FPU problem

classic Classic list List threaded Threaded
10 messages Options
Reply | Threaded
Open this post in threaded view
|

FPU problem

Takeshi Sone

artsd fails to start on dom0 with this message:

    ** ERROR **: file ../../../flow/gsl/gslopmaster.c: line 521
    (master_process_flow): assertion failed: (gsl_fpu_okround () == TRUE)
    aborting...

gsl_fpu_okround() is defined as follows:

    static inline int
    gsl_fpu_okround (void)
    {
      GslFpuState cv;

      __asm__ ("fnstcw %0"
               : "=m" (*&cv));
      return !(cv & 0x0c00);
    }

Doesn't this mean Xen fails to save/restore FPU status bits?

Sound driver itself is working well under Xen.

Below is output of cat /proc/cpuinfo, if it helps.

    processor       : 0
    vendor_id       : GenuineIntel
    cpu family      : 6
    model           : 7
    model name      : Pentium III (Katmai)
    stepping        : 2
    cpu MHz         : 501.144
    cache size      : 512 KB
    fdiv_bug        : no
    hlt_bug         : yes
    f00f_bug        : no
    coma_bug        : no
    fpu             : yes
    fpu_exception   : yes
    cpuid level     : 2
    wp              : yes
    flags           : fpu tsc msr pae mce cx8 apic mtrr mca cmov pat pse36 mmx fxsr sse
    bogomips        : 999.42

--
Takeshi

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

RE: FPU problem

Ian Pratt

> artsd fails to start on dom0 with this message:
>
>     ** ERROR **: file ../../../flow/gsl/gslopmaster.c: line 521
>     (master_process_flow): assertion failed: (gsl_fpu_okround
> () == TRUE)
>     aborting...

Which version of Xen is this? I'd be very interested to know whether you
can reproduce under 2.0.6 or unstable.

Ian
 

> gsl_fpu_okround() is defined as follows:
>
>     static inline int
>     gsl_fpu_okround (void)
>     {
>       GslFpuState cv;
>
>       __asm__ ("fnstcw %0"
>                : "=m" (*&cv));
>       return !(cv & 0x0c00);
>     }
>
> Doesn't this mean Xen fails to save/restore FPU status bits?
>
> Sound driver itself is working well under Xen.
>
> Below is output of cat /proc/cpuinfo, if it helps.
>
>     processor       : 0
>     vendor_id       : GenuineIntel
>     cpu family      : 6
>     model           : 7
>     model name      : Pentium III (Katmai)
>     stepping        : 2
>     cpu MHz         : 501.144
>     cache size      : 512 KB
>     fdiv_bug        : no
>     hlt_bug         : yes
>     f00f_bug        : no
>     coma_bug        : no
>     fpu             : yes
>     fpu_exception   : yes
>     cpuid level     : 2
>     wp              : yes
>     flags           : fpu tsc msr pae mce cx8 apic mtrr mca
> cmov pat pse36 mmx fxsr sse
>     bogomips        : 999.42
>
> --
> Takeshi
>
> _______________________________________________
> Xen-users mailing list
> [hidden email]
> http://lists.xensource.com/xen-users
>

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: FPU problem

Takeshi Sone
On Thu, May 26, 2005 at 11:58:03AM +0100, Ian Pratt wrote:

>
> > artsd fails to start on dom0 with this message:
> >
> >     ** ERROR **: file ../../../flow/gsl/gslopmaster.c: line 521
> >     (master_process_flow): assertion failed: (gsl_fpu_okround
> > () == TRUE)
> >     aborting...
>
> Which version of Xen is this? I'd be very interested to know whether you
> can reproduce under 2.0.6 or unstable.

2.0-testing as of 5-15, as reported on
http://lists.xensource.com/archives/html/xen-users/2005-05/msg00334.html

Also, I experienced the phenominon of "top" displaying "NaN" as CPU
usages of all the processes.
I'm starting to think all of the problems are related to FPU,
including X application garbages and crashes.
--
Takeshi

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

RE: FPU problem

Ian Pratt
In reply to this post by Takeshi Sone
 > > > artsd fails to start on dom0 with this message:

> > >
> > >     ** ERROR **: file ../../../flow/gsl/gslopmaster.c: line 521
> > >     (master_process_flow): assertion failed: (gsl_fpu_okround
> > > () == TRUE)
> > >     aborting...
> >
> > Which version of Xen is this? I'd be very interested to
> know whether
> > you can reproduce under 2.0.6 or unstable.
>
> 2.0-testing as of 5-15, as reported on
> http://lists.xensource.com/archives/html/xen-users/2005-05/msg
> 00334.html

Hmm, I doubt its fixed in 2.0.6 then, but it would be good if you could
verify this.

I'd be *very* interested to know if the unstable tree still has this
bug. There have been a bunch of changes to the way the FPU is handled,
so I'd be surprised if it was still present.

Also, what CPU are you using? Are you using anything that might be
making used of 3dnow/mmx/sse etc?

Ian

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: FPU problem

Takeshi Sone
On Fri, May 27, 2005 at 11:49:03AM +0100, Ian Pratt wrote:

>  > > > artsd fails to start on dom0 with this message:
> > > >
> > > >     ** ERROR **: file ../../../flow/gsl/gslopmaster.c: line 521
> > > >     (master_process_flow): assertion failed: (gsl_fpu_okround
> > > > () == TRUE)
> > > >     aborting...
> > >
> > > Which version of Xen is this? I'd be very interested to
> > know whether
> > > you can reproduce under 2.0.6 or unstable.
> >
> > 2.0-testing as of 5-15, as reported on
> > http://lists.xensource.com/archives/html/xen-users/2005-05/msg
> > 00334.html
>
> Hmm, I doubt its fixed in 2.0.6 then, but it would be good if you could
> verify this.

I just upgraded to 2.0.6, and playing around for an hour,
but can't reproduce this problem. I guess it will happen again somehow.
Other problems (X application crash) are still remain.

> I'd be *very* interested to know if the unstable tree still has this
> bug. There have been a bunch of changes to the way the FPU is handled,
> so I'd be surprised if it was still present.

I think I can try unstable in a few days.

> Also, what CPU are you using? Are you using anything that might be
> making used of 3dnow/mmx/sse etc?

Pentium III.

processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 7
model name : Pentium III (Katmai)
stepping : 2
cpu MHz : 501.147
cache size : 512 KB
fdiv_bug : no
hlt_bug : yes
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 2
wp : yes
flags : fpu tsc msr pae mce cx8 apic mtrr mca cmov pat pse36 mmx fxsr sse
bogomips : 999.42


Dom0 is running usual server software (exim, apache, etc) and
Gnome desktop, krxvt, gkrellm, and firefox.
I changed the window manager from sawfish to metacity, and it crashes
more often than sawfish.
I have only one domU (mostly idle) running bind9.
I don't know any of them use MMX/SSE, however, playing DivX movie
with mplayer, which I think uses MMX/SSE, seems to screw the sysetm badly;
it brings death to metacity, gnome-panel and applets, and firefox,
within seconds.

--
Takeshi

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

RE: FPU problem

Ian Pratt
In reply to this post by Takeshi Sone
> I just upgraded to 2.0.6, and playing around for an hour, but
> can't reproduce this problem. I guess it will happen again somehow.
> Other problems (X application crash) are still remain.

What graphics card to you have? What modules is the Xserver using?
 
> > I'd be *very* interested to know if the unstable tree still
> has this
> > bug. There have been a bunch of changes to the way the FPU
> is handled,
> > so I'd be surprised if it was still present.
>
> I think I can try unstable in a few days.

Thanks.

Ian

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: FPU problem

Takeshi Sone
On Sat, May 28, 2005 at 11:25:04PM +0100, Ian Pratt wrote:
> > I just upgraded to 2.0.6, and playing around for an hour, but
> > can't reproduce this problem. I guess it will happen again somehow.
> > Other problems (X application crash) are still remain.
>
> What graphics card to you have? What modules is the Xserver using?

I was using ATI ("radeon" X driver) and replaced with Matrox G400 ("mga"
driver) thereafter. Both cards have the same problem.

> > > I'd be *very* interested to know if the unstable tree still
> > has this
> > > bug. There have been a bunch of changes to the way the FPU
> > is handled,
> > > so I'd be surprised if it was still present.
> >
> > I think I can try unstable in a few days.
>
> Thanks.

The artsd problem didn't reproduce in this 48 hours since I
upgraded to 2.0.6.

--
Takeshi

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: FPU problem

Takeshi Sone
In reply to this post by Ian Pratt
On Fri, May 27, 2005 at 11:49:03AM +0100, Ian Pratt wrote:
> I'd be *very* interested to know if the unstable tree still has this
> bug. There have been a bunch of changes to the way the FPU is handled,
> so I'd be surprised if it was still present.

I tried unstable. It booted dom0 fine, but couldn't start domU
with error Permission denied.
xm list displayed nothing (no dom0 either).

I couldn't play around with this set up for long time because we were
loosing our name server in domU (some people depends on it),
so I couldn't say if FPU or X problem was still there.

I thought I built and installed userspace tools from unstable properly.
What was wrong?

--
Takeshi

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: FPU problem

Steven Hand

>On Fri, May 27, 2005 at 11:49:03AM +0100, Ian Pratt wrote:
>> I'd be *very* interested to know if the unstable tree still has this
>> bug. There have been a bunch of changes to the way the FPU is handled,
>> so I'd be surprised if it was still present.
>
>I tried unstable. It booted dom0 fine, but couldn't start domU
>with error Permission denied.
>xm list displayed nothing (no dom0 either).

Are you running as root?

One change in the past while is the use of unix domain sockets rather
than [loopback] network sockets as the basic comms mechanism used by
the tools, which has a (desired!) side-effect of adding basic access
control to the xm <-> xend interface.

If this is not it, can you please try either/or:
 
  - looking at the log files /var/log/xend.log /var/log/xend-debug.log
    to see if anything jumps out at you
  - running "strace xm <foo>" to see if anything's up at that side.

cheers,

S.

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: FPU problem

Takeshi Sone
On Mon, May 30, 2005 at 01:34:41AM +0100, Steven Hand wrote:

>
> >On Fri, May 27, 2005 at 11:49:03AM +0100, Ian Pratt wrote:
> >> I'd be *very* interested to know if the unstable tree still has this
> >> bug. There have been a bunch of changes to the way the FPU is handled,
> >> so I'd be surprised if it was still present.
> >
> >I tried unstable. It booted dom0 fine, but couldn't start domU
> >with error Permission denied.
> >xm list displayed nothing (no dom0 either).
>
> Are you running as root?

I'm sure "xm create" was run by root (it was from init.d script).
I don't remember if "xm list" was by root, however.

> One change in the past while is the use of unix domain sockets rather
> than [loopback] network sockets as the basic comms mechanism used by
> the tools, which has a (desired!) side-effect of adding basic access
> control to the xm <-> xend interface.

That would be good, but I don't think this is xm-xend communication
problem because it's xend that reports error. Here is snippet from
/var/log/xend.log.

[2005-05-29 23:03:50 xend] INFO (SrvDaemon:610) Xend Daemon started
[2005-05-29 23:03:50 xend] INFO (XendRoot:91) EVENT> xend.start 0
[2005-05-29 23:03:54 xend] ERROR (SrvBase:162) op=create: Error creating domain:
 (13, 'Permission denied')
Traceback (most recent call last):
  File "/usr/lib/python2.3/site-packages/xen/xend/server/SrvBase.py", line 107,
in _perform
    val = op_method(op, req)
  File "/usr/lib/python2.3/site-packages/xen/xend/server/SrvDomainDir.py", line
71, in op_create
    raise XendError("Error creating domain: " + str(ex))
XendError: Error creating domain: (13, 'Permission denied')

>   - looking at the log files /var/log/xend.log /var/log/xend-debug.log
>     to see if anything jumps out at you
>   - running "strace xm <foo>" to see if anything's up at that side.

xend-debug.log has nothing special.
I will try to see more later.


--
Takeshi

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users