Re: [Linuxarm] Re: [PATCH for-next 00/32] spin lock usage optimization for SCSI drivers

From: Xiaofei Tan
Date: Wed Feb 24 2021 - 21:38:16 EST


Hi Geert,

On 2021/2/24 17:41, Geert Uytterhoeven wrote:
Hi Xiaofei,

On Sun, Feb 7, 2021 at 12:46 PM Xiaofei Tan <tanxiaofei@xxxxxxxxxx> wrote:
Replace spin_lock_irqsave with spin_lock in hard IRQ of SCSI drivers.
There are no function changes, but may speed up if interrupt happen
too often.

I'll bite: how much does this speed up interrupt processing?
What's the typical cost of saving/disabling, and restoring interrupt
state?

It could only take a few CPU cycles. So there is little benefit for speeding up interrupt processing.You could take them as cleanup.

Is removing this cost worth the risk of introducing subtle
regressions on platforms you cannot test yourself?


Currently, only found M68K platform support that high-priority interrupt preempts low-priority. No other platform has such services. Therefore, these changes do not affect non-M68K platforms.

For M68K platform, no one report such interrupt preemption case in these SCSI drivers.

BTW, how many of these legacy SCSI controllers do you have access to?


Actually, no.

Thanks for your answers!

Gr{oetje,eeting}s,

Geert