nested kvm tests questions

[Gianluca Cecchi]

Hello,
I have a Dell R815 with plenty of memory where I'm testing nested kvm,
as the processor is
AMD Opteron(tm) Processor 6174

and nested kvm is enabled by default.
The OS is Fedora 15 x86_64 and I have installed a fully updated Centos
5.6 guest.
To be able to use full virtualization inside guests of the CentOS
guests, I have configured its cpu as "copy to cpu"
so that now I have "Opteron_G3" in this centos cpu type.
I have configured it as 2 cpu.
The F15 host is quad socket / 12 cores each

Flags on f15 host I can see are:
 flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt
pdpe1gb rdtscp lm 3dnowext 3dnow constant_tsc rep_good nopl
nonstop_tsc extd_apicid amd_dcm pni monitor cx16 popcnt lahf_lm
cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch
osvw ibs skinit wdt nodeid_msr arat npt lbrv svm_lock nrip_save
pausefilter

flags on 5.6 guest from its cpuinfo are:
processor	: 1
vendor_id	: AuthenticAMD
cpu family	: 16
model		: 2
model name	: AMD Phenom(tm) 9550 Quad-Core Processor
stepping	: 3
cpu MHz		: 2200.024
cache size	: 512 KB
fpu		: yes
fpu_exception	: yes
cpuid level	: 5
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov
pat pse36 clflush mmx fxsr sse sse2 syscall nx mmxext fxsr_opt pdpe1gb
lm pni cx16 popcnt lahf_lm cmp_legacy svm cr8_legacy altmovcr8 abm
sse4a misalignsse
bogomips	: 4400.61
TLB size	: 1024 4K pages
clflush size	: 64
cache_alignment	: 64
address sizes	: 40 bits physical, 48 bits virtual
power management:

I can successfully install linux nested guests.
But I have problems installing windows xp nested guests.
I try to configure them as "standard", so with ide cd and default network card.
The installation proceeds ok up to the first reboot, but then at that
time I only get a white cursor blinking at top left...

Any reason a win guest could have these problems?
Anyone already installed win nested guests?
Any log I can provide to understand?

I suspect that only a subset f cpu flags are "nestable" and so
probably instead of "copy to cpu" it could be better to pass only a
subset of them as it is possible from virt-manager gui... but I don't
know what they could be...

Thanks in advance for suggestions
Next step could be to install a Fedora 15 guest and try nested
virtualization with it, to understand if the problem could be
virtualization feature limitation of codebase found in CentOS 5.6
version...

BTW: I saw great animation at beginning of May related to nested
virtualization arriving at code level for Intel cpu too... any
information about expected integration in fedora qemu/kvm versions...
(or rawhide eventually)?

Gianluca

[Gianluca Cecchi]

If I run an strace against the winxp guest that remains with the
blinking cursor I get this:

Process 19352 attached - interrupt to quit
select(24, [4 9 13 17 18 20 22 23], [], [], {1, 0}) = 1 (in [20], left
{0, 967000})
read(20,
"\16\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
128) = 128
rt_sigaction(SIGALRM, NULL, {0x407a90, ~[KILL STOP RTMIN RT_1],
SA_RESTORER, 0x3b7940eb10}, 8) = 0
write(5, "\0", 1)                       = 1
read(20, 0x7fff58cc0f90, 128)           = -1 EAGAIN (Resource
temporarily unavailable)
clock_gettime(CLOCK_MONOTONIC, {118919, 862857170}) = 0
clock_gettime(CLOCK_MONOTONIC, {118919, 862906170}) = 0
clock_gettime(CLOCK_MONOTONIC, {118919, 862950170}) = 0
select(24, [4 9 13 17 18 20 22 23], [], [], {1, 0}) = 1 (in [4], left {1, 0})
read(4, "\0", 512)                      = 1
read(4, 0x7fff58cc0e10, 512)            = -1 EAGAIN (Resource
temporarily unavailable)
clock_gettime(CLOCK_MONOTONIC, {118919, 863074170}) = 0
clock_gettime(CLOCK_MONOTONIC, {118919, 863098170}) = 0
timer_gettime(0, {it_interval={0, 0}, it_value={0, 0}}) = 0
timer_settime(0, 0, {it_interval={0, 0}, it_value={0, 32000000}}, NULL) = 0
clock_gettime(CLOCK_MONOTONIC, {118919, 863179170}) = 0
clock_gettime(CLOCK_MONOTONIC, {118919, 863204170}) = 0
select(24, [4 9 13 17 18 20 22 23], [], [], {1, 0}) = 1 (in [20], left
{0, 967000})
read(20,
"\16\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
128) = 128
rt_sigaction(SIGALRM, NULL, {0x407a90, ~[KILL STOP RTMIN RT_1],
SA_RESTORER, 0x3b7940eb10}, 8) = 0
write(5, "\0", 1)                       = 1
read(20, 0x7fff58cc0f90, 128)           = -1 EAGAIN (Resource
temporarily unavailable)
clock_gettime(CLOCK_MONOTONIC, {118919, 896796170}) = 0
clock_gettime(CLOCK_MONOTONIC, {118919, 896821170}) = 0
clock_gettime(CLOCK_MONOTONIC, {118919, 896880170}) = 0
select(24, [4 9 13 17 18 20 22 23], [], [], {1, 0}) = 1 (in [4], left {1, 0})
read(4, "\0", 512)                      = 1
read(4, 0x7fff58cc0e10, 512)            = -1 EAGAIN (Resource
temporarily unavailable)
clock_gettime(CLOCK_MONOTONIC, {118919, 897001170}) = 0
clock_gettime(CLOCK_MONOTONIC, {118919, 897025170}) = 0

HIH,
Gianluca

[Dor Laor]

On 06/13/2011 10:14 AM, Gianluca Cecchi wrote:
>  If I run an strace against the winxp guest that remains with the
>  blinking cursor I get this: 
Since nested kvm is relatively new and not really a full fledged feature 
yet - it should work but as you see not as stable as regular virt. I 
suggest you'll send the questions to kvm-devel, you'll get better 
feedback there, especially since the upstream kvm got new nested vmx 
patches (+CC the nested vmx developer)

>  Process 19352 attached - interrupt to quit
>  select(24, [4 9 13 17 18 20 22 23], [], [], {1, 0}) = 1 (in [20], left
>  {0, 967000})
>  read(20,
> "\16\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>  128) = 128
>  rt_sigaction(SIGALRM, NULL, {0x407a90, ~[KILL STOP RTMIN RT_1],
>  SA_RESTORER, 0x3b7940eb10}, 8) = 0
>  write(5, "\0", 1)                       = 1
>  read(20, 0x7fff58cc0f90, 128)           = -1 EAGAIN (Resource
>  temporarily unavailable)
>  clock_gettime(CLOCK_MONOTONIC, {118919, 862857170}) = 0
>  clock_gettime(CLOCK_MONOTONIC, {118919, 862906170}) = 0
>  clock_gettime(CLOCK_MONOTONIC, {118919, 862950170}) = 0
>  select(24, [4 9 13 17 18 20 22 23], [], [], {1, 0}) = 1 (in [4], left {1, 0})
>  read(4, "\0", 512)                      = 1
>  read(4, 0x7fff58cc0e10, 512)            = -1 EAGAIN (Resource
>  temporarily unavailable)
>  clock_gettime(CLOCK_MONOTONIC, {118919, 863074170}) = 0
>  clock_gettime(CLOCK_MONOTONIC, {118919, 863098170}) = 0
>  timer_gettime(0, {it_interval={0, 0}, it_value={0, 0}}) = 0
>  timer_settime(0, 0, {it_interval={0, 0}, it_value={0, 32000000}}, NULL) = 0
>  clock_gettime(CLOCK_MONOTONIC, {118919, 863179170}) = 0
>  clock_gettime(CLOCK_MONOTONIC, {118919, 863204170}) = 0
>  select(24, [4 9 13 17 18 20 22 23], [], [], {1, 0}) = 1 (in [20], left
>  {0, 967000})
>  read(20,
> "\16\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>  128) = 128
>  rt_sigaction(SIGALRM, NULL, {0x407a90, ~[KILL STOP RTMIN RT_1],
>  SA_RESTORER, 0x3b7940eb10}, 8) = 0
>  write(5, "\0", 1)                       = 1
>  read(20, 0x7fff58cc0f90, 128)           = -1 EAGAIN (Resource
>  temporarily unavailable)
>  clock_gettime(CLOCK_MONOTONIC, {118919, 896796170}) = 0
>  clock_gettime(CLOCK_MONOTONIC, {118919, 896821170}) = 0
>  clock_gettime(CLOCK_MONOTONIC, {118919, 896880170}) = 0
>  select(24, [4 9 13 17 18 20 22 23], [], [], {1, 0}) = 1 (in [4], left {1, 0})
>  read(4, "\0", 512)                      = 1
>  read(4, 0x7fff58cc0e10, 512)            = -1 EAGAIN (Resource
>  temporarily unavailable)
>  clock_gettime(CLOCK_MONOTONIC, {118919, 897001170}) = 0
>  clock_gettime(CLOCK_MONOTONIC, {118919, 897025170}) = 0
>
>  HIH,
>  Gianluca
>  _______________________________________________
>  virt mailing list
>  virt(a)lists.fedoraproject.org
>  https://admin.fedoraproject.org/mailman/listinfo/virt

[Gianluca Cecchi]

On Mon, Jun 13, 2011 at 9:52 AM, Dor Laor <dlaor(a)redhat.com&gt; wrote:
>  On 06/13/2011 10:14 AM, Gianluca Cecchi wrote:
> >
> > If I run an strace against the winxp guest that remains with the
> > blinking cursor I get this:
>
>  Since nested kvm is relatively new and not really a full fledged feature yet
>  - it should work but as you see not as stable as regular virt. I suggest
>  you'll send the questions to kvm-devel, you'll get better feedback there,
>  especially since the upstream kvm got new nested vmx patches (+CC the nested
>  vmx developer)
>  
Ok, I will ping that list too. Thansk for remarking
Just for info: a win7 guest works instead... so it seems to be
something win xp doens't like...