I am working on a slow suspend problem on my pda running WM5.0 on Intel
PXA270.
I noticed that the function call in the power manager
FileSystemPowerFunction(FSNOTIFY_POWER_OFF);
is taking very long to return.

Does anyone has experience with this?

Thanks.

Re: File System by voidcoder

voidcoder
Fri Feb 02 02:13:47 CST 2007


Perhaps it is flushing registry. Is there any hard
activity in your flash driver during the call?


Krypton wrote:
> I am working on a slow suspend problem on my pda running WM5.0 on Intel
> PXA270.
> I noticed that the function call in the power manager
> FileSystemPowerFunction(FSNOTIFY_POWER_OFF);
> is taking very long to return.
>
> Does anyone has experience with this?
>
> Thanks.

Re: File System by Krypton

Krypton
Sun Feb 04 18:51:01 CST 2007

I am sorry I am not too sure what you meant by hard activity.
I am not clear abt which function calls will follow FileSystemPowerFunction.
This seems to be in coredll and I do not know eventually what flash driver
calls
are made. Do you know where I can get more information on this?

Thanks.

"voidcoder" wrote:

>
> Perhaps it is flushing registry. Is there any hard
> activity in your flash driver during the call?
>
>
> Krypton wrote:
> > I am working on a slow suspend problem on my pda running WM5.0 on Intel
> > PXA270.
> > I noticed that the function call in the power manager
> > FileSystemPowerFunction(FSNOTIFY_POWER_OFF);
> > is taking very long to return.
> >
> > Does anyone has experience with this?
> >
> > Thanks.
>

Re: File System by voidcoder

voidcoder
Mon Feb 05 02:17:43 CST 2007


Normally you can track such a situation using kernel profiler,
this may shed some light on where exactly it spends the time.
However suspension is quite specific situation, not sure
profiling may help you here.

To the flash driver. I think flushing registry changes to the
flash while suspending is one of the possible reasons for the
slowness. You can try to track WriteSector() and EraseSector()
callbacks in your flash driver. Just insert some debug output
to see how often they are called and how long does it take to
program/erase flash while suspending.


Krypton wrote:
> I am sorry I am not too sure what you meant by hard activity.
> I am not clear abt which function calls will follow FileSystemPowerFunction.
> This seems to be in coredll and I do not know eventually what flash driver
> calls
> are made. Do you know where I can get more information on this?
>
> Thanks.
>
> "voidcoder" wrote:
>
>> Perhaps it is flushing registry. Is there any hard
>> activity in your flash driver during the call?
>>
>>
>> Krypton wrote:
>>> I am working on a slow suspend problem on my pda running WM5.0 on Intel
>>> PXA270.
>>> I noticed that the function call in the power manager
>>> FileSystemPowerFunction(FSNOTIFY_POWER_OFF);
>>> is taking very long to return.
>>>
>>> Does anyone has experience with this?
>>>
>>> Thanks.

Re: File System by Krypton

Krypton
Tue Feb 06 00:49:00 CST 2007

Hi,

I found that the time is spent doing read, write and flush to flash.
I am using P30 with Intel IPSM.
Is this an inherent problem, the slowness of write and flush to flash
for this combination?

Thanks.

"voidcoder" wrote:

>
> Normally you can track such a situation using kernel profiler,
> this may shed some light on where exactly it spends the time.
> However suspension is quite specific situation, not sure
> profiling may help you here.
>
> To the flash driver. I think flushing registry changes to the
> flash while suspending is one of the possible reasons for the
> slowness. You can try to track WriteSector() and EraseSector()
> callbacks in your flash driver. Just insert some debug output
> to see how often they are called and how long does it take to
> program/erase flash while suspending.
>
>
> Krypton wrote:
> > I am sorry I am not too sure what you meant by hard activity.
> > I am not clear abt which function calls will follow FileSystemPowerFunction.
> > This seems to be in coredll and I do not know eventually what flash driver
> > calls
> > are made. Do you know where I can get more information on this?
> >
> > Thanks.
> >
> > "voidcoder" wrote:
> >
> >> Perhaps it is flushing registry. Is there any hard
> >> activity in your flash driver during the call?
> >>
> >>
> >> Krypton wrote:
> >>> I am working on a slow suspend problem on my pda running WM5.0 on Intel
> >>> PXA270.
> >>> I noticed that the function call in the power manager
> >>> FileSystemPowerFunction(FSNOTIFY_POWER_OFF);
> >>> is taking very long to return.
> >>>
> >>> Does anyone has experience with this?
> >>>
> >>> Thanks.
>

Re: File System by voidcoder

voidcoder
Tue Feb 06 02:55:43 CST 2007


P30 is too slow itself. For example, depending on the block
size / parameter, block erase time may vary from 0.4 up to 4.0
seconds ...


Krypton wrote:
> Hi,
>
> I found that the time is spent doing read, write and flush to flash.
> I am using P30 with Intel IPSM.
> Is this an inherent problem, the slowness of write and flush to flash
> for this combination?
>
> Thanks.
>
> "voidcoder" wrote:
>
>> Normally you can track such a situation using kernel profiler,
>> this may shed some light on where exactly it spends the time.
>> However suspension is quite specific situation, not sure
>> profiling may help you here.
>>
>> To the flash driver. I think flushing registry changes to the
>> flash while suspending is one of the possible reasons for the
>> slowness. You can try to track WriteSector() and EraseSector()
>> callbacks in your flash driver. Just insert some debug output
>> to see how often they are called and how long does it take to
>> program/erase flash while suspending.
>>
>>
>> Krypton wrote:
>>> I am sorry I am not too sure what you meant by hard activity.
>>> I am not clear abt which function calls will follow FileSystemPowerFunction.
>>> This seems to be in coredll and I do not know eventually what flash driver
>>> calls
>>> are made. Do you know where I can get more information on this?
>>>
>>> Thanks.
>>>
>>> "voidcoder" wrote:
>>>
>>>> Perhaps it is flushing registry. Is there any hard
>>>> activity in your flash driver during the call?
>>>>
>>>>
>>>> Krypton wrote:
>>>>> I am working on a slow suspend problem on my pda running WM5.0 on Intel
>>>>> PXA270.
>>>>> I noticed that the function call in the power manager
>>>>> FileSystemPowerFunction(FSNOTIFY_POWER_OFF);
>>>>> is taking very long to return.
>>>>>
>>>>> Does anyone has experience with this?
>>>>>
>>>>> Thanks.

Re: File System by Krypton

Krypton
Tue Feb 06 05:26:00 CST 2007

Thank you for the info.

Is there any way to ask the PM not to flush registry upon suspend?

I removed the call by the PM (by commenting it out) and things still
seems to work. Hopefully the OS does flush frequently enough so that
only some registry changes are not registered in flash?

"voidcoder" wrote:

>
> P30 is too slow itself. For example, depending on the block
> size / parameter, block erase time may vary from 0.4 up to 4.0
> seconds ...
>
>
> Krypton wrote:
> > Hi,
> >
> > I found that the time is spent doing read, write and flush to flash.
> > I am using P30 with Intel IPSM.
> > Is this an inherent problem, the slowness of write and flush to flash
> > for this combination?
> >
> > Thanks.
> >
> > "voidcoder" wrote:
> >
> >> Normally you can track such a situation using kernel profiler,
> >> this may shed some light on where exactly it spends the time.
> >> However suspension is quite specific situation, not sure
> >> profiling may help you here.
> >>
> >> To the flash driver. I think flushing registry changes to the
> >> flash while suspending is one of the possible reasons for the
> >> slowness. You can try to track WriteSector() and EraseSector()
> >> callbacks in your flash driver. Just insert some debug output
> >> to see how often they are called and how long does it take to
> >> program/erase flash while suspending.
> >>
> >>
> >> Krypton wrote:
> >>> I am sorry I am not too sure what you meant by hard activity.
> >>> I am not clear abt which function calls will follow FileSystemPowerFunction.
> >>> This seems to be in coredll and I do not know eventually what flash driver
> >>> calls
> >>> are made. Do you know where I can get more information on this?
> >>>
> >>> Thanks.
> >>>
> >>> "voidcoder" wrote:
> >>>
> >>>> Perhaps it is flushing registry. Is there any hard
> >>>> activity in your flash driver during the call?
> >>>>
> >>>>
> >>>> Krypton wrote:
> >>>>> I am working on a slow suspend problem on my pda running WM5.0 on Intel
> >>>>> PXA270.
> >>>>> I noticed that the function call in the power manager
> >>>>> FileSystemPowerFunction(FSNOTIFY_POWER_OFF);
> >>>>> is taking very long to return.
> >>>>>
> >>>>> Does anyone has experience with this?
> >>>>>
> >>>>> Thanks.
>

Re: File System by voidcoder

voidcoder
Tue Feb 06 06:17:02 CST 2007


Interesting, I thought IPSM is no longer used on
PPC/WM devices. Here you can find some advanced
registry flushing options:

http://msdn2.microsoft.com/en-us/library/ms885238.aspx

Actually it doesn't matter what method you will
use. The problem will popup anyway as far as your
flash is slow. If not during the power down then
at some point during the normal operation.

Btw what are you doing exactly? Are you developing
a WM design or just killing your own WM5 PDA?
In the first case you can get much better answer
through your WM support channel. I'm sure you don't
have enough freedom choosing methods and components
when developing WM5 design ...



Krypton wrote:
> Thank you for the info.
>
> Is there any way to ask the PM not to flush registry upon suspend?
>
> I removed the call by the PM (by commenting it out) and things still
> seems to work. Hopefully the OS does flush frequently enough so that
> only some registry changes are not registered in flash?
>
> "voidcoder" wrote:
>
>> P30 is too slow itself. For example, depending on the block
>> size / parameter, block erase time may vary from 0.4 up to 4.0
>> seconds ...
>>
>>
>> Krypton wrote:
>>> Hi,
>>>
>>> I found that the time is spent doing read, write and flush to flash.
>>> I am using P30 with Intel IPSM.
>>> Is this an inherent problem, the slowness of write and flush to flash
>>> for this combination?
>>>
>>> Thanks.
>>>
>>> "voidcoder" wrote:
>>>
>>>> Normally you can track such a situation using kernel profiler,
>>>> this may shed some light on where exactly it spends the time.
>>>> However suspension is quite specific situation, not sure
>>>> profiling may help you here.
>>>>
>>>> To the flash driver. I think flushing registry changes to the
>>>> flash while suspending is one of the possible reasons for the
>>>> slowness. You can try to track WriteSector() and EraseSector()
>>>> callbacks in your flash driver. Just insert some debug output
>>>> to see how often they are called and how long does it take to
>>>> program/erase flash while suspending.
>>>>
>>>>
>>>> Krypton wrote:
>>>>> I am sorry I am not too sure what you meant by hard activity.
>>>>> I am not clear abt which function calls will follow FileSystemPowerFunction.
>>>>> This seems to be in coredll and I do not know eventually what flash driver
>>>>> calls
>>>>> are made. Do you know where I can get more information on this?
>>>>>
>>>>> Thanks.
>>>>>
>>>>> "voidcoder" wrote:
>>>>>
>>>>>> Perhaps it is flushing registry. Is there any hard
>>>>>> activity in your flash driver during the call?
>>>>>>
>>>>>>
>>>>>> Krypton wrote:
>>>>>>> I am working on a slow suspend problem on my pda running WM5.0 on Intel
>>>>>>> PXA270.
>>>>>>> I noticed that the function call in the power manager
>>>>>>> FileSystemPowerFunction(FSNOTIFY_POWER_OFF);
>>>>>>> is taking very long to return.
>>>>>>>
>>>>>>> Does anyone has experience with this?
>>>>>>>
>>>>>>> Thanks.

Re: File System by Krypton

Krypton
Tue Feb 06 18:51:00 CST 2007

Thank you once again for your reply.
I am actually developing, not trying to kill my own PDA ha.
Will look up the link you provided.

"voidcoder" wrote:

>
> Interesting, I thought IPSM is no longer used on
> PPC/WM devices. Here you can find some advanced
> registry flushing options:
>
> http://msdn2.microsoft.com/en-us/library/ms885238.aspx
>
> Actually it doesn't matter what method you will
> use. The problem will popup anyway as far as your
> flash is slow. If not during the power down then
> at some point during the normal operation.
>
> Btw what are you doing exactly? Are you developing
> a WM design or just killing your own WM5 PDA?
> In the first case you can get much better answer
> through your WM support channel. I'm sure you don't
> have enough freedom choosing methods and components
> when developing WM5 design ...
>
>
>
> Krypton wrote:
> > Thank you for the info.
> >
> > Is there any way to ask the PM not to flush registry upon suspend?
> >
> > I removed the call by the PM (by commenting it out) and things still
> > seems to work. Hopefully the OS does flush frequently enough so that
> > only some registry changes are not registered in flash?
> >
> > "voidcoder" wrote:
> >
> >> P30 is too slow itself. For example, depending on the block
> >> size / parameter, block erase time may vary from 0.4 up to 4.0
> >> seconds ...
> >>
> >>
> >> Krypton wrote:
> >>> Hi,
> >>>
> >>> I found that the time is spent doing read, write and flush to flash.
> >>> I am using P30 with Intel IPSM.
> >>> Is this an inherent problem, the slowness of write and flush to flash
> >>> for this combination?
> >>>
> >>> Thanks.
> >>>
> >>> "voidcoder" wrote:
> >>>
> >>>> Normally you can track such a situation using kernel profiler,
> >>>> this may shed some light on where exactly it spends the time.
> >>>> However suspension is quite specific situation, not sure
> >>>> profiling may help you here.
> >>>>
> >>>> To the flash driver. I think flushing registry changes to the
> >>>> flash while suspending is one of the possible reasons for the
> >>>> slowness. You can try to track WriteSector() and EraseSector()
> >>>> callbacks in your flash driver. Just insert some debug output
> >>>> to see how often they are called and how long does it take to
> >>>> program/erase flash while suspending.
> >>>>
> >>>>
> >>>> Krypton wrote:
> >>>>> I am sorry I am not too sure what you meant by hard activity.
> >>>>> I am not clear abt which function calls will follow FileSystemPowerFunction.
> >>>>> This seems to be in coredll and I do not know eventually what flash driver
> >>>>> calls
> >>>>> are made. Do you know where I can get more information on this?
> >>>>>
> >>>>> Thanks.
> >>>>>
> >>>>> "voidcoder" wrote:
> >>>>>
> >>>>>> Perhaps it is flushing registry. Is there any hard
> >>>>>> activity in your flash driver during the call?
> >>>>>>
> >>>>>>
> >>>>>> Krypton wrote:
> >>>>>>> I am working on a slow suspend problem on my pda running WM5.0 on Intel
> >>>>>>> PXA270.
> >>>>>>> I noticed that the function call in the power manager
> >>>>>>> FileSystemPowerFunction(FSNOTIFY_POWER_OFF);
> >>>>>>> is taking very long to return.
> >>>>>>>
> >>>>>>> Does anyone has experience with this?
> >>>>>>>
> >>>>>>> Thanks.
>