<marcan>
huh... I guess the fix should be conditional on aarch64 then :/
<marcan>
wonder why that happens
<marcan>
thank you for figuring that out, I'm glad I didn't submit it as-is then :)
bisko has quit [Ping timeout: 480 seconds]
lrea^ has quit [Ping timeout: 480 seconds]
bisko has joined #asahi-dev
<mipi[m]>
@marcan: Yeah, seems so. I am preparing a PR for alarm.
bisko has quit [Ping timeout: 480 seconds]
bisko has joined #asahi-dev
bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
bps2 has joined #asahi-dev
bisko has joined #asahi-dev
bisko has quit []
bisko has joined #asahi-dev
Nspace has quit []
Nspace has joined #asahi-dev
bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
bisko has joined #asahi-dev
c10l7 has quit []
c10l7 has joined #asahi-dev
c10l7 has quit []
c10l7 has joined #asahi-dev
c10l7 is now known as c10l
lrea^ has joined #asahi-dev
bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
lrea^ has quit [Ping timeout: 480 seconds]
bisko has joined #asahi-dev
<amarioguy>
sven: i did some preliminary digging into what I needed to do for the i2c chip, the main instance of polling i saw seems to be in the waitready func
<amarioguy>
unless i missed something it should be relatively simple to request the irq, add a wait queue, then replace the polling with a sleep until the irq is handled i assume (the Apple specific driver didn't have any polling from what I can tell)
<amarioguy>
if i missed an instance of polling please do let me know
kov has joined #asahi-dev
<sven>
yeah, you probably want a completion instead of the wait queue
<sven>
and maybe even have the driver support both irq and pilled mode since it’s unclear if the old pasemi boards have a working irq
<amarioguy>
btw is there a particular irq # we want to use for the irq
<millenialhacker_>
Coming back after a couple of days off, not sure why my main nickname is taken, but will resume ISP work soon. I have a clear idea in how apple handles communication over all channels but need to update the tracer in my branch.
<millenialhacker_>
But first, I need to `emerge @world ` to keep my lovely Gentoo updated :D
lrea^ has joined #asahi-dev
<jannau>
amarioguy: you have to use the one specified in the ADT/FDT, not sure if we have transferred it to the device tree
<povik>
it depends on the wiring of the chip
millenialhacker has quit [Ping timeout: 480 seconds]
<povik>
and i see for i2c it's already been inserted into the devicetree
<amarioguy>
jannau: gotcha
<povik>
so all you need to do is something to the effect of platform_get_irq(pdev, 0) and it will pick up the right IRQ behind the scenes
millenialhacker_ is now known as millenialhacker
<amarioguy>
ah perfect, didn't know that was an api actually
<povik>
^-- at least that will work for the platform variant of the driver, don't know about the pci
<amarioguy>
this is in shared logic and i'm planning to keep polled mode as a fallback
<amarioguy>
since this is a fairly small change i should have the git patch ready sometime this week assuming life doesn't complicate things lol
<amarioguy>
(not sure if it's worth creating a whole branch just for this stuff)
<jannau>
platform_get_irq might has to live in the apple specific part of the driver init
<amarioguy>
makes sense
<sven>
the common code probably only needs a use_irq flag and export a irq handler while the apple specific code requests the irq and then sets use_irq to true
<sven>
and I think I already added the irq number to the device tree
the_lanetly_052__ has quit [Ping timeout: 480 seconds]
lrea^ has quit [Remote host closed the connection]
<povik>
hm, i was pretty certain to find that "ASoC: tas2764: Add workaround for spurious shutdowns on SN012776" was merely a workaround for unreliable i2c
<povik>
that would even explain why it affected only of the chips, being the chip that was in a bad spot during init ordering
<povik>
but no, turns out the workaround is still needed
<povik>
who would have guessed...
<povik>
*only one of the chips
<povik>
marcan: indeed it looks like macos always does the SHUTDOWN->ACTIVE->MUTE dance you wrote into the workaround