Merge branch 'raid56-experimental' into for-linus-3.9
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Conflicts:
fs/btrfs/ctree.h
fs/btrfs/extent-tree.c
fs/btrfs/inode.c
fs/btrfs/volumes.c
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 5cd44e2..b3ecca4 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -31,6 +31,7 @@
#include "print-tree.h"
#include "transaction.h"
#include "volumes.h"
+#include "raid56.h"
#include "locking.h"
#include "free-space-cache.h"
#include "math.h"
@@ -1852,6 +1853,8 @@
*actual_bytes = discarded_bytes;
+ if (ret == -EOPNOTSUPP)
+ ret = 0;
return ret;
}
@@ -2440,6 +2443,16 @@
return ret;
}
+static int refs_newer(struct btrfs_delayed_ref_root *delayed_refs, int seq,
+ int count)
+{
+ int val = atomic_read(&delayed_refs->ref_seq);
+
+ if (val < seq || val >= seq + count)
+ return 1;
+ return 0;
+}
+
/*
* this starts processing the delayed reference count updates and
* extent insertions we have queued up so far. count can be
@@ -2474,6 +2487,44 @@
delayed_refs = &trans->transaction->delayed_refs;
INIT_LIST_HEAD(&cluster);
+ if (count == 0) {
+ count = delayed_refs->num_entries * 2;
+ run_most = 1;
+ }
+
+ if (!run_all && !run_most) {
+ int old;
+ int seq = atomic_read(&delayed_refs->ref_seq);
+
+progress:
+ old = atomic_cmpxchg(&delayed_refs->procs_running_refs, 0, 1);
+ if (old) {
+ DEFINE_WAIT(__wait);
+ if (delayed_refs->num_entries < 16348)
+ return 0;
+
+ prepare_to_wait(&delayed_refs->wait, &__wait,
+ TASK_UNINTERRUPTIBLE);
+
+ old = atomic_cmpxchg(&delayed_refs->procs_running_refs, 0, 1);
+ if (old) {
+ schedule();
+ finish_wait(&delayed_refs->wait, &__wait);
+
+ if (!refs_newer(delayed_refs, seq, 256))
+ goto progress;
+ else
+ return 0;
+ } else {
+ finish_wait(&delayed_refs->wait, &__wait);
+ goto again;
+ }
+ }
+
+ } else {
+ atomic_inc(&delayed_refs->procs_running_refs);
+ }
+
again:
loops = 0;
spin_lock(&delayed_refs->lock);
@@ -2482,10 +2533,6 @@
delayed_refs->run_delayed_start = find_middle(&delayed_refs->root);
#endif
- if (count == 0) {
- count = delayed_refs->num_entries * 2;
- run_most = 1;
- }
while (1) {
if (!(run_all || run_most) &&
delayed_refs->num_heads_ready < 64)
@@ -2508,9 +2555,12 @@
btrfs_release_ref_cluster(&cluster);
spin_unlock(&delayed_refs->lock);
btrfs_abort_transaction(trans, root, ret);
+ atomic_dec(&delayed_refs->procs_running_refs);
return ret;
}
+ atomic_add(ret, &delayed_refs->ref_seq);
+
count -= min_t(unsigned long, ret, count);
if (count == 0)
@@ -2579,6 +2629,11 @@
goto again;
}
out:
+ atomic_dec(&delayed_refs->procs_running_refs);
+ smp_mb();
+ if (waitqueue_active(&delayed_refs->wait))
+ wake_up(&delayed_refs->wait);
+
spin_unlock(&delayed_refs->lock);
assert_qgroups_uptodate(trans);
return 0;
@@ -3284,6 +3339,7 @@
u64 num_devices = root->fs_info->fs_devices->rw_devices +
root->fs_info->fs_devices->missing_devices;
u64 target;
+ u64 tmp;
/*
* see if restripe for this chunk_type is in progress, if so
@@ -3300,30 +3356,32 @@
}
spin_unlock(&root->fs_info->balance_lock);
+ /* First, mask out the RAID levels which aren't possible */
if (num_devices == 1)
- flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
+ flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0 |
+ BTRFS_BLOCK_GROUP_RAID5);
+ if (num_devices < 3)
+ flags &= ~BTRFS_BLOCK_GROUP_RAID6;
if (num_devices < 4)
flags &= ~BTRFS_BLOCK_GROUP_RAID10;
- if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
- (flags & (BTRFS_BLOCK_GROUP_RAID1 |
- BTRFS_BLOCK_GROUP_RAID10))) {
- flags &= ~BTRFS_BLOCK_GROUP_DUP;
- }
+ tmp = flags & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID0 |
+ BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID5 |
+ BTRFS_BLOCK_GROUP_RAID6 | BTRFS_BLOCK_GROUP_RAID10);
+ flags &= ~tmp;
- if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
- (flags & BTRFS_BLOCK_GROUP_RAID10)) {
- flags &= ~BTRFS_BLOCK_GROUP_RAID1;
- }
+ if (tmp & BTRFS_BLOCK_GROUP_RAID6)
+ tmp = BTRFS_BLOCK_GROUP_RAID6;
+ else if (tmp & BTRFS_BLOCK_GROUP_RAID5)
+ tmp = BTRFS_BLOCK_GROUP_RAID5;
+ else if (tmp & BTRFS_BLOCK_GROUP_RAID10)
+ tmp = BTRFS_BLOCK_GROUP_RAID10;
+ else if (tmp & BTRFS_BLOCK_GROUP_RAID1)
+ tmp = BTRFS_BLOCK_GROUP_RAID1;
+ else if (tmp & BTRFS_BLOCK_GROUP_RAID0)
+ tmp = BTRFS_BLOCK_GROUP_RAID0;
- if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
- ((flags & BTRFS_BLOCK_GROUP_RAID1) |
- (flags & BTRFS_BLOCK_GROUP_RAID10) |
- (flags & BTRFS_BLOCK_GROUP_DUP))) {
- flags &= ~BTRFS_BLOCK_GROUP_RAID0;
- }
-
- return extended_to_chunk(flags);
+ return extended_to_chunk(flags | tmp);
}
static u64 get_alloc_profile(struct btrfs_root *root, u64 flags)
@@ -3347,6 +3405,7 @@
u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data)
{
u64 flags;
+ u64 ret;
if (data)
flags = BTRFS_BLOCK_GROUP_DATA;
@@ -3355,7 +3414,8 @@
else
flags = BTRFS_BLOCK_GROUP_METADATA;
- return get_alloc_profile(root, flags);
+ ret = get_alloc_profile(root, flags);
+ return ret;
}
/*
@@ -3530,8 +3590,10 @@
{
u64 num_dev;
- if (type & BTRFS_BLOCK_GROUP_RAID10 ||
- type & BTRFS_BLOCK_GROUP_RAID0)
+ if (type & (BTRFS_BLOCK_GROUP_RAID10 |
+ BTRFS_BLOCK_GROUP_RAID0 |
+ BTRFS_BLOCK_GROUP_RAID5 |
+ BTRFS_BLOCK_GROUP_RAID6))
num_dev = root->fs_info->fs_devices->rw_devices;
else if (type & BTRFS_BLOCK_GROUP_RAID1)
num_dev = 2;
@@ -3706,7 +3768,9 @@
/*
* If we have dup, raid1 or raid10 then only half of the free
- * space is actually useable.
+ * space is actually useable. For raid56, the space info used
+ * doesn't include the parity drive, so we don't have to
+ * change the math
*/
if (profile & (BTRFS_BLOCK_GROUP_DUP |
BTRFS_BLOCK_GROUP_RAID1 |
@@ -5539,10 +5603,14 @@
return ret;
}
-static u64 stripe_align(struct btrfs_root *root, u64 val)
+static u64 stripe_align(struct btrfs_root *root,
+ struct btrfs_block_group_cache *cache,
+ u64 val, u64 num_bytes)
{
- u64 mask = ((u64)root->stripesize - 1);
- u64 ret = (val + mask) & ~mask;
+ u64 mask;
+ u64 ret;
+ mask = ((u64)root->stripesize - 1);
+ ret = (val + mask) & ~mask;
return ret;
}
@@ -5599,8 +5667,12 @@
return BTRFS_RAID_DUP;
else if (flags & BTRFS_BLOCK_GROUP_RAID0)
return BTRFS_RAID_RAID0;
- else
- return BTRFS_RAID_SINGLE;
+ else if (flags & BTRFS_BLOCK_GROUP_RAID5)
+ return BTRFS_RAID_RAID5;
+ else if (flags & BTRFS_BLOCK_GROUP_RAID6)
+ return BTRFS_RAID_RAID6;
+
+ return BTRFS_RAID_SINGLE; /* BTRFS_BLOCK_GROUP_SINGLE */
}
static int get_block_group_index(struct btrfs_block_group_cache *cache)
@@ -5743,6 +5815,8 @@
if (!block_group_bits(block_group, data)) {
u64 extra = BTRFS_BLOCK_GROUP_DUP |
BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_RAID5 |
+ BTRFS_BLOCK_GROUP_RAID6 |
BTRFS_BLOCK_GROUP_RAID10;
/*
@@ -5771,6 +5845,7 @@
* lets look there
*/
if (last_ptr) {
+ unsigned long aligned_cluster;
/*
* the refill lock keeps out other
* people trying to start a new cluster
@@ -5837,11 +5912,15 @@
goto unclustered_alloc;
}
+ aligned_cluster = max_t(unsigned long,
+ empty_cluster + empty_size,
+ block_group->full_stripe_len);
+
/* allocate a cluster in this block group */
ret = btrfs_find_space_cluster(trans, root,
block_group, last_ptr,
search_start, num_bytes,
- empty_cluster + empty_size);
+ aligned_cluster);
if (ret == 0) {
/*
* now pull our allocation out of this
@@ -5912,7 +5991,8 @@
goto loop;
}
checks:
- search_start = stripe_align(root, offset);
+ search_start = stripe_align(root, used_block_group,
+ offset, num_bytes);
/* move on to the next group */
if (search_start + num_bytes >
@@ -7284,6 +7364,7 @@
root->fs_info->fs_devices->missing_devices;
stripped = BTRFS_BLOCK_GROUP_RAID0 |
+ BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6 |
BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
if (num_devices == 1) {
@@ -7837,7 +7918,9 @@
btrfs_release_path(path);
cache->flags = btrfs_block_group_flags(&cache->item);
cache->sectorsize = root->sectorsize;
-
+ cache->full_stripe_len = btrfs_full_stripe_len(root,
+ &root->fs_info->mapping_tree,
+ found_key.objectid);
btrfs_init_free_space_ctl(cache);
/*
@@ -7891,6 +7974,8 @@
if (!(get_alloc_profile(root, space_info->flags) &
(BTRFS_BLOCK_GROUP_RAID10 |
BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_RAID5 |
+ BTRFS_BLOCK_GROUP_RAID6 |
BTRFS_BLOCK_GROUP_DUP)))
continue;
/*
@@ -7966,6 +8051,9 @@
cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
cache->sectorsize = root->sectorsize;
cache->fs_info = root->fs_info;
+ cache->full_stripe_len = btrfs_full_stripe_len(root,
+ &root->fs_info->mapping_tree,
+ chunk_offset);
atomic_set(&cache->count, 1);
spin_lock_init(&cache->lock);