Re: [Qemu-devel] [qemu-s390x] [PATCH for-4.1 1/2] s390x/pci: add some fallthrough annotations

[Cornelia Huck]

On Tue, 9 Jul 2019 18:55:34 -0400
Collin Walling <address@hidden> wrote:

> On 7/8/19 9:23 AM, Christian Borntraeger wrote:
> > 
> > 
> > On 08.07.19 14:54, Cornelia Huck wrote:  
> >> According to the comment, the bits are supposed to accumulate.
> >>
> >> Reported-by: Stefan Weil <address@hidden>
> >> Fixes: 5d1abf234462 ("s390x/pci: enforce zPCI state checking")
> >> Signed-off-by: Cornelia Huck <address@hidden>  
> > 
> > This patch does not change behaviour, so it is certainly not wrong.
> > 
> > So lets have a look at if the bug report was actually a real bug or
> > just a missing annotation.
> >   
> >> ---
> >>   hw/s390x/s390-pci-inst.c | 2 ++
> >>   1 file changed, 2 insertions(+)
> >>
> >> diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
> >> index 61f30b8e55d2..00235148bed7 100644
> >> --- a/hw/s390x/s390-pci-inst.c
> >> +++ b/hw/s390x/s390-pci-inst.c
> >> @@ -1209,8 +1209,10 @@ int stpcifc_service_call(S390CPU *cpu, uint8_t r1, 
> >> uint64_t fiba, uint8_t ar,
> >>        * FH Enabled bit is set to one in states of ENABLED, BLOCKED or 
> >> ERROR. */
> >>       case ZPCI_FS_ERROR:
> >>           fib.fc |= 0x20;
> >> +        /* fallthrough */  
> > 
> > This is correct, in case of an error we are also blocked.
> >   
> 
> Agreed. This is definitely correct based on our architecture.
> 
> >>       case ZPCI_FS_BLOCKED:
> >>           fib.fc |= 0x40;
> >> +        /* fallthrough */  
> > 
> > I think this is also correct, but  it would be good if Collin could verify.
> >   
> 
> I failed to find anything to support setting the function control
> enabled bit when the function state is in error / blocked. I'm
> assuming this might be some QEMU hack to get things working? I'll have
> to dive further to understand why this was done this way, as it doesn't
> align with how the s390x architecture is documented. It's confusing.

Might this also be a real issue? Not matching the architecture is not a
good sign...

> 
> Functionally, this doesn't change anything... so I'll at least give it
> an ACK for now and investigate this further.

I prefer to hold off on this patch for now, unless we're really sure
that the code does not have a problem here. Actually documenting
something that might be wrong does not sound like the right thing to
do :/

> 
> >>       case ZPCI_FS_ENABLED:
> >>           fib.fc |= 0x80;
> >>           if (pbdev->iommu->enabled) {
> >>  
> > 
> >   
> 
> Acked-by: Collin Walling <address@hidden>
> 
> Side note: is there somewhere that I could access this bug report? :)

It's the build log with the extra warnings in
https://qemu.weilnetz.de/results/build-20190708.txt (referenced in
<address@hidden> on qemu-devel).