Re: [Devel] [PATCH RHEL10 COMMIT] drivers/md/dm-qcow2: fix revert_cluster_alloc() for ext_l2 case

[Konstantin Khorenko <khorenko@virtuozzo.com>]

The commit is pushed to "branch-rh10-6.12.0-211.39.1.16.x.vz10-ovz" and will appear at git@bitbucket.org:openvz/vzkernel.git
after rh10-6.12.0-211.39.1.16.11.vz10
------>
commit 96e80433d537c82f2cfb27a3e2c335882d4e15a3
Author: Andrey Zhadchenko <andrey.zhadchenko@virtuozzo.com>
Date:   Thu Aug 27 19:06:10 2026 +0300

    drivers/md/dm-qcow2: fix revert_cluster_alloc() for ext_l2 case
    
    This function walks over all changed u64 values in md. With ext_l2
    half of them holds subcluster description. Firstly it reverts these
    values to saved pe_page, which is fine, but then it tries to revert
    r1r2 changes. It makes no sense when the value is a subcluster
    description.
    Teach the function to skip subcluster descriptions: do it based on
    a new lx_level in struct wb_desc. Add new argument to
    prepare_l_entry_update() and set it there.
    The warning also could have tripped for ext_l2 entries, so drop
    it entirely.
    The function effectively reverts not only cluster alloc, so rename
    it to revert_l_entries_update() to mimic prepare_l_entry_update().
    
    Feature: dm-qcow2: block device over QCOW2 files driver
    https://virtuozzo.atlassian.net/browse/VSTOR-139406
    Signed-off-by: Andrey Zhadchenko <andrey.zhadchenko@virtuozzo.com>
    Reviewed-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
    Reviewed-by: Vasileios Almpanis <vasileios.almpanis@virtuozzo.com>
    Reviewed-by: Konstantin Khorenko <khorenko@virtuozzo.com>
---
 drivers/md/dm-qcow2-map.c    | 28 ++++++++++++++++++----------
 drivers/md/dm-qcow2-target.c |  2 +-
 drivers/md/dm-qcow2.h        |  1 +
 3 files changed, 20 insertions(+), 11 deletions(-)

diff --git a/drivers/md/dm-qcow2-map.c b/drivers/md/dm-qcow2-map.c
index 6ead06ef7b3f9..52dc68baa49be 100644
--- a/drivers/md/dm-qcow2-map.c
+++ b/drivers/md/dm-qcow2-map.c
@@ -1297,18 +1297,19 @@ static void do_md_page_read_complete(int ret, struct qcow2 *qcow2,
 }
 
 /* Be careful with dirty_or_writeback()/etc! Check races. */
-static void revert_clusters_alloc(struct qcow2 *qcow2, struct wb_desc *wbd)
+static void revert_l_entries_update(struct qcow2 *qcow2, struct wb_desc *wbd)
 {
 	struct qcow2_map_item r1, r2;
 	struct page *pe_page;
+	bool skip_odd;
 	u64 pos, old;
 	int i, ret;
 
+	skip_odd = qcow2->ext_l2 && wbd->lx_level == L2_LEVEL;
+
 	lockdep_assert_held(&qcow2->md_pages_lock);
 	for_each_set_bit(i, wbd->changed_indexes, LX_INDEXES_PER_PAGE) {
 		pos = get_u64_from_be_page(wbd->md->page, i);
-		WARN_ON_ONCE(!(pos & ~LX_REFCOUNT_EXACTLY_ONE) ||
-			     !(pos & LX_REFCOUNT_EXACTLY_ONE));
 
 		/* Here we restore prealloced and compressed clu mappings */
 		pe_page = wbd->pe_page;
@@ -1321,6 +1322,9 @@ static void revert_clusters_alloc(struct qcow2 *qcow2, struct wb_desc *wbd)
 		}
 
 		set_u64_to_be_page(wbd->md->page, i, 0);
+		if (skip_odd && (i & 1))
+			continue; /* pos contains ext_l2 part of L2 entry */
+
 		spin_unlock(&qcow2->md_pages_lock);
 		pos &= ~LX_REFCOUNT_EXACTLY_ONE;
 
@@ -1369,7 +1373,7 @@ static void complete_wbd(struct qcow2 *qcow2, struct wb_desc *wbd)
 		unsigned long flags;
 
 		spin_lock_irqsave(&qcow2->md_pages_lock, flags);
-		revert_clusters_alloc(qcow2, wbd);
+		revert_l_entries_update(qcow2, wbd);
 		clear_writeback_status(qcow2, wbd->md, wbd->ret,
 				       &wait_list, &end_list);
 		spin_unlock_irqrestore(&qcow2->md_pages_lock, flags);
@@ -2433,7 +2437,7 @@ static loff_t allocate_cluster(struct qcow2 *qcow2, struct qio *qio,
 #define LU_IGN_CHANGED_IND	(1 << 3)
 static int prepare_l_entry_update(struct qcow2 *qcow2, struct qio *qio,
 				  struct md_page *md, u32 index_in_page,
-				  u64 *pval, u32 arg_mask)
+				  u64 *pval, u32 arg_mask, u8 lx_level)
 {
 	bool wants_pe_page = (arg_mask & LU_WANTS_PE_PAGE);
 	struct wb_desc *new_wbd = NULL;
@@ -2453,6 +2457,7 @@ static int prepare_l_entry_update(struct qcow2 *qcow2, struct qio *qio,
 		if (!new_wbd)
 			return -ENOMEM;
 		new_wbd->md = md;
+		new_wbd->lx_level = lx_level;
 	} else if (wants_pe_page && !md->wbd->pe_page) {
 		pe_page = alloc_page(GFP_NOIO|__GFP_ZERO);
 		if (!pe_page)
@@ -2515,7 +2520,8 @@ static int prepare_l1l2_allocation(struct qcow2 *qcow2, struct qio *qio,
 		/* Allocate cluster for L2 entries, and prepare L1 update */
 		ret = prepare_l_entry_update(qcow2, qio, map->l1.md,
 					     map->l1.index_in_page, &val,
-					     LU_SET_ONE_MASK|LU_WANTS_ALLOC);
+					     LU_SET_ONE_MASK | LU_WANTS_ALLOC,
+					     L1_LEVEL);
 		if (ret <= 0)
 			return ret;
 
@@ -2541,7 +2547,8 @@ static int prepare_l1l2_allocation(struct qcow2 *qcow2, struct qio *qio,
 
 		ret = prepare_l_entry_update(qcow2, qio, map->l2.md,
 					     map->l2.index_in_page,
-					     &map->data_clu_pos, arg_mask);
+					     &map->data_clu_pos, arg_mask,
+					     L2_LEVEL);
 		if (ret <= 0)
 			return ret;
 
@@ -2562,7 +2569,7 @@ static int prepare_l1l2_allocation(struct qcow2 *qcow2, struct qio *qio,
 
 	return prepare_l_entry_update(qcow2, qio, map->l2.md,
 				      map->l2.index_in_page + 1,
-				      &val, arg_mask);
+				      &val, arg_mask, L2_LEVEL);
 }
 
 /*
@@ -3979,7 +3986,7 @@ static void process_cow_indexes_write(struct qcow2 *qcow2,
 		ret = prepare_l_entry_update(qcow2, qio, lx_md,
 					     ext->lx_index_in_page,
 					     &ext->allocated_clu_pos,
-					     arg_mask);
+					     arg_mask, ext->cow_level);
 		if (ret < 0) {
 			qio->bi_status = errno_to_blk_status(ret);
 			qio_endio(qio);
@@ -3990,7 +3997,8 @@ static void process_cow_indexes_write(struct qcow2 *qcow2,
 			arg_mask &= ~LU_SET_ONE_MASK;
 			ret = prepare_l_entry_update(qcow2, qio, lx_md,
 					     ext->lx_index_in_page + 1,
-					   &ext->new_ext_l2, arg_mask);
+					     &ext->new_ext_l2, arg_mask,
+					     L2_LEVEL);
 			WARN_ON_ONCE(ret < 0);
 		}
 
diff --git a/drivers/md/dm-qcow2-target.c b/drivers/md/dm-qcow2-target.c
index 2877de6f11c12..be09615eb8e2e 100644
--- a/drivers/md/dm-qcow2-target.c
+++ b/drivers/md/dm-qcow2-target.c
@@ -174,7 +174,7 @@ void qcow2_flush_deferred_activity(struct qcow2_target *tgt, struct qcow2 *qcow2
 	int i;
 
 	/*
-	 * We need second iteration, since revert_clusters_alloc()
+	 * We need second iteration, since revert_l_entries_update()
 	 * may start timer again after failed wb.
 	 */
 	for (i = 0; i < 2; i++) {
diff --git a/drivers/md/dm-qcow2.h b/drivers/md/dm-qcow2.h
index 86f0688e7345a..aa3487007523f 100644
--- a/drivers/md/dm-qcow2.h
+++ b/drivers/md/dm-qcow2.h
@@ -118,6 +118,7 @@ struct wb_desc {
 	struct list_head dependent_list;
 	int nr_submitted;
 	bool completed;
+	u8 lx_level;
 	int ret;
 };
 
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel