#
#asahi-dev · 2026-01-06 16:08
fl0_idthx for all the hints
#asahi-dev · 2026-01-06 16:08
fl0_idbut it had debug ingo
#asahi-dev · 2026-01-06 16:06
fl0_idjannau oops yeah it was compiled as efi stub ... didn't even think about that
#asahi-dev · 2026-01-06 16:05
fl0_idyeah the incremental upgrade part is why i compiled it in a linux vm on the m3, then copied to the x230...
#asahi-dev · 2026-01-06 16:04
fl0_idgood things to keep in mind, thx
#asahi-dev · 2026-01-06 16:04
nickchanunless you want to spend half an eternity and a lot of ram linking
#asahi-dev · 2026-01-06 16:04
nickchanand don't use lto
#asahi-dev · 2026-01-06 16:03
fl0_idjannau tbh I don't remember, good to check. I think I just mostly used the config linked somewhere in the docs
#asahi-dev · 2026-01-06 16:03
jannauccache could help as well
#asahi-dev · 2026-01-06 16:03
jannaualso you need to do it only once otherwise it should be most incremental compilation of 1-2 object
#asahi-dev · 2026-01-06 16:02
jannauyes. do you compile the kernel with DEBUG_INFO? and not as EFI_STUB?
#asahi-dev · 2026-01-06 16:02
svenstart with a minimal .config, you really don't need much for early bringup
#asahi-dev · 2026-01-06 16:02
fl0_idputting uart prints seems like a good way, just need to think about compiling, as my x230 rly doesn't like that / takes forever
#asahi-dev · 2026-01-06 16:00
fl0_idright. just to confirm - there should have been like actual symbol names, not just kernel_size_le_lo32 etc, and at that point it should already be past start_kernel, right?
#asahi-dev · 2026-01-06 15:59
fl0_idhe suggested adding an address manually, which I will prob try next (like where m1n1 starts the kernel)
#asahi-dev · 2026-01-06 15:59
sven*anything else
#asahi-dev · 2026-01-06 15:59
svenanything is guessing and probably won't get you anywhere
#asahi-dev · 2026-01-06 15:59
nickchan> different memory configs behaved differently - I have seen a varying number of DCS channel used depending on how much mmeory installed
#asahi-dev · 2026-01-06 15:58
svenif you can't get breakpoints to work then that leaves the uart prints in kernel code
#asahi-dev · 2026-01-06 15:58
fl0_idand the bp was accepted but it never stopped
#asahi-dev · 2026-01-06 15:58
fl0_idjannau thx
#asahi-dev · 2026-01-06 15:58
nickchanhow about the number of DCS channel?
#asahi-dev · 2026-01-06 15:58
fl0_idright. sven that's what I was doing but I think symbols were not really loaded, as I told integralpilot, it just showed _kernel_size_le_lo32+someaddress or kimage_limit+address at the end
#asahi-dev · 2026-01-06 15:43
jannaunot sure how that could be an issue in linux though. maybe pmgr-misc but that would be much later
#asahi-dev · 2026-01-06 15:41
jannauno that that I can rememeber but m3 max uses different number of memory controllers for the 14 and 16 core versions
#asahi-dev · 2026-01-06 15:38
svenno, and you really want to either single step the early kernel code or add uart prints (or anything similar) to your kernel to figure out what's going on
#asahi-dev · 2026-01-06 15:33
fl0_idbtw, was there ever a thing where different memory configs behaved differently? Like if f.e. someone tested with 64 GB memory but 96 GB failed or sth
#asahi-dev · 2026-01-06 15:27
jannaupushed to asahi-wip
#asahi-dev · 2026-01-06 15:23
jannauyes, seems to work. that should be less objectionable than `for_each_child_of_node() { of_platform_device_create(); }
#asahi-dev · 2026-01-06 15:23
sveni also wonder who should review this, it's in soc/apple but i don't feel comfortable just picking it up without understanding if what that driver doing is correct
#asahi-dev · 2026-01-06 15:21
svenit should
#asahi-dev · 2026-01-06 15:21
jannauI wonder if that's enough for us as well as mtp_hid ahas a compatible
#asahi-dev · 2026-01-06 15:20
jannauthat doesn't explictly register platform devices (ignoring the pd subdevice) but uses devm_of_platform_populate
#asahi-dev · 2026-01-06 15:15
sveni meant drivers/platform/chrome/cros_ec.c
#asahi-dev · 2026-01-06 15:13
svener wait..
#asahi-dev · 2026-01-06 15:13
svenwas about to mention that since i just found it as well :)
#asahi-dev · 2026-01-06 15:13
jannaudrivers/platform/chrome/chromeos_of_hw_prober.c seems to be the best reference
#asahi-dev · 2026-01-06 15:09
svensounds good then
#asahi-dev · 2026-01-06 15:08
chaos_princessit may be not cros-ec, but i do the same thing in aop, and there was precedent
#asahi-dev · 2026-01-06 15:07
svendidn't find that example when i was looking
#asahi-dev · 2026-01-06 15:07
svenah, if there's precedence in cros-ec then it's probably fine
#asahi-dev · 2026-01-06 15:05
chaos_princessnot that many things hang off dockchannel for a new bus type to be worth it. and i think cros-ec does platform devices too
#asahi-dev · 2026-01-06 14:42
sveni also wonder if it's correct that we just create platform child devices in dockchannel, if you look at other things like e.g. spi or i2c they create their own bus type or e.g. mfd at least creates its own device_type
#asahi-dev · 2026-01-06 14:41
sventhere's a built-time dependency and then they can both just be taken through the hid tree in the end instead of waiting a cycle or creating immutable branches
#asahi-dev · 2026-01-06 14:40
svenI think i'd actually send dockchannel and dockchannel-hid together now that i'm looking at it
#asahi-dev · 2026-01-06 13:06
chadmedthe TCONs all support it
#asahi-dev · 2026-01-06 13:06
chadmedi wonder if DCP enables PSR for us on the builtin displays
#asahi-dev · 2026-01-06 11:07
chaos_princesshey, we RE-d some pretty complex hardware, we can RE what the reviewers want even if it doesn't really match how dt is being "marketed"
#asahi-dev · 2026-01-06 11:04
chadmed~the hardest thing to get right is ensuring the reviewer woke up on the right side of the bed today~
#asahi-dev · 2026-01-06 11:01
sventhese things are annoying to get right

plus ›