#
✕ nick:marcan
#asahi-dev · 2022-02-06 12:55
marcanwhat do you mean by breaks pcie?
#asahi-dev · 2022-02-06 12:31
marcanmight want to poke that, since that blocks the new chips anyway
#asahi-dev · 2022-02-06 12:31
marcansven: we still haven't heard from the t6k DART stuff right?
#asahi-dev · 2022-02-06 12:31
marcanthen again
#asahi-dev · 2022-02-06 12:31
marcanI need to send the rest of wifi... and clean up aicv2...
#asahi-dev · 2022-02-06 05:43
marcanfyi: you can use `softwareupdate --fetch-full-installer --full-installer-version 12.1` to download a 12.1 installer, useful for updating to a specific version that is no longer the latest
#asahi-dev · 2022-02-05 04:39
marcanhence the *h variants are the same for big/little
#asahi-dev · 2022-02-05 04:39
marcan< alyssa> unless I don't understand what the &(u32) notation means? <- it means an integer in native byte order
#asahi-dev · 2022-02-04 10:28
marcanwouldn't even be that surprising if it's useless and some legacy thing even
#asahi-dev · 2022-02-04 10:28
marcankettenis: we don't, I'm just going to ignore it for now. seems to work fine without it.
#asahi-dev · 2022-02-04 08:08
marcanin principle the firmware could decide to do whatever crazy mappings it wants all within reserved memory, and we'd have to preserve that
#asahi-dev · 2022-02-04 08:08
marcanI mean, what's the advantage of putting this in the DT vs reading out mappings?
#asahi-dev · 2022-02-04 04:11
marcanand that's orthogonal to whether the mappings got inherited or not (though in practice there's no case where they aren't and the DART is locked of course)
#asahi-dev · 2022-02-04 04:10
marcanif the DART is locked we atomically copy out the top level PT on top of the existing one, then use that instead
#asahi-dev · 2022-02-04 04:10
marcanthen if the DART is not locked, we just set a TTBR as usual replacing the existing one (and from that point on we don't care about the firmware-allocated PTs)
#asahi-dev · 2022-02-04 04:10
marcanin fact this feature is sort of orthogonal to locked DART support; there should be an apple,inherit-mappings thing or something, and that would be set both for DCP and DISP0; it would mean on startup we walk the page tables and validate/absorb those mappings
#asahi-dev · 2022-02-04 04:08
marcanI added code to m1n1 to clean up DART page tables that are allocated out of the heap on exit, for DCP/DISP0
#asahi-dev · 2022-02-04 04:07
marcanfor DARTs for devices that can be shut down, the TTBRs should be torn down entirely; for DARTs that need to stay alive (DCP and DISP0), all mappings in normal RAM should be removed
#asahi-dev · 2022-02-04 04:06
marcanprior stages should always clean up
#asahi-dev · 2022-02-04 04:06
marcanfirmware leaving behind mappings in normal RAM space is a bug anyway
#asahi-dev · 2022-02-04 04:05
marcan*superset
#asahi-dev · 2022-02-04 04:05
marcan(a subset)
#asahi-dev · 2022-02-04 04:05
marcanafter all, we need this for locked DARTs anwyay
#asahi-dev · 2022-02-04 04:05
marcan(except for non-locked DARTs we can move the TTBR if we want to)
#asahi-dev · 2022-02-04 04:05
marcanthat works for both locked and non-locked DARTs then, it'd basically implement the same thing
#asahi-dev · 2022-02-04 04:04
marcanjannau, sven: why not just have a property in the DART node that tells the driver to not reset it and inherit the iova mappings (the driver should then verify that none of them are in the normal RAM range, and complain loudly about firmware bugs if they are)
#asahi-dev · 2022-02-04 04:02
marcanpushed SMC stuff I had to smc/work; going to work on hooking up the GPIOs to PCIe now
#asahi-dev · 2022-02-03 16:51
marcanthat's pretty much next
#asahi-dev · 2022-02-03 16:50
marcankettenis: mostly because it doesn't even create an ESP yet :-)
#asahi-dev · 2022-02-03 16:04
marcangmail hated me *forwarding* spam to a specific account there via an alias once, but only rejected that spam itself / stuff to that account, they didn't blacklist my whole host
#asahi-dev · 2022-02-03 16:04
marcanthe only provider that has ever given me trouble with blacklists is microsoft, but well, microsoft.
#asahi-dev · 2022-02-03 12:51
marcanRedecorating[m]: thanks, I'll fix the commit message
#asahi-dev · 2022-02-03 10:06
marcanAIUI there are a few bits of physical addr ASLR, and then a few more of virtual addr ASLR
#asahi-dev · 2022-02-03 09:34
marcanI have my own roadmap of what order I want to do things in, but it's always subject to change if someone volunteers to pick up something...
#asahi-dev · 2022-02-03 09:33
marcanalso if someone really wants to build this for me and does a good job, it might even make it into the first "release" or a nearby one and future-proof us a bit
#asahi-dev · 2022-02-03 09:32
marcankettenis: people get excited :p
#asahi-dev · 2022-02-03 09:30
marcanso one way or another there has to be code there, even if it's just "jump elsewhere"
#asahi-dev · 2022-02-03 09:30
marcanthe sleep resume vector is locked at m1n1+0 (the way we load it now)
#asahi-dev · 2022-02-03 09:30
marcanand because either way it has to go through m1n1
#asahi-dev · 2022-02-03 09:23
marcan:)
#asahi-dev · 2022-02-03 09:23
marcanChaosPrincess: the main difference with chainload.py is no mach-o nonsense; just load m1n1, then sepfw (aligned), then boot-args (aligned) as one blob, then have a copy stub (probably safest to put it right after boot-args along with everything else), and call it to clobber the original m1n1
#asahi-dev · 2022-02-03 09:21
marcanj`ey: technically we could avoid some copies by being smarter, but it's a pointless micro-optimization given how fast these things are
#asahi-dev · 2022-02-03 09:20
marcanand the ADT needs to be patched to change the sepfw address
#asahi-dev · 2022-02-03 09:20
marcanof course the pointers within that image (i.e. in the ADT and boot-args) need to point to where things will eventually end up, not where they are in the image
#asahi-dev · 2022-02-03 09:19
marcanyes
#asahi-dev · 2022-02-03 09:19
marcanyes
#asahi-dev · 2022-02-03 09:17
marcanthis is required for sleep support to work in the future
#asahi-dev · 2022-02-03 09:16
marcansee what chainload.py does
#asahi-dev · 2022-02-03 09:16
marcanChaosPrincess: not correct; you should do an in-place chainload, which means building an in-memory image with sepfw and boot-args and then calling a copy stub that puts it on top of where the original m1n1 was
#asahi-dev · 2022-02-03 08:19
marcan< maz> but I'm everyday grateful I run my own email infrastructure. <- word...

plus ›