[PATCH] nvme-pci: prevent SK Hynix PE8000 from using Write Zeroes command

[Gopal Tiwari]

Thanks Chaitanya, Sure. 

Regards,
Gopal 


----- Original Message -----
From: "Chaitanya Kulkarni" <Chaitanya.Kulkarni at wdc.com>
To: "Gopal Tiwari" <gtiwari at redhat.com>, hch at lst.de, kbusch at kernel.org, linux-nvme at lists.infradead.org
Sent: Tuesday, February 9, 2021 11:08:03 AM
Subject: Re: [PATCH] nvme-pci: prevent SK Hynix PE8000 from using Write Zeroes command

Gopal,

On 2/8/21 21:06, Gopal Tiwari wrote:
> For SK Hynix PE8000 WriteZerocommands exceed typical I/O MDTS
> (Maximum Data Transfer Size) supported by the hardware. Which
> intern fail and generate following erros :
>
> nvme nvme0: I/O 0 QID 96 timeout, aborting
> nvme nvme0: Abort status: 0x371
> nvme nvme0: Shutdown timeout set to 15 seconds
> nvme nvme0: 128/0/0 default/read/poll queues
> nvme nvme0: I/O 0 QID 96 timeout, disable controller
> nvme nvme0: I/O 0 QID 0 timeout, disable controller
> nvme nvme0: Device shutdown incomplete; abort shutdown
> nvme nvme0: failed to mark controller live state
> nvme nvme0: Removing after probe failure status: -19
>
> Firmware unable to complete any outstanding commands
> due to lack of resources, therefore placing NVMe drive
> in a condition that is unable to further process WriteZerocommands.
>
> fixing by disabling Write zeroes.

Please consider following commit log, can be used at the time applying the
patch:-

For SK Hynix PE8000 NVMe Write Zeroes Command exceed typical I/O MDTS
(Maximum Data Transfer Size) supported by the hardware. That intern
fail and generate following errors :

nvme nvme0: I/O 0 QID 96 timeout, aborting
nvme nvme0: Abort status: 0x371
nvme nvme0: Shutdown timeout set to 15 seconds
nvme nvme0: 128/0/0 default/read/poll queues
nvme nvme0: I/O 0 QID 96 timeout, disable controller
nvme nvme0: I/O 0 QID 0 timeout, disable controller
nvme nvme0: Device shutdown incomplete; abort shutdown
nvme nvme0: failed to mark controller live state
nvme nvme0: Removing after probe failure status: -19

Firmware is unable to complete any outstanding commands due to lack of
resources, therefore placing NVMe drive in a condition that is unable to
further process write-zeores
command.                                                                 

Add NVME_QUIRK_DISABLE_WRITE_ZEROES quirk for SK Hynix PE8000 that fixes
the problem.


With the modified commit log, looks good.

Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni at wdc.com>


_______________________________________________
Linux-nvme mailing list
Linux-nvme at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme