xfs: debug mode forced buffered write failure

Add a DEBUG mode-only sysfs knob to enable forced buffered write
failure. An additional side effect of this mode is brute force killing
of delayed allocation blocks in the range of the write. The latter is
the prime motiviation behind this patch, as userspace test
infrastructure requires a reliable mechanism to create and split
delalloc extents without causing extent conversion.

Certain fallocate operations (i.e., zero range) were used for this in
the past, but the implementations have changed such that delalloc
extents are flushed and converted to real blocks, rendering the test
useless.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

3 files changed, 101 insertions, 11 deletions

diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index 379c089fb051..eed0bfc4443c 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -1752,6 +1752,7 @@ xfs_vm_write_failed(
 	loff_t			from = pos & (PAGE_CACHE_SIZE - 1);
 	loff_t			to = from + len;
 	struct buffer_head	*bh, *head;
+	struct xfs_mount	*mp = XFS_I(inode)->i_mount;
 
 	/*
 	 * The request pos offset might be 32 or 64 bit, this is all fine
@@ -1786,7 +1787,8 @@ xfs_vm_write_failed(
 		if (!buffer_delay(bh))
 			continue;
 
-		if (!buffer_new(bh) && block_offset < i_size_read(inode))
+		if (!xfs_mp_fail_writes(mp) && !buffer_new(bh) &&
+		    block_offset < i_size_read(inode))
 			continue;
 
 		xfs_vm_kill_delalloc_range(inode, block_offset,
@@ -1824,6 +1826,7 @@ xfs_vm_write_begin(
 	pgoff_t			index = pos >> PAGE_CACHE_SHIFT;
 	struct page		*page;
 	int			status;
+	struct xfs_mount	*mp = XFS_I(mapping->host)->i_mount;
 
 	ASSERT(len <= PAGE_CACHE_SIZE);
 
@@ -1832,6 +1835,8 @@ xfs_vm_write_begin(
 		return -ENOMEM;
 
 	status = __block_write_begin(page, pos, len, xfs_get_blocks);
+	if (xfs_mp_fail_writes(mp))
+		status = -EIO;
 	if (unlikely(status)) {
 		struct inode	*inode = mapping->host;
 		size_t		isize = i_size_read(inode);
@@ -1844,6 +1849,8 @@ xfs_vm_write_begin(
 		 * allocated in this write, not blocks that were previously
 		 * written successfully.
 		 */
+		if (xfs_mp_fail_writes(mp))
+			isize = 0;
 		if (pos + len > isize) {
 			ssize_t start = max_t(ssize_t, pos, isize);
 
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index b57098481c10..546cf46a97ce 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -147,6 +147,17 @@ typedef struct xfs_mount {
 	 * to various other kinds of pain inflicted on the pNFS server.
 	 */
 	__uint32_t		m_generation;
+
+#ifdef DEBUG
+	/*
+	 * DEBUG mode instrumentation to test and/or trigger delayed allocation
+	 * block killing in the event of failed writes. When enabled, all
+	 * buffered writes are forced to fail. All delalloc blocks in the range
+	 * of the write (including pre-existing delalloc blocks!) are tossed as
+	 * part of the write failure error handling sequence.
+	 */
+	bool			m_fail_writes;
+#endif
 } xfs_mount_t;
 
 /*
@@ -264,6 +275,20 @@ xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
 	return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
 }
 
+#ifdef DEBUG
+static inline bool
+xfs_mp_fail_writes(struct xfs_mount *mp)
+{
+	return mp->m_fail_writes;
+}
+#else
+static inline bool
+xfs_mp_fail_writes(struct xfs_mount *mp)
+{
+	return 0;
+}
+#endif
+
 /*
  * Per-ag incore structure, copies of information in agf and agi, to improve the
  * performance of allocation group selection.
diff --git a/fs/xfs/xfs_sysfs.c b/fs/xfs/xfs_sysfs.c
index 641d625eb334..6ced4f143494 100644
--- a/fs/xfs/xfs_sysfs.c
+++ b/fs/xfs/xfs_sysfs.c
@@ -18,10 +18,13 @@
 
 #include "xfs.h"
 #include "xfs_sysfs.h"
+#include "xfs_format.h"
 #include "xfs_log_format.h"
+#include "xfs_trans_resv.h"
 #include "xfs_log.h"
 #include "xfs_log_priv.h"
 #include "xfs_stats.h"
+#include "xfs_mount.h"
 
 struct xfs_sysfs_attr {
 	struct attribute attr;
@@ -45,16 +48,6 @@ to_attr(struct attribute *attr)
 
 #define ATTR_LIST(name) &xfs_sysfs_attr_##name.attr
 
-/*
- * xfs_mount kobject. This currently has no attributes and thus no need for show
- * and store helpers. The mp kobject serves as the per-mount parent object that
- * is identified by the fsname under sysfs.
- */
-
-struct kobj_type xfs_mp_ktype = {
-	.release = xfs_sysfs_release,
-};
-
 STATIC ssize_t
 xfs_sysfs_object_show(
 	struct kobject		*kobject,
@@ -83,6 +76,71 @@ static const struct sysfs_ops xfs_sysfs_ops = {
 	.store = xfs_sysfs_object_store,
 };
 
+/*
+ * xfs_mount kobject. The mp kobject also serves as the per-mount parent object
+ * that is identified by the fsname under sysfs.
+ */
+
+static inline struct xfs_mount *
+to_mp(struct kobject *kobject)
+{
+	struct xfs_kobj *kobj = to_kobj(kobject);
+
+	return container_of(kobj, struct xfs_mount, m_kobj);
+}
+
+#ifdef DEBUG
+
+STATIC ssize_t
+fail_writes_store(
+	struct kobject		*kobject,
+	const char		*buf,
+	size_t			count)
+{
+	struct xfs_mount	*mp = to_mp(kobject);
+	int			ret;
+	int			val;
+
+	ret = kstrtoint(buf, 0, &val);
+	if (ret)
+		return ret;
+
+	if (val == 1)
+		mp->m_fail_writes = true;
+	else if (val == 0)
+		mp->m_fail_writes = false;
+	else
+		return -EINVAL;
+
+	return count;
+}
+
+STATIC ssize_t
+fail_writes_show(
+	struct kobject		*kobject,
+	char			*buf)
+{
+	struct xfs_mount	*mp = to_mp(kobject);
+
+	return snprintf(buf, PAGE_SIZE, "%d\n", mp->m_fail_writes ? 1 : 0);
+}
+XFS_SYSFS_ATTR_RW(fail_writes);
+
+#endif /* DEBUG */
+
+static struct attribute *xfs_mp_attrs[] = {
+#ifdef DEBUG
+	ATTR_LIST(fail_writes),
+#endif
+	NULL,
+};
+
+struct kobj_type xfs_mp_ktype = {
+	.release = xfs_sysfs_release,
+	.sysfs_ops = &xfs_sysfs_ops,
+	.default_attrs = xfs_mp_attrs,
+};
+
 #ifdef DEBUG
 /* debug */