Re: [BUG] rockchip: v2026.07 fails to boot on GameForce Ace (RK3588S)
Daniel Martin <[email protected]>
| Newsgroups | org.u-boot-project.lists.u-boot |
|---|---|
| Message-ID | <CAOyTmd0QrejH5tR013FDOpMJu27UXf2BE1t87f9Skxm9uyLC8Q@mail.gmail.com> |
That's great, thanks for the update Chris. So maybe I'm missing something... I didn't try `Arm Trusted Firmware` but I used rk3588_bl31_v1.54.elf without Optee. Does it need Optee? I assumed not as other RK3588 devices don't. Let me try again on the weekend with ATF & LInux 7.2. Dan On Tue, 18 Aug 2026 at 03:34, Chris Morgan <[email protected]> wrote: > > On Fri, Aug 14, 2026 at 10:02:05PM -0500, Chris Morgan wrote: > > On Sat, Aug 15, 2026 at 10:06:23AM +1000, Daniel Martin wrote: > > > Thanks for the further details Chris, > > > > > > It would indeed be great if we could move this device to mainline end-to-end. > > > Let me know how you get on with v2026.07 and if possible, test your > > > DTS with the stable kernel (7.1.8 or similar). > > > When I reverted uboot, I encountered inconsistent I/O issues with the > > > latest stable kernel, therefore I ended up going back to the BSP. > > > > > > Thanks for all your work, we appreciate the effort. > > > > > > Dan > > > > I've just tested booting the device with a completely unmodified > > version of Arm Trusted Firmware (checked out at tag v2.15.0), U-Boot > > (checked out at tag v2026.07 and changing nothing from the > > gameforce-ace-rk3588s_defconfig), loading Linux 7.2 (tag v7.2-rc7) via > > extlinux boot method and it seems to be working. I do note that it > > seems to be slower to boot since I upgraded A-TF and U-Boot to the most > > recent releases, so I'll troubleshoot that some more next week as I try > > to install a fully up to date Ubuntu 26.04 on it. > > > > I know you said it's hard to get a serial port on it, but if you > > continue to have problems booting it that might be required to > > troubleshoot it further. Otherwise I can send you my binaries if > > need be. > > > > Thank you, > > Chris > > I'm afraid I cannot seem to recreate the problem, and I'm going to need > a serial dump of the kernel logs. I'm sorry to ask that of you, but it > will help further debug what the issue may be on your end. So far I've > tested using a mainline kernel (v7.2 checked out this morning), U-Boot > v2026.07, Arm Trusted Firmware (mainline master, checked out this > morning OR the rkbin version, rk3588_bl31_v1.54.elf), optionally with > Optee (master, checked out this morning). In each case the device boots > successfully for me running Debian 13 (Forky) with my own mainline > kernel. > > I'm going to test installing Ubuntu 26.04 and fusing some OTP keys > later today, I'll let you know if I encounter further issues. > > Thank you, > Chris > > > > > > > > > > > > On Sat, 15 Aug 2026 at 07:40, Christopher Morgan > > > <[email protected]> wrote: > > > > > > > > I'm having trouble with my usual email client, so I'm sorry if this message is formatted weird. > > > > > > > > I have tested the device most recently with 2026.04 with the default defconfig options. I am using mainline Arm Trusted Firmware though, not the Rockchip versions. Using Rockchip's A-TF should work, but again it's not something I've tested. I can note that if you're using mainline A-TF along with mainline Optee, you need to pass CONFIG_SPL_ATF_NO_PLATFORM_PARAM=N, otherwise the device will fail to boot because it won't protect the Optee regions. (For mainline A-TF CONFIG_SPL_ATF_NO_PLATFORM_PARAM is optional unless you use Optee, for Rockchip's BL31 you need CONFIG_SPL_ATF_NO_PLATFORM_PARAM=Y). > > > > > > > > I'm going to test the device next week and see if I encounter any issues with it. I'll report back if I find out anything weird. > > > > > > > > Thank you, > > > > Chris > > > > ________________________________ > > > > From: Quentin Schulz <[email protected]> > > > > Sent: Wednesday, August 12, 2026 10:22 AM > > > > To: [email protected] <[email protected]>; [email protected] <[email protected]> > > > > Cc: Chris Morgan <[email protected]> > > > > Subject: Re: [BUG] rockchip: v2026.07 fails to boot on GameForce Ace (RK3588S) > > > > > > > > +Cc Chris who maintains this board and may be able to help. > > > > > > > > I somehow didn't receive this mail in my inbox, here's my attempt at > > > > quoting text from a mail I didn't receive (and it's also not available > > > > on lore.kernel.org...). > > > > > > > > > Hello, > > > > > > > > > > I am experiencing a total boot failure on the GameForce Ace handheld > > > > > console (Rockchip RK3588S) when trying to run mainline U-Boot > > > > > v2026.07. > > > > > The device fails to initialize or display any output during the > > > > > initial handshake. > > > > > Unfortunately, due to the tightly integrated form factor of this > > > > > handheld device, it is exceptionally difficult to break open the > > > > > casing and wire up a UART serial console to capture early boot logs. > > > > > The target operating system (Batocera Linux) environment is running > > > > > the mainline Linux 7.1.5 kernel. > > > > > Given the complete lack of visible console output, I wanted to check: > > > > > > > > > > 1. Has the GameForce Ace board target been actively tested with the > > > > > v2026.07 release? > > > > > 2. Are there any specific or updated build instructions, BL31/TPL/SPL > > > > > combinations, or proprietary Rockchip firmware binary requirements > > > > > (rkbin) needed for this release that may differ from previous > > > > > versions? > > > > > > > > > > For reference, the build configuration used was: gameforce-ace-> > > > > rk3588s_defconfig > > > > > TF-A / BL31 version: rk3588_ddr_lp4_2112MHz_lp5_2400MHz_v1.21.bin / > > > > > rk3588_bl31_v1.54.elf > > > > > Compiler: GCC 14.3 > > > > > Linux 7.1.5 uses the rk3588s-gameforce-ace.dts > > > > > > > > > > Any guidance on known regressions regarding this target, or > > > > > recommendations on how to isolate the stage where it hangs without a > > > > > hardware UART, would be greatly appreciated. > > > > > > > > > > > > > If you have a known working version of U-Boot, maybe start a bisect with > > > > that commit as good and 2026.07 as bad to isolate a commit that broke > > > > the board. > > > > > > > > Cheers, > > > > Quentin > > > > > > > > > > > > -- > > > > > > Kind Regards, > > > > > > Daniel -- Kind Regards, Daniel