#
✕ nick:maz
#asahi-dev · 2021-11-24 09:45
mazit keeps the top bits in some intermediate, non SW visible register, and you start getting page fault in the IOMMU.
#asahi-dev · 2021-11-24 09:44
mazgoes into a very funky state if you program it with a 64bit address and then switch it back to 32bit.
#asahi-dev · 2021-11-24 09:44
mazuPD72020x
#asahi-dev · 2021-11-24 09:42
mazI gave up on xHCI after the Renesas debacle...
#asahi-dev · 2021-11-24 09:41
mazat least it is consistent.
#asahi-dev · 2021-11-24 09:39
maztry booting with pci=nomsi, just to see...
#asahi-dev · 2021-11-24 09:39
mazsome controllers have 'interesting' implementations either way...
#asahi-dev · 2021-11-24 09:38
mazmarcan: using MSI or INTx?
#asahi-dev · 2021-11-23 17:54
mazkettenis: fired.
#asahi-dev · 2021-11-23 16:11
mazkettenis: cheers.
#asahi-dev · 2021-11-23 16:05
mazI'll post patches for that tomorrow (can't really drop the branch merging nightmare I'm in the middle of...)
#asahi-dev · 2021-11-23 15:28
maz:)
#asahi-dev · 2021-11-23 15:27
mazsven: yeah, that. yet another thing to RE, and I'm not holding my breath. Nothing a controlled PDU can't work around.
#asahi-dev · 2021-11-23 15:20
maz"PowerON, PowerOFF, Reset". that's a pretty limited LOM, but hey, I'd take that.
#asahi-dev · 2021-11-23 15:13
maz*blink* does the 10GbE mini have LOM?
#asahi-dev · 2021-11-22 10:56
mazmarcan: it's not too horrible though. can you please send it on the list? there are a number of patches already (I reposted the #PERST stuff), and I can prod Lorenzo to pick them up.
#asahi-dev · 2021-11-22 10:50
mazI really wish all this was done in core code. every single driver is reinventing this logic, because the core doesn't understand the PCIe state machine...
#asahi-dev · 2021-11-22 10:45
mazmarcan: given the level of quality the DWC driver has, I'd rather not touch it at all. I'll happily review patches that implement the speed selection though.
#asahi-dev · 2021-11-15 11:38
mazmarcan: I'm obviously partial to the PSCI solution, as it avoids reinventing another wheel, even if that's not a very nice wheel.
#asahi-dev · 2021-11-15 11:37
mazmarcan: ardb had some idea on how to fake this up, though this will require some buy-in from the usual suspects.
#asahi-dev · 2021-11-15 11:36
mazmarcan: yeah, EFI provides its own address map (which is a 1:1 mapping IIRC), which we use when calling into it.
#asahi-dev · 2021-11-13 12:28
mazkettenis: I need to entertain myself... it's otherwise very boring!
#asahi-dev · 2021-11-12 18:11
mazyup. given that PD reset is what I use all the time, that's the only solution (at least as I understand it)
#asahi-dev · 2021-11-12 18:09
mazsven: yeah, still can't be bothered. I found that I end-up hacking on the M1 when I'm not next to it. another good reason to go back into the office!
#asahi-dev · 2021-11-12 18:06
mazsven: it's the bestest!!!!
#asahi-dev · 2021-11-11 11:47
mazcan't blame him! :D
#asahi-dev · 2021-11-11 11:44
mazarnd: (and our DT parsing code is just as crufty as their DT...)
#asahi-dev · 2021-11-11 11:44
mazarnd: so I'd try and fix that first if at all possible.
#asahi-dev · 2021-11-11 11:43
mazarnd: their DT is totally mad, but I actually think this is a bug in the generic code. What I have added is supposed to handle the case where the imap points to the local node. but it also ignores the 'interrupt-controller' property in the current node if it fails to match.
#asahi-dev · 2021-11-11 10:24
mazdhewg: well, Bjorn already has queued a revert, so that's the thing I want to defuse. And I don't mind helping these guys, but I dislike the attitude.
#asahi-dev · 2021-11-11 09:39
mazI think I have an idea on how to avoid that particular case, but the idea of adding changes to the kernel for out of tree systems makes me want to puke a bit... (yes, I'm in a foul mood today).
#asahi-dev · 2021-11-11 09:38
mazbroke*
#asahi-dev · 2021-11-11 09:37
mazthe of changes in the PCI code broken this box.
#asahi-dev · 2021-11-11 09:32
mazsometimes, I feel like this is not a battle worth fighting any more.
#asahi-dev · 2021-11-11 09:31
mazso I now need to unfsck some Amiga clone that is broken by the M1 patches. and as it turns out, this machine isn't even supported by upstream...
#asahi-dev · 2021-11-10 17:53
mazmarcan: first patch seems reasonable. it's all black magic anyway.
#asahi-dev · 2021-11-10 08:38
mazkettenis: I wouldn't be surprised if that was the case. that's the 'easy' way to bolt anything on PCI, leaving to SW to deal with the resulting crap.
#asahi-dev · 2021-11-10 08:35
mazmarcan: I'll take patches for the bug!
#asahi-dev · 2021-11-09 09:23
mazalmost feels like it needs another enable to make it visible (all Fs is usually 'no device here')
#asahi-dev · 2021-11-08 15:50
mazmarcan: PCIe got merged over the weekend, it's time to break it!
#asahi-dev · 2021-11-04 10:55
mazI don't think anyone plans to port edk2...
#asahi-dev · 2021-11-04 10:51
mazinstalled his M1 by running the debian installer as a guest...
#asahi-dev · 2021-11-04 08:26
mazmarcan: fuse-ext2 ?
#asahi-dev · 2021-11-04 08:14
mazyou can always explain what is expected from a submission. they are of course free to ignore your suggestions, but you are also free to ignore them altogether in athat case.
#asahi-dev · 2021-11-03 16:47
mazj_ey: the other side isn't any better.
#asahi-dev · 2021-11-03 16:40
mazjust not the same thing as instructions
#asahi-dev · 2021-11-03 16:40
mazjannau: on the other hand, 78 (which is supposed to be AArch32) does count something.
#asahi-dev · 2021-11-03 16:38
mazjannau: 7a never seems to count anything here.
#asahi-dev · 2021-11-03 14:34
mazhopefully, this will give someone an incentive to work out the semantics of the counters...
#asahi-dev · 2021-11-03 14:33
mazhot off the press: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/log/?h=hack/m1-pmu-WIP

plus ›