From: Wen Congyang <wency@cn.fujitsu.com>
To: Jan Kiszka <jan.kiszka@siemens.com>
Cc: qemu-devel <qemu-devel@nongnu.org>,
kvm list <kvm@vger.kernel.org>,
Markus Armbruster <armbru@redhat.com>
Subject: Re: [Qemu-devel] pci-assign can not work
Date: Fri, 16 Mar 2012 16:38:36 +0800 [thread overview]
Message-ID: <4F62FC0C.9030302@cn.fujitsu.com> (raw)
In-Reply-To: <4F62F97C.5080805@siemens.com>
At 03/16/2012 04:27 PM, Jan Kiszka Wrote:
> On 2012-03-16 03:38, Wen Congyang wrote:
>> At 03/15/2012 06:21 PM, Wen Congyang Wrote:
>>> Hi all
>>>
>>> When I use pci-assign, I meet the following error:
>>>
>>> Failed to assign irq for "hostdev0": Input/output error
>>> Perhaps you are assigning a device that shares an IRQ with another device?
>>>
>>> Is it a bug or I miss something?
>>
>> Hi, Jan
>>
>> This problem is caused by your patch:
>> commit 6919115a8715c34cd80baa08422d90496f11f5d7
>> Author: Jan Kiszka <jan.kiszka@siemens.com>
>> Date: Thu Mar 8 11:10:27 2012 +0100
>>
>> pci_assign: Flip defaults of prefer_msi and share_intx
>>
>> INTx sharing is a bit more expensive than exclusive host interrupts, but
>> this channel is not supposed to be used for high-performance scenarios
>> anyway. Modern devices support MSI/MSI-X and do not depend on using INTx
>> under critical workload, real old devices do not support INTx sharing
>> anyway.
>>
>> For those in the middle, the user experience is much better if they just
>> work even when IRQ sharing is required. If there is nothing to share,
>> share_intx=off can still be applied as tuning parameter.
>>
>> With INTx sharing as default, the primary reason for prefer_msi=on is
>> gone. Make it default off, specifically as it is known to cause troubles
>> with devices that have incomplete/broken MSI support or otherwise
>> stumble if host IRQ configuration does not match guest driver
>> expectation.
>>
>> Acked-by: Alex Williamson <alex.williamson@redhat.com>
>> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
>> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
>>
>> If I revert this commit. qemu can work.
>>
>
> This should be "solvable" by passing prefer_msi=on to the pci-assign
> device, or likely by updating your host kernel to latest kvm.git (to
> enable INTx sharing).
Is there some way to find out if the kernel supports to enable INTx
sharing?
Thanks
Wen Congyang
>
> Hmm, unfortunate. We needed a conditional default for the prefer_msi
> property here. If INTx sharing doesn't work for some reason AND the user
> did not ask for disabling the host-side MSI usage, we should fall back
> to it again.
>
> Markus, is there some easy way to find out if a specific qdev property
> was set due to a command line switch or was defined by the default value?
>
> Jan
>
next prev parent reply other threads:[~2012-03-16 8:37 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-03-15 10:21 [Qemu-devel] pci-assign can not work Wen Congyang
2012-03-16 2:38 ` Wen Congyang
2012-03-16 8:27 ` Jan Kiszka
2012-03-16 8:38 ` Wen Congyang [this message]
2012-03-16 10:13 ` Jan Kiszka
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=4F62FC0C.9030302@cn.fujitsu.com \
--to=wency@cn.fujitsu.com \
--cc=armbru@redhat.com \
--cc=jan.kiszka@siemens.com \
--cc=kvm@vger.kernel.org \
--cc=qemu-devel@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).