From 62867a046a223e6eb771e23d2048e839c1d949d7 Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Sat, 12 Apr 2025 10:30:23 +0800
Subject: [PATCH 01/16] selftests: ublk: setup ring with
 IORING_SETUP_SINGLE_ISSUER/IORING_SETUP_DEFER_TASKRUN

It is observed that this way is more efficient for fast nvme backing
file.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250412023035.2649275-8-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 tools/testing/selftests/ublk/kublk.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/ublk/kublk.c b/tools/testing/selftests/ublk/kublk.c
index 381e31acaad9..c2acd874f9af 100644
--- a/tools/testing/selftests/ublk/kublk.c
+++ b/tools/testing/selftests/ublk/kublk.c
@@ -346,7 +346,9 @@ static int ublk_queue_init(struct ublk_queue *q)
 	}
 
 	ret = ublk_setup_ring(&q->ring, ring_depth, cq_depth,
-			IORING_SETUP_COOP_TASKRUN);
+			IORING_SETUP_COOP_TASKRUN |
+			IORING_SETUP_SINGLE_ISSUER |
+			IORING_SETUP_DEFER_TASKRUN);
 	if (ret < 0) {
 		ublk_err("ublk dev %d queue %d setup io_uring failed %d\n",
 				q->dev->dev_info.dev_id, q->q_id, ret);
-- 
2.51.0


From 2f0a692a93a585ead9ccffd0642694946d74411f Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Sat, 12 Apr 2025 10:30:24 +0800
Subject: [PATCH 02/16] selftests: ublk: set queue pthread's cpu affinity

In NUMA machine, ublk IO performance is very sensitive with queue
pthread's affinity setting.

Retrieve queue's affinity and select the 1st cpu as queue thread's sched
affinity, and it is observed that single cpu task affinity can get
stable & good performance if client application is put on proper cpu.

Dump this info when adding one ublk device. Use shmem to communicate
queue's tid between parent and daemon.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250412023035.2649275-9-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 tools/testing/selftests/ublk/kublk.c | 156 +++++++++++++++++++++++++--
 tools/testing/selftests/ublk/kublk.h |  11 +-
 2 files changed, 159 insertions(+), 8 deletions(-)

diff --git a/tools/testing/selftests/ublk/kublk.c b/tools/testing/selftests/ublk/kublk.c
index c2acd874f9af..1e21e1401a08 100644
--- a/tools/testing/selftests/ublk/kublk.c
+++ b/tools/testing/selftests/ublk/kublk.c
@@ -206,10 +206,73 @@ static const char *ublk_dev_state_desc(struct ublk_dev *dev)
 	};
 }
 
+static void ublk_print_cpu_set(const cpu_set_t *set, char *buf, unsigned len)
+{
+	unsigned done = 0;
+	int i;
+
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		if (CPU_ISSET(i, set))
+			done += snprintf(&buf[done], len - done, "%d ", i);
+	}
+}
+
+static void ublk_adjust_affinity(cpu_set_t *set)
+{
+	int j, updated = 0;
+
+	/*
+	 * Just keep the 1st CPU now.
+	 *
+	 * In future, auto affinity selection can be tried.
+	 */
+	for (j = 0; j < CPU_SETSIZE; j++) {
+		if (CPU_ISSET(j, set)) {
+			if (!updated) {
+				updated = 1;
+				continue;
+			}
+			CPU_CLR(j, set);
+		}
+	}
+}
+
+/* Caller must free the allocated buffer */
+static int ublk_ctrl_get_affinity(struct ublk_dev *ctrl_dev, cpu_set_t **ptr_buf)
+{
+	struct ublk_ctrl_cmd_data data = {
+		.cmd_op	= UBLK_U_CMD_GET_QUEUE_AFFINITY,
+		.flags	= CTRL_CMD_HAS_DATA | CTRL_CMD_HAS_BUF,
+	};
+	cpu_set_t *buf;
+	int i, ret;
+
+	buf = malloc(sizeof(cpu_set_t) * ctrl_dev->dev_info.nr_hw_queues);
+	if (!buf)
+		return -ENOMEM;
+
+	for (i = 0; i < ctrl_dev->dev_info.nr_hw_queues; i++) {
+		data.data[0] = i;
+		data.len = sizeof(cpu_set_t);
+		data.addr = (__u64)&buf[i];
+
+		ret = __ublk_ctrl_cmd(ctrl_dev, &data);
+		if (ret < 0) {
+			free(buf);
+			return ret;
+		}
+		ublk_adjust_affinity(&buf[i]);
+	}
+
+	*ptr_buf = buf;
+	return 0;
+}
+
 static void ublk_ctrl_dump(struct ublk_dev *dev)
 {
 	struct ublksrv_ctrl_dev_info *info = &dev->dev_info;
 	struct ublk_params p;
+	cpu_set_t *affinity;
 	int ret;
 
 	ret = ublk_ctrl_get_params(dev, &p);
@@ -218,12 +281,31 @@ static void ublk_ctrl_dump(struct ublk_dev *dev)
 		return;
 	}
 
+	ret = ublk_ctrl_get_affinity(dev, &affinity);
+	if (ret < 0) {
+		ublk_err("failed to get affinity %m\n");
+		return;
+	}
+
 	ublk_log("dev id %d: nr_hw_queues %d queue_depth %d block size %d dev_capacity %lld\n",
 			info->dev_id, info->nr_hw_queues, info->queue_depth,
 			1 << p.basic.logical_bs_shift, p.basic.dev_sectors);
 	ublk_log("\tmax rq size %d daemon pid %d flags 0x%llx state %s\n",
 			info->max_io_buf_bytes, info->ublksrv_pid, info->flags,
 			ublk_dev_state_desc(dev));
+
+	if (affinity) {
+		char buf[512];
+		int i;
+
+		for (i = 0; i < info->nr_hw_queues; i++) {
+			ublk_print_cpu_set(&affinity[i], buf, sizeof(buf));
+			printf("\tqueue %u: tid %d affinity(%s)\n",
+					i, dev->q[i].tid, buf);
+		}
+		free(affinity);
+	}
+
 	fflush(stdout);
 }
 
@@ -603,9 +685,24 @@ static int ublk_process_io(struct ublk_queue *q)
 	return reapped;
 }
 
+static void ublk_queue_set_sched_affinity(const struct ublk_queue *q,
+		cpu_set_t *cpuset)
+{
+        if (sched_setaffinity(0, sizeof(*cpuset), cpuset) < 0)
+                ublk_err("ublk dev %u queue %u set affinity failed",
+                                q->dev->dev_info.dev_id, q->q_id);
+}
+
+struct ublk_queue_info {
+	struct ublk_queue 	*q;
+	sem_t 			*queue_sem;
+	cpu_set_t 		*affinity;
+};
+
 static void *ublk_io_handler_fn(void *data)
 {
-	struct ublk_queue *q = data;
+	struct ublk_queue_info *info = data;
+	struct ublk_queue *q = info->q;
 	int dev_id = q->dev->dev_info.dev_id;
 	int ret;
 
@@ -615,6 +712,10 @@ static void *ublk_io_handler_fn(void *data)
 				dev_id, q->q_id);
 		return NULL;
 	}
+	/* IO perf is sensitive with queue pthread affinity on NUMA machine*/
+	ublk_queue_set_sched_affinity(q, info->affinity);
+	sem_post(info->queue_sem);
+
 	ublk_dbg(UBLK_DBG_QUEUE, "tid %d: ublk dev %d queue %d started\n",
 			q->tid, dev_id, q->q_id);
 
@@ -640,7 +741,7 @@ static void ublk_set_parameters(struct ublk_dev *dev)
 				dev->dev_info.dev_id, ret);
 }
 
-static int ublk_send_dev_event(const struct dev_ctx *ctx, int dev_id)
+static int ublk_send_dev_event(const struct dev_ctx *ctx, struct ublk_dev *dev, int dev_id)
 {
 	uint64_t id;
 	int evtfd = ctx->_evtfd;
@@ -653,10 +754,14 @@ static int ublk_send_dev_event(const struct dev_ctx *ctx, int dev_id)
 	else
 		id = ERROR_EVTFD_DEVID;
 
+	if (dev && ctx->shadow_dev)
+		memcpy(&ctx->shadow_dev->q, &dev->q, sizeof(dev->q));
+
 	if (write(evtfd, &id, sizeof(id)) != sizeof(id))
 		return -EINVAL;
 
 	close(evtfd);
+	shmdt(ctx->shadow_dev);
 
 	return 0;
 }
@@ -664,24 +769,46 @@ static int ublk_send_dev_event(const struct dev_ctx *ctx, int dev_id)
 
 static int ublk_start_daemon(const struct dev_ctx *ctx, struct ublk_dev *dev)
 {
-	int ret, i;
-	void *thread_ret;
 	const struct ublksrv_ctrl_dev_info *dinfo = &dev->dev_info;
+	struct ublk_queue_info *qinfo;
+	cpu_set_t *affinity_buf;
+	void *thread_ret;
+	sem_t queue_sem;
+	int ret, i;
 
 	ublk_dbg(UBLK_DBG_DEV, "%s enter\n", __func__);
 
+	qinfo = (struct ublk_queue_info *)calloc(sizeof(struct ublk_queue_info),
+			dinfo->nr_hw_queues);
+	if (!qinfo)
+		return -ENOMEM;
+
+	sem_init(&queue_sem, 0, 0);
 	ret = ublk_dev_prep(ctx, dev);
 	if (ret)
 		return ret;
 
+	ret = ublk_ctrl_get_affinity(dev, &affinity_buf);
+	if (ret)
+		return ret;
+
 	for (i = 0; i < dinfo->nr_hw_queues; i++) {
 		dev->q[i].dev = dev;
 		dev->q[i].q_id = i;
+
+		qinfo[i].q = &dev->q[i];
+		qinfo[i].queue_sem = &queue_sem;
+		qinfo[i].affinity = &affinity_buf[i];
 		pthread_create(&dev->q[i].thread, NULL,
 				ublk_io_handler_fn,
-				&dev->q[i]);
+				&qinfo[i]);
 	}
 
+	for (i = 0; i < dinfo->nr_hw_queues; i++)
+		sem_wait(&queue_sem);
+	free(qinfo);
+	free(affinity_buf);
+
 	/* everything is fine now, start us */
 	ublk_set_parameters(dev);
 	ret = ublk_ctrl_start_dev(dev, getpid());
@@ -694,7 +821,7 @@ static int ublk_start_daemon(const struct dev_ctx *ctx, struct ublk_dev *dev)
 	if (ctx->fg)
 		ublk_ctrl_dump(dev);
 	else
-		ublk_send_dev_event(ctx, dev->dev_info.dev_id);
+		ublk_send_dev_event(ctx, dev, dev->dev_info.dev_id);
 
 	/* wait until we are terminated */
 	for (i = 0; i < dinfo->nr_hw_queues; i++)
@@ -873,7 +1000,7 @@ static int __cmd_dev_add(const struct dev_ctx *ctx)
 
 fail:
 	if (ret < 0)
-		ublk_send_dev_event(ctx, -1);
+		ublk_send_dev_event(ctx, dev, -1);
 	ublk_ctrl_deinit(dev);
 	return ret;
 }
@@ -887,6 +1014,16 @@ static int cmd_dev_add(struct dev_ctx *ctx)
 	if (ctx->fg)
 		goto run;
 
+	ctx->_shmid = shmget(IPC_PRIVATE, sizeof(struct ublk_dev), IPC_CREAT | 0666);
+	if (ctx->_shmid < 0) {
+		ublk_err("%s: failed to shmget %s\n", __func__, strerror(errno));
+		exit(-1);
+	}
+	ctx->shadow_dev = (struct ublk_dev *)shmat(ctx->_shmid, NULL, 0);
+	if (ctx->shadow_dev == (struct ublk_dev *)-1) {
+		ublk_err("%s: failed to shmat %s\n", __func__, strerror(errno));
+		exit(-1);
+	}
 	ctx->_evtfd = eventfd(0, 0);
 	if (ctx->_evtfd < 0) {
 		ublk_err("%s: failed to create eventfd %s\n", __func__, strerror(errno));
@@ -922,6 +1059,8 @@ run:
 			if (__cmd_dev_list(ctx) >= 0)
 				exit_code = EXIT_SUCCESS;
 		}
+		shmdt(ctx->shadow_dev);
+		shmctl(ctx->_shmid, IPC_RMID, NULL);
 		/* wait for child and detach from it */
 		wait(NULL);
 		exit(exit_code);
@@ -988,6 +1127,9 @@ static int __cmd_dev_list(struct dev_ctx *ctx)
 			ublk_err("%s: can't get dev info from %d: %d\n",
 					__func__, ctx->dev_id, ret);
 	} else {
+		if (ctx->shadow_dev)
+			memcpy(&dev->q, ctx->shadow_dev->q, sizeof(dev->q));
+
 		ublk_ctrl_dump(dev);
 	}
 
diff --git a/tools/testing/selftests/ublk/kublk.h b/tools/testing/selftests/ublk/kublk.h
index eccf12360a14..85295d3e36cb 100644
--- a/tools/testing/selftests/ublk/kublk.h
+++ b/tools/testing/selftests/ublk/kublk.h
@@ -20,10 +20,15 @@
 #include <sys/wait.h>
 #include <sys/eventfd.h>
 #include <sys/uio.h>
+#include <sys/ipc.h>
+#include <sys/shm.h>
 #include <linux/io_uring.h>
 #include <liburing.h>
-#include <linux/ublk_cmd.h>
+#include <semaphore.h>
+
+/* allow ublk_dep.h to override ublk_cmd.h */
 #include "ublk_dep.h"
+#include <linux/ublk_cmd.h>
 
 #define __maybe_unused __attribute__((unused))
 #define MAX_BACK_FILES   4
@@ -74,6 +79,10 @@ struct dev_ctx {
 	unsigned int    chunk_size;
 
 	int _evtfd;
+	int _shmid;
+
+	/* built from shmem, only for ublk_dump_dev() */
+	struct ublk_dev *shadow_dev;
 };
 
 struct ublk_ctrl_cmd_data {
-- 
2.51.0


From 6c62fd04e8bfc06f37ccda0d12fd367591445954 Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Sat, 12 Apr 2025 10:30:25 +0800
Subject: [PATCH 03/16] selftests: ublk: increase max nr_queues and queue depth

Increase max nr_queues to 32, and queue depth to 1024.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250412023035.2649275-10-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 tools/testing/selftests/ublk/kublk.c | 2 +-
 tools/testing/selftests/ublk/kublk.h | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/ublk/kublk.c b/tools/testing/selftests/ublk/kublk.c
index 1e21e1401a08..5e805d358739 100644
--- a/tools/testing/selftests/ublk/kublk.c
+++ b/tools/testing/selftests/ublk/kublk.c
@@ -1203,7 +1203,7 @@ static int cmd_dev_get_features(void)
 static int cmd_dev_help(char *exe)
 {
 	printf("%s add -t [null|loop] [-q nr_queues] [-d depth] [-n dev_id] [backfile1] [backfile2] ...\n", exe);
-	printf("\t default: nr_queues=2(max 4), depth=128(max 128), dev_id=-1(auto allocation)\n");
+	printf("\t default: nr_queues=2(max 32), depth=128(max 1024), dev_id=-1(auto allocation)\n");
 	printf("%s del [-n dev_id] -a \n", exe);
 	printf("\t -a delete all devices -n delete specified device\n");
 	printf("%s list [-n dev_id] -a \n", exe);
diff --git a/tools/testing/selftests/ublk/kublk.h b/tools/testing/selftests/ublk/kublk.h
index 85295d3e36cb..9b77137b8700 100644
--- a/tools/testing/selftests/ublk/kublk.h
+++ b/tools/testing/selftests/ublk/kublk.h
@@ -50,8 +50,8 @@
 #define UBLKSRV_IO_IDLE_SECS		20
 
 #define UBLK_IO_MAX_BYTES               (1 << 20)
-#define UBLK_MAX_QUEUES                 4
-#define UBLK_QUEUE_DEPTH                128
+#define UBLK_MAX_QUEUES                 32
+#define UBLK_QUEUE_DEPTH                1024
 
 #define UBLK_DBG_DEV            (1U << 0)
 #define UBLK_DBG_QUEUE          (1U << 1)
-- 
2.51.0


From 810b88f3dcb6d04e274b37d05f421330e20a3714 Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Sat, 12 Apr 2025 10:30:26 +0800
Subject: [PATCH 04/16] selftests: ublk: support target specific command line

Support target specific command line for making related command line code
handling more readable & clean.

Also helps for adding new features.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250412023035.2649275-11-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 tools/testing/selftests/ublk/kublk.c  | 59 +++++++++++++++++++++++----
 tools/testing/selftests/ublk/kublk.h  | 20 +++++++--
 tools/testing/selftests/ublk/stripe.c | 28 ++++++++++++-
 3 files changed, 95 insertions(+), 12 deletions(-)

diff --git a/tools/testing/selftests/ublk/kublk.c b/tools/testing/selftests/ublk/kublk.c
index 5e805d358739..03b3d6427775 100644
--- a/tools/testing/selftests/ublk/kublk.c
+++ b/tools/testing/selftests/ublk/kublk.c
@@ -5,6 +5,8 @@
 
 #include "kublk.h"
 
+#define MAX_NR_TGT_ARG 	64
+
 unsigned int ublk_dbg_mask = UBLK_LOG;
 static const struct ublk_tgt_ops *tgt_ops_list[] = {
 	&null_tgt_ops,
@@ -1202,12 +1204,25 @@ static int cmd_dev_get_features(void)
 
 static int cmd_dev_help(char *exe)
 {
-	printf("%s add -t [null|loop] [-q nr_queues] [-d depth] [-n dev_id] [backfile1] [backfile2] ...\n", exe);
-	printf("\t default: nr_queues=2(max 32), depth=128(max 1024), dev_id=-1(auto allocation)\n");
+	int i;
+
+	printf("%s add -t [null|loop|stripe] [-q nr_queues] [-d depth] [-n dev_id]\n", exe);
+	printf("\t[--foreground] [--quiet] [-z] [--debug_mask mask]\n");
+	printf("\t[target options] [backfile1] [backfile2] ...\n");
+	printf("\tdefault: nr_queues=2(max 32), depth=128(max 1024), dev_id=-1(auto allocation)\n");
+
+	for (i = 0; i < sizeof(tgt_ops_list) / sizeof(tgt_ops_list[0]); i++) {
+		const struct ublk_tgt_ops *ops = tgt_ops_list[i];
+
+		if (ops->usage)
+			ops->usage(ops);
+	}
+	printf("\n");
+
 	printf("%s del [-n dev_id] -a \n", exe);
-	printf("\t -a delete all devices -n delete specified device\n");
+	printf("\t -a delete all devices -n delete specified device\n\n");
 	printf("%s list [-n dev_id] -a \n", exe);
-	printf("\t -a list all devices, -n list specified device, default -a \n");
+	printf("\t -a list all devices, -n list specified device, default -a \n\n");
 	printf("%s features\n", exe);
 	return 0;
 }
@@ -1224,9 +1239,9 @@ int main(int argc, char *argv[])
 		{ "quiet",		0,	NULL,  0  },
 		{ "zero_copy",          0,      NULL, 'z' },
 		{ "foreground",		0,	NULL,  0  },
-		{ "chunk_size", 	1,	NULL,  0  },
 		{ 0, 0, 0, 0 }
 	};
+	const struct ublk_tgt_ops *ops = NULL;
 	int option_idx, opt;
 	const char *cmd = argv[1];
 	struct dev_ctx ctx = {
@@ -1234,13 +1249,15 @@ int main(int argc, char *argv[])
 		.nr_hw_queues	=	2,
 		.dev_id		=	-1,
 		.tgt_type	=	"unknown",
-		.chunk_size 	= 	65536, 	/* def chunk size is 64K */
 	};
 	int ret = -EINVAL, i;
+	int tgt_argc = 1;
+	char *tgt_argv[MAX_NR_TGT_ARG] = { NULL };
 
 	if (argc == 1)
 		return ret;
 
+	opterr = 0;
 	optind = 2;
 	while ((opt = getopt_long(argc, argv, "t:n:d:q:az",
 				  longopts, &option_idx)) != -1) {
@@ -1271,8 +1288,26 @@ int main(int argc, char *argv[])
 				ublk_dbg_mask = 0;
 			if (!strcmp(longopts[option_idx].name, "foreground"))
 				ctx.fg = 1;
-			if (!strcmp(longopts[option_idx].name, "chunk_size"))
-				ctx.chunk_size = strtol(optarg, NULL, 10);
+			break;
+		case '?':
+			/*
+			 * target requires every option must have argument
+			 */
+			if (argv[optind][0] == '-' || argv[optind - 1][0] != '-') {
+				fprintf(stderr, "every target option requires argument: %s %s\n",
+						argv[optind - 1], argv[optind]);
+				exit(EXIT_FAILURE);
+			}
+
+			if (tgt_argc < (MAX_NR_TGT_ARG - 1) / 2) {
+				tgt_argv[tgt_argc++] = argv[optind - 1];
+				tgt_argv[tgt_argc++] = argv[optind];
+			} else {
+				fprintf(stderr, "too many target options\n");
+				exit(EXIT_FAILURE);
+			}
+			optind += 1;
+			break;
 		}
 	}
 
@@ -1281,6 +1316,14 @@ int main(int argc, char *argv[])
 		ctx.files[ctx.nr_files++] = argv[i++];
 	}
 
+	ops = ublk_find_tgt(ctx.tgt_type);
+	if (ops && ops->parse_cmd_line) {
+		optind = 0;
+
+		tgt_argv[0] = ctx.tgt_type;
+		ops->parse_cmd_line(&ctx, tgt_argc, tgt_argv);
+	}
+
 	if (!strcmp(cmd, "add"))
 		ret = cmd_dev_add(&ctx);
 	else if (!strcmp(cmd, "del"))
diff --git a/tools/testing/selftests/ublk/kublk.h b/tools/testing/selftests/ublk/kublk.h
index 9b77137b8700..7b7446359c8f 100644
--- a/tools/testing/selftests/ublk/kublk.h
+++ b/tools/testing/selftests/ublk/kublk.h
@@ -63,6 +63,11 @@
 struct ublk_dev;
 struct ublk_queue;
 
+struct stripe_ctx {
+	/* stripe */
+	unsigned int    chunk_size;
+};
+
 struct dev_ctx {
 	char tgt_type[16];
 	unsigned long flags;
@@ -75,14 +80,15 @@ struct dev_ctx {
 	unsigned int	all:1;
 	unsigned int	fg:1;
 
-	/* stripe */
-	unsigned int    chunk_size;
-
 	int _evtfd;
 	int _shmid;
 
 	/* built from shmem, only for ublk_dump_dev() */
 	struct ublk_dev *shadow_dev;
+
+	union {
+		struct stripe_ctx  stripe;
+	};
 };
 
 struct ublk_ctrl_cmd_data {
@@ -119,6 +125,14 @@ struct ublk_tgt_ops {
 	int (*queue_io)(struct ublk_queue *, int tag);
 	void (*tgt_io_done)(struct ublk_queue *,
 			int tag, const struct io_uring_cqe *);
+
+	/*
+	 * Target specific command line handling
+	 *
+	 * each option requires argument for target command line
+	 */
+	void (*parse_cmd_line)(struct dev_ctx *ctx, int argc, char *argv[]);
+	void (*usage)(const struct ublk_tgt_ops *ops);
 };
 
 struct ublk_tgt {
diff --git a/tools/testing/selftests/ublk/stripe.c b/tools/testing/selftests/ublk/stripe.c
index 179731c3dd6f..5dbd6392d83d 100644
--- a/tools/testing/selftests/ublk/stripe.c
+++ b/tools/testing/selftests/ublk/stripe.c
@@ -281,7 +281,7 @@ static int ublk_stripe_tgt_init(const struct dev_ctx *ctx, struct ublk_dev *dev)
 			.max_sectors = dev->dev_info.max_io_buf_bytes >> 9,
 		},
 	};
-	unsigned chunk_size = ctx->chunk_size;
+	unsigned chunk_size = ctx->stripe.chunk_size;
 	struct stripe_conf *conf;
 	unsigned chunk_shift;
 	loff_t bytes = 0;
@@ -344,10 +344,36 @@ static void ublk_stripe_tgt_deinit(struct ublk_dev *dev)
 	backing_file_tgt_deinit(dev);
 }
 
+static void ublk_stripe_cmd_line(struct dev_ctx *ctx, int argc, char *argv[])
+{
+	static const struct option longopts[] = {
+		{ "chunk_size", 	1,	NULL,  0  },
+		{ 0, 0, 0, 0 }
+	};
+	int option_idx, opt;
+
+	ctx->stripe.chunk_size = 65536;
+	while ((opt = getopt_long(argc, argv, "",
+				  longopts, &option_idx)) != -1) {
+		switch (opt) {
+		case 0:
+			if (!strcmp(longopts[option_idx].name, "chunk_size"))
+				ctx->stripe.chunk_size = strtol(optarg, NULL, 10);
+		}
+	}
+}
+
+static void ublk_stripe_usage(const struct ublk_tgt_ops *ops)
+{
+	printf("\tstripe: [--chunk_size chunk_size (default 65536)]\n");
+}
+
 const struct ublk_tgt_ops stripe_tgt_ops = {
 	.name = "stripe",
 	.init_tgt = ublk_stripe_tgt_init,
 	.deinit_tgt = ublk_stripe_tgt_deinit,
 	.queue_io = ublk_stripe_queue_io,
 	.tgt_io_done = ublk_stripe_io_done,
+	.parse_cmd_line = ublk_stripe_cmd_line,
+	.usage = ublk_stripe_usage,
 };
-- 
2.51.0


From 57e13a2e8cd208db254968631820fc1353da9db0 Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Sat, 12 Apr 2025 10:30:27 +0800
Subject: [PATCH 05/16] selftests: ublk: support user recovery

Add user recovery feature.

Meantime add user recovery test: generic_04 and generic_05(zero copy)

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250412023035.2649275-12-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 tools/testing/selftests/ublk/Makefile         |  2 +
 tools/testing/selftests/ublk/kublk.c          | 96 +++++++++++++++++--
 tools/testing/selftests/ublk/kublk.h          |  1 +
 tools/testing/selftests/ublk/test_common.sh   | 57 ++++++++++-
 .../testing/selftests/ublk/test_generic_04.sh | 40 ++++++++
 .../testing/selftests/ublk/test_generic_05.sh | 44 +++++++++
 6 files changed, 230 insertions(+), 10 deletions(-)
 create mode 100755 tools/testing/selftests/ublk/test_generic_04.sh
 create mode 100755 tools/testing/selftests/ublk/test_generic_05.sh

diff --git a/tools/testing/selftests/ublk/Makefile b/tools/testing/selftests/ublk/Makefile
index 7311e8f6bee7..d93373384e93 100644
--- a/tools/testing/selftests/ublk/Makefile
+++ b/tools/testing/selftests/ublk/Makefile
@@ -6,6 +6,8 @@ LDLIBS += -lpthread -lm -luring
 TEST_PROGS := test_generic_01.sh
 TEST_PROGS += test_generic_02.sh
 TEST_PROGS += test_generic_03.sh
+TEST_PROGS += test_generic_04.sh
+TEST_PROGS += test_generic_05.sh
 
 TEST_PROGS += test_null_01.sh
 TEST_PROGS += test_null_02.sh
diff --git a/tools/testing/selftests/ublk/kublk.c b/tools/testing/selftests/ublk/kublk.c
index 03b3d6427775..0cd6dce3f303 100644
--- a/tools/testing/selftests/ublk/kublk.c
+++ b/tools/testing/selftests/ublk/kublk.c
@@ -119,6 +119,27 @@ static int ublk_ctrl_start_dev(struct ublk_dev *dev,
 	return __ublk_ctrl_cmd(dev, &data);
 }
 
+static int ublk_ctrl_start_user_recovery(struct ublk_dev *dev)
+{
+	struct ublk_ctrl_cmd_data data = {
+		.cmd_op	= UBLK_U_CMD_START_USER_RECOVERY,
+	};
+
+	return __ublk_ctrl_cmd(dev, &data);
+}
+
+static int ublk_ctrl_end_user_recovery(struct ublk_dev *dev, int daemon_pid)
+{
+	struct ublk_ctrl_cmd_data data = {
+		.cmd_op	= UBLK_U_CMD_END_USER_RECOVERY,
+		.flags	= CTRL_CMD_HAS_DATA,
+	};
+
+	dev->dev_info.ublksrv_pid = data.data[0] = daemon_pid;
+
+	return __ublk_ctrl_cmd(dev, &data);
+}
+
 static int ublk_ctrl_add_dev(struct ublk_dev *dev)
 {
 	struct ublk_ctrl_cmd_data data = {
@@ -812,8 +833,12 @@ static int ublk_start_daemon(const struct dev_ctx *ctx, struct ublk_dev *dev)
 	free(affinity_buf);
 
 	/* everything is fine now, start us */
-	ublk_set_parameters(dev);
-	ret = ublk_ctrl_start_dev(dev, getpid());
+	if (ctx->recovery)
+		ret = ublk_ctrl_end_user_recovery(dev, getpid());
+	else {
+		ublk_set_parameters(dev);
+		ret = ublk_ctrl_start_dev(dev, getpid());
+	}
 	if (ret < 0) {
 		ublk_err("%s: ublk_ctrl_start_dev failed: %d\n", __func__, ret);
 		goto fail;
@@ -988,7 +1013,10 @@ static int __cmd_dev_add(const struct dev_ctx *ctx)
 		}
 	}
 
-	ret = ublk_ctrl_add_dev(dev);
+	if (ctx->recovery)
+		ret = ublk_ctrl_start_user_recovery(dev);
+	else
+		ret = ublk_ctrl_add_dev(dev);
 	if (ret < 0) {
 		ublk_err("%s: can't add dev id %d, type %s ret %d\n",
 				__func__, dev_id, tgt_type, ret);
@@ -1202,12 +1230,14 @@ static int cmd_dev_get_features(void)
 	return ret;
 }
 
-static int cmd_dev_help(char *exe)
+static void __cmd_create_help(char *exe, bool recovery)
 {
 	int i;
 
-	printf("%s add -t [null|loop|stripe] [-q nr_queues] [-d depth] [-n dev_id]\n", exe);
-	printf("\t[--foreground] [--quiet] [-z] [--debug_mask mask]\n");
+	printf("%s %s -t [null|loop|stripe] [-q nr_queues] [-d depth] [-n dev_id]\n",
+			exe, recovery ? "recover" : "add");
+	printf("\t[--foreground] [--quiet] [-z] [--debug_mask mask] [-r 0|1 ] [-g 0|1]\n");
+	printf("\t[-e 0|1 ] [-i 0|1]\n");
 	printf("\t[target options] [backfile1] [backfile2] ...\n");
 	printf("\tdefault: nr_queues=2(max 32), depth=128(max 1024), dev_id=-1(auto allocation)\n");
 
@@ -1217,7 +1247,25 @@ static int cmd_dev_help(char *exe)
 		if (ops->usage)
 			ops->usage(ops);
 	}
+}
+
+static void cmd_add_help(char *exe)
+{
+	__cmd_create_help(exe, false);
+	printf("\n");
+}
+
+static void cmd_recover_help(char *exe)
+{
+	__cmd_create_help(exe, true);
+	printf("\tPlease provide exact command line for creating this device with real dev_id\n");
 	printf("\n");
+}
+
+static int cmd_dev_help(char *exe)
+{
+	cmd_add_help(exe);
+	cmd_recover_help(exe);
 
 	printf("%s del [-n dev_id] -a \n", exe);
 	printf("\t -a delete all devices -n delete specified device\n\n");
@@ -1239,6 +1287,10 @@ int main(int argc, char *argv[])
 		{ "quiet",		0,	NULL,  0  },
 		{ "zero_copy",          0,      NULL, 'z' },
 		{ "foreground",		0,	NULL,  0  },
+		{ "recovery", 		1,      NULL, 'r' },
+		{ "recovery_fail_io",	1,	NULL, 'e'},
+		{ "recovery_reissue",	1,	NULL, 'i'},
+		{ "get_data",		1,	NULL, 'g'},
 		{ 0, 0, 0, 0 }
 	};
 	const struct ublk_tgt_ops *ops = NULL;
@@ -1253,13 +1305,14 @@ int main(int argc, char *argv[])
 	int ret = -EINVAL, i;
 	int tgt_argc = 1;
 	char *tgt_argv[MAX_NR_TGT_ARG] = { NULL };
+	int value;
 
 	if (argc == 1)
 		return ret;
 
 	opterr = 0;
 	optind = 2;
-	while ((opt = getopt_long(argc, argv, "t:n:d:q:az",
+	while ((opt = getopt_long(argc, argv, "t:n:d:q:r:e:i:az",
 				  longopts, &option_idx)) != -1) {
 		switch (opt) {
 		case 'a':
@@ -1281,6 +1334,25 @@ int main(int argc, char *argv[])
 		case 'z':
 			ctx.flags |= UBLK_F_SUPPORT_ZERO_COPY | UBLK_F_USER_COPY;
 			break;
+		case 'r':
+			value = strtol(optarg, NULL, 10);
+			if (value)
+				ctx.flags |= UBLK_F_USER_RECOVERY;
+			break;
+		case 'e':
+			value = strtol(optarg, NULL, 10);
+			if (value)
+				ctx.flags |= UBLK_F_USER_RECOVERY | UBLK_F_USER_RECOVERY_FAIL_IO;
+			break;
+		case 'i':
+			value = strtol(optarg, NULL, 10);
+			if (value)
+				ctx.flags |= UBLK_F_USER_RECOVERY | UBLK_F_USER_RECOVERY_REISSUE;
+			break;
+		case 'g':
+			value = strtol(optarg, NULL, 10);
+			if (value)
+				ctx.flags |= UBLK_F_NEED_GET_DATA;
 		case 0:
 			if (!strcmp(longopts[option_idx].name, "debug_mask"))
 				ublk_dbg_mask = strtol(optarg, NULL, 16);
@@ -1326,7 +1398,15 @@ int main(int argc, char *argv[])
 
 	if (!strcmp(cmd, "add"))
 		ret = cmd_dev_add(&ctx);
-	else if (!strcmp(cmd, "del"))
+	else if (!strcmp(cmd, "recover")) {
+		if (ctx.dev_id < 0) {
+			fprintf(stderr, "device id isn't provided for recovering\n");
+			ret = -EINVAL;
+		} else {
+			ctx.recovery = 1;
+			ret = cmd_dev_add(&ctx);
+		}
+	} else if (!strcmp(cmd, "del"))
 		ret = cmd_dev_del(&ctx);
 	else if (!strcmp(cmd, "list")) {
 		ctx.all = 1;
diff --git a/tools/testing/selftests/ublk/kublk.h b/tools/testing/selftests/ublk/kublk.h
index 7b7446359c8f..3d2b9f14491c 100644
--- a/tools/testing/selftests/ublk/kublk.h
+++ b/tools/testing/selftests/ublk/kublk.h
@@ -79,6 +79,7 @@ struct dev_ctx {
 	unsigned int	logging:1;
 	unsigned int	all:1;
 	unsigned int	fg:1;
+	unsigned int	recovery:1;
 
 	int _evtfd;
 	int _shmid;
diff --git a/tools/testing/selftests/ublk/test_common.sh b/tools/testing/selftests/ublk/test_common.sh
index 87fd0c824b77..e822b2a2729a 100755
--- a/tools/testing/selftests/ublk/test_common.sh
+++ b/tools/testing/selftests/ublk/test_common.sh
@@ -169,8 +169,11 @@ _have_feature()
 	return 1
 }
 
-_add_ublk_dev() {
+_create_ublk_dev() {
 	local dev_id;
+	local cmd=$1
+
+	shift 1
 
 	if [ ! -c /dev/ublk-control ]; then
 		return ${UBLK_SKIP_CODE}
@@ -181,7 +184,7 @@ _add_ublk_dev() {
 		fi
 	fi
 
-	if ! dev_id=$("${UBLK_PROG}" add "$@" | grep "dev id" | awk -F '[ :]' '{print $3}'); then
+	if ! dev_id=$("${UBLK_PROG}" "$cmd" "$@" | grep "dev id" | awk -F '[ :]' '{print $3}'); then
 		echo "fail to add ublk dev $*"
 		return 255
 	fi
@@ -194,6 +197,23 @@ _add_ublk_dev() {
 	fi
 }
 
+_add_ublk_dev() {
+	_create_ublk_dev "add" "$@"
+}
+
+_recover_ublk_dev() {
+	local dev_id
+	local state
+
+	dev_id=$(_create_ublk_dev "recover" "$@")
+	for ((j=0;j<20;j++)); do
+		state=$(_get_ublk_dev_state "${dev_id}")
+		[ "$state" == "LIVE" ] && break
+		sleep 1
+	done
+	echo "$state"
+}
+
 # kill the ublk daemon and return ublk device state
 __ublk_kill_daemon()
 {
@@ -280,6 +300,39 @@ run_io_and_kill_daemon()
 	fi
 }
 
+run_io_and_recover()
+{
+	local state
+	local dev_id
+
+	dev_id=$(_add_ublk_dev "$@")
+	_check_add_dev "$TID" $?
+
+	fio --name=job1 --filename=/dev/ublkb"${dev_id}" --ioengine=libaio \
+		--rw=readwrite --iodepth=256 --size="${size}" --numjobs=4 \
+		--runtime=20 --time_based > /dev/null 2>&1 &
+	sleep 4
+
+	state=$(__ublk_kill_daemon "${dev_id}" "QUIESCED")
+	if [ "$state" != "QUIESCED" ]; then
+		echo "device isn't quiesced($state) after killing daemon"
+		return 255
+	fi
+
+	state=$(_recover_ublk_dev -n "$dev_id" "$@")
+	if [ "$state" != "LIVE" ]; then
+		echo "faile to recover to LIVE($state)"
+		return 255
+	fi
+
+	if ! __remove_ublk_dev_return "${dev_id}"; then
+		echo "delete dev ${dev_id} failed"
+		return 255
+	fi
+	wait
+}
+
+
 _ublk_test_top_dir()
 {
 	cd "$(dirname "$0")" && pwd
diff --git a/tools/testing/selftests/ublk/test_generic_04.sh b/tools/testing/selftests/ublk/test_generic_04.sh
new file mode 100755
index 000000000000..8a3bc080c577
--- /dev/null
+++ b/tools/testing/selftests/ublk/test_generic_04.sh
@@ -0,0 +1,40 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+. "$(cd "$(dirname "$0")" && pwd)"/test_common.sh
+
+TID="generic_04"
+ERR_CODE=0
+
+ublk_run_recover_test()
+{
+	run_io_and_recover "$@"
+	ERR_CODE=$?
+	if [ ${ERR_CODE} -ne 0 ]; then
+		echo "$TID failure: $*"
+		_show_result $TID $ERR_CODE
+	fi
+}
+
+if ! _have_program fio; then
+	exit "$UBLK_SKIP_CODE"
+fi
+
+_prep_test "recover" "basic recover function verification"
+
+_create_backfile 0 256M
+_create_backfile 1 128M
+_create_backfile 2 128M
+
+ublk_run_recover_test -t null -q 2 -r 1 &
+ublk_run_recover_test -t loop -q 2 -r 1 "${UBLK_BACKFILES[0]}" &
+ublk_run_recover_test -t stripe -q 2 -r 1 "${UBLK_BACKFILES[1]}" "${UBLK_BACKFILES[2]}" &
+wait
+
+ublk_run_recover_test -t null -q 2 -r 1 -i 1 &
+ublk_run_recover_test -t loop -q 2 -r 1 -i 1 "${UBLK_BACKFILES[0]}" &
+ublk_run_recover_test -t stripe -q 2 -r 1 -i 1 "${UBLK_BACKFILES[1]}" "${UBLK_BACKFILES[2]}" &
+wait
+
+_cleanup_test "recover"
+_show_result $TID $ERR_CODE
diff --git a/tools/testing/selftests/ublk/test_generic_05.sh b/tools/testing/selftests/ublk/test_generic_05.sh
new file mode 100755
index 000000000000..714630b4b329
--- /dev/null
+++ b/tools/testing/selftests/ublk/test_generic_05.sh
@@ -0,0 +1,44 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+. "$(cd "$(dirname "$0")" && pwd)"/test_common.sh
+
+TID="generic_04"
+ERR_CODE=0
+
+ublk_run_recover_test()
+{
+	run_io_and_recover "$@"
+	ERR_CODE=$?
+	if [ ${ERR_CODE} -ne 0 ]; then
+		echo "$TID failure: $*"
+		_show_result $TID $ERR_CODE
+	fi
+}
+
+if ! _have_program fio; then
+	exit "$UBLK_SKIP_CODE"
+fi
+
+if ! _have_feature "ZERO_COPY"; then
+	exit "$UBLK_SKIP_CODE"
+fi
+
+_prep_test "recover" "basic recover function verification (zero copy)"
+
+_create_backfile 0 256M
+_create_backfile 1 128M
+_create_backfile 2 128M
+
+ublk_run_recover_test -t null -q 2 -r 1 -z &
+ublk_run_recover_test -t loop -q 2 -r 1 -z "${UBLK_BACKFILES[0]}" &
+ublk_run_recover_test -t stripe -q 2 -r 1 -z "${UBLK_BACKFILES[1]}" "${UBLK_BACKFILES[2]}" &
+wait
+
+ublk_run_recover_test -t null -q 2 -r 1 -z -i 1 &
+ublk_run_recover_test -t loop -q 2 -r 1 -z -i 1 "${UBLK_BACKFILES[0]}" &
+ublk_run_recover_test -t stripe -q 2 -r 1 -z -i 1 "${UBLK_BACKFILES[1]}" "${UBLK_BACKFILES[2]}" &
+wait
+
+_cleanup_test "recover"
+_show_result $TID $ERR_CODE
-- 
2.51.0


From 2f9a30bd16643d842da0921dc37bf00c750b0a8b Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Sat, 12 Apr 2025 10:30:28 +0800
Subject: [PATCH 06/16] selftests: ublk: add test_stress_05.sh

Add test_stress_05.sh for covering removing device with recovery
enabled.

io-hang has been observed with the following patch:

https://lore.kernel.org/linux-block/20250403-ublk_timeout-v3-1-aa09f76c7451@purestorage.com/

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250412023035.2649275-13-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 tools/testing/selftests/ublk/Makefile         |  1 +
 .../testing/selftests/ublk/test_stress_05.sh  | 64 +++++++++++++++++++
 2 files changed, 65 insertions(+)
 create mode 100755 tools/testing/selftests/ublk/test_stress_05.sh

diff --git a/tools/testing/selftests/ublk/Makefile b/tools/testing/selftests/ublk/Makefile
index d93373384e93..dddc64036aa1 100644
--- a/tools/testing/selftests/ublk/Makefile
+++ b/tools/testing/selftests/ublk/Makefile
@@ -25,6 +25,7 @@ TEST_PROGS += test_stress_01.sh
 TEST_PROGS += test_stress_02.sh
 TEST_PROGS += test_stress_03.sh
 TEST_PROGS += test_stress_04.sh
+TEST_PROGS += test_stress_05.sh
 
 TEST_GEN_PROGS_EXTENDED = kublk
 
diff --git a/tools/testing/selftests/ublk/test_stress_05.sh b/tools/testing/selftests/ublk/test_stress_05.sh
new file mode 100755
index 000000000000..a7071b10224d
--- /dev/null
+++ b/tools/testing/selftests/ublk/test_stress_05.sh
@@ -0,0 +1,64 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+. "$(cd "$(dirname "$0")" && pwd)"/test_common.sh
+TID="stress_05"
+ERR_CODE=0
+
+run_io_and_remove()
+{
+	local size=$1
+	local dev_id
+	local dev_pid
+	shift 1
+
+	dev_id=$(_add_ublk_dev "$@")
+	_check_add_dev $TID $?
+
+	[ "$UBLK_TEST_QUIET" -eq 0 ] && echo "run ublk IO vs. remove device(ublk add $*)"
+
+	fio --name=job1 --filename=/dev/ublkb"${dev_id}" --ioengine=libaio \
+		--rw=readwrite --iodepth=128 --size="${size}" --numjobs=4 \
+		--runtime=40 --time_based > /dev/null 2>&1 &
+	sleep 4
+
+	dev_pid=$(_get_ublk_daemon_pid "$dev_id")
+	kill -9 "$dev_pid"
+
+	if ! __remove_ublk_dev_return "${dev_id}"; then
+		echo "delete dev ${dev_id} failed"
+		return 255
+	fi
+}
+
+ublk_io_and_remove()
+{
+	run_io_and_remove "$@"
+	ERR_CODE=$?
+	if [ ${ERR_CODE} -ne 0 ]; then
+		echo "$TID failure: $*"
+		_show_result $TID $ERR_CODE
+	fi
+}
+
+_prep_test "stress" "run IO and remove device with recovery enabled"
+
+_create_backfile 0 256M
+_create_backfile 1 256M
+
+for reissue in $(seq 0 1); do
+	ublk_io_and_remove 8G -t null -q 4 -g 1 -r 1 -i "$reissue" &
+	ublk_io_and_remove 256M -t loop -q 4 -g 1 -r 1 -i "$reissue" "${UBLK_BACKFILES[0]}" &
+	wait
+done
+
+if _have_feature "ZERO_COPY"; then
+	for reissue in $(seq 0 1); do
+		ublk_io_and_remove 8G -t null -q 4 -g 1 -z -r 1 -i "$reissue" &
+		ublk_io_and_remove 256M -t loop -q 4 -g 1 -z -r 1 -i "$reissue" "${UBLK_BACKFILES[1]}" &
+		wait
+	done
+fi
+
+_cleanup_test "stress"
+_show_result $TID $ERR_CODE
-- 
2.51.0


From 3bf540609cab0402a7c3e40c1425532f3376318a Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Sat, 12 Apr 2025 10:30:29 +0800
Subject: [PATCH 07/16] selftests: ublk: move creating UBLK_TMP into
 _prep_test()

test may exit early because of missing program or not having required
feature before calling _prep_test(), then $UBLK_TMP isn't cleaned.

Fix it by moving creating $UBLK_TMP into _prep_test(), any resources
created since _prep_test() will be cleaned by _cleanup_test().

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250412023035.2649275-14-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 tools/testing/selftests/ublk/test_common.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/testing/selftests/ublk/test_common.sh b/tools/testing/selftests/ublk/test_common.sh
index e822b2a2729a..9fc111f64576 100755
--- a/tools/testing/selftests/ublk/test_common.sh
+++ b/tools/testing/selftests/ublk/test_common.sh
@@ -114,6 +114,7 @@ _prep_test() {
 	local type=$1
 	shift 1
 	modprobe ublk_drv > /dev/null 2>&1
+	UBLK_TMP=$(mktemp ublk_test_XXXXX)
 	[ "$UBLK_TEST_QUIET" -eq 0 ] && echo "ublk $type: $*"
 }
 
@@ -338,7 +339,6 @@ _ublk_test_top_dir()
 	cd "$(dirname "$0")" && pwd
 }
 
-UBLK_TMP=$(mktemp ublk_test_XXXXX)
 UBLK_PROG=$(_ublk_test_top_dir)/kublk
 UBLK_TEST_QUIET=1
 UBLK_TEST_SHOW_RESULT=1
-- 
2.51.0


From b69b8edfb27dfa563cd53f590ec42b481f9eb174 Mon Sep 17 00:00:00 2001
From: Uday Shankar <ushankar@purestorage.com>
Date: Wed, 16 Apr 2025 11:54:35 +0800
Subject: [PATCH 08/16] ublk: properly serialize all FETCH_REQs

Most uring_cmds issued against ublk character devices are serialized
because each command affects only one queue, and there is an early check
which only allows a single task (the queue's ubq_daemon) to issue
uring_cmds against that queue. However, this mechanism does not work for
FETCH_REQs, since they are expected before ubq_daemon is set. Since
FETCH_REQs are only used at initialization and not in the fast path,
serialize them using the per-ublk-device mutex. This fixes a number of
data races that were previously possible if a badly behaved ublk server
decided to issue multiple FETCH_REQs against the same qid/tag
concurrently.

Reported-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-2-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 drivers/block/ublk_drv.c | 77 +++++++++++++++++++++++++---------------
 1 file changed, 49 insertions(+), 28 deletions(-)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index cdb1543fa4a9..bf47f9cb8329 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -1832,8 +1832,8 @@ static void ublk_nosrv_work(struct work_struct *work)
 
 /* device can only be started after all IOs are ready */
 static void ublk_mark_io_ready(struct ublk_device *ub, struct ublk_queue *ubq)
+	__must_hold(&ub->mutex)
 {
-	mutex_lock(&ub->mutex);
 	ubq->nr_io_ready++;
 	if (ublk_queue_ready(ubq)) {
 		ubq->ubq_daemon = current;
@@ -1845,7 +1845,6 @@ static void ublk_mark_io_ready(struct ublk_device *ub, struct ublk_queue *ubq)
 	}
 	if (ub->nr_queues_ready == ub->dev_info.nr_hw_queues)
 		complete_all(&ub->completion);
-	mutex_unlock(&ub->mutex);
 }
 
 static void ublk_handle_need_get_data(struct ublk_device *ub, int q_id,
@@ -1929,6 +1928,52 @@ static int ublk_unregister_io_buf(struct io_uring_cmd *cmd,
 	return io_buffer_unregister_bvec(cmd, index, issue_flags);
 }
 
+static int ublk_fetch(struct io_uring_cmd *cmd, struct ublk_queue *ubq,
+		      struct ublk_io *io, __u64 buf_addr)
+{
+	struct ublk_device *ub = ubq->dev;
+	int ret = 0;
+
+	/*
+	 * When handling FETCH command for setting up ublk uring queue,
+	 * ub->mutex is the innermost lock, and we won't block for handling
+	 * FETCH, so it is fine even for IO_URING_F_NONBLOCK.
+	 */
+	mutex_lock(&ub->mutex);
+	/* UBLK_IO_FETCH_REQ is only allowed before queue is setup */
+	if (ublk_queue_ready(ubq)) {
+		ret = -EBUSY;
+		goto out;
+	}
+
+	/* allow each command to be FETCHed at most once */
+	if (io->flags & UBLK_IO_FLAG_ACTIVE) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	WARN_ON_ONCE(io->flags & UBLK_IO_FLAG_OWNED_BY_SRV);
+
+	if (ublk_need_map_io(ubq)) {
+		/*
+		 * FETCH_RQ has to provide IO buffer if NEED GET
+		 * DATA is not enabled
+		 */
+		if (!buf_addr && !ublk_need_get_data(ubq))
+			goto out;
+	} else if (buf_addr) {
+		/* User copy requires addr to be unset */
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ublk_fill_io_cmd(io, cmd, buf_addr);
+	ublk_mark_io_ready(ub, ubq);
+out:
+	mutex_unlock(&ub->mutex);
+	return ret;
+}
+
 static int __ublk_ch_uring_cmd(struct io_uring_cmd *cmd,
 			       unsigned int issue_flags,
 			       const struct ublksrv_io_cmd *ub_cmd)
@@ -1985,33 +2030,9 @@ static int __ublk_ch_uring_cmd(struct io_uring_cmd *cmd,
 	case UBLK_IO_UNREGISTER_IO_BUF:
 		return ublk_unregister_io_buf(cmd, ub_cmd->addr, issue_flags);
 	case UBLK_IO_FETCH_REQ:
-		/* UBLK_IO_FETCH_REQ is only allowed before queue is setup */
-		if (ublk_queue_ready(ubq)) {
-			ret = -EBUSY;
-			goto out;
-		}
-		/*
-		 * The io is being handled by server, so COMMIT_RQ is expected
-		 * instead of FETCH_REQ
-		 */
-		if (io->flags & UBLK_IO_FLAG_OWNED_BY_SRV)
-			goto out;
-
-		if (ublk_need_map_io(ubq)) {
-			/*
-			 * FETCH_RQ has to provide IO buffer if NEED GET
-			 * DATA is not enabled
-			 */
-			if (!ub_cmd->addr && !ublk_need_get_data(ubq))
-				goto out;
-		} else if (ub_cmd->addr) {
-			/* User copy requires addr to be unset */
-			ret = -EINVAL;
+		ret = ublk_fetch(cmd, ubq, io, ub_cmd->addr);
+		if (ret)
 			goto out;
-		}
-
-		ublk_fill_io_cmd(io, cmd, ub_cmd->addr);
-		ublk_mark_io_ready(ub, ubq);
 		break;
 	case UBLK_IO_COMMIT_AND_FETCH_REQ:
 		req = blk_mq_tag_to_rq(ub->tag_set.tags[ub_cmd->q_id], tag);
-- 
2.51.0


From 00b3b0d7cb454d614117c93f33351cdcd20b5b93 Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Wed, 16 Apr 2025 11:54:36 +0800
Subject: [PATCH 09/16] ublk: add ublk_force_abort_dev()

Add ublk_force_abort_dev() for handling ublk_nosrv_dev_should_queue_io()
in ublk_stop_dev(). Then queue quiesce and unquiesce can be paired in
single function.

Meantime not change device state to QUIESCED any more, since the disk is
going to be removed soon.

Reviewed-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-3-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 drivers/block/ublk_drv.c | 21 ++++++++-------------
 1 file changed, 8 insertions(+), 13 deletions(-)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index bf47f9cb8329..e1b4db2f8a56 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -1743,22 +1743,20 @@ static void __ublk_quiesce_dev(struct ublk_device *ub)
 	ub->dev_info.state = UBLK_S_DEV_QUIESCED;
 }
 
-static void ublk_unquiesce_dev(struct ublk_device *ub)
+static void ublk_force_abort_dev(struct ublk_device *ub)
 {
 	int i;
 
-	pr_devel("%s: unquiesce ub: dev_id %d state %s\n",
+	pr_devel("%s: force abort ub: dev_id %d state %s\n",
 			__func__, ub->dev_info.dev_id,
 			ub->dev_info.state == UBLK_S_DEV_LIVE ?
 			"LIVE" : "QUIESCED");
-	/* quiesce_work has run. We let requeued rqs be aborted
-	 * before running fallback_wq. "force_abort" must be seen
-	 * after request queue is unqiuesced. Then del_gendisk()
-	 * can move on.
-	 */
+	blk_mq_quiesce_queue(ub->ub_disk->queue);
+	if (ub->dev_info.state == UBLK_S_DEV_LIVE)
+		ublk_wait_tagset_rqs_idle(ub);
+
 	for (i = 0; i < ub->dev_info.nr_hw_queues; i++)
 		ublk_get_queue(ub, i)->force_abort = true;
-
 	blk_mq_unquiesce_queue(ub->ub_disk->queue);
 	/* We may have requeued some rqs in ublk_quiesce_queue() */
 	blk_mq_kick_requeue_list(ub->ub_disk->queue);
@@ -1786,11 +1784,8 @@ static void ublk_stop_dev(struct ublk_device *ub)
 	mutex_lock(&ub->mutex);
 	if (ub->dev_info.state == UBLK_S_DEV_DEAD)
 		goto unlock;
-	if (ublk_nosrv_dev_should_queue_io(ub)) {
-		if (ub->dev_info.state == UBLK_S_DEV_LIVE)
-			__ublk_quiesce_dev(ub);
-		ublk_unquiesce_dev(ub);
-	}
+	if (ublk_nosrv_dev_should_queue_io(ub))
+		ublk_force_abort_dev(ub);
 	del_gendisk(ub->ub_disk);
 	disk = ublk_detach_disk(ub);
 	put_disk(disk);
-- 
2.51.0


From 7e26cb69c5e62152a6f05a2ae23605a983a8ef31 Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Wed, 16 Apr 2025 11:54:37 +0800
Subject: [PATCH 10/16] ublk: rely on ->canceling for dealing with
 ublk_nosrv_dev_should_queue_io

Now ublk deals with ublk_nosrv_dev_should_queue_io() by keeping request
queue as quiesced. This way is fragile because queue quiesce crosses syscalls
or process contexts.

Switch to rely on ubq->canceling for dealing with
ublk_nosrv_dev_should_queue_io(), because it has been used for this purpose
during io_uring context exiting, and it can be reused before recovering too.
In ublk_queue_rq(), the request will be added to requeue list without
kicking off requeue in case of ubq->canceling, and finally requests added in
requeue list will be dispatched from either ublk_stop_dev() or
ublk_ctrl_end_recovery().

Meantime we have to move reset of ubq->canceling from ublk_ctrl_start_recovery()
to ublk_ctrl_end_recovery(), when IO handling can be recovered completely.

Then blk_mq_quiesce_queue() and blk_mq_unquiesce_queue() are always used
in same context.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Uday Shankar <ushankar@purestorage.com>
Link: https://lore.kernel.org/r/20250416035444.99569-4-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 drivers/block/ublk_drv.c | 31 +++++++++++++++++--------------
 1 file changed, 17 insertions(+), 14 deletions(-)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index e1b4db2f8a56..a479969fd77e 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -1734,13 +1734,19 @@ static void ublk_wait_tagset_rqs_idle(struct ublk_device *ub)
 
 static void __ublk_quiesce_dev(struct ublk_device *ub)
 {
+	int i;
+
 	pr_devel("%s: quiesce ub: dev_id %d state %s\n",
 			__func__, ub->dev_info.dev_id,
 			ub->dev_info.state == UBLK_S_DEV_LIVE ?
 			"LIVE" : "QUIESCED");
 	blk_mq_quiesce_queue(ub->ub_disk->queue);
+	/* mark every queue as canceling */
+	for (i = 0; i < ub->dev_info.nr_hw_queues; i++)
+		ublk_get_queue(ub, i)->canceling = true;
 	ublk_wait_tagset_rqs_idle(ub);
 	ub->dev_info.state = UBLK_S_DEV_QUIESCED;
+	blk_mq_unquiesce_queue(ub->ub_disk->queue);
 }
 
 static void ublk_force_abort_dev(struct ublk_device *ub)
@@ -2961,7 +2967,6 @@ static void ublk_queue_reinit(struct ublk_device *ub, struct ublk_queue *ubq)
 	/* We have to reset it to NULL, otherwise ub won't accept new FETCH_REQ */
 	ubq->ubq_daemon = NULL;
 	ubq->timeout = false;
-	ubq->canceling = false;
 
 	for (i = 0; i < ubq->q_depth; i++) {
 		struct ublk_io *io = &ubq->ios[i];
@@ -3048,20 +3053,18 @@ static int ublk_ctrl_end_recovery(struct ublk_device *ub,
 	pr_devel("%s: new ublksrv_pid %d, dev id %d\n",
 			__func__, ublksrv_pid, header->dev_id);
 
-	if (ublk_nosrv_dev_should_queue_io(ub)) {
-		ub->dev_info.state = UBLK_S_DEV_LIVE;
-		blk_mq_unquiesce_queue(ub->ub_disk->queue);
-		pr_devel("%s: queue unquiesced, dev id %d.\n",
-				__func__, header->dev_id);
-		blk_mq_kick_requeue_list(ub->ub_disk->queue);
-	} else {
-		blk_mq_quiesce_queue(ub->ub_disk->queue);
-		ub->dev_info.state = UBLK_S_DEV_LIVE;
-		for (i = 0; i < ub->dev_info.nr_hw_queues; i++) {
-			ublk_get_queue(ub, i)->fail_io = false;
-		}
-		blk_mq_unquiesce_queue(ub->ub_disk->queue);
+	blk_mq_quiesce_queue(ub->ub_disk->queue);
+	ub->dev_info.state = UBLK_S_DEV_LIVE;
+	for (i = 0; i < ub->dev_info.nr_hw_queues; i++) {
+		struct ublk_queue *ubq = ublk_get_queue(ub, i);
+
+		ubq->canceling = false;
+		ubq->fail_io = false;
 	}
+	blk_mq_unquiesce_queue(ub->ub_disk->queue);
+	pr_devel("%s: queue unquiesced, dev id %d.\n",
+			__func__, header->dev_id);
+	blk_mq_kick_requeue_list(ub->ub_disk->queue);
 
 	ret = 0;
  out_unlock:
-- 
2.51.0


From 728cbac5fe219d3b8a21a0688a08f2b7f8aeda2b Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Wed, 16 Apr 2025 11:54:38 +0800
Subject: [PATCH 11/16] ublk: move device reset into ublk_ch_release()

ublk_ch_release() is called after ublk char device is closed, when all
uring_cmd are done, so it is perfect fine to move ublk device reset to
ublk_ch_release() from ublk_ctrl_start_recovery().

This way can avoid to grab the exiting daemon task_struct too long.

However, reset of the following ublk IO flags has to be moved until ublk
io_uring queues are ready:

- ubq->canceling

For requeuing IO in case of ublk_nosrv_dev_should_queue_io() before device
is recovered

- ubq->fail_io

For failing IO in case of UBLK_F_USER_RECOVERY_FAIL_IO before device is
recovered

- ublk_io->flags

For preventing using io->cmd

With this way, recovery is simplified a lot.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-5-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 drivers/block/ublk_drv.c | 121 +++++++++++++++++++++++----------------
 1 file changed, 72 insertions(+), 49 deletions(-)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index a479969fd77e..1fe39cf85b2f 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -1074,7 +1074,7 @@ static inline struct ublk_uring_cmd_pdu *ublk_get_uring_cmd_pdu(
 
 static inline bool ubq_daemon_is_dying(struct ublk_queue *ubq)
 {
-	return ubq->ubq_daemon->flags & PF_EXITING;
+	return !ubq->ubq_daemon || ubq->ubq_daemon->flags & PF_EXITING;
 }
 
 /* todo: handle partial completion */
@@ -1470,6 +1470,37 @@ static const struct blk_mq_ops ublk_mq_ops = {
 	.timeout	= ublk_timeout,
 };
 
+static void ublk_queue_reinit(struct ublk_device *ub, struct ublk_queue *ubq)
+{
+	int i;
+
+	/* All old ioucmds have to be completed */
+	ubq->nr_io_ready = 0;
+
+	/*
+	 * old daemon is PF_EXITING, put it now
+	 *
+	 * It could be NULL in case of closing one quisced device.
+	 */
+	if (ubq->ubq_daemon)
+		put_task_struct(ubq->ubq_daemon);
+	/* We have to reset it to NULL, otherwise ub won't accept new FETCH_REQ */
+	ubq->ubq_daemon = NULL;
+	ubq->timeout = false;
+
+	for (i = 0; i < ubq->q_depth; i++) {
+		struct ublk_io *io = &ubq->ios[i];
+
+		/*
+		 * UBLK_IO_FLAG_CANCELED is kept for avoiding to touch
+		 * io->cmd
+		 */
+		io->flags &= UBLK_IO_FLAG_CANCELED;
+		io->cmd = NULL;
+		io->addr = 0;
+	}
+}
+
 static int ublk_ch_open(struct inode *inode, struct file *filp)
 {
 	struct ublk_device *ub = container_of(inode->i_cdev,
@@ -1481,10 +1512,26 @@ static int ublk_ch_open(struct inode *inode, struct file *filp)
 	return 0;
 }
 
+static void ublk_reset_ch_dev(struct ublk_device *ub)
+{
+	int i;
+
+	for (i = 0; i < ub->dev_info.nr_hw_queues; i++)
+		ublk_queue_reinit(ub, ublk_get_queue(ub, i));
+
+	/* set to NULL, otherwise new ubq_daemon cannot mmap the io_cmd_buf */
+	ub->mm = NULL;
+	ub->nr_queues_ready = 0;
+	ub->nr_privileged_daemon = 0;
+}
+
 static int ublk_ch_release(struct inode *inode, struct file *filp)
 {
 	struct ublk_device *ub = filp->private_data;
 
+	/* all uring_cmd has been done now, reset device & ubq */
+	ublk_reset_ch_dev(ub);
+
 	clear_bit(UB_STATE_OPEN, &ub->state);
 	return 0;
 }
@@ -1831,6 +1878,24 @@ static void ublk_nosrv_work(struct work_struct *work)
 	ublk_cancel_dev(ub);
 }
 
+/* reset ublk io_uring queue & io flags */
+static void ublk_reset_io_flags(struct ublk_device *ub)
+{
+	int i, j;
+
+	for (i = 0; i < ub->dev_info.nr_hw_queues; i++) {
+		struct ublk_queue *ubq = ublk_get_queue(ub, i);
+
+		/* UBLK_IO_FLAG_CANCELED can be cleared now */
+		spin_lock(&ubq->cancel_lock);
+		for (j = 0; j < ubq->q_depth; j++)
+			ubq->ios[j].flags &= ~UBLK_IO_FLAG_CANCELED;
+		spin_unlock(&ubq->cancel_lock);
+		ubq->canceling = false;
+		ubq->fail_io = false;
+	}
+}
+
 /* device can only be started after all IOs are ready */
 static void ublk_mark_io_ready(struct ublk_device *ub, struct ublk_queue *ubq)
 	__must_hold(&ub->mutex)
@@ -1844,8 +1909,12 @@ static void ublk_mark_io_ready(struct ublk_device *ub, struct ublk_queue *ubq)
 		if (capable(CAP_SYS_ADMIN))
 			ub->nr_privileged_daemon++;
 	}
-	if (ub->nr_queues_ready == ub->dev_info.nr_hw_queues)
+
+	if (ub->nr_queues_ready == ub->dev_info.nr_hw_queues) {
+		/* now we are ready for handling ublk io request */
+		ublk_reset_io_flags(ub);
 		complete_all(&ub->completion);
+	}
 }
 
 static void ublk_handle_need_get_data(struct ublk_device *ub, int q_id,
@@ -2954,41 +3023,14 @@ static int ublk_ctrl_set_params(struct ublk_device *ub,
 	return ret;
 }
 
-static void ublk_queue_reinit(struct ublk_device *ub, struct ublk_queue *ubq)
-{
-	int i;
-
-	WARN_ON_ONCE(!(ubq->ubq_daemon && ubq_daemon_is_dying(ubq)));
-
-	/* All old ioucmds have to be completed */
-	ubq->nr_io_ready = 0;
-	/* old daemon is PF_EXITING, put it now */
-	put_task_struct(ubq->ubq_daemon);
-	/* We have to reset it to NULL, otherwise ub won't accept new FETCH_REQ */
-	ubq->ubq_daemon = NULL;
-	ubq->timeout = false;
-
-	for (i = 0; i < ubq->q_depth; i++) {
-		struct ublk_io *io = &ubq->ios[i];
-
-		/* forget everything now and be ready for new FETCH_REQ */
-		io->flags = 0;
-		io->cmd = NULL;
-		io->addr = 0;
-	}
-}
-
 static int ublk_ctrl_start_recovery(struct ublk_device *ub,
 		const struct ublksrv_ctrl_cmd *header)
 {
 	int ret = -EINVAL;
-	int i;
 
 	mutex_lock(&ub->mutex);
 	if (ublk_nosrv_should_stop_dev(ub))
 		goto out_unlock;
-	if (!ub->nr_queues_ready)
-		goto out_unlock;
 	/*
 	 * START_RECOVERY is only allowd after:
 	 *
@@ -3012,12 +3054,6 @@ static int ublk_ctrl_start_recovery(struct ublk_device *ub,
 		goto out_unlock;
 	}
 	pr_devel("%s: start recovery for dev id %d.\n", __func__, header->dev_id);
-	for (i = 0; i < ub->dev_info.nr_hw_queues; i++)
-		ublk_queue_reinit(ub, ublk_get_queue(ub, i));
-	/* set to NULL, otherwise new ubq_daemon cannot mmap the io_cmd_buf */
-	ub->mm = NULL;
-	ub->nr_queues_ready = 0;
-	ub->nr_privileged_daemon = 0;
 	init_completion(&ub->completion);
 	ret = 0;
  out_unlock:
@@ -3030,7 +3066,6 @@ static int ublk_ctrl_end_recovery(struct ublk_device *ub,
 {
 	int ublksrv_pid = (int)header->data[0];
 	int ret = -EINVAL;
-	int i;
 
 	pr_devel("%s: Waiting for new ubq_daemons(nr: %d) are ready, dev id %d...\n",
 			__func__, ub->dev_info.nr_hw_queues, header->dev_id);
@@ -3050,22 +3085,10 @@ static int ublk_ctrl_end_recovery(struct ublk_device *ub,
 		goto out_unlock;
 	}
 	ub->dev_info.ublksrv_pid = ublksrv_pid;
+	ub->dev_info.state = UBLK_S_DEV_LIVE;
 	pr_devel("%s: new ublksrv_pid %d, dev id %d\n",
 			__func__, ublksrv_pid, header->dev_id);
-
-	blk_mq_quiesce_queue(ub->ub_disk->queue);
-	ub->dev_info.state = UBLK_S_DEV_LIVE;
-	for (i = 0; i < ub->dev_info.nr_hw_queues; i++) {
-		struct ublk_queue *ubq = ublk_get_queue(ub, i);
-
-		ubq->canceling = false;
-		ubq->fail_io = false;
-	}
-	blk_mq_unquiesce_queue(ub->ub_disk->queue);
-	pr_devel("%s: queue unquiesced, dev id %d.\n",
-			__func__, header->dev_id);
 	blk_mq_kick_requeue_list(ub->ub_disk->queue);
-
 	ret = 0;
  out_unlock:
 	mutex_unlock(&ub->mutex);
-- 
2.51.0


From 82a8a30c581bbbe653d33c6ce2ef67e3072c7f12 Mon Sep 17 00:00:00 2001
From: Uday Shankar <ushankar@purestorage.com>
Date: Wed, 16 Apr 2025 11:54:39 +0800
Subject: [PATCH 12/16] ublk: improve detection and handling of ublk server
 exit

There are currently two ways in which ublk server exit is detected by
ublk_drv:

1. uring_cmd cancellation. If there are any outstanding uring_cmds which
   have not been completed to the ublk server when it exits, io_uring
   calls the uring_cmd callback with a special cancellation flag as the
   issuing task is exiting.
2. I/O timeout. This is needed in addition to the above to handle the
   "saturated queue" case, when all I/Os for a given queue are in the
   ublk server, and therefore there are no outstanding uring_cmds to
   cancel when the ublk server exits.

There are a couple of issues with this approach:

- It is complex and inelegant to have two methods to detect the same
  condition
- The second method detects ublk server exit only after a long delay
  (~30s, the default timeout assigned by the block layer). This delays
  the nosrv behavior from kicking in and potential subsequent recovery
  of the device.

The second issue is brought to light with the new test_generic_06 which
will be added in following patch. It fails before this fix:

selftests: ublk: test_generic_06.sh
dev id is 0
dd: error writing '/dev/ublkb0': Input/output error
1+0 records in
0+0 records out
0 bytes copied, 30.0611 s, 0.0 kB/s
DEAD
dd took 31 seconds to exit (>= 5s tolerance)!
generic_06 : [FAIL]

Fix this by instead detecting and handling ublk server exit in the
character file release callback. This has several advantages:

- This one place can handle both saturated and unsaturated queues. Thus,
  it replaces both preexisting methods of detecting ublk server exit.
- It runs quickly on ublk server exit - there is no 30s delay.
- It starts the process of removing task references in ublk_drv. This is
  needed if we want to relax restrictions in the driver like letting
  only one thread serve each queue

There is also the disadvantage that the character file release callback
can also be triggered by intentional close of the file, which is a
significant behavior change. Preexisting ublk servers (libublksrv) are
dependent on the ability to open/close the file multiple times. To
address this, only transition to a nosrv state if the file is released
while the ublk device is live. This allows for programs to open/close
the file multiple times during setup. It is still a behavior change if a
ublk server decides to close/reopen the file while the device is LIVE
(i.e. while it is responsible for serving I/O), but that would be highly
unusual. This behavior is in line with what is done by FUSE, which is
very similar to ublk in that a userspace daemon is providing services
traditionally provided by the kernel.

With this change in, the new test (and all other selftests, and all
ublksrv tests) pass:

selftests: ublk: test_generic_06.sh
dev id is 0
dd: error writing '/dev/ublkb0': Input/output error
1+0 records in
0+0 records out
0 bytes copied, 0.0376731 s, 0.0 kB/s
DEAD
generic_04 : [PASS]

Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-6-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 drivers/block/ublk_drv.c | 223 +++++++++++++++++++++------------------
 1 file changed, 123 insertions(+), 100 deletions(-)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index 1fe39cf85b2f..b0a7e5acb2eb 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -199,8 +199,6 @@ struct ublk_device {
 	struct completion	completion;
 	unsigned int		nr_queues_ready;
 	unsigned int		nr_privileged_daemon;
-
-	struct work_struct	nosrv_work;
 };
 
 /* header of ublk_params */
@@ -209,8 +207,9 @@ struct ublk_params_header {
 	__u32	types;
 };
 
-static bool ublk_abort_requests(struct ublk_device *ub, struct ublk_queue *ubq);
-
+static void ublk_stop_dev_unlocked(struct ublk_device *ub);
+static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq);
+static void __ublk_quiesce_dev(struct ublk_device *ub);
 static inline struct request *__ublk_check_and_get_req(struct ublk_device *ub,
 		struct ublk_queue *ubq, int tag, size_t offset);
 static inline unsigned int ublk_req_build_flags(struct request *req);
@@ -1336,8 +1335,6 @@ static void ublk_queue_cmd_list(struct ublk_queue *ubq, struct rq_list *l)
 static enum blk_eh_timer_return ublk_timeout(struct request *rq)
 {
 	struct ublk_queue *ubq = rq->mq_hctx->driver_data;
-	unsigned int nr_inflight = 0;
-	int i;
 
 	if (ubq->flags & UBLK_F_UNPRIVILEGED_DEV) {
 		if (!ubq->timeout) {
@@ -1348,26 +1345,6 @@ static enum blk_eh_timer_return ublk_timeout(struct request *rq)
 		return BLK_EH_DONE;
 	}
 
-	if (!ubq_daemon_is_dying(ubq))
-		return BLK_EH_RESET_TIMER;
-
-	for (i = 0; i < ubq->q_depth; i++) {
-		struct ublk_io *io = &ubq->ios[i];
-
-		if (!(io->flags & UBLK_IO_FLAG_ACTIVE))
-			nr_inflight++;
-	}
-
-	/* cancelable uring_cmd can't help us if all commands are in-flight */
-	if (nr_inflight == ubq->q_depth) {
-		struct ublk_device *ub = ubq->dev;
-
-		if (ublk_abort_requests(ub, ubq)) {
-			schedule_work(&ub->nosrv_work);
-		}
-		return BLK_EH_DONE;
-	}
-
 	return BLK_EH_RESET_TIMER;
 }
 
@@ -1525,13 +1502,105 @@ static void ublk_reset_ch_dev(struct ublk_device *ub)
 	ub->nr_privileged_daemon = 0;
 }
 
+static struct gendisk *ublk_get_disk(struct ublk_device *ub)
+{
+	struct gendisk *disk;
+
+	spin_lock(&ub->lock);
+	disk = ub->ub_disk;
+	if (disk)
+		get_device(disk_to_dev(disk));
+	spin_unlock(&ub->lock);
+
+	return disk;
+}
+
+static void ublk_put_disk(struct gendisk *disk)
+{
+	if (disk)
+		put_device(disk_to_dev(disk));
+}
+
 static int ublk_ch_release(struct inode *inode, struct file *filp)
 {
 	struct ublk_device *ub = filp->private_data;
+	struct gendisk *disk;
+	int i;
+
+	/*
+	 * disk isn't attached yet, either device isn't live, or it has
+	 * been removed already, so we needn't to do anything
+	 */
+	disk = ublk_get_disk(ub);
+	if (!disk)
+		goto out;
+
+	/*
+	 * All uring_cmd are done now, so abort any request outstanding to
+	 * the ublk server
+	 *
+	 * This can be done in lockless way because ublk server has been
+	 * gone
+	 *
+	 * More importantly, we have to provide forward progress guarantee
+	 * without holding ub->mutex, otherwise control task grabbing
+	 * ub->mutex triggers deadlock
+	 *
+	 * All requests may be inflight, so ->canceling may not be set, set
+	 * it now.
+	 */
+	for (i = 0; i < ub->dev_info.nr_hw_queues; i++) {
+		struct ublk_queue *ubq = ublk_get_queue(ub, i);
+
+		ubq->canceling = true;
+		ublk_abort_queue(ub, ubq);
+	}
+	blk_mq_kick_requeue_list(disk->queue);
+
+	/*
+	 * All infligh requests have been completed or requeued and any new
+	 * request will be failed or requeued via `->canceling` now, so it is
+	 * fine to grab ub->mutex now.
+	 */
+	mutex_lock(&ub->mutex);
+
+	/* double check after grabbing lock */
+	if (!ub->ub_disk)
+		goto unlock;
+
+	/*
+	 * Transition the device to the nosrv state. What exactly this
+	 * means depends on the recovery flags
+	 */
+	blk_mq_quiesce_queue(disk->queue);
+	if (ublk_nosrv_should_stop_dev(ub)) {
+		/*
+		 * Allow any pending/future I/O to pass through quickly
+		 * with an error. This is needed because del_gendisk
+		 * waits for all pending I/O to complete
+		 */
+		for (i = 0; i < ub->dev_info.nr_hw_queues; i++)
+			ublk_get_queue(ub, i)->force_abort = true;
+		blk_mq_unquiesce_queue(disk->queue);
+
+		ublk_stop_dev_unlocked(ub);
+	} else {
+		if (ublk_nosrv_dev_should_queue_io(ub)) {
+			__ublk_quiesce_dev(ub);
+		} else {
+			ub->dev_info.state = UBLK_S_DEV_FAIL_IO;
+			for (i = 0; i < ub->dev_info.nr_hw_queues; i++)
+				ublk_get_queue(ub, i)->fail_io = true;
+		}
+		blk_mq_unquiesce_queue(disk->queue);
+	}
+unlock:
+	mutex_unlock(&ub->mutex);
+	ublk_put_disk(disk);
 
 	/* all uring_cmd has been done now, reset device & ubq */
 	ublk_reset_ch_dev(ub);
-
+out:
 	clear_bit(UB_STATE_OPEN, &ub->state);
 	return 0;
 }
@@ -1627,37 +1696,22 @@ static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq)
 }
 
 /* Must be called when queue is frozen */
-static bool ublk_mark_queue_canceling(struct ublk_queue *ubq)
+static void ublk_mark_queue_canceling(struct ublk_queue *ubq)
 {
-	bool canceled;
-
 	spin_lock(&ubq->cancel_lock);
-	canceled = ubq->canceling;
-	if (!canceled)
+	if (!ubq->canceling)
 		ubq->canceling = true;
 	spin_unlock(&ubq->cancel_lock);
-
-	return canceled;
 }
 
-static bool ublk_abort_requests(struct ublk_device *ub, struct ublk_queue *ubq)
+static void ublk_start_cancel(struct ublk_queue *ubq)
 {
-	bool was_canceled = ubq->canceling;
-	struct gendisk *disk;
-
-	if (was_canceled)
-		return false;
-
-	spin_lock(&ub->lock);
-	disk = ub->ub_disk;
-	if (disk)
-		get_device(disk_to_dev(disk));
-	spin_unlock(&ub->lock);
+	struct ublk_device *ub = ubq->dev;
+	struct gendisk *disk = ublk_get_disk(ub);
 
 	/* Our disk has been dead */
 	if (!disk)
-		return false;
-
+		return;
 	/*
 	 * Now we are serialized with ublk_queue_rq()
 	 *
@@ -1666,15 +1720,9 @@ static bool ublk_abort_requests(struct ublk_device *ub, struct ublk_queue *ubq)
 	 * touch completed uring_cmd
 	 */
 	blk_mq_quiesce_queue(disk->queue);
-	was_canceled = ublk_mark_queue_canceling(ubq);
-	if (!was_canceled) {
-		/* abort queue is for making forward progress */
-		ublk_abort_queue(ub, ubq);
-	}
+	ublk_mark_queue_canceling(ubq);
 	blk_mq_unquiesce_queue(disk->queue);
-	put_device(disk_to_dev(disk));
-
-	return !was_canceled;
+	ublk_put_disk(disk);
 }
 
 static void ublk_cancel_cmd(struct ublk_queue *ubq, struct ublk_io *io,
@@ -1698,6 +1746,17 @@ static void ublk_cancel_cmd(struct ublk_queue *ubq, struct ublk_io *io,
 /*
  * The ublk char device won't be closed when calling cancel fn, so both
  * ublk device and queue are guaranteed to be live
+ *
+ * Two-stage cancel:
+ *
+ * - make every active uring_cmd done in ->cancel_fn()
+ *
+ * - aborting inflight ublk IO requests in ublk char device release handler,
+ *   which depends on 1st stage because device can only be closed iff all
+ *   uring_cmd are done
+ *
+ * Do _not_ try to acquire ub->mutex before all inflight requests are
+ * aborted, otherwise deadlock may be caused.
  */
 static void ublk_uring_cmd_cancel_fn(struct io_uring_cmd *cmd,
 		unsigned int issue_flags)
@@ -1705,8 +1764,6 @@ static void ublk_uring_cmd_cancel_fn(struct io_uring_cmd *cmd,
 	struct ublk_uring_cmd_pdu *pdu = ublk_get_uring_cmd_pdu(cmd);
 	struct ublk_queue *ubq = pdu->ubq;
 	struct task_struct *task;
-	struct ublk_device *ub;
-	bool need_schedule;
 	struct ublk_io *io;
 
 	if (WARN_ON_ONCE(!ubq))
@@ -1719,16 +1776,12 @@ static void ublk_uring_cmd_cancel_fn(struct io_uring_cmd *cmd,
 	if (WARN_ON_ONCE(task && task != ubq->ubq_daemon))
 		return;
 
-	ub = ubq->dev;
-	need_schedule = ublk_abort_requests(ub, ubq);
+	if (!ubq->canceling)
+		ublk_start_cancel(ubq);
 
 	io = &ubq->ios[pdu->tag];
 	WARN_ON_ONCE(io->cmd != cmd);
 	ublk_cancel_cmd(ubq, io, issue_flags);
-
-	if (need_schedule) {
-		schedule_work(&ub->nosrv_work);
-	}
 }
 
 static inline bool ublk_queue_ready(struct ublk_queue *ubq)
@@ -1787,13 +1840,11 @@ static void __ublk_quiesce_dev(struct ublk_device *ub)
 			__func__, ub->dev_info.dev_id,
 			ub->dev_info.state == UBLK_S_DEV_LIVE ?
 			"LIVE" : "QUIESCED");
-	blk_mq_quiesce_queue(ub->ub_disk->queue);
 	/* mark every queue as canceling */
 	for (i = 0; i < ub->dev_info.nr_hw_queues; i++)
 		ublk_get_queue(ub, i)->canceling = true;
 	ublk_wait_tagset_rqs_idle(ub);
 	ub->dev_info.state = UBLK_S_DEV_QUIESCED;
-	blk_mq_unquiesce_queue(ub->ub_disk->queue);
 }
 
 static void ublk_force_abort_dev(struct ublk_device *ub)
@@ -1830,50 +1881,25 @@ static struct gendisk *ublk_detach_disk(struct ublk_device *ub)
 	return disk;
 }
 
-static void ublk_stop_dev(struct ublk_device *ub)
+static void ublk_stop_dev_unlocked(struct ublk_device *ub)
+	__must_hold(&ub->mutex)
 {
 	struct gendisk *disk;
 
-	mutex_lock(&ub->mutex);
 	if (ub->dev_info.state == UBLK_S_DEV_DEAD)
-		goto unlock;
+		return;
+
 	if (ublk_nosrv_dev_should_queue_io(ub))
 		ublk_force_abort_dev(ub);
 	del_gendisk(ub->ub_disk);
 	disk = ublk_detach_disk(ub);
 	put_disk(disk);
- unlock:
-	mutex_unlock(&ub->mutex);
-	ublk_cancel_dev(ub);
 }
 
-static void ublk_nosrv_work(struct work_struct *work)
+static void ublk_stop_dev(struct ublk_device *ub)
 {
-	struct ublk_device *ub =
-		container_of(work, struct ublk_device, nosrv_work);
-	int i;
-
-	if (ublk_nosrv_should_stop_dev(ub)) {
-		ublk_stop_dev(ub);
-		return;
-	}
-
 	mutex_lock(&ub->mutex);
-	if (ub->dev_info.state != UBLK_S_DEV_LIVE)
-		goto unlock;
-
-	if (ublk_nosrv_dev_should_queue_io(ub)) {
-		__ublk_quiesce_dev(ub);
-	} else {
-		blk_mq_quiesce_queue(ub->ub_disk->queue);
-		ub->dev_info.state = UBLK_S_DEV_FAIL_IO;
-		for (i = 0; i < ub->dev_info.nr_hw_queues; i++) {
-			ublk_get_queue(ub, i)->fail_io = true;
-		}
-		blk_mq_unquiesce_queue(ub->ub_disk->queue);
-	}
-
- unlock:
+	ublk_stop_dev_unlocked(ub);
 	mutex_unlock(&ub->mutex);
 	ublk_cancel_dev(ub);
 }
@@ -2502,7 +2528,6 @@ static void ublk_remove(struct ublk_device *ub)
 	bool unprivileged;
 
 	ublk_stop_dev(ub);
-	cancel_work_sync(&ub->nosrv_work);
 	cdev_device_del(&ub->cdev, &ub->cdev_dev);
 	unprivileged = ub->dev_info.flags & UBLK_F_UNPRIVILEGED_DEV;
 	ublk_put_device(ub);
@@ -2787,7 +2812,6 @@ static int ublk_ctrl_add_dev(const struct ublksrv_ctrl_cmd *header)
 		goto out_unlock;
 	mutex_init(&ub->mutex);
 	spin_lock_init(&ub->lock);
-	INIT_WORK(&ub->nosrv_work, ublk_nosrv_work);
 
 	ret = ublk_alloc_dev_number(ub, header->dev_id);
 	if (ret < 0)
@@ -2919,7 +2943,6 @@ static inline void ublk_ctrl_cmd_dump(struct io_uring_cmd *cmd)
 static int ublk_ctrl_stop_dev(struct ublk_device *ub)
 {
 	ublk_stop_dev(ub);
-	cancel_work_sync(&ub->nosrv_work);
 	return 0;
 }
 
-- 
2.51.0


From 736b005b413a172670711ee17cab3c8ccab83223 Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Wed, 16 Apr 2025 11:54:40 +0800
Subject: [PATCH 13/16] ublk: remove __ublk_quiesce_dev()

Remove __ublk_quiesce_dev() and open code for updating device state as
QUIESCED.

We needn't to drain inflight requests in __ublk_quiesce_dev() any more,
because all inflight requests are aborted in ublk char device release
handler.

Also we needn't to set ->canceling in __ublk_quiesce_dev() any more
because it is done unconditionally now in ublk_ch_release().

Reviewed-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-7-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 drivers/block/ublk_drv.c | 19 ++-----------------
 1 file changed, 2 insertions(+), 17 deletions(-)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index b0a7e5acb2eb..bf708e9e9a12 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -209,7 +209,6 @@ struct ublk_params_header {
 
 static void ublk_stop_dev_unlocked(struct ublk_device *ub);
 static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq);
-static void __ublk_quiesce_dev(struct ublk_device *ub);
 static inline struct request *__ublk_check_and_get_req(struct ublk_device *ub,
 		struct ublk_queue *ubq, int tag, size_t offset);
 static inline unsigned int ublk_req_build_flags(struct request *req);
@@ -1586,7 +1585,8 @@ static int ublk_ch_release(struct inode *inode, struct file *filp)
 		ublk_stop_dev_unlocked(ub);
 	} else {
 		if (ublk_nosrv_dev_should_queue_io(ub)) {
-			__ublk_quiesce_dev(ub);
+			/* ->canceling is set and all requests are aborted */
+			ub->dev_info.state = UBLK_S_DEV_QUIESCED;
 		} else {
 			ub->dev_info.state = UBLK_S_DEV_FAIL_IO;
 			for (i = 0; i < ub->dev_info.nr_hw_queues; i++)
@@ -1832,21 +1832,6 @@ static void ublk_wait_tagset_rqs_idle(struct ublk_device *ub)
 	}
 }
 
-static void __ublk_quiesce_dev(struct ublk_device *ub)
-{
-	int i;
-
-	pr_devel("%s: quiesce ub: dev_id %d state %s\n",
-			__func__, ub->dev_info.dev_id,
-			ub->dev_info.state == UBLK_S_DEV_LIVE ?
-			"LIVE" : "QUIESCED");
-	/* mark every queue as canceling */
-	for (i = 0; i < ub->dev_info.nr_hw_queues; i++)
-		ublk_get_queue(ub, i)->canceling = true;
-	ublk_wait_tagset_rqs_idle(ub);
-	ub->dev_info.state = UBLK_S_DEV_QUIESCED;
-}
-
 static void ublk_force_abort_dev(struct ublk_device *ub)
 {
 	int i;
-- 
2.51.0


From e63d2228ef831af36f963b3ab8604160cfff84c1 Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Wed, 16 Apr 2025 11:54:41 +0800
Subject: [PATCH 14/16] ublk: simplify aborting ublk request

Now ublk_abort_queue() is moved to ublk char device release handler,
meantime our request queue is "quiesced" because either ->canceling was
set from uring_cmd cancel function or all IOs are inflight and can't be
completed by ublk server, things becomes easy much:

- all uring_cmd are done, so we needn't to mark io as UBLK_IO_FLAG_ABORTED
for handling completion from uring_cmd

- ublk char device is closed, no one can hold IO request reference any more,
so we can simply complete this request or requeue it for ublk_nosrv_should_reissue_outstanding.

Reviewed-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-8-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 drivers/block/ublk_drv.c | 82 ++++++++++------------------------------
 1 file changed, 20 insertions(+), 62 deletions(-)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index bf708e9e9a12..2de7b2bd409d 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -122,15 +122,6 @@ struct ublk_uring_cmd_pdu {
  */
 #define UBLK_IO_FLAG_OWNED_BY_SRV 0x02
 
-/*
- * IO command is aborted, so this flag is set in case of
- * !UBLK_IO_FLAG_ACTIVE.
- *
- * After this flag is observed, any pending or new incoming request
- * associated with this io command will be failed immediately
- */
-#define UBLK_IO_FLAG_ABORTED 0x04
-
 /*
  * UBLK_IO_FLAG_NEED_GET_DATA is set because IO command requires
  * get data buffer address from ublksrv.
@@ -1083,12 +1074,6 @@ static inline void __ublk_complete_rq(struct request *req)
 	unsigned int unmapped_bytes;
 	blk_status_t res = BLK_STS_OK;
 
-	/* called from ublk_abort_queue() code path */
-	if (io->flags & UBLK_IO_FLAG_ABORTED) {
-		res = BLK_STS_IOERR;
-		goto exit;
-	}
-
 	/* failed read IO if nothing is read */
 	if (!io->res && req_op(req) == REQ_OP_READ)
 		io->res = -EIO;
@@ -1138,47 +1123,6 @@ static void ublk_complete_rq(struct kref *ref)
 	__ublk_complete_rq(req);
 }
 
-static void ublk_do_fail_rq(struct request *req)
-{
-	struct ublk_queue *ubq = req->mq_hctx->driver_data;
-
-	if (ublk_nosrv_should_reissue_outstanding(ubq->dev))
-		blk_mq_requeue_request(req, false);
-	else
-		__ublk_complete_rq(req);
-}
-
-static void ublk_fail_rq_fn(struct kref *ref)
-{
-	struct ublk_rq_data *data = container_of(ref, struct ublk_rq_data,
-			ref);
-	struct request *req = blk_mq_rq_from_pdu(data);
-
-	ublk_do_fail_rq(req);
-}
-
-/*
- * Since ublk_rq_task_work_cb always fails requests immediately during
- * exiting, __ublk_fail_req() is only called from abort context during
- * exiting. So lock is unnecessary.
- *
- * Also aborting may not be started yet, keep in mind that one failed
- * request may be issued by block layer again.
- */
-static void __ublk_fail_req(struct ublk_queue *ubq, struct ublk_io *io,
-		struct request *req)
-{
-	WARN_ON_ONCE(io->flags & UBLK_IO_FLAG_ACTIVE);
-
-	if (ublk_need_req_ref(ubq)) {
-		struct ublk_rq_data *data = blk_mq_rq_to_pdu(req);
-
-		kref_put(&data->ref, ublk_fail_rq_fn);
-	} else {
-		ublk_do_fail_rq(req);
-	}
-}
-
 static void ubq_complete_io_cmd(struct ublk_io *io, int res,
 				unsigned issue_flags)
 {
@@ -1667,10 +1611,26 @@ static void ublk_commit_completion(struct ublk_device *ub,
 		ublk_put_req_ref(ubq, req);
 }
 
+static void __ublk_fail_req(struct ublk_queue *ubq, struct ublk_io *io,
+		struct request *req)
+{
+	WARN_ON_ONCE(io->flags & UBLK_IO_FLAG_ACTIVE);
+
+	if (ublk_nosrv_should_reissue_outstanding(ubq->dev))
+		blk_mq_requeue_request(req, false);
+	else {
+		io->res = -EIO;
+		__ublk_complete_rq(req);
+	}
+}
+
 /*
- * Called from ubq_daemon context via cancel fn, meantime quiesce ublk
- * blk-mq queue, so we are called exclusively with blk-mq and ubq_daemon
- * context, so everything is serialized.
+ * Called from ublk char device release handler, when any uring_cmd is
+ * done, meantime request queue is "quiesced" since all inflight requests
+ * can't be completed because ublk server is dead.
+ *
+ * So no one can hold our request IO reference any more, simply ignore the
+ * reference, and complete the request immediately
  */
 static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq)
 {
@@ -1687,10 +1647,8 @@ static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq)
 			 * will do it
 			 */
 			rq = blk_mq_tag_to_rq(ub->tag_set.tags[ubq->q_id], i);
-			if (rq && blk_mq_request_started(rq)) {
-				io->flags |= UBLK_IO_FLAG_ABORTED;
+			if (rq && blk_mq_request_started(rq))
 				__ublk_fail_req(ubq, io, rq);
-			}
 		}
 	}
 }
-- 
2.51.0


From 81586652bb1f6c797159161db8d59c18d66b9eb3 Mon Sep 17 00:00:00 2001
From: Uday Shankar <ushankar@purestorage.com>
Date: Wed, 16 Apr 2025 11:54:42 +0800
Subject: [PATCH 15/16] selftests: ublk: add generic_06 for covering fault
 inject

Add one simple fault inject target, and verify if an application using ublk
device sees an I/O error quickly after the ublk server dies.

Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-9-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 tools/testing/selftests/ublk/Makefile         |  4 +-
 tools/testing/selftests/ublk/fault_inject.c   | 98 +++++++++++++++++++
 tools/testing/selftests/ublk/kublk.c          |  3 +-
 tools/testing/selftests/ublk/kublk.h          | 12 ++-
 .../testing/selftests/ublk/test_generic_06.sh | 41 ++++++++
 5 files changed, 155 insertions(+), 3 deletions(-)
 create mode 100644 tools/testing/selftests/ublk/fault_inject.c
 create mode 100755 tools/testing/selftests/ublk/test_generic_06.sh

diff --git a/tools/testing/selftests/ublk/Makefile b/tools/testing/selftests/ublk/Makefile
index dddc64036aa1..ec4624a283bc 100644
--- a/tools/testing/selftests/ublk/Makefile
+++ b/tools/testing/selftests/ublk/Makefile
@@ -8,6 +8,7 @@ TEST_PROGS += test_generic_02.sh
 TEST_PROGS += test_generic_03.sh
 TEST_PROGS += test_generic_04.sh
 TEST_PROGS += test_generic_05.sh
+TEST_PROGS += test_generic_06.sh
 
 TEST_PROGS += test_null_01.sh
 TEST_PROGS += test_null_02.sh
@@ -31,7 +32,8 @@ TEST_GEN_PROGS_EXTENDED = kublk
 
 include ../lib.mk
 
-$(TEST_GEN_PROGS_EXTENDED): kublk.c null.c file_backed.c common.c stripe.c
+$(TEST_GEN_PROGS_EXTENDED): kublk.c null.c file_backed.c common.c stripe.c \
+	fault_inject.c
 
 check:
 	shellcheck -x -f gcc *.sh
diff --git a/tools/testing/selftests/ublk/fault_inject.c b/tools/testing/selftests/ublk/fault_inject.c
new file mode 100644
index 000000000000..94a8e729ba4c
--- /dev/null
+++ b/tools/testing/selftests/ublk/fault_inject.c
@@ -0,0 +1,98 @@
+// SPDX-License-Identifier: GPL-2.0
+
+/*
+ * Fault injection ublk target. Hack this up however you like for
+ * testing specific behaviors of ublk_drv. Currently is a null target
+ * with a configurable delay before completing each I/O. This delay can
+ * be used to test ublk_drv's handling of I/O outstanding to the ublk
+ * server when it dies.
+ */
+
+#include "kublk.h"
+
+static int ublk_fault_inject_tgt_init(const struct dev_ctx *ctx,
+				      struct ublk_dev *dev)
+{
+	const struct ublksrv_ctrl_dev_info *info = &dev->dev_info;
+	unsigned long dev_size = 250UL << 30;
+
+	dev->tgt.dev_size = dev_size;
+	dev->tgt.params = (struct ublk_params) {
+		.types = UBLK_PARAM_TYPE_BASIC,
+		.basic = {
+			.logical_bs_shift	= 9,
+			.physical_bs_shift	= 12,
+			.io_opt_shift		= 12,
+			.io_min_shift		= 9,
+			.max_sectors		= info->max_io_buf_bytes >> 9,
+			.dev_sectors		= dev_size >> 9,
+		},
+	};
+
+	dev->private_data = (void *)(unsigned long)(ctx->fault_inject.delay_us * 1000);
+	return 0;
+}
+
+static int ublk_fault_inject_queue_io(struct ublk_queue *q, int tag)
+{
+	const struct ublksrv_io_desc *iod = ublk_get_iod(q, tag);
+	struct io_uring_sqe *sqe;
+	struct __kernel_timespec ts = {
+		.tv_nsec = (long long)q->dev->private_data,
+	};
+
+	ublk_queue_alloc_sqes(q, &sqe, 1);
+	io_uring_prep_timeout(sqe, &ts, 1, 0);
+	sqe->user_data = build_user_data(tag, ublksrv_get_op(iod), 0, 1);
+
+	ublk_queued_tgt_io(q, tag, 1);
+
+	return 0;
+}
+
+static void ublk_fault_inject_tgt_io_done(struct ublk_queue *q, int tag,
+					  const struct io_uring_cqe *cqe)
+{
+	const struct ublksrv_io_desc *iod = ublk_get_iod(q, tag);
+
+	if (cqe->res != -ETIME)
+		ublk_err("%s: unexpected cqe res %d\n", __func__, cqe->res);
+
+	if (ublk_completed_tgt_io(q, tag))
+		ublk_complete_io(q, tag, iod->nr_sectors << 9);
+	else
+		ublk_err("%s: io not complete after 1 cqe\n", __func__);
+}
+
+static void ublk_fault_inject_cmd_line(struct dev_ctx *ctx, int argc, char *argv[])
+{
+	static const struct option longopts[] = {
+		{ "delay_us", 	1,	NULL,  0  },
+		{ 0, 0, 0, 0 }
+	};
+	int option_idx, opt;
+
+	ctx->fault_inject.delay_us = 0;
+	while ((opt = getopt_long(argc, argv, "",
+				  longopts, &option_idx)) != -1) {
+		switch (opt) {
+		case 0:
+			if (!strcmp(longopts[option_idx].name, "delay_us"))
+				ctx->fault_inject.delay_us = strtoll(optarg, NULL, 10);
+		}
+	}
+}
+
+static void ublk_fault_inject_usage(const struct ublk_tgt_ops *ops)
+{
+	printf("\tfault_inject: [--delay_us us (default 0)]\n");
+}
+
+const struct ublk_tgt_ops fault_inject_tgt_ops = {
+	.name = "fault_inject",
+	.init_tgt = ublk_fault_inject_tgt_init,
+	.queue_io = ublk_fault_inject_queue_io,
+	.tgt_io_done = ublk_fault_inject_tgt_io_done,
+	.parse_cmd_line = ublk_fault_inject_cmd_line,
+	.usage = ublk_fault_inject_usage,
+};
diff --git a/tools/testing/selftests/ublk/kublk.c b/tools/testing/selftests/ublk/kublk.c
index 0cd6dce3f303..759f06637146 100644
--- a/tools/testing/selftests/ublk/kublk.c
+++ b/tools/testing/selftests/ublk/kublk.c
@@ -12,6 +12,7 @@ static const struct ublk_tgt_ops *tgt_ops_list[] = {
 	&null_tgt_ops,
 	&loop_tgt_ops,
 	&stripe_tgt_ops,
+	&fault_inject_tgt_ops,
 };
 
 static const struct ublk_tgt_ops *ublk_find_tgt(const char *name)
@@ -1234,7 +1235,7 @@ static void __cmd_create_help(char *exe, bool recovery)
 {
 	int i;
 
-	printf("%s %s -t [null|loop|stripe] [-q nr_queues] [-d depth] [-n dev_id]\n",
+	printf("%s %s -t [null|loop|stripe|fault_inject] [-q nr_queues] [-d depth] [-n dev_id]\n",
 			exe, recovery ? "recover" : "add");
 	printf("\t[--foreground] [--quiet] [-z] [--debug_mask mask] [-r 0|1 ] [-g 0|1]\n");
 	printf("\t[-e 0|1 ] [-i 0|1]\n");
diff --git a/tools/testing/selftests/ublk/kublk.h b/tools/testing/selftests/ublk/kublk.h
index 3d2b9f14491c..29571eb296f1 100644
--- a/tools/testing/selftests/ublk/kublk.h
+++ b/tools/testing/selftests/ublk/kublk.h
@@ -68,6 +68,11 @@ struct stripe_ctx {
 	unsigned int    chunk_size;
 };
 
+struct fault_inject_ctx {
+	/* fault_inject */
+	unsigned long   delay_us;
+};
+
 struct dev_ctx {
 	char tgt_type[16];
 	unsigned long flags;
@@ -81,6 +86,9 @@ struct dev_ctx {
 	unsigned int	fg:1;
 	unsigned int	recovery:1;
 
+	/* fault_inject */
+	long long	delay_us;
+
 	int _evtfd;
 	int _shmid;
 
@@ -88,7 +96,8 @@ struct dev_ctx {
 	struct ublk_dev *shadow_dev;
 
 	union {
-		struct stripe_ctx  stripe;
+		struct stripe_ctx 	stripe;
+		struct fault_inject_ctx fault_inject;
 	};
 };
 
@@ -384,6 +393,7 @@ static inline int ublk_queue_use_zc(const struct ublk_queue *q)
 extern const struct ublk_tgt_ops null_tgt_ops;
 extern const struct ublk_tgt_ops loop_tgt_ops;
 extern const struct ublk_tgt_ops stripe_tgt_ops;
+extern const struct ublk_tgt_ops fault_inject_tgt_ops;
 
 void backing_file_tgt_deinit(struct ublk_dev *dev);
 int backing_file_tgt_init(struct ublk_dev *dev);
diff --git a/tools/testing/selftests/ublk/test_generic_06.sh b/tools/testing/selftests/ublk/test_generic_06.sh
new file mode 100755
index 000000000000..b67230c42c84
--- /dev/null
+++ b/tools/testing/selftests/ublk/test_generic_06.sh
@@ -0,0 +1,41 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+. "$(cd "$(dirname "$0")" && pwd)"/test_common.sh
+
+TID="generic_06"
+ERR_CODE=0
+
+_prep_test "fault_inject" "fast cleanup when all I/Os of one hctx are in server"
+
+# configure ublk server to sleep 2s before completing each I/O
+dev_id=$(_add_ublk_dev -t fault_inject -q 2 -d 1 --delay_us 2000000)
+_check_add_dev $TID $?
+
+STARTTIME=${SECONDS}
+
+dd if=/dev/urandom of=/dev/ublkb${dev_id} oflag=direct bs=4k count=1 status=none > /dev/null 2>&1 &
+dd_pid=$!
+
+__ublk_kill_daemon ${dev_id} "DEAD"
+
+wait $dd_pid
+dd_exitcode=$?
+
+ENDTIME=${SECONDS}
+ELAPSED=$(($ENDTIME - $STARTTIME))
+
+# assert that dd sees an error and exits quickly after ublk server is
+# killed. previously this relied on seeing an I/O timeout and so would
+# take ~30s
+if [ $dd_exitcode -eq 0 ]; then
+        echo "dd unexpectedly exited successfully!"
+        ERR_CODE=255
+fi
+if [ $ELAPSED -ge 5 ]; then
+        echo "dd took $ELAPSED seconds to exit (>= 5s tolerance)!"
+        ERR_CODE=255
+fi
+
+_cleanup_test "fault_inject"
+_show_result $TID $ERR_CODE
-- 
2.51.0


From 3d7aa0c7b4e96cd460826d932e44710cdeb3378b Mon Sep 17 00:00:00 2001
From: Richard Weinberger <richard@nod.at>
Date: Fri, 18 Apr 2025 10:02:50 +0200
Subject: [PATCH 16/16] nvmet: fix out-of-bounds access in nvmet_enable_port

When trying to enable a port that has no transport configured yet,
nvmet_enable_port() uses NVMF_TRTYPE_MAX (255) to query the transports
array, causing an out-of-bounds access:

[  106.058694] BUG: KASAN: global-out-of-bounds in nvmet_enable_port+0x42/0x1da
[  106.058719] Read of size 8 at addr ffffffff89dafa58 by task ln/632
[...]
[  106.076026] nvmet: transport type 255 not supported

Since commit 200adac75888, NVMF_TRTYPE_MAX is the default state as configured by
nvmet_ports_make().
Avoid this by checking for NVMF_TRTYPE_MAX before proceeding.

Fixes: 200adac75888 ("nvme: Add PCI transport type")
Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
---
 drivers/nvme/target/core.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c
index 71f8d06998d6..245475c43127 100644
--- a/drivers/nvme/target/core.c
+++ b/drivers/nvme/target/core.c
@@ -324,6 +324,9 @@ int nvmet_enable_port(struct nvmet_port *port)
 
 	lockdep_assert_held(&nvmet_config_sem);
 
+	if (port->disc_addr.trtype == NVMF_TRTYPE_MAX)
+		return -EINVAL;
+
 	ops = nvmet_transports[port->disc_addr.trtype];
 	if (!ops) {
 		up_write(&nvmet_config_sem);
-- 
2.51.0