[Bug 293382] Dead lock and kernel crash around closefp_impl

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=293382

--- Comment #70 from Konstantin Belousov <kib@FreeBSD.org> ---
The '1024' magic number you mentioned is somewhat consistent with the idea that
there is something not right with 'rep;movsb' or like, because CPU most likely
does not check for the interrupt after each transferred byte, but only after
some
run was processed.

Anyway, the fact that replacement of memcpy() by explicit byte moving using
plain
moves did not work is perhaps enough to close that lead.

Another interesting feature of the CPUs is the support for INVLPGB.
Please keep the same patched kernel, and additionally set the tunable
vm.pmap.invlpgb_works to 0 from the loader.

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