#asahi-dev · 2026-08-15 09:10
chaos_princessif irq is the bottleneck, wouldn't using 9000mtu affect it (it didn't)#asahi-dev · 2026-08-15 09:08
svenmy current local copy is probably slightly more efficient in the irq path but that shouldn't explain >2x difference in RX vs TX#asahi-dev · 2026-08-15 09:07
svenyeah, possibly#asahi-dev · 2026-08-15 09:06
jannauI'd expect that the intel laptop uses an iommu as well so it could only be an issue how well iommu and driver handle the usage pattern#asahi-dev · 2026-08-15 09:03
svenoh well, was worth a quick try#asahi-dev · 2026-08-15 09:03
jannausven: when forcing usb4_dart to identity tbt breaks with "thunderbolt-apple-nhi 501f00000.nhi: error -ENODEV: Failed to init software connection manager"#asahi-dev · 2026-08-15 08:57
svenhates thunderbolt #asahi-dev · 2026-08-15 08:57
svenlol, fixed the thunderbolt domain leak and i get immediately greeted with a WARN now :D#asahi-dev · 2026-08-15 08:50
svenwhich with VHE gets aliased to the same in EL2#asahi-dev · 2026-08-15 08:49
svenyuka: i bet they're both the same register, some CTRR control just _EL1 and _EL2#asahi-dev · 2026-08-15 08:34
elisaadohmmmm that actually sounds useful af#asahi-dev · 2026-08-15 05:42
JamesCalligeros[m]Sounds cursed :p#asahi-dev · 2026-08-15 05:24
yukafor quickly dumping some memory area using a phone#asahi-dev · 2026-08-15 05:24
yukawebusb m1n1 proxyclient when#asahi-dev · 2026-08-15 04:21
yukaquoting one of sven's messages from april: "much of the s3_0_c11_c smells like CTRR stuff to me"#asahi-dev · 2026-08-15 04:19
yukawait, that should be s3_0_c11_c1_4 and s3_0_c11_c2_0#asahi-dev · 2026-08-15 04:09
yukathat should bring the secondary cores in line with the boot cores, and fix the boot issues#asahi-dev · 2026-08-15 04:09
yukahttps://github.com/AsahiLinux/m1n1/pull/657#issuecomment-5300454583 <-- I suppose I could send a feedback assistant requesting s3_0_c11_c0_12 and s3_0_c11_c0_16 to be zeroed on the secondary cores?#asahi-dev · 2026-08-15 02:22
chadmedsince that was the original behaviour and it got changed before being merged#asahi-dev · 2026-08-15 02:21
chadmedim not sure what to do about the shared gpio. i dont think upstream will want the shared-gpio-proxy behaviour to change to only requiring a vote when trying to deassert the "default value" #asahi-dev · 2026-08-15 00:02
chadmedmaking the cursor plane linear-only fixes all three surfaces on t600x and t8112#asahi-dev · 2026-08-15 00:01
chadmedand once again as per usual the solution is "do what macos does" #asahi-dev · 2026-08-14 21:44
svenand if the RX direction constantly churns iommu mappings and invalidates DART all the time that might just be very expensive #asahi-dev · 2026-08-14 21:44
svenmy suspicion is that usb4net was just not written with an iommu in mind#asahi-dev · 2026-08-14 21:43
svenjannau: if you can try to set the DART to bypass#asahi-dev · 2026-08-14 20:33
jannauusb4stream is faster, 13.5 gbps instead of 7.2#asahi-dev · 2026-08-14 19:43
svenfound the leak, it‘s that commit. just dunno how to fix it yet :/#asahi-dev · 2026-08-14 19:22
svenhmmmm#asahi-dev · 2026-08-14 19:20
jannauthe faster variant is then the linux side transmits data#asahi-dev · 2026-08-14 19:19
jannauthe slower side direction has tcp retransmits (50 - 300) while the faster variant has none. not sure if that is enough to explain the difference#asahi-dev · 2026-08-14 19:13
svenI wonder how well this code was tested and optimized when running with an iommu enabled #asahi-dev · 2026-08-14 19:13
svencan you put the dart in bypass mode?#asahi-dev · 2026-08-14 19:13
yukajannau: that means if I got a match for region-id-2, then its start/end value minus 1*stride should be the ones in the tz_regs?#asahi-dev · 2026-08-14 19:05
svenmust avoid being nerdsniped into figuring out why that one direction is slower #asahi-dev · 2026-08-14 19:04
svenwhoa#asahi-dev · 2026-08-14 19:03
jannaut8103/macos <> intel/linux does 30 / 24 Gbit/s in iperf3 over tbt net#asahi-dev · 2026-08-14 18:57
jannauthose entries are not enabled though. looks the same on t8122. I wouldn't be surpised if those entries even match region-id-2/4 (1 indexed)#asahi-dev · 2026-08-14 18:52
yukawell, if I assume stride = 0x14, then the entries above and below all have start = end = 0x10000000#asahi-dev · 2026-08-14 18:50
yukaand do not even make for reasonable start and end#asahi-dev · 2026-08-14 18:50
yukathe values below and above it do not match with any region-id-*#asahi-dev · 2026-08-14 18:49
yukaand what is the stride#asahi-dev · 2026-08-14 18:49
yukabut I'm not sure if this is supposed to be the first entry now#asahi-dev · 2026-08-14 18:49
yukathe mcc regs on t8132 are funny: mcc[7]+0x704 = 0x103e6da8, mcc[7]+0x708 = 0x103ffda7, which would be start and end of carveout-memory-map region-id-2#asahi-dev · 2026-08-14 18:19
svenand leaking a domain isn't fatal on intel unlike on apple#asahi-dev · 2026-08-14 18:19
sventhat was introduced after the previous tbt branch that worked and it already had a bunch of subtle fixes#asahi-dev · 2026-08-14 18:19
sveni'm willing to bet it's d6d458d42e1e ("thunderbolt: Handle DisplayPort tunnel activation asynchronously")#asahi-dev · 2026-08-14 18:00
svenyup, it's hanging at wait_for_completion(&anhi->nhi.domain_released) and something still has a reference to that domain#asahi-dev · 2026-08-14 17:50
svenhah nice #asahi-dev · 2026-08-14 17:46
Rafaysven: It was almost entirely m1n1 overhead. After not tracing acio like jannau said, I get 15Gbps which is the max the other host supports.