ia64/linux-2.6.18-xen.hg

view drivers/net/ifb.c @ 897:329ea0ccb344

balloon: try harder to balloon up under memory pressure.

Currently if the balloon driver is unable to increase the guest's
reservation it assumes the failure was due to reaching its full
allocation, gives up on the ballooning operation and records the limit
it reached as the "hard limit". The driver will not try again until
the target is set again (even to the same value).

However it is possible that ballooning has in fact failed due to
memory pressure in the host and therefore it is desirable to keep
attempting to reach the target in case memory becomes available. The
most likely scenario is that some guests are ballooning down while
others are ballooning up and therefore there is temporary memory
pressure while things stabilise. You would not expect a well behaved
toolstack to ask a domain to balloon to more than its allocation nor
would you expect it to deliberately over-commit memory by setting
balloon targets which exceed the total host memory.

This patch drops the concept of a hard limit and causes the balloon
driver to retry increasing the reservation on a timer in the same
manner as when decreasing the reservation.

Also if we partially succeed in increasing the reservation
(i.e. receive less pages than we asked for) then we may as well keep
those pages rather than returning them to Xen.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Fri Jun 05 14:01:20 2009 +0100 (2009-06-05)
parents 831230e53067
children
line source
1 /* drivers/net/ifb.c:
3 The purpose of this driver is to provide a device that allows
4 for sharing of resources:
6 1) qdiscs/policies that are per device as opposed to system wide.
7 ifb allows for a device which can be redirected to thus providing
8 an impression of sharing.
10 2) Allows for queueing incoming traffic for shaping instead of
11 dropping.
13 The original concept is based on what is known as the IMQ
14 driver initially written by Martin Devera, later rewritten
15 by Patrick McHardy and then maintained by Andre Correa.
17 You need the tc action mirror or redirect to feed this device
18 packets.
20 This program is free software; you can redistribute it and/or
21 modify it under the terms of the GNU General Public License
22 as published by the Free Software Foundation; either version
23 2 of the License, or (at your option) any later version.
25 Authors: Jamal Hadi Salim (2005)
27 */
30 #include <linux/module.h>
31 #include <linux/kernel.h>
32 #include <linux/netdevice.h>
33 #include <linux/etherdevice.h>
34 #include <linux/init.h>
35 #include <linux/moduleparam.h>
36 #include <net/pkt_sched.h>
38 #define TX_TIMEOUT (2*HZ)
40 #define TX_Q_LIMIT 32
41 struct ifb_private {
42 struct net_device_stats stats;
43 struct tasklet_struct ifb_tasklet;
44 int tasklet_pending;
45 /* mostly debug stats leave in for now */
46 unsigned long st_task_enter; /* tasklet entered */
47 unsigned long st_txq_refl_try; /* transmit queue refill attempt */
48 unsigned long st_rxq_enter; /* receive queue entered */
49 unsigned long st_rx2tx_tran; /* receive to trasmit transfers */
50 unsigned long st_rxq_notenter; /*receiveQ not entered, resched */
51 unsigned long st_rx_frm_egr; /* received from egress path */
52 unsigned long st_rx_frm_ing; /* received from ingress path */
53 unsigned long st_rxq_check;
54 unsigned long st_rxq_rsch;
55 struct sk_buff_head rq;
56 struct sk_buff_head tq;
57 };
59 static int numifbs = 2;
61 static void ri_tasklet(unsigned long dev);
62 static int ifb_xmit(struct sk_buff *skb, struct net_device *dev);
63 static struct net_device_stats *ifb_get_stats(struct net_device *dev);
64 static int ifb_open(struct net_device *dev);
65 static int ifb_close(struct net_device *dev);
67 static void ri_tasklet(unsigned long dev)
68 {
70 struct net_device *_dev = (struct net_device *)dev;
71 struct ifb_private *dp = netdev_priv(_dev);
72 struct net_device_stats *stats = &dp->stats;
73 struct sk_buff *skb;
75 dp->st_task_enter++;
76 if ((skb = skb_peek(&dp->tq)) == NULL) {
77 dp->st_txq_refl_try++;
78 if (netif_tx_trylock(_dev)) {
79 dp->st_rxq_enter++;
80 while ((skb = skb_dequeue(&dp->rq)) != NULL) {
81 skb_queue_tail(&dp->tq, skb);
82 dp->st_rx2tx_tran++;
83 }
84 netif_tx_unlock(_dev);
85 } else {
86 /* reschedule */
87 dp->st_rxq_notenter++;
88 goto resched;
89 }
90 }
92 while ((skb = skb_dequeue(&dp->tq)) != NULL) {
93 u32 from = G_TC_FROM(skb->tc_verd);
95 skb->tc_verd = 0;
96 skb->tc_verd = SET_TC_NCLS(skb->tc_verd);
97 stats->tx_packets++;
98 stats->tx_bytes +=skb->len;
99 if (from & AT_EGRESS) {
100 dp->st_rx_frm_egr++;
101 dev_queue_xmit(skb);
102 } else if (from & AT_INGRESS) {
104 dp->st_rx_frm_ing++;
105 netif_rx(skb);
106 } else {
107 dev_kfree_skb(skb);
108 stats->tx_dropped++;
109 }
110 }
112 if (netif_tx_trylock(_dev)) {
113 dp->st_rxq_check++;
114 if ((skb = skb_peek(&dp->rq)) == NULL) {
115 dp->tasklet_pending = 0;
116 if (netif_queue_stopped(_dev))
117 netif_wake_queue(_dev);
118 } else {
119 dp->st_rxq_rsch++;
120 netif_tx_unlock(_dev);
121 goto resched;
122 }
123 netif_tx_unlock(_dev);
124 } else {
125 resched:
126 dp->tasklet_pending = 1;
127 tasklet_schedule(&dp->ifb_tasklet);
128 }
130 }
132 static void __init ifb_setup(struct net_device *dev)
133 {
134 /* Initialize the device structure. */
135 dev->get_stats = ifb_get_stats;
136 dev->hard_start_xmit = ifb_xmit;
137 dev->open = &ifb_open;
138 dev->stop = &ifb_close;
140 /* Fill in device structure with ethernet-generic values. */
141 ether_setup(dev);
142 dev->tx_queue_len = TX_Q_LIMIT;
143 dev->change_mtu = NULL;
144 dev->flags |= IFF_NOARP;
145 dev->flags &= ~IFF_MULTICAST;
146 SET_MODULE_OWNER(dev);
147 random_ether_addr(dev->dev_addr);
148 }
150 static int ifb_xmit(struct sk_buff *skb, struct net_device *dev)
151 {
152 struct ifb_private *dp = netdev_priv(dev);
153 struct net_device_stats *stats = &dp->stats;
154 int ret = 0;
155 u32 from = G_TC_FROM(skb->tc_verd);
157 stats->tx_packets++;
158 stats->tx_bytes+=skb->len;
160 if (!from || !skb->input_dev) {
161 dropped:
162 dev_kfree_skb(skb);
163 stats->rx_dropped++;
164 return ret;
165 } else {
166 /*
167 * note we could be going
168 * ingress -> egress or
169 * egress -> ingress
170 */
171 skb->dev = skb->input_dev;
172 skb->input_dev = dev;
173 if (from & AT_INGRESS) {
174 skb_pull(skb, skb->dev->hard_header_len);
175 } else {
176 if (!(from & AT_EGRESS)) {
177 goto dropped;
178 }
179 }
180 }
182 if (skb_queue_len(&dp->rq) >= dev->tx_queue_len) {
183 netif_stop_queue(dev);
184 }
186 dev->trans_start = jiffies;
187 skb_queue_tail(&dp->rq, skb);
188 if (!dp->tasklet_pending) {
189 dp->tasklet_pending = 1;
190 tasklet_schedule(&dp->ifb_tasklet);
191 }
193 return ret;
194 }
196 static struct net_device_stats *ifb_get_stats(struct net_device *dev)
197 {
198 struct ifb_private *dp = netdev_priv(dev);
199 struct net_device_stats *stats = &dp->stats;
201 pr_debug("tasklets stats %ld:%ld:%ld:%ld:%ld:%ld:%ld:%ld:%ld \n",
202 dp->st_task_enter, dp->st_txq_refl_try, dp->st_rxq_enter,
203 dp->st_rx2tx_tran dp->st_rxq_notenter, dp->st_rx_frm_egr,
204 dp->st_rx_frm_ing, dp->st_rxq_check, dp->st_rxq_rsch );
206 return stats;
207 }
209 static struct net_device **ifbs;
211 /* Number of ifb devices to be set up by this module. */
212 module_param(numifbs, int, 0);
213 MODULE_PARM_DESC(numifbs, "Number of ifb devices");
215 static int ifb_close(struct net_device *dev)
216 {
217 struct ifb_private *dp = netdev_priv(dev);
219 tasklet_kill(&dp->ifb_tasklet);
220 netif_stop_queue(dev);
221 skb_queue_purge(&dp->rq);
222 skb_queue_purge(&dp->tq);
223 return 0;
224 }
226 static int ifb_open(struct net_device *dev)
227 {
228 struct ifb_private *dp = netdev_priv(dev);
230 tasklet_init(&dp->ifb_tasklet, ri_tasklet, (unsigned long)dev);
231 skb_queue_head_init(&dp->rq);
232 skb_queue_head_init(&dp->tq);
233 netif_start_queue(dev);
235 return 0;
236 }
238 static int __init ifb_init_one(int index)
239 {
240 struct net_device *dev_ifb;
241 int err;
243 dev_ifb = alloc_netdev(sizeof(struct ifb_private),
244 "ifb%d", ifb_setup);
246 if (!dev_ifb)
247 return -ENOMEM;
249 if ((err = register_netdev(dev_ifb))) {
250 free_netdev(dev_ifb);
251 dev_ifb = NULL;
252 } else {
253 ifbs[index] = dev_ifb;
254 }
256 return err;
257 }
259 static void ifb_free_one(int index)
260 {
261 unregister_netdev(ifbs[index]);
262 free_netdev(ifbs[index]);
263 }
265 static int __init ifb_init_module(void)
266 {
267 int i, err = 0;
268 ifbs = kmalloc(numifbs * sizeof(void *), GFP_KERNEL);
269 if (!ifbs)
270 return -ENOMEM;
271 for (i = 0; i < numifbs && !err; i++)
272 err = ifb_init_one(i);
273 if (err) {
274 i--;
275 while (--i >= 0)
276 ifb_free_one(i);
277 }
279 return err;
280 }
282 static void __exit ifb_cleanup_module(void)
283 {
284 int i;
286 for (i = 0; i < numifbs; i++)
287 ifb_free_one(i);
288 kfree(ifbs);
289 }
291 module_init(ifb_init_module);
292 module_exit(ifb_cleanup_module);
293 MODULE_LICENSE("GPL");
294 MODULE_AUTHOR("Jamal Hadi Salim");