Changeset 10e16a7 in mainline for generic/src/mm/slab.c

Timestamp:

2006-02-04T13:51:35Z (19 years ago)

Author:

Ondrej Palkovsky <ondrap@…>

Branches:

lfn, master, serial, ticket/834-toolchain-update, topic/msim-upgrade, topic/simplify-dev-export

Children:

428aabf

Parents:

c5613b7

Message:

Added scheduler queues output. The scheduler is buggy - on SMP
the cpus never get tu cpu_sleep, in slab2 test on 4 cpus everything
is on the first cpu.
The slab allocator passes tests in this configuration, but in slightly
different(more efficient) locking order it panics. TODO Find out why
does it panic.

File:

• generic/src/mm/slab.c (modified) (4 diffs)

generic/src/mm/slab.c

@@ -77 +77 @@
  * magazines.
  *
- * 
+ * TODO: For better CPU-scaling the magazine allocation strategy should
+ * be extended. Currently, if the cache does not have magazine, it asks
+ * for non-cpu cached magazine cache to provide one. It might be feasible
+ * to add cpu-cached magazine cache (which would allocate it's magazines
+ * from non-cpu-cached mag. cache). This would provide a nice per-cpu
+ * buffer. The other possibility is to use the per-cache 
+ * 'empty-magazine-list', which decreases competing for 1 per-system
+ * magazine cache.
+ *
  */
 
@@ -296 +304 @@
  * Free all objects in magazine and free memory associated with magazine
  *
- * Assume mag_cache[cpu].lock is locked 
+ * Assume cache->lock is held
  *
  * @return Number of freed pages
@@ -620 +628 @@
         
         spinlock_unlock(&cache->lock);
+        /* We can release the cache locks now */
         if (flags & SLAB_RECLAIM_ALL) {
                 for (i=0; i < config.cpu_count; i++)
@@ -778 +787 @@
 {
         int idx;
-
+        
         ASSERT( size && size <= (1 << SLAB_MAX_MALLOC_W));