ARM: implement the ProcessInfo methods

Information
Submitter:	Tony Gutierrez
Repository:	gem5
Branch:	default
Bugs:
Depends On:
Reviewers
Groups:	Default
People:

Description

Changeset 8919:59db10fbd30e
---------------------------
ARM: implement the ProcessInfo methods

Testing Done

Booted modified kernel (with thread info hints compiled in) while printing process names, using ProcessInfo::name(), periodically from atomic cpu's tick() method.

Issue Summary

Description	From	Last Updated	Status
Is this value correct, or is it just what Alpha used?	Gabe Black	Feb. 22, 2012, 5:32 a.m.	Resolved
I'm assuming this is copied from Alpha. It's probably not appropriate for ARM.	Gabe Black	Feb. 20, 2012, 6:50 p.m.	Open
This value is certainly wrong for ARM.	Ali Saidi	Feb. 22, 2012, 5:28 a.m.	Resolved
I'm fairly certain this isn't correct. I've got a kernel that has it at f8000000, so it's not generic. There ...	Ali Saidi	March 26, 2012, 2:19 a.m.	Open
console is an alpha concept. "Unknown" is probably better here.	Ali Saidi	March 26, 2012, 2:57 a.m.	Resolved

Don't just copy in the Alpha implementations. Either figure out what ARM should really be using, make them panic, or leave them unimplemented.

src/arch/arm/stacktrace.cc (Diff revision 1)

Is this value correct, or is it just what Alpha used?

Show all issues

Tony Gutierrez Feb. 21, 2012, 3:12 a.m. (Feb. 21, 2012, 3:12 a.m.)

You may be right. I just copied this value and it worked, i.e., I could get the process's name. I'll see if I can find the kernel base address for the RealView type of system modeled in gem5. Unless anybody knows it offhand? Ali?

src/arch/arm/stacktrace.cc (Diff revision 1)

I'm assuming this is copied from Alpha. It's probably not appropriate for ARM.

Show all issues

Tony Gutierrez Feb. 21, 2012, 3:12 a.m. (Feb. 21, 2012, 3:12 a.m.)

This is the same for MIPS/Alpha/x86. I don't really know why "console" is assumed in this case. So, I left it.

src/arch/arm/stacktrace.cc (Diff revision 1)

Does this work? I'm surprised that it would be identical for arm and alpha

Tony Gutierrez Feb. 21, 2012, 4:28 a.m. (Feb. 21, 2012, 4:28 a.m.)

Yes this works. I changed this from 0xcfff to 0x1fff because ARM has 8K pages.

Tony Gutierrez Feb. 21, 2012, 4:30 a.m. (Feb. 21, 2012, 4:30 a.m.)
```
8K stack, not pages.
```

src/arch/arm/stacktrace.cc (Diff revision 1)

This value is certainly wrong for ARM.

Show all issues

Tony Gutierrez Feb. 21, 2012, 4:28 a.m. (Feb. 21, 2012, 4:28 a.m.)

Do you know where I can find the actual value? Can't seem to locate it in the docs/kernel source.

Tony Gutierrez Feb. 22, 2012, 5:32 a.m. (Feb. 22, 2012, 5:32 a.m.)

According to the source and config phys PAGE_OFFSET = 0xc0000000 and PHYS_OFFSET = 0x00000000.

From:

config_arm_2.6.38_mp_fb from gem5.org ARM full-system files
arch/arm/mach-realview/include/mach/memory.h from linux-2.6 kernel source

Description:

~		Changeset 8849:bdd7f776a182
	~	Changeset 8849:b3fd47d9957d

~		ARM: implement the ProcessInfo methods
	~	changed kernel segment base address to 0xc0000000

Diff:

Revision 2 (+34 -3)

Show changes

src/arch/arm/stacktrace.cc

Testing Done:

Booted modified kernel (with thread info hints compiled in) while printing process names, using ProcessInfo::name(), periodically from atomic cpu's tick() method.

Description:

~		Changeset 8849:b3fd47d9957d
	~	Changeset 8917:6db3a60bd4a0

~		changed kernel segment base address to 0xc0000000
	~	changed from pointers to references for the port proxies

Diff:

Revision 3 (+30 -3)

Show changes

src/arch/arm/stacktrace.cc

Description:

		Changeset 8917:6db3a60bd4a0

~		changed from pointers to references for the port proxies
	~	changed from pointers to references for the port proxies
	+	it is still not known whether or not the base value is correct

Description:

~		Changeset 8917:6db3a60bd4a0
	~	Changeset 8917:b79bdeef440f

~		changed from pointers to references for the port proxies
	~	changed from pointers to references for the port proxies. it is still not known whether 0xffffffffc0000000 is the correct value to compare against base.
-		it is still not known whether or not the base value is correct

Description:

~		Changeset 8917:b79bdeef440f
	~	Changeset 8917:271859b2e61b

		changed from pointers to references for the port proxies. it is still not known whether 0xffffffffc0000000 is the correct value to compare against base.

Diff:

Revision 4 (+30 -3)

Show changes

src/arch/arm/stacktrace.cc

Where are you calling this these functions from? How are you getting a kernel stack pointer?

Tony Gutierrez March 26, 2012, 2:51 a.m. (March 26, 2012, 2:51 a.m.)

Whenever an instruction was fetched - so inside of cache_impl.hh access() something like this:

if (pkt->req->isInstFetch()) {
    ThreadContext *tc = system->getThreadContext(id);
    Addr stackPtr = tc->readIntReg(INTREG_SP_SVC);
    ProcessInfo procInfo(tc);

    std::cout << "process: " << procInfo.name(stackPtr) << std::endl;
}

src/arch/arm/stacktrace.cc (Diff revision 4)

I'm fairly certain this isn't correct. I've got a kernel that has it at f8000000, so it's not generic. There might be something in the kernel that tells you the start address, but either way I'm not sure what this code is doing in the the context of ARM linux. For Alpha, we knew that the initial stack frame was placed at this address, I don't think we know that for ARM.

Show all issues

Tony Gutierrez March 26, 2012, 2:50 a.m. (March 26, 2012, 2:50 a.m.)

There is little documentation on this but, from what I gathered this value is dependent on the machine type. So, I got the values from the kernel source and configs used for the RealView PBX machine. Maybe this value can be parameterized...

src/arch/arm/stacktrace.cc (Diff revision 4)

console is an alpha concept. "Unknown" is probably better here.

Show all issues

Description:

~		Changeset 8917:271859b2e61b
	~	Changeset 8919:59db10fbd30e

~		changed from pointers to references for the port proxies. it is still not known whether 0xffffffffc0000000 is the correct value to compare against base.
	~	ARM: implement the ProcessInfo methods

Diff:

Revision 5 (+30 -3)

Show changes

src/arch/arm/stacktrace.cc

Ship It!

You have a pending review.

Review Board 2.0.15

This change has been marked as submitted.

Screenshots

Files

Issue Summary

Description:

Changeset 8849:bdd7f776a182

Changeset 8849:b3fd47d9957d

Diff:

Testing Done:

Description:

Changeset 8849:b3fd47d9957d

Changeset 8917:6db3a60bd4a0

Diff:

Description:

Changeset 8917:6db3a60bd4a0

Description:

Changeset 8917:6db3a60bd4a0

Changeset 8917:b79bdeef440f

Description:

Changeset 8917:b79bdeef440f

Changeset 8917:271859b2e61b

Diff:

Description:

Changeset 8917:271859b2e61b

Changeset 8919:59db10fbd30e

Diff:

Status: Closed (submitted)