percpu: allow pcpu_alloc() to be called with IRQs off
authorJiri Kosina <jkosina@suse.cz>
Wed, 28 Oct 2009 15:25:59 +0000 (00:25 +0900)
committerTejun Heo <tj@kernel.org>
Wed, 28 Oct 2009 15:25:59 +0000 (00:25 +0900)
pcpu_alloc() and pcpu_extend_area_map() perform a series of
spin_lock_irq()/spin_unlock_irq() calls, which make them unsafe
with respect to being called from contexts which have IRQs off.

This patch converts the code to perform save/restore of flags instead,
making pcpu_alloc() (or __alloc_percpu() respectively) to be called
from early kernel startup stage, where IRQs are off.

This is needed for proper initialization of per-cpu rq_weight data from
sched_init().

tj: added comment explaining why irqsave/restore is used in alloc path.

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Tejun Heo <tj@kernel.org>
mm/percpu.c

index 6af78c1..d907971 100644 (file)
@@ -153,7 +153,10 @@ static int pcpu_reserved_chunk_limit;
  *
  * During allocation, pcpu_alloc_mutex is kept locked all the time and
  * pcpu_lock is grabbed and released as necessary.  All actual memory
- * allocations are done using GFP_KERNEL with pcpu_lock released.
+ * allocations are done using GFP_KERNEL with pcpu_lock released.  In
+ * general, percpu memory can't be allocated with irq off but
+ * irqsave/restore are still used in alloc path so that it can be used
+ * from early init path - sched_init() specifically.
  *
  * Free path accesses and alters only the index data structures, so it
  * can be safely called from atomic context.  When memory needs to be
@@ -366,7 +369,7 @@ static struct pcpu_chunk *pcpu_chunk_addr_search(void *addr)
  * RETURNS:
  * 0 if noop, 1 if successfully extended, -errno on failure.
  */
-static int pcpu_extend_area_map(struct pcpu_chunk *chunk)
+static int pcpu_extend_area_map(struct pcpu_chunk *chunk, unsigned long *flags)
 {
        int new_alloc;
        int *new;
@@ -376,7 +379,7 @@ static int pcpu_extend_area_map(struct pcpu_chunk *chunk)
        if (chunk->map_alloc >= chunk->map_used + 2)
                return 0;
 
-       spin_unlock_irq(&pcpu_lock);
+       spin_unlock_irqrestore(&pcpu_lock, *flags);
 
        new_alloc = PCPU_DFL_MAP_ALLOC;
        while (new_alloc < chunk->map_used + 2)
@@ -384,7 +387,7 @@ static int pcpu_extend_area_map(struct pcpu_chunk *chunk)
 
        new = pcpu_mem_alloc(new_alloc * sizeof(new[0]));
        if (!new) {
-               spin_lock_irq(&pcpu_lock);
+               spin_lock_irqsave(&pcpu_lock, *flags);
                return -ENOMEM;
        }
 
@@ -393,7 +396,7 @@ static int pcpu_extend_area_map(struct pcpu_chunk *chunk)
         * could have happened inbetween, so map_used couldn't have
         * grown.
         */
-       spin_lock_irq(&pcpu_lock);
+       spin_lock_irqsave(&pcpu_lock, *flags);
        BUG_ON(new_alloc < chunk->map_used + 2);
 
        size = chunk->map_alloc * sizeof(chunk->map[0]);
@@ -1047,6 +1050,7 @@ static void *pcpu_alloc(size_t size, size_t align, bool reserved)
        struct pcpu_chunk *chunk;
        const char *err;
        int slot, off;
+       unsigned long flags;
 
        if (unlikely(!size || size > PCPU_MIN_UNIT_SIZE || align > PAGE_SIZE)) {
                WARN(true, "illegal size (%zu) or align (%zu) for "
@@ -1055,13 +1059,13 @@ static void *pcpu_alloc(size_t size, size_t align, bool reserved)
        }
 
        mutex_lock(&pcpu_alloc_mutex);
-       spin_lock_irq(&pcpu_lock);
+       spin_lock_irqsave(&pcpu_lock, flags);
 
        /* serve reserved allocations from the reserved chunk if available */
        if (reserved && pcpu_reserved_chunk) {
                chunk = pcpu_reserved_chunk;
                if (size > chunk->contig_hint ||
-                   pcpu_extend_area_map(chunk) < 0) {
+                   pcpu_extend_area_map(chunk, &flags) < 0) {
                        err = "failed to extend area map of reserved chunk";
                        goto fail_unlock;
                }
@@ -1079,7 +1083,7 @@ restart:
                        if (size > chunk->contig_hint)
                                continue;
 
-                       switch (pcpu_extend_area_map(chunk)) {
+                       switch (pcpu_extend_area_map(chunk, &flags)) {
                        case 0:
                                break;
                        case 1:
@@ -1096,7 +1100,7 @@ restart:
        }
 
        /* hmmm... no space left, create a new chunk */
-       spin_unlock_irq(&pcpu_lock);
+       spin_unlock_irqrestore(&pcpu_lock, flags);
 
        chunk = alloc_pcpu_chunk();
        if (!chunk) {
@@ -1104,16 +1108,16 @@ restart:
                goto fail_unlock_mutex;
        }
 
-       spin_lock_irq(&pcpu_lock);
+       spin_lock_irqsave(&pcpu_lock, flags);
        pcpu_chunk_relocate(chunk, -1);
        goto restart;
 
 area_found:
-       spin_unlock_irq(&pcpu_lock);
+       spin_unlock_irqrestore(&pcpu_lock, flags);
 
        /* populate, map and clear the area */
        if (pcpu_populate_chunk(chunk, off, size)) {
-               spin_lock_irq(&pcpu_lock);
+               spin_lock_irqsave(&pcpu_lock, flags);
                pcpu_free_area(chunk, off);
                err = "failed to populate";
                goto fail_unlock;
@@ -1125,7 +1129,7 @@ area_found:
        return __addr_to_pcpu_ptr(chunk->base_addr + off);
 
 fail_unlock:
-       spin_unlock_irq(&pcpu_lock);
+       spin_unlock_irqrestore(&pcpu_lock, flags);
 fail_unlock_mutex:
        mutex_unlock(&pcpu_alloc_mutex);
        if (warn_limit) {