Simon.Mills++at++esa.int
Date: 03/24/2004 06:20:51
----- Forwarded by Simon C Mills/estec/ESA on 24/03/2004 15:20 -----
Simon C Mills
To: Clive Harding <clive++at++sgi.com>
24/03/2004 15:19 cc: allan++at++sgi.com
Subject: Re: [info-performer] pfuFreeAllCPUs hangs Onyx300(Document link: Simon C Mills)
Hi Clive and Allan,
The system is indeed running 6.5.21m. I'll check out those patches, thanks.
Actually it's been a problem for a while, I just never looked into it deeply
enough. Since we do not install every update it's hard to tell which update
caused it but IIRC it was some time after 6.5.17m. It was definately working
when the code was originally developed some years ago now.
Regards, Simon
_______________________________________________________________________
Simon Mills
Modelling & Simulation Section (TOS-EMM) Tel: +31 (0)71 565 3725
European Space Agency (ESA/ESTEC) Fax: +31 (0)71 565 5420
Postbus 299, 2200AG Noordwijk e-mail: Simon.Mills++at++esa.int
The Netherlands http://www.estec.esa.nl/wmwww/EMM
_______________________________________________________________________
Clive Harding
<clive++at++sgi.com> To: Simon.Mills++at++esa.int
cc: info-performer++at++sgi.com
24/03/2004 11:13 Subject: Re: [info-performer] pfuFreeAllCPUs hangs Onyx300
Hi Simon,
If you are running 6.5.21, then this sounds like bug 896223
which is
fixed in various patches. 5355 for 6.5.21m and 5383 for 6.5.21f are the
latest.
Clive.
Allan Schaffer wrote:
> Hi Simon,
>
> Thanks for the report (& workaround!). We've had a second similar case of
> this reported as well. My first suspicion is that an IRIX update might be
> causing this problem since this bit of Performer code hasn't changed in
quite
> some time. We'll look into it (SCR 911439); which version of IRIX are you
> running, and did you notice if this started after a recent upgrade?
>
> Allan
>
> Simon.Mills++at++esa.int wrote:
>
>>The call pfuFreeAllCPUs is hanging our Onyx300 with 6 CPUs and 2 IR4 pipes
>>when running a performer 3.0 application in real-time mode i.e. running as
>>root and isolating CPU's and locking processes to run on only those CPUs.
>>This call is done when exiting the application just like in perfly. The
>>symptoms are the call never returns and after some seconds the system hangs
-
>>no mouse movement, no response from console etc., all I can do is reboot.
>>
>>The same behaviour is shown by perfly if run as root with the command
'perfly
>>-L -N'.
>>
>>I have a workaround now by using a modified pfuFreeAllCPUs with the line
>>
>> sysmp(MP_EMPOWER, i); /* allow CPU to run any process */
>>
>>commented out. In fact this line would seem redundant in pfuFreeAllCPUs if
I
>>look at the sysmp() call man page (see also MP_UNISOLATE).
>>
>>Can anyone confirm the same behavior? Is it a bug in pfuFreeAllCPUs or the
>>symp call? Is it or will it be fixed in newer Performer releases?
This archive was generated by hypermail 2b29 : Wed Mar 24 2004 - 06:23:52 PST