#
#asahi-dev · 2024-06-19 09:40
kettenisSo implementing a UCLASS_RNG driver for SEP might be worth the effort
#asahi-dev · 2024-06-19 09:36
kettenisImplementing EFI_RNG_PROTOCOL would be good since it would allow the Linux kernel efistub to randomize the physical address of the kernel
#asahi-dev · 2024-06-19 09:33
ketteniszzywysm: kaslr-seed is set by m1n1 and makes it through to the kernel
#asahi-dev · 2024-06-19 00:54
zzywysmkettenis: nvm, looks like that stuff hasn't been wired up yet for Apple Silicon, which is likely why it's disabled
#asahi-dev · 2024-06-19 00:40
zzywysmkettenis: it looks like the u-boot apple_m1_defconfig doesn't enable BOARD_RNG_SEED. does that mean that the entropy chunk that m1n1 grabs from sep_get_random() doesn't make it to the linux kernel?
#asahi-dev · 2024-06-17 22:03
allmazzhey! i use macbook pro 14 m2 max, and i see that battery life on asahi much worse than on macos. are there any workarounds?
#asahi-dev · 2024-06-16 08:17
chadmeddoing uclampset improves performance a little bit but its nowhere near as massive of a difference as it was last time
#asahi-dev · 2024-06-16 08:17
chadmedthe latter, we havent wired up any uclamp/taskset stuff in mesa yet
#asahi-dev · 2024-06-16 08:02
LarstiQis that tasksetting now already done without manual intervention, or performance otherwise got so good it doesn't matter?
#asahi-dev · 2024-06-16 06:42
chadmedperformance has increased significantly and any difference between letting the scheduler do whatever it wants and tasksetting the jobs to the pcores has disappeared entirely
#asahi-dev · 2024-06-16 06:42
chadmedinteresting results repeating my glmark2 schedutil testing 2 years later
#asahi-dev · 2024-06-12 16:44
amarioguyjannau: ahh i see
#asahi-dev · 2024-06-12 11:24
jannauamarioguy:
#asahi-dev · 2024-06-12 11:24
jannaudevices are in "/soc@200000000/ and "/soc@2200000000/" on Ultra
#asahi-dev · 2024-06-12 11:22
jannauthe linux device tree omits 'die0' for things on the first die (for compatibility with the max)
#asahi-dev · 2024-06-12 11:20
jannaufdt_path_offset("dcp") without leading '/' will find what "/aliases/dcp" resolves to
#asahi-dev · 2024-06-12 11:09
amarioguyah didn't notice that one, thanks
#asahi-dev · 2024-06-12 11:08
svenfdt_get_alias
#asahi-dev · 2024-06-12 11:07
amarioguyso it's not just peripherals that are on multiple dies)
#asahi-dev · 2024-06-12 11:07
amarioguyactually had one more question, when i try to use libfdt (don't have the open firmware abstractions in my firmware the way that linux/u-boot do), seems like using aliases (example being /die0/pinctrl_ap for example, trying to use the die0 alias to refer to the first die) don't work whatsoever (fdt_path_offset returns that the paths with aliases are not found, /soc/aic doesn't work either
#asahi-dev · 2024-06-11 07:08
chadmednot that i think it would stop some of them from doing it, but at least its in writing that i said not to
#asahi-dev · 2024-06-11 07:07
chadmedi got dirty looks from some of the sales guys today because i mentioned in passing that i had banned my team from using it for anything requiring putting customer data into it
#asahi-dev · 2024-06-11 07:05
nicolas17I couldn't get it to work by the deadline, it was too unmaintainable
#asahi-dev · 2024-06-11 07:04
nicolas17a friend asked me for help updating a mini game for a conference stand, and I tried but my main roadblock was that the existing code was written with significant "help" from Copilot/ChatGPT, and it *showed*
#asahi-dev · 2024-06-11 07:02
chadmedlike just fuck off with this stuff already
#asahi-dev · 2024-06-11 07:02
chadmedlike i had to explain to a guy last week why simply copying and pasting (yes literally ctrl c and ctrl v) 40,000 rows from a spreadsheet into the web version of chatgpt wouldnt "proactively identify trends before they arise"
#asahi-dev · 2024-06-11 07:01
chadmedtrue but i work in the part of IT where the "IT professionals" clap like brain damaged seals at the mere mention of AI or ${thing}GPT so im just sick to death of it being everywhere and had hope that apple would just not entertain this bs
#asahi-dev · 2024-06-11 06:59
nicolas17tbf Siri sends more data to Google in current versions than it will send to OpenAI in the new version
#asahi-dev · 2024-06-11 06:57
nicolas17:p
#asahi-dev · 2024-06-11 06:57
chadmedi dont want their leaked ipsw if its going to have openai stankware in it now
#asahi-dev · 2024-06-11 06:55
nicolas17best theory is that they can stop us from decrypting leaked ipsws
#asahi-dev · 2024-06-11 06:55
jannaumacos 15 supports nested virtualization and hw accelerated GIC: https://mastodon.social/@never_released/112596475468895974
#asahi-dev · 2024-06-11 06:54
chadmedsmells like product managers
#asahi-dev · 2024-06-11 06:53
nicolas17fortunately asahi doesn't care about the rootfs dmg
#asahi-dev · 2024-06-11 06:52
nicolas17x_x
#asahi-dev · 2024-06-11 06:52
nicolas17they're shipping ipsws where the rootfs is in a .dmg.aea
#asahi-dev · 2024-06-11 06:52
nicolas17yes
#asahi-dev · 2024-06-11 06:52
chadmedtheyre still shipping ipsws though right
#asahi-dev · 2024-06-11 06:51
nicolas17.aea
#asahi-dev · 2024-06-11 06:51
nicolas17in macOS 15 (and iOS 18 etc), the *outer* zip is now an *encrypted* applearchive
#asahi-dev · 2024-06-11 06:50
nicolas17chadmed: OTAs in macOS 14 were zips, containing payload.nnn files, which are lzfse-compressed AppleArchive
#asahi-dev · 2024-06-11 06:48
chadmedAppleArchive being lzfse?
#asahi-dev · 2024-06-11 05:44
nicolas17well fyi macOS 15 is now distributing OTAs as AppleArchive instead of zip, so all this stuff won't work anymore for future versions
#asahi-dev · 2024-06-11 05:40
jannauiirc it was pointless since apple's "fix" in 13.6* was some bug avoidance which wouldn't have helped us
#asahi-dev · 2024-06-11 05:38
nicolas17and that ended up not working? I can't find anywhere an ota zip is actually used
#asahi-dev · 2024-06-11 05:37
jannaunicolas17: trying to avoid the dcp non-promotion init bug by installing 13.6.x for which no ipsw exists
#asahi-dev · 2024-06-11 05:06
nicolas17installing boot components from an OTA zip instead of ipsw
#asahi-dev · 2024-06-11 05:05
nicolas17https://github.com/AsahiLinux/asahi-installer/commit/2c5ab624030b721ec4910047e71c24da0c1e13a0 what was this for? (I can't remember)
#asahi-dev · 2024-06-09 19:43
amarioguyjannau: ahh i see, how do you verify that the link is properly up?
#asahi-dev · 2024-06-09 19:05
jannaualso pcie port setup is somewhat fickle. under the HV I sometimes have issues getting the link to the aquantia nic up

plus ›