arm64: fix mask in atomic_testand{set,clear}_64

[rlibby]

Summary

These macros generate both the 32- and 64-bit ops, but the mask was hard
coded for 32-bit ops, causing the 64-bit ops always to affect only the
low 32 bits.

Reported by: mmel

Test Plan

qemu-system-aarch64 -M virt -m 512m -cpu cortex-a57 -kernel obj/usr/src/freebsd/arm64.aarch64/sys/GENERIC/kernel.bin -nographic

After 942951ba46ecd5ebab18de006a24dc52e2d3f745 and before this patch, this panicked very early in boot (just after the "WITNESS option enabled" message), and now it makes it to mountroot.

[mmel]

Good catch. It never occurred to me that we could have a bug in atomics on two architectures at the same time, so I wasn't looking for a problem in my own garden :) I will try to fix arm by myself.
Tested on real HW, everything OK.

[rlibby]

In D27886#623242, @mmel wrote:

Good catch. It never occurred to me that we could have a bug in atomics on two architectures at the same time, so I wasn't looking for a problem in my own garden :) I will try to fix arm by myself.

Are you saying 32-bit ARM is also broken? Those are a little weirder, I think it uses sys/sys/_atomic_subword.h.

Tested on real HW, everything OK.

Thanks.

[rlibby]

In D27886#623414, @rlibby wrote:

In D27886#623242, @mmel wrote:

Good catch. It never occurred to me that we could have a bug in atomics on two architectures at the same time, so I wasn't looking for a problem in my own garden :) I will try to fix arm by myself.

Are you saying 32-bit ARM is also broken? Those are a little weirder, I think it uses sys/sys/_atomic_subword.h.

Okay, I can see that indeed 32-bit arm's atomic_testand{set,clear} also do not match the atomic(9) API with respect to the bit argument.

[mmel]

In D27886#623414, @rlibby wrote:

In D27886#623242, @mmel wrote:

Good catch. It never occurred to me that we could have a bug in atomics on two architectures at the same time, so I wasn't looking for a problem in my own garden :) I will try to fix arm by myself.

Are you saying 32-bit ARM is also broken? Those are a little weirder, I think it uses sys/sys/_atomic_subword.h.

Yes, it is also broken. ARM uses its own implementation, which does not do modulo for the bit position.
See https://cgit.freebsd.org/src/tree/sys/arm/include/atomic-v6.h#n862.
I will commit proper fix in next hour or so...

[rlibby]

In D27886#623424, @mmel wrote:

In D27886#623414, @rlibby wrote:

In D27886#623242, @mmel wrote:

Good catch. It never occurred to me that we could have a bug in atomics on two architectures at the same time, so I wasn't looking for a problem in my own garden :) I will try to fix arm by myself.

Are you saying 32-bit ARM is also broken? Those are a little weirder, I think it uses sys/sys/_atomic_subword.h.

Yes, it is also broken. ARM uses its own implementation, which does not do modulo for the bit position.
See https://cgit.freebsd.org/src/tree/sys/arm/include/atomic-v6.h#n862.
I will commit proper fix in next hour or so...

Yeah, I have seen this too. I have a patch, but no hardware, and no luck so far with getting qemu-system-arm to work.