| Message ID | B8376D2DEA074F45BA033984477C453E03440296@shsmsx102.ccr.corp.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Mar 30, 2016 at 02:05:28AM +0000, Hao, Xudong wrote: > > -----Original Message----- > > From: Wei Liu [mailto:wei.liu2@citrix.com] > > Sent: Wednesday, March 30, 2016 12:58 AM > > To: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > > Cc: Hao, Xudong <xudong.hao@intel.com>; wei.liu2@citrix.com; > > samuel.thibault@ens-lyon.org; stefano.stabellini@eu.citrix.com; xen- > > devel@lists.xen.org > > Subject: Re: [Xen-devel] pv-grub guest booting fail with recent qemu-xen > > > > On Mon, Mar 28, 2016 at 09:21:14AM -0400, Konrad Rzeszutek Wilk wrote: > > > On Mon, Mar 28, 2016 at 02:03:35AM +0000, Hao, Xudong wrote: > > > > > -----Original Message----- > > > > > From: Xen-devel [mailto:xen-devel-bounces@lists.xen.org] On Behalf > > > > > Of Konrad Rzeszutek Wilk > > > > > Sent: Saturday, March 26, 2016 2:58 AM > > > > > To: Hao, Xudong <xudong.hao@intel.com> > > > > > Cc: stefano.stabellini@eu.citrix.com; xen-devel@lists.xen.org > > > > > Subject: Re: [Xen-devel] pv-grub guest booting fail with recent > > > > > qemu-xen > > > > > > > > > > On Wed, Mar 02, 2016 at 07:16:40AM +0000, Hao, Xudong wrote: > > > > > > Hi, > > > > > > For Xen upstream master branch with commit 1949868d, After > > > > > > updating qemu- > > > > > xen version from fcf6ac57 to 2ce1d30e, booting a pv-grub guest will fail. > > > > > > > pv-grub should be using qemu-traditional, not qemu-xen > > > > Never hear this limitation. > > > The log message you posted in your original post doesn't seem to reveal much. > > Can you have a look at relevant QEMU logs under /var/log/xen? > > There is not valuable qemu log, only one line: "qemu: terminating on signal 1 from pid 36642". > > Bisect and the bad commit of qemu-xen is: If you use PV-grub without the framebuffer does it boot? > > commit 2ce1d30ef2858dfed72a281872579e5a26b090dd > Author: Stefano Stabellini <stefano.stabellini@eu.citrix.com> > Date: Wed Jan 6 16:32:22 2016 +0000 > > xenfb.c: avoid expensive loops when prod <= out_cons > > If the frontend sets out_cons to a value higher than out_prod, it will > cause xenfb_handle_events to loop about 2^32 times. Avoid that by using > better checks at the beginning of the function. > > upstream-commit-id: ac0487e1d2ae811cd4d035741a109a4ecfb013f1 > > Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> > Reported-by: Ling Liu <liuling-it@360.cn> > > diff --git a/hw/display/xenfb.c b/hw/display/xenfb.c > index 4e2a27a..8eb3046 100644 > --- a/hw/display/xenfb.c > +++ b/hw/display/xenfb.c > @@ -789,8 +789,9 @@ static void xenfb_handle_events(struct XenFB *xenfb) > > prod = page->out_prod; > out_cons = page->out_cons; > - if (prod == out_cons) > - return; > + if (prod - out_cons >= XENFB_OUT_RING_LEN) { > + return; > + } > xen_rmb(); /* ensure we see ring contents up to prod */ > for (cons = out_cons; cons != prod; cons++) { > union xenfb_out_event *event = &XENFB_OUT_RING_REF(page, cons); >
> -----Original Message----- > From: Xen-devel [mailto:xen-devel-bounces@lists.xen.org] On Behalf Of Konrad > Rzeszutek Wilk > Sent: Friday, April 1, 2016 11:55 PM > To: Hao, Xudong <xudong.hao@intel.com> > Cc: samuel.thibault@ens-lyon.org; xen-devel@lists.xen.org; Wei Liu > <wei.liu2@citrix.com>; stefano.stabellini@eu.citrix.com > Subject: Re: [Xen-devel] pv-grub guest booting fail with recent qemu-xen > > On Wed, Mar 30, 2016 at 02:05:28AM +0000, Hao, Xudong wrote: > > > -----Original Message----- > > > From: Wei Liu [mailto:wei.liu2@citrix.com] > > > Sent: Wednesday, March 30, 2016 12:58 AM > > > To: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > > > Cc: Hao, Xudong <xudong.hao@intel.com>; wei.liu2@citrix.com; > > > samuel.thibault@ens-lyon.org; stefano.stabellini@eu.citrix.com; xen- > > > devel@lists.xen.org > > > Subject: Re: [Xen-devel] pv-grub guest booting fail with recent > > > qemu-xen > > > > > > On Mon, Mar 28, 2016 at 09:21:14AM -0400, Konrad Rzeszutek Wilk wrote: > > > > On Mon, Mar 28, 2016 at 02:03:35AM +0000, Hao, Xudong wrote: > > > > > > -----Original Message----- > > > > > > From: Xen-devel [mailto:xen-devel-bounces@lists.xen.org] On > > > > > > Behalf Of Konrad Rzeszutek Wilk > > > > > > Sent: Saturday, March 26, 2016 2:58 AM > > > > > > To: Hao, Xudong <xudong.hao@intel.com> > > > > > > Cc: stefano.stabellini@eu.citrix.com; xen-devel@lists.xen.org > > > > > > Subject: Re: [Xen-devel] pv-grub guest booting fail with > > > > > > recent qemu-xen > > > > > > > > > > > > On Wed, Mar 02, 2016 at 07:16:40AM +0000, Hao, Xudong wrote: > > > > > > > Hi, > > > > > > > For Xen upstream master branch with commit 1949868d, After > > > > > > > updating qemu- > > > > > > xen version from fcf6ac57 to 2ce1d30e, booting a pv-grub guest will > fail. > > > > > > > > > > pv-grub should be using qemu-traditional, not qemu-xen > > > > > > > Never hear this limitation. > > > > > The log message you posted in your original post doesn't seem to reveal > much. > > > Can you have a look at relevant QEMU logs under /var/log/xen? > > > > There is not valuable qemu log, only one line: "qemu: terminating on signal 1 > from pid 36642". > > > > Bisect and the bad commit of qemu-xen is: > > If you use PV-grub without the framebuffer does it boot? How to disable framebuffer when VM booting. Reverted this patch, PV-grub guest boot successfully. > > > > > > commit 2ce1d30ef2858dfed72a281872579e5a26b090dd > > Author: Stefano Stabellini <stefano.stabellini@eu.citrix.com> > > Date: Wed Jan 6 16:32:22 2016 +0000 > > > > xenfb.c: avoid expensive loops when prod <= out_cons > > > > If the frontend sets out_cons to a value higher than out_prod, it will > > cause xenfb_handle_events to loop about 2^32 times. Avoid that by using > > better checks at the beginning of the function. > > > > upstream-commit-id: ac0487e1d2ae811cd4d035741a109a4ecfb013f1 > > > > Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> > > Reported-by: Ling Liu <liuling-it@360.cn> > > > > diff --git a/hw/display/xenfb.c b/hw/display/xenfb.c index > > 4e2a27a..8eb3046 100644 > > --- a/hw/display/xenfb.c > > +++ b/hw/display/xenfb.c > > @@ -789,8 +789,9 @@ static void xenfb_handle_events(struct XenFB > > *xenfb) > > > > prod = page->out_prod; > > out_cons = page->out_cons; > > - if (prod == out_cons) > > - return; > > + if (prod - out_cons >= XENFB_OUT_RING_LEN) { > > + return; > > + } > > xen_rmb(); /* ensure we see ring contents up to prod */ > > for (cons = out_cons; cons != prod; cons++) { > > union xenfb_out_event *event = &XENFB_OUT_RING_REF(page, > > cons); > > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel
On Tue, Apr 05, 2016 at 01:26:47AM +0000, Hao, Xudong wrote: > > -----Original Message----- > > From: Xen-devel [mailto:xen-devel-bounces@lists.xen.org] On Behalf Of Konrad > > Rzeszutek Wilk > > Sent: Friday, April 1, 2016 11:55 PM > > To: Hao, Xudong <xudong.hao@intel.com> > > Cc: samuel.thibault@ens-lyon.org; xen-devel@lists.xen.org; Wei Liu > > <wei.liu2@citrix.com>; stefano.stabellini@eu.citrix.com > > Subject: Re: [Xen-devel] pv-grub guest booting fail with recent qemu-xen > > > > On Wed, Mar 30, 2016 at 02:05:28AM +0000, Hao, Xudong wrote: > > > > -----Original Message----- > > > > From: Wei Liu [mailto:wei.liu2@citrix.com] > > > > Sent: Wednesday, March 30, 2016 12:58 AM > > > > To: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > > > > Cc: Hao, Xudong <xudong.hao@intel.com>; wei.liu2@citrix.com; > > > > samuel.thibault@ens-lyon.org; stefano.stabellini@eu.citrix.com; xen- > > > > devel@lists.xen.org > > > > Subject: Re: [Xen-devel] pv-grub guest booting fail with recent > > > > qemu-xen > > > > > > > > On Mon, Mar 28, 2016 at 09:21:14AM -0400, Konrad Rzeszutek Wilk wrote: > > > > > On Mon, Mar 28, 2016 at 02:03:35AM +0000, Hao, Xudong wrote: > > > > > > > -----Original Message----- > > > > > > > From: Xen-devel [mailto:xen-devel-bounces@lists.xen.org] On > > > > > > > Behalf Of Konrad Rzeszutek Wilk > > > > > > > Sent: Saturday, March 26, 2016 2:58 AM > > > > > > > To: Hao, Xudong <xudong.hao@intel.com> > > > > > > > Cc: stefano.stabellini@eu.citrix.com; xen-devel@lists.xen.org > > > > > > > Subject: Re: [Xen-devel] pv-grub guest booting fail with > > > > > > > recent qemu-xen > > > > > > > > > > > > > > On Wed, Mar 02, 2016 at 07:16:40AM +0000, Hao, Xudong wrote: > > > > > > > > Hi, > > > > > > > > For Xen upstream master branch with commit 1949868d, After > > > > > > > > updating qemu- > > > > > > > xen version from fcf6ac57 to 2ce1d30e, booting a pv-grub guest will > > fail. > > > > > > > > > > > > > pv-grub should be using qemu-traditional, not qemu-xen > > > > > > > > > > Never hear this limitation. > > > > > > > The log message you posted in your original post doesn't seem to reveal > > much. > > > > Can you have a look at relevant QEMU logs under /var/log/xen? > > > > > > There is not valuable qemu log, only one line: "qemu: terminating on signal 1 > > from pid 36642". > > > > > > Bisect and the bad commit of qemu-xen is: > > > > If you use PV-grub without the framebuffer does it boot? > > How to disable framebuffer when VM booting. > Reverted this patch, PV-grub guest boot successfully. > It could be the frontend that's buggy. I think mini-os's fbfront might aggressively overwrites prod without checking if there is space available. We should probably fix mini-os. On the other hand, can this check in QEMU be improved a bit so that it accommodate buggy frontend? Anthony? Wei. > > > > > > > > > > commit 2ce1d30ef2858dfed72a281872579e5a26b090dd > > > Author: Stefano Stabellini <stefano.stabellini@eu.citrix.com> > > > Date: Wed Jan 6 16:32:22 2016 +0000 > > > > > > xenfb.c: avoid expensive loops when prod <= out_cons > > > > > > If the frontend sets out_cons to a value higher than out_prod, it will > > > cause xenfb_handle_events to loop about 2^32 times. Avoid that by using > > > better checks at the beginning of the function. > > > > > > upstream-commit-id: ac0487e1d2ae811cd4d035741a109a4ecfb013f1 > > > > > > Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> > > > Reported-by: Ling Liu <liuling-it@360.cn> > > > > > > diff --git a/hw/display/xenfb.c b/hw/display/xenfb.c index > > > 4e2a27a..8eb3046 100644 > > > --- a/hw/display/xenfb.c > > > +++ b/hw/display/xenfb.c > > > @@ -789,8 +789,9 @@ static void xenfb_handle_events(struct XenFB > > > *xenfb) > > > > > > prod = page->out_prod; > > > out_cons = page->out_cons; > > > - if (prod == out_cons) > > > - return; > > > + if (prod - out_cons >= XENFB_OUT_RING_LEN) { > > > + return; > > > + } > > > xen_rmb(); /* ensure we see ring contents up to prod */ > > > for (cons = out_cons; cons != prod; cons++) { > > > union xenfb_out_event *event = &XENFB_OUT_RING_REF(page, > > > cons); > > > > > > > _______________________________________________ > > Xen-devel mailing list > > Xen-devel@lists.xen.org > > http://lists.xen.org/xen-devel
Hello, > > > > + if (prod - out_cons >= XENFB_OUT_RING_LEN) { > > > > + return; > > > > + } This test seems overzealous to me: AIUI, the producer can produce XENFB_OUT_RING_LEN events, and thus prod - out_cons is exactly XENFB_OUT_RING_LEN, i.e. there is no room left at all. The frontend part is: while (page->out_prod - page->out_cons == XENFB_OUT_RING_LEN) schedule(); I.e. it waits while the buffer is exactly full. So it seems to me the bug is at the backend side. Samuel
diff --git a/hw/display/xenfb.c b/hw/display/xenfb.c index 4e2a27a..8eb3046 100644 --- a/hw/display/xenfb.c +++ b/hw/display/xenfb.c @@ -789,8 +789,9 @@ static void xenfb_handle_events(struct XenFB *xenfb) prod = page->out_prod; out_cons = page->out_cons; - if (prod == out_cons) - return; + if (prod - out_cons >= XENFB_OUT_RING_LEN) { + return; + } xen_rmb(); /* ensure we see ring contents up to prod */ for (cons = out_cons; cons != prod; cons++) { union xenfb_out_event *event = &XENFB_OUT_RING_REF(page, cons);