QEmu Script Error Checking

QEmu Script Error Checking

[Jack Mitchell]

I have run into the following issue where the tap device is in use due 
to the QEmu machine unexpectedly crashing. The error I recieve is here:

http://i.imgur.com/5t9U1.png (appologies for the screenshot but it 
wouldn't let me copy the text)

Is there any leaway for more robust error checking or a way to forcibly 
destroy the tap node to allow a new one to be created?

Cheers,
Jack.

Re: QEmu Script Error Checking

[Khem Raj]

On Tue, Oct 11, 2011 at 8:41 AM, Jack Mitchell <ml@communistcode.co.uk> wrote:
> I have run into the following issue where the tap device is in use due to
> the QEmu machine unexpectedly crashing. The error I recieve is here:
>
> http://i.imgur.com/5t9U1.png (appologies for the screenshot but it wouldn't
> let me copy the text)
>
> Is there any leaway for more robust error checking or a way to forcibly
> destroy the tap node to allow a new one to be created?
>

run-qemu-internal shows errors in that log. Did you try to figure out
why these errors are spewing.

> Cheers,
> Jack.
> _______________________________________________
> yocto mailing list
> yocto@yoctoproject.org
> https://lists.yoctoproject.org/listinfo/yocto
>

Re: QEmu Script Error Checking

[Scott Garman]

On 10/11/2011 08:41 AM, Jack Mitchell wrote:
> I have run into the following issue where the tap device is in use due
> to the QEmu machine unexpectedly crashing. The error I recieve is here:
>
> http://i.imgur.com/5t9U1.png (appologies for the screenshot but it
> wouldn't let me copy the text)
>
> Is there any leaway for more robust error checking or a way to forcibly
> destroy the tap node to allow a new one to be created?

Hi Jack,

Since qemu is being run from a parent shell script (runqemu), if qemu 
were to crash, I would think the parent shell script would continue on 
and destroy the tap device normally. Might you be killing the runqemu 
process instead of qemu itself?

I'm not sure if there is a way we could reliably force a cleanup of tap 
devices when runqemu starts, because we need to support the case where 
multiple instances of qemu sessions are running simultaneously (each 
with their own tap device). Off the top of my head I think this would 
make the state of tap devices non-deterministic.

Furthermore, we support a mode where an administrator can set up one or 
more tap devices, allowing the runqemu user to not need sudo privileges. 
So checking for the case where a tap device exists but no qemu process 
is running wouldn't work.

Scott

-- 
Scott Garman
Embedded Linux Engineer - Yocto Project
Intel Open Source Technology Center

Re: QEmu Script Error Checking

[Jack Mitchell]

On 11/10/2011 19:02, Scott Garman wrote:
> On 10/11/2011 08:41 AM, Jack Mitchell wrote:
>> I have run into the following issue where the tap device is in use due
>> to the QEmu machine unexpectedly crashing. The error I recieve is here:
>>
>> http://i.imgur.com/5t9U1.png (appologies for the screenshot but it
>> wouldn't let me copy the text)
>>
>> Is there any leaway for more robust error checking or a way to forcibly
>> destroy the tap node to allow a new one to be created?
>
> Hi Jack,
>
> Since qemu is being run from a parent shell script (runqemu), if qemu 
> were to crash, I would think the parent shell script would continue on 
> and destroy the tap device normally. Might you be killing the runqemu 
> process instead of qemu itself?
>
> I'm not sure if there is a way we could reliably force a cleanup of 
> tap devices when runqemu starts, because we need to support the case 
> where multiple instances of qemu sessions are running simultaneously 
> (each with their own tap device). Off the top of my head I think this 
> would make the state of tap devices non-deterministic.
>
> Furthermore, we support a mode where an administrator can set up one 
> or more tap devices, allowing the runqemu user to not need sudo 
> privileges. So checking for the case where a tap device exists but no 
> qemu process is running wouldn't work.
>
> Scott
>

Ok, thank you for the information Scott. I think I found the issue today 
which is causing the QEmu script to terminate pre-maturely.

I am using the eclipse-poky master branch as my eclipse plugin and it 
threw an error today when I tried to debug - which coincided with the 
QEmu and terminal window terminating. The errror I recieved was:

!SESSION 2011-10-12 09:22:49.135 
-----------------------------------------------
eclipse.buildId=I20110613-1736
java.version=1.6.0_22
java.vendor=Sun Microsystems Inc.
BootLoader constants: OS=linux, ARCH=x86_64, WS=gtk, NL=en_GB
Framework arguments:  -product org.eclipse.sdk.ide
Command-line arguments:  -product org.eclipse.sdk.ide -data 
/home/developer/jackdaw/../runtime-EclipseApplication -dev 
file:/home/developer/jackdaw/.metadata/.plugins/org.eclipse.pde.core/Eclipse 
Application/dev.properties -os linux -ws gtk -arch x86_64 -consoleLog

!ENTRY org.eclipse.cdt.debug.mi.core 4 42 2011-10-12 09:24:47.192
!MESSAGE Internal Error
!STACK 0
org.eclipse.cdt.debug.mi.core.cdi.MI2CDIException: Inferior terminated[]
     at 
org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.getMIShareds(SharedLibraryManager.java:123)
     at 
org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.updateState(SharedLibraryManager.java:188)
     at 
org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.update(SharedLibraryManager.java:136)
     at 
org.eclipse.cdt.debug.mi.core.cdi.EventManager.processSuspendedEvent(EventManager.java:326)
     at 
org.eclipse.cdt.debug.mi.core.cdi.EventManager.update(EventManager.java:100)
     at java.util.Observable.notifyObservers(Observable.java:159)
     at 
org.eclipse.cdt.debug.mi.core.MISession.notifyObservers(MISession.java:791)
     at org.eclipse.cdt.debug.mi.core.EventThread.run(EventThread.java:46)

Which looks like a CDT bug or the interfacing with CDT. I could file a 
bug report if you think it is related to the poky-eclipse plugin?

Cheers,
Jack.

Re: QEmu Script Error Checking

[Zhang, Jessica]

Hi Jack,

When and how did you trigger the error.  It looks more like a CDT issue than a Yocto issue, but can you tell us more what exactly you were doing with Yocto plug-in while this crash happened?  So you mean when you click on the debug, it caused qemu crashes? Or the qemu crashed 1st. then you got the eclipse error msg.

Thanks,
Jessica

-----Original Message-----
From: yocto-bounces@yoctoproject.org [mailto:yocto-bounces@yoctoproject.org] On Behalf Of Jack Mitchell
Sent: Wednesday, October 12, 2011 3:12 AM
To: yocto@yoctoproject.org
Subject: Re: [yocto] QEmu Script Error Checking

On 11/10/2011 19:02, Scott Garman wrote:
> On 10/11/2011 08:41 AM, Jack Mitchell wrote:
>> I have run into the following issue where the tap device is in use due
>> to the QEmu machine unexpectedly crashing. The error I recieve is here:
>>
>> http://i.imgur.com/5t9U1.png (appologies for the screenshot but it
>> wouldn't let me copy the text)
>>
>> Is there any leaway for more robust error checking or a way to forcibly
>> destroy the tap node to allow a new one to be created?
>
> Hi Jack,
>
> Since qemu is being run from a parent shell script (runqemu), if qemu
> were to crash, I would think the parent shell script would continue on
> and destroy the tap device normally. Might you be killing the runqemu
> process instead of qemu itself?
>
> I'm not sure if there is a way we could reliably force a cleanup of
> tap devices when runqemu starts, because we need to support the case
> where multiple instances of qemu sessions are running simultaneously
> (each with their own tap device). Off the top of my head I think this
> would make the state of tap devices non-deterministic.
>
> Furthermore, we support a mode where an administrator can set up one
> or more tap devices, allowing the runqemu user to not need sudo
> privileges. So checking for the case where a tap device exists but no
> qemu process is running wouldn't work.
>
> Scott
>

Ok, thank you for the information Scott. I think I found the issue today
which is causing the QEmu script to terminate pre-maturely.

I am using the eclipse-poky master branch as my eclipse plugin and it
threw an error today when I tried to debug - which coincided with the
QEmu and terminal window terminating. The errror I recieved was:

!SESSION 2011-10-12 09:22:49.135
-----------------------------------------------
eclipse.buildId=I20110613-1736
java.version=1.6.0_22
java.vendor=Sun Microsystems Inc.
BootLoader constants: OS=linux, ARCH=x86_64, WS=gtk, NL=en_GB
Framework arguments:  -product org.eclipse.sdk.ide
Command-line arguments:  -product org.eclipse.sdk.ide -data
/home/developer/jackdaw/../runtime-EclipseApplication -dev
file:/home/developer/jackdaw/.metadata/.plugins/org.eclipse.pde.core/Eclipse
Application/dev.properties -os linux -ws gtk -arch x86_64 -consoleLog

!ENTRY org.eclipse.cdt.debug.mi.core 4 42 2011-10-12 09:24:47.192
!MESSAGE Internal Error
!STACK 0
org.eclipse.cdt.debug.mi.core.cdi.MI2CDIException: Inferior terminated[]
     at
org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.getMIShareds(SharedLibraryManager.java:123)
     at
org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.updateState(SharedLibraryManager.java:188)
     at
org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.update(SharedLibraryManager.java:136)
     at
org.eclipse.cdt.debug.mi.core.cdi.EventManager.processSuspendedEvent(EventManager.java:326)
     at
org.eclipse.cdt.debug.mi.core.cdi.EventManager.update(EventManager.java:100)
     at java.util.Observable.notifyObservers(Observable.java:159)
     at
org.eclipse.cdt.debug.mi.core.MISession.notifyObservers(MISession.java:791)
     at org.eclipse.cdt.debug.mi.core.EventThread.run(EventThread.java:46)

Which looks like a CDT bug or the interfacing with CDT. I could file a
bug report if you think it is related to the poky-eclipse plugin?

Cheers,
Jack.
_______________________________________________
yocto mailing list
yocto@yoctoproject.org
https://lists.yoctoproject.org/listinfo/yocto

Re: QEmu Script Error Checking

[Jack Mitchell]

On 12/10/2011 18:52, Zhang, Jessica wrote:
> Hi Jack,
>
> When and how did you trigger the error.  It looks more like a CDT issue than a Yocto issue, but can you tell us more what exactly you were doing with Yocto plug-in while this crash happened?  So you mean when you click on the debug, it caused qemu crashes? Or the qemu crashed 1st. then you got the eclipse error msg.
>
> Thanks,
> Jessica
>
> -----Original Message-----
> From: yocto-bounces@yoctoproject.org [mailto:yocto-bounces@yoctoproject.org] On Behalf Of Jack Mitchell
> Sent: Wednesday, October 12, 2011 3:12 AM
> To: yocto@yoctoproject.org
> Subject: Re: [yocto] QEmu Script Error Checking
>
> On 11/10/2011 19:02, Scott Garman wrote:
>> On 10/11/2011 08:41 AM, Jack Mitchell wrote:
>>> I have run into the following issue where the tap device is in use due
>>> to the QEmu machine unexpectedly crashing. The error I recieve is here:
>>>
>>> http://i.imgur.com/5t9U1.png (appologies for the screenshot but it
>>> wouldn't let me copy the text)
>>>
>>> Is there any leaway for more robust error checking or a way to forcibly
>>> destroy the tap node to allow a new one to be created?
>> Hi Jack,
>>
>> Since qemu is being run from a parent shell script (runqemu), if qemu
>> were to crash, I would think the parent shell script would continue on
>> and destroy the tap device normally. Might you be killing the runqemu
>> process instead of qemu itself?
>>
>> I'm not sure if there is a way we could reliably force a cleanup of
>> tap devices when runqemu starts, because we need to support the case
>> where multiple instances of qemu sessions are running simultaneously
>> (each with their own tap device). Off the top of my head I think this
>> would make the state of tap devices non-deterministic.
>>
>> Furthermore, we support a mode where an administrator can set up one
>> or more tap devices, allowing the runqemu user to not need sudo
>> privileges. So checking for the case where a tap device exists but no
>> qemu process is running wouldn't work.
>>
>> Scott
>>
> Ok, thank you for the information Scott. I think I found the issue today
> which is causing the QEmu script to terminate pre-maturely.
>
> I am using the eclipse-poky master branch as my eclipse plugin and it
> threw an error today when I tried to debug - which coincided with the
> QEmu and terminal window terminating. The errror I recieved was:
>
> !SESSION 2011-10-12 09:22:49.135
> -----------------------------------------------
> eclipse.buildId=I20110613-1736
> java.version=1.6.0_22
> java.vendor=Sun Microsystems Inc.
> BootLoader constants: OS=linux, ARCH=x86_64, WS=gtk, NL=en_GB
> Framework arguments:  -product org.eclipse.sdk.ide
> Command-line arguments:  -product org.eclipse.sdk.ide -data
> /home/developer/jackdaw/../runtime-EclipseApplication -dev
> file:/home/developer/jackdaw/.metadata/.plugins/org.eclipse.pde.core/Eclipse
> Application/dev.properties -os linux -ws gtk -arch x86_64 -consoleLog
>
> !ENTRY org.eclipse.cdt.debug.mi.core 4 42 2011-10-12 09:24:47.192
> !MESSAGE Internal Error
> !STACK 0
> org.eclipse.cdt.debug.mi.core.cdi.MI2CDIException: Inferior terminated[]
>       at
> org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.getMIShareds(SharedLibraryManager.java:123)
>       at
> org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.updateState(SharedLibraryManager.java:188)
>       at
> org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.update(SharedLibraryManager.java:136)
>       at
> org.eclipse.cdt.debug.mi.core.cdi.EventManager.processSuspendedEvent(EventManager.java:326)
>       at
> org.eclipse.cdt.debug.mi.core.cdi.EventManager.update(EventManager.java:100)
>       at java.util.Observable.notifyObservers(Observable.java:159)
>       at
> org.eclipse.cdt.debug.mi.core.MISession.notifyObservers(MISession.java:791)
>       at org.eclipse.cdt.debug.mi.core.EventThread.run(EventThread.java:46)
>
> Which looks like a CDT bug or the interfacing with CDT. I could file a
> bug report if you think it is related to the poky-eclipse plugin?
>
> Cheers,
> Jack.
> _______________________________________________
> yocto mailing list
> yocto@yoctoproject.org
> https://lists.yoctoproject.org/listinfo/yocto

Jessica,

I have attached another error log which seems to be more verbose when I 
had this issue earlier today. Some of the errors seem to propogate from 
my code but there are also some CDT errors in there. The log is located 
here: http://pastebin.com/JHuE838B

Regards,
Jack

Re: QEmu Script Error Checking

[Zhang, Jessica]

Hi Jack,

What is your target architecture?  There's a known issue (bug 1489) that tcf-agent may segfault on qemux86, but it's never caused the whole qemu to crash.  One way to work around your issue is before your do the remote debugging, can you bring up a terminal in qemu to make sure the tcf-agent is up and running, if it's not, you can manually bring it up then do the remote debug.

Thanks,
Jessica

-----Original Message-----
From: Jack Mitchell [mailto:ml@communistcode.co.uk]
Sent: Thursday, October 13, 2011 5:39 AM
To: Zhang, Jessica
Cc: yocto@yoctoproject.org
Subject: Re: [yocto] QEmu Script Error Checking

On 12/10/2011 18:52, Zhang, Jessica wrote:
> Hi Jack,
>
> When and how did you trigger the error.  It looks more like a CDT issue than a Yocto issue, but can you tell us more what exactly you were doing with Yocto plug-in while this crash happened?  So you mean when you click on the debug, it caused qemu crashes? Or the qemu crashed 1st. then you got the eclipse error msg.
>
> Thanks,
> Jessica
>
> -----Original Message-----
> From: yocto-bounces@yoctoproject.org [mailto:yocto-bounces@yoctoproject.org] On Behalf Of Jack Mitchell
> Sent: Wednesday, October 12, 2011 3:12 AM
> To: yocto@yoctoproject.org
> Subject: Re: [yocto] QEmu Script Error Checking
>
> On 11/10/2011 19:02, Scott Garman wrote:
>> On 10/11/2011 08:41 AM, Jack Mitchell wrote:
>>> I have run into the following issue where the tap device is in use due
>>> to the QEmu machine unexpectedly crashing. The error I recieve is here:
>>>
>>> http://i.imgur.com/5t9U1.png (appologies for the screenshot but it
>>> wouldn't let me copy the text)
>>>
>>> Is there any leaway for more robust error checking or a way to forcibly
>>> destroy the tap node to allow a new one to be created?
>> Hi Jack,
>>
>> Since qemu is being run from a parent shell script (runqemu), if qemu
>> were to crash, I would think the parent shell script would continue on
>> and destroy the tap device normally. Might you be killing the runqemu
>> process instead of qemu itself?
>>
>> I'm not sure if there is a way we could reliably force a cleanup of
>> tap devices when runqemu starts, because we need to support the case
>> where multiple instances of qemu sessions are running simultaneously
>> (each with their own tap device). Off the top of my head I think this
>> would make the state of tap devices non-deterministic.
>>
>> Furthermore, we support a mode where an administrator can set up one
>> or more tap devices, allowing the runqemu user to not need sudo
>> privileges. So checking for the case where a tap device exists but no
>> qemu process is running wouldn't work.
>>
>> Scott
>>
> Ok, thank you for the information Scott. I think I found the issue today
> which is causing the QEmu script to terminate pre-maturely.
>
> I am using the eclipse-poky master branch as my eclipse plugin and it
> threw an error today when I tried to debug - which coincided with the
> QEmu and terminal window terminating. The errror I recieved was:
>
> !SESSION 2011-10-12 09:22:49.135
> -----------------------------------------------
> eclipse.buildId=I20110613-1736
> java.version=1.6.0_22
> java.vendor=Sun Microsystems Inc.
> BootLoader constants: OS=linux, ARCH=x86_64, WS=gtk, NL=en_GB
> Framework arguments:  -product org.eclipse.sdk.ide
> Command-line arguments:  -product org.eclipse.sdk.ide -data
> /home/developer/jackdaw/../runtime-EclipseApplication -dev
> file:/home/developer/jackdaw/.metadata/.plugins/org.eclipse.pde.core/Eclipse
> Application/dev.properties -os linux -ws gtk -arch x86_64 -consoleLog
>
> !ENTRY org.eclipse.cdt.debug.mi.core 4 42 2011-10-12 09:24:47.192
> !MESSAGE Internal Error
> !STACK 0
> org.eclipse.cdt.debug.mi.core.cdi.MI2CDIException: Inferior terminated[]
>       at
> org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.getMIShareds(SharedLibraryManager.java:123)
>       at
> org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.updateState(SharedLibraryManager.java:188)
>       at
> org.eclipse.cdt.debug.mi.core.cdi.SharedLibraryManager.update(SharedLibraryManager.java:136)
>       at
> org.eclipse.cdt.debug.mi.core.cdi.EventManager.processSuspendedEvent(EventManager.java:326)
>       at
> org.eclipse.cdt.debug.mi.core.cdi.EventManager.update(EventManager.java:100)
>       at java.util.Observable.notifyObservers(Observable.java:159)
>       at
> org.eclipse.cdt.debug.mi.core.MISession.notifyObservers(MISession.java:791)
>       at org.eclipse.cdt.debug.mi.core.EventThread.run(EventThread.java:46)
>
> Which looks like a CDT bug or the interfacing with CDT. I could file a
> bug report if you think it is related to the poky-eclipse plugin?
>
> Cheers,
> Jack.
> _______________________________________________
> yocto mailing list
> yocto@yoctoproject.org
> https://lists.yoctoproject.org/listinfo/yocto

Jessica,

I have attached another error log which seems to be more verbose when I
had this issue earlier today. Some of the errors seem to propogate from
my code but there are also some CDT errors in there. The log is located
here: http://pastebin.com/JHuE838B

Regards,
Jack