From 1a5b28ad386a8caeea7a458a7fbd29d685a3ecd9 Mon Sep 17 00:00:00 2001
From: Andrea Mazzoleni <amadvance@gmail.com>
Date: Wed, 23 Sep 2026 16:11:20 +0200
Subject: [PATCH] Mark scrub/sync blocks bad when data file open fails with an
 I/O error

---
 cmdline/io.h    |  1 +
 cmdline/scrub.c | 10 ++++++++++
 cmdline/sync.c  | 14 +++++++++++---
 3 files changed, 22 insertions(+), 3 deletions(-)

diff --git a/cmdline/io.h b/cmdline/io.h
index 42e7f127..df535946 100644
--- a/cmdline/io.h
+++ b/cmdline/io.h
@@ -66,6 +66,7 @@ struct snapraid_task {
 	block_off_t file_pos;
 	ssize_t read_size; /**< Size of the data read. */
 	int is_timestamp_different; /**< Report if file has a changed timestamp. */
+	int unrelated_error; /**< Error unrelated to the current parity position. */
 };
 
 /**
diff --git a/cmdline/scrub.c b/cmdline/scrub.c
index f5d23474..094210c0 100644
--- a/cmdline/scrub.c
+++ b/cmdline/scrub.c
@@ -333,6 +333,10 @@ static void scrub_data_reader(struct snapraid_worker* worker, struct snapraid_ta
 	block_off_t blockcur = task->position;
 	unsigned char* buffer = task->buffer;
 	int ret;
+
+	/* tasks are reused, so a previous close failure must not affect this position */
+	task->unrelated_error = 0;
+
 	/* if the disk position is not used */
 	if (!disk) {
 		/* use an empty block */
@@ -370,6 +374,7 @@ static void scrub_data_reader(struct snapraid_worker* worker, struct snapraid_ta
 			log_fatal_errno(errno, disk->name);
 			log_fatal(errno, "Stopping at block %" PRIu64 "\n", blockcur);
 
+			task->unrelated_error = 1;
 			if (is_hw(errno)) {
 				task->state = TASK_STATE_IOERROR;
 			} else {
@@ -676,6 +681,11 @@ static int state_scrub_process(struct snapraid_state* state, struct snapraid_par
 			if (task->state == TASK_STATE_IOERROR) {
 				/* LCOV_EXCL_START */
 				++io_error;
+				/* closing the previous file says nothing about this parity position */
+				if (!task->unrelated_error) {
+					info_set(&state->infoarr, blockcur, info_set_bad(info));
+					state->need_write = 1;
+				}
 				goto bail;
 				/* LCOV_EXCL_STOP */
 			}
diff --git a/cmdline/sync.c b/cmdline/sync.c
index 3c7450bd..c85ccaab 100644
--- a/cmdline/sync.c
+++ b/cmdline/sync.c
@@ -762,6 +762,10 @@ static void sync_data_reader(struct snapraid_worker* worker, struct snapraid_tas
 	block_off_t blockcur = task->position;
 	unsigned char* buffer = task->buffer;
 	int ret;
+
+	/* tasks are reused, so a previous close failure must not affect this position */
+	task->unrelated_error = 0;
+
 	/* if the disk position is not used */
 	if (!disk) {
 		/* use an empty block */
@@ -802,6 +806,7 @@ static void sync_data_reader(struct snapraid_worker* worker, struct snapraid_tas
 			log_fatal_errno(errno, disk->name);
 			log_fatal(errno, "Stopping at block %" PRIu64 "\n", blockcur);
 
+			task->unrelated_error = 1;
 			if (is_hw(errno)) {
 				task->state = TASK_STATE_IOERROR;
 			} else {
@@ -1192,9 +1197,12 @@ static int state_sync_process(struct snapraid_state* state, struct snapraid_pari
 			if (task->state == TASK_STATE_IOERROR) {
 				/* LCOV_EXCL_START */
 				++io_error;
-				/* preserve the failing stripe before publishing partial sync progress */
-				info_set(&state->infoarr, blockcur, info_set_bad(info));
-				state->need_write = 1;
+				/* only errors in the current file identify this stripe as bad */
+				if (!task->unrelated_error) {
+					/* preserve the failing stripe before publishing partial sync progress */
+					info_set(&state->infoarr, blockcur, info_set_bad(info));
+					state->need_write = 1;
+				}
 				goto bail_durable;
 				/* LCOV_EXCL_STOP */
 			}