[PATCH 01/12] nvme-pci: don't call nvme_init_ctrl_finish from nvme_passthru_end

Christoph Hellwig hch at lst.de
Tue Nov 8 22:26:57 PST 2022


[note to self: the subject should say nvme instead of nvme-pci]

On Wed, Nov 09, 2022 at 04:55:47AM +0200, Sagi Grimberg wrote:
>> nvme_passthrough_end can race with a reset, so we should not call
>> nvme_init_ctrl_finish from here.  Instead just log that the controller
>> capabilities might have changed.
>
> Can you explain here what is the problem caused by calling this from
> here?

I'll add it to the next version:

 - stores to the cels xarray can race
 - the new opal initialization can race

and in general it just updates random controller fields without
a reset, which I'm worried about hitting completely untested code.



More information about the Linux-nvme mailing list