observed with acpicpu. C1 (MWAIT) is the deepest sleep currently enabled so
there should be no functional change. Also, the computed nap time is run
through hztoms() which seems incorrect as the ACPI timer is running in the
MHz range.
"build.sh release" on my test system:
- Crossing mount points during lookup is slow because the set up for, and
act of doing VFS_ROOT() is quite involved. Use the name cache to help
with this. Cache an "impossible" zero-length name with covered vnodes,
that points to the root of the file system mounted there. Use it to cross
mounts. When cache_purge() is called on either of the vnodes involved the
cache entry will disappear. All of the needed calls for that are already
in place (vnode reclaim, unmount, etc).
- In lookup_fastforward(), if the the last component has been found and the
parent directory (searchdir) is not going to be returned, then don't get a
reference to it.
- If the lookup fails return with the same lock held. There's no bug here
because either parent or child's lock is sufficient to prevent both from
disappearing into thin air, but may as well be correct.
- if FSCRED is passed in then skip the auth check.
also remove M_PCB - while it is used by SCTP, it's actually not defined
anywhere
in fact add new NOTES section explaining that the malloc types are
actually not used on NetBSD
- Switch to the C11-style atomic primitives using atomic_loadstore(9).
- npfkern: introduce the 'state.key.interface' and 'state.key.direction'
settings. Users can now choose whether the connection state should be
strictly per-interface or global at the configuration level. Keep NAT
logic to be always per-interface, though.
- npfkern: rewrite the G/C worker logic and make it self-tuning.
- npfkern and libnpf: multiple bug fixes; add param exporting; introduce
more parameters. Remove npf_nvlist_{copyin,copyout}() functions and
refactor npfctl_load_nvlist() with others; add npfctl_run_op() to have
a single entry point for operations. Introduce npf_flow_t and clean up
some code.
- npfctl: lots of fixes for the 'npfctl show' logic; make 'npfctl list'
more informative; misc usability improvements and more user-friendly
error messages.
- Amend and improve the manual pages.
0(sp) is safe here even on O32 because the only callee function
mips_cp0_cause_read() doesn't take arguments, but it's caller's
responsibility to choose proper stackframe region and the name
of CALLFRAME_S0 is just confusing. No binary change.
On O32 and O64, 0(sp) could be clobbered by callee functions
to save argument a0, so we can't use it to save caller registers.
Use CALLFRAME_SP instead because a frame pointer is not used here.
Maybe all other references of CALLFRAME_S0 might also be problematic
on O32. Discussed with soda@.
as a time.
Gather the full 512 bytes of samples and process in a single call to
rnd_add_data_sync() - about 10% faster than 4 calls to rnd_add_data_sync().
Put sample buffer in the softc to save some stack usage.
Currently NetBSD/arc kernels are loaded at 0x80200000
(on some machines ARC BIOS uses regions before it) and
ARC BIOS also uses around 0x807f0000, so bootloader fails
on loading >6MB kernels.
Note GXemul doesn't have this restriction because it loads
a kernel directly via a command line argument.
Tested by installing using sysinst on RAMDISK kernel on netbsd-9.
Should be pulled up to netbsd-9.
