Re: [Xen-devel] HVM restore broken?

[Tim Deegan <Tim.Deegan@xxxxxxxxxxxxx>]

Hi,

Cset 13766:7b72e2536065 of -unstable fixes these *-on-64 crashes for me: 
the compat-mode XLAT_foo() macros were escaping from their proper scope.

Cheers,

Tim.

At 12:57 +0100 on 30 Jan (1170161839), Petersson, Mats wrote:
> My windows problem is definitely 64b HV + 32-bit guest, and none of
> Stefan's reported error messages in my log either. 
> 
> --
> Mats
> 
> > -----Original Message-----
> > From: Zhai, Edwin [mailto:edwin.zhai@xxxxxxxxx] 
> > Sent: 30 January 2007 11:33
> > To: Stefan Berger
> > Cc: Zhai, Edwin; Petersson, Mats; Xen Development Mailing 
> > List; xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
> > Subject: Re: [Xen-devel] HVM restore broken?
> > 
> > berger,
> > 
> > you run 32 linux on 64b HV, right?
> > 
> > i can reproduce this bug in such combination on latest change 
> > set, but have no 
> > console MFN & set maxmem err msg.
> > 
> > i doubt some memory restore issue here.
> > 
> > thanks,
> > 
> > On Sat, Jan 27, 2007 at 11:57:09AM -0500, Stefan Berger wrote:
> > >    I tried suspend/resume with a Linux guest with 1 processor.
> > > 
> > >    I get the following errors in xend's log file upon resume.
> > > 
> > >    [...]
> > >    [2007-01-27 11:44:06 3629] DEBUG (XendDomainInfo:775) 
> > Storing domain
> > >    details: {'console/port': '3', 'name': 'TCG-TEST', 
> > 'console/limit':
> > >    '1048576', 'vm': 
> > '/vm/b0aaf4ff-ede8-994d-551c-23da391edbf2', 'domid': '4',
> > >    'cpu/0/availability': 'online', 'memory/target': 
> > '196608', 'store/port':
> > >    '2'}
> > >    [2007-01-27 11:44:06 3629] INFO (XendCheckpoint:207) 
> > restore hvm domain 4,
> > >    mem=192, apic=1, pae=1
> > >    [2007-01-27 11:44:06 3629] DEBUG (XendCheckpoint:226) 
> > restore:shadow=0x3,
> > >    _static_max=0xd0, _static_min=0xc0, nr_pfns=0xc000.
> > >    [2007-01-27 11:44:06 3629] DEBUG (balloon:127) Balloon: 
> > 258044 KiB free;
> > >    need 199680; done.
> > >    [2007-01-27 11:44:06 3629] DEBUG (XendCheckpoint:236) 
> > [xc_restore]:
> > >    /usr/lib/xen/bin/xc_restore 22 4 49152 2 3 192 1 1
> > >    [2007-01-27 11:44:06 3629] INFO (XendCheckpoint:340) 
> > xc_hvm_restore:dom=4,
> > >    nr_pfns=0xc000, store_evtchn=2, *store_mfn=192, console_evtchn=3,
> > >    *console_mfn=-1208080352, pae=1, apic=1.
> > >    [2007-01-27 11:44:06 3629] INFO (XendCheckpoint:340) 
> > xc_hvm_restore start:
> > >    max_pfn = c000, max_mfn = 3f700, hvirt_start=f5800000, 
> > pt_levels=3
> > >    [2007-01-27 11:44:07 3629] INFO (XendCheckpoint:340) hvm 
> > restore:calculate
> > >    new store_mfn=0xbffe,v_end=0xc000000..
> > >    [2007-01-27 11:44:07 3629] INFO (XendCheckpoint:340) hvm 
> > restore:get
> > >    nr_vcpus=1.
> > >    [2007-01-27 11:44:07 3629] INFO (XendCheckpoint:340) 
> > Restore exit with
> > >    rc=0
> > >    [2007-01-27 11:44:07 3629] DEBUG (XendCheckpoint:311) 
> > store-mfn 49150
> > >    [2007-01-27 11:44:07 3629] DEBUG (XendCheckpoint:311) console-mfn
> > >    -1208080352
> > >    [2007-01-27 11:44:07 3629] DEBUG (XendDomainInfo:1558)
> > >    XendDomainInfo.destroy: domid=4
> > >    [2007-01-27 11:44:07 3629] DEBUG (XendDomainInfo:1566)
> > >    XendDomainInfo.destroyDomain(4)
> > >    [2007-01-27 11:44:07 3629] ERROR (XendDomain:1030) Restore failed
> > >    Traceback (most recent call last):
> > >     File "//usr/lib/python/xen/xend/XendDomain.py", line 1025, in
> > >    domain_restore_fd
> > >       return XendCheckpoint.restore(self, fd, paused=paused)
> > >     File "//usr/lib/python/xen/xend/XendCheckpoint.py", 
> > line 243, in restore
> > >       raise XendError('Could not read store/console MFN')
> > >    XendError: Could not read store/console MFN
> > > 
> > >    In the qemu dm's log file I see this here  --- notice 
> > the 'error 22':
> > > 
> > >    domid: 3
> > >    qemu: the number of cpus is 1
> > >    qemu_map_cache_init nr_buckets = c00
> > >    shared page at pfn:bfff
> > >    buffered io page at pfn:bffd
> > >    xs_read(): vncpasswd get error.
> > >    /vm/b0aaf4ff-ede8-994d-551c-23da391edbf2/vncpasswd.
> > >    I/O request not ready: 0, ptr: 0, port: 0, data: 0, 
> > count: 0, size: 0
> > >    suspend sig handler called with requested=0!
> > >    device model received suspend signal!
> > >    set maxmem returned error 22
> > >    cirrus_stop_acc:unset_vram_mapping.
> > > 
> > >    'xm dmesg' shows this here (trying multiple times to resume):
> > > 
> > >    (XEN) HVM S/R Loading "xen_hvm_i8259" instance 0x20
> > >    (XEN) HVM S/R Loading "xen_hvm_i8259" instance 0xa0
> > >    (XEN) HVM S/R Loading "xen_hvm_ioapic" instance 0x0
> > >    (XEN) HVM S/R Loading "xen_hvm_cpu" instance 0x0
> > >    (XEN) HVM S/R Loading "xen_hvm_lapic" instance 0x0
> > >    (XEN) HVM S/R Loading "xen_hvm_i8254" instance 0x40
> > >    (XEN) HVM S/R Loading "xen_hvm_shpage" instance 0x10
> > >    (XEN) HVM S/R Loading "xen_hvm_i8259" instance 0x20
> > >    (XEN) HVM S/R Loading "xen_hvm_i8259" instance 0xa0
> > >    (XEN) HVM S/R Loading "xen_hvm_ioapic" instance 0x0
> > >    (XEN) HVM S/R Loading "xen_hvm_cpu" instance 0x0
> > >    (XEN) HVM S/R Loading "xen_hvm_lapic" instance 0x0
> > >    (XEN) HVM S/R Loading "xen_hvm_i8254" instance 0x40
> > >    (XEN) HVM S/R Loading "xen_hvm_shpage" instance 0x10
> > > 
> > >       Stefan
> > > 
> > >    xen-devel-bounces@xxxxxxxxxxxxxxxxxxx wrote on 
> > 01/27/2007 08:53:07 AM:
> > > 
> > >    > Mats,
> > >    >
> > >    > at least 32b up windows on 32b HV save/restore works here.
> > >    >
> > >    > what's your configuration? 32 or 64 windows/HV? up or smp?
> > >    >
> > >    > BTW, can you have a try with linux guest?
> > >    >
> > >    >
> > >    > On Fri, Jan 26, 2007 at 06:41:13PM +0100, Petersson, 
> > Mats wrote:
> > >    > > I got latest (13601) yesterday evening. This doesn't 
> > seem to work to
> > >    do
> > >    > > Restore (at least of the Windows test-image that 
> > I've been using for
> > >    > > testing previously).
> > >    > >
> > >    > > The VM restores reasonably OK, but it jumps to an 
> > invalid address
> > >    > > shortly after restoring, giving a D1 blue-screen error
> > >    > > (DRIVER_IRQL_LESS_OR_EQUAL), which turns out to be 
> > "page-fault in
> > >    > > driver" after I looked at the memory dump in windbg. 
> > (The address it
> > >    > > jumps to is consistenly a0000ca5, if that's of any 
> > meaning to anyone).
> > >    > >
> > >    > > I've compared my svm.c that I had previously and the 
> > current one that
> > >    I
> > >    > > got from mercurial, and they are identical.
> > >    > >
> > >    > > I went back to my 13568 build of the hypervisor, and 
> > it works there...
> > >    > > There's no obvious changes in between...
> > >    > >
> > >    > > Has anyone else tried this, does anyone have an idea 
> > of what's going
> > >    > > wrong?
> > >    > >
> > >    > > --
> > >    > > Mats
> > >    > >
> > >    > >
> > >    > >
> > >    > > _______________________________________________
> > >    > > Xen-devel mailing list
> > >    > > Xen-devel@xxxxxxxxxxxxxxxxxxx
> > >    > > http://lists.xensource.com/xen-devel
> > >    > >
> > >    >
> > >    > --
> > >    > best rgds,
> > >    > edwin
> > >    >
> > >    > _______________________________________________
> > >    > Xen-devel mailing list
> > >    > Xen-devel@xxxxxxxxxxxxxxxxxxx
> > >    > http://lists.xensource.com/xen-devel
> > 
> > -- 
> > best rgds,
> > edwin
> > 
> > 
> > 
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel