u-boot: 2023.07.02 -> 2023.10 #258701

jmbaur · 2023-10-03T02:42:14Z

Description of changes

Among other changes, this release moves more boards to use the new environment format where a simple text file is used under ./board rather than using CFG_EXTRA_ENV_SETTINGS. This includes the default environment for the raspberry pi, hence the change of patch.

Something I haven't been able to track down is the need for removing the DTC make flag. U-boot has a version of this (along with libfdt) in their tree that builds just fine, so I propose we use that instead.

I've tested these changes building various boards defined in nixpkgs (cross compiling from x86_64-linux) with good success. I'm not sure of the expectation that every u-boot board defined in nixpkgs build successfully, (as I saw one with an incorrect filesToInstall), so I have not tested all boards. Please test if these changes break your board!

Things done

U-boot seems to have a dtc and libfdt implementation that works better with their build system. Use that instead. Builds fail when specifying `DTC=<our dtc>`.

There does not exist a file to install at $BUILD/u-boot-spl.bin, but there is a file named $BUILD/u-boot-with-spl.bin, which is most likely the file that a user wants to boot their clearfog device with.

jmbaur · 2023-10-03T02:47:19Z

@ofborg build pkgsCross.aarch64-multiplatform.ubootRaspberryPi4_64bit

jmbaur · 2023-10-03T03:08:17Z

@ofborg build ubootTools

samueldr

(No changes requested)

All devices should be tried to be built (pkgsCross on their arch) before merging.

It is fine to assume built == valid, given the track record of U-Boot.

The mkimage tool requires openssl for the host platform.

jmbaur · 2023-10-04T03:08:38Z

I quickly built all immediately build-able (with a pkgsCross package set available) u-boot derivations and gathered the results here: https://gist.github.com/jmbaur/466df3f4200914f906c05994d3f38674. All of them that fail also fail on master for the same reason. Luckily the failure is the same, so the fix should be easy. I'm happy to try and track down the SCP binary for these boards and include that in this PR, I'm assuming this means they will need to be marked unfree. I'm also happy keeping them broken and fixing them in a separate PR, just let me know. I also fixed cross-compilation of ubootTools.

lopsided98 · 2023-10-04T22:39:43Z

This PR fixes the SCP build failures: #252111

lopsided98 · 2023-10-04T22:45:24Z

These are the attributes for the builds with question marks:

pkgsCross.raspberryPi.ubootRaspberryPi
pkgsCross.raspberryPi.ubootRaspberryPiZero
pkgsCross.sheevaplug.ubootSheevaplug
pkgsCross.sheevaplug.ubootGuruplug

jmbaur · 2023-10-04T23:43:10Z

Thanks @lopsided98! The rest of them build successfully (gist updated), and that PR you linked would cover the rest that are failing.

lopsided98 · 2023-10-05T03:06:20Z

Successfully booted:

ubootRaspberryPi2
ubootRockPro64: Initially failed with Error enabling VMMC supply : -114, fixed after applying https://patchwork.ozlabs.org/project/uboot/list/?series=364753&state=*
ubootRock64v2 (from u-boot: ROCK64 RAM init improvements #191538): ~~works with above patch applied (didn't test without, but it is probably necessary)~~ (edit: I accidentally applied the patch to ubootRock64 instead of ubootRock64v2, but ubootRock64v2 boots fine anyway).
ubootOdroidC2

Failed:

ubootRaspberryPi4_64bit: can't debug right now, but fails to boot
ubootOdroidXU3: not tested, but known bugs have not been fixed, see: uboot: 2023.01 -> 2023.07 #227947 (comment)

samueldr · 2023-10-05T03:37:44Z

When looking for Tow-Boot, I've observed that all Raspberry Pi boards were moved to bootstd. So depending on the exact setup, it might be related to that move.

lopsided98 · 2023-10-06T00:02:16Z

I thought I had a saved environment on this RPi4, which might have broken things with bootstd, but this wasn't the problem.

The real problem is that, by default, U-Boot has boot_targets=mmc usb pxe dhcp. This appears to make it try mmc 1 first, which is an unconnected interface on the RPi4. This obviously fails, but rather than trying mmc 0, it just moves on to USB and PXE. Explicitly specifying boot_targets=mmc0 makes it boot successfully. Interestingly, u-boot/u-boot@c771e5b says:

We don't need to specify the mmc devices individually, since they are
used in order from 0 to 2, and standard boot uses that order anyway.

This appears to be incorrect, and the fix is probably to set boot_targets=mmc0 mmc1 mmc2 usb pxe dhcp

lopsided98 · 2023-10-06T00:33:04Z

Another odd RPi4 bug: the 2 second boot delay takes closer to 10 seconds

jmbaur · 2023-10-07T06:47:37Z

I added a patch to address this, let me know if you think this is appropriate or not. Unfortunately I do not have an rpi to try and reproduce this, but I'll check out the behavior on qemu with bootstd.

Edit: behavior on qemu with two virtio drives looks fine, bootflow scan -lb tries the first disk before trying the second, though this really isn't representative of what it seems like is happening on the rpi.

lopsided98 · 2023-10-07T15:59:12Z

I submitted the patch upstream; we'll see what they think: https://patchwork.ozlabs.org/project/uboot/patch/[email protected]/

pkgs/misc/uboot/default.nix

lopsided98 · 2023-10-08T22:46:50Z

Upstream suggested trying this patch, which also fixes the RPi4 issue: https://patchwork.ozlabs.org/project/uboot/patch/[email protected]/

This patch enables the raspberry pi (and really any board using standard boot without BOOTSTD_FULL) to boot properly when `boot_targets` only specifies the uclass of the device to boot from (e.g. "mmc" rather than "mmc0").

jmbaur · 2023-10-09T01:07:37Z

I included the patch here: https://patchwork.ozlabs.org/project/uboot/patch/[email protected]/raw

sorki · 2023-10-18T15:29:57Z

I missed this one and created #261775 which does bunch of other things as well.

Except for DTC=dtc removal the first patch is pretty much identical

I've added another patch fixing QEMU test
We can still use external dtc but it has to be a full path otherwise it fails horribly.

@jmbaur do you mind if I pick the 3 other patches to my branch and we continue there?

jmbaur · 2023-10-20T02:00:22Z

@sorki that sounds good to me!

jmbaur · 2023-10-20T02:00:40Z

closed in favor of #261775

sorki · 2023-10-20T09:34:35Z

Much appreciated @jmbaur!

jmbaur added 2 commits October 2, 2023 19:24

u-boot: 2023.07.02 -> 2023.10

f91bec0

U-boot seems to have a dtc and libfdt implementation that works better with their build system. Use that instead. Builds fail when specifying `DTC=<our dtc>`.

ubootClearfog: fix filesToInstall

d5adf32

There does not exist a file to install at $BUILD/u-boot-spl.bin, but there is a file named $BUILD/u-boot-with-spl.bin, which is most likely the file that a user wants to boot their clearfog device with.

ofborg bot requested review from dezgeg, lopsided98 and samueldr October 3, 2023 03:02

ofborg bot added 10.rebuild-darwin: 0 This PR does not cause any packages to rebuild on Darwin 10.rebuild-linux: 1-10 labels Oct 3, 2023

samueldr suggested changes Oct 3, 2023

View reviewed changes

ubootTools: fix cross-compile

1b6f79e

The mkimage tool requires openssl for the host platform.

jmbaur requested a review from samueldr October 4, 2023 03:19

jmbaur force-pushed the uboot-2023.10 branch from 070d344 to 344c0d3 Compare October 7, 2023 06:37

samueldr reviewed Oct 7, 2023

View reviewed changes

pkgs/misc/uboot/default.nix Outdated Show resolved Hide resolved

jmbaur force-pushed the uboot-2023.10 branch 2 times, most recently from 2f48bfc to 4d2e0c3 Compare October 7, 2023 20:16

ofborg bot requested a review from samueldr October 7, 2023 21:04

ubootRaspberryPi*: add patch to fix bootstd behavior

6a12107

This patch enables the raspberry pi (and really any board using standard boot without BOOTSTD_FULL) to boot properly when `boot_targets` only specifies the uclass of the device to boot from (e.g. "mmc" rather than "mmc0").

jmbaur force-pushed the uboot-2023.10 branch from 4d2e0c3 to 6a12107 Compare October 9, 2023 01:07

jmbaur closed this Oct 20, 2023

sorki mentioned this pull request Oct 20, 2023

uboot 2023.10, timeout handling, tests #261775

Closed

12 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

u-boot: 2023.07.02 -> 2023.10 #258701

u-boot: 2023.07.02 -> 2023.10 #258701

jmbaur commented Oct 3, 2023 •

edited

Loading

jmbaur commented Oct 3, 2023

jmbaur commented Oct 3, 2023

samueldr left a comment

jmbaur commented Oct 4, 2023 •

edited

Loading

lopsided98 commented Oct 4, 2023

lopsided98 commented Oct 4, 2023

jmbaur commented Oct 4, 2023

lopsided98 commented Oct 5, 2023 •

edited

Loading

samueldr commented Oct 5, 2023

lopsided98 commented Oct 6, 2023

lopsided98 commented Oct 6, 2023

jmbaur commented Oct 7, 2023 •

edited

Loading

lopsided98 commented Oct 7, 2023

lopsided98 commented Oct 8, 2023

jmbaur commented Oct 9, 2023

sorki commented Oct 18, 2023

jmbaur commented Oct 20, 2023

jmbaur commented Oct 20, 2023

sorki commented Oct 20, 2023

u-boot: 2023.07.02 -> 2023.10 #258701

u-boot: 2023.07.02 -> 2023.10 #258701

Conversation

jmbaur commented Oct 3, 2023 • edited Loading

Description of changes

Things done

jmbaur commented Oct 3, 2023

jmbaur commented Oct 3, 2023

samueldr left a comment

Choose a reason for hiding this comment

jmbaur commented Oct 4, 2023 • edited Loading

lopsided98 commented Oct 4, 2023

lopsided98 commented Oct 4, 2023

jmbaur commented Oct 4, 2023

lopsided98 commented Oct 5, 2023 • edited Loading

samueldr commented Oct 5, 2023

lopsided98 commented Oct 6, 2023

lopsided98 commented Oct 6, 2023

jmbaur commented Oct 7, 2023 • edited Loading

lopsided98 commented Oct 7, 2023

lopsided98 commented Oct 8, 2023

jmbaur commented Oct 9, 2023

sorki commented Oct 18, 2023

jmbaur commented Oct 20, 2023

jmbaur commented Oct 20, 2023

sorki commented Oct 20, 2023

jmbaur commented Oct 3, 2023 •

edited

Loading

jmbaur commented Oct 4, 2023 •

edited

Loading

lopsided98 commented Oct 5, 2023 •

edited

Loading

jmbaur commented Oct 7, 2023 •

edited

Loading