!rsKWMJGPMsyPTTjXuh:matrix.org

Dasharo - General

275 Members
General discussion about Dasharo and open source firmware projects maintained, developed and planned by 3mdeb. Feel free to bring topics related to open source firmware, open source hardware and open instruction set architecture.41 Servers

Load older messages


SenderMessageTime
7 Dec 2021
@insurgo:matrix.orgInsurgo, Technologies Libres / Open Technologies (Timezone: EDT, UTC-5) *

tonux: There were changes on Heads' CircleCI configuration to permit seperation of per arch musl-cross-make being built first. (same logic will apply to Talos II for ppc64 musl-cross-make being built first, then workspace being passed to the actual board build which has to build under 1h).

As stated prior in other tickets (when I bumped my head into making builds work with new CircleCI free tier limitations) you might have issues being able to build KGPE-D16_workstation, since that board's linux config actually builds for all GPU out there, and last time I checked in my tests, this takes around 40 minutes even if built with 36 make_jobs?!

This is where my Make kung fu stops. It seems, for some reasons, that coreboot 4.13 boards builds are successfully building with 36 jobs under CircleCI, where 4.11 seemed to build, from the coreboot passed linux build statement, with 4 make_jobs or so.

I was not able to understand where that problem came from, and had to drop kgpe-d16 builds since the workstation was taking too much time to build and made the whole building of all other boards fail.

When you are at that stage in your testings, ping me in. I would love to have someone else's brain double check this.

Relative, and important changes in Heads master recently in regard of CircleCI and build paralellization:

  • https://github.com/osresearch/heads/commit/e492786d0a52fd1e2100629b523c7510797329bb
  • https://github.com/osresearch/heads/commit/ee5073ebe8613051ee2d233a0d4e3ccb36956aaa
21:02:12
@pietrushnic:matrix.orgpietrushnic (Old)
In reply to @insurgo:matrix.org

pietrushnic: Can we rename this room? Passed under my radar! Excited about KGPE-D16 progress!!!!!!

Channel naming advice: there is a lot of channels named general. Name containing parenthood would ease differentiation: Dasharo - General for example. No need to rename the published matrix channel, but the actual name of the rooms should be clearer!

Insurgo, Technologies Libres / Open Technologies (Timezone: EDT, UTC-5): definitely will do it right not
21:22:50
@pietrushnic:matrix.orgpietrushnic (Old)now*21:22:52
@pietrushnic:matrix.orgpietrushnic (Old) changed the room name to "Dasharo - General" from "general".21:24:36
@pietrushnic:matrix.orgpietrushnic (Old) set the room topic to "General discussion about Dasharo and open source firmware projects maintained, developed and planned by 3mdeb. Feel free to bring topics related to open source firmware, open source hardware and open instruction set architecture.".21:24:37
@tonux:matrix.orgtonux
In reply to @insurgo:matrix.org

tonux: There were changes on Heads' CircleCI configuration to permit seperation of per arch musl-cross-make being built first. (same logic will apply to Talos II for ppc64 musl-cross-make being built first, then workspace being passed to the actual board build which has to build under 1h).

As stated prior in other tickets (when I bumped my head into making builds work with new CircleCI free tier limitations) you might have issues being able to build KGPE-D16_workstation, since that board's linux config actually builds for all GPU out there, and last time I checked in my tests, this takes around 40 minutes even if built with 36 make_jobs?!

This is where my Make kung fu stops. It seems, for some reasons, that coreboot 4.13 boards builds are successfully building with 36 jobs under CircleCI, where 4.11 seemed to build, from the coreboot passed linux build statement, with 4 make_jobs or so.

I was not able to understand where that problem came from, and had to drop kgpe-d16 builds since the workstation was taking too much time to build and made the whole building of all other boards fail.

When you are at that stage in your testings, ping me in. I would love to have someone else's brain double check this.

Relative, and important changes in Heads master recently in regard of CircleCI and build paralellization:

  • https://github.com/osresearch/heads/commit/e492786d0a52fd1e2100629b523c7510797329bb
  • https://github.com/osresearch/heads/commit/ee5073ebe8613051ee2d233a0d4e3ccb36956aaa
I'm not fantastic at make to be honest, but yeah I'll ping you. I'll open a POC PR at some point but won't be too soon as still studying atm.
21:33:21
@pietrushnic:matrix.orgpietrushnic (Old) miczyg: Krystian Hebel any ideas what is going on there that coreboot build system calculate job 21:40:38
@pietrushnic:matrix.orgpietrushnic (Old)

I couldn't resist:

 _____________________________________
/ Dasharo is open source firmware     \
| distribution, we prefer clean and   |
| simple code, long term maintenance, |
| privacy-respecting implementation,  |
| liberty for the owners, and         |
\ trustworthiness for all.            /
 -------------------------------------
      \                    / \  //\
       \    |\___/|      /   \//  \\
            /0  0  \__  /    //  | \ \    
           /     /  \/_/    //   |  \  \  
           @_^_@'/   \/_   //    |   \   \ 
           //_^_/     \/_ //     |    \    \
        ( //) |        \///      |     \     \
      ( / /) _|_ /   )  //       |      \     _\
    ( // /) '/,_ _ _/  ( ; -.    |    _ _\.-~        .-~~~^-.
  (( / / )) ,-{        _      `-.|.-~-.           .~         `.
 (( // / ))  '/\      /                 ~-. _ .-~      .-~^-.  \
 (( /// ))      `.   {            }                   /      \  \
  (( / ))     .----~-.\        \-'                 .~         \  `. \^-.
             ///.----..>        \             _ -~             `.  ^-`  ^-_
               ///-._ _ _ _ _ _ _}^ - - - - ~                     ~-- ,.-~
                                                                  /.-~
22:21:30
8 Dec 2021
@pietrushnic:matrix.orgpietrushnic (Old)jeez: https://github.com/andreiw/ppc64le-edk200:53:46
@matt_b7:matrix.orgmatt_b7what.00:59:47
@matt_b7:matrix.orgmatt_b7I get that UEFI is a standard that answers a LOT of questions about how a system should come up.......but is this product of microsoft even well-defined enough to cover POWER?01:00:42
@matt_b7:matrix.orgmatt_b7petiteboot seems to have this working pretty well, even if the kexec mechanism can't pass all the same information as easily01:01:27
@matt_b7:matrix.orgmatt_b7and really, who wouldn't want to run linux on power? :P01:03:17
@matt_b7:matrix.orgmatt_b7(or BSD, etc)01:03:48
@pietrushnic:matrix.orgpietrushnic (Old):)01:09:04
@pietrushnic:matrix.orgpietrushnic (Old)There is quite a lot of work around LinuxBoot to make UEFI services work across kexec boundary01:09:32
@pietrushnic:matrix.orgpietrushnic (Old)unfortunately on OSFC Ron expressed his concerns how far kexec can go, I had hope this is new shine star, but it looks it is probably 5+ years for kexec to take over, if ever01:10:10
@pietrushnic:matrix.orgpietrushnic (Old)someone should do comprehensive OST2 training about boot protocols and existing boot paths01:10:32
@matt_b7:matrix.orgmatt_b7mmm. I guess kind of the whole point of kexec is to cut off old services, and cut into the new kernel/environment01:11:26
@matt_b7:matrix.orgmatt_b7passing it along means, at the very least, leaving it and it's stuff in memory01:11:57
@matt_b7:matrix.orgmatt_b7with hardware still untouched enough that it'll still work01:12:22
@matt_b7:matrix.orgmatt_b7I would expect it to become harder and harder to preserve old services though many stages of chainloading, as there's the possibility assumptions will be broken (either it's or something else's)01:13:59
@matt_b7:matrix.orgmatt_b7but what do I know?01:14:19
@pietrushnic:matrix.orgpietrushnic (Old)Please take a look at LPC security MC, concept of non-UEFI world is unknown to Linux maintainers, even IBM employees resist boot protocol and boot path changes 01:24:16
@pietrushnic:matrix.orgpietrushnic (Old)https://youtu.be/YpOwgZNC-jY01:25:22
@miczyg:matrix.orgmiczyg
In reply to @pietrushnic:matrix.org
miczyg: Krystian Hebel any ideas what is going on there that coreboot build system calculate job
I don't exactly understand what do you mean by build system calculate job pietrushnic could you clarify?
09:24:21
@miczyg:matrix.orgmiczygFYI guys, apu2 already fulfills the requirements for the ext releases of coreboot: https://review.coreboot.org/c/coreboot/+/52761 https://review.coreboot.org/c/coreboot/+/59808 https://review.coreboot.org/c/coreboot/+/59810 PARALLEL_MP, RESOURCE_ALLOCATOR_V4, NO_SMM (instead of SMM_TSEG, because not having SMM is an advantage) And everything on binary AGESA platform. family14 AGESA (apu1) is in a good progress to reflect apu2 support: https://review.coreboot.org/c/coreboot/+/52780 https://review.coreboot.org/c/coreboot/+/59807 https://review.coreboot.org/c/coreboot/+/59809 Hopefully we can move with fam15tn and fam16kb when time allows.09:28:18
@miczyg:matrix.orgmiczyg * FYI guys, apu2 already fulfills the requirements for the next releases of coreboot: https://review.coreboot.org/c/coreboot/+/52761 https://review.coreboot.org/c/coreboot/+/59808 https://review.coreboot.org/c/coreboot/+/59810 PARALLEL_MP, RESOURCE_ALLOCATOR_V4, NO_SMM (instead of SMM_TSEG, because not having SMM is an advantage) And everything on binary AGESA platform. family14 AGESA (apu1) is in a good progress to reflect apu2 support: https://review.coreboot.org/c/coreboot/+/52780 https://review.coreboot.org/c/coreboot/+/59807 https://review.coreboot.org/c/coreboot/+/59809 Hopefully we can move with fam15tn and fam16kb when time allows.09:28:30
@pietrushnic:matrix.orgpietrushnic (Old)
In reply to @miczyg:matrix.org
I don't exactly understand what do you mean by build system calculate job pietrushnic could you clarify?
Please check above Insurgo, Technologies Libres / Open Technologies (Timezone: EDT, UTC-5) message https://matrix.to/#/!rsKWMJGPMsyPTTjXuh:matrix.org/$9IjPLe_H_7p-l3zd_WATSMkYByFGekMAOxlozO7gutY?via=matrix.org
10:19:07
@miczyg:matrix.orgmiczygI also recall having problems with multijob coreboot builds back in 4.11. But the situation improved in newer releases (dunno why thou). My make-fu also isn't the best out there...11:30:26

There are no newer messages yet.


Back to Room ListRoom Version: 6