nfp_net_offload.c 7.8 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
/*
 * Copyright (C) 2016 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.
 */

/*
 * nfp_net_offload.c
 * Netronome network device driver: TC offload functions for PF and VF
 */

#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 <net/pkt_cls.h>
#include <net/tc_act/tc_gact.h>
#include <net/tc_act/tc_mirred.h>

#include "nfp_bpf.h"
#include "nfp_net_ctrl.h"
#include "nfp_net.h"

54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107
void nfp_net_filter_stats_timer(unsigned long data)
{
	struct nfp_net *nn = (void *)data;
	struct nfp_stat_pair latest;

	spin_lock_bh(&nn->rx_filter_lock);

	if (nn->ctrl & NFP_NET_CFG_CTRL_BPF)
		mod_timer(&nn->rx_filter_stats_timer,
			  jiffies + NFP_NET_STAT_POLL_IVL);

	spin_unlock_bh(&nn->rx_filter_lock);

	latest.pkts = nn_readq(nn, NFP_NET_CFG_STATS_APP1_FRAMES);
	latest.bytes = nn_readq(nn, NFP_NET_CFG_STATS_APP1_BYTES);

	if (latest.pkts != nn->rx_filter.pkts)
		nn->rx_filter_change = jiffies;

	nn->rx_filter = latest;
}

static void nfp_net_bpf_stats_reset(struct nfp_net *nn)
{
	nn->rx_filter.pkts = nn_readq(nn, NFP_NET_CFG_STATS_APP1_FRAMES);
	nn->rx_filter.bytes = nn_readq(nn, NFP_NET_CFG_STATS_APP1_BYTES);
	nn->rx_filter_prev = nn->rx_filter;
	nn->rx_filter_change = jiffies;
}

static int
nfp_net_bpf_stats_update(struct nfp_net *nn, struct tc_cls_bpf_offload *cls_bpf)
{
	struct tc_action *a;
	LIST_HEAD(actions);
	u64 bytes, pkts;

	pkts = nn->rx_filter.pkts - nn->rx_filter_prev.pkts;
	bytes = nn->rx_filter.bytes - nn->rx_filter_prev.bytes;
	bytes -= pkts * ETH_HLEN;

	nn->rx_filter_prev = nn->rx_filter;

	preempt_disable();

	tcf_exts_to_list(cls_bpf->exts, &actions);
	list_for_each_entry(a, &actions, list)
		tcf_action_stats_update(a, bytes, pkts, nn->rx_filter_change);

	preempt_enable();

	return 0;
}

108 109 110 111 112 113 114
static int
nfp_net_bpf_get_act(struct nfp_net *nn, struct tc_cls_bpf_offload *cls_bpf)
{
	const struct tc_action *a;
	LIST_HEAD(actions);

	/* TC direct action */
115 116 117 118
	if (cls_bpf->exts_integrated) {
		if (tc_no_actions(cls_bpf->exts))
			return NN_ACT_DIRECT;

119
		return -ENOTSUPP;
120
	}
121 122 123 124 125 126 127 128 129

	/* TC legacy mode */
	if (!tc_single_action(cls_bpf->exts))
		return -ENOTSUPP;

	tcf_exts_to_list(cls_bpf->exts, &actions);
	list_for_each_entry(a, &actions, list) {
		if (is_tcf_gact_shot(a))
			return NN_ACT_TC_DROP;
130

131
		if (is_tcf_mirred_egress_redirect(a) &&
132 133
		    tcf_mirred_ifindex(a) == nn->netdev->ifindex)
			return NN_ACT_TC_REDIR;
134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150
	}

	return -ENOTSUPP;
}

static int
nfp_net_bpf_offload_prepare(struct nfp_net *nn,
			    struct tc_cls_bpf_offload *cls_bpf,
			    struct nfp_bpf_result *res,
			    void **code, dma_addr_t *dma_addr, u16 max_instr)
{
	unsigned int code_sz = max_instr * sizeof(u64);
	enum nfp_bpf_action_type act;
	u16 start_off, done_off;
	unsigned int max_mtu;
	int ret;

151 152 153
	if (!IS_ENABLED(CONFIG_BPF_SYSCALL))
		return -ENOTSUPP;

154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214
	ret = nfp_net_bpf_get_act(nn, cls_bpf);
	if (ret < 0)
		return ret;
	act = ret;

	max_mtu = nn_readb(nn, NFP_NET_CFG_BPF_INL_MTU) * 64 - 32;
	if (max_mtu < nn->netdev->mtu) {
		nn_info(nn, "BPF offload not supported with MTU larger than HW packet split boundary\n");
		return -ENOTSUPP;
	}

	start_off = nn_readw(nn, NFP_NET_CFG_BPF_START);
	done_off = nn_readw(nn, NFP_NET_CFG_BPF_DONE);

	*code = dma_zalloc_coherent(&nn->pdev->dev, code_sz, dma_addr,
				    GFP_KERNEL);
	if (!*code)
		return -ENOMEM;

	ret = nfp_bpf_jit(cls_bpf->prog, *code, act, start_off, done_off,
			  max_instr, res);
	if (ret)
		goto out;

	return 0;

out:
	dma_free_coherent(&nn->pdev->dev, code_sz, *code, *dma_addr);
	return ret;
}

static void
nfp_net_bpf_load_and_start(struct nfp_net *nn, u32 tc_flags,
			   void *code, dma_addr_t dma_addr,
			   unsigned int code_sz, unsigned int n_instr,
			   bool dense_mode)
{
	u64 bpf_addr = dma_addr;
	int err;

	nn->bpf_offload_skip_sw = !!(tc_flags & TCA_CLS_FLAGS_SKIP_SW);

	if (dense_mode)
		bpf_addr |= NFP_NET_CFG_BPF_CFG_8CTX;

	nn_writew(nn, NFP_NET_CFG_BPF_SIZE, n_instr);
	nn_writeq(nn, NFP_NET_CFG_BPF_ADDR, bpf_addr);

	/* 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);

	/* Enable passing packets through BPF function */
	nn->ctrl |= NFP_NET_CFG_CTRL_BPF;
	nn_writel(nn, NFP_NET_CFG_CTRL, nn->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);

	dma_free_coherent(&nn->pdev->dev, code_sz, code, dma_addr);
215 216 217

	nfp_net_bpf_stats_reset(nn);
	mod_timer(&nn->rx_filter_stats_timer, jiffies + NFP_NET_STAT_POLL_IVL);
218 219 220 221 222 223 224
}

static int nfp_net_bpf_stop(struct nfp_net *nn)
{
	if (!(nn->ctrl & NFP_NET_CFG_CTRL_BPF))
		return 0;

225
	spin_lock_bh(&nn->rx_filter_lock);
226
	nn->ctrl &= ~NFP_NET_CFG_CTRL_BPF;
227
	spin_unlock_bh(&nn->rx_filter_lock);
228 229
	nn_writel(nn, NFP_NET_CFG_CTRL, nn->ctrl);

230
	del_timer_sync(&nn->rx_filter_stats_timer);
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
	nn->bpf_offload_skip_sw = 0;

	return nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_GEN);
}

int
nfp_net_bpf_offload(struct nfp_net *nn, u32 handle, __be16 proto,
		    struct tc_cls_bpf_offload *cls_bpf)
{
	struct nfp_bpf_result res;
	dma_addr_t dma_addr;
	u16 max_instr;
	void *code;
	int err;

	max_instr = nn_readw(nn, NFP_NET_CFG_BPF_MAX_LEN);

	switch (cls_bpf->command) {
	case TC_CLSBPF_REPLACE:
		/* There is nothing stopping us from implementing seamless
		 * replace but the simple method of loading I adopted in
		 * the firmware does not handle atomic replace (i.e. we have to
		 * stop the BPF offload and re-enable it).  Leaking-in a few
		 * frames which didn't have BPF applied in the hardware should
		 * be fine if software fallback is available, though.
		 */
		if (nn->bpf_offload_skip_sw)
			return -EBUSY;

		err = nfp_net_bpf_offload_prepare(nn, cls_bpf, &res, &code,
						  &dma_addr, max_instr);
		if (err)
			return err;

		nfp_net_bpf_stop(nn);
		nfp_net_bpf_load_and_start(nn, cls_bpf->gen_flags, code,
					   dma_addr, max_instr * sizeof(u64),
					   res.n_instr, res.dense_mode);
		return 0;

	case TC_CLSBPF_ADD:
		if (nn->ctrl & NFP_NET_CFG_CTRL_BPF)
			return -EBUSY;

		err = nfp_net_bpf_offload_prepare(nn, cls_bpf, &res, &code,
						  &dma_addr, max_instr);
		if (err)
			return err;

		nfp_net_bpf_load_and_start(nn, cls_bpf->gen_flags, code,
					   dma_addr, max_instr * sizeof(u64),
					   res.n_instr, res.dense_mode);
		return 0;

	case TC_CLSBPF_DESTROY:
		return nfp_net_bpf_stop(nn);

288 289 290
	case TC_CLSBPF_STATS:
		return nfp_net_bpf_stats_update(nn, cls_bpf);

291 292 293 294
	default:
		return -ENOTSUPP;
	}
}