#
…

2023-02-08

00:03Keyaedisa has quit [Ping timeout: 480 seconds]
00:11leitao has quit [Ping timeout: 480 seconds]
00:15seeeath has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
00:16seeeath has joined #asahi-dev
00:29Keyaedisa has joined #asahi-dev
00:42Keyaedisa has quit [Ping timeout: 480 seconds]
00:53seeeath has quit [Remote host closed the connection]
00:53seeeath has joined #asahi-dev
01:15Keyaedisa has joined #asahi-dev
01:25Keyaedisa has quit [Ping timeout: 480 seconds]
02:23seeeath has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
02:38seeeath has joined #asahi-dev
02:42Keyaedisa has joined #asahi-dev
02:45___nick___ has quit []
02:49___nick___ has joined #asahi-dev
02:55seeeath_ has joined #asahi-dev
02:55seeeath_ has quit []
02:58seeeath_ has joined #asahi-dev
03:00seeeath has quit [Ping timeout: 480 seconds]
03:14seeeath_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
03:16seeeath has joined #asahi-dev
03:22Keyaedisa has quit []
03:23marcanpovik: pushed the testtone/notebook stuff
03:53leitao has joined #asahi-dev
04:02leitao has quit [Ping timeout: 480 seconds]
04:45leitao has joined #asahi-dev
04:53leitao has quit [Ping timeout: 480 seconds]
05:10seeeath has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
06:07jeffmiw has joined #asahi-dev
06:19crabbedhaloablut has joined #asahi-dev
06:21loki_val has quit [Ping timeout: 480 seconds]
07:39jeffmiw has quit [Ping timeout: 480 seconds]
08:25bisko has joined #asahi-dev
08:32audiojaguar has joined #asahi-dev
08:37audiojaguar_ has quit [Ping timeout: 480 seconds]
08:46leitao has joined #asahi-dev
08:54leitao has quit [Ping timeout: 480 seconds]
08:57MajorBiscuit has joined #asahi-dev
09:06svenhrm, so the acio dart's stream 15 is set to bypass by macos
09:06sventhunderbolt still works if I don't do that
09:06svenI've enabled translation without setting up any pagetables and get these faults: apple-dart 501a80000.iommu: translation fault: status:0x8f0f0001 stream:15 code:0x1 (NO TTBR FOR IOVA) at 0x100100
09:06svenso... uh... what the hell is that stream doing?
09:10bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
09:24zef has joined #asahi-dev
09:29chadmed has quit [Remote host closed the connection]
09:33chadmed has joined #asahi-dev
09:37jluthra has quit [Remote host closed the connection]
09:37jluthra has joined #asahi-dev
09:47leitao has joined #asahi-dev
10:04jannausven: could it be a backchannel to pmgr / pmp?
10:13leitao has quit [Ping timeout: 480 seconds]
10:14chadmed has quit [Remote host closed the connection]
10:21chadmed has joined #asahi-dev
10:26jannaubt device/key parsing from nvram done. I guess I should write asahi-btsync in rust as well to avoid a stdout / shell interaction with device supplied name fields
10:27jannauto bad bluez seems to have no import device (with pairing key) functionality
10:37c10l6 has joined #asahi-dev
10:37c10l6 is now known as c10l
10:42bisko has joined #asahi-dev
10:44chadmed has quit [Remote host closed the connection]
10:46chadmed has joined #asahi-dev
10:47chadmedusing rtkit to give pipewire "proper" "realtime" support fixes all the buffer overflowing issues with the compressor
10:48chadmedit does have a higher affinity for pcores now but i think thats alright, better than choppy audio
10:49chadmedpower use is ~the same anyway
10:49chadmedmoving the cursor pushes up power consumption more than playing music does
10:55jannauthe trackpad generates ~10k irq/s
10:59povikdo we want to bring up sio to address that?
11:00povikor is there something else we can do to push the irq number down?
11:00mazdon't touch the trackpad? :D
11:00povikother than that... :P
11:01mazsome trackpads have a "precision" setting that influence the rate of updates, and thus the rate of IRQs.
11:01chadmedits quite entertaining watching the scheduler try to pass irq threads around when you move the trackpad. the thing goes from every core sitting at 0 to multiple pcores running up 5-10% utilisation
11:01maz(vague memory from a previous life)
11:01jannauadd delays in the irq hadnler
11:02* mazthrows something at jannau
11:03jannauI think aic2 helpfully distributes the irq to all cores
11:03mazyeah, it has no notion of affinity to speak of. which is an annoyance on its own (Linux is not very well equipped to deal with 1:N distribution)
11:04mazI'll eventually have to do something about that, as some GIC users are doing that too these days.
11:08jannaupovik: not sure if sio will help much. I doubt the spi transfers itself cause too much cpu load compared to the event parsing and input handling
11:17povikwouldn't there be significant overhead in handling the 10k/s irqs?
11:17poviki assume the number of events is much lower and we spend many irqs per event
11:18povik(i'm just guessing here, no hard numbers at hand)
11:29jannauthe interrupt handling in spi_hid should maybe re-thought
11:37joshtaylor has joined #asahi-dev
11:49dottedmagjannau: This happens on boot, couldn't one supply bluez with the files in /var/lib/bluetooth directly? Though the file format is documented as "discouraged to access directly", of course.
11:53jannaudottedmag: yes, I'll plan to manage /var/lib/bluetooth directly but it's annoying
11:58jnn is now known as jn
12:03svenjannau: hm... possibly. the iova which should be 1:1 to phyaddr in bypass mode is just very strange
12:07svenand ofc the thunderbolt/usb4 firmware has almost no strings :<
12:16poviki remember there could be offsets applied in bypass mode but you probably checked for those
12:16svenyeah, and there's also not DAPF behind the DART
12:38bluetail8 has joined #asahi-dev
12:45bluetail has quit [Ping timeout: 480 seconds]
13:07svenhrm, every write to a NHI ring seems to trigger that fault
13:08svenlet's see if I can just map that page to some memory and observe if it tries to write anything there
13:57leitao has joined #asahi-dev
13:58leitao has quit []
13:59leitao has joined #asahi-dev
14:37bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
14:37ncopa2 is now known as ncopa
14:42sven... that probably didn't work because stream 15 just seems to be related to an entirely different physical space afaict
14:56seeeath has joined #asahi-dev
15:38svenjannau: can you trace a acio dart on your j314? the adt you sent me some time ago seems to have a bypass-address configured for those
15:55svenohh… wait.. maybe whatever triggers that 0x100100 access actually sends a 64bit address but because once I enable translation I’m limited to 32bit I just don’t see the upper bits
16:14seeeath has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
16:18jannausven: how do you trace dart-acio? trace_dart.py accesses the dart too early
16:19povikhow come a trace script accesses the dart at all?
16:19svenjust trace_device should work
16:19sveni only care about the TCR value for stream 15
16:19svenyou need to plug in a tbt device fwiw
16:20balrog has quit [Quit: Bye]
16:24balrog has joined #asahi-dev
16:24leitao has quit [Read error: Connection reset by peer]
16:25jannausven: MMIO: W.4 0xb01a8013c (dart-acio1[0], offset 0x13c) = 0x20000
16:27kh has joined #asahi-dev
16:27svenhuh.. that doesn’t even enable bypass i think
16:29khmarcan: Are non-PSCI cpuidle drivers blocked on arm64? Sudeep doesn't seem to be against it
16:36svenweird
16:37svenit doesn’t matter much because thunderbolt works just fine even with that stream set to block all access but I’d really like to know what these 0x100100 reads/writes are supposed to do
16:41jannaudev 0f: TCR=0x20000 (BYPASS_DAPF=0, BYPASS_DART=0, TRANSLATE_ENABLE=0)
16:42svenguess we can eventually try and see what happens on t6k with my Linux driver and that stream set to translate with an empty pagetable
16:48cupermir has quit [Remote host closed the connection]
16:52cupermir has joined #asahi-dev
16:57marcankh: As far as I know, yes, certainly without significant discussion first.
17:01tino has joined #asahi-dev
17:07cupermir has quit [Ping timeout: 480 seconds]
17:19Dementor has quit [Ping timeout: 480 seconds]
17:23seeeath has joined #asahi-dev
17:24gladiac has joined #asahi-dev
17:29MajorBiscuit has quit [Ping timeout: 480 seconds]
17:35bisko has joined #asahi-dev
18:06bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
18:08lawrence has quit [Quit: The Lounge - https://thelounge.chat]
18:12lawrence has joined #asahi-dev
18:42Dementor has joined #asahi-dev
18:57kh has quit [Read error: Connection reset by peer]
18:58leitao has joined #asahi-dev
19:06Dementor has quit [Ping timeout: 480 seconds]
19:36zzywysm has joined #asahi-dev
19:57jeffmiw has joined #asahi-dev
20:00leitao has quit [Ping timeout: 480 seconds]
20:24jeffmiwdebugging some probably wrong usage of dt of_ functions, that makes a sequence of insmod/rmmod/insmod creating a cpu deadlock :) the 2nd insmod get stuck on an of_find_node_by_name for a node that does not exist ... the first insmod works well with it ...
21:08___nick___ has quit [Ping timeout: 480 seconds]
22:18gladiac has quit [Quit: k thx bye]
22:32hertz has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
23:05seeeath has quit [Ping timeout: 480 seconds]
…