#
…

2021-08-23

00:14al3xtjames has quit [Quit: Ping timeout (120 seconds)]
00:14al3xtjames has joined #asahi-re
02:56PhilippvK has joined #asahi-re
02:59phiologe has quit [Ping timeout: 480 seconds]
03:10quarkyalice has quit [Remote host closed the connection]
03:12quarkyalice has joined #asahi-re
03:18quarkyalice has quit [Remote host closed the connection]
03:19quarkyalice has joined #asahi-re
03:52marcansven: yeah, I was thinking that DCP is probably the only case where those drivers are so broken out in macos, but I think for linux those things tend to be all monolithic
03:52marcanso we can probably just drive them all from a single DCP driver eventually
04:10yuyichao has joined #asahi-re
04:45quarkyalice_ has joined #asahi-re
04:47quarkyalice has quit [Ping timeout: 480 seconds]
05:21quarkyalice has joined #asahi-re
05:48besmirich[m] has joined #asahi-re
05:56svenmarcan: yeah, so if we can really just handle every single co-processor with a single driver i think that shared rtkit library you suggested makes sense
05:59marcanyeah, tbh I don't see modularity really helping us here
06:00marcanafter all whoever wants to run these things will need the whole driver anyway
06:00marcanand if the thing crashes the whole thing crashes
06:00marcantrying to split it up into modules... would just add overhead for no reason
06:00sveni briefly looked into remoteproc but that doesn't fit all that well either
06:00marcanyeah, different story
06:00marcanI took a good look at mailbox the other day and I don't like that we can't use the channel concept, but given this situation, it's not worth making changes to that
06:00svenso just a simple mailbox driver + rtkit shared library sounds good then
06:01marcanlet's just treat it as a single-channel mailbox
06:01marcanand deal with the endpoints on our rtkit stuff
06:01marcanthe rtkit lib would of course be a module to be shared by all the drivers
06:01sventhe whole channel concepts really seems to be more about a single mailbox block having multiple FIFOs
06:01svenwhich... we don't
06:01marcanyeah, exactly
06:02marcanthough it could be adapted to ours by adding a mode to mailbox where all channels share a busy status
06:02marcanbut... not worth it, I think
06:02marcangiven we're building monolithic drivers anyway
06:02svenyeah. the core needs a small change anyway unless we want to wait for an interrupt for every single message we sent
06:02svenbut that's a small thing
06:03marcanyeah that too, the queuing
06:03marcanit would need a flag to poll after a send
06:03marcanjust once
06:03marcanso that it can keep sending
06:03marcanbut it might not matter in practice for now
06:03marcanso we can also just defer that to whenever we run into perf issues
06:04marcanoh, also: the DT bindings need some thought. I want to make sure that e.g. if we decide the mailbox framework sucks one day, that we don't run into issues with the bindings
06:04marcanwhatever we decide one needs to be fairly future-proof
06:04svenyeah, i have two variants that i implemented: one just uses the poll mode as-as. the other one convinces it to keep sending unless it gets EBUSY and request the irq to continue then
06:05marcanok, I'm going to stream a bit today, what was on the nobody's doing it list? i2c and pinctrl was it?
06:05svenclocks, i2c. i think j_ey did some work on gpio/pinctrl but i dunno how far along he is
06:05marcanah, right, clocks
06:05marcansounds good
06:06marcanwe never figured out the clock dependency stuff, right? I guess I'll just do $something and then see what upstream says
06:06svenhttps://github.com/AsahiLinux/linux/commits/nvme/new first three commits or so are the variant that keeps sending messages and requests and irq once the FIFO is full fwiw
06:06marcando we have any tooling for extracting clock trees from the adt?
06:06sveni have some hack https://github.com/AsahiLinux/linux/commits/clk/dev there but that's not great
06:06svenand also not enough, because some clocks have *two* parents
06:07marcanah that was the instance per clock model, right?
06:07marcanand yeah
06:07sventhat was a single clock controller with a clocks property that can reference the controller itself
06:08svenwhich the maintainers would probably hate and which isn't enough anyway because the two parents thing
06:09marcanI think, realistically, this needs to be a DT structure
06:09marcanwith subnodes for clocks
06:09marcanone way or another
06:09marcanI'll poke around
06:11svenmarcan: pmgr.c is missing the second parent fwiw. i think it was just after the first parent in the structure at the top
06:12marcanit's max 2?
06:13svenhaven't seen anything with more than 2
06:13sventhere's still 2 unknown bytes so it could be three i guess
07:56x56 has quit [Quit: Ծ-Ծ]
08:00x56 has joined #asahi-re
09:18chadmed has quit [Remote host closed the connection]
09:19chadmed has joined #asahi-re
09:19chadmed has quit [Remote host closed the connection]
10:22chadmed has joined #asahi-re
12:31afwolfe has joined #asahi-re
12:33afwolfe has quit []
12:44yuyichao has quit [Ping timeout: 480 seconds]
13:12jbowen has joined #asahi-re
14:11chadmed has quit [Ping timeout: 480 seconds]
14:44jeffmiw has joined #asahi-re
15:18chadmed has joined #asahi-re
16:12phiologe has joined #asahi-re
16:15PhilippvK has quit [Ping timeout: 480 seconds]
17:11jbowen has quit [Read error: Connection reset by peer]
17:17jbowen has joined #asahi-re
18:27jeffmiw has quit [Ping timeout: 480 seconds]
18:56yrlf has quit [Quit: Ping timeout (120 seconds)]
18:57yrlf has joined #asahi-re
18:59jbowen has quit [Ping timeout: 480 seconds]
19:04jbowen has joined #asahi-re
19:08jeffmiw has joined #asahi-re
19:31shaman_br[m] has joined #asahi-re
19:42jbowen has quit [Ping timeout: 480 seconds]
20:00jbowen has joined #asahi-re
20:17yuyichao has joined #asahi-re
20:31yuyichao has quit [Ping timeout: 480 seconds]
20:33yuyichao has joined #asahi-re
20:39PhilippvK has joined #asahi-re
20:42phiologe has quit [Ping timeout: 480 seconds]
21:15jbowen has quit [Ping timeout: 480 seconds]
21:28j_ey has joined #asahi-re
21:28alyssaj_ey: the "S" in "SIO" is for "Smart"
21:29j_eyactually?
21:29j_eyso what does it actually handle? (the sio asc)
21:29alyssaI just like expanding acronyms
21:30jeffmiwis this expansion of the acronym based on some reverse eng ?
21:31alyssai'll leave that one up to your imagination
21:35alyssaj_ey: It might just be the DMA controller...?
21:36alyssawell, a DMA controller
21:37j_eyit does have a dma controller subnode..
21:39jeffmiwthis is what I was thinking of, the combination of sio & dart-sio to do some dma stuff from spi .... but no real facts I can use to confirm this speculation ...
21:39j_eytrace macOS a bit?
21:48jeffmiwthis is what I'm doing to try to understand and found those sio / dart-sio acesses, thus my initial questions if anyone knows anything about them
21:49jeffmiwI'll continue try to figure it out
21:50svensee #asahi. it at least maps the aes engine
21:52jbowen has joined #asahi-re
22:06jeffmiwsven: thx. what it seems to me is that the kbd/spi patches from amw including Corellium patches are not using the same mode of spi that macos is using, but I can be mistaken.
22:06svendma doesn't make much sense for a keyboard though
22:06svenmaybe for the nor flash, but even that one is quite small and probably not read/written often
22:08jbowen has quit [Quit: leaving]
22:08svenbut anyway, if they use DMA for SPI for some reason and if that co-processor is involved you will have to reverse engineer its protocol
22:15jeffmiwtouchpad ? yeah re. eng. is needed then. I will continue dig into it.
22:17svenit's rtkit, so you could try to use the asctracer (i think it's called like that) and see what messages are exchanged
22:18svenand if you trace the DART mappings as well you should be able to see what data it exhanges
22:20j_eyyou'll have to figure out extra messages it may do, whats in the asctracer currently is only what markan figured out for dcp
22:21svenit should handle syslog and the rtkit bootstrap stuff though
22:24j_eyheh, the thing that marcan fixed today was one of the things I hit, syslog with the buffer size c39fe2fbdea0a9eaf5c32b519d674820b46509a6
22:25jeffmiw has quit [Ping timeout: 480 seconds]
22:30alyssasven: speaking of aes, I wonder how much prior art there are for things like aes and jpeg which have s5l-era compatibles in the ADT
22:31svenaes is simple enough, i reversed that one back in match or so
22:31sveni was hoping to use it to understand DART better but couldn't get it to work and gave up after 30 minutes
22:32svenor, well, i could only get it work in bypass mode
22:33sventhere's a DAPF behind that DART that needs to be configured
22:33svenno idea what that is
22:33sveni just know that the AES DART wasn't able to even read the pagetable
22:33alyssaalso, any idea what NCO is?
22:33svennon-comissioned officer? :P
22:34alyssaI meant in the adt :-p
22:34alyssa compatible = [nco,t8101, nco,s5l8960x]
22:34alyssaI guess power management related
22:35svenAppleS5L8960XNCOFrequencyFunction
22:36alyssaHow enlightening :-p
22:36* alyssawonders if jpeg/scalar are used at all in macOS or are just legacy from old iphones
22:36alyssa(too much math)
23:19skoobasteeve_ has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
23:19skoobasteeve has joined #asahi-re
23:24skoobasteeve has quit []
23:26skoobasteeve has joined #asahi-re
23:53marcanSIO apparently can handle a bunch of IO paths including SPI/UART/etc, but it looks generic
23:53marcani.e. we don't *have* to use it
23:54blazra[m] has joined #asahi-re
23:55marcan< sven> i just know that the AES DART wasn't able to even read the pagetable <- that matches the 1:1 memory map entry in the DAPF I saw, with special flags
23:55marcanalyssa: Numerically Controlled Oscillator probably
…