Re: Question: Might loader.kboot use be appropriate for booting RPi5's without U-Boot? (as an example)
- Reply: Warner Losh : "Re: Question: Might loader.kboot use be appropriate for booting RPi5's without U-Boot? (as an example)"
- Reply: Mark Millard : "Re: Question: Might loader.kboot use be appropriate for booting RPi5's without U-Boot? (as an example)"
- In reply to: Warner Losh : "Re: Question: Might loader.kboot use be appropriate for booting RPi5's without U-Boot? (as an example)"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Sun, 13 Sep 2026 00:14:29 UTC
I totally misunderstood the loader.kboot build. It was probably wishful thinking, and I think maybe loader.kboot might not be the best way to get to a good official RPi5 FreeBSD binary distribution. What I think I want is RPI5 Firmware \--> FreeBSD loader hiding in a bootable zipped Linux kernel image file \--> FreeBSD. What I think needs to be done is - figuring out how to build a FreeBSD loader image that RPi5 Firmware will happily try to boot - figuring out how to get that loader to find, and load the DT setup that RPi5 firmware provides, apply additional DT changes as needed - fixing any remaining bugs from the memory map if that didn't solve Mark's kernel boot error - working through remaining issues with FreeBSD FDT kernel until we can use critical path devices like USB and PCIe block devices and console (and probably also the thermal/fan) reliably - fixing remaining FDT device driver issues for all the other drivers. I think EDK2 stuff is cool, and it was certainly a critical enabler, but I suspect everyone would be happier with one less thing that can break. I think I want to try making loader do what's needed. On Saturday, September 12th, 2026 at 6:38 PM, Warner Losh <imp@bsdimp.com> wrote: > On Sat, Sep 12, 2026 at 4:23 PM Mark Millard <marklmi@yahoo.com> wrote: > >> On 9/11/26 10:21, Jeremy McMillan wrote: >>> >>> >>> I was planning to crack this nut open this weekend with some research on the current state of loader.kboot and smoke testing the RPi5 boot targeting loader.kboot. >>> >>> I see this as a stepping stone to proper supportable FDT kernel on RPi5, and also ideas about how to best manage custom overlays. Eventually, I want to get to a point where I can plug an I2C ADC chip module into RPi5, put together an overlay, and get a usable audio input device when I boot FreeBSD, for example. >>> >>> So, if I may follow up Mark's question, what stumbling blocks am I likely to hit first, and is there any advice inuring from past struggles that might help me out? >> >> If I gather from Warner's reply to my note correctly, the >> sequencing/flow issues would be something like (for example): >> >> RPi5 Firmware -> Host Linux Kernel -> loader.kboot -> FreeBSD kernel >> >> would not populate the information that the FreeBSD kernel needs. >> >> RPi5 Firmware -> EDK2 (UEFI/DeviceTree style) -> Host Linux Kernel -> >> loader.kboot -> FreeBSD kernel >> >> It would populate the information but the FreeBSD kernel then has the >> basically same problems it has for the original: >> >> RPi5 Firmware -> EDK2 (UEFI/DeviceTree style) -> FreeBSD kernel >> >> Thus, involving Host Linux Kernel -> loader.kboot is to no advantage for >> this type of context. >> >> Also, there would be no point for inserting such into: >> >> RPi5 Firmware -> EDK2 (UEFI/ACPI style) -> FreeBSD kernel > > Yes. LinuxBoot is for when you only have a linux kernel to launch from, > and the vast majority of those cases you have a tiny sliver of EDK2 > or similar that then loads the linux kernel as its first device driver > and takes over from there (so the board integrator doesn't have to > write two versions of board integration software, just one: Linux). > >> even if the insertion worked: no advantage to the insertion. > > I'm not seeing one for this use case, but who knows... If you really > wanted, you would have to investigate whatever interface Linux uses > to boot early and put that directly into the FreeBSD kernel, but then > you lose all the /boot/loader functionality. > > Warner > >> (EDK2 could be an over specific example for the general structure.) >> >>> --- >>> JTM >>> >>> -------- Original Message -------- >>> On Friday, 09/11/26 at 11:39 Mark Millard <marklmi@yahoo.com> wrote: >>> Hello Warner, >>> >>> This is just an informational query for my background knowledge. >>> >>> Note: I'm not dealing here with the FreeBSD kernel having support for >>> RPi5 DeviceTree content or the RP1 based structure or such. >>> >>> Context: >>> Various Linux distributions boot the RPi* families that they support >>> directly, no U-Boot or EDK2 or the like, including booting the RPi5 that >>> way. For the RPi5, they support, for example, USB booting and such that >>> U-Boot still does not support. >>> >>> That forms an example of a more general question of if loader.kboot and >>> such a linux kernel would be a viable alternative to the normal FreeBSD >>> U-Boot port style of support for aarch64? >>> >>> (As stands the RPi5 draft EDK2's provide a DeviceTree mode alternative >>> to ACPI mode for Linux booting. The EDK2's are not as limited as U-Boot >>> is for what is supported for booting [so far].) >>> >>> -- >>> === >>> Mark Millard >>> marklmi at yahoo.com >>> >>> >>> >>> >>> >> >> -- >> === >> Mark Millard >> marklmi at yahoo.com