I have several x86-32 machines (blades) which hang immediately after grub displays the modules it has loaded. No further output is displayed. This is due to ''something'' in changeset 15035 from xen-unstable.hg. Anybody else experiencing the same problems? Updating to changes 15080 does not solve the problem. Stefan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
It might be related to gcc compiler. Which one you are using? GCC 4.1.x might work better. -Wei ________________________________ From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of Stefan Berger Sent: Friday, May 25, 2007 4:10 PM To: xen-devel@lists.xensource.com Subject: [Xen-devel] Changeset 15035 causes x86-32 to hang I have several x86-32 machines (blades) which hang immediately after grub displays the modules it has loaded. No further output is displayed. This is due to ''something'' in changeset 15035 from xen-unstable.hg. Anybody else experiencing the same problems? Updating to changes 15080 does not solve the problem. Stefan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
This changeset seems to have introduced some general instability. A complete hang at boot is the best most reproducible case reported so far however. It would be great if you can add some tracing and find out what is going on. It may be that the underlying bug is also responsible for some issues which are harder to reproduce and track down. -- Keir On 25/5/07 22:09, "Stefan Berger" <stefanb@us.ibm.com> wrote:> > I have several x86-32 machines (blades) which hang immediately after grub > displays the modules it has loaded. No further output is displayed. This is > due to ''something'' in changeset 15035 from xen-unstable.hg. Anybody else > experiencing the same problems? Updating to changes 15080 does not solve the > problem. > > Stefan > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
"Huang2, Wei" <Wei.Huang2@amd.com> wrote on 05/25/2007 05:11:57 PM:> It might be related to gcc compiler. Which one you are using? GCC 4. > 1.x might work better.Scary... I am using gcc 4.1.1 20070105 (RedHat 4.1.1-51) from FC 6. Stefan> > -Wei > > From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel- > bounces@lists.xensource.com] On Behalf Of Stefan Berger > Sent: Friday, May 25, 2007 4:10 PM > To: xen-devel@lists.xensource.com > Subject: [Xen-devel] Changeset 15035 causes x86-32 to hang> > I have several x86-32 machines (blades) which hang immediately after > grub displays the modules it has loaded. No further output is > displayed. This is due to ''something'' in changeset 15035 from xen- > unstable.hg. Anybody else experiencing the same problems? Updating > to changes 15080 does not solve the problem. > > Stefan_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
xen-devel-bounces@lists.xensource.com wrote on 05/25/2007 05:42:04 PM:> > This changeset seems to have introduced some general instability. A > complete hang at boot is the best most reproducible case reported so > far however. It would be great if you can add some tracing and find > out what is going on. It may be that the underlying bug is also > responsible for some issues which are harder to reproduce and trackdown. At what point does printk() work? All the way at the entry point of __start_xen()? I have added some printks there but haven''t seen any output to the screen so far. Looking at the assembly part now that''s calling __start_xen(). I''d help, but I''ll be away for a couple of days starting tomorrow... Stefan> > -- Keir > > On 25/5/07 22:09, "Stefan Berger" <stefanb@us.ibm.com> wrote:> > I have several x86-32 machines (blades) which hang immediately after > grub displays the modules it has loaded. No further output is > displayed. This is due to ''something'' in changeset 15035 from xen- > unstable.hg. Anybody else experiencing the same problems? Updating > to changes 15080 does not solve the problem. > > Stefan > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 26/5/07 00:46, "Stefan Berger" <stefanb@us.ibm.com> wrote:>> > This changeset seems to have introduced some general instability. A >> > complete hang at boot is the best most reproducible case reported so >> > far however. It would be great if you can add some tracing and find >> > out what is going on. It may be that the underlying bug is also >> > responsible for some issues which are harder to reproduce and track down. > > At what point does printk() work? All the way at the entry point of > __start_xen()? > I have added some printks there but haven''t seen any output to the screen so > far. Looking at the assembly part now that''s calling __start_xen(). > > I''d help, but I''ll be away for a couple of days starting tomorrow...If you have a serial line attached then you can output characters as early as you like with just a couple of machine instructions. Outputting to VGA text console is similarly quite easy (just write characters to the text console at 0xb8000). It¹s not full printk, but is sufficient to track down how far your boot is getting before crash or hang. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
xen-devel-bounces@lists.xensource.com wrote on 05/26/2007 01:31:51 AM:> On 26/5/07 00:46, "Stefan Berger" <stefanb@us.ibm.com> wrote:> > This changeset seems to have introduced some general instability. A > > complete hang at boot is the best most reproducible case reported so > > far however. It would be great if you can add some tracing and find > > out what is going on. It may be that the underlying bug is also > > responsible for some issues which are harder to reproduce and trackdown.> > At what point does printk() work? All the way at the entry point of > __start_xen()? > I have added some printks there but haven''t seen any output to the > screen so far. Looking at the assembly part now that''s calling__start_xen().> > I''d help, but I''ll be away for a couple of days starting tomorrow... > > If you have a serial line attached then you can output characters as > early as you like with just a couple of machine instructions. > Outputting to VGA text console is similarly quite easy (just write > characters to the text console at 0xb8000). It?s not full printk, > but is sufficient to track down how far your boot is getting before > crash or hang.>From xen/arch/x86/boot/trampoline.S:[...] #if CONFIG_PAGING_LEVELS != 2 /* Set up EFER (Extended Feature Enable Register). */ movl $MSR_EFER,%ecx rdmsr [...] The rdmsr is causing the crash on my machines. This patch fixes the problem: Signed-off-by: Stefan Berger <stefanb@us.ibm.com> diff -r bd3d6b4c52ec xen/arch/x86/boot/trampoline.S --- a/xen/arch/x86/boot/trampoline.S Fri Jun 01 14:50:52 2007 +0100 +++ b/xen/arch/x86/boot/trampoline.S Mon Jun 04 12:41:28 2007 -0400 @@ -59,6 +59,10 @@ trampoline_protmode_entry: mov %eax,%cr3 #if CONFIG_PAGING_LEVELS != 2 + mov SYM_TRAMP_PHYS(cpuid_ext_features),%edi + btl $20,%edi /* CPUID 0x80000001, EDX[20] */ + jnc 2f + /* Set up EFER (Extended Feature Enable Register). */ movl $MSR_EFER,%ecx rdmsr @@ -66,13 +70,10 @@ trampoline_protmode_entry: btsl $_EFER_LME,%eax /* Long Mode */ btsl $_EFER_SCE,%eax /* SYSCALL/SYSRET */ #endif - mov SYM_TRAMP_PHYS(cpuid_ext_features),%edi - btl $20,%edi /* CPUID 0x80000001, EDX[20] */ - jnc 1f btsl $_EFER_NX,%eax /* No-Execute */ -1: wrmsr + wrmsr #endif - +2: mov $0x80050033,%eax /* hi-to-lo: PG,AM,WP,NE,ET,MP,PE */ mov %eax,%cr0 jmp 1f Stefan> > -- Keir_______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 4/6/07 17:58, "Stefan Berger" <stefanb@us.ibm.com> wrote:> The rdmsr is causing the crash on my machines. > This patch fixes the problem:Thanks for tracking it down. I fixed the patch to handle the (possibly non-existent) case of a machine that supports Long Mode but not NX, and checked in as 15222:0feaf2fc75d3. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Petersson, Mats
2007-Jun-05 09:34 UTC
RE: [Xen-devel] Changeset 15035 causes x86-32 to hang
> -----Original Message----- > From: xen-devel-bounces@lists.xensource.com > [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of > Keir Fraser > Sent: 05 June 2007 10:31 > To: Stefan Berger > Cc: xen-devel@lists.xensource.com > Subject: Re: [Xen-devel] Changeset 15035 causes x86-32 to hang > > On 4/6/07 17:58, "Stefan Berger" <stefanb@us.ibm.com> wrote: > > > > The rdmsr is causing the crash on my machines. > This patch fixes the problem: > > > > Thanks for tracking it down. I fixed the patch to handle the > (possibly non-existent) case of a machine that supports Long > Mode but not NX, and checked in as 15222:0feaf2fc75d3.Some early 64-bit Intel processors have this combo of LM but no NX - apparently, we kept NX sufficiently hidden that Intel didn''t manage to get it fully working for the first public release of the processor. So whilst those processors are most likely quite rare, they do exist. All AMD processors that reached the public have NX capability if they have long-mode. -- Mats> > -- Keir > >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel