#asahi-dev · 2022-09-26 13:13
mazyou just need to explain to the arm64 mm subsystem that the IOMMU only has 6 bits of ASIDs...#asahi-dev · 2022-09-26 13:12
mazI think this is an interesting avenue to explore. you need a userspace context to perform the TLBI anyway, and pinning the ASID between the CPU task and the GPU makes some sense.#asahi-dev · 2022-09-26 13:11
mazyeah, something like it#asahi-dev · 2022-09-26 13:10
mazcan't please everyone! :D#asahi-dev · 2022-09-26 13:09
mazI'm happy to limit that to 1 (or less... ;-)#asahi-dev · 2022-09-26 13:08
mazcan it do with less?#asahi-dev · 2022-09-26 13:07
mazonly problem is that M1 is pretty poor on the ASID front.#asahi-dev · 2022-09-26 13:07
mazbut it wouldn't take much to turn this into a reservation mechanism.#asahi-dev · 2022-09-26 13:06
mazcurrently, this is only used to ensure that the ASID allocated to a process doesn't change, so that SMMU and MMU are kept in sync.#asahi-dev · 2022-09-26 13:05
mazgrep for pinned_asid_map#asahi-dev · 2022-09-26 13:04
mazdigs#asahi-dev · 2022-09-26 13:04
mazthat'd be a marginally different usage of the feature, but that should hopefully help.#asahi-dev · 2022-09-26 13:03
mazyou just mark some ASIDs as resserved, IIRC.#asahi-dev · 2022-09-26 13:03
maz(sharing the userspace PTs with the SMMU)#asahi-dev · 2022-09-26 13:03
mazyeah, for SVA.#asahi-dev · 2022-09-26 13:03
mazwe already have that.#asahi-dev · 2022-09-26 13:02
mazdon't they have the notion of ASID in the GPU?#asahi-dev · 2022-09-08 13:02
mazalso, if you think os-x is better, go use it. we're not presecriptive.#asahi-dev · 2022-09-08 11:07
mazjannau: pinging Joerg to make sure that this is on his radar wouldn't hurt.#asahi-dev · 2022-09-05 14:56
mazsven: I rest my case! :D#asahi-dev · 2022-09-05 14:54
mazstill with Intel AFAIK. overall, I tend to ignore his comments for the stuff I care about, due to the very poor signal/noise ratio.#asahi-dev · 2022-09-05 14:50
mazpovik: don't attribute to malice...#asahi-dev · 2022-09-05 13:56
mazwsx[m]: xen definitely doesn't support NV on arm64. as for VMware, ask them.#asahi-dev · 2022-09-05 13:54
mazj`ey: also, debugging NV on a laptop without serial console isn't my idea of fun!#asahi-dev · 2022-09-05 13:53
mazj`ey: I need to rebase it and address the pending comments, which will likely take me another couple of weeks, at which point I'll be on holiday...#asahi-dev · 2022-09-05 13:52
mazlooks somewhere else.#asahi-dev · 2022-09-05 13:48
mazkettenis: yeah, the current trend has been worrying me for some time. the more we wait to submit this upstream, the more difficult it becomes. #asahi-dev · 2022-09-05 13:36
mazGlanzmann: hey, it's not my laptop, and I could do *without* having him on my back when KVM breaks.#asahi-dev · 2022-09-05 13:35
mazyeah. maybe I'll convince him to run arm64 guests, just to reduce the potential pain.#asahi-dev · 2022-09-05 13:34
mazplumbers, kvm forum, maintainer summit, security summit...#asahi-dev · 2022-09-05 13:32
mazwell, he's true to his earlier plan.#asahi-dev · 2022-09-05 13:29
mazgah, pinhead travelling to Dublin with his M2 only. no pressure... :-/#asahi-dev · 2022-09-05 08:40
maz-march=native doesn't mean much, really, as most of the features cannot be discovered from userspace.#asahi-dev · 2022-08-31 10:07
mazsame EFI problems, by the way -- half-baked QC stuff.#asahi-dev · 2022-08-31 10:06
mazkettenis: I still have an older c630 as a test machine, which also has a nice trackpoint. OK for email, but otherwise painful.#asahi-dev · 2022-08-30 17:37
mazsven: make it exponential!#asahi-dev · 2022-08-17 15:06
mazdoesn't u-boot have a minimal stack?#asahi-dev · 2022-08-17 09:41
mazthat's what this is on M1.#asahi-dev · 2022-08-17 09:41
maz"Instruction issued from scheduler"#asahi-dev · 2022-08-17 09:39
mazcould well be. what's the event number?#asahi-dev · 2022-08-17 09:32
mazusually, CPU folks tend to *add* them...#asahi-dev · 2022-08-17 09:32
maz_jannau_: OK. interesting that they have removed an event. I wonder if that event was actually something on M1.#asahi-dev · 2022-08-17 09:24
mazjannau: seen your branch the for M2 PMU appearing. have you checked whether the events are the same?#asahi-dev · 2022-08-16 08:21
mazmarcan: that's a QC debugging special.#asahi-dev · 2022-08-09 11:19
mazpovik: AFAIK, it is mostly used in crypto code.#asahi-dev · 2022-08-09 09:21
mazsven: don't go there, please...#asahi-dev · 2022-08-09 09:21
mazruns...#asahi-dev · 2022-08-09 09:21
mazyou mean, like a BPF program?#asahi-dev · 2022-08-08 10:52
mazmarcan: with pinhead as a customer, you want to be *very* careful when breaking backward compat.#asahi-dev · 2022-08-05 11:07
mazit looks totally unrelated.