#
✕ nick:maz
#asahi-dev · 2022-03-30 11:06
mazit is also an ABI issue. Not specifying 'exclude_guest' implies that you can count guest events should you run one. however, we don't support it.
#asahi-dev · 2022-03-30 11:04
mazYichaoYu[m]: the manpage is pretty much unmaintained (as is most of the documentation). Also, it is hard to find out whether you run a VM in a race-free way.
#asahi-dev · 2022-03-30 08:09
mazYichaoYu[m]: 'exclude_guest' is expected, we don't support guest profiling (or any sort of PMU virtualisation). and as jannau pointed out, you must explicitly select the PMUs you want to count on.
#asahi-dev · 2022-03-28 13:52
mazjannau: ah, I hadn't realised it was merged already. no need to touch that then.
#asahi-dev · 2022-03-28 13:45
mazjannau: IIRC, there was a bunch of u-boot changes to support that. I wonder if we could get rid of that by letting m1n1 switch to x8r8g8b8.
#asahi-dev · 2022-03-25 14:13
mazat least if you want it to appear as a runtime service, that is.
#asahi-dev · 2022-03-25 14:13
mazj`ey: you'll need u-boot to expose the service, no matter where it is implemented.
#asahi-dev · 2022-03-25 14:11
mazSOC_ID*
#asahi-dev · 2022-03-25 14:10
mazyeah. instead, we have ended up with a DOC_ID firmware service, which of course cannot be implemented on M1.
#asahi-dev · 2022-03-25 14:09
mazthere were discussion of an architected SOCIDR at some point, but nobody could agree on what that would return...
#asahi-dev · 2022-03-25 14:07
mazsounds like it indeed. it feels like Apple abuses the MIDR to also designate the SoC used.
#asahi-dev · 2022-03-25 12:08
mazso at the moment, we only mitigate on t8103, and a guest could take down the machine otherwise (assuming the other cores are equally broken).
#asahi-dev · 2022-03-25 12:06
mazjannau: KVM uses that to work out which CPUs need to have workarounds applied.
#asahi-dev · 2022-03-25 12:01
mazjannau: sure. but since I had my eyes on that... ;-)
#asahi-dev · 2022-03-25 12:00
mazmarcan: do you have any information about the CPUs on Ultra? they seem to be of a different model compared to the previous ones (0x28/0x29 vs 0x22/0x23).
#asahi-dev · 2022-03-25 11:34
mazsimilarly, the PMU affinities in the AIC node should reflect all the CPUs (you only have the t6001 ones atm).
#asahi-dev · 2022-03-25 11:34
mazjannau: small nit on the t6002 DT: the PMU nodes should be per CPU types, not per node (you should only have two PMU nodes in total).
#asahi-dev · 2022-03-24 18:19
mazsven: ah, true.
#asahi-dev · 2022-03-24 18:13
mazscaryh.
#asahi-dev · 2022-03-24 18:13
mazXHCI_NO_64BIT_SUPPORT
#asahi-dev · 2022-03-24 18:09
mazsven: I get these messages on the mini too. that's because there is no mapping for the ports themselves.
#asahi-dev · 2022-03-24 18:08
mazjannau: not the same controller as the mini. mini has a FL1100, while this is an ASM2142
#asahi-dev · 2022-03-24 17:57
mazASmedia controller. I was hoping some something better, given the price of the machine...
#asahi-dev · 2022-03-24 17:45
mazOK, cool.
#asahi-dev · 2022-03-24 17:42
mazjannau: do you see interrupts being taken on the CPUs sitting on the other die? other than timer interrupts, that is?
#asahi-dev · 2022-03-24 17:42
mazyup. now the fun begins to make Linux grok this crazy interrupt (non-)routing.
#asahi-dev · 2022-03-24 17:37
mazright, that's twice as fast as my current box, with a third of the CPUs. pretty much what I expected given the speed of my mini.
#asahi-dev · 2022-03-24 17:31
mazI guess I'll have to check whether the CPUs are affected by the same bug or not.
#asahi-dev · 2022-03-24 17:30
mazinteresting, the MIDRs are different. which means the SEIS workaround for the ICV registers isn't applied.
#asahi-dev · 2022-03-24 17:28
mazjannau: thanks!
#asahi-dev · 2022-03-24 17:26
mazlalala
#asahi-dev · 2022-03-24 17:26
mazit doesn't mean it's free.
#asahi-dev · 2022-03-24 17:25
mazindeed. if you don't put the NUMA description in DT, the kernel doesn't know about it.
#asahi-dev · 2022-03-24 17:23
maz(i.e. there is no magic).
#asahi-dev · 2022-03-24 17:23
maztpw_rules: I'd be very surprised if it wasn't NUMA. there is no way the interconnect doesn't add cost to the memory path if accessing cross node.
#asahi-dev · 2022-03-24 17:23
mazjannau: do you mind posting a full dmesg somewhere?
#asahi-dev · 2022-03-22 18:54
mazjannau: cheers. saves me probing the other 5 ports! :D
#asahi-dev · 2022-03-22 18:42
mazbasic m0, a stm32 knock-off that came as part of a USB-PD trigger.
#asahi-dev · 2022-03-22 18:38
mazalyssa: firmware to drive the FUSB302 that enables the UART and allows me to remotely reboot the box. I already have one on the mini, but need a spare for the Studio. and since I can't get the same parts I used the first time, I'm redoing everything from scratch.
#asahi-dev · 2022-03-22 18:31
mazin*
#asahi-dev · 2022-03-22 18:30
mazjannau: keep us posted! I hope to have my serial adapter ready by the time it lands (specially as I'm foolishly trying to write the firmware is bloody Rust...)
#asahi-dev · 2022-03-22 17:37
maz*documenting* why you use barriers or not is just as important as the barriers themselves (or lack thereof).
#asahi-dev · 2022-03-22 17:37
mazand TBF, readin Arnd's emails, I mainly see the later.
#asahi-dev · 2022-03-22 17:36
mazI'd rather people reason about what they are writing and use the right amount of synchronisation.
#asahi-dev · 2022-03-22 17:36
mazmarcan: my point is that barriers are expensive, even on shiny HW like this one, and that *removing* barriers is incredibly hard. it also plays to the mindset of "I don't know what I'm doing, so I'll be adding barriers until it behaves".
#asahi-dev · 2022-03-22 17:34
mazmarcan: we've had that battle for a long time, and I don't think we'll ever agree on that one.
#asahi-dev · 2022-03-22 13:01
mazmarcan: thanks for the braindump. it'd be good to have some if this rational as part of the cover letter or ML discussion, because IRC really isn't the right medium for that (or anything else IMO... ;-)
#asahi-dev · 2022-03-22 10:38
mazsven: I had similar questions about the rtkit driver, which smells a lot like a "standard" mailbox driver...
#asahi-dev · 2022-03-21 16:15
mazI'd tempt it, but really drop the RFC part.
#asahi-dev · 2022-03-21 16:15
mazit really depends on the maintainer, but some become pretty annoyed about large series during the merge window (yours is pretty small, to be honest).

plus ›