#asahi-re · 2021-02-18 22:32
davidrysk[m]I think A13 has improved PAC?#asahi-re · 2021-02-18 22:27
roxfanalthough that was added in A12, not A13#asahi-re · 2021-02-18 22:26
roxfanmight be PAC then#asahi-re · 2021-02-18 22:25
roxfanah wait they do#asahi-re · 2021-02-18 22:25
roxfanthey say iOS 14, not A14 or whatever so it sounds mostly software#asahi-re · 2021-02-18 21:27
modwizcodeIn general those sound a lot like some of the safety that clang has maybe augmented a bit.#asahi-re · 2021-02-18 21:27
modwizcodeThat seems odd. It doesn't really sound like something that would require a chip to support#asahi-re · 2021-02-18 20:52
never_releasedhttps://twitter.com/fayfiftynine/status/1362486597749415936?s=21#asahi-re · 2021-02-18 20:24
never_releasedWill have to test#asahi-re · 2021-02-18 20:18
modwizcodeAre the compression instructions accessible in user mode?#asahi-re · 2021-02-18 20:09
never_releasedmarcan: is the Apple one accessible from user mode#asahi-re · 2021-02-18 20:08
never_releasedthrough ACTLR or other registers#asahi-re · 2021-02-18 20:08
never_releasedSome extensions have to explicitly be enabled#asahi-re · 2021-02-18 18:34
marcanthe page compress/uncompress instructions and translate instruction might not like that#asahi-re · 2021-02-18 18:34
marcando note that the other cores have the MMU off#asahi-re · 2021-02-18 18:33
maximus64oh nice. I was hack up mini to start on other cores before. Didn't know that there are framework for that already#asahi-re · 2021-02-18 18:32
marcanfeel free to extend it with a cpu= argument where 0 is call() and others go through smp_call_sync or so#asahi-re · 2021-02-18 18:31
marcan*inst()#asahi-re · 2021-02-18 18:31
marcanto run it on other CPUs, you call p.smp_start_secondaries() first and then just copy/hack the instr() thing in util.py to use proxy.smp_call_sync() instead of proxy.call()#asahi-re · 2021-02-18 18:30
marcanthe script is just stuff like https://mrcn.st/p/sBwViUh5 and manually poking around#asahi-re · 2021-02-18 18:30
marcanI pushed a minor framework change#asahi-re · 2021-02-18 18:30
maximus64did you push the change? I can try it on my setup :)#asahi-re · 2021-02-18 18:30
marcan(has anyone with a serial cable *not* tried m1n1's python stuff? because you're totally missing out)#asahi-re · 2021-02-18 18:30
marcan(it's adding one line of code at the top of my python script...)#asahi-re · 2021-02-18 18:29
marcanI can try later anyway#asahi-re · 2021-02-18 18:29
maximus64yea haha#asahi-re · 2021-02-18 18:29
j`eythat would be bad for scheduling!#asahi-re · 2021-02-18 18:29
marcanthat would be madness#asahi-re · 2021-02-18 18:29
marcanI did not, and I sure hope there aren't#asahi-re · 2021-02-18 18:29
maximus64marcan: did you try to scan for instructions on the firestorm core? wondering if there any differences there. Unlikely but possible#asahi-re · 2021-02-18 18:16
modwizcodenot sure if you'd want to predict that kind of instruction though#asahi-re · 2021-02-18 18:16
modwizcodeApple one might be predictable?#asahi-re · 2021-02-18 18:16
modwizcodeI mean ARM also has PAC and Apple does their own thing (although apple ones make sense)#asahi-re · 2021-02-18 18:16
marcanbut the ARM instructions return via sysreg, the apple one returns via GPR#asahi-re · 2021-02-18 18:16
modwizcoderight which is what I'm confused about#asahi-re · 2021-02-18 18:16
marcanand the output format is the same even#asahi-re · 2021-02-18 18:16
marcanthe thing is ARM *already has* instructions for translate address#asahi-re · 2021-02-18 18:13
modwizcodeI wonder what the translate address instruction is used for#asahi-re · 2021-02-18 18:07
marcandavidrysk[m]: it's in there https://github.com/AsahiLinux/docs/wiki/HW:Apple-Instructions#asahi-re · 2021-02-18 17:24
modwizcodeI suppose determining what type of params an opcode uses can be a lot easier if you have detailed fault reporting to use#asahi-re · 2021-02-18 17:23
davidrysk[m]marcan: how about the memory decompression instruction? :p#asahi-re · 2021-02-18 17:22
modwizcodeAdding support for ISA extensions in qemu is going to be fun, but I don't think there's a sane framework for doing so. Which makes sort of sense, ARM was originally completely against custom vendor ISA extensions iirc.#asahi-re · 2021-02-18 17:20
marcan(could just be some hint space Apple made for themselves, or PAC related or something)#asahi-re · 2021-02-18 17:20
marcanthe previous subset all does "nothing" as far as I can tell, but does not fault; no idea what it is#asahi-re · 2021-02-18 17:20
marcanit's literally the first thing that had an obvious effect (faulting), numerically, in the range#asahi-re · 2021-02-18 17:20
marcanhttps://twitter.com/marcan42/status/1362450439845781505#asahi-re · 2021-02-18 17:15
modwizcodeI figured those would be the most confusing and hard to find#asahi-re · 2021-02-18 17:15
modwizcodethat was quick#asahi-re · 2021-02-18 17:15
modwizcodeReally?#asahi-re · 2021-02-18 17:15
marcanwell, I just found the freaking memory compression instruction