Commit 7e5ab157813993356f021757d0b0dcbdca7c55a1
Committed by
David S. Miller
1 parent
9badba2543
Exists in
master
and in
7 other branches
net_sched: minor netns related cleanup
These changes were suggested by Alexey Dobriyan <adobriyan@gmail.com>: - psched_show() does not use any private data so just pass NULL to psched_open() - remove unnecessary return statement Signed-off-by: Tom Goff <thomas.goff@boeing.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Showing 1 changed file with 1 additions and 3 deletions Inline Diff
net/sched/sch_api.c
1 | /* | 1 | /* |
2 | * net/sched/sch_api.c Packet scheduler API. | 2 | * net/sched/sch_api.c Packet scheduler API. |
3 | * | 3 | * |
4 | * This program is free software; you can redistribute it and/or | 4 | * This program is free software; you can redistribute it and/or |
5 | * modify it under the terms of the GNU General Public License | 5 | * modify it under the terms of the GNU General Public License |
6 | * as published by the Free Software Foundation; either version | 6 | * as published by the Free Software Foundation; either version |
7 | * 2 of the License, or (at your option) any later version. | 7 | * 2 of the License, or (at your option) any later version. |
8 | * | 8 | * |
9 | * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> | 9 | * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> |
10 | * | 10 | * |
11 | * Fixes: | 11 | * Fixes: |
12 | * | 12 | * |
13 | * Rani Assaf <rani@magic.metawire.com> :980802: JIFFIES and CPU clock sources are repaired. | 13 | * Rani Assaf <rani@magic.metawire.com> :980802: JIFFIES and CPU clock sources are repaired. |
14 | * Eduardo J. Blanco <ejbs@netlabs.com.uy> :990222: kmod support | 14 | * Eduardo J. Blanco <ejbs@netlabs.com.uy> :990222: kmod support |
15 | * Jamal Hadi Salim <hadi@nortelnetworks.com>: 990601: ingress support | 15 | * Jamal Hadi Salim <hadi@nortelnetworks.com>: 990601: ingress support |
16 | */ | 16 | */ |
17 | 17 | ||
18 | #include <linux/module.h> | 18 | #include <linux/module.h> |
19 | #include <linux/types.h> | 19 | #include <linux/types.h> |
20 | #include <linux/kernel.h> | 20 | #include <linux/kernel.h> |
21 | #include <linux/string.h> | 21 | #include <linux/string.h> |
22 | #include <linux/errno.h> | 22 | #include <linux/errno.h> |
23 | #include <linux/skbuff.h> | 23 | #include <linux/skbuff.h> |
24 | #include <linux/init.h> | 24 | #include <linux/init.h> |
25 | #include <linux/proc_fs.h> | 25 | #include <linux/proc_fs.h> |
26 | #include <linux/seq_file.h> | 26 | #include <linux/seq_file.h> |
27 | #include <linux/kmod.h> | 27 | #include <linux/kmod.h> |
28 | #include <linux/list.h> | 28 | #include <linux/list.h> |
29 | #include <linux/hrtimer.h> | 29 | #include <linux/hrtimer.h> |
30 | #include <linux/lockdep.h> | 30 | #include <linux/lockdep.h> |
31 | 31 | ||
32 | #include <net/net_namespace.h> | 32 | #include <net/net_namespace.h> |
33 | #include <net/sock.h> | 33 | #include <net/sock.h> |
34 | #include <net/netlink.h> | 34 | #include <net/netlink.h> |
35 | #include <net/pkt_sched.h> | 35 | #include <net/pkt_sched.h> |
36 | 36 | ||
37 | static int qdisc_notify(struct net *net, struct sk_buff *oskb, | 37 | static int qdisc_notify(struct net *net, struct sk_buff *oskb, |
38 | struct nlmsghdr *n, u32 clid, | 38 | struct nlmsghdr *n, u32 clid, |
39 | struct Qdisc *old, struct Qdisc *new); | 39 | struct Qdisc *old, struct Qdisc *new); |
40 | static int tclass_notify(struct net *net, struct sk_buff *oskb, | 40 | static int tclass_notify(struct net *net, struct sk_buff *oskb, |
41 | struct nlmsghdr *n, struct Qdisc *q, | 41 | struct nlmsghdr *n, struct Qdisc *q, |
42 | unsigned long cl, int event); | 42 | unsigned long cl, int event); |
43 | 43 | ||
44 | /* | 44 | /* |
45 | 45 | ||
46 | Short review. | 46 | Short review. |
47 | ------------- | 47 | ------------- |
48 | 48 | ||
49 | This file consists of two interrelated parts: | 49 | This file consists of two interrelated parts: |
50 | 50 | ||
51 | 1. queueing disciplines manager frontend. | 51 | 1. queueing disciplines manager frontend. |
52 | 2. traffic classes manager frontend. | 52 | 2. traffic classes manager frontend. |
53 | 53 | ||
54 | Generally, queueing discipline ("qdisc") is a black box, | 54 | Generally, queueing discipline ("qdisc") is a black box, |
55 | which is able to enqueue packets and to dequeue them (when | 55 | which is able to enqueue packets and to dequeue them (when |
56 | device is ready to send something) in order and at times | 56 | device is ready to send something) in order and at times |
57 | determined by algorithm hidden in it. | 57 | determined by algorithm hidden in it. |
58 | 58 | ||
59 | qdisc's are divided to two categories: | 59 | qdisc's are divided to two categories: |
60 | - "queues", which have no internal structure visible from outside. | 60 | - "queues", which have no internal structure visible from outside. |
61 | - "schedulers", which split all the packets to "traffic classes", | 61 | - "schedulers", which split all the packets to "traffic classes", |
62 | using "packet classifiers" (look at cls_api.c) | 62 | using "packet classifiers" (look at cls_api.c) |
63 | 63 | ||
64 | In turn, classes may have child qdiscs (as rule, queues) | 64 | In turn, classes may have child qdiscs (as rule, queues) |
65 | attached to them etc. etc. etc. | 65 | attached to them etc. etc. etc. |
66 | 66 | ||
67 | The goal of the routines in this file is to translate | 67 | The goal of the routines in this file is to translate |
68 | information supplied by user in the form of handles | 68 | information supplied by user in the form of handles |
69 | to more intelligible for kernel form, to make some sanity | 69 | to more intelligible for kernel form, to make some sanity |
70 | checks and part of work, which is common to all qdiscs | 70 | checks and part of work, which is common to all qdiscs |
71 | and to provide rtnetlink notifications. | 71 | and to provide rtnetlink notifications. |
72 | 72 | ||
73 | All real intelligent work is done inside qdisc modules. | 73 | All real intelligent work is done inside qdisc modules. |
74 | 74 | ||
75 | 75 | ||
76 | 76 | ||
77 | Every discipline has two major routines: enqueue and dequeue. | 77 | Every discipline has two major routines: enqueue and dequeue. |
78 | 78 | ||
79 | ---dequeue | 79 | ---dequeue |
80 | 80 | ||
81 | dequeue usually returns a skb to send. It is allowed to return NULL, | 81 | dequeue usually returns a skb to send. It is allowed to return NULL, |
82 | but it does not mean that queue is empty, it just means that | 82 | but it does not mean that queue is empty, it just means that |
83 | discipline does not want to send anything this time. | 83 | discipline does not want to send anything this time. |
84 | Queue is really empty if q->q.qlen == 0. | 84 | Queue is really empty if q->q.qlen == 0. |
85 | For complicated disciplines with multiple queues q->q is not | 85 | For complicated disciplines with multiple queues q->q is not |
86 | real packet queue, but however q->q.qlen must be valid. | 86 | real packet queue, but however q->q.qlen must be valid. |
87 | 87 | ||
88 | ---enqueue | 88 | ---enqueue |
89 | 89 | ||
90 | enqueue returns 0, if packet was enqueued successfully. | 90 | enqueue returns 0, if packet was enqueued successfully. |
91 | If packet (this one or another one) was dropped, it returns | 91 | If packet (this one or another one) was dropped, it returns |
92 | not zero error code. | 92 | not zero error code. |
93 | NET_XMIT_DROP - this packet dropped | 93 | NET_XMIT_DROP - this packet dropped |
94 | Expected action: do not backoff, but wait until queue will clear. | 94 | Expected action: do not backoff, but wait until queue will clear. |
95 | NET_XMIT_CN - probably this packet enqueued, but another one dropped. | 95 | NET_XMIT_CN - probably this packet enqueued, but another one dropped. |
96 | Expected action: backoff or ignore | 96 | Expected action: backoff or ignore |
97 | NET_XMIT_POLICED - dropped by police. | 97 | NET_XMIT_POLICED - dropped by police. |
98 | Expected action: backoff or error to real-time apps. | 98 | Expected action: backoff or error to real-time apps. |
99 | 99 | ||
100 | Auxiliary routines: | 100 | Auxiliary routines: |
101 | 101 | ||
102 | ---peek | 102 | ---peek |
103 | 103 | ||
104 | like dequeue but without removing a packet from the queue | 104 | like dequeue but without removing a packet from the queue |
105 | 105 | ||
106 | ---reset | 106 | ---reset |
107 | 107 | ||
108 | returns qdisc to initial state: purge all buffers, clear all | 108 | returns qdisc to initial state: purge all buffers, clear all |
109 | timers, counters (except for statistics) etc. | 109 | timers, counters (except for statistics) etc. |
110 | 110 | ||
111 | ---init | 111 | ---init |
112 | 112 | ||
113 | initializes newly created qdisc. | 113 | initializes newly created qdisc. |
114 | 114 | ||
115 | ---destroy | 115 | ---destroy |
116 | 116 | ||
117 | destroys resources allocated by init and during lifetime of qdisc. | 117 | destroys resources allocated by init and during lifetime of qdisc. |
118 | 118 | ||
119 | ---change | 119 | ---change |
120 | 120 | ||
121 | changes qdisc parameters. | 121 | changes qdisc parameters. |
122 | */ | 122 | */ |
123 | 123 | ||
124 | /* Protects list of registered TC modules. It is pure SMP lock. */ | 124 | /* Protects list of registered TC modules. It is pure SMP lock. */ |
125 | static DEFINE_RWLOCK(qdisc_mod_lock); | 125 | static DEFINE_RWLOCK(qdisc_mod_lock); |
126 | 126 | ||
127 | 127 | ||
128 | /************************************************ | 128 | /************************************************ |
129 | * Queueing disciplines manipulation. * | 129 | * Queueing disciplines manipulation. * |
130 | ************************************************/ | 130 | ************************************************/ |
131 | 131 | ||
132 | 132 | ||
133 | /* The list of all installed queueing disciplines. */ | 133 | /* The list of all installed queueing disciplines. */ |
134 | 134 | ||
135 | static struct Qdisc_ops *qdisc_base; | 135 | static struct Qdisc_ops *qdisc_base; |
136 | 136 | ||
137 | /* Register/uregister queueing discipline */ | 137 | /* Register/uregister queueing discipline */ |
138 | 138 | ||
139 | int register_qdisc(struct Qdisc_ops *qops) | 139 | int register_qdisc(struct Qdisc_ops *qops) |
140 | { | 140 | { |
141 | struct Qdisc_ops *q, **qp; | 141 | struct Qdisc_ops *q, **qp; |
142 | int rc = -EEXIST; | 142 | int rc = -EEXIST; |
143 | 143 | ||
144 | write_lock(&qdisc_mod_lock); | 144 | write_lock(&qdisc_mod_lock); |
145 | for (qp = &qdisc_base; (q = *qp) != NULL; qp = &q->next) | 145 | for (qp = &qdisc_base; (q = *qp) != NULL; qp = &q->next) |
146 | if (!strcmp(qops->id, q->id)) | 146 | if (!strcmp(qops->id, q->id)) |
147 | goto out; | 147 | goto out; |
148 | 148 | ||
149 | if (qops->enqueue == NULL) | 149 | if (qops->enqueue == NULL) |
150 | qops->enqueue = noop_qdisc_ops.enqueue; | 150 | qops->enqueue = noop_qdisc_ops.enqueue; |
151 | if (qops->peek == NULL) { | 151 | if (qops->peek == NULL) { |
152 | if (qops->dequeue == NULL) { | 152 | if (qops->dequeue == NULL) { |
153 | qops->peek = noop_qdisc_ops.peek; | 153 | qops->peek = noop_qdisc_ops.peek; |
154 | } else { | 154 | } else { |
155 | rc = -EINVAL; | 155 | rc = -EINVAL; |
156 | goto out; | 156 | goto out; |
157 | } | 157 | } |
158 | } | 158 | } |
159 | if (qops->dequeue == NULL) | 159 | if (qops->dequeue == NULL) |
160 | qops->dequeue = noop_qdisc_ops.dequeue; | 160 | qops->dequeue = noop_qdisc_ops.dequeue; |
161 | 161 | ||
162 | qops->next = NULL; | 162 | qops->next = NULL; |
163 | *qp = qops; | 163 | *qp = qops; |
164 | rc = 0; | 164 | rc = 0; |
165 | out: | 165 | out: |
166 | write_unlock(&qdisc_mod_lock); | 166 | write_unlock(&qdisc_mod_lock); |
167 | return rc; | 167 | return rc; |
168 | } | 168 | } |
169 | EXPORT_SYMBOL(register_qdisc); | 169 | EXPORT_SYMBOL(register_qdisc); |
170 | 170 | ||
171 | int unregister_qdisc(struct Qdisc_ops *qops) | 171 | int unregister_qdisc(struct Qdisc_ops *qops) |
172 | { | 172 | { |
173 | struct Qdisc_ops *q, **qp; | 173 | struct Qdisc_ops *q, **qp; |
174 | int err = -ENOENT; | 174 | int err = -ENOENT; |
175 | 175 | ||
176 | write_lock(&qdisc_mod_lock); | 176 | write_lock(&qdisc_mod_lock); |
177 | for (qp = &qdisc_base; (q=*qp)!=NULL; qp = &q->next) | 177 | for (qp = &qdisc_base; (q=*qp)!=NULL; qp = &q->next) |
178 | if (q == qops) | 178 | if (q == qops) |
179 | break; | 179 | break; |
180 | if (q) { | 180 | if (q) { |
181 | *qp = q->next; | 181 | *qp = q->next; |
182 | q->next = NULL; | 182 | q->next = NULL; |
183 | err = 0; | 183 | err = 0; |
184 | } | 184 | } |
185 | write_unlock(&qdisc_mod_lock); | 185 | write_unlock(&qdisc_mod_lock); |
186 | return err; | 186 | return err; |
187 | } | 187 | } |
188 | EXPORT_SYMBOL(unregister_qdisc); | 188 | EXPORT_SYMBOL(unregister_qdisc); |
189 | 189 | ||
190 | /* We know handle. Find qdisc among all qdisc's attached to device | 190 | /* We know handle. Find qdisc among all qdisc's attached to device |
191 | (root qdisc, all its children, children of children etc.) | 191 | (root qdisc, all its children, children of children etc.) |
192 | */ | 192 | */ |
193 | 193 | ||
194 | static struct Qdisc *qdisc_match_from_root(struct Qdisc *root, u32 handle) | 194 | static struct Qdisc *qdisc_match_from_root(struct Qdisc *root, u32 handle) |
195 | { | 195 | { |
196 | struct Qdisc *q; | 196 | struct Qdisc *q; |
197 | 197 | ||
198 | if (!(root->flags & TCQ_F_BUILTIN) && | 198 | if (!(root->flags & TCQ_F_BUILTIN) && |
199 | root->handle == handle) | 199 | root->handle == handle) |
200 | return root; | 200 | return root; |
201 | 201 | ||
202 | list_for_each_entry(q, &root->list, list) { | 202 | list_for_each_entry(q, &root->list, list) { |
203 | if (q->handle == handle) | 203 | if (q->handle == handle) |
204 | return q; | 204 | return q; |
205 | } | 205 | } |
206 | return NULL; | 206 | return NULL; |
207 | } | 207 | } |
208 | 208 | ||
209 | static void qdisc_list_add(struct Qdisc *q) | 209 | static void qdisc_list_add(struct Qdisc *q) |
210 | { | 210 | { |
211 | if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) | 211 | if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) |
212 | list_add_tail(&q->list, &qdisc_dev(q)->qdisc->list); | 212 | list_add_tail(&q->list, &qdisc_dev(q)->qdisc->list); |
213 | } | 213 | } |
214 | 214 | ||
215 | void qdisc_list_del(struct Qdisc *q) | 215 | void qdisc_list_del(struct Qdisc *q) |
216 | { | 216 | { |
217 | if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) | 217 | if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) |
218 | list_del(&q->list); | 218 | list_del(&q->list); |
219 | } | 219 | } |
220 | EXPORT_SYMBOL(qdisc_list_del); | 220 | EXPORT_SYMBOL(qdisc_list_del); |
221 | 221 | ||
222 | struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle) | 222 | struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle) |
223 | { | 223 | { |
224 | struct Qdisc *q; | 224 | struct Qdisc *q; |
225 | 225 | ||
226 | q = qdisc_match_from_root(dev->qdisc, handle); | 226 | q = qdisc_match_from_root(dev->qdisc, handle); |
227 | if (q) | 227 | if (q) |
228 | goto out; | 228 | goto out; |
229 | 229 | ||
230 | q = qdisc_match_from_root(dev->rx_queue.qdisc_sleeping, handle); | 230 | q = qdisc_match_from_root(dev->rx_queue.qdisc_sleeping, handle); |
231 | out: | 231 | out: |
232 | return q; | 232 | return q; |
233 | } | 233 | } |
234 | 234 | ||
235 | static struct Qdisc *qdisc_leaf(struct Qdisc *p, u32 classid) | 235 | static struct Qdisc *qdisc_leaf(struct Qdisc *p, u32 classid) |
236 | { | 236 | { |
237 | unsigned long cl; | 237 | unsigned long cl; |
238 | struct Qdisc *leaf; | 238 | struct Qdisc *leaf; |
239 | const struct Qdisc_class_ops *cops = p->ops->cl_ops; | 239 | const struct Qdisc_class_ops *cops = p->ops->cl_ops; |
240 | 240 | ||
241 | if (cops == NULL) | 241 | if (cops == NULL) |
242 | return NULL; | 242 | return NULL; |
243 | cl = cops->get(p, classid); | 243 | cl = cops->get(p, classid); |
244 | 244 | ||
245 | if (cl == 0) | 245 | if (cl == 0) |
246 | return NULL; | 246 | return NULL; |
247 | leaf = cops->leaf(p, cl); | 247 | leaf = cops->leaf(p, cl); |
248 | cops->put(p, cl); | 248 | cops->put(p, cl); |
249 | return leaf; | 249 | return leaf; |
250 | } | 250 | } |
251 | 251 | ||
252 | /* Find queueing discipline by name */ | 252 | /* Find queueing discipline by name */ |
253 | 253 | ||
254 | static struct Qdisc_ops *qdisc_lookup_ops(struct nlattr *kind) | 254 | static struct Qdisc_ops *qdisc_lookup_ops(struct nlattr *kind) |
255 | { | 255 | { |
256 | struct Qdisc_ops *q = NULL; | 256 | struct Qdisc_ops *q = NULL; |
257 | 257 | ||
258 | if (kind) { | 258 | if (kind) { |
259 | read_lock(&qdisc_mod_lock); | 259 | read_lock(&qdisc_mod_lock); |
260 | for (q = qdisc_base; q; q = q->next) { | 260 | for (q = qdisc_base; q; q = q->next) { |
261 | if (nla_strcmp(kind, q->id) == 0) { | 261 | if (nla_strcmp(kind, q->id) == 0) { |
262 | if (!try_module_get(q->owner)) | 262 | if (!try_module_get(q->owner)) |
263 | q = NULL; | 263 | q = NULL; |
264 | break; | 264 | break; |
265 | } | 265 | } |
266 | } | 266 | } |
267 | read_unlock(&qdisc_mod_lock); | 267 | read_unlock(&qdisc_mod_lock); |
268 | } | 268 | } |
269 | return q; | 269 | return q; |
270 | } | 270 | } |
271 | 271 | ||
272 | static struct qdisc_rate_table *qdisc_rtab_list; | 272 | static struct qdisc_rate_table *qdisc_rtab_list; |
273 | 273 | ||
274 | struct qdisc_rate_table *qdisc_get_rtab(struct tc_ratespec *r, struct nlattr *tab) | 274 | struct qdisc_rate_table *qdisc_get_rtab(struct tc_ratespec *r, struct nlattr *tab) |
275 | { | 275 | { |
276 | struct qdisc_rate_table *rtab; | 276 | struct qdisc_rate_table *rtab; |
277 | 277 | ||
278 | for (rtab = qdisc_rtab_list; rtab; rtab = rtab->next) { | 278 | for (rtab = qdisc_rtab_list; rtab; rtab = rtab->next) { |
279 | if (memcmp(&rtab->rate, r, sizeof(struct tc_ratespec)) == 0) { | 279 | if (memcmp(&rtab->rate, r, sizeof(struct tc_ratespec)) == 0) { |
280 | rtab->refcnt++; | 280 | rtab->refcnt++; |
281 | return rtab; | 281 | return rtab; |
282 | } | 282 | } |
283 | } | 283 | } |
284 | 284 | ||
285 | if (tab == NULL || r->rate == 0 || r->cell_log == 0 || | 285 | if (tab == NULL || r->rate == 0 || r->cell_log == 0 || |
286 | nla_len(tab) != TC_RTAB_SIZE) | 286 | nla_len(tab) != TC_RTAB_SIZE) |
287 | return NULL; | 287 | return NULL; |
288 | 288 | ||
289 | rtab = kmalloc(sizeof(*rtab), GFP_KERNEL); | 289 | rtab = kmalloc(sizeof(*rtab), GFP_KERNEL); |
290 | if (rtab) { | 290 | if (rtab) { |
291 | rtab->rate = *r; | 291 | rtab->rate = *r; |
292 | rtab->refcnt = 1; | 292 | rtab->refcnt = 1; |
293 | memcpy(rtab->data, nla_data(tab), 1024); | 293 | memcpy(rtab->data, nla_data(tab), 1024); |
294 | rtab->next = qdisc_rtab_list; | 294 | rtab->next = qdisc_rtab_list; |
295 | qdisc_rtab_list = rtab; | 295 | qdisc_rtab_list = rtab; |
296 | } | 296 | } |
297 | return rtab; | 297 | return rtab; |
298 | } | 298 | } |
299 | EXPORT_SYMBOL(qdisc_get_rtab); | 299 | EXPORT_SYMBOL(qdisc_get_rtab); |
300 | 300 | ||
301 | void qdisc_put_rtab(struct qdisc_rate_table *tab) | 301 | void qdisc_put_rtab(struct qdisc_rate_table *tab) |
302 | { | 302 | { |
303 | struct qdisc_rate_table *rtab, **rtabp; | 303 | struct qdisc_rate_table *rtab, **rtabp; |
304 | 304 | ||
305 | if (!tab || --tab->refcnt) | 305 | if (!tab || --tab->refcnt) |
306 | return; | 306 | return; |
307 | 307 | ||
308 | for (rtabp = &qdisc_rtab_list; (rtab=*rtabp) != NULL; rtabp = &rtab->next) { | 308 | for (rtabp = &qdisc_rtab_list; (rtab=*rtabp) != NULL; rtabp = &rtab->next) { |
309 | if (rtab == tab) { | 309 | if (rtab == tab) { |
310 | *rtabp = rtab->next; | 310 | *rtabp = rtab->next; |
311 | kfree(rtab); | 311 | kfree(rtab); |
312 | return; | 312 | return; |
313 | } | 313 | } |
314 | } | 314 | } |
315 | } | 315 | } |
316 | EXPORT_SYMBOL(qdisc_put_rtab); | 316 | EXPORT_SYMBOL(qdisc_put_rtab); |
317 | 317 | ||
318 | static LIST_HEAD(qdisc_stab_list); | 318 | static LIST_HEAD(qdisc_stab_list); |
319 | static DEFINE_SPINLOCK(qdisc_stab_lock); | 319 | static DEFINE_SPINLOCK(qdisc_stab_lock); |
320 | 320 | ||
321 | static const struct nla_policy stab_policy[TCA_STAB_MAX + 1] = { | 321 | static const struct nla_policy stab_policy[TCA_STAB_MAX + 1] = { |
322 | [TCA_STAB_BASE] = { .len = sizeof(struct tc_sizespec) }, | 322 | [TCA_STAB_BASE] = { .len = sizeof(struct tc_sizespec) }, |
323 | [TCA_STAB_DATA] = { .type = NLA_BINARY }, | 323 | [TCA_STAB_DATA] = { .type = NLA_BINARY }, |
324 | }; | 324 | }; |
325 | 325 | ||
326 | static struct qdisc_size_table *qdisc_get_stab(struct nlattr *opt) | 326 | static struct qdisc_size_table *qdisc_get_stab(struct nlattr *opt) |
327 | { | 327 | { |
328 | struct nlattr *tb[TCA_STAB_MAX + 1]; | 328 | struct nlattr *tb[TCA_STAB_MAX + 1]; |
329 | struct qdisc_size_table *stab; | 329 | struct qdisc_size_table *stab; |
330 | struct tc_sizespec *s; | 330 | struct tc_sizespec *s; |
331 | unsigned int tsize = 0; | 331 | unsigned int tsize = 0; |
332 | u16 *tab = NULL; | 332 | u16 *tab = NULL; |
333 | int err; | 333 | int err; |
334 | 334 | ||
335 | err = nla_parse_nested(tb, TCA_STAB_MAX, opt, stab_policy); | 335 | err = nla_parse_nested(tb, TCA_STAB_MAX, opt, stab_policy); |
336 | if (err < 0) | 336 | if (err < 0) |
337 | return ERR_PTR(err); | 337 | return ERR_PTR(err); |
338 | if (!tb[TCA_STAB_BASE]) | 338 | if (!tb[TCA_STAB_BASE]) |
339 | return ERR_PTR(-EINVAL); | 339 | return ERR_PTR(-EINVAL); |
340 | 340 | ||
341 | s = nla_data(tb[TCA_STAB_BASE]); | 341 | s = nla_data(tb[TCA_STAB_BASE]); |
342 | 342 | ||
343 | if (s->tsize > 0) { | 343 | if (s->tsize > 0) { |
344 | if (!tb[TCA_STAB_DATA]) | 344 | if (!tb[TCA_STAB_DATA]) |
345 | return ERR_PTR(-EINVAL); | 345 | return ERR_PTR(-EINVAL); |
346 | tab = nla_data(tb[TCA_STAB_DATA]); | 346 | tab = nla_data(tb[TCA_STAB_DATA]); |
347 | tsize = nla_len(tb[TCA_STAB_DATA]) / sizeof(u16); | 347 | tsize = nla_len(tb[TCA_STAB_DATA]) / sizeof(u16); |
348 | } | 348 | } |
349 | 349 | ||
350 | if (!s || tsize != s->tsize || (!tab && tsize > 0)) | 350 | if (!s || tsize != s->tsize || (!tab && tsize > 0)) |
351 | return ERR_PTR(-EINVAL); | 351 | return ERR_PTR(-EINVAL); |
352 | 352 | ||
353 | spin_lock(&qdisc_stab_lock); | 353 | spin_lock(&qdisc_stab_lock); |
354 | 354 | ||
355 | list_for_each_entry(stab, &qdisc_stab_list, list) { | 355 | list_for_each_entry(stab, &qdisc_stab_list, list) { |
356 | if (memcmp(&stab->szopts, s, sizeof(*s))) | 356 | if (memcmp(&stab->szopts, s, sizeof(*s))) |
357 | continue; | 357 | continue; |
358 | if (tsize > 0 && memcmp(stab->data, tab, tsize * sizeof(u16))) | 358 | if (tsize > 0 && memcmp(stab->data, tab, tsize * sizeof(u16))) |
359 | continue; | 359 | continue; |
360 | stab->refcnt++; | 360 | stab->refcnt++; |
361 | spin_unlock(&qdisc_stab_lock); | 361 | spin_unlock(&qdisc_stab_lock); |
362 | return stab; | 362 | return stab; |
363 | } | 363 | } |
364 | 364 | ||
365 | spin_unlock(&qdisc_stab_lock); | 365 | spin_unlock(&qdisc_stab_lock); |
366 | 366 | ||
367 | stab = kmalloc(sizeof(*stab) + tsize * sizeof(u16), GFP_KERNEL); | 367 | stab = kmalloc(sizeof(*stab) + tsize * sizeof(u16), GFP_KERNEL); |
368 | if (!stab) | 368 | if (!stab) |
369 | return ERR_PTR(-ENOMEM); | 369 | return ERR_PTR(-ENOMEM); |
370 | 370 | ||
371 | stab->refcnt = 1; | 371 | stab->refcnt = 1; |
372 | stab->szopts = *s; | 372 | stab->szopts = *s; |
373 | if (tsize > 0) | 373 | if (tsize > 0) |
374 | memcpy(stab->data, tab, tsize * sizeof(u16)); | 374 | memcpy(stab->data, tab, tsize * sizeof(u16)); |
375 | 375 | ||
376 | spin_lock(&qdisc_stab_lock); | 376 | spin_lock(&qdisc_stab_lock); |
377 | list_add_tail(&stab->list, &qdisc_stab_list); | 377 | list_add_tail(&stab->list, &qdisc_stab_list); |
378 | spin_unlock(&qdisc_stab_lock); | 378 | spin_unlock(&qdisc_stab_lock); |
379 | 379 | ||
380 | return stab; | 380 | return stab; |
381 | } | 381 | } |
382 | 382 | ||
383 | void qdisc_put_stab(struct qdisc_size_table *tab) | 383 | void qdisc_put_stab(struct qdisc_size_table *tab) |
384 | { | 384 | { |
385 | if (!tab) | 385 | if (!tab) |
386 | return; | 386 | return; |
387 | 387 | ||
388 | spin_lock(&qdisc_stab_lock); | 388 | spin_lock(&qdisc_stab_lock); |
389 | 389 | ||
390 | if (--tab->refcnt == 0) { | 390 | if (--tab->refcnt == 0) { |
391 | list_del(&tab->list); | 391 | list_del(&tab->list); |
392 | kfree(tab); | 392 | kfree(tab); |
393 | } | 393 | } |
394 | 394 | ||
395 | spin_unlock(&qdisc_stab_lock); | 395 | spin_unlock(&qdisc_stab_lock); |
396 | } | 396 | } |
397 | EXPORT_SYMBOL(qdisc_put_stab); | 397 | EXPORT_SYMBOL(qdisc_put_stab); |
398 | 398 | ||
399 | static int qdisc_dump_stab(struct sk_buff *skb, struct qdisc_size_table *stab) | 399 | static int qdisc_dump_stab(struct sk_buff *skb, struct qdisc_size_table *stab) |
400 | { | 400 | { |
401 | struct nlattr *nest; | 401 | struct nlattr *nest; |
402 | 402 | ||
403 | nest = nla_nest_start(skb, TCA_STAB); | 403 | nest = nla_nest_start(skb, TCA_STAB); |
404 | if (nest == NULL) | 404 | if (nest == NULL) |
405 | goto nla_put_failure; | 405 | goto nla_put_failure; |
406 | NLA_PUT(skb, TCA_STAB_BASE, sizeof(stab->szopts), &stab->szopts); | 406 | NLA_PUT(skb, TCA_STAB_BASE, sizeof(stab->szopts), &stab->szopts); |
407 | nla_nest_end(skb, nest); | 407 | nla_nest_end(skb, nest); |
408 | 408 | ||
409 | return skb->len; | 409 | return skb->len; |
410 | 410 | ||
411 | nla_put_failure: | 411 | nla_put_failure: |
412 | return -1; | 412 | return -1; |
413 | } | 413 | } |
414 | 414 | ||
415 | void qdisc_calculate_pkt_len(struct sk_buff *skb, struct qdisc_size_table *stab) | 415 | void qdisc_calculate_pkt_len(struct sk_buff *skb, struct qdisc_size_table *stab) |
416 | { | 416 | { |
417 | int pkt_len, slot; | 417 | int pkt_len, slot; |
418 | 418 | ||
419 | pkt_len = skb->len + stab->szopts.overhead; | 419 | pkt_len = skb->len + stab->szopts.overhead; |
420 | if (unlikely(!stab->szopts.tsize)) | 420 | if (unlikely(!stab->szopts.tsize)) |
421 | goto out; | 421 | goto out; |
422 | 422 | ||
423 | slot = pkt_len + stab->szopts.cell_align; | 423 | slot = pkt_len + stab->szopts.cell_align; |
424 | if (unlikely(slot < 0)) | 424 | if (unlikely(slot < 0)) |
425 | slot = 0; | 425 | slot = 0; |
426 | 426 | ||
427 | slot >>= stab->szopts.cell_log; | 427 | slot >>= stab->szopts.cell_log; |
428 | if (likely(slot < stab->szopts.tsize)) | 428 | if (likely(slot < stab->szopts.tsize)) |
429 | pkt_len = stab->data[slot]; | 429 | pkt_len = stab->data[slot]; |
430 | else | 430 | else |
431 | pkt_len = stab->data[stab->szopts.tsize - 1] * | 431 | pkt_len = stab->data[stab->szopts.tsize - 1] * |
432 | (slot / stab->szopts.tsize) + | 432 | (slot / stab->szopts.tsize) + |
433 | stab->data[slot % stab->szopts.tsize]; | 433 | stab->data[slot % stab->szopts.tsize]; |
434 | 434 | ||
435 | pkt_len <<= stab->szopts.size_log; | 435 | pkt_len <<= stab->szopts.size_log; |
436 | out: | 436 | out: |
437 | if (unlikely(pkt_len < 1)) | 437 | if (unlikely(pkt_len < 1)) |
438 | pkt_len = 1; | 438 | pkt_len = 1; |
439 | qdisc_skb_cb(skb)->pkt_len = pkt_len; | 439 | qdisc_skb_cb(skb)->pkt_len = pkt_len; |
440 | } | 440 | } |
441 | EXPORT_SYMBOL(qdisc_calculate_pkt_len); | 441 | EXPORT_SYMBOL(qdisc_calculate_pkt_len); |
442 | 442 | ||
443 | void qdisc_warn_nonwc(char *txt, struct Qdisc *qdisc) | 443 | void qdisc_warn_nonwc(char *txt, struct Qdisc *qdisc) |
444 | { | 444 | { |
445 | if (!(qdisc->flags & TCQ_F_WARN_NONWC)) { | 445 | if (!(qdisc->flags & TCQ_F_WARN_NONWC)) { |
446 | printk(KERN_WARNING | 446 | printk(KERN_WARNING |
447 | "%s: %s qdisc %X: is non-work-conserving?\n", | 447 | "%s: %s qdisc %X: is non-work-conserving?\n", |
448 | txt, qdisc->ops->id, qdisc->handle >> 16); | 448 | txt, qdisc->ops->id, qdisc->handle >> 16); |
449 | qdisc->flags |= TCQ_F_WARN_NONWC; | 449 | qdisc->flags |= TCQ_F_WARN_NONWC; |
450 | } | 450 | } |
451 | } | 451 | } |
452 | EXPORT_SYMBOL(qdisc_warn_nonwc); | 452 | EXPORT_SYMBOL(qdisc_warn_nonwc); |
453 | 453 | ||
454 | static enum hrtimer_restart qdisc_watchdog(struct hrtimer *timer) | 454 | static enum hrtimer_restart qdisc_watchdog(struct hrtimer *timer) |
455 | { | 455 | { |
456 | struct qdisc_watchdog *wd = container_of(timer, struct qdisc_watchdog, | 456 | struct qdisc_watchdog *wd = container_of(timer, struct qdisc_watchdog, |
457 | timer); | 457 | timer); |
458 | 458 | ||
459 | wd->qdisc->flags &= ~TCQ_F_THROTTLED; | 459 | wd->qdisc->flags &= ~TCQ_F_THROTTLED; |
460 | __netif_schedule(qdisc_root(wd->qdisc)); | 460 | __netif_schedule(qdisc_root(wd->qdisc)); |
461 | 461 | ||
462 | return HRTIMER_NORESTART; | 462 | return HRTIMER_NORESTART; |
463 | } | 463 | } |
464 | 464 | ||
465 | void qdisc_watchdog_init(struct qdisc_watchdog *wd, struct Qdisc *qdisc) | 465 | void qdisc_watchdog_init(struct qdisc_watchdog *wd, struct Qdisc *qdisc) |
466 | { | 466 | { |
467 | hrtimer_init(&wd->timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS); | 467 | hrtimer_init(&wd->timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS); |
468 | wd->timer.function = qdisc_watchdog; | 468 | wd->timer.function = qdisc_watchdog; |
469 | wd->qdisc = qdisc; | 469 | wd->qdisc = qdisc; |
470 | } | 470 | } |
471 | EXPORT_SYMBOL(qdisc_watchdog_init); | 471 | EXPORT_SYMBOL(qdisc_watchdog_init); |
472 | 472 | ||
473 | void qdisc_watchdog_schedule(struct qdisc_watchdog *wd, psched_time_t expires) | 473 | void qdisc_watchdog_schedule(struct qdisc_watchdog *wd, psched_time_t expires) |
474 | { | 474 | { |
475 | ktime_t time; | 475 | ktime_t time; |
476 | 476 | ||
477 | if (test_bit(__QDISC_STATE_DEACTIVATED, | 477 | if (test_bit(__QDISC_STATE_DEACTIVATED, |
478 | &qdisc_root_sleeping(wd->qdisc)->state)) | 478 | &qdisc_root_sleeping(wd->qdisc)->state)) |
479 | return; | 479 | return; |
480 | 480 | ||
481 | wd->qdisc->flags |= TCQ_F_THROTTLED; | 481 | wd->qdisc->flags |= TCQ_F_THROTTLED; |
482 | time = ktime_set(0, 0); | 482 | time = ktime_set(0, 0); |
483 | time = ktime_add_ns(time, PSCHED_TICKS2NS(expires)); | 483 | time = ktime_add_ns(time, PSCHED_TICKS2NS(expires)); |
484 | hrtimer_start(&wd->timer, time, HRTIMER_MODE_ABS); | 484 | hrtimer_start(&wd->timer, time, HRTIMER_MODE_ABS); |
485 | } | 485 | } |
486 | EXPORT_SYMBOL(qdisc_watchdog_schedule); | 486 | EXPORT_SYMBOL(qdisc_watchdog_schedule); |
487 | 487 | ||
488 | void qdisc_watchdog_cancel(struct qdisc_watchdog *wd) | 488 | void qdisc_watchdog_cancel(struct qdisc_watchdog *wd) |
489 | { | 489 | { |
490 | hrtimer_cancel(&wd->timer); | 490 | hrtimer_cancel(&wd->timer); |
491 | wd->qdisc->flags &= ~TCQ_F_THROTTLED; | 491 | wd->qdisc->flags &= ~TCQ_F_THROTTLED; |
492 | } | 492 | } |
493 | EXPORT_SYMBOL(qdisc_watchdog_cancel); | 493 | EXPORT_SYMBOL(qdisc_watchdog_cancel); |
494 | 494 | ||
495 | static struct hlist_head *qdisc_class_hash_alloc(unsigned int n) | 495 | static struct hlist_head *qdisc_class_hash_alloc(unsigned int n) |
496 | { | 496 | { |
497 | unsigned int size = n * sizeof(struct hlist_head), i; | 497 | unsigned int size = n * sizeof(struct hlist_head), i; |
498 | struct hlist_head *h; | 498 | struct hlist_head *h; |
499 | 499 | ||
500 | if (size <= PAGE_SIZE) | 500 | if (size <= PAGE_SIZE) |
501 | h = kmalloc(size, GFP_KERNEL); | 501 | h = kmalloc(size, GFP_KERNEL); |
502 | else | 502 | else |
503 | h = (struct hlist_head *) | 503 | h = (struct hlist_head *) |
504 | __get_free_pages(GFP_KERNEL, get_order(size)); | 504 | __get_free_pages(GFP_KERNEL, get_order(size)); |
505 | 505 | ||
506 | if (h != NULL) { | 506 | if (h != NULL) { |
507 | for (i = 0; i < n; i++) | 507 | for (i = 0; i < n; i++) |
508 | INIT_HLIST_HEAD(&h[i]); | 508 | INIT_HLIST_HEAD(&h[i]); |
509 | } | 509 | } |
510 | return h; | 510 | return h; |
511 | } | 511 | } |
512 | 512 | ||
513 | static void qdisc_class_hash_free(struct hlist_head *h, unsigned int n) | 513 | static void qdisc_class_hash_free(struct hlist_head *h, unsigned int n) |
514 | { | 514 | { |
515 | unsigned int size = n * sizeof(struct hlist_head); | 515 | unsigned int size = n * sizeof(struct hlist_head); |
516 | 516 | ||
517 | if (size <= PAGE_SIZE) | 517 | if (size <= PAGE_SIZE) |
518 | kfree(h); | 518 | kfree(h); |
519 | else | 519 | else |
520 | free_pages((unsigned long)h, get_order(size)); | 520 | free_pages((unsigned long)h, get_order(size)); |
521 | } | 521 | } |
522 | 522 | ||
523 | void qdisc_class_hash_grow(struct Qdisc *sch, struct Qdisc_class_hash *clhash) | 523 | void qdisc_class_hash_grow(struct Qdisc *sch, struct Qdisc_class_hash *clhash) |
524 | { | 524 | { |
525 | struct Qdisc_class_common *cl; | 525 | struct Qdisc_class_common *cl; |
526 | struct hlist_node *n, *next; | 526 | struct hlist_node *n, *next; |
527 | struct hlist_head *nhash, *ohash; | 527 | struct hlist_head *nhash, *ohash; |
528 | unsigned int nsize, nmask, osize; | 528 | unsigned int nsize, nmask, osize; |
529 | unsigned int i, h; | 529 | unsigned int i, h; |
530 | 530 | ||
531 | /* Rehash when load factor exceeds 0.75 */ | 531 | /* Rehash when load factor exceeds 0.75 */ |
532 | if (clhash->hashelems * 4 <= clhash->hashsize * 3) | 532 | if (clhash->hashelems * 4 <= clhash->hashsize * 3) |
533 | return; | 533 | return; |
534 | nsize = clhash->hashsize * 2; | 534 | nsize = clhash->hashsize * 2; |
535 | nmask = nsize - 1; | 535 | nmask = nsize - 1; |
536 | nhash = qdisc_class_hash_alloc(nsize); | 536 | nhash = qdisc_class_hash_alloc(nsize); |
537 | if (nhash == NULL) | 537 | if (nhash == NULL) |
538 | return; | 538 | return; |
539 | 539 | ||
540 | ohash = clhash->hash; | 540 | ohash = clhash->hash; |
541 | osize = clhash->hashsize; | 541 | osize = clhash->hashsize; |
542 | 542 | ||
543 | sch_tree_lock(sch); | 543 | sch_tree_lock(sch); |
544 | for (i = 0; i < osize; i++) { | 544 | for (i = 0; i < osize; i++) { |
545 | hlist_for_each_entry_safe(cl, n, next, &ohash[i], hnode) { | 545 | hlist_for_each_entry_safe(cl, n, next, &ohash[i], hnode) { |
546 | h = qdisc_class_hash(cl->classid, nmask); | 546 | h = qdisc_class_hash(cl->classid, nmask); |
547 | hlist_add_head(&cl->hnode, &nhash[h]); | 547 | hlist_add_head(&cl->hnode, &nhash[h]); |
548 | } | 548 | } |
549 | } | 549 | } |
550 | clhash->hash = nhash; | 550 | clhash->hash = nhash; |
551 | clhash->hashsize = nsize; | 551 | clhash->hashsize = nsize; |
552 | clhash->hashmask = nmask; | 552 | clhash->hashmask = nmask; |
553 | sch_tree_unlock(sch); | 553 | sch_tree_unlock(sch); |
554 | 554 | ||
555 | qdisc_class_hash_free(ohash, osize); | 555 | qdisc_class_hash_free(ohash, osize); |
556 | } | 556 | } |
557 | EXPORT_SYMBOL(qdisc_class_hash_grow); | 557 | EXPORT_SYMBOL(qdisc_class_hash_grow); |
558 | 558 | ||
559 | int qdisc_class_hash_init(struct Qdisc_class_hash *clhash) | 559 | int qdisc_class_hash_init(struct Qdisc_class_hash *clhash) |
560 | { | 560 | { |
561 | unsigned int size = 4; | 561 | unsigned int size = 4; |
562 | 562 | ||
563 | clhash->hash = qdisc_class_hash_alloc(size); | 563 | clhash->hash = qdisc_class_hash_alloc(size); |
564 | if (clhash->hash == NULL) | 564 | if (clhash->hash == NULL) |
565 | return -ENOMEM; | 565 | return -ENOMEM; |
566 | clhash->hashsize = size; | 566 | clhash->hashsize = size; |
567 | clhash->hashmask = size - 1; | 567 | clhash->hashmask = size - 1; |
568 | clhash->hashelems = 0; | 568 | clhash->hashelems = 0; |
569 | return 0; | 569 | return 0; |
570 | } | 570 | } |
571 | EXPORT_SYMBOL(qdisc_class_hash_init); | 571 | EXPORT_SYMBOL(qdisc_class_hash_init); |
572 | 572 | ||
573 | void qdisc_class_hash_destroy(struct Qdisc_class_hash *clhash) | 573 | void qdisc_class_hash_destroy(struct Qdisc_class_hash *clhash) |
574 | { | 574 | { |
575 | qdisc_class_hash_free(clhash->hash, clhash->hashsize); | 575 | qdisc_class_hash_free(clhash->hash, clhash->hashsize); |
576 | } | 576 | } |
577 | EXPORT_SYMBOL(qdisc_class_hash_destroy); | 577 | EXPORT_SYMBOL(qdisc_class_hash_destroy); |
578 | 578 | ||
579 | void qdisc_class_hash_insert(struct Qdisc_class_hash *clhash, | 579 | void qdisc_class_hash_insert(struct Qdisc_class_hash *clhash, |
580 | struct Qdisc_class_common *cl) | 580 | struct Qdisc_class_common *cl) |
581 | { | 581 | { |
582 | unsigned int h; | 582 | unsigned int h; |
583 | 583 | ||
584 | INIT_HLIST_NODE(&cl->hnode); | 584 | INIT_HLIST_NODE(&cl->hnode); |
585 | h = qdisc_class_hash(cl->classid, clhash->hashmask); | 585 | h = qdisc_class_hash(cl->classid, clhash->hashmask); |
586 | hlist_add_head(&cl->hnode, &clhash->hash[h]); | 586 | hlist_add_head(&cl->hnode, &clhash->hash[h]); |
587 | clhash->hashelems++; | 587 | clhash->hashelems++; |
588 | } | 588 | } |
589 | EXPORT_SYMBOL(qdisc_class_hash_insert); | 589 | EXPORT_SYMBOL(qdisc_class_hash_insert); |
590 | 590 | ||
591 | void qdisc_class_hash_remove(struct Qdisc_class_hash *clhash, | 591 | void qdisc_class_hash_remove(struct Qdisc_class_hash *clhash, |
592 | struct Qdisc_class_common *cl) | 592 | struct Qdisc_class_common *cl) |
593 | { | 593 | { |
594 | hlist_del(&cl->hnode); | 594 | hlist_del(&cl->hnode); |
595 | clhash->hashelems--; | 595 | clhash->hashelems--; |
596 | } | 596 | } |
597 | EXPORT_SYMBOL(qdisc_class_hash_remove); | 597 | EXPORT_SYMBOL(qdisc_class_hash_remove); |
598 | 598 | ||
599 | /* Allocate an unique handle from space managed by kernel */ | 599 | /* Allocate an unique handle from space managed by kernel */ |
600 | 600 | ||
601 | static u32 qdisc_alloc_handle(struct net_device *dev) | 601 | static u32 qdisc_alloc_handle(struct net_device *dev) |
602 | { | 602 | { |
603 | int i = 0x10000; | 603 | int i = 0x10000; |
604 | static u32 autohandle = TC_H_MAKE(0x80000000U, 0); | 604 | static u32 autohandle = TC_H_MAKE(0x80000000U, 0); |
605 | 605 | ||
606 | do { | 606 | do { |
607 | autohandle += TC_H_MAKE(0x10000U, 0); | 607 | autohandle += TC_H_MAKE(0x10000U, 0); |
608 | if (autohandle == TC_H_MAKE(TC_H_ROOT, 0)) | 608 | if (autohandle == TC_H_MAKE(TC_H_ROOT, 0)) |
609 | autohandle = TC_H_MAKE(0x80000000U, 0); | 609 | autohandle = TC_H_MAKE(0x80000000U, 0); |
610 | } while (qdisc_lookup(dev, autohandle) && --i > 0); | 610 | } while (qdisc_lookup(dev, autohandle) && --i > 0); |
611 | 611 | ||
612 | return i>0 ? autohandle : 0; | 612 | return i>0 ? autohandle : 0; |
613 | } | 613 | } |
614 | 614 | ||
615 | void qdisc_tree_decrease_qlen(struct Qdisc *sch, unsigned int n) | 615 | void qdisc_tree_decrease_qlen(struct Qdisc *sch, unsigned int n) |
616 | { | 616 | { |
617 | const struct Qdisc_class_ops *cops; | 617 | const struct Qdisc_class_ops *cops; |
618 | unsigned long cl; | 618 | unsigned long cl; |
619 | u32 parentid; | 619 | u32 parentid; |
620 | 620 | ||
621 | if (n == 0) | 621 | if (n == 0) |
622 | return; | 622 | return; |
623 | while ((parentid = sch->parent)) { | 623 | while ((parentid = sch->parent)) { |
624 | if (TC_H_MAJ(parentid) == TC_H_MAJ(TC_H_INGRESS)) | 624 | if (TC_H_MAJ(parentid) == TC_H_MAJ(TC_H_INGRESS)) |
625 | return; | 625 | return; |
626 | 626 | ||
627 | sch = qdisc_lookup(qdisc_dev(sch), TC_H_MAJ(parentid)); | 627 | sch = qdisc_lookup(qdisc_dev(sch), TC_H_MAJ(parentid)); |
628 | if (sch == NULL) { | 628 | if (sch == NULL) { |
629 | WARN_ON(parentid != TC_H_ROOT); | 629 | WARN_ON(parentid != TC_H_ROOT); |
630 | return; | 630 | return; |
631 | } | 631 | } |
632 | cops = sch->ops->cl_ops; | 632 | cops = sch->ops->cl_ops; |
633 | if (cops->qlen_notify) { | 633 | if (cops->qlen_notify) { |
634 | cl = cops->get(sch, parentid); | 634 | cl = cops->get(sch, parentid); |
635 | cops->qlen_notify(sch, cl); | 635 | cops->qlen_notify(sch, cl); |
636 | cops->put(sch, cl); | 636 | cops->put(sch, cl); |
637 | } | 637 | } |
638 | sch->q.qlen -= n; | 638 | sch->q.qlen -= n; |
639 | } | 639 | } |
640 | } | 640 | } |
641 | EXPORT_SYMBOL(qdisc_tree_decrease_qlen); | 641 | EXPORT_SYMBOL(qdisc_tree_decrease_qlen); |
642 | 642 | ||
643 | static void notify_and_destroy(struct net *net, struct sk_buff *skb, | 643 | static void notify_and_destroy(struct net *net, struct sk_buff *skb, |
644 | struct nlmsghdr *n, u32 clid, | 644 | struct nlmsghdr *n, u32 clid, |
645 | struct Qdisc *old, struct Qdisc *new) | 645 | struct Qdisc *old, struct Qdisc *new) |
646 | { | 646 | { |
647 | if (new || old) | 647 | if (new || old) |
648 | qdisc_notify(net, skb, n, clid, old, new); | 648 | qdisc_notify(net, skb, n, clid, old, new); |
649 | 649 | ||
650 | if (old) | 650 | if (old) |
651 | qdisc_destroy(old); | 651 | qdisc_destroy(old); |
652 | } | 652 | } |
653 | 653 | ||
654 | /* Graft qdisc "new" to class "classid" of qdisc "parent" or | 654 | /* Graft qdisc "new" to class "classid" of qdisc "parent" or |
655 | * to device "dev". | 655 | * to device "dev". |
656 | * | 656 | * |
657 | * When appropriate send a netlink notification using 'skb' | 657 | * When appropriate send a netlink notification using 'skb' |
658 | * and "n". | 658 | * and "n". |
659 | * | 659 | * |
660 | * On success, destroy old qdisc. | 660 | * On success, destroy old qdisc. |
661 | */ | 661 | */ |
662 | 662 | ||
663 | static int qdisc_graft(struct net_device *dev, struct Qdisc *parent, | 663 | static int qdisc_graft(struct net_device *dev, struct Qdisc *parent, |
664 | struct sk_buff *skb, struct nlmsghdr *n, u32 classid, | 664 | struct sk_buff *skb, struct nlmsghdr *n, u32 classid, |
665 | struct Qdisc *new, struct Qdisc *old) | 665 | struct Qdisc *new, struct Qdisc *old) |
666 | { | 666 | { |
667 | struct Qdisc *q = old; | 667 | struct Qdisc *q = old; |
668 | struct net *net = dev_net(dev); | 668 | struct net *net = dev_net(dev); |
669 | int err = 0; | 669 | int err = 0; |
670 | 670 | ||
671 | if (parent == NULL) { | 671 | if (parent == NULL) { |
672 | unsigned int i, num_q, ingress; | 672 | unsigned int i, num_q, ingress; |
673 | 673 | ||
674 | ingress = 0; | 674 | ingress = 0; |
675 | num_q = dev->num_tx_queues; | 675 | num_q = dev->num_tx_queues; |
676 | if ((q && q->flags & TCQ_F_INGRESS) || | 676 | if ((q && q->flags & TCQ_F_INGRESS) || |
677 | (new && new->flags & TCQ_F_INGRESS)) { | 677 | (new && new->flags & TCQ_F_INGRESS)) { |
678 | num_q = 1; | 678 | num_q = 1; |
679 | ingress = 1; | 679 | ingress = 1; |
680 | } | 680 | } |
681 | 681 | ||
682 | if (dev->flags & IFF_UP) | 682 | if (dev->flags & IFF_UP) |
683 | dev_deactivate(dev); | 683 | dev_deactivate(dev); |
684 | 684 | ||
685 | if (new && new->ops->attach) { | 685 | if (new && new->ops->attach) { |
686 | new->ops->attach(new); | 686 | new->ops->attach(new); |
687 | num_q = 0; | 687 | num_q = 0; |
688 | } | 688 | } |
689 | 689 | ||
690 | for (i = 0; i < num_q; i++) { | 690 | for (i = 0; i < num_q; i++) { |
691 | struct netdev_queue *dev_queue = &dev->rx_queue; | 691 | struct netdev_queue *dev_queue = &dev->rx_queue; |
692 | 692 | ||
693 | if (!ingress) | 693 | if (!ingress) |
694 | dev_queue = netdev_get_tx_queue(dev, i); | 694 | dev_queue = netdev_get_tx_queue(dev, i); |
695 | 695 | ||
696 | old = dev_graft_qdisc(dev_queue, new); | 696 | old = dev_graft_qdisc(dev_queue, new); |
697 | if (new && i > 0) | 697 | if (new && i > 0) |
698 | atomic_inc(&new->refcnt); | 698 | atomic_inc(&new->refcnt); |
699 | 699 | ||
700 | if (!ingress) | 700 | if (!ingress) |
701 | qdisc_destroy(old); | 701 | qdisc_destroy(old); |
702 | } | 702 | } |
703 | 703 | ||
704 | if (!ingress) { | 704 | if (!ingress) { |
705 | notify_and_destroy(net, skb, n, classid, | 705 | notify_and_destroy(net, skb, n, classid, |
706 | dev->qdisc, new); | 706 | dev->qdisc, new); |
707 | if (new && !new->ops->attach) | 707 | if (new && !new->ops->attach) |
708 | atomic_inc(&new->refcnt); | 708 | atomic_inc(&new->refcnt); |
709 | dev->qdisc = new ? : &noop_qdisc; | 709 | dev->qdisc = new ? : &noop_qdisc; |
710 | } else { | 710 | } else { |
711 | notify_and_destroy(net, skb, n, classid, old, new); | 711 | notify_and_destroy(net, skb, n, classid, old, new); |
712 | } | 712 | } |
713 | 713 | ||
714 | if (dev->flags & IFF_UP) | 714 | if (dev->flags & IFF_UP) |
715 | dev_activate(dev); | 715 | dev_activate(dev); |
716 | } else { | 716 | } else { |
717 | const struct Qdisc_class_ops *cops = parent->ops->cl_ops; | 717 | const struct Qdisc_class_ops *cops = parent->ops->cl_ops; |
718 | 718 | ||
719 | err = -EOPNOTSUPP; | 719 | err = -EOPNOTSUPP; |
720 | if (cops && cops->graft) { | 720 | if (cops && cops->graft) { |
721 | unsigned long cl = cops->get(parent, classid); | 721 | unsigned long cl = cops->get(parent, classid); |
722 | if (cl) { | 722 | if (cl) { |
723 | err = cops->graft(parent, cl, new, &old); | 723 | err = cops->graft(parent, cl, new, &old); |
724 | cops->put(parent, cl); | 724 | cops->put(parent, cl); |
725 | } else | 725 | } else |
726 | err = -ENOENT; | 726 | err = -ENOENT; |
727 | } | 727 | } |
728 | if (!err) | 728 | if (!err) |
729 | notify_and_destroy(net, skb, n, classid, old, new); | 729 | notify_and_destroy(net, skb, n, classid, old, new); |
730 | } | 730 | } |
731 | return err; | 731 | return err; |
732 | } | 732 | } |
733 | 733 | ||
734 | /* lockdep annotation is needed for ingress; egress gets it only for name */ | 734 | /* lockdep annotation is needed for ingress; egress gets it only for name */ |
735 | static struct lock_class_key qdisc_tx_lock; | 735 | static struct lock_class_key qdisc_tx_lock; |
736 | static struct lock_class_key qdisc_rx_lock; | 736 | static struct lock_class_key qdisc_rx_lock; |
737 | 737 | ||
738 | /* | 738 | /* |
739 | Allocate and initialize new qdisc. | 739 | Allocate and initialize new qdisc. |
740 | 740 | ||
741 | Parameters are passed via opt. | 741 | Parameters are passed via opt. |
742 | */ | 742 | */ |
743 | 743 | ||
744 | static struct Qdisc * | 744 | static struct Qdisc * |
745 | qdisc_create(struct net_device *dev, struct netdev_queue *dev_queue, | 745 | qdisc_create(struct net_device *dev, struct netdev_queue *dev_queue, |
746 | struct Qdisc *p, u32 parent, u32 handle, | 746 | struct Qdisc *p, u32 parent, u32 handle, |
747 | struct nlattr **tca, int *errp) | 747 | struct nlattr **tca, int *errp) |
748 | { | 748 | { |
749 | int err; | 749 | int err; |
750 | struct nlattr *kind = tca[TCA_KIND]; | 750 | struct nlattr *kind = tca[TCA_KIND]; |
751 | struct Qdisc *sch; | 751 | struct Qdisc *sch; |
752 | struct Qdisc_ops *ops; | 752 | struct Qdisc_ops *ops; |
753 | struct qdisc_size_table *stab; | 753 | struct qdisc_size_table *stab; |
754 | 754 | ||
755 | ops = qdisc_lookup_ops(kind); | 755 | ops = qdisc_lookup_ops(kind); |
756 | #ifdef CONFIG_MODULES | 756 | #ifdef CONFIG_MODULES |
757 | if (ops == NULL && kind != NULL) { | 757 | if (ops == NULL && kind != NULL) { |
758 | char name[IFNAMSIZ]; | 758 | char name[IFNAMSIZ]; |
759 | if (nla_strlcpy(name, kind, IFNAMSIZ) < IFNAMSIZ) { | 759 | if (nla_strlcpy(name, kind, IFNAMSIZ) < IFNAMSIZ) { |
760 | /* We dropped the RTNL semaphore in order to | 760 | /* We dropped the RTNL semaphore in order to |
761 | * perform the module load. So, even if we | 761 | * perform the module load. So, even if we |
762 | * succeeded in loading the module we have to | 762 | * succeeded in loading the module we have to |
763 | * tell the caller to replay the request. We | 763 | * tell the caller to replay the request. We |
764 | * indicate this using -EAGAIN. | 764 | * indicate this using -EAGAIN. |
765 | * We replay the request because the device may | 765 | * We replay the request because the device may |
766 | * go away in the mean time. | 766 | * go away in the mean time. |
767 | */ | 767 | */ |
768 | rtnl_unlock(); | 768 | rtnl_unlock(); |
769 | request_module("sch_%s", name); | 769 | request_module("sch_%s", name); |
770 | rtnl_lock(); | 770 | rtnl_lock(); |
771 | ops = qdisc_lookup_ops(kind); | 771 | ops = qdisc_lookup_ops(kind); |
772 | if (ops != NULL) { | 772 | if (ops != NULL) { |
773 | /* We will try again qdisc_lookup_ops, | 773 | /* We will try again qdisc_lookup_ops, |
774 | * so don't keep a reference. | 774 | * so don't keep a reference. |
775 | */ | 775 | */ |
776 | module_put(ops->owner); | 776 | module_put(ops->owner); |
777 | err = -EAGAIN; | 777 | err = -EAGAIN; |
778 | goto err_out; | 778 | goto err_out; |
779 | } | 779 | } |
780 | } | 780 | } |
781 | } | 781 | } |
782 | #endif | 782 | #endif |
783 | 783 | ||
784 | err = -ENOENT; | 784 | err = -ENOENT; |
785 | if (ops == NULL) | 785 | if (ops == NULL) |
786 | goto err_out; | 786 | goto err_out; |
787 | 787 | ||
788 | sch = qdisc_alloc(dev_queue, ops); | 788 | sch = qdisc_alloc(dev_queue, ops); |
789 | if (IS_ERR(sch)) { | 789 | if (IS_ERR(sch)) { |
790 | err = PTR_ERR(sch); | 790 | err = PTR_ERR(sch); |
791 | goto err_out2; | 791 | goto err_out2; |
792 | } | 792 | } |
793 | 793 | ||
794 | sch->parent = parent; | 794 | sch->parent = parent; |
795 | 795 | ||
796 | if (handle == TC_H_INGRESS) { | 796 | if (handle == TC_H_INGRESS) { |
797 | sch->flags |= TCQ_F_INGRESS; | 797 | sch->flags |= TCQ_F_INGRESS; |
798 | handle = TC_H_MAKE(TC_H_INGRESS, 0); | 798 | handle = TC_H_MAKE(TC_H_INGRESS, 0); |
799 | lockdep_set_class(qdisc_lock(sch), &qdisc_rx_lock); | 799 | lockdep_set_class(qdisc_lock(sch), &qdisc_rx_lock); |
800 | } else { | 800 | } else { |
801 | if (handle == 0) { | 801 | if (handle == 0) { |
802 | handle = qdisc_alloc_handle(dev); | 802 | handle = qdisc_alloc_handle(dev); |
803 | err = -ENOMEM; | 803 | err = -ENOMEM; |
804 | if (handle == 0) | 804 | if (handle == 0) |
805 | goto err_out3; | 805 | goto err_out3; |
806 | } | 806 | } |
807 | lockdep_set_class(qdisc_lock(sch), &qdisc_tx_lock); | 807 | lockdep_set_class(qdisc_lock(sch), &qdisc_tx_lock); |
808 | } | 808 | } |
809 | 809 | ||
810 | sch->handle = handle; | 810 | sch->handle = handle; |
811 | 811 | ||
812 | if (!ops->init || (err = ops->init(sch, tca[TCA_OPTIONS])) == 0) { | 812 | if (!ops->init || (err = ops->init(sch, tca[TCA_OPTIONS])) == 0) { |
813 | if (tca[TCA_STAB]) { | 813 | if (tca[TCA_STAB]) { |
814 | stab = qdisc_get_stab(tca[TCA_STAB]); | 814 | stab = qdisc_get_stab(tca[TCA_STAB]); |
815 | if (IS_ERR(stab)) { | 815 | if (IS_ERR(stab)) { |
816 | err = PTR_ERR(stab); | 816 | err = PTR_ERR(stab); |
817 | goto err_out4; | 817 | goto err_out4; |
818 | } | 818 | } |
819 | sch->stab = stab; | 819 | sch->stab = stab; |
820 | } | 820 | } |
821 | if (tca[TCA_RATE]) { | 821 | if (tca[TCA_RATE]) { |
822 | spinlock_t *root_lock; | 822 | spinlock_t *root_lock; |
823 | 823 | ||
824 | err = -EOPNOTSUPP; | 824 | err = -EOPNOTSUPP; |
825 | if (sch->flags & TCQ_F_MQROOT) | 825 | if (sch->flags & TCQ_F_MQROOT) |
826 | goto err_out4; | 826 | goto err_out4; |
827 | 827 | ||
828 | if ((sch->parent != TC_H_ROOT) && | 828 | if ((sch->parent != TC_H_ROOT) && |
829 | !(sch->flags & TCQ_F_INGRESS) && | 829 | !(sch->flags & TCQ_F_INGRESS) && |
830 | (!p || !(p->flags & TCQ_F_MQROOT))) | 830 | (!p || !(p->flags & TCQ_F_MQROOT))) |
831 | root_lock = qdisc_root_sleeping_lock(sch); | 831 | root_lock = qdisc_root_sleeping_lock(sch); |
832 | else | 832 | else |
833 | root_lock = qdisc_lock(sch); | 833 | root_lock = qdisc_lock(sch); |
834 | 834 | ||
835 | err = gen_new_estimator(&sch->bstats, &sch->rate_est, | 835 | err = gen_new_estimator(&sch->bstats, &sch->rate_est, |
836 | root_lock, tca[TCA_RATE]); | 836 | root_lock, tca[TCA_RATE]); |
837 | if (err) | 837 | if (err) |
838 | goto err_out4; | 838 | goto err_out4; |
839 | } | 839 | } |
840 | 840 | ||
841 | qdisc_list_add(sch); | 841 | qdisc_list_add(sch); |
842 | 842 | ||
843 | return sch; | 843 | return sch; |
844 | } | 844 | } |
845 | err_out3: | 845 | err_out3: |
846 | dev_put(dev); | 846 | dev_put(dev); |
847 | kfree((char *) sch - sch->padded); | 847 | kfree((char *) sch - sch->padded); |
848 | err_out2: | 848 | err_out2: |
849 | module_put(ops->owner); | 849 | module_put(ops->owner); |
850 | err_out: | 850 | err_out: |
851 | *errp = err; | 851 | *errp = err; |
852 | return NULL; | 852 | return NULL; |
853 | 853 | ||
854 | err_out4: | 854 | err_out4: |
855 | /* | 855 | /* |
856 | * Any broken qdiscs that would require a ops->reset() here? | 856 | * Any broken qdiscs that would require a ops->reset() here? |
857 | * The qdisc was never in action so it shouldn't be necessary. | 857 | * The qdisc was never in action so it shouldn't be necessary. |
858 | */ | 858 | */ |
859 | qdisc_put_stab(sch->stab); | 859 | qdisc_put_stab(sch->stab); |
860 | if (ops->destroy) | 860 | if (ops->destroy) |
861 | ops->destroy(sch); | 861 | ops->destroy(sch); |
862 | goto err_out3; | 862 | goto err_out3; |
863 | } | 863 | } |
864 | 864 | ||
865 | static int qdisc_change(struct Qdisc *sch, struct nlattr **tca) | 865 | static int qdisc_change(struct Qdisc *sch, struct nlattr **tca) |
866 | { | 866 | { |
867 | struct qdisc_size_table *stab = NULL; | 867 | struct qdisc_size_table *stab = NULL; |
868 | int err = 0; | 868 | int err = 0; |
869 | 869 | ||
870 | if (tca[TCA_OPTIONS]) { | 870 | if (tca[TCA_OPTIONS]) { |
871 | if (sch->ops->change == NULL) | 871 | if (sch->ops->change == NULL) |
872 | return -EINVAL; | 872 | return -EINVAL; |
873 | err = sch->ops->change(sch, tca[TCA_OPTIONS]); | 873 | err = sch->ops->change(sch, tca[TCA_OPTIONS]); |
874 | if (err) | 874 | if (err) |
875 | return err; | 875 | return err; |
876 | } | 876 | } |
877 | 877 | ||
878 | if (tca[TCA_STAB]) { | 878 | if (tca[TCA_STAB]) { |
879 | stab = qdisc_get_stab(tca[TCA_STAB]); | 879 | stab = qdisc_get_stab(tca[TCA_STAB]); |
880 | if (IS_ERR(stab)) | 880 | if (IS_ERR(stab)) |
881 | return PTR_ERR(stab); | 881 | return PTR_ERR(stab); |
882 | } | 882 | } |
883 | 883 | ||
884 | qdisc_put_stab(sch->stab); | 884 | qdisc_put_stab(sch->stab); |
885 | sch->stab = stab; | 885 | sch->stab = stab; |
886 | 886 | ||
887 | if (tca[TCA_RATE]) { | 887 | if (tca[TCA_RATE]) { |
888 | /* NB: ignores errors from replace_estimator | 888 | /* NB: ignores errors from replace_estimator |
889 | because change can't be undone. */ | 889 | because change can't be undone. */ |
890 | if (sch->flags & TCQ_F_MQROOT) | 890 | if (sch->flags & TCQ_F_MQROOT) |
891 | goto out; | 891 | goto out; |
892 | gen_replace_estimator(&sch->bstats, &sch->rate_est, | 892 | gen_replace_estimator(&sch->bstats, &sch->rate_est, |
893 | qdisc_root_sleeping_lock(sch), | 893 | qdisc_root_sleeping_lock(sch), |
894 | tca[TCA_RATE]); | 894 | tca[TCA_RATE]); |
895 | } | 895 | } |
896 | out: | 896 | out: |
897 | return 0; | 897 | return 0; |
898 | } | 898 | } |
899 | 899 | ||
900 | struct check_loop_arg | 900 | struct check_loop_arg |
901 | { | 901 | { |
902 | struct qdisc_walker w; | 902 | struct qdisc_walker w; |
903 | struct Qdisc *p; | 903 | struct Qdisc *p; |
904 | int depth; | 904 | int depth; |
905 | }; | 905 | }; |
906 | 906 | ||
907 | static int check_loop_fn(struct Qdisc *q, unsigned long cl, struct qdisc_walker *w); | 907 | static int check_loop_fn(struct Qdisc *q, unsigned long cl, struct qdisc_walker *w); |
908 | 908 | ||
909 | static int check_loop(struct Qdisc *q, struct Qdisc *p, int depth) | 909 | static int check_loop(struct Qdisc *q, struct Qdisc *p, int depth) |
910 | { | 910 | { |
911 | struct check_loop_arg arg; | 911 | struct check_loop_arg arg; |
912 | 912 | ||
913 | if (q->ops->cl_ops == NULL) | 913 | if (q->ops->cl_ops == NULL) |
914 | return 0; | 914 | return 0; |
915 | 915 | ||
916 | arg.w.stop = arg.w.skip = arg.w.count = 0; | 916 | arg.w.stop = arg.w.skip = arg.w.count = 0; |
917 | arg.w.fn = check_loop_fn; | 917 | arg.w.fn = check_loop_fn; |
918 | arg.depth = depth; | 918 | arg.depth = depth; |
919 | arg.p = p; | 919 | arg.p = p; |
920 | q->ops->cl_ops->walk(q, &arg.w); | 920 | q->ops->cl_ops->walk(q, &arg.w); |
921 | return arg.w.stop ? -ELOOP : 0; | 921 | return arg.w.stop ? -ELOOP : 0; |
922 | } | 922 | } |
923 | 923 | ||
924 | static int | 924 | static int |
925 | check_loop_fn(struct Qdisc *q, unsigned long cl, struct qdisc_walker *w) | 925 | check_loop_fn(struct Qdisc *q, unsigned long cl, struct qdisc_walker *w) |
926 | { | 926 | { |
927 | struct Qdisc *leaf; | 927 | struct Qdisc *leaf; |
928 | const struct Qdisc_class_ops *cops = q->ops->cl_ops; | 928 | const struct Qdisc_class_ops *cops = q->ops->cl_ops; |
929 | struct check_loop_arg *arg = (struct check_loop_arg *)w; | 929 | struct check_loop_arg *arg = (struct check_loop_arg *)w; |
930 | 930 | ||
931 | leaf = cops->leaf(q, cl); | 931 | leaf = cops->leaf(q, cl); |
932 | if (leaf) { | 932 | if (leaf) { |
933 | if (leaf == arg->p || arg->depth > 7) | 933 | if (leaf == arg->p || arg->depth > 7) |
934 | return -ELOOP; | 934 | return -ELOOP; |
935 | return check_loop(leaf, arg->p, arg->depth + 1); | 935 | return check_loop(leaf, arg->p, arg->depth + 1); |
936 | } | 936 | } |
937 | return 0; | 937 | return 0; |
938 | } | 938 | } |
939 | 939 | ||
940 | /* | 940 | /* |
941 | * Delete/get qdisc. | 941 | * Delete/get qdisc. |
942 | */ | 942 | */ |
943 | 943 | ||
944 | static int tc_get_qdisc(struct sk_buff *skb, struct nlmsghdr *n, void *arg) | 944 | static int tc_get_qdisc(struct sk_buff *skb, struct nlmsghdr *n, void *arg) |
945 | { | 945 | { |
946 | struct net *net = sock_net(skb->sk); | 946 | struct net *net = sock_net(skb->sk); |
947 | struct tcmsg *tcm = NLMSG_DATA(n); | 947 | struct tcmsg *tcm = NLMSG_DATA(n); |
948 | struct nlattr *tca[TCA_MAX + 1]; | 948 | struct nlattr *tca[TCA_MAX + 1]; |
949 | struct net_device *dev; | 949 | struct net_device *dev; |
950 | u32 clid = tcm->tcm_parent; | 950 | u32 clid = tcm->tcm_parent; |
951 | struct Qdisc *q = NULL; | 951 | struct Qdisc *q = NULL; |
952 | struct Qdisc *p = NULL; | 952 | struct Qdisc *p = NULL; |
953 | int err; | 953 | int err; |
954 | 954 | ||
955 | if ((dev = __dev_get_by_index(net, tcm->tcm_ifindex)) == NULL) | 955 | if ((dev = __dev_get_by_index(net, tcm->tcm_ifindex)) == NULL) |
956 | return -ENODEV; | 956 | return -ENODEV; |
957 | 957 | ||
958 | err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL); | 958 | err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL); |
959 | if (err < 0) | 959 | if (err < 0) |
960 | return err; | 960 | return err; |
961 | 961 | ||
962 | if (clid) { | 962 | if (clid) { |
963 | if (clid != TC_H_ROOT) { | 963 | if (clid != TC_H_ROOT) { |
964 | if (TC_H_MAJ(clid) != TC_H_MAJ(TC_H_INGRESS)) { | 964 | if (TC_H_MAJ(clid) != TC_H_MAJ(TC_H_INGRESS)) { |
965 | if ((p = qdisc_lookup(dev, TC_H_MAJ(clid))) == NULL) | 965 | if ((p = qdisc_lookup(dev, TC_H_MAJ(clid))) == NULL) |
966 | return -ENOENT; | 966 | return -ENOENT; |
967 | q = qdisc_leaf(p, clid); | 967 | q = qdisc_leaf(p, clid); |
968 | } else { /* ingress */ | 968 | } else { /* ingress */ |
969 | q = dev->rx_queue.qdisc_sleeping; | 969 | q = dev->rx_queue.qdisc_sleeping; |
970 | } | 970 | } |
971 | } else { | 971 | } else { |
972 | q = dev->qdisc; | 972 | q = dev->qdisc; |
973 | } | 973 | } |
974 | if (!q) | 974 | if (!q) |
975 | return -ENOENT; | 975 | return -ENOENT; |
976 | 976 | ||
977 | if (tcm->tcm_handle && q->handle != tcm->tcm_handle) | 977 | if (tcm->tcm_handle && q->handle != tcm->tcm_handle) |
978 | return -EINVAL; | 978 | return -EINVAL; |
979 | } else { | 979 | } else { |
980 | if ((q = qdisc_lookup(dev, tcm->tcm_handle)) == NULL) | 980 | if ((q = qdisc_lookup(dev, tcm->tcm_handle)) == NULL) |
981 | return -ENOENT; | 981 | return -ENOENT; |
982 | } | 982 | } |
983 | 983 | ||
984 | if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id)) | 984 | if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id)) |
985 | return -EINVAL; | 985 | return -EINVAL; |
986 | 986 | ||
987 | if (n->nlmsg_type == RTM_DELQDISC) { | 987 | if (n->nlmsg_type == RTM_DELQDISC) { |
988 | if (!clid) | 988 | if (!clid) |
989 | return -EINVAL; | 989 | return -EINVAL; |
990 | if (q->handle == 0) | 990 | if (q->handle == 0) |
991 | return -ENOENT; | 991 | return -ENOENT; |
992 | if ((err = qdisc_graft(dev, p, skb, n, clid, NULL, q)) != 0) | 992 | if ((err = qdisc_graft(dev, p, skb, n, clid, NULL, q)) != 0) |
993 | return err; | 993 | return err; |
994 | } else { | 994 | } else { |
995 | qdisc_notify(net, skb, n, clid, NULL, q); | 995 | qdisc_notify(net, skb, n, clid, NULL, q); |
996 | } | 996 | } |
997 | return 0; | 997 | return 0; |
998 | } | 998 | } |
999 | 999 | ||
1000 | /* | 1000 | /* |
1001 | Create/change qdisc. | 1001 | Create/change qdisc. |
1002 | */ | 1002 | */ |
1003 | 1003 | ||
1004 | static int tc_modify_qdisc(struct sk_buff *skb, struct nlmsghdr *n, void *arg) | 1004 | static int tc_modify_qdisc(struct sk_buff *skb, struct nlmsghdr *n, void *arg) |
1005 | { | 1005 | { |
1006 | struct net *net = sock_net(skb->sk); | 1006 | struct net *net = sock_net(skb->sk); |
1007 | struct tcmsg *tcm; | 1007 | struct tcmsg *tcm; |
1008 | struct nlattr *tca[TCA_MAX + 1]; | 1008 | struct nlattr *tca[TCA_MAX + 1]; |
1009 | struct net_device *dev; | 1009 | struct net_device *dev; |
1010 | u32 clid; | 1010 | u32 clid; |
1011 | struct Qdisc *q, *p; | 1011 | struct Qdisc *q, *p; |
1012 | int err; | 1012 | int err; |
1013 | 1013 | ||
1014 | replay: | 1014 | replay: |
1015 | /* Reinit, just in case something touches this. */ | 1015 | /* Reinit, just in case something touches this. */ |
1016 | tcm = NLMSG_DATA(n); | 1016 | tcm = NLMSG_DATA(n); |
1017 | clid = tcm->tcm_parent; | 1017 | clid = tcm->tcm_parent; |
1018 | q = p = NULL; | 1018 | q = p = NULL; |
1019 | 1019 | ||
1020 | if ((dev = __dev_get_by_index(net, tcm->tcm_ifindex)) == NULL) | 1020 | if ((dev = __dev_get_by_index(net, tcm->tcm_ifindex)) == NULL) |
1021 | return -ENODEV; | 1021 | return -ENODEV; |
1022 | 1022 | ||
1023 | err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL); | 1023 | err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL); |
1024 | if (err < 0) | 1024 | if (err < 0) |
1025 | return err; | 1025 | return err; |
1026 | 1026 | ||
1027 | if (clid) { | 1027 | if (clid) { |
1028 | if (clid != TC_H_ROOT) { | 1028 | if (clid != TC_H_ROOT) { |
1029 | if (clid != TC_H_INGRESS) { | 1029 | if (clid != TC_H_INGRESS) { |
1030 | if ((p = qdisc_lookup(dev, TC_H_MAJ(clid))) == NULL) | 1030 | if ((p = qdisc_lookup(dev, TC_H_MAJ(clid))) == NULL) |
1031 | return -ENOENT; | 1031 | return -ENOENT; |
1032 | q = qdisc_leaf(p, clid); | 1032 | q = qdisc_leaf(p, clid); |
1033 | } else { /*ingress */ | 1033 | } else { /*ingress */ |
1034 | q = dev->rx_queue.qdisc_sleeping; | 1034 | q = dev->rx_queue.qdisc_sleeping; |
1035 | } | 1035 | } |
1036 | } else { | 1036 | } else { |
1037 | q = dev->qdisc; | 1037 | q = dev->qdisc; |
1038 | } | 1038 | } |
1039 | 1039 | ||
1040 | /* It may be default qdisc, ignore it */ | 1040 | /* It may be default qdisc, ignore it */ |
1041 | if (q && q->handle == 0) | 1041 | if (q && q->handle == 0) |
1042 | q = NULL; | 1042 | q = NULL; |
1043 | 1043 | ||
1044 | if (!q || !tcm->tcm_handle || q->handle != tcm->tcm_handle) { | 1044 | if (!q || !tcm->tcm_handle || q->handle != tcm->tcm_handle) { |
1045 | if (tcm->tcm_handle) { | 1045 | if (tcm->tcm_handle) { |
1046 | if (q && !(n->nlmsg_flags&NLM_F_REPLACE)) | 1046 | if (q && !(n->nlmsg_flags&NLM_F_REPLACE)) |
1047 | return -EEXIST; | 1047 | return -EEXIST; |
1048 | if (TC_H_MIN(tcm->tcm_handle)) | 1048 | if (TC_H_MIN(tcm->tcm_handle)) |
1049 | return -EINVAL; | 1049 | return -EINVAL; |
1050 | if ((q = qdisc_lookup(dev, tcm->tcm_handle)) == NULL) | 1050 | if ((q = qdisc_lookup(dev, tcm->tcm_handle)) == NULL) |
1051 | goto create_n_graft; | 1051 | goto create_n_graft; |
1052 | if (n->nlmsg_flags&NLM_F_EXCL) | 1052 | if (n->nlmsg_flags&NLM_F_EXCL) |
1053 | return -EEXIST; | 1053 | return -EEXIST; |
1054 | if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id)) | 1054 | if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id)) |
1055 | return -EINVAL; | 1055 | return -EINVAL; |
1056 | if (q == p || | 1056 | if (q == p || |
1057 | (p && check_loop(q, p, 0))) | 1057 | (p && check_loop(q, p, 0))) |
1058 | return -ELOOP; | 1058 | return -ELOOP; |
1059 | atomic_inc(&q->refcnt); | 1059 | atomic_inc(&q->refcnt); |
1060 | goto graft; | 1060 | goto graft; |
1061 | } else { | 1061 | } else { |
1062 | if (q == NULL) | 1062 | if (q == NULL) |
1063 | goto create_n_graft; | 1063 | goto create_n_graft; |
1064 | 1064 | ||
1065 | /* This magic test requires explanation. | 1065 | /* This magic test requires explanation. |
1066 | * | 1066 | * |
1067 | * We know, that some child q is already | 1067 | * We know, that some child q is already |
1068 | * attached to this parent and have choice: | 1068 | * attached to this parent and have choice: |
1069 | * either to change it or to create/graft new one. | 1069 | * either to change it or to create/graft new one. |
1070 | * | 1070 | * |
1071 | * 1. We are allowed to create/graft only | 1071 | * 1. We are allowed to create/graft only |
1072 | * if CREATE and REPLACE flags are set. | 1072 | * if CREATE and REPLACE flags are set. |
1073 | * | 1073 | * |
1074 | * 2. If EXCL is set, requestor wanted to say, | 1074 | * 2. If EXCL is set, requestor wanted to say, |
1075 | * that qdisc tcm_handle is not expected | 1075 | * that qdisc tcm_handle is not expected |
1076 | * to exist, so that we choose create/graft too. | 1076 | * to exist, so that we choose create/graft too. |
1077 | * | 1077 | * |
1078 | * 3. The last case is when no flags are set. | 1078 | * 3. The last case is when no flags are set. |
1079 | * Alas, it is sort of hole in API, we | 1079 | * Alas, it is sort of hole in API, we |
1080 | * cannot decide what to do unambiguously. | 1080 | * cannot decide what to do unambiguously. |
1081 | * For now we select create/graft, if | 1081 | * For now we select create/graft, if |
1082 | * user gave KIND, which does not match existing. | 1082 | * user gave KIND, which does not match existing. |
1083 | */ | 1083 | */ |
1084 | if ((n->nlmsg_flags&NLM_F_CREATE) && | 1084 | if ((n->nlmsg_flags&NLM_F_CREATE) && |
1085 | (n->nlmsg_flags&NLM_F_REPLACE) && | 1085 | (n->nlmsg_flags&NLM_F_REPLACE) && |
1086 | ((n->nlmsg_flags&NLM_F_EXCL) || | 1086 | ((n->nlmsg_flags&NLM_F_EXCL) || |
1087 | (tca[TCA_KIND] && | 1087 | (tca[TCA_KIND] && |
1088 | nla_strcmp(tca[TCA_KIND], q->ops->id)))) | 1088 | nla_strcmp(tca[TCA_KIND], q->ops->id)))) |
1089 | goto create_n_graft; | 1089 | goto create_n_graft; |
1090 | } | 1090 | } |
1091 | } | 1091 | } |
1092 | } else { | 1092 | } else { |
1093 | if (!tcm->tcm_handle) | 1093 | if (!tcm->tcm_handle) |
1094 | return -EINVAL; | 1094 | return -EINVAL; |
1095 | q = qdisc_lookup(dev, tcm->tcm_handle); | 1095 | q = qdisc_lookup(dev, tcm->tcm_handle); |
1096 | } | 1096 | } |
1097 | 1097 | ||
1098 | /* Change qdisc parameters */ | 1098 | /* Change qdisc parameters */ |
1099 | if (q == NULL) | 1099 | if (q == NULL) |
1100 | return -ENOENT; | 1100 | return -ENOENT; |
1101 | if (n->nlmsg_flags&NLM_F_EXCL) | 1101 | if (n->nlmsg_flags&NLM_F_EXCL) |
1102 | return -EEXIST; | 1102 | return -EEXIST; |
1103 | if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id)) | 1103 | if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id)) |
1104 | return -EINVAL; | 1104 | return -EINVAL; |
1105 | err = qdisc_change(q, tca); | 1105 | err = qdisc_change(q, tca); |
1106 | if (err == 0) | 1106 | if (err == 0) |
1107 | qdisc_notify(net, skb, n, clid, NULL, q); | 1107 | qdisc_notify(net, skb, n, clid, NULL, q); |
1108 | return err; | 1108 | return err; |
1109 | 1109 | ||
1110 | create_n_graft: | 1110 | create_n_graft: |
1111 | if (!(n->nlmsg_flags&NLM_F_CREATE)) | 1111 | if (!(n->nlmsg_flags&NLM_F_CREATE)) |
1112 | return -ENOENT; | 1112 | return -ENOENT; |
1113 | if (clid == TC_H_INGRESS) | 1113 | if (clid == TC_H_INGRESS) |
1114 | q = qdisc_create(dev, &dev->rx_queue, p, | 1114 | q = qdisc_create(dev, &dev->rx_queue, p, |
1115 | tcm->tcm_parent, tcm->tcm_parent, | 1115 | tcm->tcm_parent, tcm->tcm_parent, |
1116 | tca, &err); | 1116 | tca, &err); |
1117 | else { | 1117 | else { |
1118 | struct netdev_queue *dev_queue; | 1118 | struct netdev_queue *dev_queue; |
1119 | 1119 | ||
1120 | if (p && p->ops->cl_ops && p->ops->cl_ops->select_queue) | 1120 | if (p && p->ops->cl_ops && p->ops->cl_ops->select_queue) |
1121 | dev_queue = p->ops->cl_ops->select_queue(p, tcm); | 1121 | dev_queue = p->ops->cl_ops->select_queue(p, tcm); |
1122 | else if (p) | 1122 | else if (p) |
1123 | dev_queue = p->dev_queue; | 1123 | dev_queue = p->dev_queue; |
1124 | else | 1124 | else |
1125 | dev_queue = netdev_get_tx_queue(dev, 0); | 1125 | dev_queue = netdev_get_tx_queue(dev, 0); |
1126 | 1126 | ||
1127 | q = qdisc_create(dev, dev_queue, p, | 1127 | q = qdisc_create(dev, dev_queue, p, |
1128 | tcm->tcm_parent, tcm->tcm_handle, | 1128 | tcm->tcm_parent, tcm->tcm_handle, |
1129 | tca, &err); | 1129 | tca, &err); |
1130 | } | 1130 | } |
1131 | if (q == NULL) { | 1131 | if (q == NULL) { |
1132 | if (err == -EAGAIN) | 1132 | if (err == -EAGAIN) |
1133 | goto replay; | 1133 | goto replay; |
1134 | return err; | 1134 | return err; |
1135 | } | 1135 | } |
1136 | 1136 | ||
1137 | graft: | 1137 | graft: |
1138 | err = qdisc_graft(dev, p, skb, n, clid, q, NULL); | 1138 | err = qdisc_graft(dev, p, skb, n, clid, q, NULL); |
1139 | if (err) { | 1139 | if (err) { |
1140 | if (q) | 1140 | if (q) |
1141 | qdisc_destroy(q); | 1141 | qdisc_destroy(q); |
1142 | return err; | 1142 | return err; |
1143 | } | 1143 | } |
1144 | 1144 | ||
1145 | return 0; | 1145 | return 0; |
1146 | } | 1146 | } |
1147 | 1147 | ||
1148 | static int tc_fill_qdisc(struct sk_buff *skb, struct Qdisc *q, u32 clid, | 1148 | static int tc_fill_qdisc(struct sk_buff *skb, struct Qdisc *q, u32 clid, |
1149 | u32 pid, u32 seq, u16 flags, int event) | 1149 | u32 pid, u32 seq, u16 flags, int event) |
1150 | { | 1150 | { |
1151 | struct tcmsg *tcm; | 1151 | struct tcmsg *tcm; |
1152 | struct nlmsghdr *nlh; | 1152 | struct nlmsghdr *nlh; |
1153 | unsigned char *b = skb_tail_pointer(skb); | 1153 | unsigned char *b = skb_tail_pointer(skb); |
1154 | struct gnet_dump d; | 1154 | struct gnet_dump d; |
1155 | 1155 | ||
1156 | nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*tcm), flags); | 1156 | nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*tcm), flags); |
1157 | tcm = NLMSG_DATA(nlh); | 1157 | tcm = NLMSG_DATA(nlh); |
1158 | tcm->tcm_family = AF_UNSPEC; | 1158 | tcm->tcm_family = AF_UNSPEC; |
1159 | tcm->tcm__pad1 = 0; | 1159 | tcm->tcm__pad1 = 0; |
1160 | tcm->tcm__pad2 = 0; | 1160 | tcm->tcm__pad2 = 0; |
1161 | tcm->tcm_ifindex = qdisc_dev(q)->ifindex; | 1161 | tcm->tcm_ifindex = qdisc_dev(q)->ifindex; |
1162 | tcm->tcm_parent = clid; | 1162 | tcm->tcm_parent = clid; |
1163 | tcm->tcm_handle = q->handle; | 1163 | tcm->tcm_handle = q->handle; |
1164 | tcm->tcm_info = atomic_read(&q->refcnt); | 1164 | tcm->tcm_info = atomic_read(&q->refcnt); |
1165 | NLA_PUT_STRING(skb, TCA_KIND, q->ops->id); | 1165 | NLA_PUT_STRING(skb, TCA_KIND, q->ops->id); |
1166 | if (q->ops->dump && q->ops->dump(q, skb) < 0) | 1166 | if (q->ops->dump && q->ops->dump(q, skb) < 0) |
1167 | goto nla_put_failure; | 1167 | goto nla_put_failure; |
1168 | q->qstats.qlen = q->q.qlen; | 1168 | q->qstats.qlen = q->q.qlen; |
1169 | 1169 | ||
1170 | if (q->stab && qdisc_dump_stab(skb, q->stab) < 0) | 1170 | if (q->stab && qdisc_dump_stab(skb, q->stab) < 0) |
1171 | goto nla_put_failure; | 1171 | goto nla_put_failure; |
1172 | 1172 | ||
1173 | if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS, | 1173 | if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS, |
1174 | qdisc_root_sleeping_lock(q), &d) < 0) | 1174 | qdisc_root_sleeping_lock(q), &d) < 0) |
1175 | goto nla_put_failure; | 1175 | goto nla_put_failure; |
1176 | 1176 | ||
1177 | if (q->ops->dump_stats && q->ops->dump_stats(q, &d) < 0) | 1177 | if (q->ops->dump_stats && q->ops->dump_stats(q, &d) < 0) |
1178 | goto nla_put_failure; | 1178 | goto nla_put_failure; |
1179 | 1179 | ||
1180 | if (gnet_stats_copy_basic(&d, &q->bstats) < 0 || | 1180 | if (gnet_stats_copy_basic(&d, &q->bstats) < 0 || |
1181 | gnet_stats_copy_rate_est(&d, &q->bstats, &q->rate_est) < 0 || | 1181 | gnet_stats_copy_rate_est(&d, &q->bstats, &q->rate_est) < 0 || |
1182 | gnet_stats_copy_queue(&d, &q->qstats) < 0) | 1182 | gnet_stats_copy_queue(&d, &q->qstats) < 0) |
1183 | goto nla_put_failure; | 1183 | goto nla_put_failure; |
1184 | 1184 | ||
1185 | if (gnet_stats_finish_copy(&d) < 0) | 1185 | if (gnet_stats_finish_copy(&d) < 0) |
1186 | goto nla_put_failure; | 1186 | goto nla_put_failure; |
1187 | 1187 | ||
1188 | nlh->nlmsg_len = skb_tail_pointer(skb) - b; | 1188 | nlh->nlmsg_len = skb_tail_pointer(skb) - b; |
1189 | return skb->len; | 1189 | return skb->len; |
1190 | 1190 | ||
1191 | nlmsg_failure: | 1191 | nlmsg_failure: |
1192 | nla_put_failure: | 1192 | nla_put_failure: |
1193 | nlmsg_trim(skb, b); | 1193 | nlmsg_trim(skb, b); |
1194 | return -1; | 1194 | return -1; |
1195 | } | 1195 | } |
1196 | 1196 | ||
1197 | static int qdisc_notify(struct net *net, struct sk_buff *oskb, | 1197 | static int qdisc_notify(struct net *net, struct sk_buff *oskb, |
1198 | struct nlmsghdr *n, u32 clid, | 1198 | struct nlmsghdr *n, u32 clid, |
1199 | struct Qdisc *old, struct Qdisc *new) | 1199 | struct Qdisc *old, struct Qdisc *new) |
1200 | { | 1200 | { |
1201 | struct sk_buff *skb; | 1201 | struct sk_buff *skb; |
1202 | u32 pid = oskb ? NETLINK_CB(oskb).pid : 0; | 1202 | u32 pid = oskb ? NETLINK_CB(oskb).pid : 0; |
1203 | 1203 | ||
1204 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); | 1204 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
1205 | if (!skb) | 1205 | if (!skb) |
1206 | return -ENOBUFS; | 1206 | return -ENOBUFS; |
1207 | 1207 | ||
1208 | if (old && old->handle) { | 1208 | if (old && old->handle) { |
1209 | if (tc_fill_qdisc(skb, old, clid, pid, n->nlmsg_seq, 0, RTM_DELQDISC) < 0) | 1209 | if (tc_fill_qdisc(skb, old, clid, pid, n->nlmsg_seq, 0, RTM_DELQDISC) < 0) |
1210 | goto err_out; | 1210 | goto err_out; |
1211 | } | 1211 | } |
1212 | if (new) { | 1212 | if (new) { |
1213 | if (tc_fill_qdisc(skb, new, clid, pid, n->nlmsg_seq, old ? NLM_F_REPLACE : 0, RTM_NEWQDISC) < 0) | 1213 | if (tc_fill_qdisc(skb, new, clid, pid, n->nlmsg_seq, old ? NLM_F_REPLACE : 0, RTM_NEWQDISC) < 0) |
1214 | goto err_out; | 1214 | goto err_out; |
1215 | } | 1215 | } |
1216 | 1216 | ||
1217 | if (skb->len) | 1217 | if (skb->len) |
1218 | return rtnetlink_send(skb, net, pid, RTNLGRP_TC, n->nlmsg_flags&NLM_F_ECHO); | 1218 | return rtnetlink_send(skb, net, pid, RTNLGRP_TC, n->nlmsg_flags&NLM_F_ECHO); |
1219 | 1219 | ||
1220 | err_out: | 1220 | err_out: |
1221 | kfree_skb(skb); | 1221 | kfree_skb(skb); |
1222 | return -EINVAL; | 1222 | return -EINVAL; |
1223 | } | 1223 | } |
1224 | 1224 | ||
1225 | static bool tc_qdisc_dump_ignore(struct Qdisc *q) | 1225 | static bool tc_qdisc_dump_ignore(struct Qdisc *q) |
1226 | { | 1226 | { |
1227 | return (q->flags & TCQ_F_BUILTIN) ? true : false; | 1227 | return (q->flags & TCQ_F_BUILTIN) ? true : false; |
1228 | } | 1228 | } |
1229 | 1229 | ||
1230 | static int tc_dump_qdisc_root(struct Qdisc *root, struct sk_buff *skb, | 1230 | static int tc_dump_qdisc_root(struct Qdisc *root, struct sk_buff *skb, |
1231 | struct netlink_callback *cb, | 1231 | struct netlink_callback *cb, |
1232 | int *q_idx_p, int s_q_idx) | 1232 | int *q_idx_p, int s_q_idx) |
1233 | { | 1233 | { |
1234 | int ret = 0, q_idx = *q_idx_p; | 1234 | int ret = 0, q_idx = *q_idx_p; |
1235 | struct Qdisc *q; | 1235 | struct Qdisc *q; |
1236 | 1236 | ||
1237 | if (!root) | 1237 | if (!root) |
1238 | return 0; | 1238 | return 0; |
1239 | 1239 | ||
1240 | q = root; | 1240 | q = root; |
1241 | if (q_idx < s_q_idx) { | 1241 | if (q_idx < s_q_idx) { |
1242 | q_idx++; | 1242 | q_idx++; |
1243 | } else { | 1243 | } else { |
1244 | if (!tc_qdisc_dump_ignore(q) && | 1244 | if (!tc_qdisc_dump_ignore(q) && |
1245 | tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).pid, | 1245 | tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).pid, |
1246 | cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWQDISC) <= 0) | 1246 | cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWQDISC) <= 0) |
1247 | goto done; | 1247 | goto done; |
1248 | q_idx++; | 1248 | q_idx++; |
1249 | } | 1249 | } |
1250 | list_for_each_entry(q, &root->list, list) { | 1250 | list_for_each_entry(q, &root->list, list) { |
1251 | if (q_idx < s_q_idx) { | 1251 | if (q_idx < s_q_idx) { |
1252 | q_idx++; | 1252 | q_idx++; |
1253 | continue; | 1253 | continue; |
1254 | } | 1254 | } |
1255 | if (!tc_qdisc_dump_ignore(q) && | 1255 | if (!tc_qdisc_dump_ignore(q) && |
1256 | tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).pid, | 1256 | tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).pid, |
1257 | cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWQDISC) <= 0) | 1257 | cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWQDISC) <= 0) |
1258 | goto done; | 1258 | goto done; |
1259 | q_idx++; | 1259 | q_idx++; |
1260 | } | 1260 | } |
1261 | 1261 | ||
1262 | out: | 1262 | out: |
1263 | *q_idx_p = q_idx; | 1263 | *q_idx_p = q_idx; |
1264 | return ret; | 1264 | return ret; |
1265 | done: | 1265 | done: |
1266 | ret = -1; | 1266 | ret = -1; |
1267 | goto out; | 1267 | goto out; |
1268 | } | 1268 | } |
1269 | 1269 | ||
1270 | static int tc_dump_qdisc(struct sk_buff *skb, struct netlink_callback *cb) | 1270 | static int tc_dump_qdisc(struct sk_buff *skb, struct netlink_callback *cb) |
1271 | { | 1271 | { |
1272 | struct net *net = sock_net(skb->sk); | 1272 | struct net *net = sock_net(skb->sk); |
1273 | int idx, q_idx; | 1273 | int idx, q_idx; |
1274 | int s_idx, s_q_idx; | 1274 | int s_idx, s_q_idx; |
1275 | struct net_device *dev; | 1275 | struct net_device *dev; |
1276 | 1276 | ||
1277 | s_idx = cb->args[0]; | 1277 | s_idx = cb->args[0]; |
1278 | s_q_idx = q_idx = cb->args[1]; | 1278 | s_q_idx = q_idx = cb->args[1]; |
1279 | 1279 | ||
1280 | rcu_read_lock(); | 1280 | rcu_read_lock(); |
1281 | idx = 0; | 1281 | idx = 0; |
1282 | for_each_netdev_rcu(net, dev) { | 1282 | for_each_netdev_rcu(net, dev) { |
1283 | struct netdev_queue *dev_queue; | 1283 | struct netdev_queue *dev_queue; |
1284 | 1284 | ||
1285 | if (idx < s_idx) | 1285 | if (idx < s_idx) |
1286 | goto cont; | 1286 | goto cont; |
1287 | if (idx > s_idx) | 1287 | if (idx > s_idx) |
1288 | s_q_idx = 0; | 1288 | s_q_idx = 0; |
1289 | q_idx = 0; | 1289 | q_idx = 0; |
1290 | 1290 | ||
1291 | if (tc_dump_qdisc_root(dev->qdisc, skb, cb, &q_idx, s_q_idx) < 0) | 1291 | if (tc_dump_qdisc_root(dev->qdisc, skb, cb, &q_idx, s_q_idx) < 0) |
1292 | goto done; | 1292 | goto done; |
1293 | 1293 | ||
1294 | dev_queue = &dev->rx_queue; | 1294 | dev_queue = &dev->rx_queue; |
1295 | if (tc_dump_qdisc_root(dev_queue->qdisc_sleeping, skb, cb, &q_idx, s_q_idx) < 0) | 1295 | if (tc_dump_qdisc_root(dev_queue->qdisc_sleeping, skb, cb, &q_idx, s_q_idx) < 0) |
1296 | goto done; | 1296 | goto done; |
1297 | 1297 | ||
1298 | cont: | 1298 | cont: |
1299 | idx++; | 1299 | idx++; |
1300 | } | 1300 | } |
1301 | 1301 | ||
1302 | done: | 1302 | done: |
1303 | rcu_read_unlock(); | 1303 | rcu_read_unlock(); |
1304 | 1304 | ||
1305 | cb->args[0] = idx; | 1305 | cb->args[0] = idx; |
1306 | cb->args[1] = q_idx; | 1306 | cb->args[1] = q_idx; |
1307 | 1307 | ||
1308 | return skb->len; | 1308 | return skb->len; |
1309 | } | 1309 | } |
1310 | 1310 | ||
1311 | 1311 | ||
1312 | 1312 | ||
1313 | /************************************************ | 1313 | /************************************************ |
1314 | * Traffic classes manipulation. * | 1314 | * Traffic classes manipulation. * |
1315 | ************************************************/ | 1315 | ************************************************/ |
1316 | 1316 | ||
1317 | 1317 | ||
1318 | 1318 | ||
1319 | static int tc_ctl_tclass(struct sk_buff *skb, struct nlmsghdr *n, void *arg) | 1319 | static int tc_ctl_tclass(struct sk_buff *skb, struct nlmsghdr *n, void *arg) |
1320 | { | 1320 | { |
1321 | struct net *net = sock_net(skb->sk); | 1321 | struct net *net = sock_net(skb->sk); |
1322 | struct tcmsg *tcm = NLMSG_DATA(n); | 1322 | struct tcmsg *tcm = NLMSG_DATA(n); |
1323 | struct nlattr *tca[TCA_MAX + 1]; | 1323 | struct nlattr *tca[TCA_MAX + 1]; |
1324 | struct net_device *dev; | 1324 | struct net_device *dev; |
1325 | struct Qdisc *q = NULL; | 1325 | struct Qdisc *q = NULL; |
1326 | const struct Qdisc_class_ops *cops; | 1326 | const struct Qdisc_class_ops *cops; |
1327 | unsigned long cl = 0; | 1327 | unsigned long cl = 0; |
1328 | unsigned long new_cl; | 1328 | unsigned long new_cl; |
1329 | u32 pid = tcm->tcm_parent; | 1329 | u32 pid = tcm->tcm_parent; |
1330 | u32 clid = tcm->tcm_handle; | 1330 | u32 clid = tcm->tcm_handle; |
1331 | u32 qid = TC_H_MAJ(clid); | 1331 | u32 qid = TC_H_MAJ(clid); |
1332 | int err; | 1332 | int err; |
1333 | 1333 | ||
1334 | if ((dev = __dev_get_by_index(net, tcm->tcm_ifindex)) == NULL) | 1334 | if ((dev = __dev_get_by_index(net, tcm->tcm_ifindex)) == NULL) |
1335 | return -ENODEV; | 1335 | return -ENODEV; |
1336 | 1336 | ||
1337 | err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL); | 1337 | err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL); |
1338 | if (err < 0) | 1338 | if (err < 0) |
1339 | return err; | 1339 | return err; |
1340 | 1340 | ||
1341 | /* | 1341 | /* |
1342 | parent == TC_H_UNSPEC - unspecified parent. | 1342 | parent == TC_H_UNSPEC - unspecified parent. |
1343 | parent == TC_H_ROOT - class is root, which has no parent. | 1343 | parent == TC_H_ROOT - class is root, which has no parent. |
1344 | parent == X:0 - parent is root class. | 1344 | parent == X:0 - parent is root class. |
1345 | parent == X:Y - parent is a node in hierarchy. | 1345 | parent == X:Y - parent is a node in hierarchy. |
1346 | parent == 0:Y - parent is X:Y, where X:0 is qdisc. | 1346 | parent == 0:Y - parent is X:Y, where X:0 is qdisc. |
1347 | 1347 | ||
1348 | handle == 0:0 - generate handle from kernel pool. | 1348 | handle == 0:0 - generate handle from kernel pool. |
1349 | handle == 0:Y - class is X:Y, where X:0 is qdisc. | 1349 | handle == 0:Y - class is X:Y, where X:0 is qdisc. |
1350 | handle == X:Y - clear. | 1350 | handle == X:Y - clear. |
1351 | handle == X:0 - root class. | 1351 | handle == X:0 - root class. |
1352 | */ | 1352 | */ |
1353 | 1353 | ||
1354 | /* Step 1. Determine qdisc handle X:0 */ | 1354 | /* Step 1. Determine qdisc handle X:0 */ |
1355 | 1355 | ||
1356 | if (pid != TC_H_ROOT) { | 1356 | if (pid != TC_H_ROOT) { |
1357 | u32 qid1 = TC_H_MAJ(pid); | 1357 | u32 qid1 = TC_H_MAJ(pid); |
1358 | 1358 | ||
1359 | if (qid && qid1) { | 1359 | if (qid && qid1) { |
1360 | /* If both majors are known, they must be identical. */ | 1360 | /* If both majors are known, they must be identical. */ |
1361 | if (qid != qid1) | 1361 | if (qid != qid1) |
1362 | return -EINVAL; | 1362 | return -EINVAL; |
1363 | } else if (qid1) { | 1363 | } else if (qid1) { |
1364 | qid = qid1; | 1364 | qid = qid1; |
1365 | } else if (qid == 0) | 1365 | } else if (qid == 0) |
1366 | qid = dev->qdisc->handle; | 1366 | qid = dev->qdisc->handle; |
1367 | 1367 | ||
1368 | /* Now qid is genuine qdisc handle consistent | 1368 | /* Now qid is genuine qdisc handle consistent |
1369 | both with parent and child. | 1369 | both with parent and child. |
1370 | 1370 | ||
1371 | TC_H_MAJ(pid) still may be unspecified, complete it now. | 1371 | TC_H_MAJ(pid) still may be unspecified, complete it now. |
1372 | */ | 1372 | */ |
1373 | if (pid) | 1373 | if (pid) |
1374 | pid = TC_H_MAKE(qid, pid); | 1374 | pid = TC_H_MAKE(qid, pid); |
1375 | } else { | 1375 | } else { |
1376 | if (qid == 0) | 1376 | if (qid == 0) |
1377 | qid = dev->qdisc->handle; | 1377 | qid = dev->qdisc->handle; |
1378 | } | 1378 | } |
1379 | 1379 | ||
1380 | /* OK. Locate qdisc */ | 1380 | /* OK. Locate qdisc */ |
1381 | if ((q = qdisc_lookup(dev, qid)) == NULL) | 1381 | if ((q = qdisc_lookup(dev, qid)) == NULL) |
1382 | return -ENOENT; | 1382 | return -ENOENT; |
1383 | 1383 | ||
1384 | /* An check that it supports classes */ | 1384 | /* An check that it supports classes */ |
1385 | cops = q->ops->cl_ops; | 1385 | cops = q->ops->cl_ops; |
1386 | if (cops == NULL) | 1386 | if (cops == NULL) |
1387 | return -EINVAL; | 1387 | return -EINVAL; |
1388 | 1388 | ||
1389 | /* Now try to get class */ | 1389 | /* Now try to get class */ |
1390 | if (clid == 0) { | 1390 | if (clid == 0) { |
1391 | if (pid == TC_H_ROOT) | 1391 | if (pid == TC_H_ROOT) |
1392 | clid = qid; | 1392 | clid = qid; |
1393 | } else | 1393 | } else |
1394 | clid = TC_H_MAKE(qid, clid); | 1394 | clid = TC_H_MAKE(qid, clid); |
1395 | 1395 | ||
1396 | if (clid) | 1396 | if (clid) |
1397 | cl = cops->get(q, clid); | 1397 | cl = cops->get(q, clid); |
1398 | 1398 | ||
1399 | if (cl == 0) { | 1399 | if (cl == 0) { |
1400 | err = -ENOENT; | 1400 | err = -ENOENT; |
1401 | if (n->nlmsg_type != RTM_NEWTCLASS || !(n->nlmsg_flags&NLM_F_CREATE)) | 1401 | if (n->nlmsg_type != RTM_NEWTCLASS || !(n->nlmsg_flags&NLM_F_CREATE)) |
1402 | goto out; | 1402 | goto out; |
1403 | } else { | 1403 | } else { |
1404 | switch (n->nlmsg_type) { | 1404 | switch (n->nlmsg_type) { |
1405 | case RTM_NEWTCLASS: | 1405 | case RTM_NEWTCLASS: |
1406 | err = -EEXIST; | 1406 | err = -EEXIST; |
1407 | if (n->nlmsg_flags&NLM_F_EXCL) | 1407 | if (n->nlmsg_flags&NLM_F_EXCL) |
1408 | goto out; | 1408 | goto out; |
1409 | break; | 1409 | break; |
1410 | case RTM_DELTCLASS: | 1410 | case RTM_DELTCLASS: |
1411 | err = -EOPNOTSUPP; | 1411 | err = -EOPNOTSUPP; |
1412 | if (cops->delete) | 1412 | if (cops->delete) |
1413 | err = cops->delete(q, cl); | 1413 | err = cops->delete(q, cl); |
1414 | if (err == 0) | 1414 | if (err == 0) |
1415 | tclass_notify(net, skb, n, q, cl, RTM_DELTCLASS); | 1415 | tclass_notify(net, skb, n, q, cl, RTM_DELTCLASS); |
1416 | goto out; | 1416 | goto out; |
1417 | case RTM_GETTCLASS: | 1417 | case RTM_GETTCLASS: |
1418 | err = tclass_notify(net, skb, n, q, cl, RTM_NEWTCLASS); | 1418 | err = tclass_notify(net, skb, n, q, cl, RTM_NEWTCLASS); |
1419 | goto out; | 1419 | goto out; |
1420 | default: | 1420 | default: |
1421 | err = -EINVAL; | 1421 | err = -EINVAL; |
1422 | goto out; | 1422 | goto out; |
1423 | } | 1423 | } |
1424 | } | 1424 | } |
1425 | 1425 | ||
1426 | new_cl = cl; | 1426 | new_cl = cl; |
1427 | err = -EOPNOTSUPP; | 1427 | err = -EOPNOTSUPP; |
1428 | if (cops->change) | 1428 | if (cops->change) |
1429 | err = cops->change(q, clid, pid, tca, &new_cl); | 1429 | err = cops->change(q, clid, pid, tca, &new_cl); |
1430 | if (err == 0) | 1430 | if (err == 0) |
1431 | tclass_notify(net, skb, n, q, new_cl, RTM_NEWTCLASS); | 1431 | tclass_notify(net, skb, n, q, new_cl, RTM_NEWTCLASS); |
1432 | 1432 | ||
1433 | out: | 1433 | out: |
1434 | if (cl) | 1434 | if (cl) |
1435 | cops->put(q, cl); | 1435 | cops->put(q, cl); |
1436 | 1436 | ||
1437 | return err; | 1437 | return err; |
1438 | } | 1438 | } |
1439 | 1439 | ||
1440 | 1440 | ||
1441 | static int tc_fill_tclass(struct sk_buff *skb, struct Qdisc *q, | 1441 | static int tc_fill_tclass(struct sk_buff *skb, struct Qdisc *q, |
1442 | unsigned long cl, | 1442 | unsigned long cl, |
1443 | u32 pid, u32 seq, u16 flags, int event) | 1443 | u32 pid, u32 seq, u16 flags, int event) |
1444 | { | 1444 | { |
1445 | struct tcmsg *tcm; | 1445 | struct tcmsg *tcm; |
1446 | struct nlmsghdr *nlh; | 1446 | struct nlmsghdr *nlh; |
1447 | unsigned char *b = skb_tail_pointer(skb); | 1447 | unsigned char *b = skb_tail_pointer(skb); |
1448 | struct gnet_dump d; | 1448 | struct gnet_dump d; |
1449 | const struct Qdisc_class_ops *cl_ops = q->ops->cl_ops; | 1449 | const struct Qdisc_class_ops *cl_ops = q->ops->cl_ops; |
1450 | 1450 | ||
1451 | nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*tcm), flags); | 1451 | nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*tcm), flags); |
1452 | tcm = NLMSG_DATA(nlh); | 1452 | tcm = NLMSG_DATA(nlh); |
1453 | tcm->tcm_family = AF_UNSPEC; | 1453 | tcm->tcm_family = AF_UNSPEC; |
1454 | tcm->tcm__pad1 = 0; | 1454 | tcm->tcm__pad1 = 0; |
1455 | tcm->tcm__pad2 = 0; | 1455 | tcm->tcm__pad2 = 0; |
1456 | tcm->tcm_ifindex = qdisc_dev(q)->ifindex; | 1456 | tcm->tcm_ifindex = qdisc_dev(q)->ifindex; |
1457 | tcm->tcm_parent = q->handle; | 1457 | tcm->tcm_parent = q->handle; |
1458 | tcm->tcm_handle = q->handle; | 1458 | tcm->tcm_handle = q->handle; |
1459 | tcm->tcm_info = 0; | 1459 | tcm->tcm_info = 0; |
1460 | NLA_PUT_STRING(skb, TCA_KIND, q->ops->id); | 1460 | NLA_PUT_STRING(skb, TCA_KIND, q->ops->id); |
1461 | if (cl_ops->dump && cl_ops->dump(q, cl, skb, tcm) < 0) | 1461 | if (cl_ops->dump && cl_ops->dump(q, cl, skb, tcm) < 0) |
1462 | goto nla_put_failure; | 1462 | goto nla_put_failure; |
1463 | 1463 | ||
1464 | if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS, | 1464 | if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS, |
1465 | qdisc_root_sleeping_lock(q), &d) < 0) | 1465 | qdisc_root_sleeping_lock(q), &d) < 0) |
1466 | goto nla_put_failure; | 1466 | goto nla_put_failure; |
1467 | 1467 | ||
1468 | if (cl_ops->dump_stats && cl_ops->dump_stats(q, cl, &d) < 0) | 1468 | if (cl_ops->dump_stats && cl_ops->dump_stats(q, cl, &d) < 0) |
1469 | goto nla_put_failure; | 1469 | goto nla_put_failure; |
1470 | 1470 | ||
1471 | if (gnet_stats_finish_copy(&d) < 0) | 1471 | if (gnet_stats_finish_copy(&d) < 0) |
1472 | goto nla_put_failure; | 1472 | goto nla_put_failure; |
1473 | 1473 | ||
1474 | nlh->nlmsg_len = skb_tail_pointer(skb) - b; | 1474 | nlh->nlmsg_len = skb_tail_pointer(skb) - b; |
1475 | return skb->len; | 1475 | return skb->len; |
1476 | 1476 | ||
1477 | nlmsg_failure: | 1477 | nlmsg_failure: |
1478 | nla_put_failure: | 1478 | nla_put_failure: |
1479 | nlmsg_trim(skb, b); | 1479 | nlmsg_trim(skb, b); |
1480 | return -1; | 1480 | return -1; |
1481 | } | 1481 | } |
1482 | 1482 | ||
1483 | static int tclass_notify(struct net *net, struct sk_buff *oskb, | 1483 | static int tclass_notify(struct net *net, struct sk_buff *oskb, |
1484 | struct nlmsghdr *n, struct Qdisc *q, | 1484 | struct nlmsghdr *n, struct Qdisc *q, |
1485 | unsigned long cl, int event) | 1485 | unsigned long cl, int event) |
1486 | { | 1486 | { |
1487 | struct sk_buff *skb; | 1487 | struct sk_buff *skb; |
1488 | u32 pid = oskb ? NETLINK_CB(oskb).pid : 0; | 1488 | u32 pid = oskb ? NETLINK_CB(oskb).pid : 0; |
1489 | 1489 | ||
1490 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); | 1490 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
1491 | if (!skb) | 1491 | if (!skb) |
1492 | return -ENOBUFS; | 1492 | return -ENOBUFS; |
1493 | 1493 | ||
1494 | if (tc_fill_tclass(skb, q, cl, pid, n->nlmsg_seq, 0, event) < 0) { | 1494 | if (tc_fill_tclass(skb, q, cl, pid, n->nlmsg_seq, 0, event) < 0) { |
1495 | kfree_skb(skb); | 1495 | kfree_skb(skb); |
1496 | return -EINVAL; | 1496 | return -EINVAL; |
1497 | } | 1497 | } |
1498 | 1498 | ||
1499 | return rtnetlink_send(skb, net, pid, RTNLGRP_TC, n->nlmsg_flags&NLM_F_ECHO); | 1499 | return rtnetlink_send(skb, net, pid, RTNLGRP_TC, n->nlmsg_flags&NLM_F_ECHO); |
1500 | } | 1500 | } |
1501 | 1501 | ||
1502 | struct qdisc_dump_args | 1502 | struct qdisc_dump_args |
1503 | { | 1503 | { |
1504 | struct qdisc_walker w; | 1504 | struct qdisc_walker w; |
1505 | struct sk_buff *skb; | 1505 | struct sk_buff *skb; |
1506 | struct netlink_callback *cb; | 1506 | struct netlink_callback *cb; |
1507 | }; | 1507 | }; |
1508 | 1508 | ||
1509 | static int qdisc_class_dump(struct Qdisc *q, unsigned long cl, struct qdisc_walker *arg) | 1509 | static int qdisc_class_dump(struct Qdisc *q, unsigned long cl, struct qdisc_walker *arg) |
1510 | { | 1510 | { |
1511 | struct qdisc_dump_args *a = (struct qdisc_dump_args *)arg; | 1511 | struct qdisc_dump_args *a = (struct qdisc_dump_args *)arg; |
1512 | 1512 | ||
1513 | return tc_fill_tclass(a->skb, q, cl, NETLINK_CB(a->cb->skb).pid, | 1513 | return tc_fill_tclass(a->skb, q, cl, NETLINK_CB(a->cb->skb).pid, |
1514 | a->cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWTCLASS); | 1514 | a->cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWTCLASS); |
1515 | } | 1515 | } |
1516 | 1516 | ||
1517 | static int tc_dump_tclass_qdisc(struct Qdisc *q, struct sk_buff *skb, | 1517 | static int tc_dump_tclass_qdisc(struct Qdisc *q, struct sk_buff *skb, |
1518 | struct tcmsg *tcm, struct netlink_callback *cb, | 1518 | struct tcmsg *tcm, struct netlink_callback *cb, |
1519 | int *t_p, int s_t) | 1519 | int *t_p, int s_t) |
1520 | { | 1520 | { |
1521 | struct qdisc_dump_args arg; | 1521 | struct qdisc_dump_args arg; |
1522 | 1522 | ||
1523 | if (tc_qdisc_dump_ignore(q) || | 1523 | if (tc_qdisc_dump_ignore(q) || |
1524 | *t_p < s_t || !q->ops->cl_ops || | 1524 | *t_p < s_t || !q->ops->cl_ops || |
1525 | (tcm->tcm_parent && | 1525 | (tcm->tcm_parent && |
1526 | TC_H_MAJ(tcm->tcm_parent) != q->handle)) { | 1526 | TC_H_MAJ(tcm->tcm_parent) != q->handle)) { |
1527 | (*t_p)++; | 1527 | (*t_p)++; |
1528 | return 0; | 1528 | return 0; |
1529 | } | 1529 | } |
1530 | if (*t_p > s_t) | 1530 | if (*t_p > s_t) |
1531 | memset(&cb->args[1], 0, sizeof(cb->args)-sizeof(cb->args[0])); | 1531 | memset(&cb->args[1], 0, sizeof(cb->args)-sizeof(cb->args[0])); |
1532 | arg.w.fn = qdisc_class_dump; | 1532 | arg.w.fn = qdisc_class_dump; |
1533 | arg.skb = skb; | 1533 | arg.skb = skb; |
1534 | arg.cb = cb; | 1534 | arg.cb = cb; |
1535 | arg.w.stop = 0; | 1535 | arg.w.stop = 0; |
1536 | arg.w.skip = cb->args[1]; | 1536 | arg.w.skip = cb->args[1]; |
1537 | arg.w.count = 0; | 1537 | arg.w.count = 0; |
1538 | q->ops->cl_ops->walk(q, &arg.w); | 1538 | q->ops->cl_ops->walk(q, &arg.w); |
1539 | cb->args[1] = arg.w.count; | 1539 | cb->args[1] = arg.w.count; |
1540 | if (arg.w.stop) | 1540 | if (arg.w.stop) |
1541 | return -1; | 1541 | return -1; |
1542 | (*t_p)++; | 1542 | (*t_p)++; |
1543 | return 0; | 1543 | return 0; |
1544 | } | 1544 | } |
1545 | 1545 | ||
1546 | static int tc_dump_tclass_root(struct Qdisc *root, struct sk_buff *skb, | 1546 | static int tc_dump_tclass_root(struct Qdisc *root, struct sk_buff *skb, |
1547 | struct tcmsg *tcm, struct netlink_callback *cb, | 1547 | struct tcmsg *tcm, struct netlink_callback *cb, |
1548 | int *t_p, int s_t) | 1548 | int *t_p, int s_t) |
1549 | { | 1549 | { |
1550 | struct Qdisc *q; | 1550 | struct Qdisc *q; |
1551 | 1551 | ||
1552 | if (!root) | 1552 | if (!root) |
1553 | return 0; | 1553 | return 0; |
1554 | 1554 | ||
1555 | if (tc_dump_tclass_qdisc(root, skb, tcm, cb, t_p, s_t) < 0) | 1555 | if (tc_dump_tclass_qdisc(root, skb, tcm, cb, t_p, s_t) < 0) |
1556 | return -1; | 1556 | return -1; |
1557 | 1557 | ||
1558 | list_for_each_entry(q, &root->list, list) { | 1558 | list_for_each_entry(q, &root->list, list) { |
1559 | if (tc_dump_tclass_qdisc(q, skb, tcm, cb, t_p, s_t) < 0) | 1559 | if (tc_dump_tclass_qdisc(q, skb, tcm, cb, t_p, s_t) < 0) |
1560 | return -1; | 1560 | return -1; |
1561 | } | 1561 | } |
1562 | 1562 | ||
1563 | return 0; | 1563 | return 0; |
1564 | } | 1564 | } |
1565 | 1565 | ||
1566 | static int tc_dump_tclass(struct sk_buff *skb, struct netlink_callback *cb) | 1566 | static int tc_dump_tclass(struct sk_buff *skb, struct netlink_callback *cb) |
1567 | { | 1567 | { |
1568 | struct tcmsg *tcm = (struct tcmsg*)NLMSG_DATA(cb->nlh); | 1568 | struct tcmsg *tcm = (struct tcmsg*)NLMSG_DATA(cb->nlh); |
1569 | struct net *net = sock_net(skb->sk); | 1569 | struct net *net = sock_net(skb->sk); |
1570 | struct netdev_queue *dev_queue; | 1570 | struct netdev_queue *dev_queue; |
1571 | struct net_device *dev; | 1571 | struct net_device *dev; |
1572 | int t, s_t; | 1572 | int t, s_t; |
1573 | 1573 | ||
1574 | if (cb->nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*tcm))) | 1574 | if (cb->nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*tcm))) |
1575 | return 0; | 1575 | return 0; |
1576 | if ((dev = dev_get_by_index(net, tcm->tcm_ifindex)) == NULL) | 1576 | if ((dev = dev_get_by_index(net, tcm->tcm_ifindex)) == NULL) |
1577 | return 0; | 1577 | return 0; |
1578 | 1578 | ||
1579 | s_t = cb->args[0]; | 1579 | s_t = cb->args[0]; |
1580 | t = 0; | 1580 | t = 0; |
1581 | 1581 | ||
1582 | if (tc_dump_tclass_root(dev->qdisc, skb, tcm, cb, &t, s_t) < 0) | 1582 | if (tc_dump_tclass_root(dev->qdisc, skb, tcm, cb, &t, s_t) < 0) |
1583 | goto done; | 1583 | goto done; |
1584 | 1584 | ||
1585 | dev_queue = &dev->rx_queue; | 1585 | dev_queue = &dev->rx_queue; |
1586 | if (tc_dump_tclass_root(dev_queue->qdisc_sleeping, skb, tcm, cb, &t, s_t) < 0) | 1586 | if (tc_dump_tclass_root(dev_queue->qdisc_sleeping, skb, tcm, cb, &t, s_t) < 0) |
1587 | goto done; | 1587 | goto done; |
1588 | 1588 | ||
1589 | done: | 1589 | done: |
1590 | cb->args[0] = t; | 1590 | cb->args[0] = t; |
1591 | 1591 | ||
1592 | dev_put(dev); | 1592 | dev_put(dev); |
1593 | return skb->len; | 1593 | return skb->len; |
1594 | } | 1594 | } |
1595 | 1595 | ||
1596 | /* Main classifier routine: scans classifier chain attached | 1596 | /* Main classifier routine: scans classifier chain attached |
1597 | to this qdisc, (optionally) tests for protocol and asks | 1597 | to this qdisc, (optionally) tests for protocol and asks |
1598 | specific classifiers. | 1598 | specific classifiers. |
1599 | */ | 1599 | */ |
1600 | int tc_classify_compat(struct sk_buff *skb, struct tcf_proto *tp, | 1600 | int tc_classify_compat(struct sk_buff *skb, struct tcf_proto *tp, |
1601 | struct tcf_result *res) | 1601 | struct tcf_result *res) |
1602 | { | 1602 | { |
1603 | __be16 protocol = skb->protocol; | 1603 | __be16 protocol = skb->protocol; |
1604 | int err = 0; | 1604 | int err = 0; |
1605 | 1605 | ||
1606 | for (; tp; tp = tp->next) { | 1606 | for (; tp; tp = tp->next) { |
1607 | if ((tp->protocol == protocol || | 1607 | if ((tp->protocol == protocol || |
1608 | tp->protocol == htons(ETH_P_ALL)) && | 1608 | tp->protocol == htons(ETH_P_ALL)) && |
1609 | (err = tp->classify(skb, tp, res)) >= 0) { | 1609 | (err = tp->classify(skb, tp, res)) >= 0) { |
1610 | #ifdef CONFIG_NET_CLS_ACT | 1610 | #ifdef CONFIG_NET_CLS_ACT |
1611 | if (err != TC_ACT_RECLASSIFY && skb->tc_verd) | 1611 | if (err != TC_ACT_RECLASSIFY && skb->tc_verd) |
1612 | skb->tc_verd = SET_TC_VERD(skb->tc_verd, 0); | 1612 | skb->tc_verd = SET_TC_VERD(skb->tc_verd, 0); |
1613 | #endif | 1613 | #endif |
1614 | return err; | 1614 | return err; |
1615 | } | 1615 | } |
1616 | } | 1616 | } |
1617 | return -1; | 1617 | return -1; |
1618 | } | 1618 | } |
1619 | EXPORT_SYMBOL(tc_classify_compat); | 1619 | EXPORT_SYMBOL(tc_classify_compat); |
1620 | 1620 | ||
1621 | int tc_classify(struct sk_buff *skb, struct tcf_proto *tp, | 1621 | int tc_classify(struct sk_buff *skb, struct tcf_proto *tp, |
1622 | struct tcf_result *res) | 1622 | struct tcf_result *res) |
1623 | { | 1623 | { |
1624 | int err = 0; | 1624 | int err = 0; |
1625 | __be16 protocol; | 1625 | __be16 protocol; |
1626 | #ifdef CONFIG_NET_CLS_ACT | 1626 | #ifdef CONFIG_NET_CLS_ACT |
1627 | struct tcf_proto *otp = tp; | 1627 | struct tcf_proto *otp = tp; |
1628 | reclassify: | 1628 | reclassify: |
1629 | #endif | 1629 | #endif |
1630 | protocol = skb->protocol; | 1630 | protocol = skb->protocol; |
1631 | 1631 | ||
1632 | err = tc_classify_compat(skb, tp, res); | 1632 | err = tc_classify_compat(skb, tp, res); |
1633 | #ifdef CONFIG_NET_CLS_ACT | 1633 | #ifdef CONFIG_NET_CLS_ACT |
1634 | if (err == TC_ACT_RECLASSIFY) { | 1634 | if (err == TC_ACT_RECLASSIFY) { |
1635 | u32 verd = G_TC_VERD(skb->tc_verd); | 1635 | u32 verd = G_TC_VERD(skb->tc_verd); |
1636 | tp = otp; | 1636 | tp = otp; |
1637 | 1637 | ||
1638 | if (verd++ >= MAX_REC_LOOP) { | 1638 | if (verd++ >= MAX_REC_LOOP) { |
1639 | printk("rule prio %u protocol %02x reclassify loop, " | 1639 | printk("rule prio %u protocol %02x reclassify loop, " |
1640 | "packet dropped\n", | 1640 | "packet dropped\n", |
1641 | tp->prio&0xffff, ntohs(tp->protocol)); | 1641 | tp->prio&0xffff, ntohs(tp->protocol)); |
1642 | return TC_ACT_SHOT; | 1642 | return TC_ACT_SHOT; |
1643 | } | 1643 | } |
1644 | skb->tc_verd = SET_TC_VERD(skb->tc_verd, verd); | 1644 | skb->tc_verd = SET_TC_VERD(skb->tc_verd, verd); |
1645 | goto reclassify; | 1645 | goto reclassify; |
1646 | } | 1646 | } |
1647 | #endif | 1647 | #endif |
1648 | return err; | 1648 | return err; |
1649 | } | 1649 | } |
1650 | EXPORT_SYMBOL(tc_classify); | 1650 | EXPORT_SYMBOL(tc_classify); |
1651 | 1651 | ||
1652 | void tcf_destroy(struct tcf_proto *tp) | 1652 | void tcf_destroy(struct tcf_proto *tp) |
1653 | { | 1653 | { |
1654 | tp->ops->destroy(tp); | 1654 | tp->ops->destroy(tp); |
1655 | module_put(tp->ops->owner); | 1655 | module_put(tp->ops->owner); |
1656 | kfree(tp); | 1656 | kfree(tp); |
1657 | } | 1657 | } |
1658 | 1658 | ||
1659 | void tcf_destroy_chain(struct tcf_proto **fl) | 1659 | void tcf_destroy_chain(struct tcf_proto **fl) |
1660 | { | 1660 | { |
1661 | struct tcf_proto *tp; | 1661 | struct tcf_proto *tp; |
1662 | 1662 | ||
1663 | while ((tp = *fl) != NULL) { | 1663 | while ((tp = *fl) != NULL) { |
1664 | *fl = tp->next; | 1664 | *fl = tp->next; |
1665 | tcf_destroy(tp); | 1665 | tcf_destroy(tp); |
1666 | } | 1666 | } |
1667 | } | 1667 | } |
1668 | EXPORT_SYMBOL(tcf_destroy_chain); | 1668 | EXPORT_SYMBOL(tcf_destroy_chain); |
1669 | 1669 | ||
1670 | #ifdef CONFIG_PROC_FS | 1670 | #ifdef CONFIG_PROC_FS |
1671 | static int psched_show(struct seq_file *seq, void *v) | 1671 | static int psched_show(struct seq_file *seq, void *v) |
1672 | { | 1672 | { |
1673 | struct timespec ts; | 1673 | struct timespec ts; |
1674 | 1674 | ||
1675 | hrtimer_get_res(CLOCK_MONOTONIC, &ts); | 1675 | hrtimer_get_res(CLOCK_MONOTONIC, &ts); |
1676 | seq_printf(seq, "%08x %08x %08x %08x\n", | 1676 | seq_printf(seq, "%08x %08x %08x %08x\n", |
1677 | (u32)NSEC_PER_USEC, (u32)PSCHED_TICKS2NS(1), | 1677 | (u32)NSEC_PER_USEC, (u32)PSCHED_TICKS2NS(1), |
1678 | 1000000, | 1678 | 1000000, |
1679 | (u32)NSEC_PER_SEC/(u32)ktime_to_ns(timespec_to_ktime(ts))); | 1679 | (u32)NSEC_PER_SEC/(u32)ktime_to_ns(timespec_to_ktime(ts))); |
1680 | 1680 | ||
1681 | return 0; | 1681 | return 0; |
1682 | } | 1682 | } |
1683 | 1683 | ||
1684 | static int psched_open(struct inode *inode, struct file *file) | 1684 | static int psched_open(struct inode *inode, struct file *file) |
1685 | { | 1685 | { |
1686 | return single_open(file, psched_show, PDE(inode)->data); | 1686 | return single_open(file, psched_show, NULL); |
1687 | } | 1687 | } |
1688 | 1688 | ||
1689 | static const struct file_operations psched_fops = { | 1689 | static const struct file_operations psched_fops = { |
1690 | .owner = THIS_MODULE, | 1690 | .owner = THIS_MODULE, |
1691 | .open = psched_open, | 1691 | .open = psched_open, |
1692 | .read = seq_read, | 1692 | .read = seq_read, |
1693 | .llseek = seq_lseek, | 1693 | .llseek = seq_lseek, |
1694 | .release = single_release, | 1694 | .release = single_release, |
1695 | }; | 1695 | }; |
1696 | 1696 | ||
1697 | static int __net_init psched_net_init(struct net *net) | 1697 | static int __net_init psched_net_init(struct net *net) |
1698 | { | 1698 | { |
1699 | struct proc_dir_entry *e; | 1699 | struct proc_dir_entry *e; |
1700 | 1700 | ||
1701 | e = proc_net_fops_create(net, "psched", 0, &psched_fops); | 1701 | e = proc_net_fops_create(net, "psched", 0, &psched_fops); |
1702 | if (e == NULL) | 1702 | if (e == NULL) |
1703 | return -ENOMEM; | 1703 | return -ENOMEM; |
1704 | 1704 | ||
1705 | return 0; | 1705 | return 0; |
1706 | } | 1706 | } |
1707 | 1707 | ||
1708 | static void __net_exit psched_net_exit(struct net *net) | 1708 | static void __net_exit psched_net_exit(struct net *net) |
1709 | { | 1709 | { |
1710 | proc_net_remove(net, "psched"); | 1710 | proc_net_remove(net, "psched"); |
1711 | |||
1712 | return; | ||
1713 | } | 1711 | } |
1714 | #else | 1712 | #else |
1715 | static int __net_init psched_net_init(struct net *net) | 1713 | static int __net_init psched_net_init(struct net *net) |
1716 | { | 1714 | { |
1717 | return 0; | 1715 | return 0; |
1718 | } | 1716 | } |
1719 | 1717 | ||
1720 | static void __net_exit psched_net_exit(struct net *net) | 1718 | static void __net_exit psched_net_exit(struct net *net) |
1721 | { | 1719 | { |
1722 | } | 1720 | } |
1723 | #endif | 1721 | #endif |
1724 | 1722 | ||
1725 | static struct pernet_operations psched_net_ops = { | 1723 | static struct pernet_operations psched_net_ops = { |
1726 | .init = psched_net_init, | 1724 | .init = psched_net_init, |
1727 | .exit = psched_net_exit, | 1725 | .exit = psched_net_exit, |
1728 | }; | 1726 | }; |
1729 | 1727 | ||
1730 | static int __init pktsched_init(void) | 1728 | static int __init pktsched_init(void) |
1731 | { | 1729 | { |
1732 | int err; | 1730 | int err; |
1733 | 1731 | ||
1734 | err = register_pernet_subsys(&psched_net_ops); | 1732 | err = register_pernet_subsys(&psched_net_ops); |
1735 | if (err) { | 1733 | if (err) { |
1736 | printk(KERN_ERR "pktsched_init: " | 1734 | printk(KERN_ERR "pktsched_init: " |
1737 | "cannot initialize per netns operations\n"); | 1735 | "cannot initialize per netns operations\n"); |
1738 | return err; | 1736 | return err; |
1739 | } | 1737 | } |
1740 | 1738 | ||
1741 | register_qdisc(&pfifo_qdisc_ops); | 1739 | register_qdisc(&pfifo_qdisc_ops); |
1742 | register_qdisc(&bfifo_qdisc_ops); | 1740 | register_qdisc(&bfifo_qdisc_ops); |
1743 | register_qdisc(&pfifo_head_drop_qdisc_ops); | 1741 | register_qdisc(&pfifo_head_drop_qdisc_ops); |
1744 | register_qdisc(&mq_qdisc_ops); | 1742 | register_qdisc(&mq_qdisc_ops); |
1745 | 1743 | ||
1746 | rtnl_register(PF_UNSPEC, RTM_NEWQDISC, tc_modify_qdisc, NULL); | 1744 | rtnl_register(PF_UNSPEC, RTM_NEWQDISC, tc_modify_qdisc, NULL); |
1747 | rtnl_register(PF_UNSPEC, RTM_DELQDISC, tc_get_qdisc, NULL); | 1745 | rtnl_register(PF_UNSPEC, RTM_DELQDISC, tc_get_qdisc, NULL); |
1748 | rtnl_register(PF_UNSPEC, RTM_GETQDISC, tc_get_qdisc, tc_dump_qdisc); | 1746 | rtnl_register(PF_UNSPEC, RTM_GETQDISC, tc_get_qdisc, tc_dump_qdisc); |
1749 | rtnl_register(PF_UNSPEC, RTM_NEWTCLASS, tc_ctl_tclass, NULL); | 1747 | rtnl_register(PF_UNSPEC, RTM_NEWTCLASS, tc_ctl_tclass, NULL); |
1750 | rtnl_register(PF_UNSPEC, RTM_DELTCLASS, tc_ctl_tclass, NULL); | 1748 | rtnl_register(PF_UNSPEC, RTM_DELTCLASS, tc_ctl_tclass, NULL); |
1751 | rtnl_register(PF_UNSPEC, RTM_GETTCLASS, tc_ctl_tclass, tc_dump_tclass); | 1749 | rtnl_register(PF_UNSPEC, RTM_GETTCLASS, tc_ctl_tclass, tc_dump_tclass); |
1752 | 1750 | ||
1753 | return 0; | 1751 | return 0; |
1754 | } | 1752 | } |
1755 | 1753 | ||
1756 | subsys_initcall(pktsched_init); | 1754 | subsys_initcall(pktsched_init); |
1757 | 1755 |