hal: k20 - flash config section addition

[0xc0170]

This is as I described in #212. Tested with pyOCD and binary mbed cmsis-dap interface (binary flashing)

[bharrisau]

LGTM. Just get @bgamari to confirm it all looks fine on his toolchain.

[bgamari]

I'll test it when I get home. I am a bit worried, however, that k20's layout.ld will now fall out of sync with the linker script used by the rest of the platforms. What could we do about this?

[0xc0170]

All kinetis device would be out of sync. I believe there are other chips which have some specific sections, so would need to override generic layout. Like this kinetis targets and I recall also lpc1768 has two ram AHBSRAM0/1.

I'll look at linkers commands if there's anything we can leverage.

[bharrisau]

Just another instance where a more flexible linker would be useful. I think
the MCHCK project is doing something to split sections up so that code can
be fit into the gap before the 0x400 FCFG field.
On 04/11/2014 4:20 pm, "Martin Kojtal" [email protected] wrote:

All kinetis device would be out of sync. I believe there are other chips
which have some specific sections, so would need to override generic
layout. Like this kinetis targets and I recall also lpc1768 has two ram
AHBSRAM0/1.

I'll look at linkers commands if there's anything we can leverage.

—
Reply to this email directly or view it on GitHub
#217 (comment).

[0xc0170]

@bharrisau yes, there are plenty of space, mainly for cortex m0 kinetis devices which have just dozen of vectors, thus I used to place there some code to fill that gap.

Can you elaborate on "another instance"?

[farcaller]

a bit worried, however, that k20's layout.ld will now fall out of sync with the linker script used by the rest of the platforms

There's really nothing we can do, layout_common is "common" just to some extent, I don't think it's possible to INCLUDE within SECTIONS, right? Extracting the discard part would be nice, otherwise.

[bgamari]

We could use a preprocessor to generate our linker scripts. If I'm not mistaken libopencm3 does something along these lines.

[0xc0170]

I propose to create an issue for generic ld, where we can continue discussing it? And can add a comment to K20 ld or TODO as I noticed are used in this project.

@bgamari looks like they do https://github.com/libopencm3/libopencm3/blob/master/ld/linker.ld.S

[bgamari]

@0xc0170 by the way, I got home a bit late last night to check the patch. Tonight I'm going to make an effort to get home before 23:00 and do some testing.

Concerning OpenOCD's behavior, the code that you cite certainly suggests that overlapping sections are handled properly. I wonder what was causing the behavior I observed earlier? I was flashing the blink_k20 example with gdb's load command and openocd.

[0xc0170]

Any progress here? Shall it be merged, to fix FCG config ?

Another pull request #219 vectors, is refactoring isr vectors for k20, I would like to see there FCG included. 😄

[bgamari]

Quite sorry about the rather ridiculous review period. Seems to work on my end.

@0xc0170 thanks for working this out!