#
…

2021-04-29

00:38gabiruh has quit [Quit: ZNC 1.7.5 - https://znc.in]
00:39gabiruh has joined #asahi
01:23tomtastic has quit [Ping timeout: 240 seconds]
01:23tomtastic has joined #asahi
02:46PhilippvK has joined #asahi
02:46phiologe has quit [Ping timeout: 276 seconds]
02:54larbob has joined #asahi
03:12marvin24 has quit [Ping timeout: 250 seconds]
03:13marvin24 has joined #asahi
03:35jeffmiw_ has joined #asahi
03:39jeffmiw_ has quit [Ping timeout: 246 seconds]
04:02odmir has quit [Remote host closed the connection]
04:34taziden has quit [Ping timeout: 260 seconds]
04:44taziden has joined #asahi
05:30VinDuv has joined #asahi
06:12bgianf has joined #asahi
06:17ephe_meral1 has joined #asahi
06:20VinDuv has quit [Quit: Leaving.]
06:41_jannau_probably outdated
06:43marcanthanks! I'll look at that again later
06:51svennice! i was planning to add sprr/gxf registers to regs.json but i like the separate file for apple-regs :)
06:54pipcet[m]marcan: the AIC driver currently sends all irq interrupts to CPU0, right?
06:55marcanpipcet[m]: no, it supports SMP
06:55pipcet[m]marcan: thanks, I'll have a look to see why it's not working here :-)
06:55pipcet[m]probably my own fault
07:11crabbedhaloablut has quit [Ping timeout: 240 seconds]
07:12crabbedhaloablut has joined #asahi
07:20zorg24 has joined #asahi
07:41flokk[m] has joined #asahi
08:03herbas has joined #asahi
08:03maknho___ has quit [Ping timeout: 252 seconds]
08:04herbas has quit [Client Quit]
08:06mazpipcet[m]: the default affinity is to CPU0. you can play with /proc/irq/*/smp_affinity* to set it to any random CPU.
08:06maknho___ has joined #asahi
08:27easylee has joined #asahi
08:28WCat[m] has joined #asahi
08:35AWanderer has joined #asahi
08:50maknho____ has joined #asahi
08:53maknho___ has quit [Ping timeout: 260 seconds]
08:58maknho____ has quit [Quit: WeeChat 2.3]
08:58maknho has joined #asahi
10:18agnem has quit [Quit: WeeChat 3.0.1]
10:19agnem has joined #asahi
10:20raster has joined #asahi
10:55tomtastic has quit [Ping timeout: 240 seconds]
10:57tomtastic has joined #asahi
10:58sven_ has joined #asahi
11:00sven has quit [Ping timeout: 248 seconds]
11:00sven_ is now known as sven
11:02pipcet[m]but just one at a time, right?
11:36JTL has quit [Ping timeout: 252 seconds]
11:38JTL has joined #asahi
11:41marcanpipcet[m]: the config lets you specify an arbitrary CPU set, but I think Linux will then pick one from that or so
11:41marcanAIC lets you set multiple bits, but I think the hardware always goes to the first CPU? I need to try again and see exactly what happens
11:41marcanthe driver currently only sets one bit
11:42pipcet[m]I think the not-to-be-mentioned old driver set several bits and there appeared to be some degree of IRQ balancing going on with it...
11:43pipcet[m]but overall I'm glad I don't have to use that anymore, the existing device drivers work fine with the new aic driver, with a few modifications :-)
11:45pipcet[m]I understand it's probably not worth it to make the change, but I do wonder why we have all the FIQ code in the AIC driver even though they appear to be totally independent. The one dependency is that we can use FIQ or IRQ IPIs, but not both, right?
11:54marcanpipcet[m]: precisely because of that: because once that's implemented, we have to pick one, and two drivers would have to coordinate that
11:54marcanbut also, things started with a common irq/fiq vector and only got split later
11:56marcanit's kind of a wash which way to do it, but if we end up with better reasons to have them merged in the future, doing so post-facto would be a lot more painful
11:56marcanso, even though it might be uglier, it seems safer to do it this way
11:57marcanand yeah, if hardware IRQ balancing works as intended we should certainly supported; I actually had that at some point but decided to play it safe for now
11:57TheJollyRoger has quit [Ping timeout: 240 seconds]
11:57marcanbut for what it's worth, setting multiple bits in smp_affinity on my x86 iMac also ends up with everything on one CPU so... :-)
11:58pipcet[m]but with the x86 there's no real difference between the cores, is there :-)
11:59TheJollyRoger has joined #asahi
11:59pipcet[m]but really, those are very minor issues. The driver appears to work, though I did hack in a few things to get the other device drivers working and need to investigate more.
12:01pipcet[m](using it with the Corellium drivers, for the time being)
12:04mazin general, dispatching on multiple CPUs has an adverse effect on the interrupt service latency.
12:05mazthat's at least what the x86 and GICv2 experience has taught us over the years.
12:06mazthings may be different on M1, but a change would need to be backed by hard numbers.
12:07kettenis1which is kinda hard to do without more device drivers ;)
12:09mazproblem solved! :D
12:12zorg24 has quit [Quit: Connection closed]
12:21pipcet[m]turns out disabling your keyboard interrupt still isn't a good idea!
12:22mazit's called secure computing.
12:35choozy has joined #asahi
12:41pipcet[m]I do wonder how the AIC selects which CPU to send the interrupt to. It appears to be mostly the first two ones to have their bits set, but once in a while it's the third.
12:41pipcet[m]so probably the first CPU to be considered "idle", by whatever standard?
12:56opticronthat standard is probably "interrupts enabled"
13:17mazthat'd mean the interrupt controller can peek into PSTATE for each CPU. doesn't scale to big systems, but certainly a good optimisation on a small system.
13:22pipcet[m]IRQs are enabled separately from FIQs, right?
13:25opticrondoes the AIC actually select a CPU or does it deliver the interrupt to all affinity-selected CPUs and then the one that gets there first sees a real interrupt while the others see a "spurious" interrupt if they catch it at the same time?
13:25mazyup. different bit in PSTATE.
13:26mazopticron: I'm sure it is extensively documented in the spec.
13:26opticronheck, I'm not even sure AIC *HAS* spurious interrupts like GICv2 and GICv3
13:27pipcet[m]opticron: good point, we should print out a message for spurious interrupts
13:27pipcet[m]so we could run userspace code with IRQs disabled, if we were using FIQ IPIs?
13:28mazerr... no.
13:28pipcet[m]on all but one core, at least :-)
13:28mazI guess you could use CPU isolation which comes very close to that.
13:29pipcet[m]just thinking about possibilities here. it'd be silly to ruin a CPU's cache if another CPU is idle
13:33tomtastic has quit [Ping timeout: 246 seconds]
13:34modrobert has quit [Read error: Connection reset by peer]
13:36TheJollyRoger has quit [Ping timeout: 240 seconds]
13:39tomtastic has joined #asahi
13:48pipcet[m]and what's wrong with running userspace code without enabling IRQs on 7 cores?
13:50mazpipcet[m]: that's fine as a special case, and you can do that from userspace by moving all the IRQs to a single core. But in general, you want to be able to place interrupts where they matter most, specially if you have to deal with devices doing back-to-back interrupts (wild guess, a router).
13:50pipcet[m]okay, thought there was a more serious issue. You're right that it's best done in userspace, though.
13:51modrobert has joined #asahi
13:53mazas I mentioned earlier, CPU isolation does exactly that to avoid any disruption to userspace performing a timing-critical task. if you're careful enough, you don't even get IPIs. of course, that comes at the expense of moving all ancillary tasks to worker CPUs which have to deal with everything the kernel normally does in the background.
13:58pipcet[m]I think you're right for real RT tasks, but I'm just thinking about a performance hack to keep the video player from being interrupted by interrupts (which we will need until we've got the hardware scaler up and running). but, again, just trying to wrap my head around this IRQ-vs-FIQ thing
14:14jeffmiw_ has joined #asahi
14:18jeffmiw_ has quit [Ping timeout: 260 seconds]
14:23maor26 has joined #asahi
14:26TheJollyRoger has joined #asahi
14:27maor26 has quit [Client Quit]
14:33AWanderer has quit [Ping timeout: 240 seconds]
14:51marcannobody cares about video players, those things have a whole 16-33ms or so to get their work done
14:51marcannow if we're talking low-latency audio, yeah, that's when starting to think about these things *might* matter
14:51marcanit's on my longer-term todo list :-)
14:52marcan(I'm probably going to be moving to linux on M1 for all my live audio stuff)
15:07kettenis1is the audio based on I2S? Wonder how much powerpc heritage there is of that's the case
15:18choozy has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
15:21ephe_meral1 has quit [Ping timeout: 252 seconds]
15:33pipcet[m]i2s is mentioned in the TI docs, yes.
15:58odmir has joined #asahi
16:07zkrx has quit [Ping timeout: 276 seconds]
16:20zkrx has joined #asahi
16:34choozy has joined #asahi
16:37odmir has quit [Remote host closed the connection]
16:38odmir has joined #asahi
16:43odmir has quit [Ping timeout: 260 seconds]
16:53modwizcode has quit [Ping timeout: 248 seconds]
16:54VinDuv has joined #asahi
16:54odmir has joined #asahi
17:22modwizcode has joined #asahi
19:26roxfan has quit [Remote host closed the connection]
19:29roxfan has joined #asahi
19:34KindTwo has joined #asahi
19:36KindOne has quit [Ping timeout: 265 seconds]
19:36KindTwo is now known as KindOne
19:39choozy has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
19:53bfredl has quit [Ping timeout: 245 seconds]
19:54ArnoldSebastian[ has quit [Ping timeout: 245 seconds]
19:54cepheus has quit [Ping timeout: 245 seconds]
19:54flokk[m] has quit [Ping timeout: 260 seconds]
19:54uartman[m] has quit [Ping timeout: 245 seconds]
19:55ArnoldSebastian[ has joined #asahi
19:55cepheus has joined #asahi
19:55flokk[m] has joined #asahi
19:55uartman[m] has joined #asahi
19:56foxlet has quit [Ping timeout: 245 seconds]
19:57foxlet has joined #asahi
19:57bfredl has joined #asahi
20:04maknho has quit [Ping timeout: 260 seconds]
20:05maknho has joined #asahi
20:16VinDuv has quit [Quit: Leaving.]
20:25odmir has quit [Remote host closed the connection]
20:26odmir has joined #asahi
20:30izzyisles[m] has joined #asahi
20:31odmir has quit [Ping timeout: 240 seconds]
20:39jeffmiw_ has joined #asahi
20:43jeffmiw_ has quit [Ping timeout: 265 seconds]
20:57odmir has joined #asahi
21:09zkrx has quit [Ping timeout: 268 seconds]
21:17KindOne has quit [Quit: K-Lined]
21:21bsandro has joined #asahi
21:22modwizcodesven: how far did you get on guarded exec stuff?
21:28svenmodwizcode: completely figured it out
21:28modwizcodethat's great!
21:28svenjust procrastinating on the documentation right now :D
21:29modwizcodewell I guess I won't bother trying to pull up the wiki then
21:29svennothing there yet
21:30svenbut i'll do it this weekend
21:30modwizcodehow about the SART?
21:30sventl;dr: lateral exc level to EL1 and EL2 with the same pagetables but different pagetable permissions
21:30modwizcodeor whatever it's called
21:30svendidn't look into that one yet
21:31maximus64wondering how hard it would be to enable guarded exec on Linux :P
21:31kettenis1it appears the SART is used to give the iop that manages the NVMe storage access to some of the DRAM on the SoC
21:31modwizcodedo we at least have any idea *why* they would bother with two (or three even?) distinct mapping formats
21:31modwizcodethat might explain the why
21:31svenmaximus64: uh... very hard probably
21:31svenit's easy from m1n1 though
21:32modwizcodeGX mode basically prevents you from changing the page tables right?
21:32sventhat's how xnu uses it
21:32modwizcodeIt's more generic then?
21:33svenit changes the meaning of AP[1:0],UXN and PXN to a table
21:33svenand then you have two different tables: one for ELn and one for GLn
21:33svenand you can essentially flip the permissions of many page at once by calling genter
21:33modwizcodeNot knowing anything (and having forgotten a lot about APRR) is there a reason they changed how that was happening?
21:33modwizcodebecause that sounds basically identical to APRR
21:33svenit's very similar
21:33modwizcodeminus the use of cpu instructions
21:34svenit feels like a small iteration of APRR
21:36modwizcodedo the APRR registers still exit too in addition to the GX thing?
21:36sventhey're gone
21:36svenor, well, they exist but are stuck at 0
21:40zkrx has joined #asahi
21:42modwizcodesven: can you give me a ping when the GX stuff goes up and then if you find out more about the SART?
21:44odmir has quit [Remote host closed the connection]
21:44odmir has joined #asahi
21:57akemin_dayo has quit [Quit: ┐(´д`🍍)┌]
21:58akemin_dayo has joined #asahi
22:05agnem has quit [Quit: WeeChat 3.1]
22:06agnem has joined #asahi
22:08agnem has quit [Client Quit]
22:09agnem has joined #asahi
22:36agnem has quit [Quit: WeeChat 3.1]
22:37agnem has joined #asahi
23:12artemist has quit [Ping timeout: 248 seconds]
23:18mrkajetanp has joined #asahi
23:30pthariensflame has joined #asahi
23:31pthariensflame has quit [Client Quit]
23:34odmir has quit [Remote host closed the connection]
23:35odmir has joined #asahi
23:40odmir has quit [Ping timeout: 240 seconds]
23:48pipcet[m]huh. So I guess the instruction and data caches are nearly coherent?
23:54Hexagon has quit [Quit: Hexagon]
23:59Hexagon has joined #asahi
23:59Hexagon has quit [Client Quit]
…