| Age | Commit message (Collapse) | Author |
|---|
|
zero cache-size.
found by william@
|
|
|
|
|
|
ok krw@
|
|
|
|
|
|
|
|
|
|
|
|
ok jakemsr@, who promises to deal with any fallout, because he's a stand-up guy
|
|
ok deraadt miod
|
|
|
|
included that way.
pointed out by william@
|
|
pv_changepte4m() since DVMA addresses are never loaded in the SRMMU; this
chunk was forgotten in the previous `don't steal DVMA out of kernel_map on 4m'
commit.
|
|
* avoid error exit code after mere warnings
* add ERROR: and FATAL: to messages when appropriate
* sort the code in mmsg() to make it easier on the eye
* make the mandocerrs[] list easier to maintain
* update a few comments in mandoc.h
ok kristaps@
|
|
requested and well explained by krw@
|
|
|
|
getpeereid(2), but also supplies the remote pid. This is supplied in
a 'struct sockpeercred' (unlike Linux -- they showed how little they
know about real unix by calling theirs 'struct ucred').
ok guenther ajacoutot
|
|
"I agree that hz / 10 is 1/10 of a second, which is 100 milliseconds" krw@
|
|
|
|
unless it also belongs to a subclassed object class (structural or
auxiliary).
|
|
ok deraadt@
|
|
namespaces. Useful for dump/restore of all namespaces.
|
|
they are supposed to do, or be silent.
ok mlarkin
|
|
We check a bunch of things, where most archs are a lot more simple.
Also, we get problems sometimes: My x201 can't map the framebuffer BAR
from the aperture with the chunk that checks the bios sections, meaning
the bios is claiming something about it.
Kettenis@ and I are pretty sure that the rest of the checks are more than
sufficient, so just nuke that one.
for the record, in the same situation i386 just checks again VGA_START,
BIOS_END and physmem.
ok kettenis@
|
|
ok marco
|
|
means.
If we hit an OOM condition, mg started to try and dump the older undo buffer
entries in order to be able to continue. OTOH, it was grabbing this entry with
LIST_END, which like all *_END() list macros evaluates to NULL.
Do what we actually want and switch that list to a TAILQ and use
TAILQ_LAST to grab the last entry.
Wrote this a loooooong time ago after a mail from Matthew Dempsky on
bugs@. ok kjell@, beck@ also looked at this months ago and thought it
was alright.
|
|
ok krw, discontent grunts from marco.
|
|
ok krw
|
|
ok krw@, marco@
|
|
ok krw
|
|
on a 128-bit boundary like we do on amd64. The padding within the struct
made sure the state was properly aligned *within* the struct, but since the
alignment restriction on the struct itself was only 64-bit, the required
alignment was not guaranteed. We just got lucky.
ok oga@, guenther@
|
|
labels we use decimal.
OK jsg@ deraadt@
|
|
broken by ryan in australia
problem found & nagging by sthen
jsg found the fix but failed to apply the cluestick correctly ;)
test & ok sthen
|
|
to not depend on it).
|
|
to better handle detaching.
ok deraadt@
|
|
Avoid null pointer dereference when reopening a namespace.
|
|
|
|
|
|
fixup arguments to the fpu_kernel_enter/exit.
from mike
|
|
|
|
|
|
ok dlg@
|
|
|
|
|
|
|
|
methods. the AD1848 DMA engine runs continuously and doesn't
need to be started/stopped or otherwise manipulated for each
block. makes the driver a bit more efficient.
* allow "Mode 2" capable chips such as the CS4231 to work in "Mode 1"
when there is only one DMA channel, instead of forcing it into
"Mode 2", which expects two DMA channels. allows recording to
work for "Mode 2" capable devices when there is only one DMA
channel.
* service both playback and recording interrupts in full-duplex
mode, instead of only playback interrupts. allows "Mode 2"
capable chips to work in full-duplex mode.
* fix checks for sample rate/format recalibration completion. gets
rid of busy loops that briefly "hang" the kernel when the device is
opened and/or reconfigured.
|
|
create verbose_system up in state, because it makes sense without verbose.
|
|
|
|
don't create updatesets directly, go through AddDelete::State.
Get all Locator/Repo access through state.
|
