nvme: enable char device per namespace

[Javier González]

On 12.01.2021 18:22, Minwoo Im wrote:
>Hello Javier,
>
>I tested this patch based on nvme-5.11:
>
>[    1.219747] BUG: unable to handle page fault for address: 0000000100000041
>[    1.220518] #PF: supervisor read access in kernel mode
>[    1.220582] #PF: error_code(0x0000) - not-present page
>[    1.220582] PGD 0 P4D 0
>[    1.220582] Oops: 0000 [#1] SMP PTI
>[    1.220582] CPU: 0 PID: 7 Comm: kworker/u16:0 Not tainted 5.11.0-rc1+ #46
>[    1.220582] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
>[    1.220582] Workqueue: nvme-wq nvme_scan_work
>[    1.220582] RIP: 0010:nvme_ns_id_attrs_are_visible+0x10f/0x152
>[    1.220582] Code: 81 7d d0 80 f9 a1 82 74 0a 48 81 7d d0 a0 f9 a1 82 75 50 48 8b 45 e8 48 89 45 f8 48 8b 45 f8 48 83 e8 60 48 8b 80 60 03 00 00 <48> 8b 40 40 48 3d e0 d1 4d 82 74 07 b8 00 00 00 00 eb 2e 48 8b 45
>[    1.220582] RSP: 0000:ffffc90000047b70 EFLAGS: 00010282
>[    1.220582] RAX: 0000000100000001 RBX: ffffffff824ddb20 RCX: 0000000000000124
>[    1.220582] RDX: ffff8881026eac00 RSI: ffffffff82a1f980 RDI: ffff888102745058
>[    1.220582] RBP: ffffc90000047ba8 R08: ffff888102948718 R09: 0000000000000000
>[    1.220582] R10: 0000000000000000 R11: ffff888100465080 R12: ffff888102745058
>[    1.220582] R13: ffff888102948600 R14: 0000000000000000 R15: ffffffff82a1f548
>[    1.220582] FS:  0000000000000000(0000) GS:ffff88842fc00000(0000) knlGS:0000000000000000
>[    1.220582] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>[    1.220582] CR2: 0000000100000041 CR3: 000000000280c001 CR4: 0000000000370ef0
>[    1.220582] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
>[    1.220582] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
>[    1.220582] Call Trace:
>[    1.220582]  internal_create_group+0xde/0x390
>[    1.220582]  internal_create_groups.part.4+0x3e/0xa0
>[    1.220582]  device_add+0x3cf/0x830
>[    1.220582]  ? cdev_get+0x20/0x20
>[    1.220582]  ? cdev_purge+0x60/0x60
>[    1.220582]  cdev_device_add+0x44/0x70
>[    1.220582]  ? cdev_init+0x50/0x60
>[    1.220582]  nvme_alloc_chardev_ns+0x187/0x1eb
>[    1.220582]  nvme_alloc_ns+0x367/0x460
>[    1.220582]  nvme_validate_or_alloc_ns+0xe2/0x139
>[    1.220582]  nvme_scan_ns_list+0x113/0x17a
>[    1.220582]  nvme_scan_work+0xa5/0x106
>[    1.220582]  process_one_work+0x1dd/0x3e0
>[    1.220582]  worker_thread+0x2d/0x3b0
>[    1.220582]  ? cancel_delayed_work+0x90/0x90
>[    1.220582]  kthread+0x117/0x130
>[    1.220582]  ? kthread_park+0x90/0x90
>[    1.220582]  ret_from_fork+0x22/0x30
>[    1.220582] Modules linked in:
>[    1.220582] CR2: 0000000100000041
>[    1.220582] ---[ end trace b1f509a1bbfbc113 ]---
>[    1.220582] RIP: 0010:nvme_ns_id_attrs_are_visible+0x10f/0x152
>[    1.220582] Code: 81 7d d0 80 f9 a1 82 74 0a 48 81 7d d0 a0 f9 a1 82 75 50 48 8b 45 e8 48 89 45 f8 48 8b 45 f8 48 83 e8 60 48 8b 80 60 03 00 00 <48> 8b 40 40 48 3d e0 d1 4d 82 74 07 b8 00 00 00 00 eb 2e 48 8b 45
>[    1.220582] RSP: 0000:ffffc90000047b70 EFLAGS: 00010282
>[    1.220582] RAX: 0000000100000001 RBX: ffffffff824ddb20 RCX: 0000000000000124
>[    1.220582] RDX: ffff8881026eac00 RSI: ffffffff82a1f980 RDI: ffff888102745058
>[    1.220582] RBP: ffffc90000047ba8 R08: ffff888102948718 R09: 0000000000000000
>[    1.220582] R10: 0000000000000000 R11: ffff888100465080 R12: ffff888102745058
>[    1.220582] R13: ffff888102948600 R14: 0000000000000000 R15: ffffffff82a1f548
>[    1.220582] FS:  0000000000000000(0000) GS:ffff88842fc00000(0000) knlGS:0000000000000000
>[    1.220582] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>[    1.220582] CR2: 0000000100000041 CR3: 000000000280c001 CR4: 0000000000370ef0
>[    1.220582] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
>[    1.220582] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
>
>And this happens when CONFIG_NVME_MULTIPATH=y configured.  Please refere
>attached log up there :)
>
>Thanks!

I have not implemented multipath support, but it should definitely not
crash like this. I'll rebase to 5.11 and test.

Christoph: Is it OK to send this without multipath or should I just send
all together?

Javier