#
✕ nick:nickchan
#asahi-dev · 2024-08-31 11:47
nickchanand then the iPod touch 7 that can only go up to 1.6 GHz
#asahi-dev · 2024-08-31 11:46
nickchana10x goes up to 2.4 GHz
#asahi-dev · 2024-08-31 11:46
nickchanthere's the normal a10 that goes up to 2.34 GHz
#asahi-dev · 2024-08-31 11:46
nickchana10(x) is a mess of cpufreq
#asahi-dev · 2024-08-31 11:45
nickchancpufreq (at least on A10) and wdt is exactly the same
#asahi-dev · 2024-08-31 11:45
nickchanbut that also does not apply to Apple TV
#asahi-dev · 2024-08-31 11:44
nickchanthe display on these devices are more similar to the z2 touchscreen stuff for the touchbar mac
#asahi-dev · 2024-08-31 11:44
nickchanDcp doesn't exist until A14
#asahi-dev · 2024-08-31 11:44
nickchanthough we did manage to test it on that
#asahi-dev · 2024-08-31 11:43
nickchansince the USB exploit can only be performed with the USB pins in the ethernet port
#asahi-dev · 2024-08-31 11:43
nickchanApple TV are also tested though Apple TV 4K is most annoying to actually use Linux with
#asahi-dev · 2024-08-31 11:41
nickchanBut for example the cellular A9X iPad Pro is not tested while the wifi one is
#asahi-dev · 2024-08-31 11:41
nickchaneg: A8 iPhone, A9X iPad Pro, A10 iPad
#asahi-dev · 2024-08-31 11:41
nickchannow because of the ridiclous amount of SKUs here - not every other SKU is actually tested, but every SoC/device type combination is tested
#asahi-dev · 2024-08-31 11:38
nickchan- the HomePod
#asahi-dev · 2024-08-31 11:38
nickchan- All T2 iBridges
#asahi-dev · 2024-08-31 11:38
nickchanfor the devicetree part it will exclude some SKUs due to testing issues:
#asahi-dev · 2024-08-31 11:37
nickchanit is not that important for the serial0 debug console but it would be problematic for the other uarts, so I would like to fix that first
#asahi-dev · 2024-08-31 11:36
nickchanon A7-A9 uart rx is partially broken, it only happens in chunks of 16 bytes
#asahi-dev · 2024-08-31 11:36
nickchanbesides the SError issue which already have a fix
#asahi-dev · 2024-08-31 11:35
nickchansven: by the way the devicetree patches might come a bit later since there's some additional troubles with the uart
#asahi-dev · 2024-08-30 10:34
nickchankettenis: the original SError is actually from trying to write to UTXH (+0x20)
#asahi-dev · 2024-08-30 09:42
nickchansven: would you mind checking if the above diff has any negative effects for early serial, serial in general on m1 and up?
#asahi-dev · 2024-08-30 09:39
nickchanInterestingly, this error only affects earlycon, normal serial is not affected but I changed the type of the normal serial for consistency
#asahi-dev · 2024-08-30 09:37
nickchanHowever, I do not know what effect this have on the M1 and up
#asahi-dev · 2024-08-30 09:37
nickchanSError on A7-A11 when trying to use serial earlycon can be fixed by changing the S5L port iotype to UPIO_MEM32: https://paste.debian.net/hidden/6cc79493/
#asahi-dev · 2024-08-28 21:35
nickchanhttps://github.com/asdfugil/linux-apple/commit/274eb4cb0b3780efc97d4e4529db3767c17eda98 - this commit adds the logic to handle the "apple,t8015-aic", global fipi only case
#asahi-dev · 2024-08-28 21:35
nickchanadditional reference points: https://github.com/asdfugil/linux-apple/commit/ee54fcfdf746ee7e7e0077db9cfeafd09dee43ef - this commit makes it so that IPI registers are not used on compatibles that doesn't actually support it
#asahi-dev · 2024-08-28 21:31
nickchanAlso, each feature level is strictly an extension of lower feature levels, that is, all "apple,t8103-aic" is also "apple,t8015-aic", "apple,aic", and all "apple,t8015-aic" is also "apple,aic"
#asahi-dev · 2024-08-28 21:25
nickchanProbably should also describe the feature levels in the description field as well...
#asahi-dev · 2024-08-28 21:25
nickchanhttps://github.com/asdfugil/linux-apple/commit/6c4d6934cbb0dd2987ed492d88940216fae333c1#diff-8035e63fcfb26a3068f4c59a9af0e278fa029ce8b69f7cef61e0897d00020128R93 here's my attempt at writing the bindings, not sure if it's right
#asahi-dev · 2024-08-28 21:24
nickchanthe driver will only recognize these compatibles, however in the device tree I plan on using a SoC-specific compatible iwith these compatibles as a fallback, like "apple,s5l8960x-aic"
#asahi-dev · 2024-08-28 21:23
nickchan"apple,t8103-aic" - Fast IPI support, global and local
#asahi-dev · 2024-08-28 21:23
nickchan"apple,t8015-aic" - Fast IPI support, global only
#asahi-dev · 2024-08-28 21:23
nickchan"apple,aic" - No fast IPI support
#asahi-dev · 2024-08-28 21:23
nickchanThere are three feature levels that needs to dealt with:
#asahi-dev · 2024-08-28 21:22
nickchanHi! I have been writing some dt-bindings for the AIC driver port to A7-A11, however I do not know if I got it right or not
#asahi-dev · 2024-08-25 10:03
nickchanunlike mac though we *can* keep rvbar of cpu0 unlocked
#asahi-dev · 2024-08-25 10:03
nickchanin the short terms this means no cpuidle in linux and no deep wfi in m1n1
#asahi-dev · 2024-08-25 10:02
nickchanwell, other than the normal wfi that every arm64 cpu has
#asahi-dev · 2024-08-25 10:02
nickchanin other words the "hypervisor would go nuts" behavior according to marcan is the only behavior here
#asahi-dev · 2024-08-25 09:58
nickchanaccroding to xnu src as well the first soc that supported state retention across deep wfi is a12
#asahi-dev · 2024-08-25 09:57
nickchani.e. after a deep wfi the cpu will always end up at rvbar
#asahi-dev · 2024-08-25 09:57
nickchanA7 - A11 does not support state retention across deep wfi
#asahi-dev · 2024-08-25 09:57
nickchanaccording to xnu src and experiments
#asahi-dev · 2024-08-25 09:18
nickchansome chicken bit is missing
#asahi-dev · 2024-08-25 09:18
nickchanperhaps
#asahi-dev · 2024-08-25 09:13
nickchanbiggest unsolved mystery so far is deep wfi making the cpu go to rvbar
#asahi-dev · 2024-08-25 08:50
nickchanjust don't even try to use the dwc2 stuffs
#asahi-dev · 2024-08-25 08:49
nickchanfor those who are interested code for the m1n1 port to a7-a11 is now available as well: https://github.com/asdfugil/m1n1-idevice

plus ›