hwmv2: qemu_x86: acrn: Port to HWMv2

[golowanow]

Port qemu_x86 to HWMv2 as well as SoCs it depends on:

• ia32
• lakemont

Several qemu_x86 board configurations have changed their identifiers to the HWMv2 scheme:

• qemu_x86_nokpti --> qemu_x86/ia32/nokpti
• qemu_x86_nommu --> qemu_x86/ia32/nommu
• qemu_x86_nopae --> qemu_x86/ia32/nopae
• qemu_x86_virt --> qemu_x86/ia32/virt
• qemu_x86_xip --> qemu_x86/ia32/xip
• qemu_x86_64_nokpti --> qemu_x86/ia32/nokpti
• qemu_x86_tiny@768 --> qemu_x86_tiny/ia32/768

Also x86/acrn boards: acrn and acrn_ehl_crb

[golowanow]

hmm, is qemu_x86_virt (and Co.) a board itself, or just a revision of qemu_x86@virt ?
currently west accepts both, but twister only as a board.

[golowanow]

@nashif, is it known and considered acceptable: with the HWMv2 board names Twister creates additional levels of directory structure under ./twister-out for board@revision/soc/variant instead of just one old board_soc_variant@revision, for example after ./scripts/twister -vv -a x86 -T tests/arch/x86/pagetables/ it looks like below

./twister-out
   |-qemu_x86
   |---ia32
   |-----nokpti
   |-------tests
   |---------arch
   |-----------x86
   |-------------pagetables
   |---------------arch.x86.pagetables
...

[nashif]

@nashif, is it known and considered acceptable: with the HWMv2 board names Twister creates additional levels of directory structure under ./twister-out for board/soc/revision instead of just one old board_soc_revision, for example after ./scripts/twister -vv -a x86 -T tests/arch/x86/pagetables/ it looks like below

./twister-out
   |-qemu_x86
   |---ia32
   |-----nokpti
   |-------tests
   |---------arch
   |-----------x86
   |-------------pagetables
   |---------------arch.x86.pagetables
...

No, was not aware of this.

This is a mess :( I do not like the slash in the board names and IMO this is going to be a problem and a blocker. FYI @tejlmand

[golowanow]

two problems with CI twister tests currently on this PR:

1. qemu_xtensa tests/subsys/llext/hello_world/llext.simple.modules_enabled - it seems not related to this change.
2. old x86\acrn can't run properly as it depends on soc/v2/ia32 here and soc/x86/elkhart_lake which is in transition at hwmv2: port intel x86 boards #67885.
   I think acrn might be moved as a separate PR when both SoC it uses migrate to the collab-hwm, or at least one of them.

[golowanow]

@tejlmand - could you please provide some feedback - is this PR on track with the general v1 to v2 migration principles ?

[nashif]

regarding twister, would be nice to patch twister to replace / with _ in the build_dir to restore existing functionality and tree structure.

[tejlmand]

looks very nice.
Only minor nits observed.

[tejlmand]

with hwmv2 it is now possible to actually implement a board revision check here:

zephyr/boards/v2/qemu/qemu_x86/revision.cmake

Lines 4 to 5 in dcd3130

	# We can't really do board_check_revision() here
	# as it will apply to all qemu_x86* boards.

because only boards which specifies revision: format: custom will now load this file.

All other boards in this folder / yaml file will ignore the revision.cmake file.

Note, this comment is not blocking for this PR, more an extra piece of info.

[tejlmand]

This is a mess :( I do not like the slash in the board names and IMO this is going to be a problem and a blocker. FYI @tejlmand

should be as simple as having twister sanitize the name before creating the folder structure, but would need to see if twister in other places than folder creation is also constructing the path based on board names.

[tejlmand]

should be as simple as having twister sanitize the name before creating the folder structure, but would need to see if twister in other places than folder creation is also constructing the path based on board names.

added as open task here: #51831

[tejlmand]

approved, nits can be addressed in followup.

[nordicjm]

These files are named wrongly, the base here is qemu_x86 which will be included for all the qemu_x86/* boards (but not qemu_x86_64/* boards), but for variants, the soc name needs to be included, so above where you have identifier: qemu_x86/ia32/nommu this can be used to work out the filename, replace / with _ to get: qemu_x86_ia32_nommu - this is the name of the files for this variant e.g. qemu_x86_ia32_nommu.dts, qemu_x86_ia32_nommu_defconfig, qemu_x86_ia32_nommu.yaml etc.
Note that for the nommu config, it will do this for deconfig: qemu_x86_defconfig + qemu_x86_ia32_defconfig + qemu_x86_ia32_nommu_defconfig and combine them in that order

[golowanow]

thank you, file names are adjusted

[golowanow]

Needs rebase,

rebased, now it needs #68801

[golowanow]

Needs rebase, and up squared 7000 board is failing CI: https://github.com/zephyrproject-rtos/zephyr/actions/runs/7843368285/job/21403615547?pr=68801

@nordicjm it seems up_squared_pro_7000 is failing because changes from #62694 were not applied to the board s migrated with collab_hwm branch on its rebase.
I'll prepare a fix for the intel boards, but isn't something wrong with that 'v2' / 'legacy' bulk move ?
@tejlmand
fix in #68814

[golowanow]

@nordicjm do you have objections to merge this PR ?

• compliance check passing if [hwm2] ci: update check_compliance to not create duplicate lines in Kconfig #68811 applied
• twister run issues are not related to this PR

[nordicjm]

@nordicjm do you have objections to merge this PR ?

* compliance check passing if [[hwm2] ci: update check_compliance to not create duplicate lines in Kconfig #68811](https://github.com/zephyrproject-rtos/zephyr/pull/68811) applied

* twister run issues are not related to this PR

Needs another approval

[golowanow]

Needs another approval

@jhedberg, @tejlmand, @nashif could you please look at this PR once more ?

[golowanow]

rebased to last changes at collab-hwm

[nordicjm]

acrn board needs fixing

[golowanow]

acrn board needs fixing

@nordicjm , it is planned as separate PR to follow on this one.

#68078 (comment)

2. old x86\acrn can't run properly as it depends on soc/v2/ia32 here and soc/x86/elkhart_lake which is in transition at hwmv2: port intel x86 boards #67885.

[nordicjm]

acrn board needs fixing

@nordicjm , it is planned as separate PR to follow on this one.

#68078 (comment)

2. old x86\acrn can't run properly as it depends on soc/v2/ia32 here and soc/x86/elkhart_lake which is in transition at hwmv2: port intel x86 boards #67885.

Can you put the commits here? Trying to fix up CI on collab-hwm not make it worse

[golowanow]

acrn board needs fixing

@nordicjm , it is planned as separate PR to follow on this one.
#68078 (comment)

2. old x86\acrn can't run properly as it depends on soc/v2/ia32 here and soc/x86/elkhart_lake which is in transition at hwmv2: port intel x86 boards #67885.

Can you put the commits here? Trying to fix up CI on collab-hwm not make it worse

@nordicjm, done, x86/acrn is also here.