aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorNick Piggin <npiggin@suse.de>2009-02-25 00:22:45 -0500
committerIngo Molnar <mingo@elte.hu>2009-02-25 06:27:08 -0500
commit15d0d3b3371227f846b9f644547fde081c7e1c0c (patch)
tree11170d41f38123923208c44fe957730350a4e31a
parentf7e603ad8f78cd3b59e33fa72707da0cbabdf699 (diff)
generic IPI: simplify barriers and locking
Simplify the barriers in generic remote function call interrupt code. Firstly, just unconditionally take the lock and check the list in the generic_call_function_single_interrupt IPI handler. As we've just taken an IPI here, the chances are fairly high that there will be work on the list for us, so do the locking unconditionally. This removes the tricky lockless list_empty check and dubious barriers. The change looks bigger than it is because it is just removing an outer loop. Secondly, clarify architecture specific IPI locking rules. Generic code has no tools to impose any sane ordering on IPIs if they go outside normal cache coherency, ergo the arch code must make them appear to obey cache coherency as a "memory operation" to initiate an IPI, and a "memory operation" to receive one. This way at least they can be reasoned about in generic code, and smp_mb used to provide ordering. The combination of these two changes means that explict barriers can be taken out of queue handling for the single case -- shared data is explicitly locked, and ipi ordering must conform to that, so no barriers needed. An extra barrier is needed in the many handler, so as to ensure we load the list element after the IPI is received. Does any architecture actually *need* these barriers? For the initiator I could see it, but for the handler I would be surprised. So the other thing we could do for simplicity is just to require that, rather than just matching with cache coherency, we just require a full barrier before generating an IPI, and after receiving an IPI. In which case, the smp_mb()s can go away. But just for now, we'll be on the safe side and use the barriers (they're in the slow case anyway). Signed-off-by: Nick Piggin <npiggin@suse.de> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: linux-arch@vger.kernel.org Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Suresh Siddha <suresh.b.siddha@intel.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
-rw-r--r--kernel/smp.c83
1 files changed, 44 insertions, 39 deletions
diff --git a/kernel/smp.c b/kernel/smp.c
index bbedbb7efe32..6ecf4b9895d4 100644
--- a/kernel/smp.c
+++ b/kernel/smp.c
@@ -74,9 +74,16 @@ static void generic_exec_single(int cpu, struct call_single_data *data)
74 spin_unlock_irqrestore(&dst->lock, flags); 74 spin_unlock_irqrestore(&dst->lock, flags);
75 75
76 /* 76 /*
77 * Make the list addition visible before sending the ipi. 77 * The list addition should be visible before sending the IPI
78 * handler locks the list to pull the entry off it because of
79 * normal cache coherency rules implied by spinlocks.
80 *
81 * If IPIs can go out of order to the cache coherency protocol
82 * in an architecture, sufficient synchronisation should be added
83 * to arch code to make it appear to obey cache coherency WRT
84 * locking and barrier primitives. Generic code isn't really equipped
85 * to do the right thing...
78 */ 86 */
79 smp_mb();
80 87
81 if (ipi) 88 if (ipi)
82 arch_send_call_function_single_ipi(cpu); 89 arch_send_call_function_single_ipi(cpu);
@@ -104,6 +111,14 @@ void generic_smp_call_function_interrupt(void)
104 int cpu = get_cpu(); 111 int cpu = get_cpu();
105 112
106 /* 113 /*
114 * Ensure entry is visible on call_function_queue after we have
115 * entered the IPI. See comment in smp_call_function_many.
116 * If we don't have this, then we may miss an entry on the list
117 * and never get another IPI to process it.
118 */
119 smp_mb();
120
121 /*
107 * It's ok to use list_for_each_rcu() here even though we may delete 122 * It's ok to use list_for_each_rcu() here even though we may delete
108 * 'pos', since list_del_rcu() doesn't clear ->next 123 * 'pos', since list_del_rcu() doesn't clear ->next
109 */ 124 */
@@ -154,49 +169,37 @@ void generic_smp_call_function_single_interrupt(void)
154{ 169{
155 struct call_single_queue *q = &__get_cpu_var(call_single_queue); 170 struct call_single_queue *q = &__get_cpu_var(call_single_queue);
156 LIST_HEAD(list); 171 LIST_HEAD(list);
172 unsigned int data_flags;
157 173
158 /* 174 spin_lock(&q->lock);
159 * Need to see other stores to list head for checking whether 175 list_replace_init(&q->list, &list);
160 * list is empty without holding q->lock 176 spin_unlock(&q->lock);
161 */
162 smp_read_barrier_depends();
163 while (!list_empty(&q->list)) {
164 unsigned int data_flags;
165
166 spin_lock(&q->lock);
167 list_replace_init(&q->list, &list);
168 spin_unlock(&q->lock);
169 177
170 while (!list_empty(&list)) { 178 while (!list_empty(&list)) {
171 struct call_single_data *data; 179 struct call_single_data *data;
172 180
173 data = list_entry(list.next, struct call_single_data, 181 data = list_entry(list.next, struct call_single_data,
174 list); 182 list);
175 list_del(&data->list); 183 list_del(&data->list);
176 184
177 /*
178 * 'data' can be invalid after this call if
179 * flags == 0 (when called through
180 * generic_exec_single(), so save them away before
181 * making the call.
182 */
183 data_flags = data->flags;
184
185 data->func(data->info);
186
187 if (data_flags & CSD_FLAG_WAIT) {
188 smp_wmb();
189 data->flags &= ~CSD_FLAG_WAIT;
190 } else if (data_flags & CSD_FLAG_LOCK) {
191 smp_wmb();
192 data->flags &= ~CSD_FLAG_LOCK;
193 } else if (data_flags & CSD_FLAG_ALLOC)
194 kfree(data);
195 }
196 /* 185 /*
197 * See comment on outer loop 186 * 'data' can be invalid after this call if
187 * flags == 0 (when called through
188 * generic_exec_single(), so save them away before
189 * making the call.
198 */ 190 */
199 smp_read_barrier_depends(); 191 data_flags = data->flags;
192
193 data->func(data->info);
194
195 if (data_flags & CSD_FLAG_WAIT) {
196 smp_wmb();
197 data->flags &= ~CSD_FLAG_WAIT;
198 } else if (data_flags & CSD_FLAG_LOCK) {
199 smp_wmb();
200 data->flags &= ~CSD_FLAG_LOCK;
201 } else if (data_flags & CSD_FLAG_ALLOC)
202 kfree(data);
200 } 203 }
201} 204}
202 205
@@ -375,6 +378,8 @@ void smp_call_function_many(const struct cpumask *mask,
375 378
376 /* 379 /*
377 * Make the list addition visible before sending the ipi. 380 * Make the list addition visible before sending the ipi.
381 * (IPIs must obey or appear to obey normal Linux cache coherency
382 * rules -- see comment in generic_exec_single).
378 */ 383 */
379 smp_mb(); 384 smp_mb();
380 385