[Bug 222234] head -r323246 aarch64 (Pine64+ 2GB) boot time context, sometimes: acquiring blockable sleep lock with spinlock or critical section held

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=222234

--- Comment #9 from Mark Millard <marklmi26-fbsd at yahoo.com> ---
(In reply to Emmanuel Vadot from comment #8)

I'll have to build a debug kernel and try using it.
(It will be later today before I get to that.) Likely
I'll test as of -r337400 since that is my established
context overall and is well after -r324207 .

As for identcpu.c 's example, I've been running with
it patched --but only because the ARM documentation
indicated to always have the dsb. I do not know what
an expected observational difference would be for the
two code variants. All I can say is that having it did
not seem to hurt anything in any obvious way. Is there
someone around that might know if there is a reason
that identcpu.c does not need the instruction, despite
what I read? Should it be updated based just one what
arm documents, even if no observational difference is
known? I leave the judgments to you but likely will
keep the patch in place if the file is not updated.

-- 
You are receiving this mail because:
You are the assignee for the bug.