#
…

2021-08-29

00:12TheLinklooks like an exiting week is coming up :)
00:17alyssaany week i take off from work (panfrost) is a good week for asahi 😋
02:31phiologe has joined #asahi-dev
02:33PhilippvK has quit [Ping timeout: 480 seconds]
03:08skipwich has quit [Ping timeout: 480 seconds]
03:43alyssa[syslog] * [FramebufferDCP.cpp:1824]virtual IOMFBStatus IOMobileFramebuffer::swap_submit_dcp(const IOMFBSwap *, IOMFB::Surface **, const uint32_t *, bool, double,
03:43alyssatell me more! :-p
03:51alyssaBy the way, since it wasn't obvious to me (and so probably isn't obvious for anyone here but marcan)--
03:51alyssaswap_enabled/swap_completed are bitfields
03:51alyssabit i says whether to enable surface i
03:51alyssa("layer" in the firmware, based on syslog)
03:53alyssa[syslog] * [UPPipeDCP_H13P.cpp:3320]IOMFB verify_active_surface_count: cannot blend more than 2 layers
03:53alyssaOh, that's a bummer :|
03:54alyssamarcan: that explains why you saw macos fuse the cursor into the main fb .... that's the only way to make room for an overlay plane for video compositing, etc
03:55marcanthen why are there 3 slots? that's weird...
03:56alyssaYeah. I'd say "just firmware, there are 4 in 12.x after all" but that doesn't explain why there seemed to be 3 slots in the registers too..
03:56alyssaUnless DCP has a particular meaning of "blend" (i.e. scaling or CSC or overlapping or something)
03:57alyssaand you get 3 as long as you aren't using fancy featur?s/ Dunno
03:57alyssa(continued in #asahi-re)
08:31aleasto has joined #asahi-dev
11:14gabuscus has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
11:14gabuscus has joined #asahi-dev
12:20al3xtjames2 has joined #asahi-dev
12:24al3xtjames has quit [Ping timeout: 480 seconds]
12:24al3xtjames2 is now known as al3xtjames
12:26aleasto has quit [Remote host closed the connection]
13:52PhilippvK has joined #asahi-dev
13:55phiologe has quit [Ping timeout: 480 seconds]
14:01alyssathis is kinda interesitng
15:00marcanalyssa: weren't those already in the list I sent back in the day?
15:01alyssamarcan: yes, but it's kinda interesting seeing it in real apple code
15:02alyssamaybe i'm a nerd :p
15:02* alyssareads the DT spec
15:21alyssaThat too
15:44bps has quit [Ping timeout: 480 seconds]
17:18alyssaj_ey: what cmdline do you use for hv?
17:19alyssalinux kernel command line I mean, in chosen.bootargs
17:19j_eyconsole=tty0
17:20* alyssatrying to debug display
17:21alyssaGuest exception: EXCEPTION_LOWER/SERROR
17:21alyssavuart works up until then, though
17:22j_eyoh, if youre trying to use vuart, dont use that console=
17:22alyssayeah, I am
17:24alyssa[ 0.116190] Unable to handle kernel NULL pointer dereference at virtual address 000000000000022c
17:24alyssaOh, finally :]
17:25j_eyfrom one error to another!
17:26alyssaj_ey: this is the one I'm interested in
17:27alyssaand answer seems to be "deferred probing" or maybe just fixing my DT bindings
17:28alyssaj_ey: thank u
17:28j_eyno thank you
17:29alyssawhat did i do
17:31alyssaok, now why is it broken
17:38chadmed has quit [Read error: Connection reset by peer]
17:46pipcet[m]blame j_ey!
17:47j_eybtw don't rebase onto a recent next, looks like some changes broke nvme
17:47pipcet[m]how broken?
17:48pipcet[m]as in, doesn't it initialize or does it randomly eat your data?
17:48j_eyformer
17:52balrog has quit [Ping timeout: 480 seconds]
18:27svenyeah.. tracking pci.c in the future is going to be fun
18:28j_eythe tag / command_id is 16 bits in the code, but we think the hw might only care about 8 bits, so therefore it's 0-ing out the 'gen' thats put into the top 4 bits
18:29svenand while we can theoretically hot plug usb-c devices now, we can't really distinguish between the two ports yet.
18:29svenso.. uh.. we were half-successful i guess :D
18:31pipcet[m]same message for both ports? d'oh.
18:32svenyeah. then we though we found some bit in the phy registers to find the correct one
18:32svenbut that didn't work either
18:33kettenis_sven: does it matter to find the correct one?
18:33svenkettenis_: yes, because otherwise the other phy will get a reset as well and whatever is plugged in there will stop working
18:34kettenis_oh, the phy actually needs a reset before it can detect a new device?
18:34svenit appears so :/
18:34sventhat's at least what mac os does
18:34sventhere's still that pd chip on i2c. that should hopefully also have some interrupt to detect a new connection
18:35kettenis_but mac os uses this SMC thing?
18:35sveni'm not entirely sure. i haven't been able to see the direct call chain in the hv
18:35j_eySMC generates a notification, but we dont know if it actually uses that
18:35j_eyor some other method
18:35sveni just see both the SMC notification and some USB PD interrupt and a few moments later it issues the PHY reset
18:36pipcet[m]sven: do you see a PD interrupt if you unplug a device from a typec-typea adapter? rather than unplugging the adapter?
18:36svenand the backtrace from the PHY reset wasn't helpful to see how it was triggered
18:37pipcet[m]is it reading something else from the SMC, maybe? maybe there's a bit in there for each port...
18:38pipcet[m]D0* / D1* would be suspicious keys...
18:43svenhttps://github.com/AsahiLinux/linux/commits/sven/20210829 too lazy to split it properly, but that contains the current WIP state. will at least put the nvme stuff back in the feature branches later this week
18:43svenj_ey: ^--
18:44j_eycool, ty!
18:52kettenis_sven: do we really need a separate node for the SART in the device tree?
18:56sveni think i'd prefer to have SART part of the ANS/NVMe node
18:57kettenis_I liked the version where you made it part of the mailbox
18:57kettenis_since it avoided the driver crosscalls
19:22balrog has joined #asahi-dev
19:56bps has joined #asahi-dev
20:11sventhat branch uses a dumb mailbox fwiw
20:11svenjust a single channel with 64+32bit messages
20:14pipcet[m]D1St and D2St are SMC keys that might correspond to device connection status, but I'm not sure that's what we want
20:14svenyeah, might be easier to just rely on the usb pd interrupts (which should exist) at that point
20:15pipcet[m]IIRC we don't get those interrupts if it's the remote end of a type-a/type-c adapter that's disconnected
20:15svendoesn't matter, those work fine
20:18pipcet[m]they do? that's weird, but if you say so they must...
20:18svenmine works fine as long as i keep the type-c end connected
20:18svenit's just disconnecting/reconnecting the type-c end that doesn't work for me
20:19alyssa^^ ditto for me
20:19alyssawhich makes sense I think
20:20alyssait's apurely physical layer issue
20:28pipcet[m]you're right, it does make sense. PD interrupts it is then :P
20:36bps has quit [Ping timeout: 480 seconds]
20:39Mary has quit [Quit: The Lounge - https://thelounge.chat]
20:42Mary has joined #asahi-dev
21:48alyssaDCP driver "works"
21:48alyssafor some very generous definition of works.
22:00yrlf has quit [Quit: The Lounge - https://thelounge.chat]
22:01balrog has quit [Quit: Bye]
22:01yrlf has joined #asahi-dev
22:14balrog has joined #asahi-dev
…