@@ -5148,9 +5148,13 @@ xfs_bmap_del_extent_real(
* the same order of operations as the data device, which is:
* Remove the file mapping, remove the reverse mapping, and
* then free the blocks. This means that we must delay the
- * freeing until after we've scheduled the rmap update.
+ * freeing until after we've scheduled the rmap update. If
+ * realtime reflink is enabled, use deferred refcount intent
+ * items to decide what to do with the extent, just like we do
+ * for the data device.
*/
- if (want_free && !xfs_has_rtrmapbt(mp)) {
+ if (want_free && !xfs_has_rtrmapbt(mp) &&
+ !xfs_has_rtreflink(mp)) {
error = xfs_rtfree_blocks(tp, del->br_startblock,
del->br_blockcount);
if (error)
@@ -1011,6 +1011,7 @@ enum xfs_dinode_fmt {
XFS_DINODE_FMT_BTREE, /* struct xfs_bmdr_block */
XFS_DINODE_FMT_UUID, /* added long ago, but never used */
XFS_DINODE_FMT_RMAP, /* reverse mapping btree */
+ XFS_DINODE_FMT_REFCOUNT, /* reference count btree */
};
#define XFS_INODE_FORMAT_STR \
@@ -1019,7 +1020,8 @@ enum xfs_dinode_fmt {
{ XFS_DINODE_FMT_EXTENTS, "extent" }, \
{ XFS_DINODE_FMT_BTREE, "btree" }, \
{ XFS_DINODE_FMT_UUID, "uuid" }, \
- { XFS_DINODE_FMT_RMAP, "rmap" }
+ { XFS_DINODE_FMT_RMAP, "rmap" }, \
+ { XFS_DINODE_FMT_REFCOUNT, "refcount" }
/*
* Max values for extnum and aextnum.
@@ -414,6 +414,12 @@ xfs_dinode_verify_fork(
if (!(dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_METADATA)))
return __this_address;
break;
+ case XFS_DINODE_FMT_REFCOUNT:
+ if (!xfs_has_rtreflink(mp))
+ return __this_address;
+ if (!(dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_METADATA)))
+ return __this_address;
+ break;
default:
return __this_address;
}
@@ -266,6 +266,11 @@ xfs_iformat_data_fork(
return -EFSCORRUPTED;
}
return xfs_iformat_rtrmap(ip, dip);
+ case XFS_DINODE_FMT_REFCOUNT:
+ if (!xfs_has_rtreflink(ip->i_mount))
+ return -EFSCORRUPTED;
+ ASSERT(0); /* to be implemented later */
+ return -EFSCORRUPTED;
default:
xfs_inode_verifier_error(ip, -EFSCORRUPTED, __func__,
dip, sizeof(*dip), __this_address);
@@ -652,6 +657,10 @@ xfs_iflush_fork(
xfs_iflush_rtrmap(ip, dip);
break;
+ case XFS_DINODE_FMT_REFCOUNT:
+ ASSERT(0); /* to be implemented later */
+ break;
+
default:
ASSERT(0);
break;
@@ -23,6 +23,9 @@ struct xfs_rtgroup {
/* reverse mapping btree inode */
struct xfs_inode *rtg_rmapip;
+ /* refcount btree inode */
+ struct xfs_inode *rtg_refcountip;
+
/* Number of blocks in this group */
xfs_rgblock_t rtg_blockcount;
@@ -26,6 +26,7 @@
#include "xfs_extent_busy.h"
#include "xfs_rtgroup.h"
#include "xfs_rtbitmap.h"
+#include "xfs_imeta.h"
static struct kmem_cache *xfs_rtrefcountbt_cur_cache;
@@ -354,6 +355,7 @@ xfs_rtrefcountbt_commit_staged_btree(
int flags = XFS_ILOG_CORE | XFS_ILOG_DBROOT;
ASSERT(cur->bc_flags & XFS_BTREE_STAGING);
+ ASSERT(ifake->if_fork->if_format == XFS_DINODE_FMT_REFCOUNT);
/*
* Free any resources hanging off the real fork, then shallow-copy the
@@ -457,3 +459,34 @@ xfs_rtrefcountbt_compute_maxlevels(
/* Add one level to handle the inode root level. */
mp->m_rtrefc_maxlevels = min(d_maxlevels, r_maxlevels) + 1;
}
+
+#define XFS_RTREFC_NAMELEN 21
+
+/* Create the metadata directory path for an rtrefcount btree inode. */
+int
+xfs_rtrefcountbt_create_path(
+ struct xfs_mount *mp,
+ xfs_rgnumber_t rgno,
+ struct xfs_imeta_path **pathp)
+{
+ struct xfs_imeta_path *path;
+ char *fname;
+ int error;
+
+ error = xfs_imeta_create_file_path(mp, 2, &path);
+ if (error)
+ return error;
+
+ fname = kmalloc(XFS_RTREFC_NAMELEN, GFP_KERNEL);
+ if (!fname) {
+ xfs_imeta_free_path(path);
+ return -ENOMEM;
+ }
+
+ snprintf(fname, XFS_RTREFC_NAMELEN, "%u.refcount", rgno);
+ path->im_path[0] = "realtime";
+ path->im_path[1] = fname;
+ path->im_dynamicmask = 0x2;
+ *pathp = path;
+ return 0;
+}
@@ -11,6 +11,7 @@ struct xfs_btree_cur;
struct xfs_mount;
struct xbtree_ifakeroot;
struct xfs_rtgroup;
+struct xfs_imeta_path;
/* refcounts only exist on crc enabled filesystems */
#define XFS_RTREFCOUNT_BLOCK_LEN XFS_BTREE_LBLOCK_CRC_LEN
@@ -68,4 +69,7 @@ unsigned int xfs_rtrefcountbt_maxlevels_ondisk(void);
int __init xfs_rtrefcountbt_init_cur_cache(void);
void xfs_rtrefcountbt_destroy_cur_cache(void);
+int xfs_rtrefcountbt_create_path(struct xfs_mount *mp, xfs_rgnumber_t rgno,
+ struct xfs_imeta_path **pathp);
+
#endif /* __XFS_RTREFCOUNT_BTREE_H__ */
@@ -2580,6 +2580,14 @@ xfs_iflush(
__func__, ip->i_ino, ip);
goto flush_out;
}
+ } else if (ip->i_df.if_format == XFS_DINODE_FMT_REFCOUNT) {
+ if (!S_ISREG(VFS_I(ip)->i_mode) ||
+ !(ip->i_diflags2 & XFS_DIFLAG2_METADATA)) {
+ xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
+ "%s: Bad rt refcountbt inode %Lu, ptr "PTR_FMT,
+ __func__, ip->i_ino, ip);
+ goto flush_out;
+ }
} else if (S_ISREG(VFS_I(ip)->i_mode)) {
if (XFS_TEST_ERROR(
ip->i_df.if_format != XFS_DINODE_FMT_EXTENTS &&
@@ -2626,6 +2634,11 @@ xfs_iflush(
"%s: rt rmapbt in inode %Lu attr fork, ptr "PTR_FMT,
__func__, ip->i_ino, ip);
goto flush_out;
+ } else if (ip->i_af.if_format == XFS_DINODE_FMT_REFCOUNT) {
+ xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
+ "%s: rt refcountbt in inode %Lu attr fork, ptr "PTR_FMT,
+ __func__, ip->i_ino, ip);
+ goto flush_out;
}
}
@@ -63,6 +63,7 @@ xfs_inode_item_data_fork_size(
break;
case XFS_DINODE_FMT_BTREE:
case XFS_DINODE_FMT_RMAP:
+ case XFS_DINODE_FMT_REFCOUNT:
if ((iip->ili_fields & XFS_ILOG_DBROOT) &&
ip->i_df.if_broot_bytes > 0) {
*nbytes += ip->i_df.if_broot_bytes;
@@ -184,6 +185,7 @@ xfs_inode_item_format_data_fork(
break;
case XFS_DINODE_FMT_BTREE:
case XFS_DINODE_FMT_RMAP:
+ case XFS_DINODE_FMT_REFCOUNT:
iip->ili_fields &=
~(XFS_ILOG_DDATA | XFS_ILOG_DEXT | XFS_ILOG_DEV);
@@ -417,6 +417,7 @@ xlog_recover_inode_commit_pass2(
if (unlikely(S_ISREG(ldip->di_mode))) {
if ((ldip->di_format != XFS_DINODE_FMT_EXTENTS) &&
(ldip->di_format != XFS_DINODE_FMT_RMAP) &&
+ (ldip->di_format != XFS_DINODE_FMT_REFCOUNT) &&
(ldip->di_format != XFS_DINODE_FMT_BTREE)) {
XFS_CORRUPTION_ERROR(
"Bad log dinode data fork format for regular file",
@@ -31,6 +31,7 @@
#include "xfs_btree.h"
#include "xfs_rmap.h"
#include "xfs_rtrmap_btree.h"
+#include "xfs_rtrefcount_btree.h"
/*
* Realtime metadata files are not quite regular files because userspace can't
@@ -42,6 +43,7 @@
static struct lock_class_key xfs_rbmip_key;
static struct lock_class_key xfs_rsumip_key;
static struct lock_class_key xfs_rrmapip_key;
+static struct lock_class_key xfs_rrefcountip_key;
/*
* Read and return the summary information for a given extent size,
@@ -1855,6 +1857,47 @@ xfs_rtmount_iread_extents(
return error;
}
+/* Load realtime refcount btree inode. */
+STATIC int
+xfs_rtmount_refcountbt(
+ struct xfs_rtgroup *rtg)
+{
+ struct xfs_mount *mp = rtg->rtg_mount;
+ struct xfs_imeta_path *path;
+ struct xfs_inode *ip;
+ xfs_ino_t ino;
+ int error;
+
+ if (!xfs_has_rtreflink(mp))
+ return 0;
+
+ error = xfs_rtrefcountbt_create_path(mp, rtg->rtg_rgno, &path);
+ if (error)
+ return error;
+
+ error = xfs_imeta_lookup(mp, path, &ino);
+ if (error)
+ goto out_path;
+
+ error = xfs_rt_iget(mp, ino, &xfs_rrefcountip_key, &ip);
+ if (error)
+ goto out_path;
+
+ if (XFS_IS_CORRUPT(mp, ip->i_df.if_format != XFS_DINODE_FMT_REFCOUNT)) {
+ error = -EFSCORRUPTED;
+ goto out_rele;
+ }
+
+ rtg->rtg_refcountip = ip;
+ ip = NULL;
+out_rele:
+ if (ip)
+ xfs_imeta_irele(ip);
+out_path:
+ xfs_imeta_free_path(path);
+ return error;
+}
+
/*
* Get the bitmap and summary inodes and the summary cache into the mount
* structure at mount time.
@@ -1902,6 +1945,10 @@ xfs_rtmount_inodes(
xfs_rtgroup_put(rtg);
goto out_rele_rtgroup;
}
+
+ error = xfs_rtmount_refcountbt(rtg);
+ if (error)
+ goto out_rele_rtgroup;
}
xfs_alloc_rsum_cache(mp, sbp->sb_rbmblocks);
@@ -1909,6 +1956,10 @@ xfs_rtmount_inodes(
out_rele_rtgroup:
for_each_rtgroup(mp, rgno, rtg) {
+ if (rtg->rtg_refcountip)
+ xfs_imeta_irele(rtg->rtg_refcountip);
+ rtg->rtg_refcountip = NULL;
+
if (rtg->rtg_rmapip)
xfs_imeta_irele(rtg->rtg_rmapip);
rtg->rtg_rmapip = NULL;
@@ -1945,6 +1996,14 @@ xfs_rtmount_dqattach(
return error;
}
}
+
+ if (rtg->rtg_refcountip) {
+ error = xfs_qm_dqattach(rtg->rtg_refcountip);
+ if (error) {
+ xfs_rtgroup_put(rtg);
+ return error;
+ }
+ }
}
return 0;
@@ -1960,6 +2019,10 @@ xfs_rtunmount_inodes(
kmem_free(mp->m_rsum_cache);
for_each_rtgroup(mp, rgno, rtg) {
+ if (rtg->rtg_refcountip)
+ xfs_imeta_irele(rtg->rtg_refcountip);
+ rtg->rtg_refcountip = NULL;
+
if (rtg->rtg_rmapip)
xfs_imeta_irele(rtg->rtg_rmapip);
rtg->rtg_rmapip = NULL;
@@ -2239,6 +2239,7 @@ TRACE_DEFINE_ENUM(XFS_DINODE_FMT_EXTENTS);
TRACE_DEFINE_ENUM(XFS_DINODE_FMT_BTREE);
TRACE_DEFINE_ENUM(XFS_DINODE_FMT_UUID);
TRACE_DEFINE_ENUM(XFS_DINODE_FMT_RMAP);
+TRACE_DEFINE_ENUM(XFS_DINODE_FMT_REFCOUNT);
DECLARE_EVENT_CLASS(xfs_swap_extent_class,
TP_PROTO(struct xfs_inode *ip, int which),