MSI allocation regression, still to be corrected in HEAD and please MFC before release/12.0 gets branched

Harry Schmalzbauer freebsd at omnilan.de
Mon Nov 12 17:03:33 UTC 2018


Am 11.06.2018 um 20:28 schrieb Harry Schmalzbauer:
> Am 05.06.2018 um 19:54 schrieb Scott Long:
>>>>>> Late in the 11.2 phase, I identified this commit as a regression 
>>>>> for MSI (non-x) alloctaion.
>>>>> I have an idea what probably causes the problem here (INTx 
>>>>> allocation, although MSI (and MSI-x) capability):
>>>>> disable_msix is not 0 (I need to disable MSI-x because of 
>>>>> ESXi-passthru…).
>>>>>
>>>>> Corresponding lines:
>>>>> {
>>>>>          device_t dev;
>>>>>          int error, msgs;
>>>>>
>>>>>          dev = sc->mps_dev;
>>>>>          error = 0;
>>>>>          msgs = 0;
>>>>>
>>>>>          if ((sc->disable_msix == 0) &&
>>>>>              ((msgs = pci_msix_count(dev)) >= MPS_MSI_COUNT))
>>>>>                  error = mps_alloc_msix(sc, MPS_MSI_COUNT);
>>>>>          if ((error != 0) && (sc->disable_msi == 0) &&
>>>>>              ((msgs = pci_msi_count(dev)) >= MPS_MSI_COUNT))
>>>>>                  error = mps_alloc_msi(sc, MPS_MSI_COUNT);
>>>>>          if (error != 0)
>>>>>                  msgs = 0;
>>>>>
>>>>>          sc->msi_msgs = msgs;
>>>>>          return (error);
>>>>> }
>>>>>
…
>>>> Hi Harry,
>>>> You are correct about the bug.  Please change the line at the top 
>>>> of the function that reads
>>>> error = 0;
>>>> to
>>>> error = ENXIO;
>>>> Let me know if that fixes the MSI problem for you.
>
>>
…
> Index: src/sys/dev/mps/mps_pci.c
> ===================================================================
> --- sys/dev/mps/mps_pci.c   (Revision 334948)
> +++ sys/dev/mps/mps_pci.c   (Arbeitskopie)
> @@ -244,7 +244,7 @@
>         int error, msgs;
>
>         dev = sc->mps_dev;
> -       error = 0;
> +       error = ENXIO;
>         msgs = 0;
>
>         if ((sc->disable_msix == 0) &&
>

To my understanding, it's obvious that the way 
mps_pci_alloc_interrupts() currently works is unintended.
This might not affect too many people, but is there a reason not to fix it?

I already created a coresponding problem report: 
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=229267
Anything else I should do?

Thanks,

-harry



More information about the freebsd-scsi mailing list