panic with the NDISulator if you did "ifconfig ndis0 10.0.0.1/24,"
whereas "ifconfig ndis0 10.0.0.1/24 up" worked fine. The double fault
was caused by the ifconfig thread running out of kernel stack space.
(This was partly due to the NDIsulator using a couple of big buffers on
the stack, but even after fixing that the double fault persisted.)
It turns out that ndis_init() is called in both cases, but in the first
case the code path passes through ieee80211_ioctl(), and it turns out
ieee80211_ioctl() consumes a whopping 2400 bytes of stack space.
Apparently, gcc -O2 causes the ieee80211_ioctl_get80211() routine to
be inlined into ieee80211_ioctl(), and for some reason which I do not
fully understand, this causes ieee80211_ioctl() to consume an extra 2K
of stack space.
To prevent this overly agressive optimization, ieee80211_ioctl_get80211()
is now declared with __attribute__ ((noinline)). With this change,
ieee80211_ioctl() now only reserves about 200 bytes of stack instead of 2400.
create kernel threads and call rfork(2) with RFTHREAD flag set in this case,
which puts parent and child into the same threading group. As a result
all threads that belong to the same program end up in the same threading
group.
This is similar to what linuxthreads port does, though in this case we don't
have a luxury of having access to the source code and there is no definite
way to differentiate linux_clone() called for threading purposes from other
uses, so that we have to resort to heuristics.
Allow SIGTHR to be delivered between all processes in the same threading
group previously it has been blocked for s[ug]id processes.
This also should improve locking of the same file descriptor from different
threads in programs running under linux compat layer.
PR: kern/72922
Reported by: Andriy Gapon <avg@icyb.net.ua>
Idea suggested by: rwatson
cojunction with -C and is used by /etc/rc.d/newsyslog.
I forgot that this was in my perforce tree and not my running system and
thus committed a non-working newsyslog script.
Reported by: des
Pointy hat: brooks
precision when IP packet may travel through internet for several seconds.
Also uptime measured in milliseconds overflows every 48+ days.
But we have to do same to keep compatibility with Cisco and flow-tools.
Make a macro MILLIUPTIME, which does overflowable multiplication to 1000.
Requested by: Sergey Ryabin, Oleg Bulyzhin
MFC after: 1 week
is terminated with a slash. Although we are not System V, ar has
been configured to put that slash in. This format allows filenames
with trailing spaces.
both consuming 1K of stack space. This is unfriendly. Allocate the buffers
off the heap instead. It's a little slower, but these aren't performance
critical routines.
Also, add a spinlock to NdisAllocatePacketPool(), NdisAllocatePacket(),
NdisFreePacketPool() and NdisFreePacket(). The pool is maintained as a
linked list. I don't know for a fact that it can be corrupted, but why
take chances.
a libalias application (e.g. natd, ppp, etc.) to crash. Note: Skinny support
is not enabled in natd or ppp by default.
Approved by: secteam (nectar)
MFC after: 1 day
Secuiryt: This fixes a remote DoS exploit
the first user/group. Caused huge fun in error messages from large script.
Old: pgrep -u root,NoSuchUser,daemon -> pgrep: unknown user `root'
Now: pgrep -u root,NoSuchUser,daemon -> pgrep: unknown user `NoSuchUser'
Obtained from: NetBSD (rev. 1.8)
MFC After: 1 week (if re@ would have approved this)
call vector which was added in rev. 1.52. This change was done way before
sparc64 switched to a 64-bit time_t so all binaries are expected to have
been recompiled by now.
aid for ABI breakages caused by system call changes. These changes were
done way before sparc64 switched to a 64-bit time_t so all binaries are
expected to have been recompiled by now.
a vlan interface attached to a fxp(4) card when it has not been
initialized yet. We now set the links from our internel TX descriptor
structure to the TX command blocks at attach time rather than at init
time. While I'm here, slightly improve the style in fxp_attach().
PR: kern/78112
Reported by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> and others
Tested by: flz, Gavin Atkinson <gavin.atkinson@ury.york.ac.uk>
MFC after: 1 week
is only the first stage and does not yet handle packages (which will move
to their own disc(s)).
- By default, include the live filesystem on disc1, so that disc1 is now
both an install disc and a fixit disc. The images can still be split
into two separate discs by specifying SEPARATE_LIVEFS=yes.
- Remove /usr/ports from the live filesystem disc as the merged images
do not fit on even i386 otherwise. Also, the merged image already
contains a tarball of the entire ports tree, so /usr/ports was a
duplicate anyway.
- Stop building port READMES for the release sinc they are no longer used
(/usr/ports in the live filesystem was the last remaining user).
- Split cdrom.1 into three targets. cdrom.1 builds the live filesystem
image. cdrom.2 builds the disc1 image. Note that if SEPARATE_LIVEFS
is not specified these both write to the same area. cdrom.3 builds
the bootonly image and the UFS miniroot.
- The various distribution directories are now in a <buildname>
subdirectory of disc1 rather than being in the root so that the merged
image's root is not quite as cluttered.
- The disc1 image is now called disc1.iso and we no longer build a
miniinst.iso image. If CD_EXTRA_BITS is defined, then the iso.1 target
will simply include those bits in the disc1 image that it builds.
MFC after: 1 day
Requested by: re (scottl, etc.)
with EBUSY and a cdrom is not mounted at /cdrom, sysinstall fails to
treat it as an error and thinks that the disk mounted ok. However, it
doesn't find a cdrom.inf file so it complains. Later when it tries to
unmount the disk due to a mediaClose() umount(2) returns an error, and it
never clears its internal mounted flag. The fix here is to properly
handle EBUSY as an error if there isn't a CD already mounted at /cdrom.
- Add a new CDROMInitQuiet variable that can be used to shut up the dialog
box about the mount(2) system call failing when trying to mount a CD-ROM.
This is used by the feature described below.
- When using a fixit CD, first try to see if we can mount the disc in the
drive now and use it as a fixit CD. If not, then prompt the user to
insert the disc and try again. If we do succeed on the first "silent"
probe then we don't ask the user to eject the disk after leaving fixit
mode.
- Add a simple file existence test to make sure that the disc that we mount
really is a livefs disc.
- Explicitly switch back to ttyv0 when using the standard console after
the fixit shell dies. Previously this behavior worked accidentally
because all the fixit modes popped up a dialog box which contained a
hidden switch to ttyv0.
MFC after: 1 day
place.
This moves the dependency on GCC's and other compiler's features into
the central sys/cdefs.h file, while the individual source files can
then refer to #ifdef __COMPILER_FEATURE_FOO where they by now used to
refer to #if __GNUC__ > 3.1415 && __BARC__ <= 42.
By now, GCC and ICC (the Intel compiler) have been actively tested on
IA32 platforms by netchild. Extension to other compilers is supposed
to be possible, of course.
Submitted by: netchild
Reviewed by: various developers on arch@, some time ago
We need to be able to test for the (possible) non-existence of the
FPSWA code.
PR: ia64/77591
Submitted by: Christian Kandeler (christian dot kandeler at hob dot de)
MFC after: 1 day
any pending HTTP request rather than calling shutdown(2) with SHUT_WR.
This makes libfetch (and thus fetch(1)) work again with Squid proxies
configured to not allow half-closed connections.
Reported by: Pawel Worach (pawel.worach AT telia DOT com)
native and foreign architectures and comparing products).
They eliminate most of the differences caused by different
object directory paths, timestamping, and identification.
(Note WORLDTMP was renamed to ${OBJTREE}${.CURDIR}/tmp.)
set the interrupt handler to be INTR_MPSAFE now that xpt_done() can be
called without Giant. Giant is still on the top half of the driver and
the timeout handlers.
paranthesis or brace) into the loop and don't leak the buffer in this
case. Remove the check for Var_Parse returning NULL - it can't.
Patch: 7.92
Submitted by: Max Okumoto <okumoto@ucsd.edu>
chroot area. This fixes make release of HEAD on systems that have a
/usr/src with bsd.endian.mk and a matching /usr/obj but haven't installed
the world in /usr/obj and thus have no bsd.endian.mk /usr/share/mk.
Return always malloc()-ed strings from VarParseShort() to get
rid of warnings when returning string constants from a non-const char *
function.
Patch: 7.90
Submitted by: Max Okumoto <okumoto@ucsd.edu>
timer case:
- Remove the virtual fooclock interrupt counters as they have served their
purpose.
- Adjust the dividers for the different clock such that profhz is now a
multiple of stathz as in the non-lapic case, and the timer now runs at
hz * 2 rather than hz * 3. With the new divisors, the default clock
rates are:
kern.clockrate: { hz = 1000, tick = 1000, profhz = 666, stathz = 133 }
sleeping, so in do_tdsignal, we no longer need to test td_waitset.
now td_waitset is only used to give a thread higher priority when
delivering signal to multithreads process.
This also fixes a bug:
when a thread in sigwait states was suspended and later resumed
by SIGCONT, it can no longer receive signals belong to waitset.
and adjust the path in the Makefile for the upgrade_checks target.
These checks are really feature upgrade checks that should be fast
and just find out whether we need to build a new make before
proceeding with other targets like buildworld. This makes the
place free for a real regression test suite in the old place.
