ThinkPad X201 compile error

[Federico Amedeo Izzo]

Hello,

I wanted to update coreboot on my X201 from the Feb 2017 source
(commit 068edc1c52cb1e5b6376ba7f296ef8797a24cd5f)
to current master
(commit f2e7b37c52cac5e7825d0e01b3c45c1506e99253)
But when compiling it gives the following error
```
src/northbridge/intel/nehalem/gma.c: In function 'gma_func0_init':
src/northbridge/intel/nehalem/gma.c:1066:5: error: implicit declaration
of function 'intel_gma_init' [-Werror=implicit-function-declaration]
     intel_gma_init(conf, res2mmio(gtt_res, 0, 0),```

Now it's configured to load a VGA blob, if I enable "native graphics
init" instead, the compile error goes away, but the resulting image does
not boot (beep and flashing leds)

I attached the configuration that gives the compilation error.

Do you have any suggestions?

Thanks,

Federico

[Nico Huber]

On 18.12.2017 11:44, Federico Amedeo Izzo wrote:
> Hello,
>
> I wanted to update coreboot on my X201 from the Feb 2017 source
> (commit 068edc1c52cb1e5b6376ba7f296ef8797a24cd5f)
> to current master
> (commit f2e7b37c52cac5e7825d0e01b3c45c1506e99253)
> But when compiling it gives the following error
> src/northbridge/intel/nehalem/gma.c: In function 'gma_func0_init':
> src/northbridge/intel/nehalem/gma.c:1066:5: error: implicit declaration
> of function 'intel_gma_init' [-Werror=implicit-function-declaration]
>      intel_gma_init(conf, res2mmio(gtt_res, 0, 0),```

Wooops, my fault! Fix here:
  https://review.coreboot.org/#/c/coreboot/+/22930

You just need to remove the #if and #endif in lines 575 and 1012,
respectively.
> Now it's configured to load a VGA blob, if I enable "native graphics
> init" instead, the compile error goes away, but the resulting image does
> not boot (beep and flashing leds)

I fear this error is not gfx related. The Nehalem code isn't widely tes-
ted, I suggest trying again with different DIMMS / DIMM combinations.

Hope that helps,
Nico

[Federico Amedeo Izzo]

On 12/18/2017 12:14 PM, Nico Huber wrote:
> On 18.12.2017 11:44, Federico Amedeo Izzo wrote:
> > Hello,
> >
> > I wanted to update coreboot on my X201 from the Feb 2017 source
> > (commit 068edc1c52cb1e5b6376ba7f296ef8797a24cd5f)
> > to current master
> > (commit f2e7b37c52cac5e7825d0e01b3c45c1506e99253)
> > But when compiling it gives the following error
> > src/northbridge/intel/nehalem/gma.c: In function 'gma_func0_init':
> > src/northbridge/intel/nehalem/gma.c:1066:5: error: implicit declaration
> > of function 'intel_gma_init' [-Werror=implicit-function-declaration]
> >      intel_gma_init(conf, res2mmio(gtt_res, 0, 0),```
>
> Wooops, my fault! Fix here:
>   https://review.coreboot.org/#/c/coreboot/+/22930
>
> You just need to remove the #if and #endif in lines 575 and 1012,
> respectively.

Thanks, the fix works, and now the master compiles again.
> > Now it's configured to load a VGA blob, if I enable "native graphics
> > init" instead, the compile error goes away, but the resulting image does
> > not boot (beep and flashing leds)
>
> I fear this error is not gfx related. The Nehalem code isn't widely tes-
> ted, I suggest trying again with different DIMMS / DIMM combinations.
>
> Hope that helps,
> Nico

The current coreboot master does not boot even with VGA blob, and it
looks like a memory problem as you were suggesting.
In fact watching the serial console it appears that the DIMM training
phase does not finish, and after a while the computer turns itself off.
I've also tried different DIMMS and combinations but without success.

I'm just finished doing a `git bisect` keeping the working commit as
Fallback and the one to test as a Normal.
This attempt was not a success, because even with the last working
commit, when selecting "Normal" mode via nvramcui, it always reboots in
Fallback mode.

Do you have any other suggestions?

Thanks,
Federico

[Nico Huber]

On 23.12.2017 17:12, Federico Amedeo Izzo wrote:
> This attempt was not a success, because even with the last working
> commit, when selecting "Normal" mode via nvramcui, it always reboots in
> Fallback mode.

You also need to clear the `reboot_counter` variable. Otherwise the
logic assumes "Normal" was already tried x times.

Nico

[Federico Amedeo Izzo]

On 12/23/2017 05:25 PM, Nico Huber wrote:
> On 23.12.2017 17:12, Federico Amedeo Izzo wrote:
> > This attempt was not a success, because even with the last working
> > commit, when selecting "Normal" mode via nvramcui, it always reboots in
> > Fallback mode.
>
> You also need to clear the `reboot_counter` variable. Otherwise the
> logic assumes "Normal" was already tried x times.

Perfect! clearing the reboot_counter I can boot in Normal mode with the
last working commit and cannot boot the Normal mode with the master commit.

I've done again the `git bisect` procedure, this time having reasonable
results:
initial tested commit: 068edc1c52, works
last working commit: 01ea8f1fcc, works
first broken commit: 806ea08bb2, not booting

Federico

[Nico Huber]

Hi Federico,

On 24.12.2017 15:53, Federico Amedeo Izzo wrote:
> last working commit: 01ea8f1fcc, works
> first broken commit: 806ea08bb2, not booting

I've just compared binaries of these two commits. It seems that the
plain addition of the unused function in 806ea08 changes the compiler's
decision when other functions in that file are inlined. That *could*
result in a stack overflow where none was before (didn't look into
callers of that code yet, if it's likely). Maybe that gives you some
clue.

See you tomorrow, I guess
Nico

PS. are you on IRC?