Hi all,
Recently, when creating dm-mpath devices based on iSCSI devices, we
encountered a 'tried to init an initialized object' error triggered
by one DM_DEV_CREATE and two DM_TABLE_LOAD operations.
// DM_DEV_CREATE
dev_create
dm_create
alloc_dev
blk_alloc_disk // alloc gendisk for md->disk
// first DM_TABLE_LOAD
table_load
dm_get_md_type // md->type is DM_TYPE_NONE
dm_setup_md_queue
dm_mq_init_request_queue
add_disk // add md->disk
device_add_disk
add_disk_fwnode
__add_disk
blk_register_queue
kobject_init
kobject_init_internal
// set kobj->state_initialized
bd_link_disk_holder
// return -ENODEV since iscsi device has been logged out
// skip setting md->type and leave it as DM_TYPE_NONE
// second DM_TABLE_LOAD
table_load
dm_get_md_type // md->type is DM_TYPE_NONE
dm_setup_md_queue
dm_mq_init_request_queue
add_disk // add md->disk
device_add_disk
add_disk_fwnode
__add_disk
blk_register_queue
kobject_init
// dump_stack since kobj->state_initialized has been set
Resetting state_initialized after the first failed DM_TABLE_LOAD might
resolve this issue, but it seems weird — I think state_initialized is
designed to never be reset during the object lifecycle.
Any suggestions would be appreciated.
Thanks,
Lingfeng
*Reproduction steps are as follows:*
Base:
master 4b290aae788e
Diff:
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index abfe0392b5a4..b41f5a6f3e72 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -2598,6 +2598,9 @@ int dm_setup_md_queue(struct mapped_device *md,
struct dm_table *t)
if (r)
return r;
+ printk("%s sleep before link disk holder\n", __func__);
+ msleep(10 * 1000);
+ printk("%s sleep done\n", __func__);
/*
* Register the holder relationship for devices added before
the disk
* was live.
Procedures:
[root@nfs_test3 test]# lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 10G 0 disk
sdb 8:16 0 10G 0 disk
sdc 8:32 0 30G 0 disk
vda 253:0 0 20G 0 disk /
[root@nfs_test3 test]# ifconfig lo:0 127.0.0.2 up
[root@nfs_test3 test]# ifconfig lo:1 127.0.0.3 up
[root@nfs_test3 test]# service tgtd restart
Redirecting to /bin/systemctl restart tgtd.service
[root@nfs_test3 test]# service iscsid restart
Redirecting to /bin/systemctl restart iscsid.service
[root@nfs_test3 test]# iscsiadm -m discovery -p 127.0.0.2 -t st
127.0.0.2:3260,1 iqn.2019-04.jenkins.disk
[root@nfs_test3 test]# iscsiadm -m discovery -p 127.0.0.3 -t st
127.0.0.3:3260,1 iqn.2019-04.jenkins.disk
[root@nfs_test3 test]# iscsiadm -m node -l
Logging in to [iface: default, target: iqn.2020-05.com.bdqn:disk1,
portal: 192.168.240.250,3260] (multiple)
Logging in to [iface: default, target: iqn.2019-04.jenkins.disk, portal:
127.0.0.2,3260] (multiple)
Logging in to [iface: default, target: iqn.2019-04.jenkins.disk, portal:
127.0.0.3,3260] (multiple)
Logging in to [iface: default, target: , portal: ,3260] (multiple)
[ 275.687170][ C10] sd 3:0:0:1: Power-on or device reset occurred
[ 275.708911][ C10] sd 4:0:0:1: Power-on or device reset occurred
iscsiadm: caught SIGINT, exiting...
[root@nfs_test3 test]# lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 10G 0 disk
sdb 8:16 0 10G 0 disk
sdc 8:32 0 30G 0 disk
sdd 8:48 0 10G 0 disk
sde 8:64 0 10G 0 disk
vda 253:0 0 20G 0 disk /
[root@nfs_test3 test]# ./dmtest &
[1] 2742
[root@nfs_test3 test]# DM_VERSION 4 50 0
[ 294.101770][ T2742] dm_setup_md_queue sleep before link disk holder
[root@nfs_test3 test]#
[root@nfs_test3 test]# iscsiadm -m node -u
Logging out of session [sid: 1, target: iqn.2019-04.jenkins.disk,
portal: 127.0.0.2,3260]
Logging out of session [sid: 2, target: iqn.2019-04.jenkins.disk,
portal: 127.0.0.3,3260]
Logout of [sid: 1, target: iqn.2019-04.jenkins.disk, portal:
127.0.0.2,3260] successful.
Logout of [sid: 2, target: iqn.2019-04.jenkins.disk, portal:
127.0.0.3,3260] successful.
[root@nfs_test3 test]#
[root@nfs_test3 test]#
[root@nfs_test3 test]# [ 304.185384][ T2742] dm_setup_md_queue sleep done
[ 304.280823][ T2742] device-mapper: ioctl: unable to set up device
queue for new table.
delay before second table_load...
[root@nfs_test3 test]#
[root@nfs_test3 test]# iscsiadm -m node -l
Logging in to [iface: default, target: iqn.2020-05.com.bdqn:disk1,
portal: 192.168.240.250,3260] (multiple)
Logging in to [iface: default, target: iqn.2019-04.jenkins.disk, portal:
127.0.0.2,3260] (multiple)
Logging in to [iface: default, target: iqn.2019-04.jenkins.disk, portal:
127.0.0.3,3260] (multiple)
iscsiadm: could not read session targetname: 5
iscsiadm: could not find session info for session3
Logging in to [iface: default, target: , portal: ,3260] (multiple)
[ 307.357847][ C10] sd 3:0:0:1: Power-on or device reset occurred
[ 307.372301][ C10] sd 4:0:0:1: Power-on or device reset occurred
iscsiadm: caught SIGINT, exiting...
[root@nfs_test3 test]#
[root@nfs_test3 test]#
[root@nfs_test3 test]# lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 10G 0 disk
sdb 8:16 0 10G 0 disk
sdc 8:32 0 30G 0 disk
sdd 8:48 0 10G 0 disk
sde 8:64 0 10G 0 disk
vda 253:0 0 20G 0 disk /
[root@nfs_test3 test]# delay done
[ 314.938868][ T2742] kobject: kobject (ffff88810c27bb90): tried to
init an initialized object, something is seriously wrong.
[ 314.940322][ T2742] CPU: 6 UID: 0 PID: 2742 Comm: dmtest Not tainted
6.16.0-04405-g4b290aae788e-dirty #21 PREEMPT(none)
[ 314.940331][ T2742] Hardware name: QEMU Standard PC (i440FX + PIIX,
1996), BIOS 1.16.3-2.fc40 04/01/2014
[ 314.940335][ T2742] Call Trace:
[ 314.940339][ T2742] <TASK>
[ 314.940345][ T2742] dump_stack_lvl+0x5b/0x80
[ 314.940370][ T2742] kobject_init.cold+0x43/0x51
[ 314.940380][ T2742] blk_register_queue+0x46/0x280
[ 314.940399][ T2742] __add_disk+0x385/0x5d0
[ 314.940413][ T2742] add_disk_fwnode+0xb5/0x280
[ 314.940425][ T2742] dm_setup_md_queue+0x194/0x1c0
[ 314.940434][ T2742] ? mempool_init_noprof+0x16/0x20
[ 314.940443][ T2742] ? __pfx_dm_setup_md_queue+0x10/0x10
[ 314.940476][ T2742] table_load+0x297/0x2d0
[ 314.940484][ T2742] ? __pfx_table_load+0x10/0x10
[ 314.940489][ T2742] ? _inline_copy_from_user+0x6c/0x90
[ 314.940499][ T2742] ? copy_params+0xa6/0x330
[ 314.940513][ T2742] ctl_ioctl+0x2a2/0x480
[ 314.940520][ T2742] ? __pfx_table_load+0x10/0x10
[ 314.940528][ T2742] ? __pfx_ctl_ioctl+0x10/0x10
[ 314.940564][ T2742] ? __free_zapped_classes+0x70/0x120
[ 314.940572][ T2742] ? exc_page_fault+0x61/0xa0
[ 314.940591][ T2742] dm_ctl_ioctl+0xe/0x20
[ 314.940598][ T2742] __x64_sys_ioctl+0xc7/0x110
[ 314.940605][ T2742] ? rcu_is_watching+0x20/0x50
[ 314.940616][ T2742] do_syscall_64+0x72/0x390
[ 314.940624][ T2742] entry_SYSCALL_64_after_hwframe+0x76/0x7e
[ 314.940631][ T2742] RIP: 0033:0x7f8ce1106577
[ 314.940639][ T2742] Code: b3 66 90 48 8b 05 11 89 2c 00 64 c7 00 26
00 00 00 48 c7 c0 ff ff ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 b8 10 00
00 00 0f 05 <48> 3d 01 f0 ff ff 73 018
[ 314.940645][ T2742] RSP: 002b:00007fff58708438 EFLAGS: 00000202
ORIG_RAX: 0000000000000010
[ 314.940652][ T2742] RAX: ffffffffffffffda RBX: 0000000000000000 RCX:
00007f8ce1106577
[ 314.940657][ T2742] RDX: 0000000008355010 RSI: 00000000c138fd09 RDI:
0000000000000003
[ 314.940661][ T2742] RBP: 00007fff587084d0 R08: 0000000000000000 R09:
0000000000000000
[ 314.940665][ T2742] R10: 0000000000400e00 R11: 0000000000000202 R12:
3120312030206874
[ 314.940669][ T2742] R13: 6563697672657320 R14: 5f65756575712031 R15:
61705f6f6e5f6669
[ 314.940686][ T2742] </TASK>
[ 315.026828][ T2742] dm_setup_md_queue sleep before link disk holder
[root@nfs_test3 test]# [ 325.177205][ T2742] dm_setup_md_queue sleep done
dmtest.c
#include <stdio.h>
#include <stdlib.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>
#include <errno.h>
#include <linux/dm-ioctl.h>
#include <sys/ioctl.h>
#define DM_EXISTS_FLAG 0x00000004
int main(int argc, char *argv[]) {
int ret, fd, index;
int in_data;
struct dm_ioctl *param = calloc(1, sizeof(struct dm_ioctl) + 16384);
struct dm_target_spec sp;
char *target = (char *)(param + 1);
char *last_param = target + sizeof(struct dm_target_spec);
struct dm_target_msg tmsg;
fd = open("/dev/mapper/control", O_RDWR);
if (fd < 0) {
printf("Failed to open device /dev/mapper/control: %s\n",
strerror(errno));
return -1;
}
// DM_VERSION
param->version[0] = 4;
param->version[1] = 0;
param->version[2] = 0;
param->data_size = 16384;
param->flags = DM_EXISTS_FLAG;
ret = ioctl(fd, DM_VERSION, param);
if (ret < 0) {
printf("Failed to send DM_VERSION: %s\n", strerror(errno));
close(fd);
return -1;
}
printf("DM_VERSION %d %d %d\n", param->version[0],
param->version[1], param->version[2]);
// DM_DEV_CREATE
param->data_start = 312;
strncpy(param->name, "multipath-dev", sizeof(param->name));
// clear
memset(target, 0, 16384);
param->data_size = 16384;
param->dev = 0;
ret = ioctl(fd, DM_DEV_CREATE, param);
if (ret < 0) {
printf("Failed to send DM_DEV_CREATE: %s\n", strerror(errno));
close(fd);
return -1;
}
// DM_TABLE_LOAD
memset(param->name, 0, sizeof(param->name));
param->flags = DM_EXISTS_FLAG;
param->data_start = 312;
param->data_size = 16384;
//param->dev = 64514; // makedev(252, 2)
param->target_count = 1;
sp.sector_start = 0;
sp.length = 20971520;
strncpy(sp.target_type, "multipath", sizeof(sp.target_type) - 1);
sp.target_type[sizeof(sp.target_type) - 1] = '\0';
memcpy(target, &sp, sizeof(struct dm_target_spec));
strcpy(last_param, "1 queue_if_no_path 0 1 1 service-time 0 1 1
8:48 1");
ret = ioctl(fd, DM_TABLE_LOAD, param);
// if (ret < 0) {
// printf("Failed to send DM_TABLE_LOAD: %s\n", strerror(errno));
// close(fd);
// return -1;
// }
printf("delay before second table_load...\n");
sleep(10);
printf("delay done\n");
memset(param->name, 0, sizeof(param->name));
param->flags = DM_EXISTS_FLAG;
param->data_start = 312;
param->data_size = 16384;
param->target_count = 1;
sp.sector_start = 0;
sp.length = 20971520;
strncpy(sp.target_type, "multipath", sizeof(sp.target_type) - 1);
sp.target_type[sizeof(sp.target_type) - 1] = '\0';
memcpy(target, &sp, sizeof(struct dm_target_spec));
strcpy(last_param, "1 queue_if_no_path 0 1 1 service-time 0 1 1
8:48 1");
ret = ioctl(fd, DM_TABLE_LOAD, param);
if (ret < 0) {
printf("Failed to send DM_TABLE_LOAD: %s\n", strerror(errno));
close(fd);
return -1;
}
// DM_DEV_SUSPEND
param->flags = DM_EXISTS_FLAG; // continue creating
param->event_nr = 6345265;
param->data_size = 16384;
param->dev = 0;
strncpy(param->name, "multipath-dev", sizeof(param->name));
ret = ioctl(fd, DM_DEV_SUSPEND, param);
if (ret < 0) {
printf("Failed to send DM_DEV_SUSPEND: %s\n", strerror(errno));
close(fd);
return -1;
}
printf("DM_DEV_SUSPEND sent successfully\n");
close(fd);
return 0;
}
On 7/30/2025 8:00 PM, Li Lingfeng wrote:
> Hi all,
>
> Recently, when creating dm-mpath devices based on iSCSI devices, we
> encountered a 'tried to init an initialized object' error triggered
> by one DM_DEV_CREATE and two DM_TABLE_LOAD operations.
>
> // DM_DEV_CREATE
> dev_create
> dm_create
> alloc_dev
> blk_alloc_disk // alloc gendisk for md->disk
>
> // first DM_TABLE_LOAD
> table_load
> dm_get_md_type // md->type is DM_TYPE_NONE
> dm_setup_md_queue
> dm_mq_init_request_queue
> add_disk // add md->disk
> device_add_disk
> add_disk_fwnode
> __add_disk
> blk_register_queue
> kobject_init
> kobject_init_internal
> // set kobj->state_initialized
> bd_link_disk_holder
> // return -ENODEV since iscsi device has been logged out
> // skip setting md->type and leave it as DM_TYPE_NONE
>
> // second DM_TABLE_LOAD
> table_load
> dm_get_md_type // md->type is DM_TYPE_NONE
> dm_setup_md_queue
> dm_mq_init_request_queue
> add_disk // add md->disk
> device_add_disk
> add_disk_fwnode
> __add_disk
> blk_register_queue
> kobject_init
> // dump_stack since kobj->state_initialized has been set
>
> Resetting state_initialized after the first failed DM_TABLE_LOAD might
> resolve this issue, but it seems weird — I think state_initialized is
> designed to never be reset during the object lifecycle.
>
cc +linux-block
The disk->queue_kobj is initialized in add_disk() and is uninitialized
in del_disk(). And it seems that blk_unregister_queue() in del_disk()
doesn't uninitialize the queue_kobj completely, because it doesn't
expect the queue_kobj will be added again. I think the right place to
fix the problem is blk_unregister_queue(). How about just memset the
queue_kobj as zero after deleting the queue_kobj in blk_unregister_queue() ?
> Any suggestions would be appreciated.
>
> Thanks,
> Lingfeng
>
> *Reproduction steps are as follows:*
> Base:
> master 4b290aae788e
>
> Diff:
> diff --git a/drivers/md/dm.c b/drivers/md/dm.c
> index abfe0392b5a4..b41f5a6f3e72 100644
> --- a/drivers/md/dm.c
> +++ b/drivers/md/dm.c
> @@ -2598,6 +2598,9 @@ int dm_setup_md_queue(struct mapped_device *md,
> struct dm_table *t)
> if (r)
> return r;
>
> + printk("%s sleep before link disk holder\n", __func__);
> + msleep(10 * 1000);
> + printk("%s sleep done\n", __func__);
> /*
> * Register the holder relationship for devices added before
> the disk
> * was live.
>
> Procedures:
> [root@nfs_test3 test]# lsblk
> NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
> sda 8:0 0 10G 0 disk
> sdb 8:16 0 10G 0 disk
> sdc 8:32 0 30G 0 disk
> vda 253:0 0 20G 0 disk /
> [root@nfs_test3 test]# ifconfig lo:0 127.0.0.2 up
> [root@nfs_test3 test]# ifconfig lo:1 127.0.0.3 up
> [root@nfs_test3 test]# service tgtd restart
> Redirecting to /bin/systemctl restart tgtd.service
> [root@nfs_test3 test]# service iscsid restart
> Redirecting to /bin/systemctl restart iscsid.service
> [root@nfs_test3 test]# iscsiadm -m discovery -p 127.0.0.2 -t st
> 127.0.0.2:3260,1 iqn.2019-04.jenkins.disk
> [root@nfs_test3 test]# iscsiadm -m discovery -p 127.0.0.3 -t st
> 127.0.0.3:3260,1 iqn.2019-04.jenkins.disk
> [root@nfs_test3 test]# iscsiadm -m node -l
> Logging in to [iface: default, target: iqn.2020-05.com.bdqn:disk1,
> portal: 192.168.240.250,3260] (multiple)
> Logging in to [iface: default, target: iqn.2019-04.jenkins.disk,
> portal: 127.0.0.2,3260] (multiple)
> Logging in to [iface: default, target: iqn.2019-04.jenkins.disk,
> portal: 127.0.0.3,3260] (multiple)
> Logging in to [iface: default, target: , portal: ,3260] (multiple)
> [ 275.687170][ C10] sd 3:0:0:1: Power-on or device reset occurred
> [ 275.708911][ C10] sd 4:0:0:1: Power-on or device reset occurred
> iscsiadm: caught SIGINT, exiting...
> [root@nfs_test3 test]# lsblk
> NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
> sda 8:0 0 10G 0 disk
> sdb 8:16 0 10G 0 disk
> sdc 8:32 0 30G 0 disk
> sdd 8:48 0 10G 0 disk
> sde 8:64 0 10G 0 disk
> vda 253:0 0 20G 0 disk /
> [root@nfs_test3 test]# ./dmtest &
> [1] 2742
> [root@nfs_test3 test]# DM_VERSION 4 50 0
> [ 294.101770][ T2742] dm_setup_md_queue sleep before link disk holder
> [root@nfs_test3 test]#
> [root@nfs_test3 test]# iscsiadm -m node -u
> Logging out of session [sid: 1, target: iqn.2019-04.jenkins.disk,
> portal: 127.0.0.2,3260]
> Logging out of session [sid: 2, target: iqn.2019-04.jenkins.disk,
> portal: 127.0.0.3,3260]
> Logout of [sid: 1, target: iqn.2019-04.jenkins.disk, portal:
> 127.0.0.2,3260] successful.
> Logout of [sid: 2, target: iqn.2019-04.jenkins.disk, portal:
> 127.0.0.3,3260] successful.
> [root@nfs_test3 test]#
> [root@nfs_test3 test]#
> [root@nfs_test3 test]# [ 304.185384][ T2742] dm_setup_md_queue sleep
> done
> [ 304.280823][ T2742] device-mapper: ioctl: unable to set up device
> queue for new table.
> delay before second table_load...
> [root@nfs_test3 test]#
> [root@nfs_test3 test]# iscsiadm -m node -l
> Logging in to [iface: default, target: iqn.2020-05.com.bdqn:disk1,
> portal: 192.168.240.250,3260] (multiple)
> Logging in to [iface: default, target: iqn.2019-04.jenkins.disk,
> portal: 127.0.0.2,3260] (multiple)
> Logging in to [iface: default, target: iqn.2019-04.jenkins.disk,
> portal: 127.0.0.3,3260] (multiple)
> iscsiadm: could not read session targetname: 5
> iscsiadm: could not find session info for session3
> Logging in to [iface: default, target: , portal: ,3260] (multiple)
> [ 307.357847][ C10] sd 3:0:0:1: Power-on or device reset occurred
> [ 307.372301][ C10] sd 4:0:0:1: Power-on or device reset occurred
> iscsiadm: caught SIGINT, exiting...
> [root@nfs_test3 test]#
> [root@nfs_test3 test]#
> [root@nfs_test3 test]# lsblk
> NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
> sda 8:0 0 10G 0 disk
> sdb 8:16 0 10G 0 disk
> sdc 8:32 0 30G 0 disk
> sdd 8:48 0 10G 0 disk
> sde 8:64 0 10G 0 disk
> vda 253:0 0 20G 0 disk /
> [root@nfs_test3 test]# delay done
> [ 314.938868][ T2742] kobject: kobject (ffff88810c27bb90): tried to
> init an initialized object, something is seriously wrong.
> [ 314.940322][ T2742] CPU: 6 UID: 0 PID: 2742 Comm: dmtest Not
> tainted 6.16.0-04405-g4b290aae788e-dirty #21 PREEMPT(none)
> [ 314.940331][ T2742] Hardware name: QEMU Standard PC (i440FX + PIIX,
> 1996), BIOS 1.16.3-2.fc40 04/01/2014
> [ 314.940335][ T2742] Call Trace:
> [ 314.940339][ T2742] <TASK>
> [ 314.940345][ T2742] dump_stack_lvl+0x5b/0x80
> [ 314.940370][ T2742] kobject_init.cold+0x43/0x51
> [ 314.940380][ T2742] blk_register_queue+0x46/0x280
> [ 314.940399][ T2742] __add_disk+0x385/0x5d0
> [ 314.940413][ T2742] add_disk_fwnode+0xb5/0x280
> [ 314.940425][ T2742] dm_setup_md_queue+0x194/0x1c0
> [ 314.940434][ T2742] ? mempool_init_noprof+0x16/0x20
> [ 314.940443][ T2742] ? __pfx_dm_setup_md_queue+0x10/0x10
> [ 314.940476][ T2742] table_load+0x297/0x2d0
> [ 314.940484][ T2742] ? __pfx_table_load+0x10/0x10
> [ 314.940489][ T2742] ? _inline_copy_from_user+0x6c/0x90
> [ 314.940499][ T2742] ? copy_params+0xa6/0x330
> [ 314.940513][ T2742] ctl_ioctl+0x2a2/0x480
> [ 314.940520][ T2742] ? __pfx_table_load+0x10/0x10
> [ 314.940528][ T2742] ? __pfx_ctl_ioctl+0x10/0x10
> [ 314.940564][ T2742] ? __free_zapped_classes+0x70/0x120
> [ 314.940572][ T2742] ? exc_page_fault+0x61/0xa0
> [ 314.940591][ T2742] dm_ctl_ioctl+0xe/0x20
> [ 314.940598][ T2742] __x64_sys_ioctl+0xc7/0x110
> [ 314.940605][ T2742] ? rcu_is_watching+0x20/0x50
> [ 314.940616][ T2742] do_syscall_64+0x72/0x390
> [ 314.940624][ T2742] entry_SYSCALL_64_after_hwframe+0x76/0x7e
> [ 314.940631][ T2742] RIP: 0033:0x7f8ce1106577
> [ 314.940639][ T2742] Code: b3 66 90 48 8b 05 11 89 2c 00 64 c7 00 26
> 00 00 00 48 c7 c0 ff ff ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 b8 10
> 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 018
> [ 314.940645][ T2742] RSP: 002b:00007fff58708438 EFLAGS: 00000202
> ORIG_RAX: 0000000000000010
> [ 314.940652][ T2742] RAX: ffffffffffffffda RBX: 0000000000000000
> RCX: 00007f8ce1106577
> [ 314.940657][ T2742] RDX: 0000000008355010 RSI: 00000000c138fd09
> RDI: 0000000000000003
> [ 314.940661][ T2742] RBP: 00007fff587084d0 R08: 0000000000000000
> R09: 0000000000000000
> [ 314.940665][ T2742] R10: 0000000000400e00 R11: 0000000000000202
> R12: 3120312030206874
> [ 314.940669][ T2742] R13: 6563697672657320 R14: 5f65756575712031
> R15: 61705f6f6e5f6669
> [ 314.940686][ T2742] </TASK>
> [ 315.026828][ T2742] dm_setup_md_queue sleep before link disk holder
> [root@nfs_test3 test]# [ 325.177205][ T2742] dm_setup_md_queue sleep
> done
>
>
> dmtest.c
> #include <stdio.h>
> #include <stdlib.h>
> #include <fcntl.h>
> #include <unistd.h>
> #include <string.h>
> #include <errno.h>
> #include <linux/dm-ioctl.h>
> #include <sys/ioctl.h>
>
> #define DM_EXISTS_FLAG 0x00000004
>
> int main(int argc, char *argv[]) {
> int ret, fd, index;
> int in_data;
>
> struct dm_ioctl *param = calloc(1, sizeof(struct dm_ioctl) + 16384);
> struct dm_target_spec sp;
> char *target = (char *)(param + 1);
> char *last_param = target + sizeof(struct dm_target_spec);
>
> struct dm_target_msg tmsg;
> fd = open("/dev/mapper/control", O_RDWR);
> if (fd < 0) {
> printf("Failed to open device /dev/mapper/control: %s\n",
> strerror(errno));
> return -1;
> }
>
> // DM_VERSION
> param->version[0] = 4;
> param->version[1] = 0;
> param->version[2] = 0;
> param->data_size = 16384;
> param->flags = DM_EXISTS_FLAG;
>
> ret = ioctl(fd, DM_VERSION, param);
> if (ret < 0) {
> printf("Failed to send DM_VERSION: %s\n", strerror(errno));
> close(fd);
> return -1;
> }
> printf("DM_VERSION %d %d %d\n", param->version[0],
> param->version[1], param->version[2]);
>
> // DM_DEV_CREATE
> param->data_start = 312;
> strncpy(param->name, "multipath-dev", sizeof(param->name));
>
> // clear
> memset(target, 0, 16384);
>
> param->data_size = 16384;
> param->dev = 0;
>
> ret = ioctl(fd, DM_DEV_CREATE, param);
> if (ret < 0) {
> printf("Failed to send DM_DEV_CREATE: %s\n", strerror(errno));
> close(fd);
> return -1;
> }
>
> // DM_TABLE_LOAD
> memset(param->name, 0, sizeof(param->name));
> param->flags = DM_EXISTS_FLAG;
> param->data_start = 312;
> param->data_size = 16384;
> //param->dev = 64514; // makedev(252, 2)
> param->target_count = 1;
>
> sp.sector_start = 0;
> sp.length = 20971520;
> strncpy(sp.target_type, "multipath", sizeof(sp.target_type) - 1);
> sp.target_type[sizeof(sp.target_type) - 1] = '\0';
>
> memcpy(target, &sp, sizeof(struct dm_target_spec));
> strcpy(last_param, "1 queue_if_no_path 0 1 1 service-time 0 1 1
> 8:48 1");
>
> ret = ioctl(fd, DM_TABLE_LOAD, param);
> // if (ret < 0) {
> // printf("Failed to send DM_TABLE_LOAD: %s\n", strerror(errno));
> // close(fd);
> // return -1;
> // }
>
> printf("delay before second table_load...\n");
> sleep(10);
> printf("delay done\n");
> memset(param->name, 0, sizeof(param->name));
> param->flags = DM_EXISTS_FLAG;
> param->data_start = 312;
> param->data_size = 16384;
> param->target_count = 1;
>
> sp.sector_start = 0;
> sp.length = 20971520;
> strncpy(sp.target_type, "multipath", sizeof(sp.target_type) - 1);
> sp.target_type[sizeof(sp.target_type) - 1] = '\0';
>
> memcpy(target, &sp, sizeof(struct dm_target_spec));
> strcpy(last_param, "1 queue_if_no_path 0 1 1 service-time 0 1 1
> 8:48 1");
>
> ret = ioctl(fd, DM_TABLE_LOAD, param);
> if (ret < 0) {
> printf("Failed to send DM_TABLE_LOAD: %s\n", strerror(errno));
> close(fd);
> return -1;
> }
>
> // DM_DEV_SUSPEND
> param->flags = DM_EXISTS_FLAG; // continue creating
> param->event_nr = 6345265;
> param->data_size = 16384;
> param->dev = 0;
> strncpy(param->name, "multipath-dev", sizeof(param->name));
>
> ret = ioctl(fd, DM_DEV_SUSPEND, param);
> if (ret < 0) {
> printf("Failed to send DM_DEV_SUSPEND: %s\n", strerror(errno));
> close(fd);
> return -1;
> }
> printf("DM_DEV_SUSPEND sent successfully\n");
> close(fd);
>
> return 0;
> }
>
>
> .
Hi, 在 2025/07/31 9:28, Hou Tao 写道: > cc +linux-block > > The disk->queue_kobj is initialized in add_disk() and is uninitialized > in del_disk(). And it seems that blk_unregister_queue() in del_disk() > doesn't uninitialize the queue_kobj completely, because it doesn't > expect the queue_kobj will be added again. I think the right place to > fix the problem is blk_unregister_queue(). How about just memset the > queue_kobj as zero after deleting the queue_kobj in blk_unregister_queue() ? I think it's better to move kobject_init to allocate disk: 1) alloc disk, kobject_init 2) register q, kobject_add 3) unregister q, kobject_del 4) free disk, kobject_put With this, it's safe to add the kobject again. Thanks, Kuai
© 2016 - 2025 Red Hat, Inc.