Re: [PATCH] PCI: Add pci shutdown ability

[Alexander Nyberg]

tis 2005-04-26 klockan 11:11 -0400 skrev Alan Stern:
> On Mon, 25 Apr 2005, Andrew Morton wrote:
> 
> > I have vague memories of this being discussed at some length last year. 
> > Nothing comprehensive came of it, except that perhaps the kdump code should
> > spin with irqs off for a couple of seconds so the DMA and IRQs stop.
> 
> Like Pavel said, this won't work.
> 
> > (Ongoing DMA is not a problem actually, because the kdump kernel won't be
> > using that memory anyway)
> 
> For PCI devices at least, DMA _can_ be disabled in a uniform way as 
> devices are discovered.  Some platforms might not want to do this for fear 
> it would kill the initial console display.
> 
> IRQs _cannot_ be disabled in a uniform way.  So they remain a problem.
> 
> > For kdump we really don't want to be doing fancy driver shutdown inside a
> > crashed kernel.  It would be better to just jump to the new kernel and
> > to reset the hardware from there.  DMA doesn't matter, and maybe IRQs can
> > be handled with a sustained local_irq_disable() (hard).
> 
> But at some point you have to enable local IRQs, and then you're in
> trouble if a device without a driver is generating requests.  Unless the 
> new kernel can run with interrupts entirely disabled?  Seems pretty 
> unlikely.

At least on x86 & x64 both i8259A and optional IOAPIC are initially
fully masked until a driver decides to open up a line.

If driver initialization fails then it should never open up the line on
the interrupt controller. So this shouldn't be a problem with interrupts
or am I missing something?
Shared interrupt lines do present a problem however as someone else gets
the chance to scream on an open line ultimately killing it, hmmm.

> The real problem is that, in general, hardware _can't_ be reset properly
> by a new kernel.  There are things that can help, like the PCI USB quirks
> code.  That might be enough to handle the most pressing existing problems;  
> certainly it would avoid the USB issues we've seen.  (But it needs to be
> updated to avoid interfering with normal operations during
> resume-from-disk.)



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/