BUG/MEDIUM: memory: Add a rwlock before freeing memory.

When using lockless pools, add a new rwlock, flush_pool. read-lock it when
getting memory from the pool, so that concurrenct access are still
authorized, but write-lock it when we're about to free memory, in
pool_flush() and pool_gc().
The problem is, when removing an item from the pool, we unreference it
to get the next one, however, that pointer may have been free'd in the
meanwhile, and that could provoke a crash if the pointer has been unmapped.
It should be OK to use a rwlock, as normal operations will still be able
to access the pool concurrently, and calls to pool_flush() and pool_gc()
should be pretty rare.

This should be backported to 2.1, 2.0 and 1.9.
This commit is contained in:
Olivier Houchard 2020-02-01 17:49:31 +01:00
parent 8af97eb4a1
commit 04f5fe87d3
2 changed files with 11 additions and 2 deletions

View File

@ -78,6 +78,7 @@ struct pool_head {
void **free_list;
#ifdef CONFIG_HAP_LOCKLESS_POOLS
uintptr_t seq;
HA_RWLOCK_T flush_lock;
#else
__decl_hathreads(HA_SPINLOCK_T lock); /* the spin lock */
#endif
@ -221,6 +222,7 @@ static inline void *__pool_get_first(struct pool_head *pool)
cmp.seq = pool->seq;
__ha_barrier_load();
HA_RWLOCK_RDLOCK(POOL_LOCK, &pool->flush_lock);
cmp.free_list = pool->free_list;
do {
if (cmp.free_list == NULL)
@ -230,6 +232,7 @@ static inline void *__pool_get_first(struct pool_head *pool)
new.free_list = *POOL_LINK(pool, cmp.free_list);
} while (HA_ATOMIC_DWCAS((void *)&pool->free_list, (void *)&cmp, (void *)&new) == 0);
__ha_barrier_atomic_store();
HA_RWLOCK_RDUNLOCK(POOL_LOCK, &pool->flush_lock);
_HA_ATOMIC_ADD(&pool->used, 1);
#ifdef DEBUG_MEMORY_POOLS

View File

@ -141,6 +141,8 @@ struct pool_head *create_pool(char *name, unsigned int size, unsigned int flags)
}
#ifndef CONFIG_HAP_LOCKLESS_POOLS
HA_SPIN_INIT(&pool->lock);
#else
HA_RWLOCK_INIT(&pool->flush_lock);
#endif
}
pool->users++;
@ -223,6 +225,7 @@ void pool_flush(struct pool_head *pool)
if (!pool)
return;
HA_RWLOCK_WRLOCK(POOL_LOCK, &pool->flush_lock);
do {
cmp.free_list = pool->free_list;
cmp.seq = pool->seq;
@ -230,6 +233,7 @@ void pool_flush(struct pool_head *pool)
new.seq = cmp.seq + 1;
} while (!_HA_ATOMIC_DWCAS(&pool->free_list, &cmp, &new));
__ha_barrier_atomic_store();
HA_RWLOCK_WRUNLOCK(POOL_LOCK, &pool->flush_lock);
next = cmp.free_list;
while (next) {
temp = next;
@ -259,6 +263,7 @@ void pool_gc(struct pool_head *pool_ctx)
return;
list_for_each_entry(entry, &pools, list) {
HA_RWLOCK_WRLOCK(POOL_LOCK, &entry->flush_lock);
while ((int)((volatile int)entry->allocated - (volatile int)entry->used) > (int)entry->minavail) {
struct pool_free_list cmp, new;
@ -275,6 +280,7 @@ void pool_gc(struct pool_head *pool_ctx)
free(cmp.free_list);
_HA_ATOMIC_SUB(&entry->allocated, 1);
}
HA_RWLOCK_WRUNLOCK(POOL_LOCK, &entry->flush_lock);
}
_HA_ATOMIC_STORE(&recurse, 0);
@ -629,7 +635,7 @@ int mem_should_fail(const struct pool_head *pool)
else
ret = 0;
}
HA_SPIN_LOCK(OTHER_LOCK, &mem_fail_lock);
HA_SPIN_LOCK(POOL_LOCK, &mem_fail_lock);
n = snprintf(&mem_fail_str[mem_fail_cur_idx * MEM_FAIL_MAX_CHAR],
MEM_FAIL_MAX_CHAR - 2,
"%d %.18s %d %d", mem_fail_cur_idx, pool->name, ret, tid);
@ -642,7 +648,7 @@ int mem_should_fail(const struct pool_head *pool)
mem_fail_cur_idx++;
if (mem_fail_cur_idx == MEM_FAIL_MAX_STR)
mem_fail_cur_idx = 0;
HA_SPIN_UNLOCK(OTHER_LOCK, &mem_fail_lock);
HA_SPIN_UNLOCK(POOL_LOCK, &mem_fail_lock);
return ret;
}