AMD-V is disabled in the BIOS
log in

Advanced search

Message boards : Number crunching : AMD-V is disabled in the BIOS

1 · 2 · Next
Author Message
Trog Dog
Send message
Joined: 25 Dec 11
Posts: 30
Credit: 78,173
RAC: 55
Message 125 - Posted: 26 Dec 2011, 12:38:22 UTC

http://boinc.berkeley.edu/vbox/result.php?resultid=604
What's the cause of this error?

Is the AMD-V the real error? if so is it refering to the virtual machine or the real machine, or do I have to recompile the kernel?

Ageless
Volunteer moderator
Avatar
Send message
Joined: 23 Dec 11
Posts: 124
Credit: 11,435
RAC: 0
Message 130 - Posted: 26 Dec 2011, 14:53:00 UTC - in response to Message 125.
Last modified: 26 Dec 2011, 14:54:18 UTC

Rom is busy with compiling a new application that will address this issue as well. In the mean time, see Richard's thread. The VT-x problem he runs into, is essentially what you run into with AMD-v.

But you shouldn't go reboot and turn things on in your BIOS. Not yet.
Let's await the new application first that should detect whether the option is on, and if not, not use it.
____________
Jord.

I used to do T4T like you. Then I took an arrow in the knee...

Trog Dog
Send message
Joined: 25 Dec 11
Posts: 30
Credit: 78,173
RAC: 55
Message 131 - Posted: 26 Dec 2011, 15:47:51 UTC

too late - I've already rebooted and gone into bios and enabled something like secure virtualisation, but as you can see from this result it hasn't made any difference. I'll wait for the new app before I do any more fiddling and if need be I'll disable that setting again. I'll have a look through the kernel configs in the meantime.

Rom Walton
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 21 Dec 11
Posts: 367
Credit: 42,702
RAC: 0
Message 152 - Posted: 26 Dec 2011, 22:27:06 UTC

I think we now have something in place for this issue.

The latest linux wrapper and BOINC client should be making the right decisions now.

----- Rom

Trog Dog
Send message
Joined: 25 Dec 11
Posts: 30
Credit: 78,173
RAC: 55
Message 176 - Posted: 27 Dec 2011, 1:54:17 UTC

There's actually a v0.9 running on this machine as we speak and I haven't fiddled with it any further.

http://boinc.berkeley.edu/vbox/result.php?resultid=617

Rom Walton
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 21 Dec 11
Posts: 367
Credit: 42,702
RAC: 0
Message 177 - Posted: 27 Dec 2011, 2:12:21 UTC - in response to Message 176.

There's actually a v0.9 running on this machine as we speak and I haven't fiddled with it any further.

http://boinc.berkeley.edu/vbox/result.php?resultid=617


Cool.

----- Rom

Fibber McGee
Send message
Joined: 23 Dec 11
Posts: 117
Credit: 27,964
RAC: 0
Message 190 - Posted: 27 Dec 2011, 10:47:54 UTC - in response to Message 177.

I've turned off hardware virtualization on thisLinux host. It started and successfully ran a task on 1 CPU with BOINC 6.12.34. Later I updated the host to BOINC 7.0.2 and it received and is now running another 1 CPU task.

Trog Dog
Send message
Joined: 25 Dec 11
Posts: 30
Credit: 78,173
RAC: 55
Message 191 - Posted: 27 Dec 2011, 13:12:29 UTC

just googling it looks like AMD-V and virtualbox is hit and miss, there are reports that bios are broken, others that the kernel (kvm) is broken, and then others that you need to compile in kvm only to disable it because that then clears the relevant flag...

Fibber McGee
Send message
Joined: 23 Dec 11
Posts: 117
Credit: 27,964
RAC: 0
Message 203 - Posted: 27 Dec 2011, 22:23:12 UTC - in response to Message 191.

I haven't googled but maybe VT-x is hit and miss too. Or perhaps it has something to do with sandboxing/daemon. I reconfigured my host #10 to start BOINC as a daemon and am able to crunch tasks but they use only 1 CPU. I aborted a few to see if I could get one to start on 2 CPUs but it only uses 1. I double checked so I know I have 8 cores allotted to BOINC, VT-x is definitely turned on in BIOS and this project is the only project I've added to BOINC so far so tasks should be able to use 2 cores. I have HT turned on too.

When I was starting BOINC on my own account (not daemon) it started at least 1 task with 2 cores but who knows, maybe if I had run more tasks that way before switching to a daemon start it might have started a task with just 1 core. So is it due to daemon mode or is due to VT-x being hit amd miss, that's the question.

I could switch back to non-daemon but I accidentally deleted that installation so I would have to download the 1.61 GB file again which will take some time because my pathetic ISP isn't open today and won't be open until Friday so I still have the wrong modem and slow dl speed. ARRRGH! Maybe I can just copy the file to the non-daemon install and trick BOINC into thinking it dl'd it.

BOINC reports the following:

No config file found - using defaults
Starting BOINC client version 7.0.2 for x86_64-pc-linux-gnu
This a development version of BOINC and may not function properly
log flags: file_xfer, sched_ops, task
Libraries: libcurl/7.21.6 OpenSSL/1.0.0e zlib/1.2.3.4 libidn/1.22 librtmp/2.3
Data directory: /var/lib/boinc-client
Processor: 8 GenuineIntel Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz [Family 6 Model 42 Stepping 7]
Processor: 8.00 MB cache
Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf pni pclmulqd
OS: Linux: 3.0.0-14-generic
Memory: 7.80 GB physical, 7.98 GB virtual
Disk: 450.59 GB total, 264.37 GB free
VirtualBox version: 4.1.8r75467

It doesn't mention VT-x in the list of processor features but it didn't even when the 1 task started and ran on 2 cores. Maybe it has a different name? I've never turned VT-x off on that CPU so it isn't a case of BIOS failing to turn it back on.

Rom Walton
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 21 Dec 11
Posts: 367
Credit: 42,702
RAC: 0
Message 204 - Posted: 27 Dec 2011, 23:53:40 UTC - in response to Message 203.
Last modified: 28 Dec 2011, 0:00:54 UTC

Copying the file from another location should work. BOINC only cares about the MD5 checksum. How the file ended up in the project directory it isn't to worried about.

As far as processor features are concerned. Linux names the VT-x feature set 'vmx' and AMD-V is called 'svm'.

What I do not know is if those features are bubbled up through to guest operating systems. IIRC, you are running a few instances of BOINC in child VMs.

----- Rom

Richard Long
Send message
Joined: 25 Dec 11
Posts: 17
Credit: 41,420
RAC: 0
Message 206 - Posted: 28 Dec 2011, 0:28:08 UTC - in response to Message 203.

Is it turned on in the bios, i thought this project was turning it off earlier

Rom Walton
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 21 Dec 11
Posts: 367
Credit: 42,702
RAC: 0
Message 209 - Posted: 28 Dec 2011, 0:51:47 UTC - in response to Message 206.

Is it turned on in the bios, i thought this project was turning it off earlier


Just for clarification:

I was, for one version of the wrapper, disabling their use by the VM. I didn't have the wrapper tweak the BIOS setting.

Based on Fibber's feedback about the feature being required by VirtualBox SMP support, vboxwrapper no longer disables VirtualBox's use of those extensions.

With the exception that the BOINC Server detects that the BOINC client did not report having those features and so only assigned a single-core VM for the client to run. Then it is treated almost like an insurance policy in that I do it just in case something is tweaked in VirtualBox and it believes it needs those extensions.

----- Rom

Fibber McGee
Send message
Joined: 23 Dec 11
Posts: 117
Credit: 27,964
RAC: 0
Message 215 - Posted: 28 Dec 2011, 2:02:25 UTC - in response to Message 204.

As far as processor features are concerned. Linux names the VT-x feature set 'vmx' and AMD-V is called 'svm'.

What I do not know is if those features are bubbled up through to guest operating systems. IIRC, you are running a few instances of BOINC in child VMs.

----- Rom


OK, I don't see vmx in the list of processor features in my previous post so I guess that's why BOINC is giving the task only 1 core. That task is running on the host OS not a guest OS. BOINC running on a guest OS doesn't show vmx either in fact the list of features is missing several features that are in the list from BOINC running on the host OS. So some features bubble up but not all.

I'll investigate more thoroughly the possibility that running BOINC as a daemon is causing VT-x to disappear and get back to you. I'll investigate the same issue on my other host as well, it has an AMD CPU.

Fibber McGee
Send message
Joined: 23 Dec 11
Posts: 117
Credit: 27,964
RAC: 0
Message 217 - Posted: 28 Dec 2011, 10:11:28 UTC - in response to Message 215.

Did some of the investigation I promised, all tests were done on Intel i7, Kubuntu 11.10 and CERN Virtual Machine 1.00 (vbox64). Now I'm not so sure I ever had a task running on 2 cores on my Intel host. I switched from a daemon BOINC 7.02 to a non-daemon BOINC 7.02 to a non-daemon BOINC 6.12.43. I also tried turning VT-x on and off in BIOS a few times but all I got were single CPU tasks.

Below is the list of processor features reported by the two BOINCs I tested with. Neither of them reported vmx when it was turned on (or off) in BIOS. Interesting though is the fact that 6.12.43 reports 2 features, highlighted in red, that 7.0.2 does not report. That makes the whole feature reporting/recording process seem a little quirky.

BOINC 6.12.43 Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 mo

BOINC 7.0.2 Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf pni pclmulqd

Tomorrow I will investigate more on my AMD host. That host runs Kubuntu 10.10, not 11.10 and it runs 2 CPU tasks. Could be the difference in BIOS, the CPU brand or maybe even the Linux or kernel version. I might setup an 11.10/10.10 dual boot on the Intel host to see if that makes a difference and check for a BIOS update for the motherboard.

Ageless
Volunteer moderator
Avatar
Send message
Joined: 23 Dec 11
Posts: 124
Credit: 11,435
RAC: 0
Message 218 - Posted: 28 Dec 2011, 12:44:06 UTC - in response to Message 217.
Last modified: 28 Dec 2011, 12:44:28 UTC

Looks more like they're cut short, as the 'mo' should be monitor.

These are the full features for your i7:
fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt aes xsave avx lahf_lm ida arat epb xsaveopt pln pts dts tpr_shadow vnmi flexpriority ept vpid
____________
Jord.

I used to do T4T like you. Then I took an arrow in the knee...

Rom Walton
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 21 Dec 11
Posts: 367
Credit: 42,702
RAC: 0
Message 219 - Posted: 28 Dec 2011, 12:45:55 UTC

Well, on Linux we are just copying what is reported as the CPU features from:
/proc/cpuinfo

Windows is the complicated platform as far as CPU Info is concerned.

----- Rom

Rom Walton
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 21 Dec 11
Posts: 367
Credit: 42,702
RAC: 0
Message 220 - Posted: 28 Dec 2011, 12:51:38 UTC - in response to Message 218.

Good catch Jord, I found and fixed that bug.

----- Rom

Rom Walton
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 21 Dec 11
Posts: 367
Credit: 42,702
RAC: 0
Message 221 - Posted: 28 Dec 2011, 12:52:12 UTC - in response to Message 220.

It'll appear in the next client release.

----- Rom

Trog Dog
Send message
Joined: 25 Dec 11
Posts: 30
Credit: 78,173
RAC: 55
Message 222 - Posted: 28 Dec 2011, 13:06:02 UTC

first task completed http://boinc.berkeley.edu/vbox/result.php?resultid=617 it only ran on one core even though I turned on secure virtualisation (or something to that effect in the bios)

cat /proc/cpuinfo lists the following

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 pni monitor cx16 popcnt lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt npt lbrv svm_lock nrip_save


I have svm and svm_lock listed. I'll have a look and play with the kernel configs (but not tonight)

Rom Walton
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 21 Dec 11
Posts: 367
Credit: 42,702
RAC: 0
Message 223 - Posted: 28 Dec 2011, 13:14:12 UTC - in response to Message 222.

first task completed http://boinc.berkeley.edu/vbox/result.php?resultid=617 it only ran on one core even though I turned on secure virtualisation (or something to that effect in the bios)

cat /proc/cpuinfo lists the following

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 pni monitor cx16 popcnt lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt npt lbrv svm_lock nrip_save


I have svm and svm_lock listed. I'll have a look and play with the kernel configs (but not tonight)


I'll get a new client out the door today. You guy are tripping up on a client bug.

----- Rom

1 · 2 · Next
Post to thread

Message boards : Number crunching : AMD-V is disabled in the BIOS


Main page · Your account · Message boards


Copyright © 2014 U.C. Berkeley