Get mips64(el)-linux working and start testing it
#21095
Conversation
|
This should be ready for review. |
alexrp
force-pushed
the
mips64-tests
branch
2 times, most recently
from
dd9545f to
cb069d3
Compare
The kernel sets r7 to 0 (success) or -1 (error), and stores the result in r2. When r7 is -1 and the result is positive, it needs to be negated to get the errno value that higher-level code, such as errnoFromSyscall(), expects to see. The old code was missing the check that r2 is positive, but was also doing the r7 check incorrectly; since it can only be set to 0 or -1, the blez instruction would always branch. In practice, this fix is necessary for e.g. the ENOSYS error to be interpreted correctly. This manifested as hitting an unreachable branch when calling process_vm_readv() in std.debug.MemoryAccessor.
It is usually generic, so no point having it in arch bits.
These take 3-4 minutes to run on my machine, i.e. they're not affected by the LLVM compilation speed bug that affects mips32.
| \\ syscall | ||
| \\ blez $7, 1f | ||
| \\ beq $7, $zero, 1f | ||
| \\ blez $2, 1f | ||
| \\ dsubu $2, $0, $2 | ||
| \\ 1: |
There was a problem hiding this comment.
The equivalent musl code for all these syscall functions has only one daddu instruction followed by syscall. Any reason we can't do that?
There was a problem hiding this comment.
The reason we don't have that is because musl mixes inline assembly and C logic: https://git.musl-libc.org/cgit/musl/tree/arch/mips64/syscall_arch.h#n38
We could of course do the same, but I thought I'd just keep things simple.
There was a problem hiding this comment.
it's not really related to C logic, it's just using C syntax to get values in and out of registers. Zig's constraint strings are powerful enough to do the same. You're already doing it with ={$2} to get the value out. You can do the same thing with the input values.
register long r4 __asm__("$4") = a;
is just
: "[a] {$4}" (a),
in zig.
@andrewrk that's because the CI results here were from before #21125 was merged. The extreme slowdown observed for MIPS is specific to O32; N64 runs in a normal amount of time. |
|
I see, thank you for pointing that out. |
No description provided.