Merge branch 'xfs-commit-cleanup' into for-next
Conflicts:
fs/xfs/xfs_attr_inactive.c
diff --git a/Documentation/filesystems/xfs.txt b/Documentation/filesystems/xfs.txt
index 5a5a055..8146e9f 100644
--- a/Documentation/filesystems/xfs.txt
+++ b/Documentation/filesystems/xfs.txt
@@ -236,10 +236,10 @@
Name Removed
---- -------
- delaylog/nodelaylog v3.20
- ihashsize v3.20
- irixsgid v3.20
- osyncisdsync/osyncisosync v3.20
+ delaylog/nodelaylog v4.0
+ ihashsize v4.0
+ irixsgid v4.0
+ osyncisdsync/osyncisosync v4.0
sysctls
@@ -346,5 +346,5 @@
Name Removed
---- -------
- fs.xfs.xfsbufd_centisec v3.20
- fs.xfs.age_buffer_centisecs v3.20
+ fs.xfs.xfsbufd_centisec v4.0
+ fs.xfs.age_buffer_centisecs v4.0
diff --git a/fs/dax.c b/fs/dax.c
index 6f65f00..99b5fbc 100644
--- a/fs/dax.c
+++ b/fs/dax.c
@@ -309,14 +309,21 @@
out:
i_mmap_unlock_read(mapping);
- if (bh->b_end_io)
- bh->b_end_io(bh, 1);
-
return error;
}
-static int do_dax_fault(struct vm_area_struct *vma, struct vm_fault *vmf,
- get_block_t get_block)
+/**
+ * __dax_fault - handle a page fault on a DAX file
+ * @vma: The virtual memory area where the fault occurred
+ * @vmf: The description of the fault
+ * @get_block: The filesystem method used to translate file offsets to blocks
+ *
+ * When a page fault occurs, filesystems may call this helper in their
+ * fault handler for DAX files. __dax_fault() assumes the caller has done all
+ * the necessary locking for the page fault to proceed successfully.
+ */
+int __dax_fault(struct vm_area_struct *vma, struct vm_fault *vmf,
+ get_block_t get_block, dax_iodone_t complete_unwritten)
{
struct file *file = vma->vm_file;
struct address_space *mapping = file->f_mapping;
@@ -417,7 +424,19 @@
page_cache_release(page);
}
+ /*
+ * If we successfully insert the new mapping over an unwritten extent,
+ * we need to ensure we convert the unwritten extent. If there is an
+ * error inserting the mapping, the filesystem needs to leave it as
+ * unwritten to prevent exposure of the stale underlying data to
+ * userspace, but we still need to call the completion function so
+ * the private resources on the mapping buffer can be released. We
+ * indicate what the callback should do via the uptodate variable, same
+ * as for normal BH based IO completions.
+ */
error = dax_insert_mapping(inode, &bh, vma, vmf);
+ if (buffer_unwritten(&bh))
+ complete_unwritten(&bh, !error);
out:
if (error == -ENOMEM)
@@ -434,6 +453,7 @@
}
goto out;
}
+EXPORT_SYMBOL(__dax_fault);
/**
* dax_fault - handle a page fault on a DAX file
@@ -445,7 +465,7 @@
* fault handler for DAX files.
*/
int dax_fault(struct vm_area_struct *vma, struct vm_fault *vmf,
- get_block_t get_block)
+ get_block_t get_block, dax_iodone_t complete_unwritten)
{
int result;
struct super_block *sb = file_inode(vma->vm_file)->i_sb;
@@ -454,7 +474,7 @@
sb_start_pagefault(sb);
file_update_time(vma->vm_file);
}
- result = do_dax_fault(vma, vmf, get_block);
+ result = __dax_fault(vma, vmf, get_block, complete_unwritten);
if (vmf->flags & FAULT_FLAG_WRITE)
sb_end_pagefault(sb);
diff --git a/fs/ext2/file.c b/fs/ext2/file.c
index 3a0a6c6..3b57c9f 100644
--- a/fs/ext2/file.c
+++ b/fs/ext2/file.c
@@ -28,12 +28,12 @@
#ifdef CONFIG_FS_DAX
static int ext2_dax_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
- return dax_fault(vma, vmf, ext2_get_block);
+ return dax_fault(vma, vmf, ext2_get_block, NULL);
}
static int ext2_dax_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
{
- return dax_mkwrite(vma, vmf, ext2_get_block);
+ return dax_mkwrite(vma, vmf, ext2_get_block, NULL);
}
static const struct vm_operations_struct ext2_dax_vm_ops = {
diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index 0613c25..f713cfcc 100644
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -192,15 +192,27 @@
}
#ifdef CONFIG_FS_DAX
+static void ext4_end_io_unwritten(struct buffer_head *bh, int uptodate)
+{
+ struct inode *inode = bh->b_assoc_map->host;
+ /* XXX: breaks on 32-bit > 16GB. Is that even supported? */
+ loff_t offset = (loff_t)(uintptr_t)bh->b_private << inode->i_blkbits;
+ int err;
+ if (!uptodate)
+ return;
+ WARN_ON(!buffer_unwritten(bh));
+ err = ext4_convert_unwritten_extents(NULL, inode, offset, bh->b_size);
+}
+
static int ext4_dax_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
- return dax_fault(vma, vmf, ext4_get_block);
+ return dax_fault(vma, vmf, ext4_get_block, ext4_end_io_unwritten);
/* Is this the right get_block? */
}
static int ext4_dax_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
{
- return dax_mkwrite(vma, vmf, ext4_get_block);
+ return dax_mkwrite(vma, vmf, ext4_get_block, ext4_end_io_unwritten);
}
static const struct vm_operations_struct ext4_dax_vm_ops = {
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 55b187c..7c38ed3 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -656,18 +656,6 @@
return retval;
}
-static void ext4_end_io_unwritten(struct buffer_head *bh, int uptodate)
-{
- struct inode *inode = bh->b_assoc_map->host;
- /* XXX: breaks on 32-bit > 16GB. Is that even supported? */
- loff_t offset = (loff_t)(uintptr_t)bh->b_private << inode->i_blkbits;
- int err;
- if (!uptodate)
- return;
- WARN_ON(!buffer_unwritten(bh));
- err = ext4_convert_unwritten_extents(NULL, inode, offset, bh->b_size);
-}
-
/* Maximum number of blocks we map for direct IO at once. */
#define DIO_MAX_BLOCKS 4096
@@ -705,10 +693,15 @@
map_bh(bh, inode->i_sb, map.m_pblk);
bh->b_state = (bh->b_state & ~EXT4_MAP_FLAGS) | map.m_flags;
- if (IS_DAX(inode) && buffer_unwritten(bh) && !io_end) {
+ if (IS_DAX(inode) && buffer_unwritten(bh)) {
+ /*
+ * dgc: I suspect unwritten conversion on ext4+DAX is
+ * fundamentally broken here when there are concurrent
+ * read/write in progress on this inode.
+ */
+ WARN_ON_ONCE(io_end);
bh->b_assoc_map = inode->i_mapping;
bh->b_private = (void *)(unsigned long)iblock;
- bh->b_end_io = ext4_end_io_unwritten;
}
if (io_end && io_end->flag & EXT4_IO_END_UNWRITTEN)
set_buffer_defer_completion(bh);
diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
index 516162b..bc78ac0 100644
--- a/fs/xfs/libxfs/xfs_alloc.c
+++ b/fs/xfs/libxfs/xfs_alloc.c
@@ -149,13 +149,27 @@
{
xfs_agblock_t bno;
xfs_extlen_t len;
+ xfs_extlen_t diff;
/* Trim busy sections out of found extent */
xfs_extent_busy_trim(args, foundbno, foundlen, &bno, &len);
+ /*
+ * If we have a largish extent that happens to start before min_agbno,
+ * see if we can shift it into range...
+ */
+ if (bno < args->min_agbno && bno + len > args->min_agbno) {
+ diff = args->min_agbno - bno;
+ if (len > diff) {
+ bno += diff;
+ len -= diff;
+ }
+ }
+
if (args->alignment > 1 && len >= args->minlen) {
xfs_agblock_t aligned_bno = roundup(bno, args->alignment);
- xfs_extlen_t diff = aligned_bno - bno;
+
+ diff = aligned_bno - bno;
*resbno = aligned_bno;
*reslen = diff >= len ? 0 : len - diff;
@@ -795,9 +809,13 @@
* The good extent is closer than this one.
*/
if (!dir) {
+ if (*sbnoa > args->max_agbno)
+ goto out_use_good;
if (*sbnoa >= args->agbno + gdiff)
goto out_use_good;
} else {
+ if (*sbnoa < args->min_agbno)
+ goto out_use_good;
if (*sbnoa <= args->agbno - gdiff)
goto out_use_good;
}
@@ -884,6 +902,17 @@
dofirst = prandom_u32() & 1;
#endif
+ /* handle unitialized agbno range so caller doesn't have to */
+ if (!args->min_agbno && !args->max_agbno)
+ args->max_agbno = args->mp->m_sb.sb_agblocks - 1;
+ ASSERT(args->min_agbno <= args->max_agbno);
+
+ /* clamp agbno to the range if it's outside */
+ if (args->agbno < args->min_agbno)
+ args->agbno = args->min_agbno;
+ if (args->agbno > args->max_agbno)
+ args->agbno = args->max_agbno;
+
restart:
bno_cur_lt = NULL;
bno_cur_gt = NULL;
@@ -976,6 +1005,8 @@
<bnoa, <lena);
if (ltlena < args->minlen)
continue;
+ if (ltbnoa < args->min_agbno || ltbnoa > args->max_agbno)
+ continue;
args->len = XFS_EXTLEN_MIN(ltlena, args->maxlen);
xfs_alloc_fix_len(args);
ASSERT(args->len >= args->minlen);
@@ -1096,11 +1127,11 @@
XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
xfs_alloc_compute_aligned(args, ltbno, ltlen,
<bnoa, <lena);
- if (ltlena >= args->minlen)
+ if (ltlena >= args->minlen && ltbnoa >= args->min_agbno)
break;
if ((error = xfs_btree_decrement(bno_cur_lt, 0, &i)))
goto error0;
- if (!i) {
+ if (!i || ltbnoa < args->min_agbno) {
xfs_btree_del_cursor(bno_cur_lt,
XFS_BTREE_NOERROR);
bno_cur_lt = NULL;
@@ -1112,11 +1143,11 @@
XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
xfs_alloc_compute_aligned(args, gtbno, gtlen,
>bnoa, >lena);
- if (gtlena >= args->minlen)
+ if (gtlena >= args->minlen && gtbnoa <= args->max_agbno)
break;
if ((error = xfs_btree_increment(bno_cur_gt, 0, &i)))
goto error0;
- if (!i) {
+ if (!i || gtbnoa > args->max_agbno) {
xfs_btree_del_cursor(bno_cur_gt,
XFS_BTREE_NOERROR);
bno_cur_gt = NULL;
@@ -1216,6 +1247,7 @@
ASSERT(ltnew >= ltbno);
ASSERT(ltnew + rlen <= ltbnoa + ltlena);
ASSERT(ltnew + rlen <= be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
+ ASSERT(ltnew >= args->min_agbno && ltnew <= args->max_agbno);
args->agbno = ltnew;
if ((error = xfs_alloc_fixup_trees(cnt_cur, bno_cur_lt, ltbno, ltlen,
diff --git a/fs/xfs/libxfs/xfs_alloc.h b/fs/xfs/libxfs/xfs_alloc.h
index d1b4b6a..29f27b2 100644
--- a/fs/xfs/libxfs/xfs_alloc.h
+++ b/fs/xfs/libxfs/xfs_alloc.h
@@ -112,6 +112,8 @@
xfs_extlen_t total; /* total blocks needed in xaction */
xfs_extlen_t alignment; /* align answer to multiple of this */
xfs_extlen_t minalignslop; /* slop for minlen+alignment calcs */
+ xfs_agblock_t min_agbno; /* set an agbno range for NEAR allocs */
+ xfs_agblock_t max_agbno; /* ... */
xfs_extlen_t len; /* output: actual size of extent */
xfs_alloctype_t type; /* allocation type XFS_ALLOCTYPE_... */
xfs_alloctype_t otype; /* original allocation type */
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
index 04e79d5..e9d401c 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.c
+++ b/fs/xfs/libxfs/xfs_attr_leaf.c
@@ -574,8 +574,8 @@
* After the last attribute is removed revert to original inode format,
* making all literal area available to the data fork once more.
*/
-STATIC void
-xfs_attr_fork_reset(
+void
+xfs_attr_fork_remove(
struct xfs_inode *ip,
struct xfs_trans *tp)
{
@@ -641,7 +641,7 @@
(mp->m_flags & XFS_MOUNT_ATTR2) &&
(dp->i_d.di_format != XFS_DINODE_FMT_BTREE) &&
!(args->op_flags & XFS_DA_OP_ADDNAME)) {
- xfs_attr_fork_reset(dp, args->trans);
+ xfs_attr_fork_remove(dp, args->trans);
} else {
xfs_idata_realloc(dp, -size, XFS_ATTR_FORK);
dp->i_d.di_forkoff = xfs_attr_shortform_bytesfit(dp, totsize);
@@ -905,7 +905,7 @@
if (forkoff == -1) {
ASSERT(dp->i_mount->m_flags & XFS_MOUNT_ATTR2);
ASSERT(dp->i_d.di_format != XFS_DINODE_FMT_BTREE);
- xfs_attr_fork_reset(dp, args->trans);
+ xfs_attr_fork_remove(dp, args->trans);
goto out;
}
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.h b/fs/xfs/libxfs/xfs_attr_leaf.h
index 025c4b8..882c8d3 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.h
+++ b/fs/xfs/libxfs/xfs_attr_leaf.h
@@ -53,7 +53,7 @@
int xfs_attr_shortform_list(struct xfs_attr_list_context *context);
int xfs_attr_shortform_allfit(struct xfs_buf *bp, struct xfs_inode *dp);
int xfs_attr_shortform_bytesfit(xfs_inode_t *dp, int bytes);
-
+void xfs_attr_fork_remove(struct xfs_inode *ip, struct xfs_trans *tp);
/*
* Internal routines when attribute fork size == XFS_LBSIZE(mp).
diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
index caca2c5..a334633 100644
--- a/fs/xfs/libxfs/xfs_bmap.c
+++ b/fs/xfs/libxfs/xfs_bmap.c
@@ -3221,12 +3221,24 @@
align_alen += temp;
align_off -= temp;
}
- /*
- * Same adjustment for the end of the requested area.
- */
- if ((temp = (align_alen % extsz))) {
+
+ /* Same adjustment for the end of the requested area. */
+ temp = (align_alen % extsz);
+ if (temp)
align_alen += extsz - temp;
- }
+
+ /*
+ * For large extent hint sizes, the aligned extent might be larger than
+ * MAXEXTLEN. In that case, reduce the size by an extsz so that it pulls
+ * the length back under MAXEXTLEN. The outer allocation loops handle
+ * short allocation just fine, so it is safe to do this. We only want to
+ * do it when we are forced to, though, because it means more allocation
+ * operations are required.
+ */
+ while (align_alen > MAXEXTLEN)
+ align_alen -= extsz;
+ ASSERT(align_alen <= MAXEXTLEN);
+
/*
* If the previous block overlaps with this proposed allocation
* then move the start forward without adjusting the length.
@@ -3315,7 +3327,9 @@
return -EINVAL;
} else {
ASSERT(orig_off >= align_off);
- ASSERT(orig_end <= align_off + align_alen);
+ /* see MAXEXTLEN handling above */
+ ASSERT(orig_end <= align_off + align_alen ||
+ align_alen + extsz > MAXEXTLEN);
}
#ifdef DEBUG
@@ -4096,13 +4110,6 @@
/* Figure out the extent size, adjust alen */
extsz = xfs_get_extsz_hint(ip);
if (extsz) {
- /*
- * Make sure we don't exceed a single extent length when we
- * align the extent by reducing length we are going to
- * allocate by the maximum amount extent size aligment may
- * require.
- */
- alen = XFS_FILBLKS_MIN(len, MAXEXTLEN - (2 * extsz - 1));
error = xfs_bmap_extsize_align(mp, got, prev, extsz, rt, eof,
1, 0, &aoff, &alen);
ASSERT(!error);
@@ -4414,7 +4421,15 @@
error = xfs_bmap_add_extent_unwritten_real(bma->tp, bma->ip, &bma->idx,
&bma->cur, mval, bma->firstblock, bma->flist,
&tmp_logflags);
- bma->logflags |= tmp_logflags;
+ /*
+ * Log the inode core unconditionally in the unwritten extent conversion
+ * path because the conversion might not have done so (e.g., if the
+ * extent count hasn't changed). We need to make sure the inode is dirty
+ * in the transaction for the sake of fsync(), even if nothing has
+ * changed, because fsync() will not force the log for this transaction
+ * unless it sees the inode pinned.
+ */
+ bma->logflags |= tmp_logflags | XFS_ILOG_CORE;
if (error)
return error;
diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h
index 4daaa66..815f61b 100644
--- a/fs/xfs/libxfs/xfs_format.h
+++ b/fs/xfs/libxfs/xfs_format.h
@@ -170,7 +170,7 @@
__uint32_t sb_features_log_incompat;
__uint32_t sb_crc; /* superblock crc */
- __uint32_t sb_pad;
+ xfs_extlen_t sb_spino_align; /* sparse inode chunk alignment */
xfs_ino_t sb_pquotino; /* project quota inode */
xfs_lsn_t sb_lsn; /* last write sequence */
@@ -256,7 +256,7 @@
__be32 sb_features_log_incompat;
__le32 sb_crc; /* superblock crc */
- __be32 sb_pad;
+ __be32 sb_spino_align; /* sparse inode chunk alignment */
__be64 sb_pquotino; /* project quota inode */
__be64 sb_lsn; /* last write sequence */
@@ -457,8 +457,10 @@
}
#define XFS_SB_FEAT_INCOMPAT_FTYPE (1 << 0) /* filetype in dirent */
+#define XFS_SB_FEAT_INCOMPAT_SPINODES (1 << 1) /* sparse inode chunks */
#define XFS_SB_FEAT_INCOMPAT_ALL \
- (XFS_SB_FEAT_INCOMPAT_FTYPE)
+ (XFS_SB_FEAT_INCOMPAT_FTYPE| \
+ XFS_SB_FEAT_INCOMPAT_SPINODES)
#define XFS_SB_FEAT_INCOMPAT_UNKNOWN ~XFS_SB_FEAT_INCOMPAT_ALL
static inline bool
@@ -506,6 +508,12 @@
(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT);
}
+static inline bool xfs_sb_version_hassparseinodes(struct xfs_sb *sbp)
+{
+ return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
+ xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_SPINODES);
+}
+
/*
* end of superblock version macros
*/
@@ -1216,26 +1224,54 @@
#define XFS_INOBT_ALL_FREE ((xfs_inofree_t)-1)
#define XFS_INOBT_MASK(i) ((xfs_inofree_t)1 << (i))
+#define XFS_INOBT_HOLEMASK_FULL 0 /* holemask for full chunk */
+#define XFS_INOBT_HOLEMASK_BITS (NBBY * sizeof(__uint16_t))
+#define XFS_INODES_PER_HOLEMASK_BIT \
+ (XFS_INODES_PER_CHUNK / (NBBY * sizeof(__uint16_t)))
+
static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
{
return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
}
/*
- * Data record structure
+ * The on-disk inode record structure has two formats. The original "full"
+ * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
+ * and replaces the 3 high-order freecount bytes wth the holemask and inode
+ * count.
+ *
+ * The holemask of the sparse record format allows an inode chunk to have holes
+ * that refer to blocks not owned by the inode record. This facilitates inode
+ * allocation in the event of severe free space fragmentation.
*/
typedef struct xfs_inobt_rec {
__be32 ir_startino; /* starting inode number */
- __be32 ir_freecount; /* count of free inodes (set bits) */
+ union {
+ struct {
+ __be32 ir_freecount; /* count of free inodes */
+ } f;
+ struct {
+ __be16 ir_holemask;/* hole mask for sparse chunks */
+ __u8 ir_count; /* total inode count */
+ __u8 ir_freecount; /* count of free inodes */
+ } sp;
+ } ir_u;
__be64 ir_free; /* free inode mask */
} xfs_inobt_rec_t;
typedef struct xfs_inobt_rec_incore {
xfs_agino_t ir_startino; /* starting inode number */
- __int32_t ir_freecount; /* count of free inodes (set bits) */
+ __uint16_t ir_holemask; /* hole mask for sparse chunks */
+ __uint8_t ir_count; /* total inode count */
+ __uint8_t ir_freecount; /* count of free inodes (set bits) */
xfs_inofree_t ir_free; /* free inode mask */
} xfs_inobt_rec_incore_t;
+static inline bool xfs_inobt_issparse(uint16_t holemask)
+{
+ /* non-zero holemask represents a sparse rec. */
+ return holemask;
+}
/*
* Key structure
@@ -1453,8 +1489,8 @@
sizeof(struct xfs_acl_entry) * XFS_ACL_MAX_ENTRIES((mp)))
/* On-disk XFS extended attribute names */
-#define SGI_ACL_FILE (unsigned char *)"SGI_ACL_FILE"
-#define SGI_ACL_DEFAULT (unsigned char *)"SGI_ACL_DEFAULT"
+#define SGI_ACL_FILE "SGI_ACL_FILE"
+#define SGI_ACL_DEFAULT "SGI_ACL_DEFAULT"
#define SGI_ACL_FILE_SIZE (sizeof(SGI_ACL_FILE)-1)
#define SGI_ACL_DEFAULT_SIZE (sizeof(SGI_ACL_DEFAULT)-1)
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 18dc721..89689c6 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -239,6 +239,7 @@
#define XFS_FSOP_GEOM_FLAGS_V5SB 0x8000 /* version 5 superblock */
#define XFS_FSOP_GEOM_FLAGS_FTYPE 0x10000 /* inode directory types */
#define XFS_FSOP_GEOM_FLAGS_FINOBT 0x20000 /* free inode btree */
+#define XFS_FSOP_GEOM_FLAGS_SPINODES 0x40000 /* sparse inode chunks */
/*
* Minimum and maximum sizes need for growth checks.
diff --git a/fs/xfs/libxfs/xfs_ialloc.c b/fs/xfs/libxfs/xfs_ialloc.c
index 07349a1..66efc70 100644
--- a/fs/xfs/libxfs/xfs_ialloc.c
+++ b/fs/xfs/libxfs/xfs_ialloc.c
@@ -65,6 +65,8 @@
int *stat) /* success/failure */
{
cur->bc_rec.i.ir_startino = ino;
+ cur->bc_rec.i.ir_holemask = 0;
+ cur->bc_rec.i.ir_count = 0;
cur->bc_rec.i.ir_freecount = 0;
cur->bc_rec.i.ir_free = 0;
return xfs_btree_lookup(cur, dir, stat);
@@ -82,7 +84,14 @@
union xfs_btree_rec rec;
rec.inobt.ir_startino = cpu_to_be32(irec->ir_startino);
- rec.inobt.ir_freecount = cpu_to_be32(irec->ir_freecount);
+ if (xfs_sb_version_hassparseinodes(&cur->bc_mp->m_sb)) {
+ rec.inobt.ir_u.sp.ir_holemask = cpu_to_be16(irec->ir_holemask);
+ rec.inobt.ir_u.sp.ir_count = irec->ir_count;
+ rec.inobt.ir_u.sp.ir_freecount = irec->ir_freecount;
+ } else {
+ /* ir_holemask/ir_count not supported on-disk */
+ rec.inobt.ir_u.f.ir_freecount = cpu_to_be32(irec->ir_freecount);
+ }
rec.inobt.ir_free = cpu_to_be64(irec->ir_free);
return xfs_btree_update(cur, &rec);
}
@@ -100,12 +109,27 @@
int error;
error = xfs_btree_get_rec(cur, &rec, stat);
- if (!error && *stat == 1) {
- irec->ir_startino = be32_to_cpu(rec->inobt.ir_startino);
- irec->ir_freecount = be32_to_cpu(rec->inobt.ir_freecount);
- irec->ir_free = be64_to_cpu(rec->inobt.ir_free);
+ if (error || *stat == 0)
+ return error;
+
+ irec->ir_startino = be32_to_cpu(rec->inobt.ir_startino);
+ if (xfs_sb_version_hassparseinodes(&cur->bc_mp->m_sb)) {
+ irec->ir_holemask = be16_to_cpu(rec->inobt.ir_u.sp.ir_holemask);
+ irec->ir_count = rec->inobt.ir_u.sp.ir_count;
+ irec->ir_freecount = rec->inobt.ir_u.sp.ir_freecount;
+ } else {
+ /*
+ * ir_holemask/ir_count not supported on-disk. Fill in hardcoded
+ * values for full inode chunks.
+ */
+ irec->ir_holemask = XFS_INOBT_HOLEMASK_FULL;
+ irec->ir_count = XFS_INODES_PER_CHUNK;
+ irec->ir_freecount =
+ be32_to_cpu(rec->inobt.ir_u.f.ir_freecount);
}
- return error;
+ irec->ir_free = be64_to_cpu(rec->inobt.ir_free);
+
+ return 0;
}
/*
@@ -114,10 +138,14 @@
STATIC int
xfs_inobt_insert_rec(
struct xfs_btree_cur *cur,
+ __uint16_t holemask,
+ __uint8_t count,
__int32_t freecount,
xfs_inofree_t free,
int *stat)
{
+ cur->bc_rec.i.ir_holemask = holemask;
+ cur->bc_rec.i.ir_count = count;
cur->bc_rec.i.ir_freecount = freecount;
cur->bc_rec.i.ir_free = free;
return xfs_btree_insert(cur, stat);
@@ -154,7 +182,9 @@
}
ASSERT(i == 0);
- error = xfs_inobt_insert_rec(cur, XFS_INODES_PER_CHUNK,
+ error = xfs_inobt_insert_rec(cur, XFS_INOBT_HOLEMASK_FULL,
+ XFS_INODES_PER_CHUNK,
+ XFS_INODES_PER_CHUNK,
XFS_INOBT_ALL_FREE, &i);
if (error) {
xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
@@ -220,6 +250,7 @@
struct xfs_mount *mp,
struct xfs_trans *tp,
struct list_head *buffer_list,
+ int icount,
xfs_agnumber_t agno,
xfs_agblock_t agbno,
xfs_agblock_t length,
@@ -275,7 +306,7 @@
* they track in the AIL as if they were physically logged.
*/
if (tp)
- xfs_icreate_log(tp, agno, agbno, mp->m_ialloc_inos,
+ xfs_icreate_log(tp, agno, agbno, icount,
mp->m_sb.sb_inodesize, length, gen);
} else
version = 2;
@@ -347,6 +378,214 @@
}
/*
+ * Align startino and allocmask for a recently allocated sparse chunk such that
+ * they are fit for insertion (or merge) into the on-disk inode btrees.
+ *
+ * Background:
+ *
+ * When enabled, sparse inode support increases the inode alignment from cluster
+ * size to inode chunk size. This means that the minimum range between two
+ * non-adjacent inode records in the inobt is large enough for a full inode
+ * record. This allows for cluster sized, cluster aligned block allocation
+ * without need to worry about whether the resulting inode record overlaps with
+ * another record in the tree. Without this basic rule, we would have to deal
+ * with the consequences of overlap by potentially undoing recent allocations in
+ * the inode allocation codepath.
+ *
+ * Because of this alignment rule (which is enforced on mount), there are two
+ * inobt possibilities for newly allocated sparse chunks. One is that the
+ * aligned inode record for the chunk covers a range of inodes not already
+ * covered in the inobt (i.e., it is safe to insert a new sparse record). The
+ * other is that a record already exists at the aligned startino that considers
+ * the newly allocated range as sparse. In the latter case, record content is
+ * merged in hope that sparse inode chunks fill to full chunks over time.
+ */
+STATIC void
+xfs_align_sparse_ino(
+ struct xfs_mount *mp,
+ xfs_agino_t *startino,
+ uint16_t *allocmask)
+{
+ xfs_agblock_t agbno;
+ xfs_agblock_t mod;
+ int offset;
+
+ agbno = XFS_AGINO_TO_AGBNO(mp, *startino);
+ mod = agbno % mp->m_sb.sb_inoalignmt;
+ if (!mod)
+ return;
+
+ /* calculate the inode offset and align startino */
+ offset = mod << mp->m_sb.sb_inopblog;
+ *startino -= offset;
+
+ /*
+ * Since startino has been aligned down, left shift allocmask such that
+ * it continues to represent the same physical inodes relative to the
+ * new startino.
+ */
+ *allocmask <<= offset / XFS_INODES_PER_HOLEMASK_BIT;
+}
+
+/*
+ * Determine whether the source inode record can merge into the target. Both
+ * records must be sparse, the inode ranges must match and there must be no
+ * allocation overlap between the records.
+ */
+STATIC bool
+__xfs_inobt_can_merge(
+ struct xfs_inobt_rec_incore *trec, /* tgt record */
+ struct xfs_inobt_rec_incore *srec) /* src record */
+{
+ uint64_t talloc;
+ uint64_t salloc;
+
+ /* records must cover the same inode range */
+ if (trec->ir_startino != srec->ir_startino)
+ return false;
+
+ /* both records must be sparse */
+ if (!xfs_inobt_issparse(trec->ir_holemask) ||
+ !xfs_inobt_issparse(srec->ir_holemask))
+ return false;
+
+ /* both records must track some inodes */
+ if (!trec->ir_count || !srec->ir_count)
+ return false;
+
+ /* can't exceed capacity of a full record */
+ if (trec->ir_count + srec->ir_count > XFS_INODES_PER_CHUNK)
+ return false;
+
+ /* verify there is no allocation overlap */
+ talloc = xfs_inobt_irec_to_allocmask(trec);
+ salloc = xfs_inobt_irec_to_allocmask(srec);
+ if (talloc & salloc)
+ return false;
+
+ return true;
+}
+
+/*
+ * Merge the source inode record into the target. The caller must call
+ * __xfs_inobt_can_merge() to ensure the merge is valid.
+ */
+STATIC void
+__xfs_inobt_rec_merge(
+ struct xfs_inobt_rec_incore *trec, /* target */
+ struct xfs_inobt_rec_incore *srec) /* src */
+{
+ ASSERT(trec->ir_startino == srec->ir_startino);
+
+ /* combine the counts */
+ trec->ir_count += srec->ir_count;
+ trec->ir_freecount += srec->ir_freecount;
+
+ /*
+ * Merge the holemask and free mask. For both fields, 0 bits refer to
+ * allocated inodes. We combine the allocated ranges with bitwise AND.
+ */
+ trec->ir_holemask &= srec->ir_holemask;
+ trec->ir_free &= srec->ir_free;
+}
+
+/*
+ * Insert a new sparse inode chunk into the associated inode btree. The inode
+ * record for the sparse chunk is pre-aligned to a startino that should match
+ * any pre-existing sparse inode record in the tree. This allows sparse chunks
+ * to fill over time.
+ *
+ * This function supports two modes of handling preexisting records depending on
+ * the merge flag. If merge is true, the provided record is merged with the
+ * existing record and updated in place. The merged record is returned in nrec.
+ * If merge is false, an existing record is replaced with the provided record.
+ * If no preexisting record exists, the provided record is always inserted.
+ *
+ * It is considered corruption if a merge is requested and not possible. Given
+ * the sparse inode alignment constraints, this should never happen.
+ */
+STATIC int
+xfs_inobt_insert_sprec(
+ struct xfs_mount *mp,
+ struct xfs_trans *tp,
+ struct xfs_buf *agbp,
+ int btnum,
+ struct xfs_inobt_rec_incore *nrec, /* in/out: new/merged rec. */
+ bool merge) /* merge or replace */
+{
+ struct xfs_btree_cur *cur;
+ struct xfs_agi *agi = XFS_BUF_TO_AGI(agbp);
+ xfs_agnumber_t agno = be32_to_cpu(agi->agi_seqno);
+ int error;
+ int i;
+ struct xfs_inobt_rec_incore rec;
+
+ cur = xfs_inobt_init_cursor(mp, tp, agbp, agno, btnum);
+
+ /* the new record is pre-aligned so we know where to look */
+ error = xfs_inobt_lookup(cur, nrec->ir_startino, XFS_LOOKUP_EQ, &i);
+ if (error)
+ goto error;
+ /* if nothing there, insert a new record and return */
+ if (i == 0) {
+ error = xfs_inobt_insert_rec(cur, nrec->ir_holemask,
+ nrec->ir_count, nrec->ir_freecount,
+ nrec->ir_free, &i);
+ if (error)
+ goto error;
+ XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error);
+
+ goto out;
+ }
+
+ /*
+ * A record exists at this startino. Merge or replace the record
+ * depending on what we've been asked to do.
+ */
+ if (merge) {
+ error = xfs_inobt_get_rec(cur, &rec, &i);
+ if (error)
+ goto error;
+ XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error);
+ XFS_WANT_CORRUPTED_GOTO(mp,
+ rec.ir_startino == nrec->ir_startino,
+ error);
+
+ /*
+ * This should never fail. If we have coexisting records that
+ * cannot merge, something is seriously wrong.
+ */
+ XFS_WANT_CORRUPTED_GOTO(mp, __xfs_inobt_can_merge(nrec, &rec),
+ error);
+
+ trace_xfs_irec_merge_pre(mp, agno, rec.ir_startino,
+ rec.ir_holemask, nrec->ir_startino,
+ nrec->ir_holemask);
+
+ /* merge to nrec to output the updated record */
+ __xfs_inobt_rec_merge(nrec, &rec);
+
+ trace_xfs_irec_merge_post(mp, agno, nrec->ir_startino,
+ nrec->ir_holemask);
+
+ error = xfs_inobt_rec_check_count(mp, nrec);
+ if (error)
+ goto error;
+ }
+
+ error = xfs_inobt_update(cur, nrec);
+ if (error)
+ goto error;
+
+out:
+ xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+ return 0;
+error:
+ xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
+ return error;
+}
+
+/*
* Allocate new inodes in the allocation group specified by agbp.
* Return 0 for success, else error code.
*/
@@ -364,11 +603,22 @@
xfs_agino_t newlen; /* new number of inodes */
int isaligned = 0; /* inode allocation at stripe unit */
/* boundary */
+ uint16_t allocmask = (uint16_t) -1; /* init. to full chunk */
+ struct xfs_inobt_rec_incore rec;
struct xfs_perag *pag;
+ int do_sparse = 0;
memset(&args, 0, sizeof(args));
args.tp = tp;
args.mp = tp->t_mountp;
+ args.fsbno = NULLFSBLOCK;
+
+#ifdef DEBUG
+ /* randomly do sparse inode allocations */
+ if (xfs_sb_version_hassparseinodes(&tp->t_mountp->m_sb) &&
+ args.mp->m_ialloc_min_blks < args.mp->m_ialloc_blks)
+ do_sparse = prandom_u32() & 1;
+#endif
/*
* Locking will ensure that we don't have two callers in here
@@ -376,7 +626,7 @@
*/
newlen = args.mp->m_ialloc_inos;
if (args.mp->m_maxicount &&
- percpu_counter_read(&args.mp->m_icount) + newlen >
+ percpu_counter_read_positive(&args.mp->m_icount) + newlen >
args.mp->m_maxicount)
return -ENOSPC;
args.minlen = args.maxlen = args.mp->m_ialloc_blks;
@@ -390,6 +640,8 @@
agno = be32_to_cpu(agi->agi_seqno);
args.agbno = XFS_AGINO_TO_AGBNO(args.mp, newino) +
args.mp->m_ialloc_blks;
+ if (do_sparse)
+ goto sparse_alloc;
if (likely(newino != NULLAGINO &&
(args.agbno < be32_to_cpu(agi->agi_length)))) {
args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno);
@@ -428,8 +680,7 @@
* subsequent requests.
*/
args.minalignslop = 0;
- } else
- args.fsbno = NULLFSBLOCK;
+ }
if (unlikely(args.fsbno == NULLFSBLOCK)) {
/*
@@ -480,6 +731,47 @@
return error;
}
+ /*
+ * Finally, try a sparse allocation if the filesystem supports it and
+ * the sparse allocation length is smaller than a full chunk.
+ */
+ if (xfs_sb_version_hassparseinodes(&args.mp->m_sb) &&
+ args.mp->m_ialloc_min_blks < args.mp->m_ialloc_blks &&
+ args.fsbno == NULLFSBLOCK) {
+sparse_alloc:
+ args.type = XFS_ALLOCTYPE_NEAR_BNO;
+ args.agbno = be32_to_cpu(agi->agi_root);
+ args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno);
+ args.alignment = args.mp->m_sb.sb_spino_align;
+ args.prod = 1;
+
+ args.minlen = args.mp->m_ialloc_min_blks;
+ args.maxlen = args.minlen;
+
+ /*
+ * The inode record will be aligned to full chunk size. We must
+ * prevent sparse allocation from AG boundaries that result in
+ * invalid inode records, such as records that start at agbno 0
+ * or extend beyond the AG.
+ *
+ * Set min agbno to the first aligned, non-zero agbno and max to
+ * the last aligned agbno that is at least one full chunk from
+ * the end of the AG.
+ */
+ args.min_agbno = args.mp->m_sb.sb_inoalignmt;
+ args.max_agbno = round_down(args.mp->m_sb.sb_agblocks,
+ args.mp->m_sb.sb_inoalignmt) -
+ args.mp->m_ialloc_blks;
+
+ error = xfs_alloc_vextent(&args);
+ if (error)
+ return error;
+
+ newlen = args.len << args.mp->m_sb.sb_inopblog;
+ ASSERT(newlen <= XFS_INODES_PER_CHUNK);
+ allocmask = (1 << (newlen / XFS_INODES_PER_HOLEMASK_BIT)) - 1;
+ }
+
if (args.fsbno == NULLFSBLOCK) {
*alloc = 0;
return 0;
@@ -495,8 +787,8 @@
* rather than a linear progression to prevent the next generation
* number from being easily guessable.
*/
- error = xfs_ialloc_inode_init(args.mp, tp, NULL, agno, args.agbno,
- args.len, prandom_u32());
+ error = xfs_ialloc_inode_init(args.mp, tp, NULL, newlen, agno,
+ args.agbno, args.len, prandom_u32());
if (error)
return error;
@@ -504,6 +796,73 @@
* Convert the results.
*/
newino = XFS_OFFBNO_TO_AGINO(args.mp, args.agbno, 0);
+
+ if (xfs_inobt_issparse(~allocmask)) {
+ /*
+ * We've allocated a sparse chunk. Align the startino and mask.
+ */
+ xfs_align_sparse_ino(args.mp, &newino, &allocmask);
+
+ rec.ir_startino = newino;
+ rec.ir_holemask = ~allocmask;
+ rec.ir_count = newlen;
+ rec.ir_freecount = newlen;
+ rec.ir_free = XFS_INOBT_ALL_FREE;
+
+ /*
+ * Insert the sparse record into the inobt and allow for a merge
+ * if necessary. If a merge does occur, rec is updated to the
+ * merged record.
+ */
+ error = xfs_inobt_insert_sprec(args.mp, tp, agbp, XFS_BTNUM_INO,
+ &rec, true);
+ if (error == -EFSCORRUPTED) {
+ xfs_alert(args.mp,
+ "invalid sparse inode record: ino 0x%llx holemask 0x%x count %u",
+ XFS_AGINO_TO_INO(args.mp, agno,
+ rec.ir_startino),
+ rec.ir_holemask, rec.ir_count);
+ xfs_force_shutdown(args.mp, SHUTDOWN_CORRUPT_INCORE);
+ }
+ if (error)
+ return error;
+
+ /*
+ * We can't merge the part we've just allocated as for the inobt
+ * due to finobt semantics. The original record may or may not
+ * exist independent of whether physical inodes exist in this
+ * sparse chunk.
+ *
+ * We must update the finobt record based on the inobt record.
+ * rec contains the fully merged and up to date inobt record
+ * from the previous call. Set merge false to replace any
+ * existing record with this one.
+ */
+ if (xfs_sb_version_hasfinobt(&args.mp->m_sb)) {
+ error = xfs_inobt_insert_sprec(args.mp, tp, agbp,
+ XFS_BTNUM_FINO, &rec,
+ false);
+ if (error)
+ return error;
+ }
+ } else {
+ /* full chunk - insert new records to both btrees */
+ error = xfs_inobt_insert(args.mp, tp, agbp, newino, newlen,
+ XFS_BTNUM_INO);
+ if (error)
+ return error;
+
+ if (xfs_sb_version_hasfinobt(&args.mp->m_sb)) {
+ error = xfs_inobt_insert(args.mp, tp, agbp, newino,
+ newlen, XFS_BTNUM_FINO);
+ if (error)
+ return error;
+ }
+ }
+
+ /*
+ * Update AGI counts and newino.
+ */
be32_add_cpu(&agi->agi_count, newlen);
be32_add_cpu(&agi->agi_freecount, newlen);
pag = xfs_perag_get(args.mp, agno);
@@ -512,20 +871,6 @@
agi->agi_newino = cpu_to_be32(newino);
/*
- * Insert records describing the new inode chunk into the btrees.
- */
- error = xfs_inobt_insert(args.mp, tp, agbp, newino, newlen,
- XFS_BTNUM_INO);
- if (error)
- return error;
-
- if (xfs_sb_version_hasfinobt(&args.mp->m_sb)) {
- error = xfs_inobt_insert(args.mp, tp, agbp, newino, newlen,
- XFS_BTNUM_FINO);
- if (error)
- return error;
- }
- /*
* Log allocation group header fields
*/
xfs_ialloc_log_agi(tp, agbp,
@@ -645,7 +990,7 @@
* if we fail allocation due to alignment issues then it is most
* likely a real ENOSPC condition.
*/
- ineed = mp->m_ialloc_blks;
+ ineed = mp->m_ialloc_min_blks;
if (flags && ineed > 1)
ineed += xfs_ialloc_cluster_alignment(mp);
longest = pag->pagf_longest;
@@ -732,6 +1077,27 @@
}
/*
+ * Return the offset of the first free inode in the record. If the inode chunk
+ * is sparsely allocated, we convert the record holemask to inode granularity
+ * and mask off the unallocated regions from the inode free mask.
+ */
+STATIC int
+xfs_inobt_first_free_inode(
+ struct xfs_inobt_rec_incore *rec)
+{
+ xfs_inofree_t realfree;
+
+ /* if there are no holes, return the first available offset */
+ if (!xfs_inobt_issparse(rec->ir_holemask))
+ return xfs_lowbit64(rec->ir_free);
+
+ realfree = xfs_inobt_irec_to_allocmask(rec);
+ realfree &= rec->ir_free;
+
+ return xfs_lowbit64(realfree);
+}
+
+/*
* Allocate an inode using the inobt-only algorithm.
*/
STATIC int
@@ -961,7 +1327,7 @@
}
alloc_inode:
- offset = xfs_lowbit64(rec.ir_free);
+ offset = xfs_inobt_first_free_inode(&rec);
ASSERT(offset >= 0);
ASSERT(offset < XFS_INODES_PER_CHUNK);
ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) %
@@ -1210,7 +1576,7 @@
if (error)
goto error_cur;
- offset = xfs_lowbit64(rec.ir_free);
+ offset = xfs_inobt_first_free_inode(&rec);
ASSERT(offset >= 0);
ASSERT(offset < XFS_INODES_PER_CHUNK);
ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) %
@@ -1339,10 +1705,13 @@
* If we have already hit the ceiling of inode blocks then clear
* okalloc so we scan all available agi structures for a free
* inode.
+ *
+ * Read rough value of mp->m_icount by percpu_counter_read_positive,
+ * which will sacrifice the preciseness but improve the performance.
*/
if (mp->m_maxicount &&
- percpu_counter_read(&mp->m_icount) + mp->m_ialloc_inos >
- mp->m_maxicount) {
+ percpu_counter_read_positive(&mp->m_icount) + mp->m_ialloc_inos
+ > mp->m_maxicount) {
noroom = 1;
okalloc = 0;
}
@@ -1436,6 +1805,83 @@
return error;
}
+/*
+ * Free the blocks of an inode chunk. We must consider that the inode chunk
+ * might be sparse and only free the regions that are allocated as part of the
+ * chunk.
+ */
+STATIC void
+xfs_difree_inode_chunk(
+ struct xfs_mount *mp,
+ xfs_agnumber_t agno,
+ struct xfs_inobt_rec_incore *rec,
+ struct xfs_bmap_free *flist)
+{
+ xfs_agblock_t sagbno = XFS_AGINO_TO_AGBNO(mp, rec->ir_startino);
+ int startidx, endidx;
+ int nextbit;
+ xfs_agblock_t agbno;
+ int contigblk;
+ DECLARE_BITMAP(holemask, XFS_INOBT_HOLEMASK_BITS);
+
+ if (!xfs_inobt_issparse(rec->ir_holemask)) {
+ /* not sparse, calculate extent info directly */
+ xfs_bmap_add_free(XFS_AGB_TO_FSB(mp, agno,
+ XFS_AGINO_TO_AGBNO(mp, rec->ir_startino)),
+ mp->m_ialloc_blks, flist, mp);
+ return;
+ }
+
+ /* holemask is only 16-bits (fits in an unsigned long) */
+ ASSERT(sizeof(rec->ir_holemask) <= sizeof(holemask[0]));
+ holemask[0] = rec->ir_holemask;
+
+ /*
+ * Find contiguous ranges of zeroes (i.e., allocated regions) in the
+ * holemask and convert the start/end index of each range to an extent.
+ * We start with the start and end index both pointing at the first 0 in
+ * the mask.
+ */
+ startidx = endidx = find_first_zero_bit(holemask,
+ XFS_INOBT_HOLEMASK_BITS);
+ nextbit = startidx + 1;
+ while (startidx < XFS_INOBT_HOLEMASK_BITS) {
+ nextbit = find_next_zero_bit(holemask, XFS_INOBT_HOLEMASK_BITS,
+ nextbit);
+ /*
+ * If the next zero bit is contiguous, update the end index of
+ * the current range and continue.
+ */
+ if (nextbit != XFS_INOBT_HOLEMASK_BITS &&
+ nextbit == endidx + 1) {
+ endidx = nextbit;
+ goto next;
+ }
+
+ /*
+ * nextbit is not contiguous with the current end index. Convert
+ * the current start/end to an extent and add it to the free
+ * list.
+ */
+ agbno = sagbno + (startidx * XFS_INODES_PER_HOLEMASK_BIT) /
+ mp->m_sb.sb_inopblock;
+ contigblk = ((endidx - startidx + 1) *
+ XFS_INODES_PER_HOLEMASK_BIT) /
+ mp->m_sb.sb_inopblock;
+
+ ASSERT(agbno % mp->m_sb.sb_spino_align == 0);
+ ASSERT(contigblk % mp->m_sb.sb_spino_align == 0);
+ xfs_bmap_add_free(XFS_AGB_TO_FSB(mp, agno, agbno), contigblk,
+ flist, mp);
+
+ /* reset range to current bit and carry on... */
+ startidx = endidx = nextbit;
+
+next:
+ nextbit++;
+ }
+}
+
STATIC int
xfs_difree_inobt(
struct xfs_mount *mp,
@@ -1443,8 +1889,7 @@
struct xfs_buf *agbp,
xfs_agino_t agino,
struct xfs_bmap_free *flist,
- int *deleted,
- xfs_ino_t *first_ino,
+ struct xfs_icluster *xic,
struct xfs_inobt_rec_incore *orec)
{
struct xfs_agi *agi = XFS_BUF_TO_AGI(agbp);
@@ -1498,20 +1943,23 @@
rec.ir_freecount++;
/*
- * When an inode cluster is free, it becomes eligible for removal
+ * When an inode chunk is free, it becomes eligible for removal. Don't
+ * remove the chunk if the block size is large enough for multiple inode
+ * chunks (that might not be free).
*/
if (!(mp->m_flags & XFS_MOUNT_IKEEP) &&
- (rec.ir_freecount == mp->m_ialloc_inos)) {
-
- *deleted = 1;
- *first_ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino);
+ rec.ir_free == XFS_INOBT_ALL_FREE &&
+ mp->m_sb.sb_inopblock <= XFS_INODES_PER_CHUNK) {
+ xic->deleted = 1;
+ xic->first_ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino);
+ xic->alloc = xfs_inobt_irec_to_allocmask(&rec);
/*
* Remove the inode cluster from the AGI B+Tree, adjust the
* AGI and Superblock inode counts, and mark the disk space
* to be freed when the transaction is committed.
*/
- ilen = mp->m_ialloc_inos;
+ ilen = rec.ir_freecount;
be32_add_cpu(&agi->agi_count, -ilen);
be32_add_cpu(&agi->agi_freecount, -(ilen - 1));
xfs_ialloc_log_agi(tp, agbp, XFS_AGI_COUNT | XFS_AGI_FREECOUNT);
@@ -1527,11 +1975,9 @@
goto error0;
}
- xfs_bmap_add_free(XFS_AGB_TO_FSB(mp, agno,
- XFS_AGINO_TO_AGBNO(mp, rec.ir_startino)),
- mp->m_ialloc_blks, flist, mp);
+ xfs_difree_inode_chunk(mp, agno, &rec, flist);
} else {
- *deleted = 0;
+ xic->deleted = 0;
error = xfs_inobt_update(cur, &rec);
if (error) {
@@ -1596,7 +2042,9 @@
*/
XFS_WANT_CORRUPTED_GOTO(mp, ibtrec->ir_freecount == 1, error);
- error = xfs_inobt_insert_rec(cur, ibtrec->ir_freecount,
+ error = xfs_inobt_insert_rec(cur, ibtrec->ir_holemask,
+ ibtrec->ir_count,
+ ibtrec->ir_freecount,
ibtrec->ir_free, &i);
if (error)
goto error;
@@ -1631,8 +2079,13 @@
* free inode. Hence, if all of the inodes are free and we aren't
* keeping inode chunks permanently on disk, remove the record.
* Otherwise, update the record with the new information.
+ *
+ * Note that we currently can't free chunks when the block size is large
+ * enough for multiple chunks. Leave the finobt record to remain in sync
+ * with the inobt.
*/
- if (rec.ir_freecount == mp->m_ialloc_inos &&
+ if (rec.ir_free == XFS_INOBT_ALL_FREE &&
+ mp->m_sb.sb_inopblock <= XFS_INODES_PER_CHUNK &&
!(mp->m_flags & XFS_MOUNT_IKEEP)) {
error = xfs_btree_delete(cur, &i);
if (error)
@@ -1668,8 +2121,7 @@
struct xfs_trans *tp, /* transaction pointer */
xfs_ino_t inode, /* inode to be freed */
struct xfs_bmap_free *flist, /* extents to free */
- int *deleted,/* set if inode cluster was deleted */
- xfs_ino_t *first_ino)/* first inode in deleted cluster */
+ struct xfs_icluster *xic) /* cluster info if deleted */
{
/* REFERENCED */
xfs_agblock_t agbno; /* block number containing inode */
@@ -1720,8 +2172,7 @@
/*
* Fix up the inode allocation btree.
*/
- error = xfs_difree_inobt(mp, tp, agbp, agino, flist, deleted, first_ino,
- &rec);
+ error = xfs_difree_inobt(mp, tp, agbp, agino, flist, xic, &rec);
if (error)
goto error0;
diff --git a/fs/xfs/libxfs/xfs_ialloc.h b/fs/xfs/libxfs/xfs_ialloc.h
index 100007d..12401fe 100644
--- a/fs/xfs/libxfs/xfs_ialloc.h
+++ b/fs/xfs/libxfs/xfs_ialloc.h
@@ -28,6 +28,13 @@
/* Move inodes in clusters of this size */
#define XFS_INODE_BIG_CLUSTER_SIZE 8192
+struct xfs_icluster {
+ bool deleted; /* record is deleted */
+ xfs_ino_t first_ino; /* first inode number */
+ uint64_t alloc; /* inode phys. allocation bitmap for
+ * sparse chunks */
+};
+
/* Calculate and return the number of filesystem blocks per inode cluster */
static inline int
xfs_icluster_size_fsb(
@@ -90,8 +97,7 @@
struct xfs_trans *tp, /* transaction pointer */
xfs_ino_t inode, /* inode to be freed */
struct xfs_bmap_free *flist, /* extents to free */
- int *deleted, /* set if inode cluster was deleted */
- xfs_ino_t *first_ino); /* first inode in deleted cluster */
+ struct xfs_icluster *ifree); /* cluster info if deleted */
/*
* Return the location of the inode in imap, for mapping it into a buffer.
@@ -156,7 +162,7 @@
* Inode chunk initialisation routine
*/
int xfs_ialloc_inode_init(struct xfs_mount *mp, struct xfs_trans *tp,
- struct list_head *buffer_list,
+ struct list_head *buffer_list, int icount,
xfs_agnumber_t agno, xfs_agblock_t agbno,
xfs_agblock_t length, unsigned int gen);
diff --git a/fs/xfs/libxfs/xfs_ialloc_btree.c b/fs/xfs/libxfs/xfs_ialloc_btree.c
index 964c465..674ad8f 100644
--- a/fs/xfs/libxfs/xfs_ialloc_btree.c
+++ b/fs/xfs/libxfs/xfs_ialloc_btree.c
@@ -167,7 +167,16 @@
union xfs_btree_rec *rec)
{
rec->inobt.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
- rec->inobt.ir_freecount = cpu_to_be32(cur->bc_rec.i.ir_freecount);
+ if (xfs_sb_version_hassparseinodes(&cur->bc_mp->m_sb)) {
+ rec->inobt.ir_u.sp.ir_holemask =
+ cpu_to_be16(cur->bc_rec.i.ir_holemask);
+ rec->inobt.ir_u.sp.ir_count = cur->bc_rec.i.ir_count;
+ rec->inobt.ir_u.sp.ir_freecount = cur->bc_rec.i.ir_freecount;
+ } else {
+ /* ir_holemask/ir_count not supported on-disk */
+ rec->inobt.ir_u.f.ir_freecount =
+ cpu_to_be32(cur->bc_rec.i.ir_freecount);
+ }
rec->inobt.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
}
@@ -418,3 +427,85 @@
return blocklen / sizeof(xfs_inobt_rec_t);
return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
}
+
+/*
+ * Convert the inode record holemask to an inode allocation bitmap. The inode
+ * allocation bitmap is inode granularity and specifies whether an inode is
+ * physically allocated on disk (not whether the inode is considered allocated
+ * or free by the fs).
+ *
+ * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
+ */
+uint64_t
+xfs_inobt_irec_to_allocmask(
+ struct xfs_inobt_rec_incore *rec)
+{
+ uint64_t bitmap = 0;
+ uint64_t inodespbit;
+ int nextbit;
+ uint allocbitmap;
+
+ /*
+ * The holemask has 16-bits for a 64 inode record. Therefore each
+ * holemask bit represents multiple inodes. Create a mask of bits to set
+ * in the allocmask for each holemask bit.
+ */
+ inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
+
+ /*
+ * Allocated inodes are represented by 0 bits in holemask. Invert the 0
+ * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
+ * anything beyond the 16 holemask bits since this casts to a larger
+ * type.
+ */
+ allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
+
+ /*
+ * allocbitmap is the inverted holemask so every set bit represents
+ * allocated inodes. To expand from 16-bit holemask granularity to
+ * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
+ * bitmap for every holemask bit.
+ */
+ nextbit = xfs_next_bit(&allocbitmap, 1, 0);
+ while (nextbit != -1) {
+ ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
+
+ bitmap |= (inodespbit <<
+ (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
+
+ nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
+ }
+
+ return bitmap;
+}
+
+#if defined(DEBUG) || defined(XFS_WARN)
+/*
+ * Verify that an in-core inode record has a valid inode count.
+ */
+int
+xfs_inobt_rec_check_count(
+ struct xfs_mount *mp,
+ struct xfs_inobt_rec_incore *rec)
+{
+ int inocount = 0;
+ int nextbit = 0;
+ uint64_t allocbmap;
+ int wordsz;
+
+ wordsz = sizeof(allocbmap) / sizeof(unsigned int);
+ allocbmap = xfs_inobt_irec_to_allocmask(rec);
+
+ nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
+ while (nextbit != -1) {
+ inocount++;
+ nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
+ nextbit + 1);
+ }
+
+ if (inocount != rec->ir_count)
+ return -EFSCORRUPTED;
+
+ return 0;
+}
+#endif /* DEBUG */
diff --git a/fs/xfs/libxfs/xfs_ialloc_btree.h b/fs/xfs/libxfs/xfs_ialloc_btree.h
index d7ebea7..bd88453 100644
--- a/fs/xfs/libxfs/xfs_ialloc_btree.h
+++ b/fs/xfs/libxfs/xfs_ialloc_btree.h
@@ -62,4 +62,14 @@
xfs_btnum_t);
extern int xfs_inobt_maxrecs(struct xfs_mount *, int, int);
+/* ir_holemask to inode allocation bitmap conversion */
+uint64_t xfs_inobt_irec_to_allocmask(struct xfs_inobt_rec_incore *);
+
+#if defined(DEBUG) || defined(XFS_WARN)
+int xfs_inobt_rec_check_count(struct xfs_mount *,
+ struct xfs_inobt_rec_incore *);
+#else
+#define xfs_inobt_rec_check_count(mp, rec) 0
+#endif /* DEBUG */
+
#endif /* __XFS_IALLOC_BTREE_H__ */
diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c
index fcc151f..df9851c 100644
--- a/fs/xfs/libxfs/xfs_sb.c
+++ b/fs/xfs/libxfs/xfs_sb.c
@@ -174,6 +174,27 @@
return -EFSCORRUPTED;
}
+ /*
+ * Full inode chunks must be aligned to inode chunk size when
+ * sparse inodes are enabled to support the sparse chunk
+ * allocation algorithm and prevent overlapping inode records.
+ */
+ if (xfs_sb_version_hassparseinodes(sbp)) {
+ uint32_t align;
+
+ xfs_alert(mp,
+ "EXPERIMENTAL sparse inode feature enabled. Use at your own risk!");
+
+ align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
+ >> sbp->sb_blocklog;
+ if (sbp->sb_inoalignmt != align) {
+ xfs_warn(mp,
+"Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
+ sbp->sb_inoalignmt, align);
+ return -EINVAL;
+ }
+ }
+
if (unlikely(
sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
xfs_warn(mp,
@@ -374,7 +395,7 @@
be32_to_cpu(from->sb_features_log_incompat);
/* crc is only used on disk, not in memory; just init to 0 here. */
to->sb_crc = 0;
- to->sb_pad = 0;
+ to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
to->sb_lsn = be64_to_cpu(from->sb_lsn);
/* Convert on-disk flags to in-memory flags? */
@@ -516,7 +537,7 @@
cpu_to_be32(from->sb_features_incompat);
to->sb_features_log_incompat =
cpu_to_be32(from->sb_features_log_incompat);
- to->sb_pad = 0;
+ to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
to->sb_lsn = cpu_to_be64(from->sb_lsn);
}
}
@@ -689,6 +710,11 @@
mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
sbp->sb_inopblock);
mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
+
+ if (sbp->sb_spino_align)
+ mp->m_ialloc_min_blks = sbp->sb_spino_align;
+ else
+ mp->m_ialloc_min_blks = mp->m_ialloc_blks;
}
/*
diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index 7246a39..dc52698 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -1349,7 +1349,7 @@
sector_t iblock,
struct buffer_head *bh_result,
int create,
- int direct)
+ bool direct)
{
struct xfs_inode *ip = XFS_I(inode);
struct xfs_mount *mp = ip->i_mount;
@@ -1414,6 +1414,7 @@
if (error)
return error;
new = 1;
+
} else {
/*
* Delalloc reservations do not require a transaction,
@@ -1508,49 +1509,29 @@
struct buffer_head *bh_result,
int create)
{
- return __xfs_get_blocks(inode, iblock, bh_result, create, 0);
+ return __xfs_get_blocks(inode, iblock, bh_result, create, false);
}
-STATIC int
+int
xfs_get_blocks_direct(
struct inode *inode,
sector_t iblock,
struct buffer_head *bh_result,
int create)
{
- return __xfs_get_blocks(inode, iblock, bh_result, create, 1);
+ return __xfs_get_blocks(inode, iblock, bh_result, create, true);
}
-/*
- * Complete a direct I/O write request.
- *
- * The ioend structure is passed from __xfs_get_blocks() to tell us what to do.
- * If no ioend exists (i.e. @private == NULL) then the write IO is an overwrite
- * wholly within the EOF and so there is nothing for us to do. Note that in this
- * case the completion can be called in interrupt context, whereas if we have an
- * ioend we will always be called in task context (i.e. from a workqueue).
- */
-STATIC void
-xfs_end_io_direct_write(
- struct kiocb *iocb,
+static void
+__xfs_end_io_direct_write(
+ struct inode *inode,
+ struct xfs_ioend *ioend,
loff_t offset,
- ssize_t size,
- void *private)
+ ssize_t size)
{
- struct inode *inode = file_inode(iocb->ki_filp);
- struct xfs_inode *ip = XFS_I(inode);
- struct xfs_mount *mp = ip->i_mount;
- struct xfs_ioend *ioend = private;
+ struct xfs_mount *mp = XFS_I(inode)->i_mount;
- trace_xfs_gbmap_direct_endio(ip, offset, size,
- ioend ? ioend->io_type : 0, NULL);
-
- if (!ioend) {
- ASSERT(offset + size <= i_size_read(inode));
- return;
- }
-
- if (XFS_FORCED_SHUTDOWN(mp))
+ if (XFS_FORCED_SHUTDOWN(mp) || ioend->io_error)
goto out_end_io;
/*
@@ -1587,10 +1568,10 @@
* here can result in EOF moving backwards and Bad Things Happen when
* that occurs.
*/
- spin_lock(&ip->i_flags_lock);
+ spin_lock(&XFS_I(inode)->i_flags_lock);
if (offset + size > i_size_read(inode))
i_size_write(inode, offset + size);
- spin_unlock(&ip->i_flags_lock);
+ spin_unlock(&XFS_I(inode)->i_flags_lock);
/*
* If we are doing an append IO that needs to update the EOF on disk,
@@ -1607,6 +1588,98 @@
return;
}
+/*
+ * Complete a direct I/O write request.
+ *
+ * The ioend structure is passed from __xfs_get_blocks() to tell us what to do.
+ * If no ioend exists (i.e. @private == NULL) then the write IO is an overwrite
+ * wholly within the EOF and so there is nothing for us to do. Note that in this
+ * case the completion can be called in interrupt context, whereas if we have an
+ * ioend we will always be called in task context (i.e. from a workqueue).
+ */
+STATIC void
+xfs_end_io_direct_write(
+ struct kiocb *iocb,
+ loff_t offset,
+ ssize_t size,
+ void *private)
+{
+ struct inode *inode = file_inode(iocb->ki_filp);
+ struct xfs_ioend *ioend = private;
+
+ trace_xfs_gbmap_direct_endio(XFS_I(inode), offset, size,
+ ioend ? ioend->io_type : 0, NULL);
+
+ if (!ioend) {
+ ASSERT(offset + size <= i_size_read(inode));
+ return;
+ }
+
+ __xfs_end_io_direct_write(inode, ioend, offset, size);
+}
+
+/*
+ * For DAX we need a mapping buffer callback for unwritten extent conversion
+ * when page faults allocate blocks and then zero them. Note that in this
+ * case the mapping indicated by the ioend may extend beyond EOF. We most
+ * definitely do not want to extend EOF here, so we trim back the ioend size to
+ * EOF.
+ */
+#ifdef CONFIG_FS_DAX
+void
+xfs_end_io_dax_write(
+ struct buffer_head *bh,
+ int uptodate)
+{
+ struct xfs_ioend *ioend = bh->b_private;
+ struct inode *inode = ioend->io_inode;
+ ssize_t size = ioend->io_size;
+
+ ASSERT(IS_DAX(ioend->io_inode));
+
+ /* if there was an error zeroing, then don't convert it */
+ if (!uptodate)
+ ioend->io_error = -EIO;
+
+ /*
+ * Trim update to EOF, so we don't extend EOF during unwritten extent
+ * conversion of partial EOF blocks.
+ */
+ spin_lock(&XFS_I(inode)->i_flags_lock);
+ if (ioend->io_offset + size > i_size_read(inode))
+ size = i_size_read(inode) - ioend->io_offset;
+ spin_unlock(&XFS_I(inode)->i_flags_lock);
+
+ __xfs_end_io_direct_write(inode, ioend, ioend->io_offset, size);
+
+}
+#else
+void xfs_end_io_dax_write(struct buffer_head *bh, int uptodate) { }
+#endif
+
+static inline ssize_t
+xfs_vm_do_dio(
+ struct inode *inode,
+ struct kiocb *iocb,
+ struct iov_iter *iter,
+ loff_t offset,
+ void (*endio)(struct kiocb *iocb,
+ loff_t offset,
+ ssize_t size,
+ void *private),
+ int flags)
+{
+ struct block_device *bdev;
+
+ if (IS_DAX(inode))
+ return dax_do_io(iocb, inode, iter, offset,
+ xfs_get_blocks_direct, endio, 0);
+
+ bdev = xfs_find_bdev_for_inode(inode);
+ return __blockdev_direct_IO(iocb, inode, bdev, iter, offset,
+ xfs_get_blocks_direct, endio, NULL, flags);
+}
+
STATIC ssize_t
xfs_vm_direct_IO(
struct kiocb *iocb,
@@ -1614,16 +1687,11 @@
loff_t offset)
{
struct inode *inode = iocb->ki_filp->f_mapping->host;
- struct block_device *bdev = xfs_find_bdev_for_inode(inode);
- if (iov_iter_rw(iter) == WRITE) {
- return __blockdev_direct_IO(iocb, inode, bdev, iter, offset,
- xfs_get_blocks_direct,
- xfs_end_io_direct_write, NULL,
- DIO_ASYNC_EXTEND);
- }
- return __blockdev_direct_IO(iocb, inode, bdev, iter, offset,
- xfs_get_blocks_direct, NULL, NULL, 0);
+ if (iov_iter_rw(iter) == WRITE)
+ return xfs_vm_do_dio(inode, iocb, iter, offset,
+ xfs_end_io_direct_write, DIO_ASYNC_EXTEND);
+ return xfs_vm_do_dio(inode, iocb, iter, offset, NULL, 0);
}
/*
diff --git a/fs/xfs/xfs_aops.h b/fs/xfs/xfs_aops.h
index ac644e0..86afd1a 100644
--- a/fs/xfs/xfs_aops.h
+++ b/fs/xfs/xfs_aops.h
@@ -53,7 +53,12 @@
} xfs_ioend_t;
extern const struct address_space_operations xfs_address_space_operations;
-extern int xfs_get_blocks(struct inode *, sector_t, struct buffer_head *, int);
+
+int xfs_get_blocks(struct inode *inode, sector_t offset,
+ struct buffer_head *map_bh, int create);
+int xfs_get_blocks_direct(struct inode *inode, sector_t offset,
+ struct buffer_head *map_bh, int create);
+void xfs_end_io_dax_write(struct buffer_head *bh, int uptodate);
extern void xfs_count_page_state(struct page *, int *, int *);
diff --git a/fs/xfs/xfs_attr_inactive.c b/fs/xfs/xfs_attr_inactive.c
index 48f26ff..69a154c 100644
--- a/fs/xfs/xfs_attr_inactive.c
+++ b/fs/xfs/xfs_attr_inactive.c
@@ -380,23 +380,30 @@
return error;
}
+/*
+ * xfs_attr_inactive kills all traces of an attribute fork on an inode. It
+ * removes both the on-disk and in-memory inode fork. Note that this also has to
+ * handle the condition of inodes without attributes but with an attribute fork
+ * configured, so we can't use xfs_inode_hasattr() here.
+ *
+ * The in-memory attribute fork is removed even on error.
+ */
int
-xfs_attr_inactive(xfs_inode_t *dp)
+xfs_attr_inactive(
+ struct xfs_inode *dp)
{
- xfs_trans_t *trans;
- xfs_mount_t *mp;
- int error;
+ struct xfs_trans *trans;
+ struct xfs_mount *mp;
+ int lock_mode = XFS_ILOCK_SHARED;
+ int error = 0;
mp = dp->i_mount;
ASSERT(! XFS_NOT_DQATTACHED(mp, dp));
- xfs_ilock(dp, XFS_ILOCK_SHARED);
- if (!xfs_inode_hasattr(dp) ||
- dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
- xfs_iunlock(dp, XFS_ILOCK_SHARED);
- return 0;
- }
- xfs_iunlock(dp, XFS_ILOCK_SHARED);
+ xfs_ilock(dp, lock_mode);
+ if (!XFS_IFORK_Q(dp))
+ goto out_destroy_fork;
+ xfs_iunlock(dp, lock_mode);
/*
* Start our first transaction of the day.
@@ -408,13 +415,17 @@
* the inode in every transaction to let it float upward through
* the log.
*/
+ lock_mode = 0;
trans = xfs_trans_alloc(mp, XFS_TRANS_ATTRINVAL);
error = xfs_trans_reserve(trans, &M_RES(mp)->tr_attrinval, 0, 0);
- if (error) {
- xfs_trans_cancel(trans);
- return error;
- }
- xfs_ilock(dp, XFS_ILOCK_EXCL);
+ if (error)
+ goto out_cancel;
+
+ lock_mode = XFS_ILOCK_EXCL;
+ xfs_ilock(dp, lock_mode);
+
+ if (!XFS_IFORK_Q(dp))
+ goto out_cancel;
/*
* No need to make quota reservations here. We expect to release some
@@ -422,29 +433,31 @@
*/
xfs_trans_ijoin(trans, dp, 0);
- /*
- * Decide on what work routines to call based on the inode size.
- */
- if (!xfs_inode_hasattr(dp) ||
- dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
- error = 0;
- goto out;
- }
- error = xfs_attr3_root_inactive(&trans, dp);
- if (error)
- goto out;
+ /* invalidate and truncate the attribute fork extents */
+ if (dp->i_d.di_aformat != XFS_DINODE_FMT_LOCAL) {
+ error = xfs_attr3_root_inactive(&trans, dp);
+ if (error)
+ goto out_cancel;
- error = xfs_itruncate_extents(&trans, dp, XFS_ATTR_FORK, 0);
- if (error)
- goto out;
+ error = xfs_itruncate_extents(&trans, dp, XFS_ATTR_FORK, 0);
+ if (error)
+ goto out_cancel;
+ }
+
+ /* Reset the attribute fork - this also destroys the in-core fork */
+ xfs_attr_fork_remove(dp, trans);
error = xfs_trans_commit(trans);
- xfs_iunlock(dp, XFS_ILOCK_EXCL);
-
+ xfs_iunlock(dp, lock_mode);
return error;
-out:
+out_cancel:
xfs_trans_cancel(trans);
- xfs_iunlock(dp, XFS_ILOCK_EXCL);
+out_destroy_fork:
+ /* kill the in-core attr fork before we drop the inode lock */
+ if (dp->i_afp)
+ xfs_idestroy_fork(dp, XFS_ATTR_FORK);
+ if (lock_mode)
+ xfs_iunlock(dp, lock_mode);
return error;
}
diff --git a/fs/xfs/xfs_bmap_util.c b/fs/xfs/xfs_bmap_util.c
index 1f0215d..0f34886c 100644
--- a/fs/xfs/xfs_bmap_util.c
+++ b/fs/xfs/xfs_bmap_util.c
@@ -1113,14 +1113,29 @@
break;
ASSERT(imap.br_blockcount >= 1);
ASSERT(imap.br_startoff == offset_fsb);
+ ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
+
+ if (imap.br_startblock == HOLESTARTBLOCK ||
+ imap.br_state == XFS_EXT_UNWRITTEN) {
+ /* skip the entire extent */
+ lastoffset = XFS_FSB_TO_B(mp, imap.br_startoff +
+ imap.br_blockcount) - 1;
+ continue;
+ }
+
lastoffset = XFS_FSB_TO_B(mp, imap.br_startoff + 1) - 1;
if (lastoffset > endoff)
lastoffset = endoff;
- if (imap.br_startblock == HOLESTARTBLOCK)
+
+ /* DAX can just zero the backing device directly */
+ if (IS_DAX(VFS_I(ip))) {
+ error = dax_zero_page_range(VFS_I(ip), offset,
+ lastoffset - offset + 1,
+ xfs_get_blocks_direct);
+ if (error)
+ return error;
continue;
- ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
- if (imap.br_state == XFS_EXT_UNWRITTEN)
- continue;
+ }
error = xfs_buf_read_uncached(XFS_IS_REALTIME_INODE(ip) ?
mp->m_rtdev_targp : mp->m_ddev_targp,
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 0dec858..97d92c1 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -79,14 +79,15 @@
}
/*
- * xfs_iozero
+ * xfs_iozero clears the specified range supplied via the page cache (except in
+ * the DAX case). Writes through the page cache will allocate blocks over holes,
+ * though the callers usually map the holes first and avoid them. If a block is
+ * not completely zeroed, then it will be read from disk before being partially
+ * zeroed.
*
- * xfs_iozero clears the specified range of buffer supplied,
- * and marks all the affected blocks as valid and modified. If
- * an affected block is not allocated, it will be allocated. If
- * an affected block is not completely overwritten, and is not
- * valid before the operation, it will be read from disk before
- * being partially zeroed.
+ * In the DAX case, we can just directly write to the underlying pages. This
+ * will not allocate blocks, but will avoid holes and unwritten extents and so
+ * not do unnecessary work.
*/
int
xfs_iozero(
@@ -96,7 +97,8 @@
{
struct page *page;
struct address_space *mapping;
- int status;
+ int status = 0;
+
mapping = VFS_I(ip)->i_mapping;
do {
@@ -108,23 +110,30 @@
if (bytes > count)
bytes = count;
- status = pagecache_write_begin(NULL, mapping, pos, bytes,
- AOP_FLAG_UNINTERRUPTIBLE,
- &page, &fsdata);
- if (status)
- break;
+ if (IS_DAX(VFS_I(ip))) {
+ status = dax_zero_page_range(VFS_I(ip), pos, bytes,
+ xfs_get_blocks_direct);
+ if (status)
+ break;
+ } else {
+ status = pagecache_write_begin(NULL, mapping, pos, bytes,
+ AOP_FLAG_UNINTERRUPTIBLE,
+ &page, &fsdata);
+ if (status)
+ break;
- zero_user(page, offset, bytes);
+ zero_user(page, offset, bytes);
- status = pagecache_write_end(NULL, mapping, pos, bytes, bytes,
- page, fsdata);
- WARN_ON(status <= 0); /* can't return less than zero! */
+ status = pagecache_write_end(NULL, mapping, pos, bytes,
+ bytes, page, fsdata);
+ WARN_ON(status <= 0); /* can't return less than zero! */
+ status = 0;
+ }
pos += bytes;
count -= bytes;
- status = 0;
} while (count);
- return (-status);
+ return status;
}
int
@@ -284,7 +293,7 @@
if (file->f_mode & FMODE_NOCMTIME)
ioflags |= XFS_IO_INVIS;
- if (unlikely(ioflags & XFS_IO_ISDIRECT)) {
+ if ((ioflags & XFS_IO_ISDIRECT) && !IS_DAX(inode)) {
xfs_buftarg_t *target =
XFS_IS_REALTIME_INODE(ip) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
@@ -378,7 +387,11 @@
trace_xfs_file_splice_read(ip, count, *ppos, ioflags);
- ret = generic_file_splice_read(infilp, ppos, pipe, count, flags);
+ /* for dax, we need to avoid the page cache */
+ if (IS_DAX(VFS_I(ip)))
+ ret = default_file_splice_read(infilp, ppos, pipe, count, flags);
+ else
+ ret = generic_file_splice_read(infilp, ppos, pipe, count, flags);
if (ret > 0)
XFS_STATS_ADD(xs_read_bytes, ret);
@@ -672,7 +685,7 @@
mp->m_rtdev_targp : mp->m_ddev_targp;
/* DIO must be aligned to device logical sector size */
- if ((pos | count) & target->bt_logical_sectormask)
+ if (!IS_DAX(inode) && ((pos | count) & target->bt_logical_sectormask))
return -EINVAL;
/* "unaligned" here means not aligned to a filesystem block */
@@ -758,8 +771,11 @@
out:
xfs_rw_iunlock(ip, iolock);
- /* No fallback to buffered IO on errors for XFS. */
- ASSERT(ret < 0 || ret == count);
+ /*
+ * No fallback to buffered IO on errors for XFS. DAX can result in
+ * partial writes, but direct IO will either complete fully or fail.
+ */
+ ASSERT(ret < 0 || ret == count || IS_DAX(VFS_I(ip)));
return ret;
}
@@ -842,7 +858,7 @@
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
return -EIO;
- if (unlikely(iocb->ki_flags & IOCB_DIRECT))
+ if ((iocb->ki_flags & IOCB_DIRECT) || IS_DAX(inode))
ret = xfs_file_dio_aio_write(iocb, from);
else
ret = xfs_file_buffered_aio_write(iocb, from);
@@ -1063,17 +1079,6 @@
return xfs_readdir(ip, ctx, bufsize);
}
-STATIC int
-xfs_file_mmap(
- struct file *filp,
- struct vm_area_struct *vma)
-{
- vma->vm_ops = &xfs_file_vm_ops;
-
- file_accessed(filp);
- return 0;
-}
-
/*
* This type is designed to indicate the type of offset we would like
* to search from page cache for xfs_seek_hole_data().
@@ -1454,26 +1459,11 @@
* ordering of:
*
* mmap_sem (MM)
- * i_mmap_lock (XFS - truncate serialisation)
- * page_lock (MM)
- * i_lock (XFS - extent map serialisation)
+ * sb_start_pagefault(vfs, freeze)
+ * i_mmap_lock (XFS - truncate serialisation)
+ * page_lock (MM)
+ * i_lock (XFS - extent map serialisation)
*/
-STATIC int
-xfs_filemap_fault(
- struct vm_area_struct *vma,
- struct vm_fault *vmf)
-{
- struct xfs_inode *ip = XFS_I(vma->vm_file->f_mapping->host);
- int error;
-
- trace_xfs_filemap_fault(ip);
-
- xfs_ilock(ip, XFS_MMAPLOCK_SHARED);
- error = filemap_fault(vma, vmf);
- xfs_iunlock(ip, XFS_MMAPLOCK_SHARED);
-
- return error;
-}
/*
* mmap()d file has taken write protection fault and is being made writable. We
@@ -1486,16 +1476,66 @@
struct vm_area_struct *vma,
struct vm_fault *vmf)
{
- struct xfs_inode *ip = XFS_I(vma->vm_file->f_mapping->host);
- int error;
+ struct inode *inode = file_inode(vma->vm_file);
+ int ret;
- trace_xfs_filemap_page_mkwrite(ip);
+ trace_xfs_filemap_page_mkwrite(XFS_I(inode));
+
+ sb_start_pagefault(inode->i_sb);
+ file_update_time(vma->vm_file);
+ xfs_ilock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
+
+ if (IS_DAX(inode)) {
+ ret = __dax_mkwrite(vma, vmf, xfs_get_blocks_direct,
+ xfs_end_io_dax_write);
+ } else {
+ ret = __block_page_mkwrite(vma, vmf, xfs_get_blocks);
+ ret = block_page_mkwrite_return(ret);
+ }
+
+ xfs_iunlock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
+ sb_end_pagefault(inode->i_sb);
+
+ return ret;
+}
+
+STATIC int
+xfs_filemap_fault(
+ struct vm_area_struct *vma,
+ struct vm_fault *vmf)
+{
+ struct xfs_inode *ip = XFS_I(file_inode(vma->vm_file));
+ int ret;
+
+ trace_xfs_filemap_fault(ip);
+
+ /* DAX can shortcut the normal fault path on write faults! */
+ if ((vmf->flags & FAULT_FLAG_WRITE) && IS_DAX(VFS_I(ip)))
+ return xfs_filemap_page_mkwrite(vma, vmf);
xfs_ilock(ip, XFS_MMAPLOCK_SHARED);
- error = block_page_mkwrite(vma, vmf, xfs_get_blocks);
+ ret = filemap_fault(vma, vmf);
xfs_iunlock(ip, XFS_MMAPLOCK_SHARED);
- return error;
+ return ret;
+}
+
+static const struct vm_operations_struct xfs_file_vm_ops = {
+ .fault = xfs_filemap_fault,
+ .map_pages = filemap_map_pages,
+ .page_mkwrite = xfs_filemap_page_mkwrite,
+};
+
+STATIC int
+xfs_file_mmap(
+ struct file *filp,
+ struct vm_area_struct *vma)
+{
+ file_accessed(filp);
+ vma->vm_ops = &xfs_file_vm_ops;
+ if (IS_DAX(file_inode(filp)))
+ vma->vm_flags |= VM_MIXEDMAP;
+ return 0;
}
const struct file_operations xfs_file_operations = {
@@ -1526,9 +1566,3 @@
#endif
.fsync = xfs_dir_fsync,
};
-
-static const struct vm_operations_struct xfs_file_vm_ops = {
- .fault = xfs_filemap_fault,
- .map_pages = filemap_map_pages,
- .page_mkwrite = xfs_filemap_page_mkwrite,
-};
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index 0932c15..9b3438a 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -101,7 +101,9 @@
(xfs_sb_version_hasftype(&mp->m_sb) ?
XFS_FSOP_GEOM_FLAGS_FTYPE : 0) |
(xfs_sb_version_hasfinobt(&mp->m_sb) ?
- XFS_FSOP_GEOM_FLAGS_FINOBT : 0);
+ XFS_FSOP_GEOM_FLAGS_FINOBT : 0) |
+ (xfs_sb_version_hassparseinodes(&mp->m_sb) ?
+ XFS_FSOP_GEOM_FLAGS_SPINODES : 0);
geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
mp->m_sb.sb_logsectsize : BBSIZE;
geo->rtsectsize = mp->m_sb.sb_blocksize;
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 63cd400..a37a101 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -1871,21 +1871,17 @@
/*
* If there are attributes associated with the file then blow them away
* now. The code calls a routine that recursively deconstructs the
- * attribute fork. We need to just commit the current transaction
- * because we can't use it for xfs_attr_inactive().
+ * attribute fork. If also blows away the in-core attribute fork.
*/
- if (ip->i_d.di_anextents > 0) {
- ASSERT(ip->i_d.di_forkoff != 0);
-
+ if (XFS_IFORK_Q(ip)) {
error = xfs_attr_inactive(ip);
if (error)
return;
}
- if (ip->i_afp)
- xfs_idestroy_fork(ip, XFS_ATTR_FORK);
-
+ ASSERT(!ip->i_afp);
ASSERT(ip->i_d.di_anextents == 0);
+ ASSERT(ip->i_d.di_forkoff == 0);
/*
* Free the inode.
@@ -2164,28 +2160,42 @@
*/
STATIC int
xfs_ifree_cluster(
- xfs_inode_t *free_ip,
- xfs_trans_t *tp,
- xfs_ino_t inum)
+ xfs_inode_t *free_ip,
+ xfs_trans_t *tp,
+ struct xfs_icluster *xic)
{
xfs_mount_t *mp = free_ip->i_mount;
int blks_per_cluster;
int inodes_per_cluster;
int nbufs;
int i, j;
+ int ioffset;
xfs_daddr_t blkno;
xfs_buf_t *bp;
xfs_inode_t *ip;
xfs_inode_log_item_t *iip;
xfs_log_item_t *lip;
struct xfs_perag *pag;
+ xfs_ino_t inum;
+ inum = xic->first_ino;
pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, inum));
blks_per_cluster = xfs_icluster_size_fsb(mp);
inodes_per_cluster = blks_per_cluster << mp->m_sb.sb_inopblog;
nbufs = mp->m_ialloc_blks / blks_per_cluster;
for (j = 0; j < nbufs; j++, inum += inodes_per_cluster) {
+ /*
+ * The allocation bitmap tells us which inodes of the chunk were
+ * physically allocated. Skip the cluster if an inode falls into
+ * a sparse region.
+ */
+ ioffset = inum - xic->first_ino;
+ if ((xic->alloc & XFS_INOBT_MASK(ioffset)) == 0) {
+ ASSERT(do_mod(ioffset, inodes_per_cluster) == 0);
+ continue;
+ }
+
blkno = XFS_AGB_TO_DADDR(mp, XFS_INO_TO_AGNO(mp, inum),
XFS_INO_TO_AGBNO(mp, inum));
@@ -2343,8 +2353,7 @@
xfs_bmap_free_t *flist)
{
int error;
- int delete;
- xfs_ino_t first_ino;
+ struct xfs_icluster xic = { 0 };
ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
ASSERT(ip->i_d.di_nlink == 0);
@@ -2360,7 +2369,7 @@
if (error)
return error;
- error = xfs_difree(tp, ip->i_ino, flist, &delete, &first_ino);
+ error = xfs_difree(tp, ip->i_ino, flist, &xic);
if (error)
return error;
@@ -2377,8 +2386,8 @@
ip->i_d.di_gen++;
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
- if (delete)
- error = xfs_ifree_cluster(ip, tp, first_ino);
+ if (xic.deleted)
+ error = xfs_ifree_cluster(ip, tp, &xic);
return error;
}
@@ -2804,7 +2813,13 @@
if (error)
return error;
- /* Satisfy xfs_bumplink that this is a real tmpfile */
+ /*
+ * Prepare the tmpfile inode as if it were created through the VFS.
+ * Otherwise, the link increment paths will complain about nlink 0->1.
+ * Drop the link count as done by d_tmpfile(), complete the inode setup
+ * and flag it as linkable.
+ */
+ drop_nlink(VFS_I(tmpfile));
xfs_finish_inode_setup(tmpfile);
VFS_I(tmpfile)->i_state |= I_LINKABLE;
@@ -3068,7 +3083,7 @@
* intermediate state on disk.
*/
if (wip) {
- ASSERT(wip->i_d.di_nlink == 0);
+ ASSERT(VFS_I(wip)->i_nlink == 0 && wip->i_d.di_nlink == 0);
error = xfs_bumplink(tp, wip);
if (error)
goto out_bmap_cancel;
diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index e440aed..2923419 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -850,7 +850,11 @@
* to hope that the caller sees ENOMEM and retries the truncate
* operation.
*/
- error = block_truncate_page(inode->i_mapping, newsize, xfs_get_blocks);
+ if (IS_DAX(inode))
+ error = dax_truncate_page(inode, newsize, xfs_get_blocks_direct);
+ else
+ error = block_truncate_page(inode->i_mapping, newsize,
+ xfs_get_blocks);
if (error)
return error;
truncate_setsize(inode, newsize);
@@ -1187,22 +1191,22 @@
struct inode *inode,
struct xfs_inode *ip)
{
- if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
+ uint16_t flags = ip->i_d.di_flags;
+
+ inode->i_flags &= ~(S_IMMUTABLE | S_APPEND | S_SYNC |
+ S_NOATIME | S_DAX);
+
+ if (flags & XFS_DIFLAG_IMMUTABLE)
inode->i_flags |= S_IMMUTABLE;
- else
- inode->i_flags &= ~S_IMMUTABLE;
- if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
+ if (flags & XFS_DIFLAG_APPEND)
inode->i_flags |= S_APPEND;
- else
- inode->i_flags &= ~S_APPEND;
- if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
+ if (flags & XFS_DIFLAG_SYNC)
inode->i_flags |= S_SYNC;
- else
- inode->i_flags &= ~S_SYNC;
- if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
+ if (flags & XFS_DIFLAG_NOATIME)
inode->i_flags |= S_NOATIME;
- else
- inode->i_flags &= ~S_NOATIME;
+ /* XXX: Also needs an on-disk per inode flag! */
+ if (ip->i_mount->m_flags & XFS_MOUNT_DAX)
+ inode->i_flags |= S_DAX;
}
/*
diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c
index 8042989..f41b0c3 100644
--- a/fs/xfs/xfs_itable.c
+++ b/fs/xfs/xfs_itable.c
@@ -252,7 +252,7 @@
}
irec->ir_free |= xfs_inobt_maskn(0, idx);
- *icount = XFS_INODES_PER_CHUNK - irec->ir_freecount;
+ *icount = irec->ir_count - irec->ir_freecount;
}
return 0;
@@ -415,6 +415,8 @@
goto del_cursor;
if (icount) {
irbp->ir_startino = r.ir_startino;
+ irbp->ir_holemask = r.ir_holemask;
+ irbp->ir_count = r.ir_count;
irbp->ir_freecount = r.ir_freecount;
irbp->ir_free = r.ir_free;
irbp++;
@@ -447,13 +449,15 @@
* If this chunk has any allocated inodes, save it.
* Also start read-ahead now for this chunk.
*/
- if (r.ir_freecount < XFS_INODES_PER_CHUNK) {
+ if (r.ir_freecount < r.ir_count) {
xfs_bulkstat_ichunk_ra(mp, agno, &r);
irbp->ir_startino = r.ir_startino;
+ irbp->ir_holemask = r.ir_holemask;
+ irbp->ir_count = r.ir_count;
irbp->ir_freecount = r.ir_freecount;
irbp->ir_free = r.ir_free;
irbp++;
- icount += XFS_INODES_PER_CHUNK - r.ir_freecount;
+ icount += r.ir_count - r.ir_freecount;
}
error = xfs_btree_increment(cur, 0, &stat);
if (error || stat == 0) {
@@ -599,8 +603,7 @@
agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1;
buffer[bufidx].xi_startino =
XFS_AGINO_TO_INO(mp, agno, r.ir_startino);
- buffer[bufidx].xi_alloccount =
- XFS_INODES_PER_CHUNK - r.ir_freecount;
+ buffer[bufidx].xi_alloccount = r.ir_count - r.ir_freecount;
buffer[bufidx].xi_allocmask = ~r.ir_free;
if (++bufidx == bcount) {
long written;
diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index 599de72..299fbaf 100644
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -3068,12 +3068,22 @@
return -EINVAL;
}
- /* existing allocation is fixed value */
- ASSERT(count == mp->m_ialloc_inos);
- ASSERT(length == mp->m_ialloc_blks);
- if (count != mp->m_ialloc_inos ||
- length != mp->m_ialloc_blks) {
- xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad count 2");
+ /*
+ * The inode chunk is either full or sparse and we only support
+ * m_ialloc_min_blks sized sparse allocations at this time.
+ */
+ if (length != mp->m_ialloc_blks &&
+ length != mp->m_ialloc_min_blks) {
+ xfs_warn(log->l_mp,
+ "%s: unsupported chunk length", __FUNCTION__);
+ return -EINVAL;
+ }
+
+ /* verify inode count is consistent with extent length */
+ if ((count >> mp->m_sb.sb_inopblog) != length) {
+ xfs_warn(log->l_mp,
+ "%s: inconsistent inode count and chunk length",
+ __FUNCTION__);
return -EINVAL;
}
@@ -3091,8 +3101,8 @@
XFS_AGB_TO_DADDR(mp, agno, agbno), length, 0))
return 0;
- xfs_ialloc_inode_init(mp, NULL, buffer_list, agno, agbno, length,
- be32_to_cpu(icl->icl_gen));
+ xfs_ialloc_inode_init(mp, NULL, buffer_list, count, agno, agbno, length,
+ be32_to_cpu(icl->icl_gen));
return 0;
}
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index 2ce7ee3..461e791 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -725,6 +725,22 @@
}
/*
+ * If enabled, sparse inode chunk alignment is expected to match the
+ * cluster size. Full inode chunk alignment must match the chunk size,
+ * but that is checked on sb read verification...
+ */
+ if (xfs_sb_version_hassparseinodes(&mp->m_sb) &&
+ mp->m_sb.sb_spino_align !=
+ XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size)) {
+ xfs_warn(mp,
+ "Sparse inode block alignment (%u) must match cluster size (%llu).",
+ mp->m_sb.sb_spino_align,
+ XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size));
+ error = -EINVAL;
+ goto out_remove_uuid;
+ }
+
+ /*
* Set inode alignment fields
*/
xfs_set_inoalignment(mp);
@@ -1084,14 +1100,18 @@
return xfs_sync_sb(mp, true);
}
+/*
+ * Deltas for the inode count are +/-64, hence we use a large batch size
+ * of 128 so we don't need to take the counter lock on every update.
+ */
+#define XFS_ICOUNT_BATCH 128
int
xfs_mod_icount(
struct xfs_mount *mp,
int64_t delta)
{
- /* deltas are +/-64, hence the large batch size of 128. */
- __percpu_counter_add(&mp->m_icount, delta, 128);
- if (percpu_counter_compare(&mp->m_icount, 0) < 0) {
+ __percpu_counter_add(&mp->m_icount, delta, XFS_ICOUNT_BATCH);
+ if (__percpu_counter_compare(&mp->m_icount, 0, XFS_ICOUNT_BATCH) < 0) {
ASSERT(0);
percpu_counter_add(&mp->m_icount, -delta);
return -EINVAL;
@@ -1113,6 +1133,14 @@
return 0;
}
+/*
+ * Deltas for the block count can vary from 1 to very large, but lock contention
+ * only occurs on frequent small block count updates such as in the delayed
+ * allocation path for buffered writes (page a time updates). Hence we set
+ * a large batch count (1024) to minimise global counter updates except when
+ * we get near to ENOSPC and we have to be very accurate with our updates.
+ */
+#define XFS_FDBLOCKS_BATCH 1024
int
xfs_mod_fdblocks(
struct xfs_mount *mp,
@@ -1151,25 +1179,19 @@
* Taking blocks away, need to be more accurate the closer we
* are to zero.
*
- * batch size is set to a maximum of 1024 blocks - if we are
- * allocating of freeing extents larger than this then we aren't
- * going to be hammering the counter lock so a lock per update
- * is not a problem.
- *
* If the counter has a value of less than 2 * max batch size,
* then make everything serialise as we are real close to
* ENOSPC.
*/
-#define __BATCH 1024
- if (percpu_counter_compare(&mp->m_fdblocks, 2 * __BATCH) < 0)
+ if (__percpu_counter_compare(&mp->m_fdblocks, 2 * XFS_FDBLOCKS_BATCH,
+ XFS_FDBLOCKS_BATCH) < 0)
batch = 1;
else
- batch = __BATCH;
-#undef __BATCH
+ batch = XFS_FDBLOCKS_BATCH;
__percpu_counter_add(&mp->m_fdblocks, delta, batch);
- if (percpu_counter_compare(&mp->m_fdblocks,
- XFS_ALLOC_SET_ASIDE(mp)) >= 0) {
+ if (__percpu_counter_compare(&mp->m_fdblocks, XFS_ALLOC_SET_ASIDE(mp),
+ XFS_FDBLOCKS_BATCH) >= 0) {
/* we had space! */
return 0;
}
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 8c995a2..7999e91 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -101,6 +101,8 @@
__uint64_t m_flags; /* global mount flags */
int m_ialloc_inos; /* inodes in inode allocation */
int m_ialloc_blks; /* blocks in inode allocation */
+ int m_ialloc_min_blks;/* min blocks in sparse inode
+ * allocation */
int m_inoalign_mask;/* mask sb_inoalignmt if used */
uint m_qflags; /* quota status flags */
struct xfs_trans_resv m_resv; /* precomputed res values */
@@ -179,6 +181,8 @@
allocator */
#define XFS_MOUNT_NOATTR2 (1ULL << 25) /* disable use of attr2 format */
+#define XFS_MOUNT_DAX (1ULL << 62) /* TEST ONLY! */
+
/*
* Default minimum read and write sizes.
diff --git a/fs/xfs/xfs_quota.h b/fs/xfs/xfs_quota.h
index 5376dd4..ce6506a 100644
--- a/fs/xfs/xfs_quota.h
+++ b/fs/xfs/xfs_quota.h
@@ -55,7 +55,6 @@
typedef struct xfs_dqtrx {
struct xfs_dquot *qt_dquot; /* the dquot this refers to */
ulong qt_blk_res; /* blks reserved on a dquot */
- ulong qt_blk_res_used; /* blks used from the reservation */
ulong qt_ino_res; /* inode reserved on a dquot */
ulong qt_ino_res_used; /* inodes used from the reservation */
long qt_bcount_delta; /* dquot blk count changes */
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 858e1e6..1fb16562 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -112,6 +112,8 @@
#define MNTOPT_DISCARD "discard" /* Discard unused blocks */
#define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
+#define MNTOPT_DAX "dax" /* Enable direct access to bdev pages */
+
/*
* Table driven mount option parser.
*
@@ -363,6 +365,10 @@
mp->m_flags |= XFS_MOUNT_DISCARD;
} else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
mp->m_flags &= ~XFS_MOUNT_DISCARD;
+#ifdef CONFIG_FS_DAX
+ } else if (!strcmp(this_char, MNTOPT_DAX)) {
+ mp->m_flags |= XFS_MOUNT_DAX;
+#endif
} else {
xfs_warn(mp, "unknown mount option [%s].", this_char);
return -EINVAL;
@@ -452,8 +458,8 @@
}
struct proc_xfs_info {
- int flag;
- char *str;
+ uint64_t flag;
+ char *str;
};
STATIC int
@@ -474,6 +480,7 @@
{ XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
{ XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
{ XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
+ { XFS_MOUNT_DAX, "," MNTOPT_DAX },
{ 0, NULL }
};
static struct proc_xfs_info xfs_info_unset[] = {
@@ -1507,6 +1514,20 @@
if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
sb->s_flags |= MS_I_VERSION;
+ if (mp->m_flags & XFS_MOUNT_DAX) {
+ xfs_warn(mp,
+ "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
+ if (sb->s_blocksize != PAGE_SIZE) {
+ xfs_alert(mp,
+ "Filesystem block size invalid for DAX Turning DAX off.");
+ mp->m_flags &= ~XFS_MOUNT_DAX;
+ } else if (!sb->s_bdev->bd_disk->fops->direct_access) {
+ xfs_alert(mp,
+ "Block device does not support DAX Turning DAX off.");
+ mp->m_flags &= ~XFS_MOUNT_DAX;
+ }
+ }
+
error = xfs_mountfs(mp);
if (error)
goto out_filestream_unmount;
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 615781b..8d916d3 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -738,6 +738,53 @@
__entry->blocks, __entry->shift, __entry->writeio_blocks)
)
+TRACE_EVENT(xfs_irec_merge_pre,
+ TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agino_t agino,
+ uint16_t holemask, xfs_agino_t nagino, uint16_t nholemask),
+ TP_ARGS(mp, agno, agino, holemask, nagino, nholemask),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(xfs_agnumber_t, agno)
+ __field(xfs_agino_t, agino)
+ __field(uint16_t, holemask)
+ __field(xfs_agino_t, nagino)
+ __field(uint16_t, nholemask)
+ ),
+ TP_fast_assign(
+ __entry->dev = mp->m_super->s_dev;
+ __entry->agno = agno;
+ __entry->agino = agino;
+ __entry->holemask = holemask;
+ __entry->nagino = nagino;
+ __entry->nholemask = holemask;
+ ),
+ TP_printk("dev %d:%d agno %d inobt (%u:0x%x) new (%u:0x%x)",
+ MAJOR(__entry->dev), MINOR(__entry->dev), __entry->agno,
+ __entry->agino, __entry->holemask, __entry->nagino,
+ __entry->nholemask)
+)
+
+TRACE_EVENT(xfs_irec_merge_post,
+ TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agino_t agino,
+ uint16_t holemask),
+ TP_ARGS(mp, agno, agino, holemask),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(xfs_agnumber_t, agno)
+ __field(xfs_agino_t, agino)
+ __field(uint16_t, holemask)
+ ),
+ TP_fast_assign(
+ __entry->dev = mp->m_super->s_dev;
+ __entry->agno = agno;
+ __entry->agino = agino;
+ __entry->holemask = holemask;
+ ),
+ TP_printk("dev %d:%d agno %d inobt (%u:0x%x)", MAJOR(__entry->dev),
+ MINOR(__entry->dev), __entry->agno, __entry->agino,
+ __entry->holemask)
+)
+
#define DEFINE_IREF_EVENT(name) \
DEFINE_EVENT(xfs_iref_class, name, \
TP_PROTO(struct xfs_inode *ip, unsigned long caller_ip), \
diff --git a/fs/xfs/xfs_trans_dquot.c b/fs/xfs/xfs_trans_dquot.c
index 76a16df..ce78534 100644
--- a/fs/xfs/xfs_trans_dquot.c
+++ b/fs/xfs/xfs_trans_dquot.c
@@ -90,8 +90,9 @@
xfs_trans_t *ntp)
{
xfs_dqtrx_t *oq, *nq;
- int i,j;
+ int i, j;
xfs_dqtrx_t *oqa, *nqa;
+ ulong blk_res_used;
if (!otp->t_dqinfo)
return;
@@ -102,18 +103,23 @@
* Because the quota blk reservation is carried forward,
* it is also necessary to carry forward the DQ_DIRTY flag.
*/
- if(otp->t_flags & XFS_TRANS_DQ_DIRTY)
+ if (otp->t_flags & XFS_TRANS_DQ_DIRTY)
ntp->t_flags |= XFS_TRANS_DQ_DIRTY;
for (j = 0; j < XFS_QM_TRANS_DQTYPES; j++) {
oqa = otp->t_dqinfo->dqs[j];
nqa = ntp->t_dqinfo->dqs[j];
for (i = 0; i < XFS_QM_TRANS_MAXDQS; i++) {
+ blk_res_used = 0;
+
if (oqa[i].qt_dquot == NULL)
break;
oq = &oqa[i];
nq = &nqa[i];
+ if (oq->qt_blk_res && oq->qt_bcount_delta > 0)
+ blk_res_used = oq->qt_bcount_delta;
+
nq->qt_dquot = oq->qt_dquot;
nq->qt_bcount_delta = nq->qt_icount_delta = 0;
nq->qt_rtbcount_delta = 0;
@@ -121,8 +127,8 @@
/*
* Transfer whatever is left of the reservations.
*/
- nq->qt_blk_res = oq->qt_blk_res - oq->qt_blk_res_used;
- oq->qt_blk_res = oq->qt_blk_res_used;
+ nq->qt_blk_res = oq->qt_blk_res - blk_res_used;
+ oq->qt_blk_res = blk_res_used;
nq->qt_rtblk_res = oq->qt_rtblk_res -
oq->qt_rtblk_res_used;
@@ -239,10 +245,6 @@
* disk blocks used.
*/
case XFS_TRANS_DQ_BCOUNT:
- if (qtrx->qt_blk_res && delta > 0) {
- qtrx->qt_blk_res_used += (ulong)delta;
- ASSERT(qtrx->qt_blk_res >= qtrx->qt_blk_res_used);
- }
qtrx->qt_bcount_delta += delta;
break;
@@ -423,15 +425,19 @@
* reservation that a transaction structure knows of.
*/
if (qtrx->qt_blk_res != 0) {
- if (qtrx->qt_blk_res != qtrx->qt_blk_res_used) {
- if (qtrx->qt_blk_res >
- qtrx->qt_blk_res_used)
+ ulong blk_res_used = 0;
+
+ if (qtrx->qt_bcount_delta > 0)
+ blk_res_used = qtrx->qt_bcount_delta;
+
+ if (qtrx->qt_blk_res != blk_res_used) {
+ if (qtrx->qt_blk_res > blk_res_used)
dqp->q_res_bcount -= (xfs_qcnt_t)
(qtrx->qt_blk_res -
- qtrx->qt_blk_res_used);
+ blk_res_used);
else
dqp->q_res_bcount -= (xfs_qcnt_t)
- (qtrx->qt_blk_res_used -
+ (blk_res_used -
qtrx->qt_blk_res);
}
} else {
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 35ec87e..5784377 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -70,6 +70,7 @@
struct buffer_head *bh_result, int create);
typedef void (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
ssize_t bytes, void *private);
+typedef void (dax_iodone_t)(struct buffer_head *bh_map, int uptodate);
#define MAY_EXEC 0x00000001
#define MAY_WRITE 0x00000002
@@ -2627,9 +2628,13 @@
int dax_clear_blocks(struct inode *, sector_t block, long size);
int dax_zero_page_range(struct inode *, loff_t from, unsigned len, get_block_t);
int dax_truncate_page(struct inode *, loff_t from, get_block_t);
-int dax_fault(struct vm_area_struct *, struct vm_fault *, get_block_t);
+int dax_fault(struct vm_area_struct *, struct vm_fault *, get_block_t,
+ dax_iodone_t);
+int __dax_fault(struct vm_area_struct *, struct vm_fault *, get_block_t,
+ dax_iodone_t);
int dax_pfn_mkwrite(struct vm_area_struct *, struct vm_fault *);
-#define dax_mkwrite(vma, vmf, gb) dax_fault(vma, vmf, gb)
+#define dax_mkwrite(vma, vmf, gb, iod) dax_fault(vma, vmf, gb, iod)
+#define __dax_mkwrite(vma, vmf, gb, iod) __dax_fault(vma, vmf, gb, iod)
#ifdef CONFIG_BLOCK
typedef void (dio_submit_t)(int rw, struct bio *bio, struct inode *inode,
diff --git a/include/linux/percpu_counter.h b/include/linux/percpu_counter.h
index 50e5009..84a1094 100644
--- a/include/linux/percpu_counter.h
+++ b/include/linux/percpu_counter.h
@@ -41,7 +41,12 @@
void percpu_counter_set(struct percpu_counter *fbc, s64 amount);
void __percpu_counter_add(struct percpu_counter *fbc, s64 amount, s32 batch);
s64 __percpu_counter_sum(struct percpu_counter *fbc);
-int percpu_counter_compare(struct percpu_counter *fbc, s64 rhs);
+int __percpu_counter_compare(struct percpu_counter *fbc, s64 rhs, s32 batch);
+
+static inline int percpu_counter_compare(struct percpu_counter *fbc, s64 rhs)
+{
+ return __percpu_counter_compare(fbc, rhs, percpu_counter_batch);
+}
static inline void percpu_counter_add(struct percpu_counter *fbc, s64 amount)
{
@@ -116,6 +121,12 @@
return 0;
}
+static inline int
+__percpu_counter_compare(struct percpu_counter *fbc, s64 rhs, s32 batch)
+{
+ return percpu_counter_compare(fbc, rhs);
+}
+
static inline void
percpu_counter_add(struct percpu_counter *fbc, s64 amount)
{
diff --git a/lib/percpu_counter.c b/lib/percpu_counter.c
index 48144cd..f051d69 100644
--- a/lib/percpu_counter.c
+++ b/lib/percpu_counter.c
@@ -197,13 +197,13 @@
* Compare counter against given value.
* Return 1 if greater, 0 if equal and -1 if less
*/
-int percpu_counter_compare(struct percpu_counter *fbc, s64 rhs)
+int __percpu_counter_compare(struct percpu_counter *fbc, s64 rhs, s32 batch)
{
s64 count;
count = percpu_counter_read(fbc);
/* Check to see if rough count will be sufficient for comparison */
- if (abs(count - rhs) > (percpu_counter_batch*num_online_cpus())) {
+ if (abs(count - rhs) > (batch * num_online_cpus())) {
if (count > rhs)
return 1;
else
@@ -218,7 +218,7 @@
else
return 0;
}
-EXPORT_SYMBOL(percpu_counter_compare);
+EXPORT_SYMBOL(__percpu_counter_compare);
static int __init percpu_counter_startup(void)
{