1 /* flow.c: Generic flow cache.
3 * Copyright (C) 2003 Alexey N. Kuznetsov (kuznet@ms2.inr.ac.ru)
4 * Copyright (C) 2003 David S. Miller (davem@redhat.com)
7 #include <linux/kernel.h>
8 #include <linux/module.h>
9 #include <linux/list.h>
10 #include <linux/jhash.h>
11 #include <linux/interrupt.h>
13 #include <linux/random.h>
14 #include <linux/init.h>
15 #include <linux/slab.h>
16 #include <linux/smp.h>
17 #include <linux/completion.h>
18 #include <linux/percpu.h>
19 #include <linux/bitops.h>
20 #include <linux/notifier.h>
21 #include <linux/cpu.h>
22 #include <linux/cpumask.h>
24 #include <asm/atomic.h>
25 #include <asm/semaphore.h>
27 struct flow_cache_entry {
28 struct flow_cache_entry *next;
37 atomic_t flow_cache_genid = ATOMIC_INIT(0);
39 static u32 flow_hash_shift;
40 #define flow_hash_size (1 << flow_hash_shift)
41 static DEFINE_PER_CPU(struct flow_cache_entry **, flow_tables) = { NULL };
43 #define flow_table(cpu) (per_cpu(flow_tables, cpu))
45 static kmem_cache_t *flow_cachep;
47 static int flow_lwm, flow_hwm;
49 struct flow_percpu_info {
53 } ____cacheline_aligned;
54 static DEFINE_PER_CPU(struct flow_percpu_info, flow_hash_info) = { 0 };
56 #define flow_hash_rnd_recalc(cpu) \
57 (per_cpu(flow_hash_info, cpu).hash_rnd_recalc)
58 #define flow_hash_rnd(cpu) \
59 (per_cpu(flow_hash_info, cpu).hash_rnd)
60 #define flow_count(cpu) \
61 (per_cpu(flow_hash_info, cpu).count)
63 static struct timer_list flow_hash_rnd_timer;
65 #define FLOW_HASH_RND_PERIOD (10 * 60 * HZ)
67 struct flow_flush_info {
69 struct completion completion;
71 static DEFINE_PER_CPU(struct tasklet_struct, flow_flush_tasklets) = { NULL };
73 #define flow_flush_tasklet(cpu) (&per_cpu(flow_flush_tasklets, cpu))
75 static void flow_cache_new_hashrnd(unsigned long arg)
80 flow_hash_rnd_recalc(i) = 1;
82 flow_hash_rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
83 add_timer(&flow_hash_rnd_timer);
86 static void __flow_cache_shrink(int cpu, int shrink_to)
88 struct flow_cache_entry *fle, **flp;
91 for (i = 0; i < flow_hash_size; i++) {
94 flp = &flow_table(cpu)[i];
95 while ((fle = *flp) != NULL && k < shrink_to) {
99 while ((fle = *flp) != NULL) {
102 atomic_dec(fle->object_ref);
103 kmem_cache_free(flow_cachep, fle);
109 static void flow_cache_shrink(int cpu)
111 int shrink_to = flow_lwm / flow_hash_size;
113 __flow_cache_shrink(cpu, shrink_to);
116 static void flow_new_hash_rnd(int cpu)
118 get_random_bytes(&flow_hash_rnd(cpu), sizeof(u32));
119 flow_hash_rnd_recalc(cpu) = 0;
121 __flow_cache_shrink(cpu, 0);
124 static u32 flow_hash_code(struct flowi *key, int cpu)
126 u32 *k = (u32 *) key;
128 return (jhash2(k, (sizeof(*key) / sizeof(u32)), flow_hash_rnd(cpu)) &
129 (flow_hash_size - 1));
132 #if (BITS_PER_LONG == 64)
133 typedef u64 flow_compare_t;
135 typedef u32 flow_compare_t;
138 extern void flowi_is_missized(void);
140 /* I hear what you're saying, use memcmp. But memcmp cannot make
141 * important assumptions that we can here, such as alignment and
144 static int flow_key_compare(struct flowi *key1, struct flowi *key2)
146 flow_compare_t *k1, *k1_lim, *k2;
147 const int n_elem = sizeof(struct flowi) / sizeof(flow_compare_t);
149 if (sizeof(struct flowi) % sizeof(flow_compare_t))
152 k1 = (flow_compare_t *) key1;
153 k1_lim = k1 + n_elem;
155 k2 = (flow_compare_t *) key2;
160 } while (k1 < k1_lim);
165 void *flow_cache_lookup(struct flowi *key, u16 family, u8 dir,
166 flow_resolve_t resolver)
168 struct flow_cache_entry *fle, **head;
173 cpu = smp_processor_id();
176 /* Packet really early in init? Making flow_cache_init a
177 * pre-smp initcall would solve this. --RR */
178 if (!flow_table(cpu))
181 if (flow_hash_rnd_recalc(cpu))
182 flow_new_hash_rnd(cpu);
183 hash = flow_hash_code(key, cpu);
185 head = &flow_table(cpu)[hash];
186 for (fle = *head; fle; fle = fle->next) {
187 if (fle->family == family &&
189 flow_key_compare(key, &fle->key) == 0) {
190 if (fle->genid == atomic_read(&flow_cache_genid)) {
191 void *ret = fle->object;
194 atomic_inc(fle->object_ref);
204 if (flow_count(cpu) > flow_hwm)
205 flow_cache_shrink(cpu);
207 fle = kmem_cache_alloc(flow_cachep, SLAB_ATOMIC);
211 fle->family = family;
213 memcpy(&fle->key, key, sizeof(*key));
224 resolver(key, family, dir, &obj, &obj_ref);
227 fle->genid = atomic_read(&flow_cache_genid);
230 atomic_dec(fle->object_ref);
233 fle->object_ref = obj_ref;
235 atomic_inc(fle->object_ref);
243 static void flow_cache_flush_tasklet(unsigned long data)
245 struct flow_flush_info *info = (void *)data;
249 cpu = smp_processor_id();
250 for (i = 0; i < flow_hash_size; i++) {
251 struct flow_cache_entry *fle;
253 fle = flow_table(cpu)[i];
254 for (; fle; fle = fle->next) {
255 unsigned genid = atomic_read(&flow_cache_genid);
257 if (!fle->object || fle->genid == genid)
261 atomic_dec(fle->object_ref);
265 if (atomic_dec_and_test(&info->cpuleft))
266 complete(&info->completion);
269 static void flow_cache_flush_per_cpu(void *) __attribute__((__unused__));
270 static void flow_cache_flush_per_cpu(void *data)
272 struct flow_flush_info *info = data;
274 struct tasklet_struct *tasklet;
276 cpu = smp_processor_id();
278 tasklet = flow_flush_tasklet(cpu);
279 tasklet->data = (unsigned long)info;
280 tasklet_schedule(tasklet);
283 void flow_cache_flush(void)
285 struct flow_flush_info info;
286 static DECLARE_MUTEX(flow_flush_sem);
288 /* Don't want cpus going down or up during this. */
290 down(&flow_flush_sem);
291 atomic_set(&info.cpuleft, num_online_cpus());
292 init_completion(&info.completion);
295 smp_call_function(flow_cache_flush_per_cpu, &info, 1, 0);
296 flow_cache_flush_tasklet((unsigned long)&info);
299 wait_for_completion(&info.completion);
301 unlock_cpu_hotplug();
304 static void __devinit flow_cache_cpu_prepare(int cpu)
306 struct tasklet_struct *tasklet;
310 (PAGE_SIZE << order) <
311 (sizeof(struct flow_cache_entry *)*flow_hash_size);
315 flow_table(cpu) = (struct flow_cache_entry **)
316 __get_free_pages(GFP_KERNEL, order);
317 if (!flow_table(cpu))
318 panic("NET: failed to allocate flow cache order %lu\n", order);
320 memset(flow_table(cpu), 0, PAGE_SIZE << order);
322 flow_hash_rnd_recalc(cpu) = 1;
325 tasklet = flow_flush_tasklet(cpu);
326 tasklet_init(tasklet, flow_cache_flush_tasklet, 0);
329 #ifdef CONFIG_HOTPLUG_CPU
330 static int flow_cache_cpu(struct notifier_block *nfb,
331 unsigned long action,
334 if (action == CPU_DEAD)
335 __flow_cache_shrink((unsigned long)hcpu, 0);
338 #endif /* CONFIG_HOTPLUG_CPU */
340 static int __init flow_cache_init(void)
344 flow_cachep = kmem_cache_create("flow_cache",
345 sizeof(struct flow_cache_entry),
346 0, SLAB_HWCACHE_ALIGN,
350 panic("NET: failed to allocate flow cache slab\n");
352 flow_hash_shift = 10;
353 flow_lwm = 2 * flow_hash_size;
354 flow_hwm = 4 * flow_hash_size;
356 init_timer(&flow_hash_rnd_timer);
357 flow_hash_rnd_timer.function = flow_cache_new_hashrnd;
358 flow_hash_rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
359 add_timer(&flow_hash_rnd_timer);
362 flow_cache_cpu_prepare(i);
364 hotcpu_notifier(flow_cache_cpu, 0);
368 module_init(flow_cache_init);
370 EXPORT_SYMBOL(flow_cache_genid);
371 EXPORT_SYMBOL(flow_cache_lookup);