[dpdk-dev] [PATCH] eal: forcing IOVA as PA in ppc

David Marchand david.marchand at redhat.com
Wed Jul 31 21:44:56 CEST 2019


On Wed, Jul 31, 2019 at 12:44 PM Takeshi T Yoshimura <TYOS at jp.ibm.com> wrote:
>
> -----David Marchand <david.marchand at redhat.com> wrote: -----
>
> >To: Takeshi Yoshimura <tyos at jp.ibm.com>
> >From: David Marchand <david.marchand at redhat.com>
> >Date: 07/31/2019 06:29PM
> >Cc: dev <dev at dpdk.org>
> >Subject: [EXTERNAL] Re: [dpdk-dev] [PATCH] eal: forcing IOVA as PA in
> >ppc
> >
> >On Wed, Jul 31, 2019 at 5:36 AM Takeshi Yoshimura <tyos at jp.ibm.com>
> >wrote:
> >>
> >> Commit b76fafb174d2 ("eal: fix IOVA mode selection as VA for PCI
> >> drivers") breaks ppc apps with no IOVA configs (or RTE_IOVA_DC)
> >> because of the inconsistency of user's request and the result of
> >> device capability for IOVA mode. I updated the code to force IOVA
> >as
> >> PA in ppc as before because current ppc driver does not support VA
> >> mode.
> >
> >What is the "ppc driver" you are referring to?
> >Do you mean the vfio implementation for ppc?
> >
> >>
> >> Theoretically, ppc can support VA mode, but I suspect that ppc with
> >> VA mode may have performance issues to create a big DMA window
> >> (VA often uses higher addresses than PA). So, I didn't change the
> >> code to check device capability in ppc.
> >
> >Sorry, I am confused.
> >Is VA mode not working at all? or is it a performance issue?
> >
> >
> >--
> >David Marchand
> >
> >
>
> Ah, I overlooked your first question. The "ppc driver" is in
> drivers/bus/pci/linux/pci.c. In pci.c, pci_one_device_iommu_support_va()
> always returns false in ppc. so, the driver does not allow VA mode in ppc.
>
> This return value is passed to rte_pci_probe_one_driver()
> in drivers/bus/pci/pci_common.c. And the error log appeared like:
>
> EAL: PCI device 0000:01:00.0 on NUMA socket 0
> EAL:   probe driver: 144d:a822 spdk_nvme
> EAL:   Expecting 'PA' IOVA mode but current mode is 'VA', not initializing

Can you confirm on which kernel driver this device was bound ?
I'd like to be sure this is the same issue than David.


-- 
David Marchand


More information about the dev mailing list