On 04/06/2020 17:46, Stefano Stabellini wrote: > On Thu, 4 Jun 2020, André Przywara wrote: >> On 04/06/2020 17:24, Stefano Stabellini wrote: >>> On Thu, 4 Jun 2020, André Przywara wrote: >>>> On 04/06/2020 09:48, Julien Grall wrote: >>>> >>>> Hi, >>>> >>>>> On 03/06/2020 23:31, Stefano Stabellini wrote: >>>>>> Touching the watchdog is required to be able to reboot the board. >>>>> >>>>> In general the preferred method is PSCI. Does it mean RPI 4 doesn't >>>>> support PSCI at all? >>>> >>>> There is mainline Trusted Firmware (TF-A) support for the RPi4 for a few >>>> months now, which includes proper PSCI support (both for SMP bringup and >>>> system reset/shutdown). At least that should work, if not, it's a bug. >>>> An EDK-2 build for RPi4 bundles TF-A automatically, but you can use TF-A >>>> without it, with or without U-Boot: It works as a drop-in replacement >>>> for armstub.bin. Instruction for building it (one line!) are here: >>>> https://git.trustedfirmware.org/TF-A/trusted-firmware-a.git/tree/docs/plat/rpi4.rst >>>> >>>>>> >>>>>> The implementation is based on >>>>>> drivers/watchdog/bcm2835_wdt.c:__bcm2835_restart in Linux. >>>>> >>>>> Can you give the baseline? This would allow us to track an issue and >>>>> port them. >>>> >>>> Given the above I don't think it's a good idea to add extra platform >>>> specific code to Xen. >>> >>> The RPi4, at least the one I have, doesn't come with any TF, and it >> >> My RPi4 didn't come with anything, actually ;-) It's just a matter of >> what you put in the uSD card slot. >> >>> doesn't come with PSCI in device tree. >> >> TF-A patches the PSCI nodes in: >> https://git.trustedfirmware.org/TF-A/trusted-firmware-a.git/commit/plat/rpi/rpi4?id=f67fa69cb6937a7fc559bbec4a7acce5edefa888 >> >>> As a user, I would rather have >>> this patch (even downstream) than having to introduce TF in my build and >>> deployment just to be able to reboot. >> >> I get your point, but would rather put more pressure on people using >> TF-A. After all you run without CPU hotplug, A72 errata workarounds and >> without Spectre/Meltdown fixes. What was the IP address of your board >> again? ;-) > > Please send a pull request to remove __bcm2835_restart from the Linux > kernel, once it is removed from there I'd be happy to take it away from > Xen too ;-)
The kernel needs to support all RPi models, so we definitely need this. Also it's already in there, so removing it is more churn. The reason I am bringing this up is that we should get away from those platform specific files in Xen at all. The only reason we have it for the RPi4 is the non-page-aligned MMIO regions and overlaps, which could actually be determined much better at runtime ... > I know I am being cheeky but we have enough battles to fight and enough > problems with Xen -- I don't think we should use the hypervisor as a > leverage to get people to use or upgrade TF. We just need to get good > functionalities to our users with the less amount of friction possible. As I said: it's not my call, just pointing that out. It's just sad that people everywhere work around the limited firmware instead of doing it properly. > Everything you mentioned are good reason to use TF, and this patch does > not take anything away from it. My suggestion would be to work with > raspberrypi.org to have TF installed by default by the Raspberry Pi > Imager. As far as I know there are (were?) efforts underway. For years ;-) Cheers, Andre
