lost ticks in linux

lost ticks in linux

[Emil Tantilov]

Hi all,

Just recently started playing with kvm and used my
linux image which was previously running with
vmplayer. From what I can tell so far kvm runs
noticeably faster compared to vmplayer on the same
image. 

The only issue I've seen so far is with the timer - I
get lost ticks, which was not a problem with vmplayer:

...
TSC appears to be running slowly. Marking it unstable.
Time: acpi_pm clocksource has been installed.
...

I don't have this problem if I run it without loading
kvm-intel.

Also would appreciate suggestions about the kernel
config inside KVM - CPU etc. 

My CPU Core Duo E6600 btw.

Thanks,
Emil


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: lost ticks in linux

[Lonnie Mendez]

On Wed, 2006-12-20 at 18:47 -0800, Emil Tantilov wrote:
> The only issue I've seen so far is with the timer - I
> get lost ticks, which was not a problem with vmplayer:
> 
> ...
> TSC appears to be running slowly. Marking it unstable.
> Time: acpi_pm clocksource has been installed.
> ...

Does passing "clock=pit" in the kernel parameters line for the guest
help?  You can set this permanently in the guest's bootloader
configuration (grub,lilo,etc).


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: lost ticks in linux

[Emil Tantilov]

Also getting the following panic on init 6:

Restarting system.
divide error: 0000 [#1]
SMP
Modules linked in: ne2k_pci 8390
CPU:    0
EIP:    0060:[<c022d884>]    Not tainted VLI
EFLAGS: 00000297   (2.6.19 #11)
EIP is at delay_tsc+0x14/0x20
eax: c6b4fe16   ebx: 0003aa4d   ecx: c6b41dc1   edx:
0000002a
esi: 01234567   edi: b7f8eff4   ebp: df8a8000   esp:
df8a9e8c
ds: 007b   es: 007b   ss: 0068
Process reboot (pid: 4563, ti=df8a8000 task=c16c6070
task.ti=df8a8000)
Stack: 00000004 c010cfc4 0003aa4d b7f8eff4 28121969
c012aa73 00000000 ffffffff
       ffffffff 00000000 c170f70c 00000000 c170f70c
c01472a3 c170f710 000000b6
       c102f300 c014a086 c170f70c 000000b6 c018c693
dfe25b28 dfa64e3c dfa08888
Call Trace:
 [<c010cfc4>] machine_emergency_restart+0x84/0x100
 [<c012aa73>] sys_reboot+0x183/0x1a0
 [<c01472a3>] find_get_page+0x23/0x50
 [<c014a086>] filemap_nopage+0x336/0x3d0
 [<c018c693>] invalidate_inode_buffers+0x13/0xc0
 [<c01552ac>] __handle_mm_fault+0x54c/0x980
 [<c02fb88b>] sock_destroy_inode+0x1b/0x20
 [<c0179b48>] dentry_iput+0x68/0xc0
 [<c02288df>] _atomic_dec_and_lock+0x3f/0x70
 [<c01142c3>] do_page_fault+0x533/0x690
 [<c0103115>] sysenter_past_esp+0x56/0x79
 =======================
Code: 8d bc 27 00 00 00 00 48 79 fd c3 8d b6 00 00 00
00 8d bf 00 00 00 00 53 8b 5c 24 08 0f 31 89 c1 8d b4
26 00 00 00 00 f3 90 0f 31 <29> c8 39 d8 72 f6 5b c3
8d 74 26 00 b8 70 d8 22 c0 a3 28 0e 3e
EIP: [<c022d884>] delay_tsc+0x14/0x20 SS:ESP
0068:df8a9e8c

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: lost ticks in linux

[Avi Kivity]

Emil Tantilov wrote:
> Also getting the following panic on init 6:
>
>   

I presume (from the module list) that this is on the guest, right?

Guest shutdown/restart is not a well supported area at the moment.


> Restarting system.
> divide error: 0000 [#1]
> SMP
> Modules linked in: ne2k_pci 8390
>   

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: lost ticks in linux

[Emil Tantilov]

Yep - that is corect. init 0 works OK, but init 6
panics.

EIP points to delay_tsc, so I think it might be
related to the issue with the clock (TSC running
slow).

Emil

--- Avi Kivity <avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org> wrote:

> Emil Tantilov wrote:
> > Also getting the following panic on init 6:
> >
> >   
> 
> I presume (from the module list) that this is on the
> guest, right?
> 
> Guest shutdown/restart is not a well supported area
> at the moment.
> 
> 
> > Restarting system.
> > divide error: 0000 [#1]
> > SMP
> > Modules linked in: ne2k_pci 8390
> >   
> 
> -- 
> error compiling committee.c: too many arguments to
> function
> 
> 


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: lost ticks in linux

[Avi Kivity]

Emil Tantilov wrote:
> Yep - that is corect. init 0 works OK, but init 6
> panics.
>
> EIP points to delay_tsc, so I think it might be
> related to the issue with the clock (TSC running
> slow).
>
>   

Well, the EIP points at a sub instruction, which has no business 
generating divide exceptions.

  b:   8d b6 00 00 00 00       lea    0x0(%esi),%esi
  11:   8d bf 00 00 00 00       lea    0x0(%edi),%edi
  17:   53                      push   %ebx
  18:   8b 5c 24 08             mov    0x8(%esp),%ebx
  1c:   0f 31                   rdtsc
  1e:   89 c1                   mov    %eax,%ecx
  20:   8d b4 26 00 00 00 00    lea    0x0(%esi),%esi
  27:   f3 90                   pause
  29:   0f 31                   rdtsc
exception here -->
  2b:   29 c8                   sub    %ecx,%eax
  2d:   39 d8                   cmp    %ebx,%eax
  2f:   72 f6                   jb     0x27
  31:   5b                      pop    %ebx
  32:   c3                      ret

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV