qemu-ppc
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-ppc] [PULL 15/16] spapr_pci: Add a 64-bit MMIO window


From: David Gibson
Subject: Re: [Qemu-ppc] [PULL 15/16] spapr_pci: Add a 64-bit MMIO window
Date: Wed, 9 Nov 2016 14:42:35 +1100
User-agent: Mutt/1.7.1 (2016-10-04)

On Tue, Nov 08, 2016 at 02:59:30PM +1100, Alexey Kardashevskiy wrote:
> On 08/11/16 12:16, David Gibson wrote:
> > On Fri, Nov 04, 2016 at 04:03:31PM +1100, Alexey Kardashevskiy wrote:
> >> On 17/10/16 13:43, David Gibson wrote:
> >>> On real hardware, and under pHyp, the PCI host bridges on Power machines
> >>> typically advertise two outbound MMIO windows from the guest's physical
> >>> memory space to PCI memory space:
> >>>   - A 32-bit window which maps onto 2GiB..4GiB in the PCI address space
> >>>   - A 64-bit window which maps onto a large region somewhere high in PCI
> >>>     address space (traditionally this used an identity mapping from guest
> >>>     physical address to PCI address, but that's not always the case)
> >>>
> >>> The qemu implementation in spapr-pci-host-bridge, however, only supports a
> >>> single outbound MMIO window, however.  At least some Linux versions expect
> >>> the two windows however, so we arranged this window to map onto the PCI
> >>> memory space from 2 GiB..~64 GiB, then advertised it as two contiguous
> >>> windows, the "32-bit" window from 2G..4G and the "64-bit" window from
> >>> 4G..~64G.
> >>>
> >>> This approach means, however, that the 64G window is not naturally 
> >>> aligned.
> >>> In turn this limits the size of the largest BAR we can map (which does 
> >>> have
> >>> to be naturally aligned) to roughly half of the total window.  With some
> >>> large nVidia GPGPU cards which have huge memory BARs, this is starting to
> >>> be a problem.
> >>>
> >>> This patch adds true support for separate 32-bit and 64-bit outbound MMIO
> >>> windows to the spapr-pci-host-bridge implementation, each of which can
> >>> be independently configured.  The 32-bit window always maps to 2G.. in PCI
> >>> space, but the PCI address of the 64-bit window can be configured (it
> >>> defaults to the same as the guest physical address).
> >>>
> >>> So as not to break possible existing configurations, as long as a 64-bit
> >>> window is not specified, a large single window can be specified.  This
> >>> will appear the same way to the guest as the old approach, although it's
> >>> now implemented by two contiguous memory regions rather than a single one.
> >>>
> >>> For now, this only adds the possibility of 64-bit windows.  The default
> >>> configuration still uses the legacy mode.
> >>
> >>
> >> This breaks migration to QEMU v2.7, the destination reports:
> >>
> >> address@hidden:vmstate_load spapr_pci, spapr_pci
> >> address@hidden:vmstate_load_field_error field "mem_win_size" load
> >> failed, ret = -22
> >> qemu-hostos1: error while loading state for instance 0x0 of device 
> >> 'spapr_pci'
> >> address@hidden:migrate_set_state new state 7
> >> qemu-hostos1: load of migration failed: Invalid argument
> >>
> >>
> >> mem_win_size decreased from 0xf80000000 to 0x80000000.
> >>
> >> I'd think it should be allowed to migrate like this.
> > 
> > AIUI, we don't generally care (upstream) about migration from newer to
> > older qemu, only from older to newer. 
> 
> Older (v2.7.0) to newer (current upstream with -machine pseries-2.7) does
> not work either with the exact same symptom.

Drat.  Ok.. I see why.  I was converting the old style property into
new-style meaning during property parsing, but the "raw" property
value was still being sent and compared in the migration stream.

It would be possible to fix it both ways, by keeping around the "raw"
mem_window_size parameter and having some pre_save / post_load logic
to shuffle the various possibilities.  But that's going to leave ugly
cruft around indefinitely.  I think it's preferable to just bump the
version number and drop those more-trouble-than-they're-worth
VMSTATE_EQUAL fields.

Patch coming shortly.

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

Attachment: signature.asc
Description: PGP signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]