summaryrefslogtreecommitdiffstats
path: root/fs/bio.c
diff options
context:
space:
mode:
authorIngo Molnar <mingo@elte.hu>2008-12-31 08:19:48 +0100
committerIngo Molnar <mingo@elte.hu>2008-12-31 08:19:48 +0100
commit818fa7f3908c7bd6c0045e9d94dc23a899ef6144 (patch)
treead3435c3f57c8222ad61709b716168932f13be6c /fs/bio.c
parent3fd4bc015ef879a7d2b955ce97fb125e3a51ba7e (diff)
parent5fdf7e5975a0b0f6a0370655612c5dca3fd6311b (diff)
downloadop-kernel-dev-818fa7f3908c7bd6c0045e9d94dc23a899ef6144.zip
op-kernel-dev-818fa7f3908c7bd6c0045e9d94dc23a899ef6144.tar.gz
Merge branch 'tracing/kmemtrace' into tracing/kmemtrace2
Diffstat (limited to 'fs/bio.c')
-rw-r--r--fs/bio.c320
1 files changed, 243 insertions, 77 deletions
diff --git a/fs/bio.c b/fs/bio.c
index df99c88..711cee1 100644
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -31,7 +31,11 @@
DEFINE_TRACE(block_split);
-static struct kmem_cache *bio_slab __read_mostly;
+/*
+ * Test patch to inline a certain number of bi_io_vec's inside the bio
+ * itself, to shrink a bio data allocation from two mempool calls to one
+ */
+#define BIO_INLINE_VECS 4
static mempool_t *bio_split_pool __read_mostly;
@@ -40,9 +44,8 @@ static mempool_t *bio_split_pool __read_mostly;
* break badly! cannot be bigger than what you can fit into an
* unsigned short
*/
-
#define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
-static struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = {
+struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = {
BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES),
};
#undef BV
@@ -53,12 +56,121 @@ static struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = {
*/
struct bio_set *fs_bio_set;
+/*
+ * Our slab pool management
+ */
+struct bio_slab {
+ struct kmem_cache *slab;
+ unsigned int slab_ref;
+ unsigned int slab_size;
+ char name[8];
+};
+static DEFINE_MUTEX(bio_slab_lock);
+static struct bio_slab *bio_slabs;
+static unsigned int bio_slab_nr, bio_slab_max;
+
+static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size)
+{
+ unsigned int sz = sizeof(struct bio) + extra_size;
+ struct kmem_cache *slab = NULL;
+ struct bio_slab *bslab;
+ unsigned int i, entry = -1;
+
+ mutex_lock(&bio_slab_lock);
+
+ i = 0;
+ while (i < bio_slab_nr) {
+ struct bio_slab *bslab = &bio_slabs[i];
+
+ if (!bslab->slab && entry == -1)
+ entry = i;
+ else if (bslab->slab_size == sz) {
+ slab = bslab->slab;
+ bslab->slab_ref++;
+ break;
+ }
+ i++;
+ }
+
+ if (slab)
+ goto out_unlock;
+
+ if (bio_slab_nr == bio_slab_max && entry == -1) {
+ bio_slab_max <<= 1;
+ bio_slabs = krealloc(bio_slabs,
+ bio_slab_max * sizeof(struct bio_slab),
+ GFP_KERNEL);
+ if (!bio_slabs)
+ goto out_unlock;
+ }
+ if (entry == -1)
+ entry = bio_slab_nr++;
+
+ bslab = &bio_slabs[entry];
+
+ snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry);
+ slab = kmem_cache_create(bslab->name, sz, 0, SLAB_HWCACHE_ALIGN, NULL);
+ if (!slab)
+ goto out_unlock;
+
+ printk("bio: create slab <%s> at %d\n", bslab->name, entry);
+ bslab->slab = slab;
+ bslab->slab_ref = 1;
+ bslab->slab_size = sz;
+out_unlock:
+ mutex_unlock(&bio_slab_lock);
+ return slab;
+}
+
+static void bio_put_slab(struct bio_set *bs)
+{
+ struct bio_slab *bslab = NULL;
+ unsigned int i;
+
+ mutex_lock(&bio_slab_lock);
+
+ for (i = 0; i < bio_slab_nr; i++) {
+ if (bs->bio_slab == bio_slabs[i].slab) {
+ bslab = &bio_slabs[i];
+ break;
+ }
+ }
+
+ if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n"))
+ goto out;
+
+ WARN_ON(!bslab->slab_ref);
+
+ if (--bslab->slab_ref)
+ goto out;
+
+ kmem_cache_destroy(bslab->slab);
+ bslab->slab = NULL;
+
+out:
+ mutex_unlock(&bio_slab_lock);
+}
+
unsigned int bvec_nr_vecs(unsigned short idx)
{
return bvec_slabs[idx].nr_vecs;
}
-struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx, struct bio_set *bs)
+void bvec_free_bs(struct bio_set *bs, struct bio_vec *bv, unsigned int idx)
+{
+ BIO_BUG_ON(idx >= BIOVEC_NR_POOLS);
+
+ if (idx == BIOVEC_MAX_IDX)
+ mempool_free(bv, bs->bvec_pool);
+ else {
+ struct biovec_slab *bvs = bvec_slabs + idx;
+
+ kmem_cache_free(bvs->slab, bv);
+ }
+}
+
+struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx,
+ struct bio_set *bs)
{
struct bio_vec *bvl;
@@ -67,60 +179,85 @@ struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx, struct
* If not, this is a bio_kmalloc() allocation and just do a
* kzalloc() for the exact number of vecs right away.
*/
- if (bs) {
+ if (!bs)
+ bvl = kmalloc(nr * sizeof(struct bio_vec), gfp_mask);
+
+ /*
+ * see comment near bvec_array define!
+ */
+ switch (nr) {
+ case 1:
+ *idx = 0;
+ break;
+ case 2 ... 4:
+ *idx = 1;
+ break;
+ case 5 ... 16:
+ *idx = 2;
+ break;
+ case 17 ... 64:
+ *idx = 3;
+ break;
+ case 65 ... 128:
+ *idx = 4;
+ break;
+ case 129 ... BIO_MAX_PAGES:
+ *idx = 5;
+ break;
+ default:
+ return NULL;
+ }
+
+ /*
+ * idx now points to the pool we want to allocate from. only the
+ * 1-vec entry pool is mempool backed.
+ */
+ if (*idx == BIOVEC_MAX_IDX) {
+fallback:
+ bvl = mempool_alloc(bs->bvec_pool, gfp_mask);
+ } else {
+ struct biovec_slab *bvs = bvec_slabs + *idx;
+ gfp_t __gfp_mask = gfp_mask & ~(__GFP_WAIT | __GFP_IO);
+
/*
- * see comment near bvec_array define!
+ * Make this allocation restricted and don't dump info on
+ * allocation failures, since we'll fallback to the mempool
+ * in case of failure.
*/
- switch (nr) {
- case 1:
- *idx = 0;
- break;
- case 2 ... 4:
- *idx = 1;
- break;
- case 5 ... 16:
- *idx = 2;
- break;
- case 17 ... 64:
- *idx = 3;
- break;
- case 65 ... 128:
- *idx = 4;
- break;
- case 129 ... BIO_MAX_PAGES:
- *idx = 5;
- break;
- default:
- return NULL;
- }
+ __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN;
/*
- * idx now points to the pool we want to allocate from
+ * Try a slab allocation. If this fails and __GFP_WAIT
+ * is set, retry with the 1-entry mempool
*/
- bvl = mempool_alloc(bs->bvec_pools[*idx], gfp_mask);
- if (bvl)
- memset(bvl, 0,
- bvec_nr_vecs(*idx) * sizeof(struct bio_vec));
- } else
- bvl = kzalloc(nr * sizeof(struct bio_vec), gfp_mask);
+ bvl = kmem_cache_alloc(bvs->slab, __gfp_mask);
+ if (unlikely(!bvl && (gfp_mask & __GFP_WAIT))) {
+ *idx = BIOVEC_MAX_IDX;
+ goto fallback;
+ }
+ }
return bvl;
}
-void bio_free(struct bio *bio, struct bio_set *bio_set)
+void bio_free(struct bio *bio, struct bio_set *bs)
{
- if (bio->bi_io_vec) {
- const int pool_idx = BIO_POOL_IDX(bio);
+ void *p;
- BIO_BUG_ON(pool_idx >= BIOVEC_NR_POOLS);
-
- mempool_free(bio->bi_io_vec, bio_set->bvec_pools[pool_idx]);
- }
+ if (bio_has_allocated_vec(bio))
+ bvec_free_bs(bs, bio->bi_io_vec, BIO_POOL_IDX(bio));
if (bio_integrity(bio))
- bio_integrity_free(bio, bio_set);
+ bio_integrity_free(bio, bs);
+
+ /*
+ * If we have front padding, adjust the bio pointer before freeing
+ */
+ p = bio;
+ if (bs->front_pad)
+ p -= bs->front_pad;
- mempool_free(bio, bio_set->bio_pool);
+ mempool_free(p, bs->bio_pool);
}
/*
@@ -133,7 +270,8 @@ static void bio_fs_destructor(struct bio *bio)
static void bio_kmalloc_destructor(struct bio *bio)
{
- kfree(bio->bi_io_vec);
+ if (bio_has_allocated_vec(bio))
+ kfree(bio->bi_io_vec);
kfree(bio);
}
@@ -157,16 +295,20 @@ void bio_init(struct bio *bio)
* for a &struct bio to become free. If a %NULL @bs is passed in, we will
* fall back to just using @kmalloc to allocate the required memory.
*
- * allocate bio and iovecs from the memory pools specified by the
- * bio_set structure, or @kmalloc if none given.
+ * Note that the caller must set ->bi_destructor on succesful return
+ * of a bio, to do the appropriate freeing of the bio once the reference
+ * count drops to zero.
**/
struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
{
- struct bio *bio;
+ struct bio *bio = NULL;
+
+ if (bs) {
+ void *p = mempool_alloc(bs->bio_pool, gfp_mask);
- if (bs)
- bio = mempool_alloc(bs->bio_pool, gfp_mask);
- else
+ if (p)
+ bio = p + bs->front_pad;
+ } else
bio = kmalloc(sizeof(*bio), gfp_mask);
if (likely(bio)) {
@@ -176,7 +318,15 @@ struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
if (likely(nr_iovecs)) {
unsigned long uninitialized_var(idx);
- bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs);
+ if (nr_iovecs <= BIO_INLINE_VECS) {
+ idx = 0;
+ bvl = bio->bi_inline_vecs;
+ nr_iovecs = BIO_INLINE_VECS;
+ } else {
+ bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx,
+ bs);
+ nr_iovecs = bvec_nr_vecs(idx);
+ }
if (unlikely(!bvl)) {
if (bs)
mempool_free(bio, bs->bio_pool);
@@ -186,7 +336,7 @@ struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
goto out;
}
bio->bi_flags |= idx << BIO_POOL_OFFSET;
- bio->bi_max_vecs = bvec_nr_vecs(idx);
+ bio->bi_max_vecs = nr_iovecs;
}
bio->bi_io_vec = bvl;
}
@@ -1346,30 +1496,18 @@ EXPORT_SYMBOL(bio_sector_offset);
*/
static int biovec_create_pools(struct bio_set *bs, int pool_entries)
{
- int i;
+ struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX;
- for (i = 0; i < BIOVEC_NR_POOLS; i++) {
- struct biovec_slab *bp = bvec_slabs + i;
- mempool_t **bvp = bs->bvec_pools + i;
+ bs->bvec_pool = mempool_create_slab_pool(pool_entries, bp->slab);
+ if (!bs->bvec_pool)
+ return -ENOMEM;
- *bvp = mempool_create_slab_pool(pool_entries, bp->slab);
- if (!*bvp)
- return -ENOMEM;
- }
return 0;
}
static void biovec_free_pools(struct bio_set *bs)
{
- int i;
-
- for (i = 0; i < BIOVEC_NR_POOLS; i++) {
- mempool_t *bvp = bs->bvec_pools[i];
-
- if (bvp)
- mempool_destroy(bvp);
- }
-
+ mempool_destroy(bs->bvec_pool);
}
void bioset_free(struct bio_set *bs)
@@ -1379,25 +1517,49 @@ void bioset_free(struct bio_set *bs)
bioset_integrity_free(bs);
biovec_free_pools(bs);
+ bio_put_slab(bs);
kfree(bs);
}
-struct bio_set *bioset_create(int bio_pool_size, int bvec_pool_size)
+/**
+ * bioset_create - Create a bio_set
+ * @pool_size: Number of bio and bio_vecs to cache in the mempool
+ * @front_pad: Number of bytes to allocate in front of the returned bio
+ *
+ * Description:
+ * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller
+ * to ask for a number of bytes to be allocated in front of the bio.
+ * Front pad allocation is useful for embedding the bio inside
+ * another structure, to avoid allocating extra data to go with the bio.
+ * Note that the bio must be embedded at the END of that structure always,
+ * or things will break badly.
+ */
+struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad)
{
- struct bio_set *bs = kzalloc(sizeof(*bs), GFP_KERNEL);
+ unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec);
+ struct bio_set *bs;
+ bs = kzalloc(sizeof(*bs), GFP_KERNEL);
if (!bs)
return NULL;
- bs->bio_pool = mempool_create_slab_pool(bio_pool_size, bio_slab);
+ bs->front_pad = front_pad;
+
+ bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad);
+ if (!bs->bio_slab) {
+ kfree(bs);
+ return NULL;
+ }
+
+ bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab);
if (!bs->bio_pool)
goto bad;
- if (bioset_integrity_create(bs, bio_pool_size))
+ if (bioset_integrity_create(bs, pool_size))
goto bad;
- if (!biovec_create_pools(bs, bvec_pool_size))
+ if (!biovec_create_pools(bs, pool_size))
return bs;
bad:
@@ -1421,12 +1583,16 @@ static void __init biovec_init_slabs(void)
static int __init init_bio(void)
{
- bio_slab = KMEM_CACHE(bio, SLAB_HWCACHE_ALIGN|SLAB_PANIC);
+ bio_slab_max = 2;
+ bio_slab_nr = 0;
+ bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL);
+ if (!bio_slabs)
+ panic("bio: can't allocate bios\n");
bio_integrity_init_slab();
biovec_init_slabs();
- fs_bio_set = bioset_create(BIO_POOL_SIZE, 2);
+ fs_bio_set = bioset_create(BIO_POOL_SIZE, 0);
if (!fs_bio_set)
panic("bio: can't allocate bios\n");
OpenPOWER on IntegriCloud