#
✕ nick:marcan
#asahi-dev · 2022-08-08 10:51
marcanit's early days enough that maybe we can afford to break people, especially with something as important as DART
#asahi-dev · 2022-08-08 10:50
marcanand I'm not sure if new compatibles is really the best option here (cc robher_ tl;dr older kernels with drivers missing features that cause things to explode when we add new devices to the DT)
#asahi-dev · 2022-08-08 10:49
marcanjannau: do we really need a new compatible for that? we're going to run into this problem (older kernels borking on new DTs) more than this time... there's no "good" solution
#asahi-dev · 2022-08-08 10:47
marcanyup, the bikeshedding as usual
#asahi-dev · 2022-08-08 05:50
marcanthat, ext display bikeshedding, the IOMMU handoff stuff, making it work on all machines, what else is needed?
#asahi-dev · 2022-08-08 05:49
marcanjannau/(alyssa if she sees the log): the new DRM backlight stuff seems to be moving slowly, so let's go with good old backlight device support for now for DCP.
#asahi-dev · 2022-08-06 09:31
marcanwouldn't be the first time I send them a patch...
#asahi-dev · 2022-08-06 09:31
marcanjannau: AIUI ceph is supposed to work nicely on arm64, but I'll have to try it
#asahi-dev · 2022-08-05 05:34
marcan(since we ended up using the linux ones in the distro, mostly to keep it synced with the kernel)
#asahi-dev · 2022-08-05 05:34
marcankettenis: probably a good idea for you to keep the u-boot DTSes up to date with an eye on BSD usage too, so then if you package that up you don't have to worry about me accidentally breaking you
#asahi-dev · 2022-08-05 04:43
marcanor if one of them kills the output mid-cycle and the other always waits for the end of a cycle, one is a hard/global enable (or reset?) and the other a softer enable
#asahi-dev · 2022-08-05 04:42
marcantry slowing it down and toggling them individually - if you get it stuck on at some point, that bit is a clock enable
#asahi-dev · 2022-08-05 04:42
marcanj`ey: two bits to enable isn't that odd, one might enable the peripheral as a whole or be a clock enable and the other actually enable the output
#asahi-dev · 2022-08-03 09:51
marcanmaybe, but that's more fragile anyway. glad I didn't so this happened and we caught it :)
#asahi-dev · 2022-08-03 09:49
marcantl;dr C is evil let's move to rust
#asahi-dev · 2022-08-03 09:49
marcanso they were getting the wrong pointer and casting it; it worked for MTP because the inner priv was the first member
#asahi-dev · 2022-08-03 09:49
marcankettenis: didn't pass through the priv into leaf functions when I moved the kbd stuff to another file
#asahi-dev · 2022-08-03 09:39
marcanuboot on spi keyboards is fixed
#asahi-dev · 2022-08-03 08:59
marcanyeah, I just had the same idea while writing that doc earlier :)
#asahi-dev · 2022-08-03 05:39
marcanwhoops, let me see what that's about
#asahi-dev · 2022-08-02 12:25
marcanbut yeah, probably doesn't make a huge difference
#asahi-dev · 2022-08-02 12:24
marcanthere's also UCLASS_NOP
#asahi-dev · 2022-08-02 12:24
marcankettenis: I wasn't sure about the uclass, by the way (e.g. misc has a ton of very generic device ops I'm all ignoring)
#asahi-dev · 2022-08-02 12:23
marcanmakes sense
#asahi-dev · 2022-08-02 12:18
marcanthanks :)
#asahi-dev · 2022-08-02 12:16
marcan-is
#asahi-dev · 2022-08-02 12:16
marcanis the merge plan for that is going straight through Linus? or is there some subsystem where it fits?
#asahi-dev · 2022-08-02 10:08
marcanas if Apple can't just make JTAG-enabled or development-cert testing devices at will
#asahi-dev · 2022-08-02 10:07
marcanbut yes, people claim it's "for internal testing of something" and other nonsense like that
#asahi-dev · 2022-08-02 10:07
marcanyeah, no, it isn't
#asahi-dev · 2022-08-02 09:16
marcanyou can point the "apple evil" spammers at that one from now on
#asahi-dev · 2022-08-02 09:16
marcanFor the record: https://twitter.com/marcan42/status/1554395176025849856
#asahi-dev · 2022-08-02 06:48
marcanwe probably won't move back to linux-next though, at least not for end users (might still do some merges based on that if something interesting happens though)
#asahi-dev · 2022-08-02 06:47
marcanno, we'll probably keep tracking -rcs, at least for dev kernels, though I might start keeping a separate release branch if warranted
#asahi-dev · 2022-08-02 06:44
marcanlinux-asahi is also now based on release 5.19
#asahi-dev · 2022-08-02 06:44
marcanjannau: fixed, pushed, and tested
#asahi-dev · 2022-08-02 06:35
marcanthat's what I get for trying to release something on stream late at night :p
#asahi-dev · 2022-08-02 06:23
marcanah, and the DT changes too...
#asahi-dev · 2022-08-02 06:07
marcanI knew I was forgetting something, the defconfig!
#asahi-dev · 2022-08-02 06:07
marcanlet me fix that
#asahi-dev · 2022-08-02 06:07
marcanoops :)
#asahi-dev · 2022-08-02 06:07
marcan.config issue.
#asahi-dev · 2022-08-02 06:07
marcanoh derp, I know what happened
#asahi-dev · 2022-08-02 06:04
marcanjannau: just m1pro, about to test installed on m2 air
#asahi-dev · 2022-07-31 09:30
marcan(actually it seems to be ASM235CM or so, since the individual drives also show up as Gen2)
#asahi-dev · 2022-07-31 08:04
marcan(and power off needs a long press, so no danger of accidents)
#asahi-dev · 2022-07-31 08:03
marcanalso the power buttons power cycle the interfaces (not just the drives), so toggling them is equivalent to a hotplug of the entire bridge chip, which is good in case they bug / for drive hotplug
#asahi-dev · 2022-07-31 08:03
marcanalmost 800MB/s across all drives on Gen2. Sold. This thing does the job.
#asahi-dev · 2022-07-31 08:00
marcan200MB/s reads from a single spinning rust drive, about 400MB/s aggregate across all 4 drives. That adds up to a Gen1 port. I think I have a Gen2 port on this box, let me move it...
#asahi-dev · 2022-07-31 07:55
marcanjannau: that logic says it's ASM1153, and I do get UAS. four of those behind a VIA hub. going to test performace now.

plus ›