#asahi-dev · 2023-11-21 11:32
marcanI'll bisect this.#asahi-dev · 2023-11-21 11:32
marcanperf 6.4 works. perf 6.5 does not. nor does 6.6, nor 6.7-rc2.#asahi-dev · 2023-11-21 11:24
marcanI think we might be onto something here...#asahi-dev · 2023-11-21 11:24
marcanI have 6.5.4 on Fedora boxes and 6.5-1 on Arch boxes#asahi-dev · 2023-11-21 11:24
marcanmaz: perf --version?#asahi-dev · 2023-11-21 11:23
marcanbut I think I've updated all these machines relatively recently#asahi-dev · 2023-11-21 11:23
marcanI'm pretty sure this used to work and it's not lining up with any kernel changes#asahi-dev · 2023-11-21 11:23
marcanis it possible this is a userspace update that broke things?#asahi-dev · 2023-11-21 11:22
marcanbut it also happens on m2 max#asahi-dev · 2023-11-21 11:22
marcanyes#asahi-dev · 2023-11-21 11:22
marcanm2#asahi-dev · 2023-11-21 11:22
marcanmaz: and on a t8112 running 6.3.0-asahi-13-1-edge-ARCH it's backwards - I only get counts on the ecores, not the pcores#asahi-dev · 2023-11-21 11:14
marcanhaven't tested 6.6 yet though#asahi-dev · 2023-11-21 11:08
marcanmaz: repros everywhere for me, including an Arch ARM m1 mini currently running a random kernel (6.5.0-asahi-00774-g0f47112a2d71) and another M1 machine running 6.5.0-asahi-11-1-edge-ARCH#asahi-dev · 2023-11-21 10:54
marcanyeah I dunno :)#asahi-dev · 2023-11-21 10:54
marcanit's more like the counters are being read from the wrong CPUs?#asahi-dev · 2023-11-21 10:52
marcanit's just a support flag#asahi-dev · 2023-11-21 10:52
marcanarm64 always selects IRQ_FORCED_THREADING#asahi-dev · 2023-11-21 10:51
marcanwhich I don't have#asahi-dev · 2023-11-21 10:51
marcanalso that still needs the bootarg right?#asahi-dev · 2023-11-21 10:50
marcanyeah that sounds likely#asahi-dev · 2023-11-21 10:50
marcanthat?#asahi-dev · 2023-11-21 10:50
marcanCONFIG_IRQ_FORCED_THREADING=y#asahi-dev · 2023-11-21 10:48
marcanbut I get the same as you on an Ultra#asahi-dev · 2023-11-21 10:48
marcanno, the first one was on M1 Pro#asahi-dev · 2023-11-21 10:47
marcanyes, same here#asahi-dev · 2023-11-21 10:47
marcan2-9#asahi-dev · 2023-11-21 10:47
marcan$ cat /sys/devices/apple_firestorm_pmu/cpus#asahi-dev · 2023-11-21 10:47
marcanwell, we just had a case where a kernel config exposed uninitialized stack memory usage in brcmfmac we'd never noticed before so... :)#asahi-dev · 2023-11-21 10:46
marcan/sys does show the PMUs assigned to the right CPUs though#asahi-dev · 2023-11-21 10:44
marcanmaz: this repros on M1 Ultra (6.5.0-asahi-00777-g9bf71751b732-dirty) and M1 Pro (6.5.11-404.asahi.fc39.aarch64+16k) on Fedora. maybe it's something about the Fedora kernel config?#asahi-dev · 2023-11-21 07:11
marcanNefsen402: once it's ready for testing we'll say so, it's not useful for people to jump on dev branches and report all the same issues#asahi-dev · 2023-11-21 06:26
marcanNefsen402: that branch is broken and not what you want.#asahi-dev · 2023-11-21 06:23
marcanI get counts for *both* PMUs (~identical) when pinned to a P-core, and zero everywhere when pinned to an E-core#asahi-dev · 2023-11-21 06:22
marcanmaz: I think something is wacky with the PMUs again. On t600x 6.5.11-404.asahi.fc39.aarch64+16k, using sudo taskset -c 2 perf stat -e apple_icestorm_pmu/cycles/ -e apple_firestorm_pmu/cycles/ -e cycles echo#asahi-dev · 2023-11-07 15:53
marcanthanks :)#asahi-dev · 2023-11-07 15:53
marcangoing to sleep now but I'll queue them up tomorrow or thursday#asahi-dev · 2023-11-07 15:53
marcanack#asahi-dev · 2023-11-07 08:36
marcangrabs logs etc automatically#asahi-dev · 2023-11-07 08:36
marcanbug reporting tool thing#asahi-dev · 2023-11-07 06:58
marcanit makes abrt refuse to consider *any* kernel oops diagnosable which is kind of silly#asahi-dev · 2023-11-07 06:57
marcanmaz: do we really want to taint the kernel for the vgic thing, on startup? maybe change it to first usage of kvm?#asahi-dev · 2023-11-06 17:21
marcanawesome :D#asahi-dev · 2023-11-02 17:42
marcanwrote usbmuxd a long, long time ago#asahi-dev · 2023-11-02 17:42
marcanin principle it should be possible to run usbmuxd in-process with some rework, if someone really cares. you could probably just spawn it as a thread and leave the socket stuff in place even if it all goes back to the same process. but then again at that point it's probably easier to just have a shell script that spawns it before idevicerestore.#asahi-dev · 2023-11-02 17:40
marcanhowever you should *also* be able to just use the iTunes usbmuxd on Windows, usbmuxd is designed to use the same protocol as Apple's implementation for that reason#asahi-dev · 2023-11-02 17:40
marcanyou do need newer usbmuxd to get the PIDs for newer machines#asahi-dev · 2023-11-02 17:39
marcanyou don't need udev activation for usbmuxd, you can just run it standalone, but you do need it to be running#asahi-dev · 2023-10-31 08:08
marcanoof#asahi-dev · 2023-10-31 00:44
marcanpossibly