Re: [RFC] How to fix an async scan - rmmod race?

[Mike Christie <michaelc@cs.wisc.edu>]

On 04/11/2012 01:17 PM, Mike Christie wrote:
> For example FC drivers will call fc_remove_host which sets things up so
> no new IO can be sent to the device after that function has completed
> and will call some callouts like dev_loss_tmo_callbk to cleanup
> outstanding IO. The problem is that when it calls scsi_remove_host that

One correction. I meant to write scsi_target_remove above.

fc_remove_host sets the rport state then calls rport_delete_work which
then calls scsi_target_remove and it does callbacks like
dev_loss_tmo_callbk and terminate_rport_io. So scsi_remove_host would
not cause commands like sync cache to fail. That would be
scsi_target_remove. By the time the driver calls scsi_remove_host the
scsi_target_remove calls would have already removed the devices.

Same thing happens though. Because fc_remove_host sets the rport state
to FC_PORTSTATE_DELETED when scsi_remove_target removes the devices and
the sd.c shutdown code is run, IO from that would get failed by the fc
check ready function.


> could cause commands like cache sync from sd.c's shutdown to get failed
> (the queuecommand function will call the fc check ready function which
> will fail the IO immediately).
>