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
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
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
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
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
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
|
#include <linux/if.h>
#include <linux/if_ether.h>
#include <linux/if_link.h>
#include <linux/netdevice.h>
#include <linux/in.h>
#include <linux/types.h>
#include <linux/skbuff.h>
#include <net/flow_keys.h>
#include "enic_res.h"
#include "enic_clsf.h"
/* enic_addfltr_5t - Add ipv4 5tuple filter
* @enic: enic struct of vnic
* @keys: flow_keys of ipv4 5tuple
* @rq: rq number to steer to
*
* This function returns filter_id(hardware_id) of the filter
* added. In case of error it returns an negative number.
*/
int enic_addfltr_5t(struct enic *enic, struct flow_keys *keys, u16 rq)
{
int res;
struct filter data;
switch (keys->ip_proto) {
case IPPROTO_TCP:
data.u.ipv4.protocol = PROTO_TCP;
break;
case IPPROTO_UDP:
data.u.ipv4.protocol = PROTO_UDP;
break;
default:
return -EPROTONOSUPPORT;
};
data.type = FILTER_IPV4_5TUPLE;
data.u.ipv4.src_addr = ntohl(keys->src);
data.u.ipv4.dst_addr = ntohl(keys->dst);
data.u.ipv4.src_port = ntohs(keys->port16[0]);
data.u.ipv4.dst_port = ntohs(keys->port16[1]);
data.u.ipv4.flags = FILTER_FIELDS_IPV4_5TUPLE;
spin_lock_bh(&enic->devcmd_lock);
res = vnic_dev_classifier(enic->vdev, CLSF_ADD, &rq, &data);
spin_unlock_bh(&enic->devcmd_lock);
res = (res == 0) ? rq : res;
return res;
}
/* enic_delfltr - Delete clsf filter
* @enic: enic struct of vnic
* @filter_id: filter_is(hardware_id) of filter to be deleted
*
* This function returns zero in case of success, negative number incase of
* error.
*/
int enic_delfltr(struct enic *enic, u16 filter_id)
{
int ret;
spin_lock_bh(&enic->devcmd_lock);
ret = vnic_dev_classifier(enic->vdev, CLSF_DEL, &filter_id, NULL);
spin_unlock_bh(&enic->devcmd_lock);
return ret;
}
#ifdef CONFIG_RFS_ACCEL
void enic_flow_may_expire(unsigned long data)
{
struct enic *enic = (struct enic *)data;
bool res;
int j;
spin_lock(&enic->rfs_h.lock);
for (j = 0; j < ENIC_CLSF_EXPIRE_COUNT; j++) {
struct hlist_head *hhead;
struct hlist_node *tmp;
struct enic_rfs_fltr_node *n;
hhead = &enic->rfs_h.ht_head[enic->rfs_h.toclean++];
hlist_for_each_entry_safe(n, tmp, hhead, node) {
res = rps_may_expire_flow(enic->netdev, n->rq_id,
n->flow_id, n->fltr_id);
if (res) {
res = enic_delfltr(enic, n->fltr_id);
if (unlikely(res))
continue;
hlist_del(&n->node);
kfree(n);
enic->rfs_h.free++;
}
}
}
spin_unlock(&enic->rfs_h.lock);
mod_timer(&enic->rfs_h.rfs_may_expire, jiffies + HZ/4);
}
/* enic_rfs_flw_tbl_init - initialize enic->rfs_h members
* @enic: enic data
*/
void enic_rfs_flw_tbl_init(struct enic *enic)
{
int i;
spin_lock_init(&enic->rfs_h.lock);
for (i = 0; i <= ENIC_RFS_FLW_MASK; i++)
INIT_HLIST_HEAD(&enic->rfs_h.ht_head[i]);
enic->rfs_h.max = enic->config.num_arfs;
enic->rfs_h.free = enic->rfs_h.max;
enic->rfs_h.toclean = 0;
init_timer(&enic->rfs_h.rfs_may_expire);
enic->rfs_h.rfs_may_expire.function = enic_flow_may_expire;
enic->rfs_h.rfs_may_expire.data = (unsigned long)enic;
mod_timer(&enic->rfs_h.rfs_may_expire, jiffies + HZ/4);
}
void enic_rfs_flw_tbl_free(struct enic *enic)
{
int i, res;
del_timer_sync(&enic->rfs_h.rfs_may_expire);
spin_lock(&enic->rfs_h.lock);
enic->rfs_h.free = 0;
for (i = 0; i < (1 << ENIC_RFS_FLW_BITSHIFT); i++) {
struct hlist_head *hhead;
struct hlist_node *tmp;
struct enic_rfs_fltr_node *n;
hhead = &enic->rfs_h.ht_head[i];
hlist_for_each_entry_safe(n, tmp, hhead, node) {
enic_delfltr(enic, n->fltr_id);
hlist_del(&n->node);
kfree(n);
}
}
spin_unlock(&enic->rfs_h.lock);
}
static struct enic_rfs_fltr_node *htbl_key_search(struct hlist_head *h,
struct flow_keys *k)
{
struct enic_rfs_fltr_node *tpos;
hlist_for_each_entry(tpos, h, node)
if (tpos->keys.src == k->src &&
tpos->keys.dst == k->dst &&
tpos->keys.ports == k->ports &&
tpos->keys.ip_proto == k->ip_proto &&
tpos->keys.n_proto == k->n_proto)
return tpos;
return NULL;
}
int enic_rx_flow_steer(struct net_device *dev, const struct sk_buff *skb,
u16 rxq_index, u32 flow_id)
{
struct flow_keys keys;
struct enic_rfs_fltr_node *n;
struct enic *enic;
u16 tbl_idx;
int res, i;
enic = netdev_priv(dev);
res = skb_flow_dissect(skb, &keys);
if (!res || keys.n_proto != htons(ETH_P_IP) ||
(keys.ip_proto != IPPROTO_TCP && keys.ip_proto != IPPROTO_UDP))
return -EPROTONOSUPPORT;
tbl_idx = skb_get_hash_raw(skb) & ENIC_RFS_FLW_MASK;
spin_lock(&enic->rfs_h.lock);
n = htbl_key_search(&enic->rfs_h.ht_head[tbl_idx], &keys);
if (n) { /* entry already present */
if (rxq_index == n->rq_id) {
res = -EEXIST;
goto ret_unlock;
}
/* desired rq changed for the flow, we need to delete
* old fltr and add new one
*
* The moment we delete the fltr, the upcoming pkts
* are put it default rq based on rss. When we add
* new filter, upcoming pkts are put in desired queue.
* This could cause ooo pkts.
*
* Lets 1st try adding new fltr and then del old one.
*/
i = --enic->rfs_h.free;
/* clsf tbl is full, we have to del old fltr first*/
if (unlikely(i < 0)) {
enic->rfs_h.free++;
res = enic_delfltr(enic, n->fltr_id);
if (unlikely(res < 0))
goto ret_unlock;
res = enic_addfltr_5t(enic, &keys, rxq_index);
if (res < 0) {
hlist_del(&n->node);
enic->rfs_h.free++;
goto ret_unlock;
}
/* add new fltr 1st then del old fltr */
} else {
int ret;
res = enic_addfltr_5t(enic, &keys, rxq_index);
if (res < 0) {
enic->rfs_h.free++;
goto ret_unlock;
}
ret = enic_delfltr(enic, n->fltr_id);
/* deleting old fltr failed. Add old fltr to list.
* enic_flow_may_expire() will try to delete it later.
*/
if (unlikely(ret < 0)) {
struct enic_rfs_fltr_node *d;
struct hlist_head *head;
head = &enic->rfs_h.ht_head[tbl_idx];
d = kmalloc(sizeof(*d), GFP_ATOMIC);
if (d) {
d->fltr_id = n->fltr_id;
INIT_HLIST_NODE(&d->node);
hlist_add_head(&d->node, head);
}
} else {
enic->rfs_h.free++;
}
}
n->rq_id = rxq_index;
n->fltr_id = res;
n->flow_id = flow_id;
/* entry not present */
} else {
i = --enic->rfs_h.free;
if (i <= 0) {
enic->rfs_h.free++;
res = -EBUSY;
goto ret_unlock;
}
n = kmalloc(sizeof(*n), GFP_ATOMIC);
if (!n) {
res = -ENOMEM;
enic->rfs_h.free++;
goto ret_unlock;
}
res = enic_addfltr_5t(enic, &keys, rxq_index);
if (res < 0) {
kfree(n);
enic->rfs_h.free++;
goto ret_unlock;
}
n->rq_id = rxq_index;
n->fltr_id = res;
n->flow_id = flow_id;
n->keys = keys;
INIT_HLIST_NODE(&n->node);
hlist_add_head(&n->node, &enic->rfs_h.ht_head[tbl_idx]);
}
ret_unlock:
spin_unlock(&enic->rfs_h.lock);
return res;
}
#else
void enic_rfs_flw_tbl_init(struct enic *enic)
{
}
void enic_rfs_flw_tbl_free(struct enic *enic)
{
}
#endif /* CONFIG_RFS_ACCEL */
|
