On 24.09.2019 11:20, Paul Durrant wrote: >> -----Original Message----- >> From: Jan Beulich <jbeul...@suse.com> >> Sent: 24 September 2019 10:10 >> To: Paul Durrant <paul.durr...@citrix.com>; xen-devel@lists.xenproject.org >> Cc: Suravee Suthikulpanit <suravee.suthikulpa...@amd.com>; Andrew Cooper >> <andrew.coop...@citrix.com> >> Subject: Re: [Xen-devel] [PATCH v6 7/8] AMD/IOMMU: allocate one device table >> per PCI segment >> >> On 23.09.2019 18:30, Paul Durrant wrote: >>>> -----Original Message----- >>>> From: Xen-devel <xen-devel-boun...@lists.xenproject.org> On Behalf Of Jan >>>> Beulich >>>> Sent: 19 September 2019 14:25 >>>> To: xen-devel@lists.xenproject.org >>>> Cc: Andrew Cooper <andrew.coop...@citrix.com>; Suravee Suthikulpanit >> <suravee.suthikulpa...@amd.com> >>>> Subject: [Xen-devel] [PATCH v6 7/8] AMD/IOMMU: allocate one device table >>>> per PCI segment >>>> >>>> Having a single device table for all segments can't possibly be right. >>> >>> The copy of the spec. I have says (on page 253: Fixed-Length IVHD >>> Blocks) that IVHD entries must have a segment group of 0, so can't >>> the code just require iommu->seg == 0? >> >> The wording in my version is "At this time, only PCI Segment Group 0 is >> supported." This suggests to me that it is not a good idea to have logic >> baked in that depends on this remaining true. I realize though that there >> are more places than just this one where we (have to) assume segment 0 >> (all in iommu_acpi.c, and all marked with an XXX comment). >> > > Ok. Fair enough. I just wasn't sure it was worth doing this change at the > moment;
The thing that I found really disgusting was the global (i.e. not even static) variable "device_table". And simply making it static would still have left things in too ugly a state for my taste. > but it doesn't hurt, so you can add my R-b. Thanks. Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xenproject.org https://lists.xenproject.org/mailman/listinfo/xen-devel
