]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoFix typo.
mux [Mon, 7 Mar 2005 00:24:52 +0000 (00:24 +0000)]
Fix typo.

19 years agoRegen after addition of linux_nosys handler.
sobomax [Mon, 7 Mar 2005 00:23:58 +0000 (00:23 +0000)]
Regen after addition of linux_nosys handler.

19 years agoHandle unimplemented syscall by instantly returning ENOSYS instead of sending
sobomax [Mon, 7 Mar 2005 00:18:06 +0000 (00:18 +0000)]
Handle unimplemented syscall by instantly returning ENOSYS instead of sending
signal first and only then returning ENOSYS to match what real linux does.

PR: kern/74302
Submitted by: Travis Poppe <tlp@LiquidX.org>

19 years agoWhen neither of supported frame type is enabled via kernel options enable
sobomax [Sun, 6 Mar 2005 23:03:58 +0000 (23:03 +0000)]
When neither of supported frame type is enabled via kernel options enable
them all, otherwise the driver will be useless and will only confuse user
as manual page says nothing about the need to enable one of those frame
types explicitly in the kernel config.

PR: kern/47152
Submitted by: Andriy Gapon <avg@icyb.net.ua>
MFC after: 3 days

19 years agoFix ef(4) driver when kernel compiled w/o IPX.
sobomax [Sun, 6 Mar 2005 22:59:40 +0000 (22:59 +0000)]
Fix ef(4) driver when kernel compiled w/o IPX.

MFC after: 3 days

19 years agoAlways produce cpuX entries, even in the case when there is only one CPU
sobomax [Sun, 6 Mar 2005 22:28:14 +0000 (22:28 +0000)]
Always produce cpuX entries, even in the case when there is only one CPU
in the system. This is consistent with what real linuxes do.

PR: kern/75848
Submitted by: Andriy Gapon <avg@icyb.net.ua>
MFC after: 3 days

19 years agoUse [ldr|str]t instead of [ldr|str] when accessing ARM_TP_ADDRESS.
cognet [Sun, 6 Mar 2005 21:12:03 +0000 (21:12 +0000)]
Use [ldr|str]t instead of [ldr|str] when accessing ARM_TP_ADDRESS.

19 years agoMake sure ARM_TP_ADDRESS is accessible right now by calling pmap_fault_fixup,
cognet [Sun, 6 Mar 2005 21:10:31 +0000 (21:10 +0000)]
Make sure ARM_TP_ADDRESS is accessible right now by calling pmap_fault_fixup,
as we can't rely on a trap happening, as it is done normally.
While I'm there, uncomment the call to cpu_dcache_wbinv_range() in
pmap_kenter_internal, as we don't call cpu_dcache_wbinv_all() there anymore.

19 years agoUse correct flags for bus_dma_tag_create().
scottl [Sun, 6 Mar 2005 20:57:54 +0000 (20:57 +0000)]
Use correct flags for bus_dma_tag_create().

19 years agoDeclare as volatile the memory location referenced by a pointer rather than
alc [Sun, 6 Mar 2005 20:57:08 +0000 (20:57 +0000)]
Declare as volatile the memory location referenced by a pointer rather than
the pointer's value.

19 years agoUse the correct flags for bus_dma_tag_create().
scottl [Sun, 6 Mar 2005 20:56:16 +0000 (20:56 +0000)]
Use the correct flags for bus_dma_tag_create().

19 years agoThe m_ext reference counts are potentially shared and modified
alc [Sun, 6 Mar 2005 20:09:00 +0000 (20:09 +0000)]
The m_ext reference counts are potentially shared and modified
asynchronously by different threads.  Thus, declare as volatile the
reference count that is accessed through m_ext's pointer, ref_cnt.
Revert the previous change, revision 1.144, that casts as volatile a
single dereference of ref_cnt.

Reviewed by: bmilekic, dwhite
Problem reported by: kris
MFC after: 3 days

19 years agoAlways go to standby efter each call.
phk [Sun, 6 Mar 2005 19:43:12 +0000 (19:43 +0000)]
Always go to standby efter each call.

19 years agoDon't disable interrupts on a stray interrupt.
phk [Sun, 6 Mar 2005 19:42:32 +0000 (19:42 +0000)]
Don't disable interrupts on a stray interrupt.

19 years agoNew release notes:
hrs [Sun, 6 Mar 2005 19:08:51 +0000 (19:08 +0000)]
New release notes:
atkbd(4) 0x08 flag,
ipnat(8) redirect rule supports non-TCP/UDP packets, and
sppp(4) MPSAFE.

Update release notes:
pppd(8) incorrect CBCP response fix (MFC).

19 years agoFix a mistake in the exponent range.
das [Sun, 6 Mar 2005 19:08:18 +0000 (19:08 +0000)]
Fix a mistake in the exponent range.

19 years agoNew release notes:
hrs [Sun, 6 Mar 2005 18:01:05 +0000 (18:01 +0000)]
New release notes:
Beastie boot menu disabled,
acpi(4) turns ACPI and PCI devices off or to a lower
power state in suspend,
acpi_ibm driver added,
ed(4) ALTQ support,
ipfw(4) ucred-based rules can be used with debug.mpsafenet=1,
TCP-MD5 implementation in KAME IPv4 IPsec,
ftpd(8) 212 and 213 status code support,
gvinum checkparity/rebuildparity/setstate subcommand support,
periodic(8) security report now includes blocked packet
counts by pf(4),
ppp(8) NAS-IP-Address/NAS-Identifier options,
pppd(8) incorrect CBCP response fix, and
rescue(8) now includes BSD tar.

Update release notes:
rc.conf(5) network interface renaming support (MFC), and
markup fix in the entry of systat(1) IPv6 support.

19 years agos/AST/ARNET/ where appropriate.
rushani [Sun, 6 Mar 2005 14:50:42 +0000 (14:50 +0000)]
s/AST/ARNET/ where appropriate.

19 years agoFix typos in a comment.
stefanf [Sun, 6 Mar 2005 13:53:05 +0000 (13:53 +0000)]
Fix typos in a comment.

19 years agoUnlike NetBSD's bcopy(), our bcopy allows the two strings to overlap, even in
cognet [Sun, 6 Mar 2005 13:11:32 +0000 (13:11 +0000)]
Unlike NetBSD's bcopy(), our bcopy allows the two strings to overlap, even in
kernel. So bring in the userland version, instead of just calling memcpy.

19 years agoMFi386: revisions 1.609 and 1.610.
nyan [Sun, 6 Mar 2005 12:59:59 +0000 (12:59 +0000)]
MFi386: revisions 1.609 and 1.610.

19 years agoo s/opt_ifpw.h/opt_ipfw.h/ in the previous commit.
maxim [Sun, 6 Mar 2005 11:22:49 +0000 (11:22 +0000)]
o s/opt_ifpw.h/opt_ipfw.h/ in the previous commit.

Submitted by: YONETANI Tomokazu

19 years agoWhitespace nit. Clarifies which body this line belongs to.
njl [Sun, 6 Mar 2005 10:17:30 +0000 (10:17 +0000)]
Whitespace nit.  Clarifies which body this line belongs to.

19 years agoIncrease the default stacksizes:
marcus [Sun, 6 Mar 2005 07:56:18 +0000 (07:56 +0000)]
Increase the default stacksizes:

32-bit 64-bit
main thread 2 MB 4 MB
other threads 1 MB 2 MB

Approved by: mtm
Adapted from: libpthread

19 years agoBack out previous commit. The description didn't match the commit,
imp [Sun, 6 Mar 2005 06:59:15 +0000 (06:59 +0000)]
Back out previous commit.  The description didn't match the commit,
and marking this interrupt as fast isn't quite right.  It also needs
to be MP Safe, and I've not done that work (yet).

19 years agoUse BUS_PROBE_DEFAULT
imp [Sun, 6 Mar 2005 06:55:11 +0000 (06:55 +0000)]
Use BUS_PROBE_DEFAULT

19 years agoCleanup: u_intXX_t -> uintX_t conversion.
mux [Sun, 6 Mar 2005 05:07:26 +0000 (05:07 +0000)]
Cleanup: u_intXX_t -> uintX_t conversion.

19 years agoUnbreak strsignal(). This is an MFC candidate.
mux [Sun, 6 Mar 2005 03:19:19 +0000 (03:19 +0000)]
Unbreak strsignal().  This is an MFC candidate.

Reported by: Sean McNeil <sean@mcneil.com>

19 years agoWork around a gcc bug. This fixes feholdexcept() et al. at -O1.
das [Sat, 5 Mar 2005 20:34:45 +0000 (20:34 +0000)]
Work around a gcc bug.  This fixes feholdexcept() et al. at -O1.
Symptoms of the problem included assembler warnings and
nondeterministic runtime behavior when a fe*() call that affects the
fpsr is closely followed by a float point op.

The bug (at least, I think it's a bug) is that gcc does not insert a
break between a volatile asm and a dependent instruction if the
volatile asm came from an inlined function.  Volatile asms seem to be
fine in other circumstances, even without -mvolatile-asm-stop, so
perhaps the compiler adds the stop bits before inlining takes place.
The problem does not occur at -O0 because inlining is disabled, and it
doesn't happen at -O2 because -fschedule-insns2 knows better.

19 years agoUse BUS_PROBE_DEFAULT for pci probe return value
imp [Sat, 5 Mar 2005 19:24:22 +0000 (19:24 +0000)]
Use BUS_PROBE_DEFAULT for pci probe return value

19 years agoUse BUS_PROBE_DEFAULT for pci probe return value
imp [Sat, 5 Mar 2005 19:06:12 +0000 (19:06 +0000)]
Use BUS_PROBE_DEFAULT for pci probe return value

19 years agoMFi386: use TUNABLE_ULONG_FETCH to retrieve hw.physmem.
des [Sat, 5 Mar 2005 18:57:30 +0000 (18:57 +0000)]
MFi386: use TUNABLE_ULONG_FETCH to retrieve hw.physmem.

19 years agoUse BUS_PROBE_DEFAULT for pci probe return value
imp [Sat, 5 Mar 2005 18:30:12 +0000 (18:30 +0000)]
Use BUS_PROBE_DEFAULT for pci probe return value

19 years agoUse BUS_PROBE_DEFAULT for pci probe return value
imp [Sat, 5 Mar 2005 18:17:35 +0000 (18:17 +0000)]
Use BUS_PROBE_DEFAULT for pci probe return value

19 years agoReplace goto with continue.
des [Sat, 5 Mar 2005 18:14:44 +0000 (18:14 +0000)]
Replace goto with continue.

19 years agoUse BUS_PROBE_DEFAULT for pci probe return value
imp [Sat, 5 Mar 2005 18:10:49 +0000 (18:10 +0000)]
Use BUS_PROBE_DEFAULT for pci probe return value

19 years agoUse TUNABLE_ULONG_FETCH to retrieve hw.physmem; getenv_quad() will take
des [Sat, 5 Mar 2005 18:10:04 +0000 (18:10 +0000)]
Use TUNABLE_ULONG_FETCH to retrieve hw.physmem; getenv_quad() will take
care of the multiplier suffix.

19 years agoReplace goto with continue.
des [Sat, 5 Mar 2005 18:02:21 +0000 (18:02 +0000)]
Replace goto with continue.

19 years agoRevise the minimum RAM requirements to 24MB for i386 (and implicitly
bmah [Sat, 5 Mar 2005 16:49:01 +0000 (16:49 +0000)]
Revise the minimum RAM requirements to 24MB for i386 (and implicitly
pc98).

(While here, remove mention of 80386 custom kernels since support for the
80386 has been removed from CURRENT.)

Feedback from: bde, des, imp, jhb

19 years agoNew release notes: Background fsck(8) summary recomputation (+MFC),
bmah [Sat, 5 Mar 2005 16:26:32 +0000 (16:26 +0000)]
New release notes:  Background fsck(8) summary recomputation (+MFC),
newfs(8) snapshot warning (+MFC), newsyslog(8) -N, systat(1) -netstat
IPv6-ware (+MFC), merged install/fixit CDROMs (+MFC).

MFCs noted:  IPFIREWALL_FORWARD_EXTENDED, strftime(3) GNU extensions,
rc.conf(5) tmpmfs_flags and varmfs_flags.

19 years agoTeach getenv_quad() to recognize k/m/g/t suffixes in both lower- and
des [Sat, 5 Mar 2005 15:52:12 +0000 (15:52 +0000)]
Teach getenv_quad() to recognize k/m/g/t suffixes in both lower- and
upper-case.  This means (almost) all tunables now support those suffixes.

19 years agoAdd support for fxp(4) cards found in Sony FS570 laptops.
mux [Sat, 5 Mar 2005 13:52:28 +0000 (13:52 +0000)]
Add support for fxp(4) cards found in Sony FS570 laptops.

Submitted by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk>

19 years agoAllocate umtx_q from heap instead of stack, this avoids
davidxu [Sat, 5 Mar 2005 09:15:03 +0000 (09:15 +0000)]
Allocate umtx_q from heap instead of stack, this avoids
page fault panic in kernel under heavy swapping.

19 years agoFix the truly bewildering sorting order near the middle of the
bmah [Sat, 5 Mar 2005 04:58:55 +0000 (04:58 +0000)]
Fix the truly bewildering sorting order near the middle of the
"Userland Changes" section.  I'm pretty sure this is all my
fault...only a native English^H^H^H^H^H^H^HAmerican speaker could mess
it up this badly.

No content changes.

19 years agoFix a few nits here and there.
bmah [Sat, 5 Mar 2005 04:53:59 +0000 (04:53 +0000)]
Fix a few nits here and there.

19 years agoAdd comments on setting resource limits.
grog [Sat, 5 Mar 2005 01:04:18 +0000 (01:04 +0000)]
Add comments on setting resource limits.

19 years agoAdd targets to build cscope name files and TAGS file for etags.
grog [Sat, 5 Mar 2005 00:56:15 +0000 (00:56 +0000)]
Add targets to build cscope name files and TAGS file for etags.

19 years agoThe td_waitset is pointing to a stack address when thread is waiting
davidxu [Fri, 4 Mar 2005 22:46:31 +0000 (22:46 +0000)]
The td_waitset is pointing to a stack address when thread is waiting
for a signal, because kernel stack is swappable, this causes page fault
in kernel under heavy swapping case. Fix this bug by eliminating unneeded
code.

19 years agoRemove superflous .Pp calls.
brueffer [Fri, 4 Mar 2005 22:34:57 +0000 (22:34 +0000)]
Remove superflous .Pp calls.

Submitted by: Joel Dahl <joel@automatvapen.se>
MFC after: 3 days

19 years ago- sparc64/fhc/fhc.c:
marius [Fri, 4 Mar 2005 22:23:21 +0000 (22:23 +0000)]
- sparc64/fhc/fhc.c:
  Change fhc(4) to use IRQ numbers instead of RIDs for allocating the
  IRQs of children. This works similar to e.g. sbus(4), i.e. add the
  IRQ resources as fully specified to the resource lists of the children,
  allocate them like normal. When establishing the interrupt search the
  interrupt maps of the children for a matching INO to determine which
  map we need to write the fully specified interrupt number to and to
  enable the mapping (before the RID was used to indicate which interrupt
  map to use).

- dev/puc/puc.c:
  Revert rev. 1.38, with the above change fhc(4) no longer needs special
  treatment for allocating IRQs.

Thanks to: joerg for providing access to an E3500

19 years agoLet central(4) manage the resources of its children so they don't need
marius [Fri, 4 Mar 2005 22:21:11 +0000 (22:21 +0000)]
Let central(4) manage the resources of its children so they don't need
to fiddle with OFW themselves.

Thanks to: joerg for providing access to an E3500

19 years agoAddendum to netchild's C compiler abstraction mega-patch which somehow
joerg [Fri, 4 Mar 2005 21:26:07 +0000 (21:26 +0000)]
Addendum to netchild's C compiler abstraction mega-patch which somehow
have been forgotten in my previous commit.

Submitted by: netchild

19 years agoAlways use a separate disc for the livefs on ia64 since the merged image
jhb [Fri, 4 Mar 2005 20:37:20 +0000 (20:37 +0000)]
Always use a separate disc for the livefs on ia64 since the merged image
doesn't fit on one disc.

Tested by: marcel
X-MFC after: 30 seconds

19 years agoDo not require the pty(4) majors to be anything in particular.
green [Fri, 4 Mar 2005 20:23:32 +0000 (20:23 +0000)]
Do not require the pty(4) majors to be anything in particular.

19 years agoNew release notes:
hrs [Fri, 4 Mar 2005 19:52:34 +0000 (19:52 +0000)]
New release notes:
vkbd(4) added,
acpi_snc added,
gx(4) removed,
old vinum(4) removed,
libxpg4 removed,
a separate directory for named(8) dynamic zones added,
rm(1) SUSv3 conformance,
strftime(3) GNU extensions added, and
rc.conf(5) variables: tmpmfs_flags and varmfs_flags.

Update release notes:
Fix a typo (s/machine-indepedent/machine-independent/).

19 years agoOn i386 platforms at least, ACPI throttling is accomplished by the chipset
njl [Fri, 4 Mar 2005 18:59:54 +0000 (18:59 +0000)]
On i386 platforms at least, ACPI throttling is accomplished by the chipset
modulating the STPCLK# pin based on the duty cycle.  Since p4tcc uses the
same mechanism (but internal to the CPU), we triggered a hang on some
systems at low frequencies when both were in use.  Now, disable
acpi_throttle when p4tcc is also present.

Tested by: Kevin Oberman

19 years agoDon't allow to synchronize a plex that is already sychronizing.
le [Fri, 4 Mar 2005 16:43:40 +0000 (16:43 +0000)]
Don't allow to synchronize a plex that is already sychronizing.

Reset the 'syncing' flag in case of errors, too.

Some cosmetics.

19 years agoMinor changes:
marius [Fri, 4 Mar 2005 16:01:57 +0000 (16:01 +0000)]
Minor changes:
- Use FBSDID.
- Remove unused macro.
- Use auto-generated typedefs for the prototypes of the bus and device
  interface functions.
- Terminate the output of device_printf(9) with a newline char.
- Honour the return values of malloc(), OF_getprop(), etc.
- Use __func__ instead of hardcoded function names.
- Print the physical slot number and the board model on attach.

MFC after: 1 month

19 years agoMinor changes:
marius [Fri, 4 Mar 2005 15:17:05 +0000 (15:17 +0000)]
Minor changes:
- Use FBSDID.
- Remove an unused include.
- Use auto-generated typedefs for the prototypes of the device interface
  functions.
- Terminate the output of device_printf(9) with a newline char.
- Honour the return value of malloc(3).

MFC after: 1 month

19 years agoDon't use the ptoa() to set the 'realmem' variable. Because F/pc98's policy
nyan [Fri, 4 Mar 2005 14:26:23 +0000 (14:26 +0000)]
Don't use the ptoa() to set the 'realmem' variable.  Because F/pc98's policy
is to keep the same as F/i386.

19 years agoNew Ukrainian locale: uk_UA.CP1251.
ru [Fri, 4 Mar 2005 14:24:30 +0000 (14:24 +0000)]
New Ukrainian locale: uk_UA.CP1251.

Submitted by: Alexander Peresunko

19 years agoUse ``.Pq Ql'' to quote single characters, instead of ``.Pq Li''.
keramida [Fri, 4 Mar 2005 13:16:50 +0000 (13:16 +0000)]
Use ``.Pq Ql'' to quote single characters, instead of ``.Pq Li''.
This makes the resulting output more aesthetically pleasing in
text-only terminals:

Reviewed by: ru

19 years agoAdd a section describing the exit status of the shell.
keramida [Fri, 4 Mar 2005 12:40:07 +0000 (12:40 +0000)]
Add a section describing the exit status of the shell.

Obtained from: NetBSD

19 years ago- Readd HARDWARE section to unbreak Hardware Notes.
simon [Fri, 4 Mar 2005 10:03:40 +0000 (10:03 +0000)]
- Readd HARDWARE section to unbreak Hardware Notes.
- Cleanup of mdoc markup.

19 years agoAnother good Dijkstra quote.
phk [Fri, 4 Mar 2005 09:47:22 +0000 (09:47 +0000)]
Another good Dijkstra quote.

19 years agoFix the ioctl path by ensuring that amr_start1() gets called for commands
scottl [Fri, 4 Mar 2005 06:11:00 +0000 (06:11 +0000)]
Fix the ioctl path by ensuring that amr_start1() gets called for commands
with no associated data.  Also revert previous changes that allocate off
of the stack instead of using malloc, as it's not needed.  Many thanks to
LSI for investigating and fixing these problems.

Submitted by: rajeshpr @ lsil . com

19 years agoAdd and clarify the names of many supported cards, and tweak some wording.
scottl [Fri, 4 Mar 2005 06:02:13 +0000 (06:02 +0000)]
Add and clarify the names of many supported cards, and tweak some wording.

Submitted by: rajeshpr @ lsil . com

19 years agoCheck for some impossible frequencies that some systems use to indicate
njl [Fri, 4 Mar 2005 05:29:56 +0000 (05:29 +0000)]
Check for some impossible frequencies that some systems use to indicate
they don't actually support Px states.

19 years agoAllocate and map a 12k data buffer such that it is contiguous in
iedowse [Fri, 4 Mar 2005 00:53:03 +0000 (00:53 +0000)]
Allocate and map a 12k data buffer such that it is contiguous in
both the kernel and vm86 virtual address spaces. Use this to increase
the maximum VESA save state size we can handle.

19 years agoHandle endianness correctly.
cognet [Thu, 3 Mar 2005 23:27:09 +0000 (23:27 +0000)]
Handle endianness correctly.

Spotted out by: jmg

19 years agoAdd missing quote.
phk [Thu, 3 Mar 2005 18:32:50 +0000 (18:32 +0000)]
Add missing quote.

19 years agoWisdom from 1972.
phk [Thu, 3 Mar 2005 18:07:23 +0000 (18:07 +0000)]
Wisdom from 1972.

19 years agoRecently, it was reported to me that you could provoke a double fault
wpaul [Thu, 3 Mar 2005 17:35:05 +0000 (17:35 +0000)]
Recently, it was reported to me that you could provoke a double fault
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.

19 years agoChange fix a bug where the length of data written by snprintf was
alfred [Thu, 3 Mar 2005 17:20:36 +0000 (17:20 +0000)]
Change fix a bug where the length of data written by snprintf was
being mishandled by using accumulation (into an uninitialized
variable) instead of direct assignment.

Reviewed by: scottl

19 years agoIn linux emulation layer try to detect attempt to use linux_clone() to
sobomax [Thu, 3 Mar 2005 16:57:55 +0000 (16:57 +0000)]
In linux emulation layer try to detect attempt to use linux_clone() to
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

19 years agoInstall rlogind(8). Pointy hat is available on request.
ru [Thu, 3 Mar 2005 15:59:32 +0000 (15:59 +0000)]
Install rlogind(8).  Pointy hat is available on request.

19 years agoAdd the -N option to not rotate any files. This is to be used in
brooks [Thu, 3 Mar 2005 15:47:32 +0000 (15:47 +0000)]
Add the -N option to not rotate any files.  This is to be used in
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

19 years agoStyle: remove unneeded parantheses in conditionals.
harti [Thu, 3 Mar 2005 11:34:04 +0000 (11:34 +0000)]
Style: remove unneeded parantheses in conditionals.

19 years agoConstify the code paths for the 'M' and 'N' modifiers.
harti [Thu, 3 Mar 2005 11:26:28 +0000 (11:26 +0000)]
Constify the code paths for the 'M' and 'N' modifiers.

Patch: 7.100

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoCisco uses milliseconds for uptime. This is stupid. Nobody cares of such
glebius [Thu, 3 Mar 2005 11:01:05 +0000 (11:01 +0000)]
Cisco uses milliseconds for uptime. This is stupid. Nobody cares of such
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

19 years agoFix parsing of archives: in System V archive format the member names
harti [Thu, 3 Mar 2005 10:46:23 +0000 (10:46 +0000)]
Fix parsing of archives: in System V archive format the member names
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.

19 years agoFix parsing of archive specifications on the target side of
harti [Thu, 3 Mar 2005 10:09:39 +0000 (10:09 +0000)]
Fix parsing of archive specifications on the target side of
dependency lines.  It seems that nobody is actually is using
the archive-feature of make.

19 years agoUpdate the description comment of the function to reflect the
harti [Thu, 3 Mar 2005 10:07:28 +0000 (10:07 +0000)]
Update the description comment of the function to reflect the
previous change.

19 years agoRemove important entry being X that's 9 months old
imp [Thu, 3 Mar 2005 08:44:33 +0000 (08:44 +0000)]
Remove important entry being X that's 9 months old

19 years agoRemove references to 4.x upgrades, since those no longer are possible
imp [Thu, 3 Mar 2005 08:43:33 +0000 (08:43 +0000)]
Remove references to 4.x upgrades, since those no longer are possible
(one must upgrade to 5.3 before first jumping to current).

19 years agoTransform VarParseLong() so that the lifetime and the constness of
harti [Thu, 3 Mar 2005 07:25:04 +0000 (07:25 +0000)]
Transform VarParseLong() so that the lifetime and the constness of
variables gets clearer. Transform if() conditions to make the flow
clearer.

Patch: 7.93-99

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoMarkup nits.
des [Thu, 3 Mar 2005 07:20:57 +0000 (07:20 +0000)]
Markup nits.

19 years agoCorrect the default value for %%TINDERBOX%%.
des [Thu, 3 Mar 2005 07:18:01 +0000 (07:18 +0000)]
Correct the default value for %%TINDERBOX%%.

19 years agoDoh, erase unused variable...
grehan [Thu, 3 Mar 2005 06:38:00 +0000 (06:38 +0000)]
Doh, erase unused variable...

19 years agoMAXPATHLEN is 1024, which means NdisOpenFile() and ndis_find_sym() were
wpaul [Thu, 3 Mar 2005 03:51:02 +0000 (03:51 +0000)]
MAXPATHLEN is 1024, which means NdisOpenFile() and ndis_find_sym() were
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.

19 years agoFix a problem in the Skinny ALG where a specially crafted packet could cause
marcus [Thu, 3 Mar 2005 03:06:37 +0000 (03:06 +0000)]
Fix a problem in the Skinny ALG where a specially crafted packet could cause
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

19 years agoInsert volatile cast to discourage gcc from optimizing the read outside
dwhite [Thu, 3 Mar 2005 02:41:37 +0000 (02:41 +0000)]
Insert volatile cast to discourage gcc from optimizing the read outside
of the while loop.

Suggested by: alc
MFC after: 1 day

19 years agoIf a user or group is not known, report the problem user/group, rather than
delphij [Thu, 3 Mar 2005 02:17:20 +0000 (02:17 +0000)]
If a user or group is not known, report the problem user/group, rather than
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)

19 years agoRemove the transition aid for the change of the sparc64 default system
marius [Thu, 3 Mar 2005 01:26:40 +0000 (01:26 +0000)]
Remove the transition aid for the change of the sparc64 default system
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.

19 years agoRemove COMPAT_FREEBSD4 again. It was added in rev. 1.41 as a transition
marius [Thu, 3 Mar 2005 01:25:33 +0000 (01:25 +0000)]
Remove COMPAT_FREEBSD4 again. It was added in rev. 1.41 as a transition
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.

19 years agoFix the panic recently reported on -current@ occuring when configuring
mux [Thu, 3 Mar 2005 00:22:59 +0000 (00:22 +0000)]
Fix the panic recently reported on -current@ occuring when configuring
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

19 years agoAutomatically switch to ttyv3 where the fixit shell is running when using
jhb [Wed, 2 Mar 2005 22:48:13 +0000 (22:48 +0000)]
Automatically switch to ttyv3 where the fixit shell is running when using
the standard console and any fixit mode.

MFC after: 1 day

19 years agoRework how we layout the CD/DVD ISO images for releases. Note that this
jhb [Wed, 2 Mar 2005 22:46:07 +0000 (22:46 +0000)]
Rework how we layout the CD/DVD ISO images for releases.  Note that this
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.)

19 years ago- Fix a bug in sysinstall related to mounting CD-ROMs. If mount(2) fails
jhb [Wed, 2 Mar 2005 22:27:22 +0000 (22:27 +0000)]
- Fix a bug in sysinstall related to mounting CD-ROMs.  If mount(2) fails
  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

19 years agofix a bug where bpf would try to wakeup before updating the state.. This
jmg [Wed, 2 Mar 2005 21:59:39 +0000 (21:59 +0000)]
fix a bug where bpf would try to wakeup before updating the state.. This
was causing kqueue not to see the correct state and not wake up a process
that is waiting...

Submitted by: nCircle Network Security, Inc.