#asahi-dev · 2026-04-11 11:09
jannauit was fine yesterday when testing on fedora under the hypervisor. on gentoo I experienced stutter with 2 different bt audio devices initially. It's fine now for minutes though.#asahi-dev · 2026-04-11 10:59
jannauchaos_princess: did you still got occasional bt audio artifacts during scanning? I initially got some but after reconnecting it's fine for minutes#asahi-dev · 2026-04-10 22:50
jannauI've tagged and pushed asahi-6.19.11-3 to github. Only change is bt audio stream priority which fixes audio stuttering, new config CONFIG_BT_BRCMEXT=y#asahi-dev · 2026-04-10 22:06
jannau+0x2000 looks to be wdt's raw counter, it appears to be read-only#asahi-dev · 2026-04-10 20:34
jannauI've seen it at 0xb1f..... just before reboot. reset either through a reset signal or power-gating#asahi-dev · 2026-04-10 20:21
jannaupmu.info-fault_log is all zero even after an wdt timeout. possible that iboot resets that#asahi-dev · 2026-04-10 20:15
jannaureboot is ~125 seconds after reset (assuming wdt.reg[0] + 0x2000 resets to 0)#asahi-dev · 2026-04-10 19:39
jannauto prove that it isn't the watchdog you could do wdt_reboot() but keep WDT_CTL as zero#asahi-dev · 2026-04-10 19:37
jannauif 0x1c is not 0 then the wdt_disable() either isn't done or does not work#asahi-dev · 2026-04-10 16:20
jannausee https://paste.debian.net/hidden/7c8d4a7e for monitoring, use `mon.poll()` from the shell for updates#asahi-dev · 2026-04-10 16:18
jannauwdt-version = 2 is constant since m1, compatible is [wdt,t8101, wdt,s5l8960x] on m1, m2, a18 pro#asahi-dev · 2026-04-10 16:11
jannauon a18 pro#asahi-dev · 2026-04-10 16:06
jannauwdt.c doesn't use the compatible but the path "/arm-io/wdt" and the compatible hasn't changed. used regmonitor to watch its mmio and as far as I can see all watchdogs are disabled#asahi-dev · 2026-04-10 16:01
jannauyes, m1n1 disables the watchdog and reenables it in kboot#asahi-dev · 2026-04-10 15:58
jannauis it still rebooting aftyer 2 minutes? I don't think this is a problem on m1/m2. m1n1 can sit there idle in proxy mode for days without resetting#asahi-dev · 2026-04-10 15:50
jannauyou can use proxyclient/tools/reset_panic_counter.py after each reset to make it less likely to run into it#asahi-dev · 2026-04-10 15:49
jannauyuka: I don't think it's possible without kvm. you have to either reset the startup disk or enter recovery to reset the panic count#asahi-dev · 2026-04-07 12:57
jannaus/clocks/power/ in main#asahi-dev · 2026-04-07 12:46
jannauchaos_princess: do you have a user space change for the brcm channel priority change? I'd expect that'll be one of the first questions. User space facing changes are expected to refer to user space implementation#asahi-dev · 2026-04-05 13:38
jannauah, copied from the wrong tab#asahi-dev · 2026-04-05 12:59
jannaujemalloc changed the default page size to 64k on aarch64#asahi-dev · 2026-04-05 12:59
jannauhttps://github.com/facebook/jemalloc/commit/e70a50f9e425c755c96857e9bba0b2879aada896#asahi-dev · 2026-04-03 17:22
jannauand still strange where the old size is coming from#asahi-dev · 2026-04-03 17:20
jannaubut I guess it should be enough that io_addr_assert() is#asahi-dev · 2026-04-03 17:18
jannaunoopwafel: iomap_sized in rust/kernel/io/mem.rs isn't inlined#asahi-dev · 2026-04-03 17:14
jannau1048576 was the old PMP_MMIO_SIZE before I reduced it to 0x80000 for m2#asahi-dev · 2026-04-03 17:12
jannauthat was only my 1.92 test build because I forgot `rustup component add rust-src` asfter switching to it#asahi-dev · 2026-04-03 17:08
jannauargh, I'm building without PMP#asahi-dev · 2026-04-03 17:05
jannauyes, sounds more like a missing dependency or other build config issue#asahi-dev · 2026-04-03 17:01
jannaubuilds here with 1.85.1 and 1.94#asahi-dev · 2026-04-03 16:58
jannauwhich rustc?#asahi-dev · 2026-04-03 16:53
jannauyes#asahi-dev · 2026-04-03 16:53
jannausomewhat usable on a M3 macbook pro#asahi-dev · 2026-04-03 16:41
jannauI've tagged asahi-6.19.11-1 with additional M3 support (NVMe, PCIe, smc (input/reboot/rtc)), mtp)#asahi-dev · 2026-04-03 12:44
jannauI do not need the spmi patches for the 14-inch mbp. that still uses i2c cd321x#asahi-dev · 2026-04-03 12:34
jannauok, I'll skip it for now then. Incompatible in just a few details or something requiring larger changes#asahi-dev · 2026-04-03 12:24
jannaudo you have a devicetree with atc/dwc nodes#asahi-dev · 2026-04-03 12:23
jannaunot just predicable but can be derived from the adt assuming the per port registers use the same pattern and/or are identical to t6020#asahi-dev · 2026-04-03 12:20
jannauthanks#asahi-dev · 2026-04-03 12:17
jannaua M3 imac would be handy to test additional pcie ports#asahi-dev · 2026-04-03 12:16
jannauI'll test as well but I'm currently still wrangling other bits#asahi-dev · 2026-04-03 12:12
jannauyes, refclk req/ack uses the per port phyX entries. so I think 0x59E00C000 is correct#asahi-dev · 2026-04-03 11:55
jannauIt looks to me as if t602x has the same mmio layout#asahi-dev · 2026-04-03 11:54
jannaufrom what I deduced from t602x it's two reg entries after portX reg entry#asahi-dev · 2026-04-03 11:52
jannauno, that is the per port phy from the linux dt#asahi-dev · 2026-04-03 11:35
jannaunoopwafel: are you sure about the phy0 reg value for t8122's pcie node? I think it should be 0x59E00C000 based on t6020#asahi-dev · 2026-03-31 19:52
jannaumaz: same as m4? as in I would be very surprised if not#asahi-dev · 2026-03-30 20:04
jannaunoopwafel: thanks, looks much better now. I think there's one small copy-n-paste error#asahi-dev · 2026-03-30 10:24
jannauwe should start preparing the m1n1 release. I think we should get the pmgr in first, m3 pcie would be nice#asahi-dev · 2026-03-30 10:21
jannauit is not. the upstream patches will be based on starlight