[Crash-utility] Re: [PATCH] x86_64: rhel 9.3: "crash" doesn't handle all IRQ exception properly

Hi Aureau, On Tue, Apr 16, 2024 at 5:14 PM Aureau, Georges (Kernel Tools ERT) <georges.aureau(a)hpe.com&gt; wrote: > > Hello Tao Liu, > > We don't update machdep->machspec->irq_eframe_link as it is now more a reference point for trying various link offsets (+8, -8, +16, -16) around it. > This reference point should be viewed as the most likely link offset, and if we fail, we try around it, but without updating this reference point, > otherwise we lose our reference point and the offsets we would be trying next would have a different meaning. > Different processors may be caught in different IRQ/SYSVEC functions with different framesizes, > so there is no longer a unique (same for all IRQ/SYSVEC functions) link offset, so we can't compute one eframe link for all. > We are adapting the eframe link offset on the fly based on possible IRQ/SYSVEC function framesizes, ie. > most likely -32, then try -40, then -24, then -48, then -16. > If we would update as we find a link, let's say -16, then next time around, > most likely -16, then try -24, then -8, then -32, then 0, eg. we would lose the -40 and -48. > So if we have a most likely eframe link as a reference point, we should not move this point. > The comment in x86_64_irq_eframe_link() for the patch I'm suggesting is saying: > "keep ms->irq_eframe_link as the most likely value, and try a few sizes around it." > See below: Thanks for the explanation, it sounds reasonable. OK I have no other comments for this patch, so ack.