Re: [PATCH v2] PCI/MSI: Don't touch MSI bits when the PCI device is disconnected

[Keith Busch]

On Thu, Nov 08, 2018 at 02:42:55PM -0800, Greg Kroah-Hartman wrote:
> On Thu, Nov 08, 2018 at 03:32:58PM -0700, Keith Busch wrote:
> > On Thu, Nov 08, 2018 at 02:01:17PM -0800, Greg Kroah-Hartman wrote:
> > > On Thu, Nov 08, 2018 at 02:09:17PM -0600, Bjorn Helgaas wrote:
> > > > I'm having second thoughts about this.  One thing I'm uncomfortable
> > > > with is that sprinkling pci_dev_is_disconnected() around feels ad hoc
> > > > instead of systematic, in the sense that I don't know how we convince
> > > > ourselves that this (and only this) is the correct place to put it.
> > > 
> > > I think my stance always has been that this call is not good at all
> > > because once you call it you never really know if it is still true as
> > > the device could have been removed right afterward.
> > > 
> > > So almost any code that relies on it is broken, there is no locking and
> > > it can and will race and you will loose.
> > 
> > AIUI, we're not trying to create code to rely on this. This more about
> > reducing reliance on hardware. If the software misses the race once and
> > accesses disconnected device memory, that's usually not a big deal to
> > let hardware sort it out, but the point is not to push our luck.
> 
> Then why even care about this call at all?  If you need to really know
> if the read worked, you have to check the value.  If the value is FF
> then you have a huge hint that the hardware is now gone.  And you can
> rely on it being gone, you can never rely on making the call to the
> function to check if the hardware is there to be still valid any point
> in time after the call returns.
> 
> > Surprise hot remove is empirically more reliable the less we interact
> > with hardware and firmware. That shouldn't be necessary, but is just an
> > unfortunate reality.
> 
> You are not "interacting", you are reading/writing to the hardware, as
> you have to do so.  So I really don't understand what you are talking
> about here, sorry.

We're reading hardware memory, yes, but the hardware isn't there.
Something obviously needs to return FF, so we are indirectly interacting
with whatever mechanism handles that. Sometimes that mechanism doesn't
handle it gracefully and instead of having FF to consider, you have a
machine check rebooting your system.