On Fri, Jan 17, 2014 at 01:52:43PM -0500, Anthony DeRobertis wrote:
> On 01/17/2014 09:08 AM, Neil Horman wrote:
> 
> >If disable_irq_remap had been set by nointremap, that function
> >should have returned immediately, but in test 2 you got the "This
> >system BIOS.." warning, indicating that disable_irq_remap was not
> >set.
> 
> No, that's not right...
> 
> anthony@Zia:/tmp/iommu$ ls -l
> total 216
> -rw-r--r-- 1 root root 71141 Jan 16 14:40 no-patch,vt-d=off
> -rw-r--r-- 1 root root 71449 Jan 16 14:37 no-patch,vt-d=on
> -rw-r--r-- 1 root root 72746 Jan 16 14:54 
> no-patch,vt-d=on,nointremap,with-error
> anthony@Zia:/tmp/iommu$ grep 'system BIOS' *
> no-patch,vt-d=on:[    0.066268] This system BIOS has enabled interrupt 
> remapping
> 
> 
> So, when I look at my copies of the files, its as expected.
> 
> I put the test cases in the order 1, 3, 2 when pasting into the
> email—the labels are right, but that order was rather stupid on my
> part. But you must be looking at the right one, because you see the
> kernel option, so I'm confused.
> 
> I have no idea what the mailing list will do with this attachment,
> and I apologize in advance if you don't like binary attachments, but
> here they are as a .tgz. Should at least take care of this
> confusion.


Ok, thats a bit more clear, thank you.  The order of tests did confuse me
apparently.

So that clarifies the results somewhat.  nointremap does in fact do a software
disable of irq remapping, just like the pci qurik does.  Thats good.  Whats less
good is that you still get the irq vector error with nointremap specified.
Whiel that supports the assertion that there is nothing wrong with the way the
pci quirk disables irq remapping, it does suggest that the listed symptom (the
no vector for irq (-1) error) is being triggered by some other root cause.  And
I am unfortunately unsure of what that might be.  I would suggest you look
through that errata sheet that I referenced in the initial commit to fix the
remapping problem, to see if there are other chip errata who's symptoms lead to
irq drainage failures.  We may need to add another quirk to this chipset.

Regards
Neil

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

Reply via email to