SLUB: Fix early boot GFP_DMA allocations

Recent change to use slab allocations earlier exposed a bug where
SLUB can call schedule_work and try to call sysfs before it is
safe to do so.

Reported-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Tested-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>

1 files changed, 14 insertions, 3 deletions

diff --git a/mm/slub.c b/mm/slub.c
index 30354bfeb43d..dcbfda0b02ed 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -2610,6 +2610,7 @@ static noinline struct kmem_cache *dma_kmalloc_cache(int index, gfp_t flags)
         struct kmem_cache *s;
         char *text;
         size_t realsize;
+        unsigned long slabflags;
 
         s = kmalloc_caches_dma[index];
         if (s)
@@ -2631,9 +2632,18 @@ static noinline struct kmem_cache *dma_kmalloc_cache(int index, gfp_t flags)
                          (unsigned int)realsize);
         s = kmalloc(kmem_size, flags & ~SLUB_DMA);
 
+        /*
+         * Must defer sysfs creation to a workqueue because we don't know
+         * what context we are called from. Before sysfs comes up, we don't
+         * need to do anything because our sysfs initcall will start by
+         * adding all existing slabs to sysfs.
+         */
+        slabflags = SLAB_CACHE_DMA;
+        if (slab_state >= SYSFS)
+                slabflags |= __SYSFS_ADD_DEFERRED;
+
         if (!s || !text || !kmem_cache_open(s, flags, text,
-                        realsize, ARCH_KMALLOC_MINALIGN,
-                        SLAB_CACHE_DMA|__SYSFS_ADD_DEFERRED, NULL)) {
+                        realsize, ARCH_KMALLOC_MINALIGN, slabflags, NULL)) {
                 kfree(s);
                 kfree(text);
                 goto unlock_out;
@@ -2642,7 +2652,8 @@ static noinline struct kmem_cache *dma_kmalloc_cache(int index, gfp_t flags)
         list_add(&s->list, &slab_caches);
         kmalloc_caches_dma[index] = s;
 
-        schedule_work(&sysfs_add_work);
+        if (slab_state >= SYSFS)
+                schedule_work(&sysfs_add_work);
 
 unlock_out:
         up_write(&slub_lock);