Re: [PATCH v2 12/15] PCI/MSI: Let PCI host bridges declare their reliance on MSI domains

[Marc Zyngier]

On Wed, 24 Mar 2021 13:19:38 +0000,
Lorenzo Pieralisi <lorenzo.pieralisi@xxxxxxx> wrote:
> 
> On Tue, Mar 23, 2021 at 06:09:36PM +0000, Marc Zyngier wrote:
> > Hi Robin,
> > 
> > On Tue, 23 Mar 2021 11:45:02 +0000,
> > Robin Murphy <robin.murphy@xxxxxxx> wrote:
> > > 
> > > On 2021-03-22 18:46, Marc Zyngier wrote:
> > > > The new 'no_msi' attribute solves the problem of advertising the lack
> > > > of MSI capability for host bridges that know for sure that there will
> > > > be no MSI for their end-points.
> > > > 
> > > > However, there is a whole class of host bridges that cannot know
> > > > whether MSIs will be provided or not, as they rely on other blocks
> > > > to provide the MSI functionnality, using MSI domains.  This is
> > > > the case for example on systems that use the ARM GIC architecture.
> > > > 
> > > > Introduce a new attribute ('msi_domain') indicating that implicit
> > > > dependency, and use this property to set the NO_MSI flag when
> > > > no MSI domain is found at probe time.
> > > > 
> > > > Acked-by: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>
> > > > Signed-off-by: Marc Zyngier <maz@xxxxxxxxxx>
> > > > ---
> > > >   drivers/pci/probe.c | 2 +-
> > > >   include/linux/pci.h | 1 +
> > > >   2 files changed, 2 insertions(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
> > > > index 146bd85c037e..bac9f69a06a8 100644
> > > > --- a/drivers/pci/probe.c
> > > > +++ b/drivers/pci/probe.c
> > > > @@ -925,7 +925,7 @@ static int pci_register_host_bridge(struct pci_host_bridge *bridge)
> > > >   	device_enable_async_suspend(bus->bridge);
> > > >   	pci_set_bus_of_node(bus);
> > > >   	pci_set_bus_msi_domain(bus);
> > > > -	if (bridge->no_msi)
> > > > +	if (bridge->no_msi || (bridge->msi_domain && !bus->dev.msi_domain))
> > > >   		bus->bus_flags |= PCI_BUS_FLAGS_NO_MSI;
> > > >     	if (!parent)
> > > > diff --git a/include/linux/pci.h b/include/linux/pci.h
> > > > index 48605cca82ae..d322d00db432 100644
> > > > --- a/include/linux/pci.h
> > > > +++ b/include/linux/pci.h
> > > > @@ -551,6 +551,7 @@ struct pci_host_bridge {
> > > >   	unsigned int	preserve_config:1;	/* Preserve FW resource setup */
> > > >   	unsigned int	size_windows:1;		/* Enable root bus sizing */
> > > >   	unsigned int	no_msi:1;		/* Bridge has no MSI support */
> > > > +	unsigned int	msi_domain:1;		/* Bridge wants MSI domain */
> > > 
> > > Aren't these really the same thing? Either way we're saying the bridge
> > > itself doesn't handle MSIs, it's just in one case we're effectively
> > > encoding a platform-specific assumption that an external domain won't
> > > be provided. I can't help wondering whether that distinction is really
> > > necessary...
> > 
> > There is a subtle difference: no_msi indicates that there is no way
> > *any* MSI can be dealt with whatsoever (maybe because the RC doesn't
> > forward the corresponding TLPs?). msi_domain says "no MSI unless...".
> > 
> > We could implement the former with the latter, but I have the feeling
> > that's not totally bullet proof. Happy to revisit this if you think it
> > really matters.
> 
> IIUC msi_domain == 1 means: this host bridge needs an msi_domain to enable
> MSIs, which in turn means that there are bridges that do _not_ require
> an msi_domain to enable MSIs. I don't know how other arches handle the 
> msi_domain pointer but I am asking whether making:
> 
> if (bridge->no_msi || !bus->dev.msi_domain))
> 	bus->bus_flags |= PCI_BUS_FLAGS_NO_MSI;
> 
> is a possibility (removing the need for the msi_domain flag).
> 
> At least this looks more like an arch property than a host bridge
> specific property (eg patch [13] pci_host_common_probe() may be used on
> arches other than ARM where it is not necessary true that it requires an
> msi_domain to enable MSIs).
> 
> I agree that's complicated to untangle - just asking if there is way
> to simplify it.

I tried to simplify that in the past (see the original discussion at
[1]), and tglx reported some breakages on systems that do not use MSI
domains, which is why we ended up with an explicit flag.

What I have done for now is to go with Robin's proposal of dropping
'no_msi' and rely on solely on 'msi_domain' to set
PCI_BUS_FLAGS_NO_MSI when no domain is found.

Note that if we indeed have a host bridge that uses
pci_host_common_probe() that doesn't use MSI domains, we may indeed
run into problems. I don't have a good way around that, unfortunately.

Thanks,

	M.

[1] https://lore.kernel.org/r/20201031140330.83768-1-linux@xxxxxxxxx

-- 
Without deviation from the norm, progress is not possible.