[Devel] [PATCH hci-8.1 v7 5/6] dm-qcow2: keep the dirty bit of an image opened dirty

[Alexander Ivanov]

With kernel_sets_dirty_bit, postsuspend clears the dirty bit
unconditionally. An image that is dirty at open, e.g. after a crash,
has unrepaired refcounts, which may be below real usage, yet it is
marked clean on disk after the first suspend. A userspace check before
the next open then has no reason to repair it.

Clear the bit only if the image was clean before the kernel set it.
The parsed header can't tell this: it is reread on every preresume and
may carry the kernel's own bit, e.g. on the lower image of a backward
merge cancelled by suspend in waiting_completion or of a merge with
depth > 0. So track two flags per image: kernel_dirty_bit is set only
while the last header write was a successful set and the reread header
still has the bit, opened_dirty is set when the header has the bit and
kernel_dirty_bit is not set. Clearing the bit of an image opened dirty
is left to userspace check.

After a failed write the bit found on disk is taken as opened dirty,
which costs an extra check, rather than as the kernel's, which could
clear a bit set by userspace while the device was suspended.

This only preserves the on-disk marker. The driver still accepts a
dirty image read-write and allocates from its refcounts, so
find_unused_cluster() may hand out clusters that L2 still references
until the image is checked.

https://virtuozzo.atlassian.net/browse/VSTOR-148171
Feature: dm-qcow2: block device over QCOW2 files driver
Signed-off-by: Alexander Ivanov <alexander.ivanov at virtuozzo.com>
---
 drivers/md/dm-qcow2-target.c | 15 +++++++++++++--
 drivers/md/dm-qcow2.h        |  2 ++
 2 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/drivers/md/dm-qcow2-target.c b/drivers/md/dm-qcow2-target.c
index b667adb5472a..2d69e39a3530 100644
--- a/drivers/md/dm-qcow2-target.c
+++ b/drivers/md/dm-qcow2-target.c
@@ -659,6 +659,7 @@ int qcow2_set_image_file_features(struct qcow2 *qcow2, bool dirty)
 	struct QCowHeader *raw_hdr;
 	struct md_page *md;
 	bool is_ro;
+	int ret;
 
 	if (qcow2->hdr.version ==  2)
 		return 0;
@@ -676,12 +677,15 @@ int qcow2_set_image_file_features(struct qcow2 *qcow2, bool dirty)
 	if (kernel_sets_dirty_bit) {
 		if (dirty)
 			raw_hdr->incompatible_features |= dirty_mask;
-		else
+		else if (!qcow2->opened_dirty)
 			raw_hdr->incompatible_features &= ~dirty_mask;
 	}
 	kunmap(md->page);
 
-	return qcow2_rw_page_sync(WRITE, qcow2, md->id, md->page);
+	ret = qcow2_rw_page_sync(WRITE, qcow2, md->id, md->page);
+	if (kernel_sets_dirty_bit)
+		qcow2->kernel_dirty_bit = dirty && !ret;
+	return ret;
 }
 ALLOW_ERROR_INJECTION(qcow2_set_image_file_features, ERRNO);
 
@@ -903,6 +907,13 @@ static int qcow2_parse_header(struct dm_target *ti, struct qcow2 *qcow2,
 		goto out;
 
 	qcow2_calc_cached_parameters(qcow2, hdr);
+	/* The header is reread on resume and may carry our own dirty bit */
+	if (!(hdr->incompatible_features & INCOMPATIBLE_FEATURES_DIRTY_BIT)) {
+		qcow2->opened_dirty = false;
+		qcow2->kernel_dirty_bit = false;
+	} else if (!qcow2->kernel_dirty_bit) {
+		qcow2->opened_dirty = true;
+	}
 	ret = -EOPNOTSUPP;
 	if (qcow2->clu_size < PAGE_SIZE ||
 	    (qcow2->ext_l2 && qcow2->clu_size < PAGE_SIZE * 32))
diff --git a/drivers/md/dm-qcow2.h b/drivers/md/dm-qcow2.h
index 69d3ae6fca8a..a2d951ee5541 100644
--- a/drivers/md/dm-qcow2.h
+++ b/drivers/md/dm-qcow2.h
@@ -257,6 +257,8 @@ struct qcow2 {
 	u32 refblock_bits;
 	u32 refblock_entries;
 	bool ext_l2;
+	bool opened_dirty; /* Dirty bit was set before we set it */
+	bool kernel_dirty_bit; /* We set the dirty bit on disk */
 
 	bool pause_submitting_qios; /* This is used only on top qcow2 image */
 	bool backward_merge_in_process;
-- 
2.53.0