This patch fixes ZFS support on aarch64. As the issue #1131 explains,
the ZFS page scanner logic clears the access flag of PTEs of relevant
memory-mapped chunks of the files. On Intel, the cpu automatically sets
the flags on first access (read or write) to those pages of memory.
But on ARM it may
This patch implements the signal handling on aarch64. I will not
be repeating the details of what it changes and why as it is quite
well explained in the code changes.
But in essence, this patch updates the build_signal_frame() which
I believe was based on the x86_64 version of it with the
When testing the tst-map.cc and tst-elf-permissions.cc with the upcoming
patch to add signals support on aarch64, I noticed that sometimes they
would kind of "hang" for a while and eventually complete successfully.
This would happen especially when running in non-SMP mode (1 CPU). After more
The assembler on aarch64 uses different characters - '//' - for the end
of line comment (see
https://sourceware.org/binutils/docs-2.26/as/i386_002dChars.html#i386_002dChars
and
https://sourceware.org/binutils/docs-2.26/as/AArch64_002dChars.html#AArch64_002dChars
respectively). So we add the
This patch does not change the ring_spsc template in any significant way
but merely makes the counter type (of _begin and _end) a parameter. This
allows us to use smaller type - unsigned short (2 bytes) - in the unit test
tst-ring-spsc-wraparound.cc to make it execute more rapidly.
Before this