When 'mtip_block_initialize' fails at 'mtip_hw_get_identify', a series
of cleanup operations will be performed. But when the execution reaches
'put_disk', it will cause refcount underflow. The reason for this error
is that after cleaning 'dd->queue', 'dd->disk->queue' was not set to
null at the same time, which caused repeated cleanup work.
Fix this by set 'dd->disk->queue' to null after cleaning
This log reveals it:
[ 59.590163] refcount_t: underflow; use-after-free.
[ 59.591650] Modules linked in:
[ 59.591867] CPU: 5 PID: 1 Comm: swapper/0 Not tainted 5.12.4-g70e7f0549188-dirty #137
[ 59.592407] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
[ 59.593178] RIP: 0010:refcount_warn_saturate+0x140/0x150
[ 59.593551] Code: 05 d1 3b dd 04 01 e8 af d4 5f ff 0f 0b e9 13 ff ff ff e8 b3 75 73 ff 48 c7 c7 30 31 df 85 c6 05 b4 3b dd 04 01 e8 90 d4 5f ff <0f> 0b e9 f4 fe ff ff 66 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 55
[ 59.594942] RSP: 0000:ffffc90000017918 EFLAGS: 00010286
[ 59.595357] RAX: 0000000000000000 RBX: 0000000000000003 RCX: 0000000000000000
[ 59.595858] RDX: 0000000000000000 RSI: ffffffff8123f301 RDI: 00000000ffffffff
[ 59.596346] RBP: ffffc90000017928 R08: 0000000000000001 R09: 0000000000000001
[ 59.596926] R10: 0000000000000000 R11: 0000000000000001 R12: ffff888105494270
[ 59.597429] R13: ffff888105494270 R14: ffffffff82498b30 R15: 0000000000000000
[ 59.597931] FS: 0000000000000000(0000) GS:ffff88817bd40000(0000) knlGS:0000000000000000
[ 59.598500] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 59.598899] CR2: 0000000000000000 CR3: 000000000642e000 CR4: 00000000000006e0
[ 59.599401] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 59.599900] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 59.600400] Call Trace:
[ 59.600579] kobject_put+0x1b0/0x2e0
[ 59.600839] ? dev_attr_show+0x70/0x70
[ 59.601112] blk_put_queue+0x19/0x20
[ 59.601372] disk_release+0xb7/0xf0
[ 59.601628] ? show_partition_start+0x80/0x80
[ 59.601943] device_release+0x40/0xd0
[ 59.602207] kobject_put+0x10b/0x2e0
[ 59.602468] put_device+0x1f/0x30
[ 59.602708] put_disk+0x2a/0x40
[ 59.602938] mtip_block_initialize+0x35f/0x1570
[ 59.603264] ? __pci_enable_msi_range+0x32c/0x470
[ 59.603606] mtip_pci_probe+0x92a/0xc80
[ 59.603899] local_pci_probe+0x4a/0xb0
[ 59.604173] pci_device_probe+0x126/0x1d0
[ 59.604478] ? pci_device_remove+0x100/0x100
[ 59.604790] really_probe+0x27e/0x650
[ 59.605059] driver_probe_device+0x84/0x1d0
[ 59.605359] ? mutex_lock_nested+0x16/0x20
[ 59.605660] device_driver_attach+0x63/0x70
[ 59.605963] __driver_attach+0x117/0x1a0
[ 59.606247] ? device_driver_attach+0x70/0x70
[ 59.606607] bus_for_each_dev+0xb6/0x110
[ 59.606919] ? rdinit_setup+0x40/0x40
[ 59.607177] driver_attach+0x22/0x30
[ 59.607431] bus_add_driver+0x1e6/0x2a0
[ 59.607703] driver_register+0xa4/0x180
[ 59.607974] __pci_register_driver+0x77/0x80
[ 59.608273] ? drbd_debugfs_init+0x78/0x78
[ 59.608560] mtip_init+0x15c/0x18f
[ 59.608820] do_one_initcall+0x7a/0x3d0
[ 59.609140] ? rdinit_setup+0x40/0x40
[ 59.609464] ? rcu_read_lock_sched_held+0x4a/0x70
[ 59.609879] kernel_init_freeable+0x2a7/0x2f9
[ 59.610268] ? rest_init+0x2c0/0x2c0
[ 59.610561] kernel_init+0x13/0x180
[ 59.610807] ? rest_init+0x2c0/0x2c0
[ 59.611058] ? rest_init+0x2c0/0x2c0
[ 59.611312] ret_from_fork+0x1f/0x30
[ 59.611574] Kernel panic - not syncing: panic_on_warn set ...
[ 59.611973] CPU: 5 PID: 1 Comm: swapper/0 Not tainted 5.12.4-g70e7f0549188-dirty #137
[ 59.612514] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
[ 59.613303] Call Trace:
[ 59.613476] dump_stack+0xba/0xf5
[ 59.613718] ? refcount_warn_saturate+0x140/0x150
[ 59.614055] panic+0x155/0x3ed
[ 59.614281] ? __warn+0xed/0x150
[ 59.614468] ? refcount_warn_saturate+0x140/0x150
[ 59.614468] __warn+0x103/0x150
[ 59.614468] ? refcount_warn_saturate+0x140/0x150
[ 59.614468] report_bug+0x119/0x1c0
[ 59.614468] handle_bug+0x3b/0x80
[ 59.614468] exc_invalid_op+0x18/0x70
[ 59.614468] asm_exc_invalid_op+0x12/0x20
[ 59.614468] RIP: 0010:refcount_warn_saturate+0x140/0x150
[ 59.614468] Code: 05 d1 3b dd 04 01 e8 af d4 5f ff 0f 0b e9 13 ff ff ff e8 b3 75 73 ff 48 c7 c7 30 31 df 85 c6 05 b4 3b dd 04 01 e8 90 d4 5f ff <0f> 0b e9 f4 fe ff ff 66 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 55
[ 59.614468] RSP: 0000:ffffc90000017918 EFLAGS: 00010286
[ 59.614468] RAX: 0000000000000000 RBX: 0000000000000003 RCX: 0000000000000000
[ 59.614468] RDX: 0000000000000000 RSI: ffffffff8123f301 RDI: 00000000ffffffff
[ 59.614468] RBP: ffffc90000017928 R08: 0000000000000001 R09: 0000000000000001
[ 59.614468] R10: 0000000000000000 R11: 0000000000000001 R12: ffff888105494270
[ 59.614468] R13: ffff888105494270 R14: ffffffff82498b30 R15: 0000000000000000
[ 59.614468] ? dev_attr_show+0x70/0x70
[ 59.614468] ? vprintk_func+0x71/0x110
[ 59.614468] ? refcount_warn_saturate+0x140/0x150
[ 59.614468] kobject_put+0x1b0/0x2e0
[ 59.614468] ? dev_attr_show+0x70/0x70
[ 59.614468] blk_put_queue+0x19/0x20
[ 59.614468] disk_release+0xb7/0xf0
[ 59.614468] ? show_partition_start+0x80/0x80
[ 59.614468] device_release+0x40/0xd0
[ 59.614468] kobject_put+0x10b/0x2e0
[ 59.614468] put_device+0x1f/0x30
[ 59.614468] put_disk+0x2a/0x40
[ 59.614468] mtip_block_initialize+0x35f/0x1570
[ 59.614468] ? __pci_enable_msi_range+0x32c/0x470
[ 59.614468] mtip_pci_probe+0x92a/0xc80
[ 59.614468] local_pci_probe+0x4a/0xb0
[ 59.614468] pci_device_probe+0x126/0x1d0
[ 59.614468] ? pci_device_remove+0x100/0x100
[ 59.614468] really_probe+0x27e/0x650
[ 59.614468] driver_probe_device+0x84/0x1d0
[ 59.614468] ? mutex_lock_nested+0x16/0x20
[ 59.614468] device_driver_attach+0x63/0x70
[ 59.614468] __driver_attach+0x117/0x1a0
[ 59.614468] ? device_driver_attach+0x70/0x70
[ 59.614468] bus_for_each_dev+0xb6/0x110
[ 59.614468] ? rdinit_setup+0x40/0x40
[ 59.614468] driver_attach+0x22/0x30
[ 59.614468] bus_add_driver+0x1e6/0x2a0
[ 59.614468] driver_register+0xa4/0x180
[ 59.614468] __pci_register_driver+0x77/0x80
[ 59.614468] ? drbd_debugfs_init+0x78/0x78
[ 59.614468] mtip_init+0x15c/0x18f
[ 59.614468] do_one_initcall+0x7a/0x3d0
[ 59.614468] ? rdinit_setup+0x40/0x40
[ 59.614468] ? rcu_read_lock_sched_held+0x4a/0x70
[ 59.614468] kernel_init_freeable+0x2a7/0x2f9
[ 59.614468] ? rest_init+0x2c0/0x2c0
[ 59.614468] kernel_init+0x13/0x180
[ 59.614468] ? rest_init+0x2c0/0x2c0
[ 59.614468] ? rest_init+0x2c0/0x2c0
[ 59.614468] ret_from_fork+0x1f/0x30
[ 59.614468] Dumping ftrace buffer:
[ 59.614468] (ftrace buffer empty)
[ 59.614468] Kernel Offset: disabled
[ 59.614468] Rebooting in 1 seconds..
Signed-off-by: Zheyu Ma <[email protected]>
---
drivers/block/mtip32xx/mtip32xx.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/block/mtip32xx/mtip32xx.c b/drivers/block/mtip32xx/mtip32xx.c
index 589cb0f1e030..305be06868df 100644
--- a/drivers/block/mtip32xx/mtip32xx.c
+++ b/drivers/block/mtip32xx/mtip32xx.c
@@ -3726,6 +3726,7 @@ static int mtip_block_initialize(struct driver_data *dd)
read_capacity_error:
init_hw_cmds_error:
blk_cleanup_queue(dd->queue);
+ dd->disk->queue = NULL;
block_queue_alloc_init_error:
blk_mq_free_tag_set(&dd->tags);
block_queue_alloc_tag_error:
--
2.17.6
On Sun, Jun 20, 2021 at 12:19:17PM +0000, Zheyu Ma wrote:
> When 'mtip_block_initialize' fails at 'mtip_hw_get_identify', a series
> of cleanup operations will be performed. But when the execution reaches
> 'put_disk', it will cause refcount underflow. The reason for this error
> is that after cleaning 'dd->queue', 'dd->disk->queue' was not set to
> null at the same time, which caused repeated cleanup work.
>
> Fix this by set 'dd->disk->queue' to null after cleaning
This should all be fixed by the switch to blk_cleanup_disk and the
preparation for it in linux-next.
On Mon, Jun 21, 2021 at 2:16 PM Christoph Hellwig <[email protected]> wrote:
>
> On Sun, Jun 20, 2021 at 12:19:17PM +0000, Zheyu Ma wrote:
> > When 'mtip_block_initialize' fails at 'mtip_hw_get_identify', a series
> > of cleanup operations will be performed. But when the execution reaches
> > 'put_disk', it will cause refcount underflow. The reason for this error
> > is that after cleaning 'dd->queue', 'dd->disk->queue' was not set to
> > null at the same time, which caused repeated cleanup work.
> >
> > Fix this by set 'dd->disk->queue' to null after cleaning
>
> This should all be fixed by the switch to blk_cleanup_disk and the
> preparation for it in linux-next.
Thanks for your reply, I have switched to the linux-next branch.
Best regards,
Zheyu Ma