#
✕ nick:maz
#asahi · 2021-02-24 10:57
mazI can see why they'd leave that off in their hypervisor.
#asahi · 2021-02-24 10:57
mazmarcan: what would be interesting is to find out whether HCR_EL2.TIDCP is honored when these non-architectural features are used at EL0/EL1.
#asahi · 2021-02-24 10:56
mazmarcan: here's to hope! :D
#asahi · 2021-02-24 10:55
mazfor TGE==0, this should already trap to EL2 when HCR_EL2.TIDCP is set.
#asahi · 2021-02-24 10:53
mazmarcan: if there is no way to disable it from EL2 when HCR_EL2.TGE==1, we'll need to wipe this register on context switch. really ugly.
#asahi · 2021-02-22 12:35
maznever_released: ARMv8.5 minus BTI, minus architecture compliance. At which stage, it doesn't mean much...
#asahi · 2021-02-22 11:41
mazbecause*
#asahi · 2021-02-22 11:41
maz(vecause the guest could run with MMU off or a NC mapping)
#asahi · 2021-02-22 11:41
mazthe killer is the PoC flush on each page being mapped into the guest.
#asahi · 2021-02-22 11:41
mazicache*
#asahi · 2021-02-22 11:41
mazcicache coherency is isn't that big a deal.
#asahi · 2021-02-22 11:39
mazno more silly cache maintenance!
#asahi · 2021-02-22 11:39
mazmarcan: I'm not surprised. they have also implemented ARMv8.4-FWB, which is going to be a godsend for KVM.
#asahi · 2021-02-22 09:36
mazit may be worth looking at what x86 has done in their own corner: Documentation/x86/pat.rst
#asahi · 2021-02-22 09:32
mazsounds very good to me! cc Jon Corbet for this.
#asahi · 2021-02-22 09:31
mazmarcan: but if you can start with something that also describe the MAIR settings for arm64 as an example of an architectural implementqwtion, that'd be great!
#asahi · 2021-02-22 09:30
mazmarcan: I don't think this should be arm64-specific, as the ioremap API should have a consistent meaning across architectures (after all, that's how we write "portable" code).
#asahi · 2021-02-22 09:25
mazdhewg: well, fruit had a meaning before dev boards, so I feel perfectly fine with my comment! :-)
#asahi · 2021-02-22 09:24
mazmarcan: it'd be good if they had a similar uarch. the traditional BL crap is awful to deal with.
#asahi · 2021-02-22 09:23
mazdhewg: RPi doesn't design CPUs.
#asahi · 2021-02-22 09:23
mazjn__: feel free to point this out on the list in reply to https://lore.kernel.org/kvmarm/20210208095732.3267263-24-maz@kernel.org/
#asahi · 2021-02-22 09:20
mazthe really ugly part is the interaction with the cpufeature override.
#asahi · 2021-02-22 09:19
mazmarcan: I'm happy with it for now, and we can always have some discussion on the list. I expect Mark to nitpick at it, but that's the rule of the game!
#asahi · 2021-02-22 09:18
mazmarcan: my sentiment as well. you don't have this "bug" by accident, you have to design it as such. and write all your SW accordingly.
#asahi · 2021-02-22 09:16
mazjn__: no idea.
#asahi · 2021-02-22 09:06
mazmarcan: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit/?h=hack/arm64-early-cpufeature&id=d4d26a1a61a81d4b3eecd5fb784d0f512bc36c7d
#asahi · 2021-02-22 09:06
mazmarcan: the delayed VHE setup code has gone in mainline, so you probably want to take this patch as part of your next round of patches:
#asahi · 2021-02-19 10:40
mazthough it isn't clear whether the SBU pins are wired anywhere. you'd probably need a USB-C breakout board to extract the uart pins.
#asahi · 2021-02-19 09:59
mazdamn. it's cheap, even...
#asahi · 2021-02-16 17:28
mazmodwizcode: the spectrum of the architecture is wide enough to avoid adding silly implementors' hacks.
#asahi · 2021-02-16 17:28
mazmodwizcode: it's a deliberate choice.
#asahi · 2021-02-16 17:27
mazmarcan: they see the real MIDR value to be able to apply workarounds.
#asahi · 2021-02-16 17:27
mazso there will be no PMU for M1 guests under KVM.
#asahi · 2021-02-16 17:26
mazKVM only presents architectural features to guest.
#asahi · 2021-02-16 17:07
maznever_released: well, it would turn a panic in a livelock, but this is a pretty theoretical case. it is likely that we'll do without all of that anyway.
#asahi · 2021-02-16 17:05
maznever_released: what difference? VM asks for a group-0 interrupt, VM gets a group-0 interrupt...
#asahi · 2021-02-16 17:04
maznever_released: only if these VMs get a FIQ injected, which shouldn't happen if they keep their Group-0 disabled.
#asahi · 2021-02-15 13:44
mazat*
#asahi · 2021-02-15 13:44
mazwe're really good an ignoring email! :D
#asahi · 2021-02-12 18:22
mazmarcan: do the scaling according the the dmips-capacity! :D
#asahi · 2021-02-12 13:04
mazmarcan: plus, I can't see penguins over a serial console (please, don't... ;-)
#asahi · 2021-02-12 13:03
mazmarcan: I don't have much hate left in me, so the little I have I keep it for politicians ;-)
#asahi · 2021-02-12 11:15
mazGlanzmann: darn.
#asahi · 2021-02-12 10:49
mazarnd: I have one of those, but then the next obstacle is keyboard input. Not to mention that logging kernel messages and grepping through them becomes interesting with video capture... ;-)
#asahi · 2021-02-12 10:45
mazarnd: framebuffer isn't really an option for me, as the box will live in the rack with the rest of the zoo, where there isn't any display available.
#asahi · 2021-02-12 10:45
mazarnd: I know about this. Started ordering the various bits to put together a UBS-C dongle as described, but bits are slow to trickle in (and my M1 lands on Monday...).
#asahi · 2021-02-12 10:40
mazjix: ah! a bit over the top for me! :D
#asahi · 2021-02-12 10:36
mazGlanzmann: is that an off-the-shelf serial port, or something built for you?
#asahi · 2021-02-11 14:08
mazmarcan: aliasing with different attributes can cause havoc in some interconnects.
#asahi · 2021-02-11 14:05
mazI'd be tempted to use nGnRnE for fixmap device mappings by default if I was sure it wouldn't introduce any regression...

plus ›