]> git.kernelconcepts.de Git - karo-tx-linux.git/blobdiff - fs/btrfs/inode.c
Btrfs: look for acls during btrfs_read_locked_inode
[karo-tx-linux.git] / fs / btrfs / inode.c
index a0d1dd492a58a57bdf23464340fd6312c192fca9..90c23eb28829932a5577c09ba728ad4156a9a936 100644 (file)
@@ -70,7 +70,6 @@ static struct extent_io_ops btrfs_extent_io_ops;
 static struct kmem_cache *btrfs_inode_cachep;
 struct kmem_cache *btrfs_trans_handle_cachep;
 struct kmem_cache *btrfs_transaction_cachep;
-struct kmem_cache *btrfs_bit_radix_cachep;
 struct kmem_cache *btrfs_path_cachep;
 
 #define S_SHIFT 12
@@ -234,7 +233,7 @@ static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
        }
 
        ret = btrfs_drop_extents(trans, root, inode, start,
-                                aligned_end, start, &hint_byte);
+                                aligned_end, aligned_end, start, &hint_byte);
        BUG_ON(ret);
 
        if (isize > actual_end)
@@ -1439,6 +1438,7 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
                                       struct inode *inode, u64 file_pos,
                                       u64 disk_bytenr, u64 disk_num_bytes,
                                       u64 num_bytes, u64 ram_bytes,
+                                      u64 locked_end,
                                       u8 compression, u8 encryption,
                                       u16 other_encoding, int extent_type)
 {
@@ -1455,7 +1455,8 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
 
        path->leave_spinning = 1;
        ret = btrfs_drop_extents(trans, root, inode, file_pos,
-                                file_pos + num_bytes, file_pos, &hint);
+                                file_pos + num_bytes, locked_end,
+                                file_pos, &hint);
        BUG_ON(ret);
 
        ins.objectid = inode->i_ino;
@@ -1590,6 +1591,8 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
                                                ordered_extent->disk_len,
                                                ordered_extent->len,
                                                ordered_extent->len,
+                                               ordered_extent->file_offset +
+                                               ordered_extent->len,
                                                compressed, 0, 0,
                                                BTRFS_FILE_EXTENT_REG);
                BUG_ON(ret);
@@ -1819,10 +1822,12 @@ good:
        return 0;
 
 zeroit:
-       printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u "
-              "private %llu\n", page->mapping->host->i_ino,
-              (unsigned long long)start, csum,
-              (unsigned long long)private);
+       if (printk_ratelimit()) {
+               printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u "
+                      "private %llu\n", page->mapping->host->i_ino,
+                      (unsigned long long)start, csum,
+                      (unsigned long long)private);
+       }
        memset(kaddr + offset, 1, end - start + 1);
        flush_dcache_page(page);
        kunmap_atomic(kaddr, KM_USER0);
@@ -2010,6 +2015,57 @@ void btrfs_orphan_cleanup(struct btrfs_root *root)
        btrfs_free_path(path);
 }
 
+/*
+ * very simple check to peek ahead in the leaf looking for xattrs.  If we
+ * don't find any xattrs, we know there can't be any acls.
+ *
+ * slot is the slot the inode is in, objectid is the objectid of the inode
+ */
+static noinline int acls_after_inode_item(struct extent_buffer *leaf,
+                                         int slot, u64 objectid)
+{
+       u32 nritems = btrfs_header_nritems(leaf);
+       struct btrfs_key found_key;
+       int scanned = 0;
+
+       slot++;
+       while (slot < nritems) {
+               btrfs_item_key_to_cpu(leaf, &found_key, slot);
+
+               /* we found a different objectid, there must not be acls */
+               if (found_key.objectid != objectid)
+                       return 0;
+
+               /* we found an xattr, assume we've got an acl */
+               if (found_key.type == BTRFS_XATTR_ITEM_KEY)
+                       return 1;
+
+               /*
+                * we found a key greater than an xattr key, there can't
+                * be any acls later on
+                */
+               if (found_key.type > BTRFS_XATTR_ITEM_KEY)
+                       return 0;
+
+               slot++;
+               scanned++;
+
+               /*
+                * it goes inode, inode backrefs, xattrs, extents,
+                * so if there are a ton of hard links to an inode there can
+                * be a lot of backrefs.  Don't waste time searching too hard,
+                * this is just an optimization
+                */
+               if (scanned >= 8)
+                       break;
+       }
+       /* we hit the end of the leaf before we found an xattr or
+        * something larger than an xattr.  We have to assume the inode
+        * has acls
+        */
+       return 1;
+}
+
 /*
  * read an inode from the btree into the in-memory inode
  */
@@ -2021,6 +2077,7 @@ void btrfs_read_locked_inode(struct inode *inode)
        struct btrfs_timespec *tspec;
        struct btrfs_root *root = BTRFS_I(inode)->root;
        struct btrfs_key location;
+       int maybe_acls;
        u64 alloc_group_block;
        u32 rdev;
        int ret;
@@ -2067,6 +2124,16 @@ void btrfs_read_locked_inode(struct inode *inode)
 
        alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
 
+       /*
+        * try to precache a NULL acl entry for files that don't have
+        * any xattrs or acls
+        */
+       maybe_acls = acls_after_inode_item(leaf, path->slots[0], inode->i_ino);
+       if (!maybe_acls) {
+               BTRFS_I(inode)->i_acl = NULL;
+               BTRFS_I(inode)->i_default_acl = NULL;
+       }
+
        BTRFS_I(inode)->block_group = btrfs_find_block_group(root, 0,
                                                alloc_group_block, 0);
        btrfs_free_path(path);
@@ -2877,6 +2944,7 @@ int btrfs_cont_expand(struct inode *inode, loff_t size)
                        err = btrfs_drop_extents(trans, root, inode,
                                                 cur_offset,
                                                 cur_offset + hole_size,
+                                                block_end,
                                                 cur_offset, &hint_byte);
                        if (err)
                                break;
@@ -3041,8 +3109,8 @@ static noinline void init_btrfs_i(struct inode *inode)
 {
        struct btrfs_inode *bi = BTRFS_I(inode);
 
-       bi->i_acl = NULL;
-       bi->i_default_acl = NULL;
+       bi->i_acl = BTRFS_ACL_NOT_CACHED;
+       bi->i_default_acl = BTRFS_ACL_NOT_CACHED;
 
        bi->generation = 0;
        bi->sequence = 0;
@@ -4634,47 +4702,36 @@ void btrfs_destroy_cachep(void)
                kmem_cache_destroy(btrfs_trans_handle_cachep);
        if (btrfs_transaction_cachep)
                kmem_cache_destroy(btrfs_transaction_cachep);
-       if (btrfs_bit_radix_cachep)
-               kmem_cache_destroy(btrfs_bit_radix_cachep);
        if (btrfs_path_cachep)
                kmem_cache_destroy(btrfs_path_cachep);
 }
 
-struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
-                                      unsigned long extra_flags,
-                                      void (*ctor)(void *))
-{
-       return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
-                                SLAB_MEM_SPREAD | extra_flags), ctor);
-}
-
 int btrfs_init_cachep(void)
 {
-       btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
-                                         sizeof(struct btrfs_inode),
-                                         0, init_once);
+       btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
+                       sizeof(struct btrfs_inode), 0,
+                       SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, init_once);
        if (!btrfs_inode_cachep)
                goto fail;
-       btrfs_trans_handle_cachep =
-                       btrfs_cache_create("btrfs_trans_handle_cache",
-                                          sizeof(struct btrfs_trans_handle),
-                                          0, NULL);
+
+       btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
+                       sizeof(struct btrfs_trans_handle), 0,
+                       SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
        if (!btrfs_trans_handle_cachep)
                goto fail;
-       btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
-                                            sizeof(struct btrfs_transaction),
-                                            0, NULL);
+
+       btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
+                       sizeof(struct btrfs_transaction), 0,
+                       SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
        if (!btrfs_transaction_cachep)
                goto fail;
-       btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
-                                        sizeof(struct btrfs_path),
-                                        0, NULL);
+
+       btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
+                       sizeof(struct btrfs_path), 0,
+                       SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
        if (!btrfs_path_cachep)
                goto fail;
-       btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
-                                             SLAB_DESTROY_BY_RCU, NULL);
-       if (!btrfs_bit_radix_cachep)
-               goto fail;
+
        return 0;
 fail:
        btrfs_destroy_cachep();
@@ -4970,10 +5027,10 @@ out_fail:
        return err;
 }
 
-static int prealloc_file_range(struct inode *inode, u64 start, u64 end,
-                              u64 alloc_hint, int mode)
+static int prealloc_file_range(struct btrfs_trans_handle *trans,
+                              struct inode *inode, u64 start, u64 end,
+                              u64 locked_end, u64 alloc_hint, int mode)
 {
-       struct btrfs_trans_handle *trans;
        struct btrfs_root *root = BTRFS_I(inode)->root;
        struct btrfs_key ins;
        u64 alloc_size;
@@ -4981,10 +5038,6 @@ static int prealloc_file_range(struct inode *inode, u64 start, u64 end,
        u64 num_bytes = end - start;
        int ret = 0;
 
-       trans = btrfs_join_transaction(root, 1);
-       BUG_ON(!trans);
-       btrfs_set_trans_block_group(trans, inode);
-
        while (num_bytes > 0) {
                alloc_size = min(num_bytes, root->fs_info->max_extent);
                ret = btrfs_reserve_extent(trans, root, alloc_size,
@@ -4997,7 +5050,8 @@ static int prealloc_file_range(struct inode *inode, u64 start, u64 end,
                ret = insert_reserved_file_extent(trans, inode,
                                                  cur_offset, ins.objectid,
                                                  ins.offset, ins.offset,
-                                                 ins.offset, 0, 0, 0,
+                                                 ins.offset, locked_end,
+                                                 0, 0, 0,
                                                  BTRFS_FILE_EXTENT_PREALLOC);
                BUG_ON(ret);
                num_bytes -= ins.offset;
@@ -5015,7 +5069,6 @@ out:
                BUG_ON(ret);
        }
 
-       btrfs_end_transaction(trans, root);
        return ret;
 }
 
@@ -5027,13 +5080,21 @@ static long btrfs_fallocate(struct inode *inode, int mode,
        u64 alloc_start;
        u64 alloc_end;
        u64 alloc_hint = 0;
+       u64 locked_end;
        u64 mask = BTRFS_I(inode)->root->sectorsize - 1;
        struct extent_map *em;
+       struct btrfs_trans_handle *trans;
        int ret;
 
        alloc_start = offset & ~mask;
        alloc_end =  (offset + len + mask) & ~mask;
 
+       /*
+        * wait for ordered IO before we have any locks.  We'll loop again
+        * below with the locks held.
+        */
+       btrfs_wait_ordered_range(inode, alloc_start, alloc_end - alloc_start);
+
        mutex_lock(&inode->i_mutex);
        if (alloc_start > inode->i_size) {
                ret = btrfs_cont_expand(inode, alloc_start);
@@ -5041,10 +5102,21 @@ static long btrfs_fallocate(struct inode *inode, int mode,
                        goto out;
        }
 
+       locked_end = alloc_end - 1;
        while (1) {
                struct btrfs_ordered_extent *ordered;
-               lock_extent(&BTRFS_I(inode)->io_tree, alloc_start,
-                           alloc_end - 1, GFP_NOFS);
+
+               trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1);
+               if (!trans) {
+                       ret = -EIO;
+                       goto out;
+               }
+
+               /* the extent lock is ordered inside the running
+                * transaction
+                */
+               lock_extent(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
+                           GFP_NOFS);
                ordered = btrfs_lookup_first_ordered_extent(inode,
                                                            alloc_end - 1);
                if (ordered &&
@@ -5052,7 +5124,13 @@ static long btrfs_fallocate(struct inode *inode, int mode,
                    ordered->file_offset < alloc_end) {
                        btrfs_put_ordered_extent(ordered);
                        unlock_extent(&BTRFS_I(inode)->io_tree,
-                                     alloc_start, alloc_end - 1, GFP_NOFS);
+                                     alloc_start, locked_end, GFP_NOFS);
+                       btrfs_end_transaction(trans, BTRFS_I(inode)->root);
+
+                       /*
+                        * we can't wait on the range with the transaction
+                        * running or with the extent lock held
+                        */
                        btrfs_wait_ordered_range(inode, alloc_start,
                                                 alloc_end - alloc_start);
                } else {
@@ -5070,8 +5148,9 @@ static long btrfs_fallocate(struct inode *inode, int mode,
                last_byte = min(extent_map_end(em), alloc_end);
                last_byte = (last_byte + mask) & ~mask;
                if (em->block_start == EXTENT_MAP_HOLE) {
-                       ret = prealloc_file_range(inode, cur_offset,
-                                       last_byte, alloc_hint, mode);
+                       ret = prealloc_file_range(trans, inode, cur_offset,
+                                       last_byte, locked_end + 1,
+                                       alloc_hint, mode);
                        if (ret < 0) {
                                free_extent_map(em);
                                break;
@@ -5087,8 +5166,10 @@ static long btrfs_fallocate(struct inode *inode, int mode,
                        break;
                }
        }
-       unlock_extent(&BTRFS_I(inode)->io_tree, alloc_start, alloc_end - 1,
+       unlock_extent(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
                      GFP_NOFS);
+
+       btrfs_end_transaction(trans, BTRFS_I(inode)->root);
 out:
        mutex_unlock(&inode->i_mutex);
        return ret;