#
…

2026-08-19

00:01hdbngr has quit [Ping timeout: 480 seconds]
00:09hdbngr has joined #asahi-dev
00:16amarioguy has joined #asahi-dev
00:17hdbngr has quit [Ping timeout: 480 seconds]
00:35Rafay has quit [Quit: Connection closed for inactivity]
00:37aurora has quit [Remote host closed the connection]
00:37aurora has joined #asahi-dev
00:38tjemke771 has joined #asahi-dev
00:45tjemke77 has quit [Ping timeout: 480 seconds]
00:45tjemke771 is now known as tjemke77
00:55aurora has quit [Remote host closed the connection]
00:55aurora has joined #asahi-dev
01:44hdbngr has joined #asahi-dev
01:53hdbngr has quit [Ping timeout: 480 seconds]
01:55pb17 has quit [Ping timeout: 480 seconds]
01:58pb17 has joined #asahi-dev
01:59___nick___ has joined #asahi-dev
02:00yurix has quit [Remote host closed the connection]
02:00yurix has joined #asahi-dev
02:00___nick___ has quit []
02:02___nick___ has joined #asahi-dev
02:36hdbngr has joined #asahi-dev
02:40tobhe_ has joined #asahi-dev
02:45hdbngr has quit [Ping timeout: 480 seconds]
02:46tobhe has quit [Ping timeout: 480 seconds]
03:02tjemke770 has joined #asahi-dev
03:07tjemke77 has quit [Ping timeout: 480 seconds]
03:07tjemke770 is now known as tjemke77
03:12hdbngr has joined #asahi-dev
03:14nora has joined #asahi-dev
03:16nora_ has quit [Ping timeout: 480 seconds]
03:20hdbngr has quit [Ping timeout: 480 seconds]
04:29nela has quit [Ping timeout: 480 seconds]
04:35nela has joined #asahi-dev
04:43Acelogic_ has quit [Remote host closed the connection]
05:00hdbngr has joined #asahi-dev
05:08hdbngr has quit [Ping timeout: 480 seconds]
05:16hdbngr has joined #asahi-dev
05:24hdbngr has quit [Ping timeout: 480 seconds]
05:47hdbngr has joined #asahi-dev
05:56hdbngr has quit [Ping timeout: 480 seconds]
05:59nela has quit [Quit: Ping timeout (120 seconds)]
05:59nela has joined #asahi-dev
06:01croissant` has joined #asahi-dev
06:08croissant has quit [Ping timeout: 480 seconds]
06:25svencould still be worth a try
06:25svenand at least it’s on the correct node
06:29xal has quit []
06:31xal has joined #asahi-dev
06:35hdbngr has joined #asahi-dev
06:36croissant has joined #asahi-dev
06:38croissant has quit []
06:39jannauI've started implementing it
06:39croissant has joined #asahi-dev
06:42croissant` has quit [Ping timeout: 480 seconds]
06:49bgb has quit [Remote host closed the connection]
06:53bgb has joined #asahi-dev
06:53hdbngr has quit [Ping timeout: 480 seconds]
06:58hdbngr has joined #asahi-dev
07:09pb17 has quit [Ping timeout: 480 seconds]
07:09hdbngr has quit [Ping timeout: 480 seconds]
07:17hdbngr has joined #asahi-dev
07:17pb17 has joined #asahi-dev
07:25hdbngr has quit [Ping timeout: 480 seconds]
07:28midou has quit [Remote host closed the connection]
07:29hardy94 has quit [Remote host closed the connection]
07:31midou has joined #asahi-dev
07:35hdbngr has joined #asahi-dev
07:38Acelogic_ has joined #asahi-dev
07:53Acelogic_ has quit [Ping timeout: 480 seconds]
07:54hdbngr has quit [Remote host closed the connection]
07:54hdbngr has joined #asahi-dev
08:03hdbngr has quit [Ping timeout: 480 seconds]
08:04svenlooks like the xdomain uuid macOS used changes each boot
08:05svenprobably just random then so the fff… fallback linux uses is fine then I guess
08:12jannaumy intel notebook generates the same fff… xdomain uuid suffix
08:14hdbngr has joined #asahi-dev
08:26hdbngr has quit [Ping timeout: 480 seconds]
08:32hdbngr has joined #asahi-dev
08:45hdbngr has quit [Ping timeout: 480 seconds]
08:50hdbngr has joined #asahi-dev
09:09hdbngr has quit [Ping timeout: 480 seconds]
09:17xal has quit [Read error: Connection reset by peer]
09:17xal has joined #asahi-dev
09:19hdbngr has joined #asahi-dev
09:24svennice, thanks!
09:25svenmuch better than that old property :)
09:27hdbngr has quit [Ping timeout: 480 seconds]
09:32svenack, will merge then this evening after $work and pull your series in as well
09:34svenI need to fix m1n1 to write to both _ and - for now and then after a bit more testing of the new cable VSE nonsense we can also start to think about shipping this in the downstream kernel
09:46jannauwill be for the 7.2 rebase of the downstream kernel. I have no motivation to backport the thunderbolt PCIe decoupling changes
09:57hdbngr has joined #asahi-dev
10:06hdbngr has quit [Ping timeout: 480 seconds]
10:14hdbngr has joined #asahi-dev
10:30hdbngr has quit [Ping timeout: 480 seconds]
10:36SimonBlaue[m] has joined #asahi-dev
10:38nicolas17 has quit [Ping timeout: 480 seconds]
10:41hdbngr has joined #asahi-dev
10:53svenI wonder how hard it is to get usb4 tunneling to work
10:53sventhat should be easier than either DP or PCIe
10:55hdbngr has quit [Ping timeout: 480 seconds]
10:59nela has quit [Quit: Ping timeout (120 seconds)]
10:59nela has joined #asahi-dev
11:00picnoir has quit [Ping timeout: 480 seconds]
11:03hdbngr has joined #asahi-dev
11:08picnoir has joined #asahi-dev
11:08aissen has joined #asahi-dev
11:15nela has quit [Quit: Ping timeout (120 seconds)]
11:15nela has joined #asahi-dev
11:19hdbngr has quit [Ping timeout: 480 seconds]
11:20hdbngr has joined #asahi-dev
11:26Acelogic_ has joined #asahi-dev
11:28hdbngr has quit [Ping timeout: 480 seconds]
11:35hdbngr has joined #asahi-dev
11:37Acelogic_ has quit [Remote host closed the connection]
11:47hdbngr has quit [Ping timeout: 480 seconds]
11:52Acelogic_ has joined #asahi-dev
11:54hdbngr has joined #asahi-dev
12:06nicolas17 has joined #asahi-dev
12:10hdbngr has quit [Ping timeout: 480 seconds]
12:11hdbngr has joined #asahi-dev
12:32hdbngr has quit [Ping timeout: 480 seconds]
12:42hdbngr has joined #asahi-dev
12:52hdbngr has quit [Ping timeout: 480 seconds]
13:00hdbngr has joined #asahi-dev
13:03sofuschadmed: for interchange i think need to change the meta_bytes calculation, in the avd driver, it looks like this:
13:05chadmedthat seems right, same as agx and dcp
13:05sofusright now it looks like dcp uses a static meta_bytes=8
13:06sofus(the naming might be a bit confusing)
13:07chadmedmeta_bytes is the bytes of metadata per tile
13:07chadmedsorry, yeah thats what the 8 is
13:08chadmedmeta_bytes is the total metadata size of the image
13:09hdbngr has quit [Ping timeout: 480 seconds]
13:09sofusi based my name on what the tracer called it, so metadata per tile. And thats what i think needs to change (at least it does for avd)
13:10sofus32 for the y plane and 8 for uv
13:10chadmedthe tiles are bigger for y too right
13:10chadmedor is it just the metadata
13:12sofusyes, its only the metadata thats not the same as the dcp driver
13:13chadmedmaybe the dcp driver is wrong for yuv too then
13:15sofusim not sure, i have not been able to test it yet
13:18hdbngr has joined #asahi-dev
13:19sofusi find it kinda annoying that i have to add a new pixel format just to describe "this format but compressed"
13:32xal_ has joined #asahi-dev
13:35xal has quit [Ping timeout: 480 seconds]
13:35hdbngr has quit [Ping timeout: 480 seconds]
13:38hdbngr has joined #asahi-dev
13:47hdbngr has quit [Ping timeout: 480 seconds]
13:47sofusthat looks right, im also seing the CompressionMetadataPatternSignallingConstant in the metadata
13:47weirdtreething has quit [Ping timeout: 480 seconds]
13:49hdbngr has joined #asahi-dev
13:58hdbngr has quit [Remote host closed the connection]
13:58hdbngr has joined #asahi-dev
14:08turboretor1 has quit [Ping timeout: 480 seconds]
14:09hdbngr has quit [Ping timeout: 480 seconds]
14:12hdbngr has joined #asahi-dev
14:17jannau(mostly) everything still works after replacing "apple,dma-range"
14:29sven„mostly“? :D
14:31hdbngr has quit [Ping timeout: 480 seconds]
14:33hdbngr has joined #asahi-dev
14:36svengithub.com/AsahiLinux/linux/tree/tbt-reset-wip compile tested only, merged janne's commits so that this should work on m2 pro/... now as well
14:47jannautwo early. old version contains a severe bug, fix pushed
14:51hdbngr has quit [Ping timeout: 480 seconds]
14:52svenugh, okay. just the iommu series commits?
14:53jannauyes, nothing catastrophic just a "
14:54jannau"||" vs. "&&" mixup in "if (args->args_count != 1 && args->args_count != 5)"
14:54svenack, pushed again with the new version
14:55jannauseries sent
14:55svenexcept for dt-bindings i'm mostly happy with that branch nw
15:00jannauworks on m2 pro
15:04chadmed has quit [Read error: Connection reset by peer]
15:04chadmed has joined #asahi-dev
15:05turboretor1 has joined #asahi-dev
15:06chadmed has quit [Read error: Connection reset by peer]
15:06chadmed has joined #asahi-dev
15:15hdbngr has joined #asahi-dev
15:27nela has quit [Quit: Ping timeout (120 seconds)]
15:27nela has joined #asahi-dev
15:32hdbngr has quit [Ping timeout: 480 seconds]
15:37turboretor1 has quit [Read error: Connection reset by peer]
15:37turboretor1 has joined #asahi-dev
15:40hdbngr has joined #asahi-dev
15:46weirdtreething has joined #asahi-dev
15:48hdbngr has quit [Ping timeout: 480 seconds]
15:57hdbngr has joined #asahi-dev
16:02svenjannau: did you mention that usb3 devices after a tbt connection don't work anymore? i can't seem to reproduce that
16:06jannausven: usb3 works now after tbt. I thought your thunderbolt fixes were supposed to fix it
16:06turboretor1 has quit [Ping timeout: 480 seconds]
16:07svenmaybe they fixed that by accident
16:07svenwhat they were supposed to fix was tbt devices preventing acio shutdown
16:07pb17 has quit [Ping timeout: 480 seconds]
16:08sventhey shouldn't affect xdomain connections I think
16:08svenbut either, happy it's fixed :D
16:11jannaustill hits "phy-apple-atc f03000000.phy: Pipehandler lock not acked" and WARN_ON_ONCE(atcphy->pipehandler_up); in atcphy_mux_set() on usb3 connect
16:11hdbngr has quit [Ping timeout: 480 seconds]
16:11svenhm, i see that on tbt disconnect here
16:11jannaudevices works though
16:11turboretor1 has joined #asahi-dev
16:12jannauthe "phy-apple-atc f03000000.phy: Pipehandler lock not acked" was from the usb3 disconenct of a different port
16:13svenyeah, i hit that as well
16:13svenmaybe we don't need the lock dance on disconnect? we're about to reset the entire thing anyway
16:15jannautiming wise the WARN_ON_ONCE() appears to be on the newly connected port (tbt disconnect, usb3 connect)
16:16sveni'm not too concerned if it works because pipehandler is just weird, but i thought we got rid of most of these warnings by now
16:18pb17 has joined #asahi-dev
16:25hdbngr has joined #asahi-dev
16:30turboretor1 has quit [Ping timeout: 480 seconds]
16:35jannauwe're hitting the same issues in asahi-wip so unrelated
16:38hdbngr has quit [Remote host closed the connection]
16:39hdbngr has joined #asahi-dev
16:40turboretor1 has joined #asahi-dev
16:49hdbngr has quit [Ping timeout: 480 seconds]
16:51sven"good"
16:53hdbngr has joined #asahi-dev
16:58svenchaos_princess: if you have some time, can you try with your cables again as well? i've completely rewritten the cable vse nonsense and it should match whatever xnu does now
16:59chaos_princesssure, will do later today
17:01svenxnu takes it directly from datastatus/status but i can't pass those through to anything outside tipd
17:01svenbut in theory datastatus/status should come from that EUDO message so this should hopefully be fine
17:34hdbngr has quit [Ping timeout: 480 seconds]
17:34hdbngr has joined #asahi-dev
17:51hdbngr has quit [Ping timeout: 480 seconds]
17:56hdbngr has joined #asahi-dev
18:08jannaucommenting apple_dart_domain_flush_tlb in apple_dart_iotlb_sync_map increases usb4net througput by 30%
18:09hdbngr has quit [Ping timeout: 480 seconds]
18:09svenat least for the cpu pagetables you don't need to invalidate anything after adding new mappings, only when deleting old ones
18:10hdbngr has joined #asahi-dev
18:10nela has quit [Ping timeout: 480 seconds]
18:16jannauDMA-FQ helps now more. 13.2 Gbits/sec for Linux as receiver in perf3
18:16nela has joined #asahi-dev
18:19hdbngr has quit [Ping timeout: 480 seconds]
18:19sveni guess we should add support for dma-fq to dart anyway
18:19sveni think all we need to do it set some bit in capabilities somewhere since all we can do is "invalidate everything" anyway
18:21hardy94 has joined #asahi-dev
18:23jannauisn't that just "iommu.strict" config symbol or command line option?
18:24svenlast i checked that won't work because we claim we can't do dma-fq
18:24sventhat sysfs write just bypasses that check :D
18:27hdbngr has joined #asahi-dev
18:28sveni think we need to add a capable function that returns true for IOMMU_CAP_DEFERRED_FLUSH
18:35hdbngr has quit [Ping timeout: 480 seconds]
18:43hdbngr has joined #asahi-dev
18:45jannauyes, that does it. I thought strict was the default but fedora uses lazy
18:47chaos_princesssven: both work
18:47svennice, thanks for confirming!
18:56hdbngr has quit [Remote host closed the connection]
18:56hdbngr has joined #asahi-dev
19:12hdbngr has quit [Ping timeout: 480 seconds]
19:25pjv has joined #asahi-dev
19:28hdbngr has joined #asahi-dev
19:36hdbngr has quit [Ping timeout: 480 seconds]
19:39hdbngr has joined #asahi-dev
19:49hdbngr has quit [Ping timeout: 480 seconds]
19:55jannauusing DART's stream remap instead of 2 iommus doesn't increase the iperf3 throughput on t6020
19:57sveni wonder if we can make the RX path re-use buffers that stay mapped the entire time as well
19:58sveni guess stream map internally then doesn't adjust the stream id but just mirrors the configuration (and invalidations) :(
19:58sven*remap
19:59jannauI think it is the way it is to support zero-copy network packet handling
20:00___nick___ has quit [Ping timeout: 480 seconds]
20:00chaos_princessso, is tx on stream 1 and rx on stream 2?
20:21jannau16.0 Gbits/sec / 23.5 Gbits/sec with 13 additional chars: "dma-coherent;"
20:21svenhah, nice!
20:22___nick___ has joined #asahi-dev
20:24pjv has quit [Ping timeout: 480 seconds]
20:26svenchaos_princess: i think both rx and tx are on stream 2 and just the control ring is on stream 1
20:26jannauand the linux transmit case could be faster than 23.5gbps. I'm blaming retransmits new: https://paste.centos.org/view/5d2bd097
20:26svenotherwise we should've seen traffic in one direction before we found that secondary stream
20:29svenwow, 27.5gbps during peaks
20:29sven28gbps even
20:30svenhow fast is this on non-apple machines again? :>
20:33chaos_princessisn't it "as fast as pcie 3x4 + overhead"?
20:34jannauhasn't intel integrated tbt controller in recent CPUs?
20:40chaos_princessapparently hangs off the chipset
20:41chaos_princessactually, ignore me, i can't read, it is on cpu
20:41chadmed has quit [Read error: Connection reset by peer]
20:41chadmed has joined #asahi-dev
20:44jannaucurrently seeing 18.0 / 21.2 gbps between 13th gen core i7 and t8103 but I think I saw more a couple of days ago
20:47jannau"30 / 24 Gbit/s" from the backlog
20:49jannaustill only "RTKit: syslog message: high_speed_lane.c:289: Gen2/3 link error. lane=1, error=89" spam when connecting the intel notebook to t6020/linux
20:51jannauwould an acio trace of intel/macos be helpful?
20:55svenyes, from the macos side
20:55svenand then Intel/Linux where that thing above happens from the Linux side as well
20:56svenand then we get to play another game of spot the difference that’s gonna end up in something dumb like the second dart stream 🙃
21:02jannauin turn of events macOS logs "[ACIO1:high_speed_lane.c:289] Gen2/3 link error. lane=1, error=89"
21:12jannauworks with a different cable
21:13barusu_ has joined #asahi-dev
21:13barusu_ has quit [Remote host closed the connection]
21:14barusu_ has joined #asahi-dev
21:16jannauboth with macos and linux. intel <> t6020/linux iperf throughput is 17.4 / 20.5 Gbits/sec
21:18barusu has quit [Ping timeout: 480 seconds]
21:22barusu_ is now known as barusu
21:34bgb has quit [Remote host closed the connection]
21:34bgb has joined #asahi-dev
22:56hardy94 has quit [Quit: hardy94]
22:59hardy94 has joined #asahi-dev
…