#asahi-dev · 2022-01-06 13:41
svenyes#asahi-dev · 2022-01-06 13:41
svenwithout that there’s also no NVMMU#asahi-dev · 2022-01-06 13:41
svenwe have to enable it by flipping a bit to 1#asahi-dev · 2022-01-06 13:41
svenI’ll have to check#asahi-dev · 2022-01-06 13:40
sven🤷🏻♂️#asahi-dev · 2022-01-06 13:39
sveni think we had this discussion before ;)#asahi-dev · 2022-01-06 13:39
svenbehaves like the T2 chip then#asahi-dev · 2022-01-06 13:39
svenyup#asahi-dev · 2022-01-06 13:39
svenit then uses the classes called ANS2#asahi-dev · 2022-01-06 13:39
svenyou can disable linear-sq with a boot arg#asahi-dev · 2022-01-06 13:39
svenyeah, boot arg#asahi-dev · 2022-01-06 13:38
svenand some drivers bind to node names#asahi-dev · 2022-01-06 13:38
svenMarcan: ans2 vs 3 is decided by the enable-linear-sq boot arg fwiw#asahi-dev · 2022-01-06 13:37
svensure#asahi-dev · 2022-01-06 13:30
svenI don’t care either way#asahi-dev · 2022-01-06 13:30
svenor maybe even ans3 if ioreg is to be trusted #asahi-dev · 2022-01-06 13:29
sveni was just typing that :D#asahi-dev · 2022-01-06 13:29
svenYeah#asahi-dev · 2022-01-06 13:29
svenuh, sure#asahi-dev · 2021-12-31 10:45
svenI still wonder what that hack is all about#asahi-dev · 2021-12-31 10:41
svenoh, I think as long as there is that bypass property in the ADR the DART should still have that mode #asahi-dev · 2021-12-31 10:40
svenI wonder if they still have that stream 15 bypass hack(?) then#asahi-dev · 2021-12-31 07:20
svenI didn’t even realize that’s a standard reg#asahi-dev · 2021-12-30 22:22
svenhttps://github.com/AsahiLinux/m1n1/blob/main/src/hv_exc.c#L142 That should be the fastest way to shut up#asahi-dev · 2021-12-30 11:57
svennice, thanks for testing!#asahi-dev · 2021-12-29 22:52
sven(and ofc whatever the nvme maintainers will request)#asahi-dev · 2021-12-29 22:52
svenjust bugfixes, im happy with it otherwise#asahi-dev · 2021-12-29 15:56
svenyeah, I might’ve dropped that by accident when removing all pci specific code#asahi-dev · 2021-12-29 12:52
svensounds like something I should do in the Linux driver as well#asahi-dev · 2021-12-29 12:08
svenI’m pretty sure I setup those three early on though #asahi-dev · 2021-12-29 12:07
svenI’m starting to wonder if that reset actually resets anything at all. #asahi-dev · 2021-12-29 11:02
svenit looks like something goes wrong with intitializing the admin queue though. in the crashlog the IO queue still looked as if it was the one used by uboot i think#asahi-dev · 2021-12-29 11:01
sventpw_rules: haven’t looked at it yet#asahi-dev · 2021-12-24 13:05
svenIf you flip those everything in dwc3 will reset to its defaults and the port will work again#asahi-dev · 2021-12-24 13:05
sven0x382A8401C bit 0 and 4 I think #asahi-dev · 2021-12-24 13:01
sventhe thing apple calls “ATC PHY”#asahi-dev · 2021-12-24 13:01
svenso essentially tear everything down, flip reset bit somewhere in the phy and then bring everything up again#asahi-dev · 2021-12-24 13:00
svenyou’ll also have to do a hard reset in the PHY#asahi-dev · 2021-12-24 13:00
sventhat’s not even enough :(#asahi-dev · 2021-12-24 00:02
svenanyway, hope your holidays are enjoyable as well :)#asahi-dev · 2021-12-24 00:01
svenbut that stuff still shouldn’t be happening. It appears as if it never gets the irq for the first admin command. It’s strange it tries to identify before setting up the new io queue though #asahi-dev · 2021-12-24 00:00
svenit looks like the controller isn’t completely reset (the old io queue still seems to exist)#asahi-dev · 2021-12-23 23:53
svenHmm… looks like a bug anyway. I’ll take a look after christmas #asahi-dev · 2021-12-23 23:43
svenAnd which machine is that? M1?#asahi-dev · 2021-12-23 23:43
svenuh, dmesg please. #asahi-dev · 2021-12-23 14:41
sventhat’s good and bad news I guess :D#asahi-dev · 2021-12-23 13:46
sventhanks, that's already all i wanted to know :)#asahi-dev · 2021-12-23 13:45
svendon't think so#asahi-dev · 2021-12-23 13:45
svenyou can just ctrl+c then. looks like it behaves as bad as it does on the M1#asahi-dev · 2021-12-23 13:39
svenit'll benchmark nvme performance for random writes w/ flushes which are still very slow on the M1 with my driver