[256] = { [BPF_JMP | BPF_JGE | BPF_X] = jge_reg, [BPF_JMP | BPF_JLT | BPF_X] = jlt_reg, [BPF_JMP | BPF_JLE | BPF_X] = jle_reg, + [BPF_JMP | BPF_JSGT | BPF_X] = jsgt_reg, + [BPF_JMP | BPF_JSGE | BPF_X] = jsge_reg, + [BPF_JMP | BPF_JSLT | BPF_X] = jslt_reg, + [BPF_JMP | BPF_JSLE | BPF_X] = jsle_reg, [BPF_JMP | BPF_JSET | BPF_X] = jset_reg, [BPF_JMP | BPF_JNE | BPF_X] = jne_reg, + [BPF_JMP | BPF_CALL] = call, [BPF_JMP | BPF_EXIT] = goto_out, }; -/* --- Misc code --- */ -static void br_set_offset(u64 *instr, u16 offset) -{ - u16 addr_lo, addr_hi; - - addr_lo = offset & (OP_BR_ADDR_LO >> __bf_shf(OP_BR_ADDR_LO)); - addr_hi = offset != addr_lo; - *instr &= ~(OP_BR_ADDR_HI | OP_BR_ADDR_LO); - *instr |= FIELD_PREP(OP_BR_ADDR_HI, addr_hi); - *instr |= FIELD_PREP(OP_BR_ADDR_LO, addr_lo); -} - /* --- Assembler logic --- */ static int nfp_fixup_branches(struct nfp_prog *nfp_prog) { - struct nfp_insn_meta *meta, *next; - u32 off, br_idx; - u32 idx; + struct nfp_insn_meta *meta, *jmp_dst; + u32 idx, br_idx; - nfp_for_each_insn_walk2(nfp_prog, meta, next) { + list_for_each_entry(meta, &nfp_prog->insns, l) { if (meta->skip) continue; + if (meta->insn.code == (BPF_JMP | BPF_CALL)) + continue; if (BPF_CLASS(meta->insn.code) != BPF_JMP) continue; - br_idx = nfp_prog_offset_to_index(nfp_prog, next->off) - 1; + if (list_is_last(&meta->l, &nfp_prog->insns)) + br_idx = nfp_prog->last_bpf_off; + else + br_idx = list_next_entry(meta, l)->off - 1; + if (!nfp_is_br(nfp_prog->prog[br_idx])) { pr_err("Fixup found block not ending in branch %d %02x %016llx!!\n", br_idx, meta->insn.code, nfp_prog->prog[br_idx]); return -ELOOP; } /* Leave special branches for later */ - if (FIELD_GET(OP_BR_SPECIAL, nfp_prog->prog[br_idx])) + if (FIELD_GET(OP_RELO_TYPE, nfp_prog->prog[br_idx]) != + RELO_BR_REL) continue; - /* Find the target offset in assembler realm */ - off = meta->insn.off; - if (!off) { - pr_err("Fixup found zero offset!!\n"); + if (!meta->jmp_dst) { + pr_err("Non-exit jump doesn't have destination info recorded!!\n"); return -ELOOP; } - while (off && nfp_meta_has_next(nfp_prog, next)) { - next = nfp_meta_next(next); - off--; - } - if (off) { - pr_err("Fixup found too large jump!! %d\n", off); - return -ELOOP; - } + jmp_dst = meta->jmp_dst; - if (next->skip) { + if (jmp_dst->skip) { pr_err("Branch landing on removed instruction!!\n"); return -ELOOP; } - for (idx = nfp_prog_offset_to_index(nfp_prog, meta->off); - idx <= br_idx; idx++) { + for (idx = meta->off; idx <= br_idx; idx++) { if (!nfp_is_br(nfp_prog->prog[idx])) continue; - br_set_offset(&nfp_prog->prog[idx], next->off); + br_set_offset(&nfp_prog->prog[idx], jmp_dst->off); } } - /* Fixup 'goto out's separately, they can be scattered around */ - for (br_idx = 0; br_idx < nfp_prog->prog_len; br_idx++) { - enum br_special special; - - if ((nfp_prog->prog[br_idx] & OP_BR_BASE_MASK) != OP_BR_BASE) - continue; - - special = FIELD_GET(OP_BR_SPECIAL, nfp_prog->prog[br_idx]); - switch (special) { - case OP_BR_NORMAL: - break; - case OP_BR_GO_OUT: - br_set_offset(&nfp_prog->prog[br_idx], - nfp_prog->tgt_out); - break; - case OP_BR_GO_ABORT: - br_set_offset(&nfp_prog->prog[br_idx], - nfp_prog->tgt_abort); - break; - } - - nfp_prog->prog[br_idx] &= ~OP_BR_SPECIAL; - } - return 0; } @@ -1987,7 +2349,7 @@ static void nfp_outro_tc_da(struct nfp_prog *nfp_prog) /* Target for aborts */ nfp_prog->tgt_abort = nfp_prog_current_offset(nfp_prog); - emit_br_def(nfp_prog, nfp_prog->tgt_done, 2); + emit_br_relo(nfp_prog, BR_UNC, BR_OFF_RELO, 2, RELO_BR_NEXT_PKT); wrp_mov(nfp_prog, reg_a(0), NFP_BPF_ABI_FLAGS); emit_ld_field(nfp_prog, reg_a(0), 0xc, reg_imm(0x11), SHF_SC_L_SHF, 16); @@ -2014,7 +2376,7 @@ static void nfp_outro_tc_da(struct nfp_prog *nfp_prog) emit_shf(nfp_prog, reg_b(2), reg_imm(0xf), SHF_OP_AND, reg_b(3), SHF_SC_R_SHF, 0); - emit_br_def(nfp_prog, nfp_prog->tgt_done, 2); + emit_br_relo(nfp_prog, BR_UNC, BR_OFF_RELO, 2, RELO_BR_NEXT_PKT); emit_shf(nfp_prog, reg_b(2), reg_a(2), SHF_OP_OR, reg_b(2), SHF_SC_L_SHF, 4); @@ -2033,7 +2395,7 @@ static void nfp_outro_xdp(struct nfp_prog *nfp_prog) /* Target for aborts */ nfp_prog->tgt_abort = nfp_prog_current_offset(nfp_prog); - emit_br_def(nfp_prog, nfp_prog->tgt_done, 2); + emit_br_relo(nfp_prog, BR_UNC, BR_OFF_RELO, 2, RELO_BR_NEXT_PKT); wrp_mov(nfp_prog, reg_a(0), NFP_BPF_ABI_FLAGS); emit_ld_field(nfp_prog, reg_a(0), 0xc, reg_imm(0x82), SHF_SC_L_SHF, 16); @@ -2054,7 +2416,7 @@ static void nfp_outro_xdp(struct nfp_prog *nfp_prog) emit_shf(nfp_prog, reg_b(2), reg_imm(0xff), SHF_OP_AND, reg_b(2), SHF_SC_R_SHF, 0); - emit_br_def(nfp_prog, nfp_prog->tgt_done, 2); + emit_br_relo(nfp_prog, BR_UNC, BR_OFF_RELO, 2, RELO_BR_NEXT_PKT); wrp_mov(nfp_prog, reg_a(0), NFP_BPF_ABI_FLAGS); emit_ld_field(nfp_prog, reg_a(0), 0xc, reg_b(2), SHF_SC_L_SHF, 16); @@ -2105,6 +2467,8 @@ static int nfp_translate(struct nfp_prog *nfp_prog) nfp_prog->n_translated++; } + nfp_prog->last_bpf_off = nfp_prog_current_offset(nfp_prog) - 1; + nfp_outro(nfp_prog); if (nfp_prog->error) return nfp_prog->error; @@ -2173,6 +2537,9 @@ static void nfp_bpf_opt_ld_mask(struct nfp_prog *nfp_prog) if (next.src_reg || next.dst_reg) continue; + if (meta2->flags & FLAG_INSN_IS_JUMP_DST) + continue; + meta2->skip = true; } } @@ -2209,40 +2576,294 @@ static void nfp_bpf_opt_ld_shift(struct nfp_prog *nfp_prog) if (next1.imm != 0x20 || next2.imm != 0x20) continue; + if (meta2->flags & FLAG_INSN_IS_JUMP_DST || + meta3->flags & FLAG_INSN_IS_JUMP_DST) + continue; + meta2->skip = true; meta3->skip = true; } } +/* load/store pair that forms memory copy sould look like the following: + * + * ld_width R, [addr_src + offset_src] + * st_width [addr_dest + offset_dest], R + * + * The destination register of load and source register of store should + * be the same, load and store should also perform at the same width. + * If either of addr_src or addr_dest is stack pointer, we don't do the + * CPP optimization as stack is modelled by registers on NFP. + */ +static bool +curr_pair_is_memcpy(struct nfp_insn_meta *ld_meta, + struct nfp_insn_meta *st_meta) +{ + struct bpf_insn *ld = &ld_meta->insn; + struct bpf_insn *st = &st_meta->insn; + + if (!is_mbpf_load(ld_meta) || !is_mbpf_store(st_meta)) + return false; + + if (ld_meta->ptr.type != PTR_TO_PACKET) + return false; + + if (st_meta->ptr.type != PTR_TO_PACKET) + return false; + + if (BPF_SIZE(ld->code) != BPF_SIZE(st->code)) + return false; + + if (ld->dst_reg != st->src_reg) + return false; + + /* There is jump to the store insn in this pair. */ + if (st_meta->flags & FLAG_INSN_IS_JUMP_DST) + return false; + + return true; +} + +/* Currently, we only support chaining load/store pairs if: + * + * - Their address base registers are the same. + * - Their address offsets are in the same order. + * - They operate at the same memory width. + * - There is no jump into the middle of them. + */ +static bool +curr_pair_chain_with_previous(struct nfp_insn_meta *ld_meta, + struct nfp_insn_meta *st_meta, + struct bpf_insn *prev_ld, + struct bpf_insn *prev_st) +{ + u8 prev_size, curr_size, prev_ld_base, prev_st_base, prev_ld_dst; + struct bpf_insn *ld = &ld_meta->insn; + struct bpf_insn *st = &st_meta->insn; + s16 prev_ld_off, prev_st_off; + + /* This pair is the start pair. */ + if (!prev_ld) + return true; + + prev_size = BPF_LDST_BYTES(prev_ld); + curr_size = BPF_LDST_BYTES(ld); + prev_ld_base = prev_ld->src_reg; + prev_st_base = prev_st->dst_reg; + prev_ld_dst = prev_ld->dst_reg; + prev_ld_off = prev_ld->off; + prev_st_off = prev_st->off; + + if (ld->dst_reg != prev_ld_dst) + return false; + + if (ld->src_reg != prev_ld_base || st->dst_reg != prev_st_base) + return false; + + if (curr_size != prev_size) + return false; + + /* There is jump to the head of this pair. */ + if (ld_meta->flags & FLAG_INSN_IS_JUMP_DST) + return false; + + /* Both in ascending order. */ + if (prev_ld_off + prev_size == ld->off && + prev_st_off + prev_size == st->off) + return true; + + /* Both in descending order. */ + if (ld->off + curr_size == prev_ld_off && + st->off + curr_size == prev_st_off) + return true; + + return false; +} + +/* Return TRUE if cross memory access happens. Cross memory access means + * store area is overlapping with load area that a later load might load + * the value from previous store, for this case we can't treat the sequence + * as an memory copy. + */ +static bool +cross_mem_access(struct bpf_insn *ld, struct nfp_insn_meta *head_ld_meta, + struct nfp_insn_meta *head_st_meta) +{ + s16 head_ld_off, head_st_off, ld_off; + + /* Different pointer types does not overlap. */ + if (head_ld_meta->ptr.type != head_st_meta->ptr.type) + return false; + + /* load and store are both PTR_TO_PACKET, check ID info. */ + if (head_ld_meta->ptr.id != head_st_meta->ptr.id) + return true; + + /* Canonicalize the offsets. Turn all of them against the original + * base register. + */ + head_ld_off = head_ld_meta->insn.off + head_ld_meta->ptr.off; + head_st_off = head_st_meta->insn.off + head_st_meta->ptr.off; + ld_off = ld->off + head_ld_meta->ptr.off; + + /* Ascending order cross. */ + if (ld_off > head_ld_off && + head_ld_off < head_st_off && ld_off >= head_st_off) + return true; + + /* Descending order cross. */ + if (ld_off < head_ld_off && + head_ld_off > head_st_off && ld_off <= head_st_off) + return true; + + return false; +} + +/* This pass try to identify the following instructoin sequences. + * + * load R, [regA + offA] + * store [regB + offB], R + * load R, [regA + offA + const_imm_A] + * store [regB + offB + const_imm_A], R + * load R, [regA + offA + 2 * const_imm_A] + * store [regB + offB + 2 * const_imm_A], R + * ... + * + * Above sequence is typically generated by compiler when lowering + * memcpy. NFP prefer using CPP instructions to accelerate it. + */ +static void nfp_bpf_opt_ldst_gather(struct nfp_prog *nfp_prog) +{ + struct nfp_insn_meta *head_ld_meta = NULL; + struct nfp_insn_meta *head_st_meta = NULL; + struct nfp_insn_meta *meta1, *meta2; + struct bpf_insn *prev_ld = NULL; + struct bpf_insn *prev_st = NULL; + u8 count = 0; + + nfp_for_each_insn_walk2(nfp_prog, meta1, meta2) { + struct bpf_insn *ld = &meta1->insn; + struct bpf_insn *st = &meta2->insn; + + /* Reset record status if any of the following if true: + * - The current insn pair is not load/store. + * - The load/store pair doesn't chain with previous one. + * - The chained load/store pair crossed with previous pair. + * - The chained load/store pair has a total size of memory + * copy beyond 128 bytes which is the maximum length a + * single NFP CPP command can transfer. + */ + if (!curr_pair_is_memcpy(meta1, meta2) || + !curr_pair_chain_with_previous(meta1, meta2, prev_ld, + prev_st) || + (head_ld_meta && (cross_mem_access(ld, head_ld_meta, + head_st_meta) || + head_ld_meta->ldst_gather_len >= 128))) { + if (!count) + continue; + + if (count > 1) { + s16 prev_ld_off = prev_ld->off; + s16 prev_st_off = prev_st->off; + s16 head_ld_off = head_ld_meta->insn.off; + + if (prev_ld_off < head_ld_off) { + head_ld_meta->insn.off = prev_ld_off; + head_st_meta->insn.off = prev_st_off; + head_ld_meta->ldst_gather_len = + -head_ld_meta->ldst_gather_len; + } + + head_ld_meta->paired_st = &head_st_meta->insn; + head_st_meta->skip = true; + } else { + head_ld_meta->ldst_gather_len = 0; + } + + /* If the chain is ended by an load/store pair then this + * could serve as the new head of the the next chain. + */ + if (curr_pair_is_memcpy(meta1, meta2)) { + head_ld_meta = meta1; + head_st_meta = meta2; + head_ld_meta->ldst_gather_len = + BPF_LDST_BYTES(ld); + meta1 = nfp_meta_next(meta1); + meta2 = nfp_meta_next(meta2); + prev_ld = ld; + prev_st = st; + count = 1; + } else { + head_ld_meta = NULL; + head_st_meta = NULL; + prev_ld = NULL; + prev_st = NULL; + count = 0; + } + + continue; + } + + if (!head_ld_meta) { + head_ld_meta = meta1; + head_st_meta = meta2; + } else { + meta1->skip = true; + meta2->skip = true; + } + + head_ld_meta->ldst_gather_len += BPF_LDST_BYTES(ld); + meta1 = nfp_meta_next(meta1); + meta2 = nfp_meta_next(meta2); + prev_ld = ld; + prev_st = st; + count++; + } +} + static int nfp_bpf_optimize(struct nfp_prog *nfp_prog) { nfp_bpf_opt_reg_init(nfp_prog); nfp_bpf_opt_ld_mask(nfp_prog); nfp_bpf_opt_ld_shift(nfp_prog); + nfp_bpf_opt_ldst_gather(nfp_prog); return 0; } -static int nfp_bpf_ustore_calc(struct nfp_prog *nfp_prog, __le64 *ustore) +static int nfp_bpf_ustore_calc(u64 *prog, unsigned int len) { + __le64 *ustore = (__force __le64 *)prog; int i; - for (i = 0; i < nfp_prog->prog_len; i++) { + for (i = 0; i < len; i++) { int err; - err = nfp_ustore_check_valid_no_ecc(nfp_prog->prog[i]); + err = nfp_ustore_check_valid_no_ecc(prog[i]); if (err) return err; - nfp_prog->prog[i] = nfp_ustore_calc_ecc_insn(nfp_prog->prog[i]); - - ustore[i] = cpu_to_le64(nfp_prog->prog[i]); + ustore[i] = cpu_to_le64(nfp_ustore_calc_ecc_insn(prog[i])); } return 0; } +static void nfp_bpf_prog_trim(struct nfp_prog *nfp_prog) +{ + void *prog; + + prog = kvmalloc_array(nfp_prog->prog_len, sizeof(u64), GFP_KERNEL); + if (!prog) + return; + + nfp_prog->__prog_alloc_len = nfp_prog->prog_len * sizeof(u64); + memcpy(prog, nfp_prog->prog, nfp_prog->__prog_alloc_len); + kvfree(nfp_prog->prog); + nfp_prog->prog = prog; +} + int nfp_bpf_jit(struct nfp_prog *nfp_prog) { int ret; @@ -2258,5 +2879,102 @@ int nfp_bpf_jit(struct nfp_prog *nfp_prog) return -EINVAL; } - return nfp_bpf_ustore_calc(nfp_prog, (__force __le64 *)nfp_prog->prog); + nfp_bpf_prog_trim(nfp_prog); + + return ret; +} + +void nfp_bpf_jit_prepare(struct nfp_prog *nfp_prog, unsigned int cnt) +{ + struct nfp_insn_meta *meta; + + /* Another pass to record jump information. */ + list_for_each_entry(meta, &nfp_prog->insns, l) { + u64 code = meta->insn.code; + + if (BPF_CLASS(code) == BPF_JMP && BPF_OP(code) != BPF_EXIT && + BPF_OP(code) != BPF_CALL) { + struct nfp_insn_meta *dst_meta; + unsigned short dst_indx; + + dst_indx = meta->n + 1 + meta->insn.off; + dst_meta = nfp_bpf_goto_meta(nfp_prog, meta, dst_indx, + cnt); + + meta->jmp_dst = dst_meta; + dst_meta->flags |= FLAG_INSN_IS_JUMP_DST; + } + } +} + +bool nfp_bpf_supported_opcode(u8 code) +{ + return !!instr_cb[code]; +} + +void *nfp_bpf_relo_for_vnic(struct nfp_prog *nfp_prog, struct nfp_bpf_vnic *bv) +{ + unsigned int i; + u64 *prog; + int err; + + prog = kmemdup(nfp_prog->prog, nfp_prog->prog_len * sizeof(u64), + GFP_KERNEL); + if (!prog) + return ERR_PTR(-ENOMEM); + + for (i = 0; i < nfp_prog->prog_len; i++) { + enum nfp_relo_type special; + u32 val; + + special = FIELD_GET(OP_RELO_TYPE, prog[i]); + switch (special) { + case RELO_NONE: + continue; + case RELO_BR_REL: + br_add_offset(&prog[i], bv->start_off); + break; + case RELO_BR_GO_OUT: + br_set_offset(&prog[i], + nfp_prog->tgt_out + bv->start_off); + break; + case RELO_BR_GO_ABORT: + br_set_offset(&prog[i], + nfp_prog->tgt_abort + bv->start_off); + break; + case RELO_BR_NEXT_PKT: + br_set_offset(&prog[i], bv->tgt_done); + break; + case RELO_BR_HELPER: + val = br_get_offset(prog[i]); + val -= BR_OFF_RELO; + switch (val) { + case BPF_FUNC_map_lookup_elem: + val = nfp_prog->bpf->helpers.map_lookup; + break; + default: + pr_err("relocation of unknown helper %d\n", + val); + err = -EINVAL; + goto err_free_prog; + } + br_set_offset(&prog[i], val); + break; + case RELO_IMMED_REL: + immed_add_value(&prog[i], bv->start_off); + break; + } + + prog[i] &= ~OP_RELO_TYPE; + } + + err = nfp_bpf_ustore_calc(prog, nfp_prog->prog_len); + if (err) + goto err_free_prog; + + return prog; + +err_free_prog: + kfree(prog); + return ERR_PTR(err); } diff --git a/drivers/net/ethernet/netronome/nfp/bpf/main.c b/drivers/net/ethernet/netronome/nfp/bpf/main.c index 13190aa09faf..322027792fe8 100644 --- a/drivers/net/ethernet/netronome/nfp/bpf/main.c +++ b/drivers/net/ethernet/netronome/nfp/bpf/main.c @@ -34,10 +34,12 @@ #include <net/pkt_cls.h> #include "../nfpcore/nfp_cpp.h" +#include "../nfpcore/nfp_nffw.h" #include "../nfp_app.h" #include "../nfp_main.h" #include "../nfp_net.h" #include "../nfp_port.h" +#include "fw.h" #include "main.h" static bool nfp_net_ebpf_capable(struct nfp_net *nn) @@ -52,7 +54,7 @@ static bool nfp_net_ebpf_capable(struct nfp_net *nn) static int nfp_bpf_xdp_offload(struct nfp_app *app, struct nfp_net *nn, - struct bpf_prog *prog) + struct bpf_prog *prog, struct netlink_ext_ack *extack) { bool running, xdp_running; int ret; @@ -68,10 +70,10 @@ nfp_bpf_xdp_offload(struct nfp_app *app, struct nfp_net *nn, if (prog && running && !xdp_running) return -EBUSY; - ret = nfp_net_bpf_offload(nn, prog, running); + ret = nfp_net_bpf_offload(nn, prog, running, extack); /* Stop offload if replace not possible */ if (ret && prog) - nfp_bpf_xdp_offload(app, nn, NULL); + nfp_bpf_xdp_offload(app, nn, NULL, extack); nn->dp.bpf_offload_xdp = prog && !ret; return ret; @@ -85,16 +87,21 @@ static const char *nfp_bpf_extra_cap(struct nfp_app *app, struct nfp_net *nn) static int nfp_bpf_vnic_alloc(struct nfp_app *app, struct nfp_net *nn, unsigned int id) { + struct nfp_bpf_vnic *bv; int err; - nn->app_priv = kzalloc(sizeof(struct nfp_bpf_vnic), GFP_KERNEL); - if (!nn->app_priv) + bv = kzalloc(sizeof(*bv), GFP_KERNEL); + if (!bv) return -ENOMEM; + nn->app_priv = bv; err = nfp_app_nic_vnic_alloc(app, nn, id); if (err) goto err_free_priv; + bv->start_off = nn_readw(nn, NFP_NET_CFG_BPF_START); + bv->tgt_done = nn_readw(nn, NFP_NET_CFG_BPF_DONE); + return 0; err_free_priv: kfree(nn->app_priv); @@ -105,8 +112,6 @@ static void nfp_bpf_vnic_free(struct nfp_app *app, struct nfp_net *nn) { struct nfp_bpf_vnic *bv = nn->app_priv; - if (nn->dp.bpf_offload_xdp) - nfp_bpf_xdp_offload(app, nn, NULL); WARN_ON(bv->tc_prog); kfree(bv); } @@ -120,17 +125,29 @@ static int nfp_bpf_setup_tc_block_cb(enum tc_setup_type type, struct nfp_bpf_vnic *bv; int err; - if (type != TC_SETUP_CLSBPF || - !tc_can_offload(nn->dp.netdev) || - !nfp_net_ebpf_capable(nn) || - cls_bpf->common.protocol != htons(ETH_P_ALL) || - cls_bpf->common.chain_index) + if (type != TC_SETUP_CLSBPF) { + NL_SET_ERR_MSG_MOD(cls_bpf->common.extack, + "only offload of BPF classifiers supported"); + return -EOPNOTSUPP; + } + if (!tc_cls_can_offload_and_chain0(nn->dp.netdev, &cls_bpf->common)) + return -EOPNOTSUPP; + if (!nfp_net_ebpf_capable(nn)) { + NL_SET_ERR_MSG_MOD(cls_bpf->common.extack, + "NFP firmware does not support eBPF offload"); return -EOPNOTSUPP; + } + if (cls_bpf->common.protocol != htons(ETH_P_ALL)) { + NL_SET_ERR_MSG_MOD(cls_bpf->common.extack, + "only ETH_P_ALL supported as filter protocol"); + return -EOPNOTSUPP; + } /* Only support TC direct action */ if (!cls_bpf->exts_integrated || tcf_exts_has_actions(cls_bpf->exts)) { - nn_err(nn, "only direct action with no legacy actions supported\n"); + NL_SET_ERR_MSG_MOD(cls_bpf->common.extack, + "only direct action with no legacy actions supported"); return -EOPNOTSUPP; } @@ -147,7 +164,8 @@ static int nfp_bpf_setup_tc_block_cb(enum tc_setup_type type, return 0; } - err = nfp_net_bpf_offload(nn, cls_bpf->prog, oldprog); + err = nfp_net_bpf_offload(nn, cls_bpf->prog, oldprog, + cls_bpf->common.extack); if (err) return err; @@ -191,23 +209,215 @@ static int nfp_bpf_setup_tc(struct nfp_app *app, struct net_device *netdev, static bool nfp_bpf_tc_busy(struct nfp_app *app, struct nfp_net *nn) { - return nn->dp.ctrl & NFP_NET_CFG_CTRL_BPF; + struct nfp_bpf_vnic *bv = nn->app_priv; + + return !!bv->tc_prog; +} + +static int +nfp_bpf_change_mtu(struct nfp_app *app, struct net_device *netdev, int new_mtu) +{ + struct nfp_net *nn = netdev_priv(netdev); + unsigned int max_mtu; + + if (~nn->dp.ctrl & NFP_NET_CFG_CTRL_BPF) + return 0; + + max_mtu = nn_readb(nn, NFP_NET_CFG_BPF_INL_MTU) * 64 - 32; + if (new_mtu > max_mtu) { + nn_info(nn, "BPF offload active, MTU over %u not supported\n", + max_mtu); + return -EBUSY; + } + return 0; +} + +static int +nfp_bpf_parse_cap_adjust_head(struct nfp_app_bpf *bpf, void __iomem *value, + u32 length) +{ + struct nfp_bpf_cap_tlv_adjust_head __iomem *cap = value; + struct nfp_cpp *cpp = bpf->app->pf->cpp; + + if (length < sizeof(*cap)) { + nfp_err(cpp, "truncated adjust_head TLV: %d\n", length); + return -EINVAL; + } + + bpf->adjust_head.flags = readl(&cap->flags); + bpf->adjust_head.off_min = readl(&cap->off_min); + bpf->adjust_head.off_max = readl(&cap->off_max); + bpf->adjust_head.guaranteed_sub = readl(&cap->guaranteed_sub); + bpf->adjust_head.guaranteed_add = readl(&cap->guaranteed_add); + + if (bpf->adjust_head.off_min > bpf->adjust_head.off_max) { + nfp_err(cpp, "invalid adjust_head TLV: min > max\n"); + return -EINVAL; + } + if (!FIELD_FIT(UR_REG_IMM_MAX, bpf->adjust_head.off_min) || + !FIELD_FIT(UR_REG_IMM_MAX, bpf->adjust_head.off_max)) { + nfp_warn(cpp, "disabling adjust_head - driver expects min/max to fit in as immediates\n"); + memset(&bpf->adjust_head, 0, sizeof(bpf->adjust_head)); + return 0; + } + + return 0; +} + +static int +nfp_bpf_parse_cap_func(struct nfp_app_bpf *bpf, void __iomem *value, u32 length) +{ + struct nfp_bpf_cap_tlv_func __iomem *cap = value; + + if (length < sizeof(*cap)) { + nfp_err(bpf->app->cpp, "truncated function TLV: %d\n", length); + return -EINVAL; + } + + switch (readl(&cap->func_id)) { + case BPF_FUNC_map_lookup_elem: + bpf->helpers.map_lookup = readl(&cap->func_addr); + break; + } + + return 0; +} + +static int +nfp_bpf_parse_cap_maps(struct nfp_app_bpf *bpf, void __iomem *value, u32 length) +{ + struct nfp_bpf_cap_tlv_maps __iomem *cap = value; + + if (length < sizeof(*cap)) { + nfp_err(bpf->app->cpp, "truncated maps TLV: %d\n", length); + return -EINVAL; + } + + bpf->maps.types = readl(&cap->types); + bpf->maps.max_maps = readl(&cap->max_maps); + bpf->maps.max_elems = readl(&cap->max_elems); + bpf->maps.max_key_sz = readl(&cap->max_key_sz); + bpf->maps.max_val_sz = readl(&cap->max_val_sz); + bpf->maps.max_elem_sz = readl(&cap->max_elem_sz); + + return 0; +} + +static int nfp_bpf_parse_capabilities(struct nfp_app *app) +{ + struct nfp_cpp *cpp = app->pf->cpp; + struct nfp_cpp_area *area; + u8 __iomem *mem, *start; + + mem = nfp_rtsym_map(app->pf->rtbl, "_abi_bpf_capabilities", "bpf.cap", + 8, &area); + if (IS_ERR(mem)) + return PTR_ERR(mem) == -ENOENT ? 0 : PTR_ERR(mem); + + start = mem; + while (mem - start + 8 < nfp_cpp_area_size(area)) { + u8 __iomem *value; + u32 type, length; + + type = readl(mem); + length = readl(mem + 4); + value = mem + 8; + + mem += 8 + length; + if (mem - start > nfp_cpp_area_size(area)) + goto err_release_free; + + switch (type) { + case NFP_BPF_CAP_TYPE_FUNC: + if (nfp_bpf_parse_cap_func(app->priv, value, length)) + goto err_release_free; + break; + case NFP_BPF_CAP_TYPE_ADJUST_HEAD: + if (nfp_bpf_parse_cap_adjust_head(app->priv, value, + length)) + goto err_release_free; + break; + case NFP_BPF_CAP_TYPE_MAPS: + if (nfp_bpf_parse_cap_maps(app->priv, value, length)) + goto err_release_free; + break; + default: + nfp_dbg(cpp, "unknown BPF capability: %d\n", type); + break; + } + } + if (mem - start != nfp_cpp_area_size(area)) { + nfp_err(cpp, "BPF capabilities left after parsing, parsed:%zd total length:%zu\n", + mem - start, nfp_cpp_area_size(area)); + goto err_release_free; + } + + nfp_cpp_area_release_free(area); + + return 0; + +err_release_free: + nfp_err(cpp, "invalid BPF capabilities at offset:%zd\n", mem - start); + nfp_cpp_area_release_free(area); + return -EINVAL; +} + +static int nfp_bpf_init(struct nfp_app *app) +{ + struct nfp_app_bpf *bpf; + int err; + + bpf = kzalloc(sizeof(*bpf), GFP_KERNEL); + if (!bpf) + return -ENOMEM; + bpf->app = app; + app->priv = bpf; + + skb_queue_head_init(&bpf->cmsg_replies); + init_waitqueue_head(&bpf->cmsg_wq); + INIT_LIST_HEAD(&bpf->map_list); + + err = nfp_bpf_parse_capabilities(app); + if (err) + goto err_free_bpf; + + return 0; + +err_free_bpf: + kfree(bpf); + return err; +} + +static void nfp_bpf_clean(struct nfp_app *app) +{ + struct nfp_app_bpf *bpf = app->priv; + + WARN_ON(!skb_queue_empty(&bpf->cmsg_replies)); + WARN_ON(!list_empty(&bpf->map_list)); + WARN_ON(bpf->maps_in_use || bpf->map_elems_in_use); + kfree(bpf); } const struct nfp_app_type app_bpf = { .id = NFP_APP_BPF_NIC, .name = "ebpf", + .ctrl_cap_mask = 0, + + .init = nfp_bpf_init, + .clean = nfp_bpf_clean, + + .change_mtu = nfp_bpf_change_mtu, + .extra_cap = nfp_bpf_extra_cap, .vnic_alloc = nfp_bpf_vnic_alloc, .vnic_free = nfp_bpf_vnic_free, + .ctrl_msg_rx = nfp_bpf_ctrl_msg_rx, + .setup_tc = nfp_bpf_setup_tc, .tc_busy = nfp_bpf_tc_busy, + .bpf = nfp_ndo_bpf, .xdp_offload = nfp_bpf_xdp_offload, - - .bpf_verifier_prep = nfp_bpf_verifier_prep, - .bpf_translate = nfp_bpf_translate, - .bpf_destroy = nfp_bpf_destroy, }; diff --git a/drivers/net/ethernet/netronome/nfp/bpf/main.h b/drivers/net/ethernet/netronome/nfp/bpf/main.h index 57b6043177a3..424fe8338105 100644 --- a/drivers/net/ethernet/netronome/nfp/bpf/main.h +++ b/drivers/net/ethernet/netronome/nfp/bpf/main.h @@ -1,5 +1,5 @@ /* - * Copyright (C) 2016 Netronome Systems, Inc. + * Copyright (C) 2016-2017 Netronome Systems, Inc. * * This software is dual licensed under the GNU General License Version 2, * June 1991 as shown in the file COPYING in the top-level directory of this @@ -37,22 +37,40 @@ #include <linux/bitfield.h> #include <linux/bpf.h> #include <linux/bpf_verifier.h> +#include <linux/kernel.h> #include <linux/list.h> +#include <linux/skbuff.h> #include <linux/types.h> +#include <linux/wait.h> #include "../nfp_asm.h" +#include "fw.h" -/* For branch fixup logic use up-most byte of branch instruction as scratch +/* For relocation logic use up-most byte of branch instruction as scratch * area. Remember to clear this before sending instructions to HW! */ -#define OP_BR_SPECIAL 0xff00000000000000ULL - -enum br_special { - OP_BR_NORMAL = 0, - OP_BR_GO_OUT, - OP_BR_GO_ABORT, +#define OP_RELO_TYPE 0xff00000000000000ULL + +enum nfp_relo_type { + RELO_NONE = 0, + /* standard internal jumps */ + RELO_BR_REL, + /* internal jumps to parts of the outro */ + RELO_BR_GO_OUT, + RELO_BR_GO_ABORT, + /* external jumps to fixed addresses */ + RELO_BR_NEXT_PKT, + RELO_BR_HELPER, + /* immediate relocation against load address */ + RELO_IMMED_REL, }; +/* To make absolute relocated branches (branches other than RELO_BR_REL) + * distinguishable in user space dumps from normal jumps, add a large offset + * to them. + */ +#define BR_OFF_RELO 15000 + enum static_regs { STATIC_REG_IMM = 21, /* Bank AB */ STATIC_REG_STACK = 22, /* Bank A */ @@ -78,6 +96,89 @@ enum pkt_vec { #define NFP_BPF_ABI_FLAGS reg_imm(0) #define NFP_BPF_ABI_FLAG_MARK 1 +/** + * struct nfp_app_bpf - bpf app priv structure + * @app: backpointer to the app + * + * @tag_allocator: bitmap of control message tags in use + * @tag_alloc_next: next tag bit to allocate + * @tag_alloc_last: next tag bit to be freed + * + * @cmsg_replies: received cmsg replies waiting to be consumed + * @cmsg_wq: work queue for waiting for cmsg replies + * + * @map_list: list of offloaded maps + * @maps_in_use: number of currently offloaded maps + * @map_elems_in_use: number of elements allocated to offloaded maps + * + * @adjust_head: adjust head capability + * @flags: extra flags for adjust head + * @off_min: minimal packet offset within buffer required + * @off_max: maximum packet offset within buffer required + * @guaranteed_sub: amount of negative adjustment guaranteed possible + * @guaranteed_add: amount of positive adjustment guaranteed possible + * + * @maps: map capability + * @types: supported map types + * @max_maps: max number of maps supported + * @max_elems: max number of entries in each map + * @max_key_sz: max size of map key + * @max_val_sz: max size of map value + * @max_elem_sz: max size of map entry (key + value) + * + * @helpers: helper addressess for various calls + * @map_lookup: map lookup helper address + */ +struct nfp_app_bpf { + struct nfp_app *app; + + DECLARE_BITMAP(tag_allocator, U16_MAX + 1); + u16 tag_alloc_next; + u16 tag_alloc_last; + + struct sk_buff_head cmsg_replies; + struct wait_queue_head cmsg_wq; + + struct list_head map_list; + unsigned int maps_in_use; + unsigned int map_elems_in_use; + + struct nfp_bpf_cap_adjust_head { + u32 flags; + int off_min; + int off_max; + int guaranteed_sub; + int guaranteed_add; + } adjust_head; + + struct { + u32 types; + u32 max_maps; + u32 max_elems; + u32 max_key_sz; + u32 max_val_sz; + u32 max_elem_sz; + } maps; + + struct { + u32 map_lookup; + } helpers; +}; + +/** + * struct nfp_bpf_map - private per-map data attached to BPF maps for offload + * @offmap: pointer to the offloaded BPF map + * @bpf: back pointer to bpf app private structure + * @tid: table id identifying map on datapath + * @l: link on the nfp_app_bpf->map_list list + */ +struct nfp_bpf_map { + struct bpf_offloaded_map *offmap; + struct nfp_app_bpf *bpf; + u32 tid; + struct list_head l; +}; + struct nfp_prog; struct nfp_insn_meta; typedef int (*instr_cb_t)(struct nfp_prog *, struct nfp_insn_meta *); @@ -89,23 +190,47 @@ typedef int (*instr_cb_t)(struct nfp_prog *, struct nfp_insn_meta *); #define nfp_meta_next(meta) list_next_entry(meta, l) #define nfp_meta_prev(meta) list_prev_entry(meta, l) +#define FLAG_INSN_IS_JUMP_DST BIT(0) + /** * struct nfp_insn_meta - BPF instruction wrapper * @insn: BPF instruction * @ptr: pointer type for memory operations + * @ldst_gather_len: memcpy length gathered from load/store sequence + * @paired_st: the paired store insn at the head of the sequence * @ptr_not_const: pointer is not always constant + * @jmp_dst: destination info for jump instructions + * @func_id: function id for call instructions + * @arg1: arg1 for call instructions + * @arg2: arg2 for call instructions + * @arg2_var_off: arg2 changes stack offset on different paths * @off: index of first generated machine instruction (in nfp_prog.prog) * @n: eBPF instruction number + * @flags: eBPF instruction extra optimization flags * @skip: skip this instruction (optimized out) * @double_cb: callback for second part of the instruction * @l: link on nfp_prog->insns list */ struct nfp_insn_meta { struct bpf_insn insn; - struct bpf_reg_state ptr; - bool ptr_not_const; + union { + struct { + struct bpf_reg_state ptr; + struct bpf_insn *paired_st; + s16 ldst_gather_len; + bool ptr_not_const; + }; + struct nfp_insn_meta *jmp_dst; + struct { + u32 func_id; + struct bpf_reg_state arg1; + struct bpf_reg_state arg2; + bool arg2_var_off; + }; + }; unsigned int off; unsigned short n; + unsigned short flags; bool skip; instr_cb_t double_cb; @@ -134,23 +259,36 @@ static inline u8 mbpf_mode(const struct nfp_insn_meta *meta) return BPF_MODE(meta->insn.code); } +static inline bool is_mbpf_load(const struct nfp_insn_meta *meta) +{ + return (meta->insn.code & ~BPF_SIZE_MASK) == (BPF_LDX | BPF_MEM); +} + +static inline bool is_mbpf_store(const struct nfp_insn_meta *meta) +{ + return (meta->insn.code & ~BPF_SIZE_MASK) == (BPF_STX | BPF_MEM); +} + /** * struct nfp_prog - nfp BPF program + * @bpf: backpointer to the bpf app priv structure * @prog: machine code * @prog_len: number of valid instructions in @prog array * @__prog_alloc_len: alloc size of @prog array * @verifier_meta: temporary storage for verifier's insn meta * @type: BPF program type - * @start_off: address of the first instruction in the memory + * @last_bpf_off: address of the last instruction translated from BPF * @tgt_out: jump target for normal exit * @tgt_abort: jump target for abort (e.g. access outside of packet buffer) - * @tgt_done: jump target to get the next packet * @n_translated: number of successfully translated instructions (for errors) * @error: error code if something went wrong * @stack_depth: max stack depth from the verifier + * @adjust_head_location: if program has single adjust head call - the insn no. * @insns: list of BPF instruction wrappers (struct nfp_insn_meta) */ struct nfp_prog { + struct nfp_app_bpf *bpf; + u64 *prog; unsigned int prog_len; unsigned int __prog_alloc_len; @@ -159,15 +297,15 @@ struct nfp_prog { enum bpf_prog_type type; - unsigned int start_off; + unsigned int last_bpf_off; unsigned int tgt_out; unsigned int tgt_abort; - unsigned int tgt_done; unsigned int n_translated; int error; unsigned int stack_depth; + unsigned int adjust_head_location; struct list_head insns; }; @@ -175,26 +313,49 @@ struct nfp_prog { /** * struct nfp_bpf_vnic - per-vNIC BPF priv structure * @tc_prog: currently loaded cls_bpf program + * @start_off: address of the first instruction in the memory + * @tgt_done: jump target to get the next packet */ struct nfp_bpf_vnic { struct bpf_prog *tc_prog; + unsigned int start_off; + unsigned int tgt_done; }; +void nfp_bpf_jit_prepare(struct nfp_prog *nfp_prog, unsigned int cnt); int nfp_bpf_jit(struct nfp_prog *prog); +bool nfp_bpf_supported_opcode(u8 code); -extern const struct bpf_ext_analyzer_ops nfp_bpf_analyzer_ops; +extern const struct bpf_prog_offload_ops nfp_bpf_analyzer_ops; struct netdev_bpf; struct nfp_app; struct nfp_net; +int nfp_ndo_bpf(struct nfp_app *app, struct nfp_net *nn, + struct netdev_bpf *bpf); int nfp_net_bpf_offload(struct nfp_net *nn, struct bpf_prog *prog, - bool old_prog); - -int nfp_bpf_verifier_prep(struct nfp_app *app, struct nfp_net *nn, - struct netdev_bpf *bpf); -int nfp_bpf_translate(struct nfp_app *app, struct nfp_net *nn, - struct bpf_prog *prog); -int nfp_bpf_destroy(struct nfp_app *app, struct nfp_net *nn, - struct bpf_prog *prog); + bool old_prog, struct netlink_ext_ack *extack); + +struct nfp_insn_meta * +nfp_bpf_goto_meta(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, + unsigned int insn_idx, unsigned int n_insns); + +void *nfp_bpf_relo_for_vnic(struct nfp_prog *nfp_prog, struct nfp_bpf_vnic *bv); + +long long int +nfp_bpf_ctrl_alloc_map(struct nfp_app_bpf *bpf, struct bpf_map *map); +void +nfp_bpf_ctrl_free_map(struct nfp_app_bpf *bpf, struct nfp_bpf_map *nfp_map); +int nfp_bpf_ctrl_getfirst_entry(struct bpf_offloaded_map *offmap, + void *next_key); +int nfp_bpf_ctrl_update_entry(struct bpf_offloaded_map *offmap, + void *key, void *value, u64 flags); +int nfp_bpf_ctrl_del_entry(struct bpf_offloaded_map *offmap, void *key); +int nfp_bpf_ctrl_lookup_entry(struct bpf_offloaded_map *offmap, + void *key, void *value); +int nfp_bpf_ctrl_getnext_entry(struct bpf_offloaded_map *offmap, + void *key, void *next_key); + +void nfp_bpf_ctrl_msg_rx(struct nfp_app *app, struct sk_buff *skb); #endif diff --git a/drivers/net/ethernet/netronome/nfp/bpf/offload.c b/drivers/net/ethernet/netronome/nfp/bpf/offload.c index bc879aeb62d4..0a7732385469 100644 --- a/drivers/net/ethernet/netronome/nfp/bpf/offload.c +++ b/drivers/net/ethernet/netronome/nfp/bpf/offload.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2016 Netronome Systems, Inc. + * Copyright (C) 2016-2017 Netronome Systems, Inc. * * This software is dual licensed under the GNU General License Version 2, * June 1991 as shown in the file COPYING in the top-level directory of this @@ -36,18 +36,23 @@ * Netronome network device driver: TC offload functions for PF and VF */ +#define pr_fmt(fmt) "NFP net bpf: " fmt + +#include <linux/bpf.h> #include <linux/kernel.h> #include <linux/netdevice.h> #include <linux/pci.h> #include <linux/jiffies.h> #include <linux/timer.h> #include <linux/list.h> +#include <linux/mm.h> #include <net/pkt_cls.h> #include <net/tc_act/tc_gact.h> #include <net/tc_act/tc_mirred.h> #include "main.h" +#include "../nfp_app.h" #include "../nfp_net_ctrl.h" #include "../nfp_net.h" @@ -55,11 +60,10 @@ static int nfp_prog_prepare(struct nfp_prog *nfp_prog, const struct bpf_insn *prog, unsigned int cnt) { + struct nfp_insn_meta *meta; unsigned int i; for (i = 0; i < cnt; i++) { - struct nfp_insn_meta *meta; - meta = kzalloc(sizeof(*meta), GFP_KERNEL); if (!meta) return -ENOMEM; @@ -70,6 +74,8 @@ nfp_prog_prepare(struct nfp_prog *nfp_prog, const struct bpf_insn *prog, list_add_tail(&meta->l, &nfp_prog->insns); } + nfp_bpf_jit_prepare(nfp_prog, cnt); + return 0; } @@ -84,8 +90,9 @@ static void nfp_prog_free(struct nfp_prog *nfp_prog) kfree(nfp_prog); } -int nfp_bpf_verifier_prep(struct nfp_app *app, struct nfp_net *nn, - struct netdev_bpf *bpf) +static int +nfp_bpf_verifier_prep(struct nfp_app *app, struct nfp_net *nn, + struct netdev_bpf *bpf) { struct bpf_prog *prog = bpf->verifier.prog; struct nfp_prog *nfp_prog; @@ -98,6 +105,7 @@ int nfp_bpf_verifier_prep(struct nfp_app *app, struct nfp_net *nn, INIT_LIST_HEAD(&nfp_prog->insns); nfp_prog->type = prog->type; + nfp_prog->bpf = app->priv; ret = nfp_prog_prepare(nfp_prog, prog->insnsi, prog->len); if (ret) @@ -114,12 +122,12 @@ err_free: return ret; } -int nfp_bpf_translate(struct nfp_app *app, struct nfp_net *nn, - struct bpf_prog *prog) +static int nfp_bpf_translate(struct nfp_net *nn, struct bpf_prog *prog) { struct nfp_prog *nfp_prog = prog->aux->offload->dev_priv; unsigned int stack_size; unsigned int max_instr; + int err; stack_size = nn_readb(nn, NFP_NET_CFG_BPF_STACK_SZ) * 64; if (prog->aux->stack_depth > stack_size) { @@ -127,50 +135,179 @@ int nfp_bpf_translate(struct nfp_app *app, struct nfp_net *nn, prog->aux->stack_depth, stack_size); return -EOPNOTSUPP; } - - nfp_prog->stack_depth = prog->aux->stack_depth; - nfp_prog->start_off = nn_readw(nn, NFP_NET_CFG_BPF_START); - nfp_prog->tgt_done = nn_readw(nn, NFP_NET_CFG_BPF_DONE); + nfp_prog->stack_depth = round_up(prog->aux->stack_depth, 4); max_instr = nn_readw(nn, NFP_NET_CFG_BPF_MAX_LEN); nfp_prog->__prog_alloc_len = max_instr * sizeof(u64); - nfp_prog->prog = kmalloc(nfp_prog->__prog_alloc_len, GFP_KERNEL); + nfp_prog->prog = kvmalloc(nfp_prog->__prog_alloc_len, GFP_KERNEL); if (!nfp_prog->prog) return -ENOMEM; - return nfp_bpf_jit(nfp_prog); + err = nfp_bpf_jit(nfp_prog); + if (err) + return err; + + prog->aux->offload->jited_len = nfp_prog->prog_len * sizeof(u64); + prog->aux->offload->jited_image = nfp_prog->prog; + + return 0; } -int nfp_bpf_destroy(struct nfp_app *app, struct nfp_net *nn, - struct bpf_prog *prog) +static int nfp_bpf_destroy(struct nfp_net *nn, struct bpf_prog *prog) { struct nfp_prog *nfp_prog = prog->aux->offload->dev_priv; - kfree(nfp_prog->prog); + kvfree(nfp_prog->prog); nfp_prog_free(nfp_prog); return 0; } -static int nfp_net_bpf_load(struct nfp_net *nn, struct bpf_prog *prog) +static int +nfp_bpf_map_get_next_key(struct bpf_offloaded_map *offmap, + void *key, void *next_key) +{ + if (!key) + return nfp_bpf_ctrl_getfirst_entry(offmap, next_key); + return nfp_bpf_ctrl_getnext_entry(offmap, key, next_key); +} + +static int +nfp_bpf_map_delete_elem(struct bpf_offloaded_map *offmap, void *key) +{ + if (offmap->map.map_type == BPF_MAP_TYPE_ARRAY) + return -EINVAL; + return nfp_bpf_ctrl_del_entry(offmap, key); +} + +static const struct bpf_map_dev_ops nfp_bpf_map_ops = { + .map_get_next_key = nfp_bpf_map_get_next_key, + .map_lookup_elem = nfp_bpf_ctrl_lookup_entry, + .map_update_elem = nfp_bpf_ctrl_update_entry, + .map_delete_elem = nfp_bpf_map_delete_elem, +}; + +static int +nfp_bpf_map_alloc(struct nfp_app_bpf *bpf, struct bpf_offloaded_map *offmap) +{ + struct nfp_bpf_map *nfp_map; + long long int res; + + if (!bpf->maps.types) + return -EOPNOTSUPP; + + if (offmap->map.map_flags || + offmap->map.numa_node != NUMA_NO_NODE) { + pr_info("map flags are not supported\n"); + return -EINVAL; + } + + if (!(bpf->maps.types & 1 << offmap->map.map_type)) { + pr_info("map type not supported\n"); + return -EOPNOTSUPP; + } + if (bpf->maps.max_maps == bpf->maps_in_use) { + pr_info("too many maps for a device\n"); + return -ENOMEM; + } + if (bpf->maps.max_elems - bpf->map_elems_in_use < + offmap->map.max_entries) { + pr_info("map with too many elements: %u, left: %u\n", + offmap->map.max_entries, + bpf->maps.max_elems - bpf->map_elems_in_use); + return -ENOMEM; + } + if (offmap->map.key_size > bpf->maps.max_key_sz || + offmap->map.value_size > bpf->maps.max_val_sz || + round_up(offmap->map.key_size, 8) + + round_up(offmap->map.value_size, 8) > bpf->maps.max_elem_sz) { + pr_info("elements don't fit in device constraints\n"); + return -ENOMEM; + } + + nfp_map = kzalloc(sizeof(*nfp_map), GFP_USER); + if (!nfp_map) + return -ENOMEM; + + offmap->dev_priv = nfp_map; + nfp_map->offmap = offmap; + nfp_map->bpf = bpf; + + res = nfp_bpf_ctrl_alloc_map(bpf, &offmap->map); + if (res < 0) { + kfree(nfp_map); + return res; + } + + nfp_map->tid = res; + offmap->dev_ops = &nfp_bpf_map_ops; + bpf->maps_in_use++; + bpf->map_elems_in_use += offmap->map.max_entries; + list_add_tail(&nfp_map->l, &bpf->map_list); + + return 0; +} + +static int +nfp_bpf_map_free(struct nfp_app_bpf *bpf, struct bpf_offloaded_map *offmap) +{ + struct nfp_bpf_map *nfp_map = offmap->dev_priv; + + nfp_bpf_ctrl_free_map(bpf, nfp_map); + list_del_init(&nfp_map->l); + bpf->map_elems_in_use -= offmap->map.max_entries; + bpf->maps_in_use--; + kfree(nfp_map); + + return 0; +} + +int nfp_ndo_bpf(struct nfp_app *app, struct nfp_net *nn, struct netdev_bpf *bpf) +{ + switch (bpf->command) { + case BPF_OFFLOAD_VERIFIER_PREP: + return nfp_bpf_verifier_prep(app, nn, bpf); + case BPF_OFFLOAD_TRANSLATE: + return nfp_bpf_translate(nn, bpf->offload.prog); + case BPF_OFFLOAD_DESTROY: + return nfp_bpf_destroy(nn, bpf->offload.prog); + case BPF_OFFLOAD_MAP_ALLOC: + return nfp_bpf_map_alloc(app->priv, bpf->offmap); + case BPF_OFFLOAD_MAP_FREE: + return nfp_bpf_map_free(app->priv, bpf->offmap); + default: + return -EINVAL; + } +} + +static int +nfp_net_bpf_load(struct nfp_net *nn, struct bpf_prog *prog, + struct netlink_ext_ack *extack) { struct nfp_prog *nfp_prog = prog->aux->offload->dev_priv; unsigned int max_mtu; dma_addr_t dma_addr; + void *img; int err; max_mtu = nn_readb(nn, NFP_NET_CFG_BPF_INL_MTU) * 64 - 32; if (max_mtu < nn->dp.netdev->mtu) { - nn_info(nn, "BPF offload not supported with MTU larger than HW packet split boundary\n"); + NL_SET_ERR_MSG_MOD(extack, "BPF offload not supported with MTU larger than HW packet split boundary"); return -EOPNOTSUPP; } - dma_addr = dma_map_single(nn->dp.dev, nfp_prog->prog, + img = nfp_bpf_relo_for_vnic(nfp_prog, nn->app_priv); + if (IS_ERR(img)) + return PTR_ERR(img); + + dma_addr = dma_map_single(nn->dp.dev, img, nfp_prog->prog_len * sizeof(u64), DMA_TO_DEVICE); - if (dma_mapping_error(nn->dp.dev, dma_addr)) + if (dma_mapping_error(nn->dp.dev, dma_addr)) { + kfree(img); return -ENOMEM; + } nn_writew(nn, NFP_NET_CFG_BPF_SIZE, nfp_prog->prog_len); nn_writeq(nn, NFP_NET_CFG_BPF_ADDR, dma_addr); @@ -178,15 +315,18 @@ static int nfp_net_bpf_load(struct nfp_net *nn, struct bpf_prog *prog) /* Load up the JITed code */ err = nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_BPF); if (err) - nn_err(nn, "FW command error while loading BPF: %d\n", err); + NL_SET_ERR_MSG_MOD(extack, + "FW command error while loading BPF"); dma_unmap_single(nn->dp.dev, dma_addr, nfp_prog->prog_len * sizeof(u64), DMA_TO_DEVICE); + kfree(img); return err; } -static void nfp_net_bpf_start(struct nfp_net *nn) +static void +nfp_net_bpf_start(struct nfp_net *nn, struct netlink_ext_ack *extack) { int err; @@ -195,7 +335,8 @@ static void nfp_net_bpf_start(struct nfp_net *nn) nn_writel(nn, NFP_NET_CFG_CTRL, nn->dp.ctrl); err = nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_GEN); if (err) - nn_err(nn, "FW command error while enabling BPF: %d\n", err); + NL_SET_ERR_MSG_MOD(extack, + "FW command error while enabling BPF"); } static int nfp_net_bpf_stop(struct nfp_net *nn) @@ -210,12 +351,12 @@ static int nfp_net_bpf_stop(struct nfp_net *nn) } int nfp_net_bpf_offload(struct nfp_net *nn, struct bpf_prog *prog, - bool old_prog) + bool old_prog, struct netlink_ext_ack *extack) { int err; if (prog) { - struct bpf_dev_offload *offload = prog->aux->offload; + struct bpf_prog_offload *offload = prog->aux->offload; if (!offload) return -EINVAL; @@ -228,7 +369,8 @@ int nfp_net_bpf_offload(struct nfp_net *nn, struct bpf_prog *prog, cap = nn_readb(nn, NFP_NET_CFG_BPF_CAP); if (!(cap & NFP_NET_BPF_CAP_RELO)) { - nn_err(nn, "FW does not support live reload\n"); + NL_SET_ERR_MSG_MOD(extack, + "FW does not support live reload"); return -EBUSY; } } @@ -240,12 +382,12 @@ int nfp_net_bpf_offload(struct nfp_net *nn, struct bpf_prog *prog, if (old_prog && !prog) return nfp_net_bpf_stop(nn); - err = nfp_net_bpf_load(nn, prog); + err = nfp_net_bpf_load(nn, prog, extack); if (err) return err; if (!old_prog) - nfp_net_bpf_start(nn); + nfp_net_bpf_start(nn, extack); return 0; } diff --git a/drivers/net/ethernet/netronome/nfp/bpf/verifier.c b/drivers/net/ethernet/netronome/nfp/bpf/verifier.c index 8d43491ddd6b..479f602887e9 100644 --- a/drivers/net/ethernet/netronome/nfp/bpf/verifier.c +++ b/drivers/net/ethernet/netronome/nfp/bpf/verifier.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2016 Netronome Systems, Inc. + * Copyright (C) 2016-2017 Netronome Systems, Inc. * * This software is dual licensed under the GNU General License Version 2, * June 1991 as shown in the file COPYING in the top-level directory of this @@ -31,16 +31,18 @@ * SOFTWARE. */ -#define pr_fmt(fmt) "NFP net bpf: " fmt - #include <linux/bpf.h> #include <linux/bpf_verifier.h> #include <linux/kernel.h> #include <linux/pkt_cls.h> +#include "fw.h" #include "main.h" -static struct nfp_insn_meta * +#define pr_vlog(env, fmt, ...) \ + bpf_verifier_log_write(env, "[nfp] " fmt, ##__VA_ARGS__) + +struct nfp_insn_meta * nfp_bpf_goto_meta(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, unsigned int insn_idx, unsigned int n_insns) { @@ -68,6 +70,114 @@ nfp_bpf_goto_meta(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, return meta; } +static void +nfp_record_adjust_head(struct nfp_app_bpf *bpf, struct nfp_prog *nfp_prog, + struct nfp_insn_meta *meta, + const struct bpf_reg_state *reg2) +{ + unsigned int location = UINT_MAX; + int imm; + + /* Datapath usually can give us guarantees on how much adjust head + * can be done without the need for any checks. Optimize the simple + * case where there is only one adjust head by a constant. + */ + if (reg2->type != SCALAR_VALUE || !tnum_is_const(reg2->var_off)) + goto exit_set_location; + imm = reg2->var_off.value; + /* Translator will skip all checks, we need to guarantee min pkt len */ + if (imm > ETH_ZLEN - ETH_HLEN) + goto exit_set_location; + if (imm > (int)bpf->adjust_head.guaranteed_add || + imm < -bpf->adjust_head.guaranteed_sub) + goto exit_set_location; + + if (nfp_prog->adjust_head_location) { + /* Only one call per program allowed */ + if (nfp_prog->adjust_head_location != meta->n) + goto exit_set_location; + + if (meta->arg2.var_off.value != imm) + goto exit_set_location; + } + + location = meta->n; +exit_set_location: + nfp_prog->adjust_head_location = location; +} + +static int +nfp_bpf_check_call(struct nfp_prog *nfp_prog, struct bpf_verifier_env *env, + struct nfp_insn_meta *meta) +{ + const struct bpf_reg_state *reg1 = cur_regs(env) + BPF_REG_1; + const struct bpf_reg_state *reg2 = cur_regs(env) + BPF_REG_2; + struct nfp_app_bpf *bpf = nfp_prog->bpf; + u32 func_id = meta->insn.imm; + s64 off, old_off; + + switch (func_id) { + case BPF_FUNC_xdp_adjust_head: + if (!bpf->adjust_head.off_max) { + pr_vlog(env, "adjust_head not supported by FW\n"); + return -EOPNOTSUPP; + } + if (!(bpf->adjust_head.flags & NFP_BPF_ADJUST_HEAD_NO_META)) { + pr_vlog(env, "adjust_head: FW requires shifting metadata, not supported by the driver\n"); + return -EOPNOTSUPP; + } + + nfp_record_adjust_head(bpf, nfp_prog, meta, reg2); + break; + + case BPF_FUNC_map_lookup_elem: + if (!bpf->helpers.map_lookup) { + pr_vlog(env, "map_lookup: not supported by FW\n"); + return -EOPNOTSUPP; + } + if (reg2->type != PTR_TO_STACK) { + pr_vlog(env, + "map_lookup: unsupported key ptr type %d\n", + reg2->type); + return -EOPNOTSUPP; + } + if (!tnum_is_const(reg2->var_off)) { + pr_vlog(env, "map_lookup: variable key pointer\n"); + return -EOPNOTSUPP; + } + + off = reg2->var_off.value + reg2->off; + if (-off % 4) { + pr_vlog(env, + "map_lookup: unaligned stack pointer %lld\n", + -off); + return -EOPNOTSUPP; + } + + /* Rest of the checks is only if we re-parse the same insn */ + if (!meta->func_id) + break; + + old_off = meta->arg2.var_off.value + meta->arg2.off; + meta->arg2_var_off |= off != old_off; + + if (meta->arg1.map_ptr != reg1->map_ptr) { + pr_vlog(env, "map_lookup: called for different map\n"); + return -EOPNOTSUPP; + } + break; + default: + pr_vlog(env, "unsupported function id: %d\n", func_id); + return -EOPNOTSUPP; + } + + meta->func_id = func_id; + meta->arg1 = *reg1; + meta->arg2 = *reg2; + + return 0; +} + static int nfp_bpf_check_exit(struct nfp_prog *nfp_prog, struct bpf_verifier_env *env) @@ -82,7 +192,7 @@ nfp_bpf_check_exit(struct nfp_prog *nfp_prog, char tn_buf[48]; tnum_strn(tn_buf, sizeof(tn_buf), reg0->var_off); - pr_info("unsupported exit state: %d, var_off: %s\n", + pr_vlog(env, "unsupported exit state: %d, var_off: %s\n", reg0->type, tn_buf); return -EINVAL; } @@ -92,7 +202,7 @@ nfp_bpf_check_exit(struct nfp_prog *nfp_prog, imm <= TC_ACT_REDIRECT && imm != TC_ACT_SHOT && imm != TC_ACT_STOLEN && imm != TC_ACT_QUEUED) { - pr_info("unsupported exit state: %d, imm: %llx\n", + pr_vlog(env, "unsupported exit state: %d, imm: %llx\n", reg0->type, imm); return -EINVAL; } @@ -103,12 +213,13 @@ nfp_bpf_check_exit(struct nfp_prog *nfp_prog, static int nfp_bpf_check_stack_access(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, - const struct bpf_reg_state *reg) + const struct bpf_reg_state *reg, + struct bpf_verifier_env *env) { s32 old_off, new_off; if (!tnum_is_const(reg->var_off)) { - pr_info("variable ptr stack access\n"); + pr_vlog(env, "variable ptr stack access\n"); return -EINVAL; } @@ -126,7 +237,7 @@ nfp_bpf_check_stack_access(struct nfp_prog *nfp_prog, if (old_off % 4 == new_off % 4) return 0; - pr_info("stack access changed location was:%d is:%d\n", + pr_vlog(env, "stack access changed location was:%d is:%d\n", old_off, new_off); return -EINVAL; } @@ -140,19 +251,27 @@ nfp_bpf_check_ptr(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, if (reg->type != PTR_TO_CTX && reg->type != PTR_TO_STACK && + reg->type != PTR_TO_MAP_VALUE && reg->type != PTR_TO_PACKET) { - pr_info("unsupported ptr type: %d\n", reg->type); + pr_vlog(env, "unsupported ptr type: %d\n", reg->type); return -EINVAL; } if (reg->type == PTR_TO_STACK) { - err = nfp_bpf_check_stack_access(nfp_prog, meta, reg); + err = nfp_bpf_check_stack_access(nfp_prog, meta, reg, env); if (err) return err; } + if (reg->type == PTR_TO_MAP_VALUE) { + if (is_mbpf_store(meta)) { + pr_vlog(env, "map writes not supported\n"); + return -EOPNOTSUPP; + } + } + if (meta->ptr.type != NOT_INIT && meta->ptr.type != reg->type) { - pr_info("ptr type changed for instruction %d -> %d\n", + pr_vlog(env, "ptr type changed for instruction %d -> %d\n", meta->ptr.type, reg->type); return -EINVAL; } @@ -171,25 +290,33 @@ nfp_verify_insn(struct bpf_verifier_env *env, int insn_idx, int prev_insn_idx) meta = nfp_bpf_goto_meta(nfp_prog, meta, insn_idx, env->prog->len); nfp_prog->verifier_meta = meta; + if (!nfp_bpf_supported_opcode(meta->insn.code)) { + pr_vlog(env, "instruction %#02x not supported\n", + meta->insn.code); + return -EINVAL; + } + if (meta->insn.src_reg >= MAX_BPF_REG || meta->insn.dst_reg >= MAX_BPF_REG) { - pr_err("program uses extended registers - jit hardening?\n"); + pr_vlog(env, "program uses extended registers - jit hardening?\n"); return -EINVAL; } + if (meta->insn.code == (BPF_JMP | BPF_CALL)) + return nfp_bpf_check_call(nfp_prog, env, meta); if (meta->insn.code == (BPF_JMP | BPF_EXIT)) return nfp_bpf_check_exit(nfp_prog, env); - if ((meta->insn.code & ~BPF_SIZE_MASK) == (BPF_LDX | BPF_MEM)) + if (is_mbpf_load(meta)) return nfp_bpf_check_ptr(nfp_prog, meta, env, meta->insn.src_reg); - if ((meta->insn.code & ~BPF_SIZE_MASK) == (BPF_STX | BPF_MEM)) + if (is_mbpf_store(meta)) return nfp_bpf_check_ptr(nfp_prog, meta, env, meta->insn.dst_reg); return 0; } -const struct bpf_ext_analyzer_ops nfp_bpf_analyzer_ops = { +const struct bpf_prog_offload_ops nfp_bpf_analyzer_ops = { .insn_hook = nfp_verify_insn, }; |