From: David Gibson <david@gibson.dropbear.id.au>
To: Alexey Kardashevskiy <aik@ozlabs.ru>
Cc: qemu-ppc@nongnu.org, qemu-devel@nongnu.org,
Alexander Graf <agraf@suse.de>
Subject: Re: [Qemu-devel] [PATCH qemu] spapr-pci: Make MMIO spacing a machine property and increase it
Date: Wed, 9 Mar 2016 12:04:10 +1100 [thread overview]
Message-ID: <20160309010410.GG22546@voom.fritz.box> (raw)
In-Reply-To: <56DE13DB.7020000@ozlabs.ru>
[-- Attachment #1: Type: text/plain, Size: 2082 bytes --]
On Tue, Mar 08, 2016 at 10:50:51AM +1100, Alexey Kardashevskiy wrote:
> On 03/04/2016 03:13 PM, Alexey Kardashevskiy wrote:
> >On 03/04/2016 02:39 PM, David Gibson wrote:
> >>On Thu, Mar 03, 2016 at 12:42:53PM +1100, Alexey Kardashevskiy wrote:
> >>>The pseries machine supports multiple PHBs. Each PHB's MMIO/IO space is
> >>>mapped to the CPU address space starting at SPAPR_PCI_WINDOW_BASE plus
> >>>some offset which is calculated from PHB's index and
> >>>SPAPR_PCI_WINDOW_SPACING which is defined now as 64GB.
> >>>
> >>>Since the default 32bit DMA window is using first 2GB of MMIO space,
> >>>the amount of MMIO which the PCI devices can actually use is reduced
> >>>to 62GB. This is a problem if the user wants to use devices with
> >>>huge BARs.
> >>>
> >>>For example, 2 PCI functions of a NVIDIA K80 adapter being passed through
> >>>will exceed this limit as they have 16M + 16G + 32M BARs which
> >>>(when aligned) will need 64GB.
> >>>
> >>>This converts SPAPR_PCI_WINDOW_BASE and SPAPR_PCI_WINDOW_SPACING to
> >>>sPAPRMachineState properties. This uses old values for pseries machine
> >>>before 2.6 and increases the spacing to 128GB so MMIO space becomes 126GB.
> >>>
> >>>This changes the default value of sPAPRPHBState::mem_win_size to -1 for
> >>>pseries-2.6 and adds setup to spapr_phb_realize.
> >>>
> >>>Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> >>
> >>So, in theory I dislike the spapr_pci device reaching into the machine
> >>type to get the spacing configuration. But.. I don't know of a better
> >>way to achieve the desired outcome.
> >
> >
> >We could drop @index and spacing; and request the user to specify the MMIO
> >window start (at least) for every additional PHB.
>
> So what is the decision? :)
There isn't one. I really don't know how to handle this, trying to
talk to some people for ideas.
--
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 #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]
next prev parent reply other threads:[~2016-03-09 1:14 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-03-03 1:42 [Qemu-devel] [PATCH qemu] spapr-pci: Make MMIO spacing a machine property and increase it Alexey Kardashevskiy
2016-03-04 3:39 ` David Gibson
2016-03-04 4:13 ` Alexey Kardashevskiy
2016-03-07 23:50 ` Alexey Kardashevskiy
2016-03-09 1:04 ` David Gibson [this message]
2016-03-21 2:15 ` Alexey Kardashevskiy
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20160309010410.GG22546@voom.fritz.box \
--to=david@gibson.dropbear.id.au \
--cc=agraf@suse.de \
--cc=aik@ozlabs.ru \
--cc=qemu-devel@nongnu.org \
--cc=qemu-ppc@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).