#
✕ nick:marcan
#asahi-dev · 2022-08-09 14:24
marcansven: there isn't, but that doesn't look too hard to add
#asahi-dev · 2022-08-09 14:23
marcanj`ey: yeah, it can just be checked for NULL and thrown away, some other drivers do that
#asahi-dev · 2022-08-09 13:34
marcanif we lose a little loudness/headroom to safety, so be it
#asahi-dev · 2022-08-09 13:33
marcanhonestly, I get the feeling that with per-speaker this and a volume cap in the DT, we can probably deal with most of the blowing up speakers problem
#asahi-dev · 2022-08-09 13:32
marcanwouldn't be hard to test
#asahi-dev · 2022-08-09 13:31
marcanI assume it's a 1-pole filter
#asahi-dev · 2022-08-09 13:10
marcanchadmed: that sounds good enough to protect tweeters?
#asahi-dev · 2022-08-09 13:10
marcanpovik: TAS2764 DC_BLK0.HPF_FREQ_PB can highpass at up to 800Hz
#asahi-dev · 2022-08-09 13:05
marcan(so even macOS isn't immune to this...)
#asahi-dev · 2022-08-09 13:04
marcan(on another note, a friend just got their T2 Mac back from Apple after it blew its speakers up today. with macOS.)
#asahi-dev · 2022-08-09 13:03
marcanit would be interesting to find out if e.g. the feedback stuff can detect the speaker diaphragm hitting mechanical limits (which might cause a detectable current pattern)
#asahi-dev · 2022-08-09 13:03
marcanreminder that I have a pile of machines and AppleCare on all of the expensive ones, so if I *have* to blow some speakers up to make sure nobody else does, I can do that.
#asahi-dev · 2022-08-09 13:02
marcanwhat about the monitoring features, can the kernel use those to detect badness happening and kill things before physical damage happens?
#asahi-dev · 2022-08-09 13:02
marcancan they implement e.g. a simple highpass?
#asahi-dev · 2022-08-09 13:02
marcanI'd be more interested in just how well can we configure the speaker amps to provide basic safety
#asahi-dev · 2022-08-09 13:01
marcanbut yeah, no, I don't see in-kernel DSP flying any time soon
#asahi-dev · 2022-08-09 13:01
marcan(my piano keyboard runs linux and I know too much about it for my own good)
#asahi-dev · 2022-08-09 13:00
marcanpovik: KORG put an entire multi-mode synth in the kernel, but then again they also violate the GPL left and right, so probably not an example to follow
#asahi-dev · 2022-08-09 01:28
marcanso that needs to be fixed either way (what I suggested was building the L2 PT then doing an in-place swap)
#asahi-dev · 2022-08-09 01:28
marcanI wouldn't trust that that works for dcp either
#asahi-dev · 2022-08-08 16:00
marcanbut I think we found a viable workaround to make old drivers fail to probe instead of exploding
#asahi-dev · 2022-08-08 15:59
marcanrobher_: in this case it's more like an existing driver will do entirely the wrong thing with some hardware instances (not due to hardware differences, due to usage/configuration differences) on probe, and break existing stuff
#asahi-dev · 2022-08-08 11:37
marcanyuup...
#asahi-dev · 2022-08-08 11:37
marcanyup...
#asahi-dev · 2022-08-08 11:35
marcan32MB is almost big enough for a 4K framebuffer but not quite; would fit the screen FB I think. so it might be a FB.
#asahi-dev · 2022-08-08 11:34
marcancool
#asahi-dev · 2022-08-08 11:31
marcan_jannau_: btw, is the driver ported to 12.3, or still 12.0? (or did it not change?)
#asahi-dev · 2022-08-08 11:25
marcanre pro/max, I suspect some of these regions have something to do with the LED backlight
#asahi-dev · 2022-08-08 11:24
marcanI think I'd rather inject that in m1n1 entirely, especially for the rando-numbered ones that aren't something explicit like vram
#asahi-dev · 2022-08-08 11:23
marcanwhat I'm not so sure about is the templated nodes, since that could change based on firmware version potentially
#asahi-dev · 2022-08-08 11:23
marcanI'm fine with your approach
#asahi-dev · 2022-08-08 11:22
marcanright, I was reluctant to touch those reserved regions but they do look static, probably just enum members they never remove
#asahi-dev · 2022-08-08 11:04
marcancool, I'll take a look once I'm done with the random low hanging fruit business
#asahi-dev · 2022-08-08 11:00
marcanbut this solves the immediate compat issue
#asahi-dev · 2022-08-08 11:00
marcanwe can always unlock it in the future once old kernels die off for a while
#asahi-dev · 2022-08-08 10:59
marcanand yeah, let's just lock it and keep the compat then
#asahi-dev · 2022-08-08 10:59
marcanso it should be a no-op on laptops
#asahi-dev · 2022-08-08 10:58
marcandisplay_init() does nothing if the display is internal and already initialized, and display_shutdown() does nothing if iboot was not initialized
#asahi-dev · 2022-08-08 10:57
marcanand then older kernels will just refuse to probe
#asahi-dev · 2022-08-08 10:57
marcanI mean, ultimately it's like DCP in that not handling it delicately will make things explode too, and once we implement support for that, who cares if we use it twice...
#asahi-dev · 2022-08-08 10:57
marcanbut what do you think about the locking disp0 trick?
#asahi-dev · 2022-08-08 10:56
marcan(I'm not sure if shutting it down on laptops is a good idea, I think it does more stuff there)
#asahi-dev · 2022-08-08 10:56
marcanonly desktops
#asahi-dev · 2022-08-08 10:56
marcandcp isn't shutdown on laptops
#asahi-dev · 2022-08-08 10:54
marcanmaybe we can lock it in m1n1 for a while, as a workaround, since we need to implement that for DCP anyway :p
#asahi-dev · 2022-08-08 10:53
marcanoh wait, DCP DART is locked, but DISP0 isn't, right? sigh...
#asahi-dev · 2022-08-08 10:53
marcanmaz: I told him about this and that I'd warn him when it happens (it will)
#asahi-dev · 2022-08-08 10:52
marcanapple_dart_hw_reset punts if the DART is locked
#asahi-dev · 2022-08-08 10:52
marcandon't we already have that anyway?
#asahi-dev · 2022-08-08 10:51
marcanwe can still have some transition kernels where we implement a probe failure before everything else is ready

plus ›