#
✕ nick:sven
#asahi-dev · 2026-08-13 19:51
svenand iirc dart only faults on enabled streams
#asahi-dev · 2026-08-13 19:51
sventhat's more than enough, each entry is 16 bytes and i only need the first one
#asahi-dev · 2026-08-13 19:43
svenas expected :(
#asahi-dev · 2026-08-13 19:39
sven*tries
#asahi-dev · 2026-08-13 19:39
sveni'd kinda also expect acio to just panic if DART doesn't allow a write it tires
#asahi-dev · 2026-08-13 19:38
svenlet's just enable every single stream
#asahi-dev · 2026-08-13 19:38
svenchaos_princess: ah.. :(
#asahi-dev · 2026-08-13 19:37
svenand i can see from those hw regs if something arrived or was picked up by acio
#asahi-dev · 2026-08-13 19:37
sveni expect that the moment we fix this stupid crap it's just gonna work like m1/m1
#asahi-dev · 2026-08-13 19:36
svenwaaaait. can you hack the dart driver to enable all streams so that we see faults?
#asahi-dev · 2026-08-13 19:36
sveni would've expected 0xff there if nothing ever arrived
#asahi-dev · 2026-08-13 19:35
sven0x100ff
#asahi-dev · 2026-08-13 19:35
sveni wanna see if there's actually a single entry in that ring due to that here >>> read32(0xb01f84008)
#asahi-dev · 2026-08-13 19:35
svenack
#asahi-dev · 2026-08-13 19:35
svenah, then that please for the dart :D
#asahi-dev · 2026-08-13 19:34
svencan you do iova = read32(0xb01f84004) << 32 | read32(0xb01f84000) ... ah crap, do we have something to read a mapped buffer through dart?
#asahi-dev · 2026-08-13 19:33
svenyeah, because the kernel put more stuff in there for the hardware to ignore :(
#asahi-dev · 2026-08-13 19:30
svenwhich is weird
#asahi-dev · 2026-08-13 19:30
sventhird one shows the kernel prepared 0xff buffers and uhh.. the hardware sent at most one
#asahi-dev · 2026-08-13 19:29
svennext one shows the acio never fetched even a single entry from the txring even though the kernel sent 0xe in there :/
#asahi-dev · 2026-08-13 19:29
svenfirst one is txring0 and PROD=CONS=1 so it's working
#asahi-dev · 2026-08-13 19:27
svenlast one confirms there's no interrupt pending so the issue isn't a wrong irq number in the dt
#asahi-dev · 2026-08-13 19:26
svenif we saw interrupts then the packets would arrive and something else would be broken
#asahi-dev · 2026-08-13 19:25
svenbut we should see interrupts on ring1 and chaos' /proc/interrupts showed there was nothing there
#asahi-dev · 2026-08-13 19:24
svenit's essentially ethernet frames with an additional header iirc
#asahi-dev · 2026-08-13 19:22
svenon the t6001
#asahi-dev · 2026-08-13 19:22
svenread32(0xb01f84008)
#asahi-dev · 2026-08-13 19:22
svenplease :)
#asahi-dev · 2026-08-13 19:22
svenread32(0xb01fd0000)
#asahi-dev · 2026-08-13 19:22
svenread32(0xb01f14008)
#asahi-dev · 2026-08-13 19:22
svenread32(0xb01f10008)
#asahi-dev · 2026-08-13 19:20
svenif yes, which acio port are you using?
#asahi-dev · 2026-08-13 19:20
svenare you still in the hv on the t6001?
#asahi-dev · 2026-08-13 19:15
svenand i assume you also see no packets in tcpdump on either side?
#asahi-dev · 2026-08-13 19:15
sveni'm really running out of ideas at this point
#asahi-dev · 2026-08-13 19:14
svencan you compare the atcphy bringup that macos does vs. what we do?
#asahi-dev · 2026-08-13 19:12
sven:/
#asahi-dev · 2026-08-13 18:50
svenas long as it kinda-sorta-works i'm happy to send a v1 even without lane bonding though
#asahi-dev · 2026-08-13 18:50
svenbut will have to check again
#asahi-dev · 2026-08-13 18:50
sveni thought i saw in one trace that it failed to negotiate lane bonding and i also saw some syslog messages
#asahi-dev · 2026-08-13 18:49
svenit just wasn't the root cause of the m1 pro issue
#asahi-dev · 2026-08-13 18:49
svenoh right, that was broken for you as well
#asahi-dev · 2026-08-13 18:45
svennow there's that m1 pro mystery left and then cleaning all this up. and dt-bindings ofc...
#asahi-dev · 2026-08-13 18:44
sven:)
#asahi-dev · 2026-08-13 18:18
sven*your
#asahi-dev · 2026-08-13 18:18
svenhttps://lore.kernel.org/asahi/20260813-b4-tipd-vdo-fix-v1-1-70317f2cd554@kernel.org/T/#u tipd fix, if you wanna send you tested-by Rafay. i removed one redundant line from the version you tested fwiw
#asahi-dev · 2026-08-13 18:07
svenack
#asahi-dev · 2026-08-13 18:07
sveni don't think any of the "fixes" we tried did anything
#asahi-dev · 2026-08-13 18:04
svenfabric firewall that's only enabled on his machine in raw boot mode? :P
#asahi-dev · 2026-08-13 18:02
svenuh oh

plus ›