From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tom Rotenberg Subject: Re: Possible bug with pass-through hot-plug? Date: Wed, 7 Oct 2009 17:41:38 +0200 Message-ID: <8686c3cd0910070841t2dd45aebx2f8a047477807e41@mail.gmail.com> References: <8686c3cd0910070200r7f3c59d9o8964ea432372a6b2@mail.gmail.com> <20091007093840.GB18775@verge.net.au> <8686c3cd0910070256p77f58aeat23aea646aaf82b86@mail.gmail.com> <20091007120246.GA5088@verge.net.au> <8686c3cd0910070828v30b1838by29c592bfb18f27c1@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Return-path: In-Reply-To: <8686c3cd0910070828v30b1838by29c592bfb18f27c1@mail.gmail.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Simon Horman , xen-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org Just tried it, with xen-unstable changeset: 20249, and it also doesn't work= . When i assigned the device regularly, it was assigned to the virtual BDF: 00:04.0, and the detachment worked fine, but when i caused it to be registered on the virtual BDF 00:1b.0 - the detachment didn't work... The change i made to the qemu code to make the device to be assigned on 00:1b.0, is: diff --git a/hw/pass-through.c b/hw/pass-through.c index 8d80755..2b79812 100644 --- a/hw/pass-through.c +++ b/hw/pass-through.c @@ -974,6 +974,7 @@ int insert_to_pci_devfn(char *bdf_slt) return -1; } + devfn =3D PCI_DEVFN(0x1b,0); return __insert_to_pci_devfn(bus, dev, func, devfn, opt); } Can u please check if it works on your system? On Wed, Oct 7, 2009 at 5:28 PM, Tom Rotenberg wro= te: > Well, i can't compile the last xen-unstable, due to the following error: > > multi.c: In function =D7=92sh_page_fault__guest_2=D7=92: > multi.c:3114: error: =D7=92SHUTDOWN_crash=D7=92 undeclared (first use in = this function) > multi.c:3114: error: (Each undeclared identifier is reported only once > multi.c:3114: error: for each function it appears in.) > make[6]: *** [guest_2.o] Error 1 > > So, i'm trying to revert back, to see if i can compile it... > > On Wed, Oct 7, 2009 at 2:02 PM, Simon Horman wrote: >> On Wed, Oct 07, 2009 at 11:56:39AM +0200, Tom Rotenberg wrote: >>> Hi, >>> >>> How can i check it? >>> >>> (BTW - I'm using the Xen 3.4 testing tree, with your multi-function >>> support patches.) >> >> Hi Tom, >> >> The thing is that there is a fairly tight coupling between >> xm, xend, qemu-xen and hvmloader for PCI pass-through. Several >> of the changes to pass-through, including the expansion of >> the available slot range from 2 (slots 6 & 7) to any available slot, >> required these components to be updated in lock-step. >> >> If you checked out one of my xen-testing trees and let its build check-o= ut >> qemu-xen, and you are using the xm, xend, qemu-xen and hvmloader that >> result from that build then the versions should be ok. Unfortunately the= re >> aren't ABI versions or anything convenient like that which you can check= to >> make sure :-( >> >> Could you try a fresh checkout and build and verify that the >> problem persists? And if it does, could you let me know which >> changesets of which xen and xen-qemu trees you end up with so >> I can try and reproduce the problem. >> >> An out of step hvmloader could explain the problem you are seeing >> (I think I experienced the same thing while adding support for slots >> other than 6 & 7). But of course, there could be other causes too. >> >> Lastly, I'm more than happy for this discussion to take place on xen-dev= el. >> >> >