Cedric,

On Tue, Dec 07 2021 at 18:42, Cédric Le Goater wrote:

This is breaking nvme on pseries but it's probably one of the previous
patches. I haven't figured out what's wrong yet. Here is the oops FYI.

Hrm.

[   32.494562] WARNING: CPU: 26 PID: 658 at kernel/irq/chip.c:210 irq_startup+0x1c0/0x1e0

This complains about a manual enable_irq() on a managed interrupt.

[   32.494575] Modules linked in: ibmvscsi ibmveth scsi_transport_srp bnx2x ipr libata xhci_pci xhci_hcd nvme xts vmx_crypto nvme_core mdio t10_pi libcrc32c dm_mirror dm_region_hash dm_log dm_mod
[   32.494601] CPU: 26 PID: 658 Comm: kworker/26:1H Not tainted 5.16.0-rc4-clg+ #54
[   32.494607] Workqueue: kblockd blk_mq_timeout_work
[   32.494681] NIP [c000000000206f00] irq_startup+0x1c0/0x1e0
[   32.494686] LR [c000000000206df0] irq_startup+0xb0/0x1e0
[   32.494690] Call Trace:
[   32.494692] [c0000018050f38b0] [c000000000206df0] irq_startup+0xb0/0x1e0 (unreliable)
[   32.494699] [c0000018050f38f0] [c00000000020155c] __enable_irq+0x9c/0xb0
[   32.494705] [c0000018050f3950] [c0000000002015d0] enable_irq+0x60/0xc0
[   32.494710] [c0000018050f39d0] [c008000014a54ae8] nvme_poll_irqdisable+0x80/0xc0 [nvme]
[   32.494719] [c0000018050f3a00] [c008000014a55824] nvme_timeout+0x18c/0x420 [nvme]
[   32.494726] [c0000018050f3ae0] [c00000000076e1b8] blk_mq_check_expired+0xa8/0x130
[   32.494732] [c0000018050f3b10] [c0000000007793e8] bt_iter+0xd8/0x120
[   32.494737] [c0000018050f3b60] [c00000000077a34c] blk_mq_queue_tag_busy_iter+0x25c/0x3f0
[   32.494742] [c0000018050f3c20] [c00000000076ffa4] blk_mq_timeout_work+0x84/0x1a0
[   32.494747] [c0000018050f3c70] [c000000000182a78] process_one_work+0x2a8/0x5a0

Confused. I diffed against v1, but could not spot anything except that
properties issue which you found too.

Thanks,

        tglx