| Age | Commit message (Collapse) | Author |
|---|
|
|
|
by adding a sb_datacc count to sockbuf that counts data excluding
MT_CONTROL and MT_SONAME mbuf types. w/help from deraadt@.
okay deraadt@ claudio@
|
|
wrong order. M_DUP_PKTHDR needs to be called with an empty mbuf.
Allocating an mbuf cluster beforehand is not allowed as the resulting mbuf is
no longer considered empty (part of the header is initialized).
The correct order is to allocate an mbuf via MGETHDR(), copy the packet header
and as last step allocate the cluster.
Issue found by JINMEI Tatuya.
OK canacar@ deraadt@ mglocker@ additional input itojun@
|
|
ok art thib
|
|
spotted by art, ok dlg art
|
|
from juan hernandez gonzalez; tested by bluhm@
|
|
ok kettenis@ cloder@ tom@ henning@
|
|
|
|
the problem with lost signals in MP kernels.
miod@, kettenis@ ok
|
|
- LK_EXCLUPGRADE is never used.
- LK_REENABLE is never used.
- LK_SETRECURSE is never used. Because of this, the lk_recurselevel
field is always zero, so it can be removed to.
- the spinlock version (and LK_SPIN) is never used, since it was decided
to use different locking structure for MP-safe protection.
Tested by many
|
|
compile and work. need to register pfkey whenever tcp md5 or ipsec is
defined, and the various ipsec encapsulations only if ipsec is defined.
ok theo
|
|
signal to is the curproc, check if it's SONPROC. That way we'll also
signotify processes on other CPUs. This gives MD code the opportunity
to actually deliver signals to other CPUs faster than the clock
interrupt (in the worst case).
This is a NOP until MD code is fixed.
miod@ ok
|
|
later when more of it's resources have been allocated and thus
kill(2)ing such a process has more predictable results.
now w/ a couple of kettenis remarks; kettenis@ miod@ ok
|
|
|
|
effectively been a no-op for quite some time now,
without promise for future usage.
ok pedro@
Testing by krw@ (earlier diff)
and Johan Mson Lindman (tybollt@solace.miun.se)
|
|
In this commit:
- gdt lock on amd64
- sysctl lock
- malloc sysctl lock
- disk sysctl lock
- swap syscall lock
miod@, pedro@ ok (and "looks good" others@)
|
|
getitimer when the clock on the machine has been reset.
miod@ ok
|
|
to the m_tag_delete_chain() function.
m_tag_free() and m_tag_unlink() are only used by m_tag_delete() and I see
no need to have these functions around.
m_tag_delete_chain() has a second argument to specifiy a starting point
from where the chain should be deleted. This feature is never used and is
more complex because it is not possible to remove a SLIST element without
doing a list walk. Instead use SLIST_FIRST() and SLIST_REMOVE_HEAD() to
remove all items from the list. OK dhartmei@
|
|
|
|
|
|
|
|
frame, use it to report profiling information via addupc_intr().
|
|
to addupc_task() in ADDUPROF(). From NetBSD via art@.
|
|
|
|
Constantine A. Murenin
|
|
|
|
set to 0, the watchdog will not be retriggered by the kernel *and* it will not
be disabled at system shutdown time (before it got disabled at system shutdown
time.)
ok markus mickey mk
|
|
|
|
Spotted by Constantine A. Murenin <cnst+openbsd@bugmail.mojo.ru>; thanks.
|
|
errors to userland: make all cpu_setperf functions return void.
Tested by many, ok gwk@
|
|
|
|
the higher level sysctl code passing them to userland. Distinct diff from
the one I passed around internally with less crazy semantics. Also nuke
some trailing spaces in a few places.
"just go ahead" deraadt@
|
|
ok jmc
|
|
ok brad
|
|
devices with sectorsizes other than 512. e.g. cd's. Fixes PR #5235
from Paul Stoeber with a slightly tweaked diff. NetBSD did the same
with their r1.59 in 2001, closing their PR#3261 and PR#14026.
tweak suggestions and ok pedro@
|
|
diff received from matthieu.
ok matthieu
|
|
|
|
should never be referenced outside the context of the process to which
this stack belongs unless we do the PHOLD/PRELE dance. Loads of code
doesn't follow the rules here. Instead of trying to track down all
offenders and fix this hairy situation, it makes much more sense
to not swap kernel stacks.
From art@, tested by many some time ago.
|
|
that provides a PPS signal, more precision and less jitter can be expected.
With my test setting (a Garmin GPS 18 LVC hooked up to a i386 machine) I
observed a skew of 0.03 to 0.05 seconds between the tty timestamp and the soft
timestamp.
"put it in" deraadt
|
|
M_CANFAIL, idea from miod@, okay deraadt@
|
|
field unless it's root that's asking. This far, this has been done by
putting a struct statfs on the stack and modifying it.
struct statfs is large. Large things on the stack are bad. Create
copyout_statfs() that copies out a struct statfs to userland, does the
necessary root check and gives userland what it needs withtout wasting
stack space.
pedro@, deraadt@ ok.
|
|
return NULL instead of panic()'ing.
ok pedro@, deraadt@
|
|
original diff from) Jake < antipsychic (at) hotmail.com >. Discussed
with Mickey and Miod.
ok miod@ pedro@
|
|
|
|
|
|
|
|
|
|
the euid/egid of the server it connects to. The euid and egid are captured in
the bind() call -- this is the equivalent of the client behaviour where the
same thing is done in connect().
tested and ok espie@, ok henning@, go for it deraadt@
|
|
|
|
kern_descrip.c change ok deraadt
|
