These annotations help to mitigate false sharing on multiprocessor
systems.
Variables annotated with __cacheline_aligned are placed into the
.data.cacheline_aligned section in the kernel. Each item in this
section is aligned on a cachline boundary - this avoids false
sharing. Highly contended global locks are a good candidate for
__cacheline_aligned annotation.
Variables annotated with __read_mostly are packed together tightly
into a .data.read_mostly section in the kernel. The idea here is that
we can pack infrequently modified data items into a cacheline and
avoid having to purge the cache, which would happen if read mostly
data and write mostly data shared a cachline. Initialisation variables
are a prime candiate for __read_mostly annotations.
using an EXTERN_INLINE definition for functions that are defined as
inline but provide an externally callable reference.
(these are externally called in ipftest)
having a system definition for alloca is counter productive as it seems
fairly common to use
#ifdef __GNUC__
#define alloca ..
#endif
in user-headers (eg see usr.sbin/gspa/gspa/gspa_ass.h)
So, defang this definition as we don't otherwise provide alloca on NetBSD
Changes since 3.99.2:
+ avoid possible free() of new value passed to netpgp_setvar(),
with thanks to Anon Ymous.
+ netpgpkeys(1): print keys to stdout, not stderr - reported by Anon
Ymous.
+ fix DSA signatures and verification
+ simplify and shorten the internals of packet processing by getting rid of
the intermediate pseudo-abstraction layer, which detracted from understanding
and had no benefit whatsoever. Rename some enums and some definitions.
+ add some checking to new key generation, and don't try to read in
the keys after writing them - reported by Tyler Retzlaff
don't try to re-read the key after writing it - that's done by a separate
function. Problem found by Tyler Retzlaff, fixed in a different way.
check that keyrings are non-NULL before attempting to free them - from a
nudge by Tyler Retzlaff.
Since they are practically never used (only when prehistoric code
uses simple_lock()), their efficiency doesn't matter that much and
we can simply adapt the versions from x86 lock.h.
by defining 'options ACPIVERBOSE' in the kernel config file (no change
from current behavior), or it can be loaded at boot time on those
architectures that support the boot loader's "load" command.
ACPIVERBOSE also includes code to dump acpi/wmi data.
has no support for it. For example, in Kurt Schreiner's Shuttle G61:
Device (FAN)
{
Name (_HID, EisaId ("PNP0C0B"))
Method (_INI, 0, NotSerialized)
{
Store (TP1H, CTOS)
Store (TP1L, CTHY)
}
}
Obviously acpitz(4) has little use for a device that only contains _INI.
(The handle of this device is referenced incorrectly in _AL0.)
Make single stepping a system call work. Instead of single stepping
through the syscall gateway page, which doesn't work since that page is
shared between process, this makes us step over that bit by setting a
breakpoint on the instruction where the system call returns.
mjf