Re: [RFC PATCH] scsi: mvsas: add suspend/resume support

[John Garry <john.garry@linux.dev>]

On 9/5/26 23:12, FVMA wrote:
> The driver has no PM callbacks. After a system suspend the controller 
> returns with its registers at reset values while the driver state 
> assumes a configured chip, so the first command after resume times out 
> and libsas error handling wedges the host.
> 
> Add dev_pm_ops. On suspend, quiesce libsas with sas_suspend_ha(), mask 
> controller interrupts, drain the IRQ and the tasklet, and cancel pending 
> deferred phy events. On resume, reset the ring indices and the STP 
> register set bookkeeping that chip_init() reinitialises in hardware, run 
> chip_init(), re-enable interrupts, report the attached phys through 
> mvs_scan_start() and hand the host back to libsas with sas_resume_ha().
> 
> Signed-off-by: Francois <fvma@protonmail.com <mailto:fvma@protonmail.com>>
> Assisted-by: Claude:claude-fable-5-1
> ---
> Compile-tested against v7.1.13 and current mainline; not yet exercised
> on hardware (88SE9485). 

What is your motivation in posting this change?

You don't even have access to the relevant HW.