Debugger OS-awareness

Debugger OS-awareness

[ulfh]

Dear maintainers

iSystem is a company supplying emulators to the embedded community.
Presently we support 600+ MCUs with our line of three emulators.

We are now interested in adding OS awareness to our universal development
software winIDEA, for eCos, and need some basic input on the following
points:

- getting us a good contact with the OS vendor, not some salesman 
- getting the specs on how to reach kernel object info from freeze mode 
- getting a sample application running on one of the common EVBs that are
availlable to iSystem.

I am shure that the first point is taken care of with this mail, but the
rest... Please direct any information on the subjects to
janez.ulcakar@isystem.com. 

Best regards

Ulf Hegelund
iSYSTEM AB
Slottagårdsgatan 9
S-235 35  Vellinge
Phone: +46 40 45 95 71
Fax: +46 40 45 93 38
ulfh@ese.se
www.isystem.se

Re: Debugger OS-awareness

[Jonathan Larmour]

ulfh@ese.se wrote:
> Dear maintainers
> 
> iSystem is a company supplying emulators to the embedded community.
> Presently we support 600+ MCUs with our line of three emulators.
> 
> We are now interested in adding OS awareness to our universal development
> software winIDEA, for eCos, and need some basic input on the following
> points:
> 
> - getting us a good contact with the OS vendor, not some salesman 

Well, eCos doesn't have any single vendor as such. 
http://sources.redhat.com/ecos/ is the master site and the maintainers 
(from a number of different companies) make (very!) occasional public 
releases.

There are a number of companies that do do support/custom development, but 
if it's not involving a sale, I'm not sure what you'd be wanting to 
achieve... perhaps that does mean these lists are the best place to talk 
about what you want.

> - getting the specs on how to reach kernel object info from freeze mode 

freeze mode? We don't have much in the way of targets supporting power 
management at all (although the support exists) so I'm not clear what 
you're getting at.

> - getting a sample application running on one of the common EVBs that are
> availlable to iSystem.

You'll have to tell us then :-).

Jifl
-- 
eCosCentric    http://www.eCosCentric.com/    The eCos and RedBoot experts
--[ "You can complain because roses have thorns, or you ]--
--[  can rejoice because thorns have roses." -Lincoln   ]-- Opinions==mine

Re: Debugger OS-awareness

[Andrew Lunn]

> >- getting the specs on how to reach kernel object info from freeze mode 
> 
> freeze mode? We don't have much in the way of targets supporting power 
> management at all (although the support exists) so I'm not clear what 
> you're getting at.

This is running on an emulator. So i guess he means the emulator has
stopped executing instructions. He then wants to know how to poke
around the emulated memory to find out about tasks.

He basically needs to implement in the emulator the functions Nick
added a few weeks ago to find out about current threads etc.

      Andrew

Re: Debugger OS-awareness

[Jonathan Larmour]

Andrew Lunn wrote:
>>>- getting the specs on how to reach kernel object info from freeze mode 
>>
>>freeze mode? We don't have much in the way of targets supporting power 
>>management at all (although the support exists) so I'm not clear what 
>>you're getting at.
> 
> 
> This is running on an emulator. So i guess he means the emulator has
> stopped executing instructions. He then wants to know how to poke
> around the emulated memory to find out about tasks.
> 
> He basically needs to implement in the emulator the functions Nick
> added a few weeks ago to find out about current threads etc.

True, but that can be problematic with a configurable operating system 
where the thread structure layout can change!

There have been two approaches to this in the past. The first one, which 
has been preferred in the two examples (ICE support) so far that needed 
this support, was to save the target context in the ICE, set the PC to a 
special function pointed to from a well defined location, start the target 
going again and that function will gather the data required for the 
debugger, and then stop the target again when the function returns.

The second approach, used by some simulators, is to have good enough 
support in the simulator and/or the port to run a ROM monitor, then 
communicate with it using the GDB remote protocol! This is probably 
slightly easier but has more overhead.

Jifl
-- 
eCosCentric    http://www.eCosCentric.com/    The eCos and RedBoot experts
--[ "You can complain because roses have thorns, or you ]--
--[  can rejoice because thorns have roses." -Lincoln   ]-- Opinions==mine

RE: Debugger OS-awareness

[Janez Ulcakar]

Hi,

You're correct in your assumptions. We've encountered that 'special function
spawned by the ICE' approach before, and I think it's better suited for an
ICE.

a) Does eCos already provide such a function (how's it called, where does it
take it's params from, and where do we read the results from and what
format)

or

b) should a debugger (we) provide a C source that the user includes in his
build

regards
janez

-----Original Message-----
From: Jonathan Larmour [mailto:jifl@eCosCentric.com]
Sent: Friday, February 28, 2003 16:51
To: Andrew Lunn
Cc: janez.ulcakar@isystem.com; ulfh@ese.se;
ecos-maintainers@sources.redhat.com
Subject: Re: Debugger OS-awareness


Andrew Lunn wrote:
>>>- getting the specs on how to reach kernel object info from freeze mode
>>
>>freeze mode? We don't have much in the way of targets supporting power
>>management at all (although the support exists) so I'm not clear what
>>you're getting at.
>
>
> This is running on an emulator. So i guess he means the emulator has
> stopped executing instructions. He then wants to know how to poke
> around the emulated memory to find out about tasks.
>
> He basically needs to implement in the emulator the functions Nick
> added a few weeks ago to find out about current threads etc.

True, but that can be problematic with a configurable operating system
where the thread structure layout can change!

There have been two approaches to this in the past. The first one, which
has been preferred in the two examples (ICE support) so far that needed
this support, was to save the target context in the ICE, set the PC to a
special function pointed to from a well defined location, start the target
going again and that function will gather the data required for the
debugger, and then stop the target again when the function returns.

The second approach, used by some simulators, is to have good enough
support in the simulator and/or the port to run a ROM monitor, then
communicate with it using the GDB remote protocol! This is probably
slightly easier but has more overhead.

Jifl
--
eCosCentric    http://www.eCosCentric.com/    The eCos and RedBoot experts
--[ "You can complain because roses have thorns, or you ]--
--[  can rejoice because thorns have roses." -Lincoln   ]-- Opinions==mine

Re: Debugger OS-awareness

[Jonathan Larmour]

Janez Ulcakar wrote:
> Hi,
> 
> You're correct in your assumptions. We've encountered that 'special function
> spawned by the ICE' approach before, and I think it's better suited for an
> ICE.
> 
> a) Does eCos already provide such a function (how's it called, where does it
> take it's params from, and where do we read the results from and what
> format)

It wasn't made generic but there's one that was *intended* to be made 
generic in the hal/v85x ports. Search for CYGDBG_HAL_V850_ICE, and in 
particular look at hal/v85x/v850/current/src/v850_ice.cxx.

There's also one for the ARM at hal/arm/arch/current/src/arm_stub.c but 
that isn't quite as good as it requires stubs to be included in applications.

> b) should a debugger (we) provide a C source that the user includes in his
> build

Ideally such support should be built into the eCos port! If it's for an 
existing target, contributing it back to the community would be the best 
thing for users. Of course the support would want to be default off for 
users who don't need it.

Jifl
-- 
eCosCentric    http://www.eCosCentric.com/    The eCos and RedBoot experts
--[ "You can complain because roses have thorns, or you ]--
--[  can rejoice because thorns have roses." -Lincoln   ]-- Opinions==mine

RE: Debugger OS-awareness

[Janez Ulcakar]

Thanks so far. I think we'll be able to delve in next month.

-----Original Message-----
From: Jonathan Larmour [mailto:jifl@eCosCentric.com]
Sent: Friday, February 28, 2003 17:20
To: Janez Ulcakar
Cc: Andrew Lunn; ecos-maintainers@sources.redhat.com
Subject: Re: Debugger OS-awareness


Janez Ulcakar wrote:
> Hi,
>
> You're correct in your assumptions. We've encountered that 'special
function
> spawned by the ICE' approach before, and I think it's better suited for an
> ICE.
>
> a) Does eCos already provide such a function (how's it called, where does
it
> take it's params from, and where do we read the results from and what
> format)

It wasn't made generic but there's one that was *intended* to be made
generic in the hal/v85x ports. Search for CYGDBG_HAL_V850_ICE, and in
particular look at hal/v85x/v850/current/src/v850_ice.cxx.

There's also one for the ARM at hal/arm/arch/current/src/arm_stub.c but
that isn't quite as good as it requires stubs to be included in
applications.

> b) should a debugger (we) provide a C source that the user includes in his
> build

Ideally such support should be built into the eCos port! If it's for an
existing target, contributing it back to the community would be the best
thing for users. Of course the support would want to be default off for
users who don't need it.

Jifl
--
eCosCentric    http://www.eCosCentric.com/    The eCos and RedBoot experts
--[ "You can complain because roses have thorns, or you ]--
--[  can rejoice because thorns have roses." -Lincoln   ]-- Opinions==mine