|
|
New to Adaptive Partitioning, problems on PXA270
|
|
10/23/2008 1:23 PM
post15459
|
New to Adaptive Partitioning, problems on PXA270
As a first task I'd like to set up a debug partition so a user can telnet and log into a heavily loaded target.
A debug partition was set up : aps -b10 DevDebug
Then io-net and inetd are launched into the DevDebug partition.
pidin sched shows that io-net and inetd are running in the new partition.
After telneting into an unloaded system pidin shows that telnetd and sh are also assigned to the new partition.
But... when the system is heavily loaded we cannot telnet into the target. If we telnet into an unloaded target, then
apply the load, the shell becomes unresponsive until the system again becomes unloaded. If the load continues telnet
eventually disconnects.
It sort of seems like AP isn't working. I thought that maybe the flash driver and inflator needed to also be in the new
partition so commands could be loaded from flash, but shouldn't those threads inherit the new partition?
Heres the output of aps show:
+---- CPU Time ----+-- Critical Time --
Partition name id | Budget | Used | Budget | Used
--------------------+------------------+-------------------
System 0 | 90% | 37.26% | 100ms | 0.000ms
DevDebug 1 | 10% | 62.69% | 0ms | 0.000ms
--------------------+------------------+-------------------
Total | 100% | 99.95% |
Here's the output of pidin sched:
pid tid name prio cpu ExtSched STATE
1 1 procnto 0f 0 System READY
1 2 procnto 70r 0 System RECEIVE
1 3 procnto 10r 0 System RECEIVE
1 4 procnto 10r 0 System RECEIVE
1 5 procnto 10r 0 System RECEIVE
1 6 procnto 10r 0 System RECEIVE
1 7 procnto 70r 0 System RECEIVE
1 8 procnto 10r 0 DevDebug RUNNING
1 9 procnto 10r 0 System RECEIVE
1 10 procnto 10r 0 System RECEIVE
2 1 ot/devf-dbpxa270dp 10r 0 System SIGWAITINFO
2 2 ot/devf-dbpxa270dp 10r 0 System RECEIVE
2 3 ot/devf-dbpxa270dp 10r 0 System RECEIVE
2 5 ot/devf-dbpxa270dp 10r 0 System RECEIVE
2 6 ot/devf-dbpxa270dp 10r 0 System RECEIVE
3 1 oot/devc-serpxa250 10r 0 System RECEIVE
4 1 proc/boot/slogger 30r 0 System RECEIVE
5 1 proc/boot/pipe 10r 0 System SIGWAITINFO
5 2 proc/boot/pipe 10r 0 System RECEIVE
5 3 proc/boot/pipe 10r 0 System RECEIVE
5 4 proc/boot/pipe 10r 0 System RECEIVE
5 5 proc/boot/pipe 10r 0 System RECEIVE
8199 1 sbin/inflator 10r 0 System RECEIVE
8199 2 sbin/inflator 10r 0 System RECEIVE
8199 3 sbin/inflator 10r 0 System RECEIVE
8199 4 sbin/inflator 24r 0 System RECEIVE
53256 1 sbin/io-net 10r 0 DevDebug SIGWAITINFO
53256 2 sbin/io-net 21r 0 DevDebug RECEIVE
53256 3 sbin/io-net 18r 0 DevDebug RECEIVE
53256 4 sbin/io-net 10r 0 DevDebug RECEIVE
53256 5 sbin/io-net 21r 0 DevDebug RECEIVE
53256 6 sbin/io-net 17f 0 DevDebug CONDVAR
53256 7 sbin/io-net 10r 0 DevDebug RECEIVE
53256 8 sbin/io-net 18r 0 DevDebug ...
View Full Message
|
|
|
|
|
|
|
Re: New to Adaptive Partitioning, problems on PXA270
|
|
10/23/2008 3:06 PM
post15465
|
Re: New to Adaptive Partitioning, problems on PXA270
Err. I'm confused. Is the debug partition on the target machine you're telnetting into, or is it on the machine you're
telnetting from?
Ken Schumm wrote:
> As a first task I'd like to set up a debug partition so a user can telnet and log into a heavily loaded target.
>
> A debug partition was set up : aps -b10 DevDebug
>
> Then io-net and inetd are launched into the DevDebug partition.
>
> pidin sched shows that io-net and inetd are running in the new partition.
>
> After telneting into an unloaded system pidin shows that telnetd and sh are also assigned to the new partition.
>
> But... when the system is heavily loaded we cannot telnet into the target. If we telnet into an unloaded target, then
apply the load, the shell becomes unresponsive until the system again becomes unloaded. If the load continues telnet
eventually disconnects.
>
> It sort of seems like AP isn't working. I thought that maybe the flash driver and inflator needed to also be in the
new partition so commands could be loaded from flash, but shouldn't those threads inherit the new partition?
>
> Heres the output of aps show:
> +---- CPU Time ----+-- Critical Time --
> Partition name id | Budget | Used | Budget | Used
> --------------------+------------------+-------------------
> System 0 | 90% | 37.26% | 100ms | 0.000ms
> DevDebug 1 | 10% | 62.69% | 0ms | 0.000ms
> --------------------+------------------+-------------------
> Total | 100% | 99.95% |
>
> Here's the output of pidin sched:
>
> pid tid name prio cpu ExtSched STATE
> 1 1 procnto 0f 0 System READY
> 1 2 procnto 70r 0 System RECEIVE
> 1 3 procnto 10r 0 System RECEIVE
> 1 4 procnto 10r 0 System RECEIVE
> 1 5 procnto 10r 0 System RECEIVE
> 1 6 procnto 10r 0 System RECEIVE
> 1 7 procnto 70r 0 System RECEIVE
> 1 8 procnto 10r 0 DevDebug RUNNING
> 1 9 procnto 10r 0 System RECEIVE
> 1 10 procnto 10r 0 System RECEIVE
> 2 1 ot/devf-dbpxa270dp 10r 0 System SIGWAITINFO
> 2 2 ot/devf-dbpxa270dp 10r 0 System RECEIVE
> 2 3 ot/devf-dbpxa270dp 10r 0 System RECEIVE
> 2 5 ot/devf-dbpxa270dp 10r 0 System RECEIVE
> 2 6 ot/devf-dbpxa270dp 10r 0 System RECEIVE
> 3 1 oot/devc-serpxa250 10r 0 System RECEIVE
> 4 1 proc/boot/slogger 30r 0 System RECEIVE
> 5 1 proc/boot/pipe 10r 0 System SIGWAITINFO
> 5 2 proc/boot/pipe 10r 0 System RECEIVE
> 5 3 proc/boot/pipe 10r 0 System RECEIVE
> 5 4 proc/boot/pipe 10r 0 System RECEIVE
> 5 5 proc/boot/pipe 10r 0 System RECEIVE
> 8199 1 sbin/inflator 10r 0 System RECEIVE
> 8199 2 sbin/inflator 10r 0 System RECEIVE
> 8199 3 sbin/inflator 10r 0 System RECEIVE
> 8199 4 sbin/inflator 24r 0 System RECEIVE
> 53256 1 sbin/io-net 10r 0 DevDebug SIGWAITINFO
> 53256 2 sbin/io-net 21r 0 DevDebug RECEIVE ...
View Full Message
|
|
|
|
|
|
|
Re: New to Adaptive Partitioning, problems on PXA270
|
|
10/23/2008 3:24 PM
post15467
|
Re: New to Adaptive Partitioning, problems on PXA270
I'm telnetting from a Windows host to a QNX target. Of course this means the partition is also on the QNX target.
|
|
|
|
|
|
|
RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/23/2008 3:38 PM
post15469
|
RE: New to Adaptive Partitioning, problems on PXA270
One thing I saw was that your devc-pty is not running in the debug
partition. That is required for telnet AFAIK.
--
Cheers,
Adam
QNX Software Systems
[ amallory@harman.com ]
---------------------------------------------------
With a PC, I always felt limited by the software available.
On Unix, I am limited only by my knowledge.
--Peter J. Schoenster
> -----Original Message-----
> From: Ken Schumm [mailto:community-noreply@qnx.com]
> Sent: Thursday, October 23, 2008 3:25 PM
> To: ostech-core_os
> Subject: Re: New to Adaptive Partitioning, problems on PXA270
>
> I'm telnetting from a Windows host to a QNX target. Of course this
means
> the partition is also on the QNX target.
>
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post15467
|
|
|
|
|
|
|
Re: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/23/2008 3:53 PM
post15471
|
Re: RE: New to Adaptive Partitioning, problems on PXA270
> One thing I saw was that your devc-pty is not running in the debug
> partition. That is required for telnet AFAIK.
>
>
> --
> Cheers,
> Adam
Thanks, but moving devc-pty to the debug partition didn't help.
It shouldn't matter since the server thread should inherit cycles from the partition being served, right?
|
|
|
|
|
|
|
RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/23/2008 4:26 PM
post15472
|
RE: RE: New to Adaptive Partitioning, problems on PXA270
> It shouldn't matter since the server thread should inherit cycles from
the
> partition being served, right?
I just wanted to mention the pty requirement for telnet. You should
capture a kernel instrumentation log using tracelogger to gain more
insight on what is going on and hopefully answer why.
--
Cheers,
Adam
QNX Software Systems
[ amallory@harman.com ]
---------------------------------------------------
With a PC, I always felt limited by the software available.
On Unix, I am limited only by my knowledge.
--Peter J. Schoenster
|
|
|
|
|
|
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 11:08 AM
post15564
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
To simplify the problem I started a login shell on the serial port. The shell is running in the DevDebug partition with
10% budget. When the target is heavily loaded any command issued on this shell hangs as if it cannot be loaded from
flash. This is the same behavior that occurs on a telnet shell. So it looks the flash driver is not inheriting the cpu
budget from the partition.
|
|
|
|
|
|
|
RE: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 11:12 AM
post15565
|
RE: RE: RE: New to Adaptive Partitioning, problems on PXA270
Ah, that's different from your original post. I interpreted that you
could not type at all (ie. no character echo). Not being able to start
commands is a different ball game.
What shell are you using?
Can you copy the same command/binary to /dev/shmem and attempt to
execute it to see if you can still reproduce the issue?
--
Cheers,
Adam
QNX Software Systems
[ amallory@harman.com ]
---------------------------------------------------
With a PC, I always felt limited by the software available.
On Unix, I am limited only by my knowledge.
--Peter J. Schoenster
> -----Original Message-----
> From: Ken Schumm [mailto:community-noreply@qnx.com]
> Sent: Monday, October 27, 2008 11:09 AM
> To: ostech-core_os
> Subject: Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
>
> To simplify the problem I started a login shell on the serial port.
The
> shell is running in the DevDebug partition with 10% budget. When the
> target is heavily loaded any command issued on this shell hangs as if
it
> cannot be loaded from flash. This is the same behavior that occurs on
a
> telnet shell. So it looks the flash driver is not inheriting the cpu
> budget from the partition.
>
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post15564
|
|
|
|
|
|
|
Re: RE: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 11:27 AM
post15566
|
Re: RE: RE: RE: New to Adaptive Partitioning, problems on PXA270
Sorry if my original post was not clear, the word "unresponsive" was not specific. Keystrokes always echo, but commands
do not load and execute when the target is under load.
I copied pidin to /dev/shmem, made it executable, applied a load to the target, then tried to run /dev/shmem/pidin from
the serial shell and it would not load and execute.
After reducing the load on the target it ran just fine.
I double checked and the shell is definitely running in the 10% budget DevDebug partition.
|
|
|
|
|
|
|
RE: RE: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 1:04 PM
post15571
|
RE: RE: RE: RE: New to Adaptive Partitioning, problems on PXA270
What is priority is the CPU load you're running? Is it a straight CPU
burning example (have you tried that as well?)?
--
Cheers,
Adam
QNX Software Systems
[ amallory@harman.com ]
---------------------------------------------------
With a PC, I always felt limited by the software available.
On Unix, I am limited only by my knowledge.
--Peter J. Schoenster
> -----Original Message-----
> From: Ken Schumm [mailto:community-noreply@qnx.com]
> Sent: Monday, October 27, 2008 11:28 AM
> To: ostech-core_os
> Subject: Re: RE: RE: RE: New to Adaptive Partitioning, problems on
PXA270
>
> Sorry if my original post was not clear, the word "unresponsive" was
not
> specific. Keystrokes always echo, but commands do not load and execute
> when the target is under load.
>
> I copied pidin to /dev/shmem, made it executable, applied a load to
the
> target, then tried to run /dev/shmem/pidin from the serial shell and
it
> would not load and execute.
>
> After reducing the load on the target it ran just fine.
>
> I double checked and the shell is definitely running in the 10% budget
> DevDebug partition.
>
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post15566
|
|
|
|
|
|
|
Re: RE: RE: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 1:37 PM
post15572
|
Re: RE: RE: RE: RE: New to Adaptive Partitioning, problems on PXA270
The shell is 10r, so anything it loads would be the same.
I wrote a quick program to delay(500) then do a printf and it continues just fine under load. Under load you can also
type at the shell prompt, or just hit enter, and the prompt comes right back.
The problem seems to occur when commands are loaded when the system is very busy.
Even at 10r the partition should get it's budget of cpu cycles, right?
Thanks,
Ken
> What is priority is the CPU load you're running? Is it a straight CPU
> burning example (have you tried that as well?)?
>
>
> --
> Cheers,
> Adam
>
> QNX Software Systems
> [ amallory@harman.com ]
> ---------------------------------------------------
> With a PC, I always felt limited by the software available.
> On Unix, I am limited only by my knowledge.
> --Peter J. Schoenster
>
> > -----Original Message-----
> > From: Ken Schumm [mailto:community-noreply@qnx.com]
> > Sent: Monday, October 27, 2008 11:28 AM
> > To: ostech-core_os
> > Subject: Re: RE: RE: RE: New to Adaptive Partitioning, problems on
> PXA270
> >
> > Sorry if my original post was not clear, the word "unresponsive" was
> not
> > specific. Keystrokes always echo, but commands do not load and execute
> > when the target is under load.
> >
> > I copied pidin to /dev/shmem, made it executable, applied a load to
> the
> > target, then tried to run /dev/shmem/pidin from the serial shell and
> it
> > would not load and execute.
> >
> > After reducing the load on the target it ran just fine.
> >
> > I double checked and the shell is definitely running in the 10% budget
> > DevDebug partition.
> >
> > _______________________________________________
> > OSTech
> > http://community.qnx.com/sf/go/post15566
|
|
|
|
|
|
|
RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 1:44 PM
post15573
|
RE: New to Adaptive Partitioning, problems on PXA270
For giggles can you boost the priority of the shell in your debug
partition to 11 (or another higher priority that no other tasks in ANY
partition are running at)?
Have you taken a kernel trace of the scenario yet?
--
Cheers,
Adam
QNX Software Systems
[ amallory@harman.com ]
---------------------------------------------------
With a PC, I always felt limited by the software available.
On Unix, I am limited only by my knowledge.
--Peter J. Schoenster
> -----Original Message-----
> From: Ken Schumm [mailto:community-noreply@qnx.com]
> Sent: Monday, October 27, 2008 1:38 PM
> To: ostech-core_os
> Subject: Re: RE: RE: RE: RE: New to Adaptive Partitioning, problems on
> PXA270
>
> The shell is 10r, so anything it loads would be the same.
>
> I wrote a quick program to delay(500) then do a printf and it
continues
> just fine under load. Under load you can also type at the shell
prompt, or
> just hit enter, and the prompt comes right back.
>
> The problem seems to occur when commands are loaded when the system is
> very busy.
>
> Even at 10r the partition should get it's budget of cpu cycles, right?
>
> Thanks,
> Ken
>
> > What is priority is the CPU load you're running? Is it a straight
CPU
> > burning example (have you tried that as well?)?
> >
> >
> > --
> > Cheers,
> > Adam
> >
> > QNX Software Systems
> > [ amallory@harman.com ]
> > ---------------------------------------------------
> > With a PC, I always felt limited by the software available.
> > On Unix, I am limited only by my knowledge.
> > --Peter J. Schoenster
> >
> > > -----Original Message-----
> > > From: Ken Schumm [mailto:community-noreply@qnx.com]
> > > Sent: Monday, October 27, 2008 11:28 AM
> > > To: ostech-core_os
> > > Subject: Re: RE: RE: RE: New to Adaptive Partitioning, problems on
> > PXA270
> > >
> > > Sorry if my original post was not clear, the word "unresponsive"
was
> > not
> > > specific. Keystrokes always echo, but commands do not load and
execute
> > > when the target is under load.
> > >
> > > I copied pidin to /dev/shmem, made it executable, applied a load
to
> > the
> > > target, then tried to run /dev/shmem/pidin from the serial shell
and
> > it
> > > would not load and execute.
> > >
> > > After reducing the load on the target it ran just fine.
> > >
> > > I double checked and the shell is definitely running in the 10%
budget
> > > DevDebug partition.
> > >
> > > _______________________________________________
> > > OSTech
> > > http://community.qnx.com/sf/go/post15566
>
>
>
>
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post15572
|
|
|
|
|
|
|
Re: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 1:55 PM
post15574
|
Re: RE: New to Adaptive Partitioning, problems on PXA270
Our highest priority process runs at 80 so I boosted the shell to 81.
It worked.
I'm not quite sure how to capture a kernel trace when this problem occurs, there are no cycles left to do it (can't
connect via tcp/ip or issue commands from the shell when this happens).
> For giggles can you boost the priority of the shell in your debug
> partition to 11 (or another higher priority that no other tasks in ANY
> partition are running at)?
>
> Have you taken a kernel trace of the scenario yet?
>
>
> --
> Cheers,
> Adam
>
> QNX Software Systems
> [ amallory@harman.com ]
> ---------------------------------------------------
> With a PC, I always felt limited by the software available.
> On Unix, I am limited only by my knowledge.
> --Peter J. Schoenster
[...]
|
|
|
|
|
|
|
RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 2:02 PM
post15578
|
RE: RE: New to Adaptive Partitioning, problems on PXA270
Maybe you could do something like this on your serial terminal:
1. # tracelogger -P 81 -c -n0 &
2. put system under load
3. try to run pidin:
# pidin
4. remove load from system
5. # slay -f tracelogger
Ideally the time between 1. and 5. will be no more than a couple of seconds.
--
Ryan J. Allen
QNX Software Systems
-----Original Message-----
From: Ken Schumm [mailto:community-noreply@qnx.com]
Sent: Mon 2008-10-27 18:55
To: ostech-core_os
Subject: Re: RE: New to Adaptive Partitioning, problems on PXA270
Our highest priority process runs at 80 so I boosted the shell to 81.
It worked.
I'm not quite sure how to capture a kernel trace when this problem occurs, there are no cycles left to do it (can't connect via tcp/ip or issue commands from the shell when this happens).
> For giggles can you boost the priority of the shell in your debug
> partition to 11 (or another higher priority that no other tasks in ANY
> partition are running at)?
>
> Have you taken a kernel trace of the scenario yet?
>
>
> --
> Cheers,
> Adam
>
> QNX Software Systems
> [ amallory@harman.com ]
> ---------------------------------------------------
> With a PC, I always felt limited by the software available.
> On Unix, I am limited only by my knowledge.
> --Peter J. Schoenster
[...]
_______________________________________________
OSTech
http://community.qnx.com/sf/go/post15574
|
|
|
|
|
|
|
Re: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 2:42 PM
post15590
|
Re: New to Adaptive Partitioning, problems on PXA270
-P is not a 6.3.2 tracelogger option - use on to set the priority.
I would also recommend using direct mapped mode, eg
tracelogger -S32M -M -n0
Ryan Allen wrote:
> Maybe you could do something like this on your serial terminal:
> 1. # tracelogger -P 81 -c -n0 &
> 2. put system under load
> 3. try to run pidin:
> # pidin
> 4. remove load from system
> 5. # slay -f tracelogger
>
> Ideally the time between 1. and 5. will be no more than a couple of seconds.
>
--
cburgess@qnx.com
|
|
|
|
|
|
|
Re: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 3:41 PM
post15593
|
Re: New to Adaptive Partitioning, problems on PXA270
> -P is not a 6.3.2 tracelogger option - use on to set the priority.
>
> I would also recommend using direct mapped mode, eg
>
> tracelogger -S32M -M -n0
If it was easy everyone would be doing it :)
Using Colin's suggestion there is now a 12MB event file in /dev/shmem on the target. Unfortunately, any attempt to ftp
it to the host for examination results in an ftp error "550 Not a plain file". I've tried binary and ascii modes with
the same result. Turns out the file looks like a named special file (n set in the permissions mask). If it can't be
ftp'd over to the host it's not of much use :(
|
|
|
|
|
|
|
Re: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 3:47 PM
post15594
|
Re: New to Adaptive Partitioning, problems on PXA270
The IDE target filesystem view can transfer it (albeit slowly). Do you have fs-nfs or fs-cifs on the target -that would
be fastest.
Colin
Ken Schumm wrote:
>> -P is not a 6.3.2 tracelogger option - use on to set the priority.
>>
>> I would also recommend using direct mapped mode, eg
>>
>> tracelogger -S32M -M -n0
>
> If it was easy everyone would be doing it :)
>
> Using Colin's suggestion there is now a 12MB event file in /dev/shmem on the target. Unfortunately, any attempt to ftp
it to the host for examination results in an ftp error "550 Not a plain file". I've tried binary and ascii modes with
the same result. Turns out the file looks like a named special file (n set in the permissions mask). If it can't be
ftp'd over to the host it's not of much use :(
>
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post15593
>
--
cburgess@qnx.com
|
|
|
|
|
|
|
Re: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 4:00 PM
post15596
|
Re: New to Adaptive Partitioning, problems on PXA270
The IDE transfer wasn't so bad, maybe a minute.
Thanks.
> The IDE target filesystem view can transfer it (albeit slowly). Do you have
> fs-nfs or fs-cifs on the target -that would be fastest.
>
> Colin
>
> Ken Schumm wrote:
> >> -P is not a 6.3.2 tracelogger option - use on to set the priority.
> >>
> >> I would also recommend using direct mapped mode, eg
> >>
> >> tracelogger -S32M -M -n0
> >
> > If it was easy everyone would be doing it :)
> >
> > Using Colin's suggestion there is now a 12MB event file in /dev/shmem on the
> target. Unfortunately, any attempt to ftp it to the host for examination
> results in an ftp error "550 Not a plain file". I've tried binary and ascii
> modes with the same result. Turns out the file looks like a named special file
> (n set in the permissions mask). If it can't be ftp'd over to the host it's
> not of much use :(
> >
> > _______________________________________________
> > OSTech
> > http://community.qnx.com/sf/go/post15593
> >
>
> --
> cburgess@qnx.com
|
|
|
|
|
|
|
RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 2:02 PM
post15579
|
RE: RE: New to Adaptive Partitioning, problems on PXA270
> Our highest priority process runs at 80 so I boosted the shell to 81.
>
> It worked.
Sorry that isn't quite what I meant. I still wanted a lowerish priority
but NOT with any other tasks at that priority (even from different
partitions), so I mentioned 11 as the next available, but only you could
really know the answer here.
> I'm not quite sure how to capture a kernel trace when this problem
occurs,
> there are no cycles left to do it (can't connect via tcp/ip or issue
> commands from the shell when this happens).
Tracelogger at priority 81 sounds like it should work. You could log 5
seconds worth of data - so start tracelogger, apply the CPU load and
attempt to telnet in during that time. After, just remove the load and
transfer the resulting .kev file (default location is /dev/shmem) off
the board and into your IDE for visualization.
--
Cheers,
Adam
QNX Software Systems
[ amallory@harman.com ]
---------------------------------------------------
With a PC, I always felt limited by the software available.
On Unix, I am limited only by my knowledge.
--Peter J. Schoenster
>
> > For giggles can you boost the priority of the shell in your debug
> > partition to 11 (or another higher priority that no other tasks in
ANY
> > partition are running at)?
> >
> > Have you taken a kernel trace of the scenario yet?
> >
> >
> > --
> > Cheers,
> > Adam
> >
> > QNX Software Systems
> > [ amallory@harman.com ]
> > ---------------------------------------------------
> > With a PC, I always felt limited by the software available.
> > On Unix, I am limited only by my knowledge.
> > --Peter J. Schoenster
> [...]
>
>
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post15574
|
|
|
|
|
|
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 2:22 PM
post15584
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
OK, I first tried the shell priority at 15, which is above photon, graphics, and our gui and below most everything else.
That didn't work.
Next the shell priority was set to 40, which is above everything except our watchdog kicker and two interrupt threads.
That didn't work either. The watchdog kicker is at priority 80 and only wakes up once a second and writes to a port. One
of the interrupt threads is at priority 70 to manage the power key (to power the unit off even if the software goes
nuts). The other interrupt thread handles a DSP interrupt at up to 10kHz. This data interrupt is what we can vary with
an ARB to increase/decrease the target load. At 10 kHz there are enough cycles for the GUI to process data and update
the screen at priority 11.
I'll do an event analysis after lunch.
|
|
|
|
|
|
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 2:35 PM
post15587
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
Tracelogger reports a whole bunch of messages like this:
"Help, we're dropping buffers! <736 dropped so far>"
It was running at priority 81.
|
|
|
|
|
|
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
10/27/2008 6:14 PM
post15600
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
[...]
>
> I'll do an event analysis after lunch.
After examining the event log, what appears to be happening in the case of the serial port shell is that keystrokes from
the host are sent to devc-serpxa250 (the interrupts show characters arriving), and the driver is echoing the characters
back to the host terminal, but when the unit is under load the typed in command string is never passed to the shell
(there seems to be no reply done from the serial port driver to a shell).
|
|
|
|
|
|
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/03/2008 3:10 PM
post15874
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
are there any news available on this issue?
this is a realy intressting thread, cause we are also using APS
and there isn't realy an explanation for this behaviour.
/hp
|
|
|
|
|
|
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/03/2008 3:46 PM
post15878
|
Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
> are there any news available on this issue?
> this is a realy intressting thread, cause we are also using APS
> and there isn't realy an explanation for this behaviour.
> /hp
Are you having the same problem? I thought maybe our situation was unusual.
|
|
|
|
|
|
|
AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/07/2008 1:43 AM
post16153
|
AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
no, we do not see exactly the same problem.
>-----Ursprüngliche Nachricht-----
>Von: Ken Schumm [mailto:community-noreply@qnx.com]
>Gesendet: Montag, 3. November 2008 21:46
>An: ostech-core_os
>Betreff: Re: RE: RE: New to Adaptive Partitioning, problems on PXA270
>
>> are there any news available on this issue?
>> this is a realy intressting thread, cause we are also using APS and
>> there isn't realy an explanation for this behaviour.
>> /hp
>
>
>Are you having the same problem? I thought maybe our
>situation was unusual.
>
>_______________________________________________
>OSTech
>http://community.qnx.com/sf/go/post15878
>
>
*******************************************
Harman Becker Automotive Systems GmbH
Management Board: Dr. Klaus Blickle (Chairman), Dr. Udo Hüls, Michael Mauser, Regis Baudot
Chairman of the Supervisory Board: Ansgar Rempp | Domicile: Karlsbad |
Local Court Mannheim: Register No. 361395
*******************************************
Diese E-Mail enthaelt vertrauliche und/oder rechtlich geschuetzte Informationen. Wenn Sie nicht der richtige Adressat
sind oder diese E-Mail irrtuemlich erhalten haben, informieren Sie bitte sofort den Absender und loeschen Sie diese Mail
. Das unerlaubte Kopieren sowie die unbefugte Weitergabe dieser Mail ist nicht gestattet.
This e-mail may contain confidential and/or privileged information. If you are not the intended recipient (or have
received this e-mail in error) please notify the sender immediately and delete this e-mail. Any unauthorized copying,
disclosure or distribution of the contents in this e-mail is strictly forbidden.
*******************************************
|
|
|
|
|
|
|
Re: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/13/2008 3:28 PM
post16533
|
Re: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
It was requested that I post a kernel event trace so here it is. This is the trace from the session mentioned above
where characters were typed at the host PC, received by devc-serpxa250 and never sent to sh which is running in a
partition that was budgeted 10% of cpu IIRC.
|
|
|
|
|
|
|
Re: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/14/2008 8:46 AM
post16577
|
Re: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
> It was requested that I post a kernel event trace so here it is. This is the
> trace from the session mentioned above where characters were typed at the host
> PC, received by devc-serpxa250 and never sent to sh which is running in a
> partition that was budgeted 10% of cpu IIRC.
Is the shell in question that with pid 426014 ?
Following is my view about what happened:
shell is receive blocked.
The uart indicates your devc-ser that data is there.
IntrHandler is invoked event is returned and delivered to devc-ser..
Event, Time, Owner, Type, Data
160144, 4s 883ms 976us, Interrupt 0x16, Entry, interrupt 0x16 ip 0x10e830
160149, 4s 884ms 29us, Interrupt 0x16 devc-serpxa250, Handler Entry, \
pid 3 interrupt 0x16 ip 0x1022f8 area 0x11d010 process devc-serpxa250
160150, 4s 884ms 68us, Interrupt 0x16 devc-serpxa250, Sigevent Pulse,\
scoid 0x40000003 pid 3 process devc-serpxa250
160151, 4s 884ms 83us, devc-serpxa250 Thread 1, Ready, pid 3 tid 1
160152, 4s 884ms 91us, Interrupt 0x16 devc-serpxa250, Handler Exit,\
interrupt 0x16 sigev_notify 4
Event, Time, Owner, Type, Data
160153, 4s 884ms 98us, Interrupt 0x16, Exit, interrupt 0x16 flags 0x1
..
..
..
Event, Time, Owner, Type, Data
160247, 4s 885ms 825us, devc-serpxa250 Thread 1, Receive Pulse, scoid \
0x40000003 pid 3 process devc-serpxa250
...
...
...
...
...
Event, Time, Owner, Type, Data
161064, 4s 902ms 277us, devc-serpxa250 Thread 1, Running, pid 3 tid 1
161065, 4s 902ms 289us, devc-serpxa250 Thread 1, MsgReplyv Enter, \ rcvid 0x7 status 0x1
161070, 4s 902ms 373us, devc-serpxa250 Thread 1, Reply, tid 1 pid 426014 process sh
161071, 4s 902ms 386us, sh Thread 1, Ready, pid 426014 tid 1
Then Communication between those both goes ahead sh writes something ....devc replies and then sh forks
something and goes to sigsuspend probably waiting for some child to exit ?
For me it is interresting why Your shell is going to sigsupend blocked state....??
For whom is it waiting for ?
I do not see that this has at all some thing to do with APS etc..
Perhaps somebody else sees more and more clear...
Jeevan
|
|
|
|
|
|
|
Re: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/14/2008 11:41 AM
post16586
|
Re: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
[...]
>
> Then Communication between those both goes ahead sh writes something ....devc
> replies and then sh forks
> something and goes to sigsuspend probably waiting for some child to exit ?
>
> For me it is interresting why Your shell is going to sigsupend blocked state..
> ..??
>
> For whom is it waiting for ?
>
> I do not see that this has at all some thing to do with APS etc..
The sigsuspend is not important, what you observed is how the application works. There is one serial port. It is one of
three available interface ports (serial, gpib, usb) that can be used from a host PC to our unit. The host PC normally
sends SCPI commands to our unit for handling by a communication process.
We needed a backdoor way to take over the serial port for debug purposes. When a special command is sent a process is
spawned to take over the serial port, spawns a child shell that inherits the serial port fds, and goes into sigsuspend
waiting for the child shell to exit. At this point any commands sent over the serial port are directed to the child
shell. If 'exit' is sent then the child shell exits, the sigsuspend cleans up, and command is returned to the normal
SCPI command parser.
It's simply a way to take over the serial port for debug.
The APS problem occurs *after* the child shell is up and running. If the system is loaded, and a command is sent over
the serial port to the child shell, *that* command is not executed due to lack of CPU cycles. It works fine on an
unloaded system, it fails on a heavily loaded system.
If this is too confusing I can go back to using telnet which has the same problem. Shell commands issued to a telnet
shell don't work either when the system is loaded even though all inetd/telnetd/sh related processes are in their own AP
partition with a 10% budget. In fact, any existing telnet shell with become unresponsive when the system is loaded, and
eventually the host will disconnect.
The serial case seemed simpler since it didn't have as many processes and config/password files to handle.
|
|
|
|
|
|
|
AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/14/2008 2:40 PM
post16603
|
AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
Hi Ken,
can you provide a trace done with wide mode tracing (-w)?
Would be good to see more details on the pulse object that is send at event 160150
At which time/point do you initialize your pulse object?
If you use SIGEV_PULSE_PRIO_INHERIT, then the pulse gets' the process' initial priority information when sending the
pulse, not the one the thread had when you created the pulse object (check nano_event.c).
So maybe your IR handler is running in a different partition when the pulse is registered for the ISR.
Checking nano_event.c .... I guess that you should use PULSE_DELIVER_APS_CRITICAL_FLAG
For your pulse object your registring on the ISR.
Any comments from you QNX guys?
/hp
-----Ursprüngliche Nachricht-----
Von: Ken Schumm [mailto:community-noreply@qnx.com]
Gesendet: Fr 14.11.2008 17:41
An: ostech-core_os
Cc:
Betreff: Re: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
[...]
>
> Then Communication between those both goes ahead sh writes something ....devc
> replies and then sh forks
> something and goes to sigsuspend probably waiting for some child to exit ?
>
> For me it is interresting why Your shell is going to sigsupend blocked state..
> ..??
>
> For whom is it waiting for ?
>
> I do not see that this has at all some thing to do with APS etc..
The sigsuspend is not important, what you observed is how the application works. There is one serial port. It is one of
three available interface ports (serial, gpib, usb) that can be used from a host PC to our unit. The host PC normally
sends SCPI commands to our unit for handling by a communication process.
We needed a backdoor way to take over the serial port for debug purposes. When a special command is sent a process is
spawned to take over the serial port, spawns a child shell that inherits the serial port fds, and goes into sigsuspend
waiting for the child shell to exit. At this point any commands sent over the serial port are directed to the child
shell. If 'exit' is sent then the child shell exits, the sigsuspend cleans up, and command is returned to the normal
SCPI command parser.
It's simply a way to take over the serial port for debug.
The APS problem occurs *after* the child shell is up and running. If the system is loaded, and a command is sent over
the serial port to the child shell, *that* command is not executed due to lack of CPU cycles. It works fine on an
unloaded system, it fails on a heavily loaded system.
If this is too confusing I can go back to using telnet which has the same problem. Shell commands issued to a telnet
shell don't work either when the system is loaded even though all inetd/telnetd/sh related processes are in their own AP
partition with a 10% budget. In fact, any existing telnet shell with become unresponsive when the system is loaded, and
eventually the host will disconnect.
The serial case seemed simpler since it didn't have as many processes and config/password files to handle.
_______________________________________________
OSTech
http://community.qnx.com/sf/go/post16586
*******************************************
Harman Becker Automotive Systems GmbH
Management Board: Dr. Klaus Blickle (Chairman), Dr. Udo Hüls, Michael Mauser, Regis Baudot
Chairman of the Supervisory Board: Ansgar Rempp | Domicile: Karlsbad |
Local Court Mannheim: Register No. 361395
*******************************************
Diese E-Mail enthaelt vertrauliche und/oder rechtlich geschuetzte Informationen. Wenn Sie nicht der richtige Adressat
sind oder diese E-Mail irrtuemlich erhalten haben, informieren Sie bitte sofort den Absender und loeschen Sie diese Mail
. Das unerlaubte Kopieren sowie die unbefugte Weitergabe dieser Mail ist nicht gestattet.
This e-mail may contain confidential and/or privileged information. If you are not the intended recipient (or have
received this e-mail in error) please notify the sender immediately and...
View Full Message
|
|
|
|
|
|
|
Re: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/14/2008 3:01 PM
post16604
|
Re: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
I'm off today but will capture a wide trace next week.
All I'm trying to do is set up a 10% budget partition that we can login to to run QNX utilities like pidin, sloginfo,
etc.
In the telnet case there is not a single piece of our code running in the AP partition and any call to a QNX driver
(disk, character) fails when the system is loaded. Telnet cannot even maintain a connection when the system is loaded
and it is in the 10% partition.
It could be a high interrupt load but that wouldn't explain why other components of ours continue to work (GUI refreshes
and is responsive, the SCPI host commands all work, writing to flash works, etc).
|
|
|
|
|
|
|
Re: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/17/2008 10:37 AM
post16675
|
Re: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
I ran another test today and captured a wide event trace log (attached).
A pidin sched prior to the test shows this:
------------------------------------------------------------------------------------------------------
pid tid name prio cpu ExtSched STATE
1 1 procnto 0f 0 System READY
1 2 procnto 10r 0 System RECEIVE
1 3 procnto 10r 0 System RECEIVE
1 4 procnto 10r 0 System RECEIVE
1 5 procnto 10r 0 System RECEIVE
1 6 procnto 10r 0 System RECEIVE
1 7 procnto 10r 0 System RECEIVE
1 8 procnto 10r 0 System RECEIVE
1 9 procnto 10r 0 System RECEIVE
1 10 procnto 10r 0 DevDebug RUNNING
1 11 procnto 10r 0 System RECEIVE
2 1 ot/devf-dbpxa270dp 10r 0 System SIGWAITINFO
2 2 ot/devf-dbpxa270dp 10r 0 System RECEIVE
2 3 ot/devf-dbpxa270dp 30r 0 System RECEIVE
2 4 ot/devf-dbpxa270dp 10r 0 System RECEIVE
2 5 ot/devf-dbpxa270dp 10r 0 System RECEIVE
3 1 oot/devc-serpxa250 24r 0 System RECEIVE
4 1 proc/boot/slogger 30r 0 System RECEIVE
5 1 proc/boot/pipe 10r 0 System SIGWAITINFO
5 2 proc/boot/pipe 10r 0 System RECEIVE
5 3 proc/boot/pipe 10r 0 System RECEIVE
5 4 proc/boot/pipe 10r 0 System RECEIVE
5 5 proc/boot/pipe 10r 0 System RECEIVE
8199 1 sbin/inflator 10r 0 System RECEIVE
8199 2 sbin/inflator 10r 0 System RECEIVE
8199 3 sbin/inflator 10r 0 System RECEIVE
8199 4 sbin/inflator 10r 0 System RECEIVE
53256 1 sbin/io-net 10r 0 DevDebug SIGWAITINFO
53256 2 sbin/io-net 21r 0 DevDebug RECEIVE
53256 3 sbin/io-net 10r 0 DevDebug RECEIVE
53256 4 sbin/io-net 10r 0 DevDebug RECEIVE
53256 5 sbin/io-net 21r 0 DevDebug RECEIVE
53256 6 sbin/io-net 17f 0 DevDebug CONDVAR
53256 7 sbin/io-net 10r 0 DevDebug RECEIVE
53256 8 sbin/io-net 10r 0 DevDebug RECEIVE
53256 10 sbin/io-net 18f 0 DevDebug CONDVAR
77834 1 usr/sbin/inetd 10r 0 DevDebug SIGWAITINFO
81929 1 sbin/devc-pty 10r 0 DevDebug RECEIVE
106507 1 usr/sbin/qconn 10r 0 DevDebug SIGWAITINFO
106507 2 usr/sbin/qconn 10r 0 DevDebug CONDVAR
106507 3 usr/sbin/qconn 10r 0 DevDebug RECEIVE
106507 4 usr/sbin/qconn 10r 0 DevDebug RECEIVE
139276 1 abmax/bin/overseer 75r 0 System RECEIVE
176141 1 bmax/bin/LabMaxGUI 11r 0 System REPLY
176141 2 bmax/bin/LabMaxGUI 11r 0 System RECEIVE
176141 3 bmax/bin/LabMaxGUI 11r 0 System REPLY
192527 1 /photon/bin/Photon 11r 0 System RECEIVE
192528 1 on/bin/io-graphics 12r 0 System RECEIVE
192528 2 on/bin/io-graphics 11r 0 System RECEIVE
192528 3...
View Full Message
|
|
|
|
|
|
|
AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/17/2008 12:07 PM
post16682
|
AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
this is HOT!!!
your pidin request is about to start
within this it is doing its normal statup stuff
ending in a request to proc thread 2 which itself is doing a fseek64() that is serviced by proc thread 10
now thread 10 is trying to answer using MsgReply() but it cant't
whenever proc thread 10 is trying to do MsgReply() it gets interrupted by IR 0x10 which is handled by LMDispatcher in
partition 0
due to the fact that you have a real high IR rate of 3000Hz the MsgReply() can't finish.
Hey QNX guys wake up this is somethig to work on.
@Colin: do you remember that beta hotfix you've done for the MMI3G project where we had a quite simillar problem?
/hp
>-----Ursprüngliche Nachricht-----
>Von: Ken Schumm [mailto:community-noreply@qnx.com]
>Gesendet: Montag, 17. November 2008 16:39
>An: ostech-core_os
>Betreff: Re: AW: AW: RE: RE: New to Adaptive Partitioning,
>problems on PXA270
>
>I ran another test today and captured a wide event trace log
>(attached).
>
>A pidin sched prior to the test shows this:
>---------------------------------------------------------------
>---------------------------------------
> pid tid name prio cpu ExtSched
>STATE
> 1 1 procnto 0f 0 System
>READY
> 1 2 procnto 10r 0 System
>RECEIVE
> 1 3 procnto 10r 0 System
>RECEIVE
> 1 4 procnto 10r 0 System
>RECEIVE
> 1 5 procnto 10r 0 System
>RECEIVE
> 1 6 procnto 10r 0 System
>RECEIVE
> 1 7 procnto 10r 0 System
>RECEIVE
> 1 8 procnto 10r 0 System
>RECEIVE
> 1 9 procnto 10r 0 System
>RECEIVE
> 1 10 procnto 10r 0 DevDebug
>RUNNING
> 1 11 procnto 10r 0 System
>RECEIVE
> 2 1 ot/devf-dbpxa270dp 10r 0 System
>SIGWAITINFO
> 2 2 ot/devf-dbpxa270dp 10r 0 System
>RECEIVE
> 2 3 ot/devf-dbpxa270dp 30r 0 System
>RECEIVE
> 2 4 ot/devf-dbpxa270dp 10r 0 System
>RECEIVE
> 2 5 ot/devf-dbpxa270dp 10r 0 System
>RECEIVE
> 3 1 oot/devc-serpxa250 24r 0 System
>RECEIVE
> 4 1 proc/boot/slogger 30r 0 System
>RECEIVE
> 5 1 proc/boot/pipe 10r 0 System
>SIGWAITINFO
> 5 2 proc/boot/pipe 10r 0 System
>RECEIVE
> 5 3 proc/boot/pipe 10r 0 System
>RECEIVE
> 5 4 proc/boot/pipe 10r 0 System
>RECEIVE
> 5 5 proc/boot/pipe 10r 0 System
>RECEIVE
> 8199 1 sbin/inflator 10r 0 System
>RECEIVE
> 8199 2 sbin/inflator 10r 0 System
>RECEIVE
> 8199 3 sbin/inflator 10r 0 System
>RECEIVE
> 8199 4 sbin/inflator 10r 0 System
>RECEIVE
> 53256 1 sbin/io-net 10r 0 DevDebug
>SIGWAITINFO
> 53256 2 sbin/io-net 21r 0 DevDebug
>RECEIVE
> 53256 3 sbin/io-net 10r 0 DevDebug
>RECEIVE
> 53256 4 sbin/io-net 10r 0 DevDebug
>RECEIVE
> 53256 5 sbin/io-net 21r 0 DevDebug
>RECEIVE
>...
View Full Message
|
|
|
|
|
|
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on
PXA270
|
|
11/17/2008 12:35 PM
post16684
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on
PXA270
Thanks for the analysis, HP.
Ken, you could check this is the case by turning kernel call premption off (-p switch to procnto)
Hans-Peter Reichert wrote:
> this is HOT!!!
> your pidin request is about to start
> within this it is doing its normal statup stuff
> ending in a request to proc thread 2 which itself is doing a fseek64() that is serviced by proc thread 10
> now thread 10 is trying to answer using MsgReply() but it cant't
> whenever proc thread 10 is trying to do MsgReply() it gets interrupted by IR 0x10 which is handled by LMDispatcher in
partition 0
> due to the fact that you have a real high IR rate of 3000Hz the MsgReply() can't finish.
>
> Hey QNX guys wake up this is somethig to work on.
> @Colin: do you remember that beta hotfix you've done for the MMI3G project where we had a quite simillar problem?
>
> /hp
>
>
>> -----Ursprüngliche Nachricht-----
>> Von: Ken Schumm [mailto:community-noreply@qnx.com]
>> Gesendet: Montag, 17. November 2008 16:39
>> An: ostech-core_os
>> Betreff: Re: AW: AW: RE: RE: New to Adaptive Partitioning,
>> problems on PXA270
>>
>> I ran another test today and captured a wide event trace log
>> (attached).
>>
>> A pidin sched prior to the test shows this:
>> ---------------------------------------------------------------
>> ---------------------------------------
>> pid tid name prio cpu ExtSched
>> STATE
>> 1 1 procnto 0f 0 System
>> READY
>> 1 2 procnto 10r 0 System
>> RECEIVE
>> 1 3 procnto 10r 0 System
>> RECEIVE
>> 1 4 procnto 10r 0 System
>> RECEIVE
>> 1 5 procnto 10r 0 System
>> RECEIVE
>> 1 6 procnto 10r 0 System
>> RECEIVE
>> 1 7 procnto 10r 0 System
>> RECEIVE
>> 1 8 procnto 10r 0 System
>> RECEIVE
>> 1 9 procnto 10r 0 System
>> RECEIVE
>> 1 10 procnto 10r 0 DevDebug
>> RUNNING
>> 1 11 procnto 10r 0 System
>> RECEIVE
>> 2 1 ot/devf-dbpxa270dp 10r 0 System
>> SIGWAITINFO
>> 2 2 ot/devf-dbpxa270dp 10r 0 System
>> RECEIVE
>> 2 3 ot/devf-dbpxa270dp 30r 0 System
>> RECEIVE
>> 2 4 ot/devf-dbpxa270dp 10r 0 System
>> RECEIVE
>> 2 5 ot/devf-dbpxa270dp 10r 0 System
>> RECEIVE
>> 3 1 oot/devc-serpxa250 24r 0 System
>> RECEIVE
>> 4 1 proc/boot/slogger 30r 0 System
>> RECEIVE
>> 5 1 proc/boot/pipe 10r 0 System
>> SIGWAITINFO
>> 5 2 proc/boot/pipe 10r 0 System
>> RECEIVE
>> 5 3 proc/boot/pipe 10r 0 System
>> RECEIVE
>> 5 4 proc/boot/pipe 10r 0 System
>> RECEIVE
>> 5 5 proc/boot/pipe 10r 0 System
>> RECEIVE
>> 8199 1 sbin/inflator 10r 0 System
>> RECEIVE
>> 8199 2 sbin/inflator 10r 0 System
>> RECEIVE
>> 8199 3 sbin/inflator 10r 0 System ...
View Full Message
|
|
|
|
|
|
|
AW: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/17/2008 1:10 PM
post16686
|
AW: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
I've red the 6.4 release notes but I didn't recognize that -P (capital)
option to procnto - COOOL
/hp
-----Ursprüngliche Nachricht-----
Von: Colin Burgess [mailto:community-noreply@qnx.com]
Gesendet: Mo 17.11.2008 18:35
An: ostech-core_os
Cc:
Betreff: Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
Thanks for the analysis, HP.
Ken, you could check this is the case by turning kernel call premption off (-p switch to procnto)
Hans-Peter Reichert wrote:
> this is HOT!!!
> your pidin request is about to start
> within this it is doing its normal statup stuff
> ending in a request to proc thread 2 which itself is doing a fseek64() that is serviced by proc thread 10
> now thread 10 is trying to answer using MsgReply() but it cant't
> whenever proc thread 10 is trying to do MsgReply() it gets interrupted by IR 0x10 which is handled by LMDispatcher in
partition 0
> due to the fact that you have a real high IR rate of 3000Hz the MsgReply() can't finish.
>
> Hey QNX guys wake up this is somethig to work on.
> @Colin: do you remember that beta hotfix you've done for the MMI3G project where we had a quite simillar problem?
>
> /hp
>
>
>> -----Ursprüngliche Nachricht-----
>> Von: Ken Schumm [mailto:community-noreply@qnx.com]
>> Gesendet: Montag, 17. November 2008 16:39
>> An: ostech-core_os
>> Betreff: Re: AW: AW: RE: RE: New to Adaptive Partitioning,
>> problems on PXA270
>>
>> I ran another test today and captured a wide event trace log
>> (attached).
>>
>> A pidin sched prior to the test shows this:
>> ---------------------------------------------------------------
>> ---------------------------------------
>> pid tid name prio cpu ExtSched
>> STATE
>> 1 1 procnto 0f 0 System
>> READY
>> 1 2 procnto 10r 0 System
>> RECEIVE
>> 1 3 procnto 10r 0 System
>> RECEIVE
>> 1 4 procnto 10r 0 System
>> RECEIVE
>> 1 5 procnto 10r 0 System
>> RECEIVE
>> 1 6 procnto 10r 0 System
>> RECEIVE
>> 1 7 procnto 10r 0 System
>> RECEIVE
>> 1 8 procnto 10r 0 System
>> RECEIVE
>> 1 9 procnto 10r 0 System
>> RECEIVE
>> 1 10 procnto 10r 0 DevDebug
>> RUNNING
>> 1 11 procnto 10r 0 System
>> RECEIVE
>> 2 1 ot/devf-dbpxa270dp 10r 0 System
>> SIGWAITINFO
>> 2 2 ot/devf-dbpxa270dp 10r 0 System
>> RECEIVE
>> 2 3 ot/devf-dbpxa270dp 30r 0 System
>> RECEIVE
>> 2 4 ot/devf-dbpxa270dp 10r 0 System
>> RECEIVE
>> 2 5 ot/devf-dbpxa270dp 10r 0 System
>> RECEIVE
>> 3 1 oot/devc-serpxa250 24r 0 System
>> RECEIVE
>> 4 1 proc/boot/slogger 30r 0 System
>> RECEIVE
>> 5 1 proc/boot/pipe 10r 0 System
>> SIGWAITINFO
>> 5 2 proc/boot/pipe 10r 0 System
>> RECEIVE
>> 5 3 proc/boot/pipe 10r 0 System
>> RECEIVE
>> 5 4 proc/boot/pipe 10r 0 System
>> RECEIVE
>> ...
View Full Message
|
|
|
|
|
|
|
Re: AW: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems
on PXA270
|
|
11/17/2008 1:15 PM
post16687
|
Re: AW: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems
on PXA270
You mean the Priv prio switch?
Hans-Peter Reichert wrote:
>
> I've red the 6.4 release notes but I didn't recognize that -P (capital)
> option to procnto - COOOL
> /hp
>
> -----Ursprüngliche Nachricht-----
> Von: Colin Burgess [mailto:community-noreply@qnx.com]
> Gesendet: Mo 17.11.2008 18:35
> An: ostech-core_os
> Cc:
> Betreff: Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning,
> problems on PXA270
>
> Thanks for the analysis, HP.
>
> Ken, you could check this is the case by turning kernel call premption
> off (-p switch to procnto)
>
> Hans-Peter Reichert wrote:
> > this is HOT!!!
> > your pidin request is about to start
> > within this it is doing its normal statup stuff
> > ending in a request to proc thread 2 which itself is doing a
> fseek64() that is serviced by proc thread 10
> > now thread 10 is trying to answer using MsgReply() but it cant't
> > whenever proc thread 10 is trying to do MsgReply() it gets
> interrupted by IR 0x10 which is handled by LMDispatcher in partition 0
> > due to the fact that you have a real high IR rate of 3000Hz the
> MsgReply() can't finish.
> >
> > Hey QNX guys wake up this is somethig to work on.
> > @Colin: do you remember that beta hotfix you've done for the MMI3G
> project where we had a quite simillar problem?
> >
> > /hp
> >
> >
> >> -----Ursprüngliche Nachricht-----
> >> Von: Ken Schumm [mailto:community-noreply@qnx.com]
> >> Gesendet: Montag, 17. November 2008 16:39
> >> An: ostech-core_os
> >> Betreff: Re: AW: AW: RE: RE: New to Adaptive Partitioning,
> >> problems on PXA270
> >>
> >> I ran another test today and captured a wide event trace log
> >> (attached).
> >>
> >> A pidin sched prior to the test shows this:
> >> ---------------------------------------------------------------
> >> ---------------------------------------
> >> pid tid name prio cpu ExtSched
> >> STATE
> >> 1 1 procnto 0f 0 System
> >> READY
> >> 1 2 procnto 10r 0 System
> >> RECEIVE
> >> 1 3 procnto 10r 0 System
> >> RECEIVE
> >> 1 4 procnto 10r 0 System
> >> RECEIVE
> >> 1 5 procnto 10r 0 System
> >> RECEIVE
> >> 1 6 procnto 10r 0 System
> >> RECEIVE
> >> 1 7 procnto 10r 0 System
> >> RECEIVE
> >> 1 8 procnto 10r 0 System
> >> RECEIVE
> >> 1 9 procnto 10r 0 System
> >> RECEIVE
> >> 1 10 procnto 10r 0 DevDebug
> >> RUNNING
> >> 1 11 procnto 10r 0 System
> >> RECEIVE
> >> 2 1 ot/devf-dbpxa270dp 10r 0 System
> >> SIGWAITINFO
> >> 2 2 ot/devf-dbpxa270dp 10r 0 System
> >> RECEIVE
> >> 2 3 ot/devf-dbpxa270dp 30r 0 System
> >> RECEIVE
> >> 2 4 ot/devf-dbpxa270dp 10r 0 System
> >> RECEIVE
> >> 2 5 ot/devf-dbpxa270dp 10r 0 System
> >> RECEIVE
> >> 3 1...
View Full Message
|
|
|
|
|
|
|
AW: AW: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/17/2008 1:17 PM
post16688
|
AW: AW: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
yepp, but I've jusst seen that this not new.... hmm
grübel grübel
-----Ursprüngliche Nachricht-----
Von: Colin Burgess [mailto:community-noreply@qnx.com]
Gesendet: Mo 17.11.2008 19:15
An: ostech-core_os
Cc:
Betreff: Re: AW: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
You mean the Priv prio switch?
Hans-Peter Reichert wrote:
>
> I've red the 6.4 release notes but I didn't recognize that -P (capital)
> option to procnto - COOOL
> /hp
>
> -----Ursprüngliche Nachricht-----
> Von: Colin Burgess [mailto:community-noreply@qnx.com]
> Gesendet: Mo 17.11.2008 18:35
> An: ostech-core_os
> Cc:
> Betreff: Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning,
> problems on PXA270
>
> Thanks for the analysis, HP.
>
> Ken, you could check this is the case by turning kernel call premption
> off (-p switch to procnto)
>
> Hans-Peter Reichert wrote:
> > this is HOT!!!
> > your pidin request is about to start
> > within this it is doing its normal statup stuff
> > ending in a request to proc thread 2 which itself is doing a
> fseek64() that is serviced by proc thread 10
> > now thread 10 is trying to answer using MsgReply() but it cant't
> > whenever proc thread 10 is trying to do MsgReply() it gets
> interrupted by IR 0x10 which is handled by LMDispatcher in partition 0
> > due to the fact that you have a real high IR rate of 3000Hz the
> MsgReply() can't finish.
> >
> > Hey QNX guys wake up this is somethig to work on.
> > @Colin: do you remember that beta hotfix you've done for the MMI3G
> project where we had a quite simillar problem?
> >
> > /hp
> >
> >
> >> -----Ursprüngliche Nachricht-----
> >> Von: Ken Schumm [mailto:community-noreply@qnx.com]
> >> Gesendet: Montag, 17. November 2008 16:39
> >> An: ostech-core_os
> >> Betreff: Re: AW: AW: RE: RE: New to Adaptive Partitioning,
> >> problems on PXA270
> >>
> >> I ran another test today and captured a wide event trace log
> >> (attached).
> >>
> >> A pidin sched prior to the test shows this:
> >> ---------------------------------------------------------------
> >> ---------------------------------------
> >> pid tid name prio cpu ExtSched
> >> STATE
> >> 1 1 procnto 0f 0 System
> >> READY
> >> 1 2 procnto 10r 0 System
> >> RECEIVE
> >> 1 3 procnto 10r 0 System
> >> RECEIVE
> >> 1 4 procnto 10r 0 System
> >> RECEIVE
> >> 1 5 procnto 10r 0 System
> >> RECEIVE
> >> 1 6 procnto 10r 0 System
> >> RECEIVE
> >> 1 7 procnto 10r 0 System
> >> RECEIVE
> >> 1 8 procnto 10r 0 System
> >> RECEIVE
> >> 1 9 procnto 10r 0 System
> >> RECEIVE
> >> 1 10 procnto 10r 0 DevDebug
> >> RUNNING
> >> 1 11 procnto 10r 0 System
> >> RECEIVE
> >> 2 1 ot/devf-dbpxa270dp 10r 0 System
> >> SIGWAITINFO
> >> 2 2 ot/devf-dbpxa270dp 10r 0 System
> >> RECEIVE
> >> 2 3...
View Full Message
|
|
|
|
|
|
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
11/17/2008 1:37 PM
post16689
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
You're onto something, with the -p option to procnto it worked.
I've attached another event log for the -p test if you're interested.
PS - I never noticed that the procnto options don't appear in 'pidin arg' before, that had me confused for awhile (easy
to do).
> Thanks for the analysis, HP.
>
> Ken, you could check this is the case by turning kernel call premption off (-p
> switch to procnto)
[...]
|
|
|
|
|
|
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
12/01/2008 1:42 PM
post17668
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
> You're onto something, with the -p option to procnto it worked.
>
> I've attached another event log for the -p test if you're interested.
>
> PS - I never noticed that the procnto options don't appear in 'pidin arg'
> before, that had me confused for awhile (easy to do).
>
> > Thanks for the analysis, HP.
> >
> > Ken, you could check this is the case by turning kernel call premption off (
> -p
> > switch to procnto)
>
> [...]
I haven't heard anything else on this and was wondering - do you consider disabling kernel preemption to be an official
solution in this case? Or are you folks thinking about a fix?
Please note that this is not a showstopper for us. I was evaluating APS for possible use in current and future products
but it is not currently in a shipping product.
|
|
|
|
|
|
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on
PXA270
|
|
12/01/2008 1:56 PM
post17670
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on
PXA270
Hi Ken,
We do consider it a problem, but it's not a simple one to fix. However we are considering solutions
to this issue.
I'll try and keep you updated as we make progress.
Colin
Ken Schumm wrote:
>> You're onto something, with the -p option to procnto it worked.
>>
>> I've attached another event log for the -p test if you're interested.
>>
>> PS - I never noticed that the procnto options don't appear in 'pidin arg'
>> before, that had me confused for awhile (easy to do).
>>
>>> Thanks for the analysis, HP.
>>>
>>> Ken, you could check this is the case by turning kernel call premption off (
>> -p
>>> switch to procnto)
>> [...]
>
> I haven't heard anything else on this and was wondering - do you consider disabling kernel preemption to be an
official solution in this case? Or are you folks thinking about a fix?
>
> Please note that this is not a showstopper for us. I was evaluating APS for possible use in current and future
products but it is not currently in a shipping product.
>
>
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post17668
>
--
cburgess@qnx.com
|
|
|
|
|
|
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
12/01/2008 2:12 PM
post17671
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
> Hi Ken,
>
> We do consider it a problem, but it's not a simple one to fix. However we are
> considering solutions
> to this issue.
>
> I'll try and keep you updated as we make progress.
>
> Colin
Thanks Colin, I appreciate it.
It doesn't sound at all like an easy fix, so good luck!
|
|
|
|
|
|
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
|
|
12/03/2008 2:15 AM
post17816
|
Re: AW: AW: AW: RE: RE: New to Adaptive Partitioning, problems on PXA270
strange my replies through do not make it, this is the second ....
remember that this is not only an APS issue, it is independant from APS just remember the MMI3G project.
I would say that the way you've done it the first time (assuming I remember this right) was not to bad - try to finish
the k-call n times and then force the completion.
What about your scheduler gurus - do they hav any further ideas?
/hp
|
|
|
|
|
|