Re: [PATCH] x86/PCI: Fully disable devices before releasing IRQ resource

["Rafael J. Wysocki" <rjw@rjwysocki.net>]

On Thursday, March 12, 2015 09:41:21 AM Jiang Liu wrote:
> On 2015/3/12 9:17, Rafael J. Wysocki wrote:
> > On Wednesday, March 11, 2015 10:04:42 PM Luck, Tony wrote:
> >>>> Unfortunately there's a long standing comment in pci_device_remove():
> >>>>
> >>>>         /*
> >>>>          * We would love to complain here if pci_dev->is_enabled is set, that
> >>>>          * the driver should have called pci_disable_device(), but the
> >>>>          * unfortunate fact is there are too many odd BIOS and bridge setups
> >>>>          * that don't like drivers doing that all of the time.
> >>>>          * Oh well, we can dream of sane hardware when we sleep, no matter how
> >>>>          * horrible the crap we have to deal with is when we are awake...
> >>>>          */
> >>>>
> >>>> So, unless we can somehow ignore that comment, I suspect forcing the
> >>>> device to be disabled on driver remove, whether done from pci-core or
> >>>> from x86/pci, is going to cause all sorts of breakage.  Are the
> >>>> expectations set by b4b55cda5874 really valid?  It seems like something
> >>>> needs to be done to allow the IRQ to be automatically re-established on
> >>>> x86 regardless of the driver doing the right thing when releasing the
> >>>> device.  We're still looking at a regression for v4.0 as a result of
> >>>> b4b55cda5874.
> >>>
> >>> In which case we probably should revert commit b4b55cda5874 for the time being.
> >>>
> >>> At least I'd be very nervous about any ad-hoc fixes at this stage of the cycle.
> >>
> >> The comment goes back to the dawn of "git" time ... not sure how much further
> >> back.
> >>
> >> Is this actually still an issue on modern systems?  Maybe we need a black list
> >> or white list to separate the good from bad systems?
> > 
> > The answer to that is "We don't know" and in my not so humble opinion it is too
> > risky to try to find out at the end of the cycle.
> Hi Rafael and Alex,
> How about a patch which:
> 1) gives a warning if PCI device is still enabled when unloading driver

That may become sort of noisy.  I really would prefer to introduce things like
that by the beginning of the cycle, not by the end of it.

> 2) release PCI interrupt only if PCI device is disabled.
> By this way, we could support IOAPIC hot-removal on latest platforms and
> avoid regressions on old platforms.

Well, please submit a patch for discussion.

I would like to know Bjorn's opinion about that too at least.

Rafael