]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoUpdate usage for the addition of -n in rev 1.35 of mv.c.
johan [Tue, 9 Jul 2002 17:45:13 +0000 (17:45 +0000)]
Update usage for the addition of -n in rev 1.35 of mv.c.

Approved by: obrien, sheldonh (mentor).

22 years agoRevert previous delta; warnings fixed by obrien:
sheldonh [Tue, 9 Jul 2002 17:29:37 +0000 (17:29 +0000)]
Revert previous delta; warnings fixed by obrien:

rev 1.5 rc/contrib/gdb/gdb/gdbserver/gdbreplay.c

22 years agoReplace the 'at least the amount of physical memory + 64K' with a better
dillon [Tue, 9 Jul 2002 17:27:08 +0000 (17:27 +0000)]
Replace the 'at least the amount of physical memory + 64K' with a better
explanation in regards to sizing the dump device.

22 years agoQuiet warnings by making this ANSI-C clean.
obrien [Tue, 9 Jul 2002 17:01:47 +0000 (17:01 +0000)]
Quiet warnings by making this ANSI-C clean.

22 years agoRegen for SMC 2206USB/ETH EZ Connect adapter.
joe [Tue, 9 Jul 2002 16:46:31 +0000 (16:46 +0000)]
Regen for SMC 2206USB/ETH EZ Connect adapter.

22 years agoMFNetBSD: Add support for the SMC 2206USB/ETH EZ Connect adapter (Pegasus II)
joe [Tue, 9 Jul 2002 16:45:03 +0000 (16:45 +0000)]
MFNetBSD: Add support for the SMC 2206USB/ETH EZ Connect adapter (Pegasus II)

22 years agoEnable emulation of the F_GETLK64, F_SETLK64, and F_SETLKW64
robert [Tue, 9 Jul 2002 15:57:12 +0000 (15:57 +0000)]
Enable emulation of the F_GETLK64, F_SETLK64, and F_SETLKW64
lock commands arguments to linux_fcntl64().

22 years agoThe comment marked with XXX was right: emulate SVR4 for
robert [Tue, 9 Jul 2002 14:59:27 +0000 (14:59 +0000)]
The comment marked with XXX was right: emulate SVR4 for
ELF binaries branded with ELFOSABI_SYSV, this is reported
to work and brandelf(1) puts this type into files if "SVR4"
was specified.

22 years agoBring this in line with what I'm using.
benno [Tue, 9 Jul 2002 13:46:27 +0000 (13:46 +0000)]
Bring this in line with what I'm using.

22 years agoAdd an implementation for pmap_zero_page_area.
benno [Tue, 9 Jul 2002 13:44:24 +0000 (13:44 +0000)]
Add an implementation for pmap_zero_page_area.

22 years agoAdd the OF_getetheraddr function required by if_gem.
benno [Tue, 9 Jul 2002 13:43:06 +0000 (13:43 +0000)]
Add the OF_getetheraddr function required by if_gem.

22 years agoTidy up trap vector and external interrupt setup.
benno [Tue, 9 Jul 2002 13:40:37 +0000 (13:40 +0000)]
Tidy up trap vector and external interrupt setup.

22 years agoDriver for the Apple UniNorth Host-PCI bridge.
benno [Tue, 9 Jul 2002 13:34:09 +0000 (13:34 +0000)]
Driver for the Apple UniNorth Host-PCI bridge.

This is in a PowerMac-specific subdirectory as it is hoped that we will support
more than just the PowerMac platform.

22 years agoAdd ofw_pci.c in the pci case.
benno [Tue, 9 Jul 2002 13:29:18 +0000 (13:29 +0000)]
Add ofw_pci.c in the pci case.

22 years agoOpenFirmware PCI support code.
benno [Tue, 9 Jul 2002 13:27:58 +0000 (13:27 +0000)]
OpenFirmware PCI support code.

This and the sparc64 equivalent should probably be merged at some point.

22 years agoOops, forgot to set the suspended flag for threads that are created
deischen [Tue, 9 Jul 2002 13:24:52 +0000 (13:24 +0000)]
Oops, forgot to set the suspended flag for threads that are created
initially suspended.  This was preventing such threads from getting
resumed.

Reported by: Bill Huey <billh@gnuppy.monkey.org>

22 years agoChanges for KSE3.
benno [Tue, 9 Jul 2002 12:57:23 +0000 (12:57 +0000)]
Changes for KSE3.

Submitted by: Peter Grehan <peterg@ptree32.com.au>

22 years agoAdd this file, which I forgot in a previous commit.
benno [Tue, 9 Jul 2002 12:50:58 +0000 (12:50 +0000)]
Add this file, which I forgot in a previous commit.

This relates to the trap/interrupt cleanup.

Submitted by: Peter Grehan <peterg@ptree32.com.au>

22 years ago1) Add busdma machdep code.
benno [Tue, 9 Jul 2002 12:47:14 +0000 (12:47 +0000)]
1) Add busdma machdep code.
2) Add bus_pio.h and bus_memio.h (which do nothing).

Submitted by: Peter Grehan <peterg@ptree32.com.au> (1)

22 years agoRegen:
sheldonh [Tue, 9 Jul 2002 11:50:14 +0000 (11:50 +0000)]
Regen:

* Hart: rev 313 of pcidevs.txt (2002-07-05)
* Boemler: vendors.txt (2002-06-24)

22 years agoDriver for OpenPIC compatible interrupt controllers.
benno [Tue, 9 Jul 2002 11:26:10 +0000 (11:26 +0000)]
Driver for OpenPIC compatible interrupt controllers.
It's fairly PowerMac specific at the moment, but that should be fixable.

22 years agoAdd definitions for the industry-standard DLT density codes.
joerg [Tue, 9 Jul 2002 11:26:09 +0000 (11:26 +0000)]
Add definitions for the industry-standard DLT density codes.

Obtained from: Quantum DLT8000 manual
MFC after: 1 week

22 years ago- Add the "compatible" property to the list that we keep in ivars.
benno [Tue, 9 Jul 2002 11:13:33 +0000 (11:13 +0000)]
- Add the "compatible" property to the list that we keep in ivars.
- Add interrupt alloc/setup/teardown/dealloc support, via whichever PIC
  OpenFirmware gives us.

22 years agoAdd interrupt handling support code.
benno [Tue, 9 Jul 2002 11:12:20 +0000 (11:12 +0000)]
Add interrupt handling support code.

I've tried to make this fairly platform-independant as some PowerPC platforms
may not have openpic-style interrupt controllers.  This may not have the best
performance but it works for now.

22 years agoClose file descriptors when [n]>&- and [n]<&- redirections are used.
tjr [Tue, 9 Jul 2002 08:55:46 +0000 (08:55 +0000)]
Close file descriptors when [n]>&- and [n]<&- redirections are used.
This was broken by rev. 1.16.

PR: 40334
MFC after: 1 week

22 years agoThe rate between EUR and USD has never been as large as our old value
joerg [Tue, 9 Jul 2002 08:47:03 +0000 (08:47 +0000)]
The rate between EUR and USD has never been as large as our old value
(1.12) since the introductions as a common EU currency.  Anyway, since
the rate now seems to stabilize just around 1.00, fix up our value
here.

MFC after: 1 day

22 years agoDefine NO_WERROR to unbreak world until the maintainer has had time
sheldonh [Tue, 9 Jul 2002 08:37:36 +0000 (08:37 +0000)]
Define NO_WERROR to unbreak world until the maintainer has had time
to coordinate fixes for this contrib code.

22 years agoChange one variable to make it easier to switch between ipfw and ipfw2
luigi [Tue, 9 Jul 2002 06:53:38 +0000 (06:53 +0000)]
Change one variable to make it easier to switch between ipfw and ipfw2

22 years agoFix whitespace botch in previous commit.
peter [Tue, 9 Jul 2002 06:47:29 +0000 (06:47 +0000)]
Fix whitespace botch in previous commit.

22 years agoDon't define wchar_t if we are a C++ compiler.
obrien [Tue, 9 Jul 2002 05:13:30 +0000 (05:13 +0000)]
Don't define wchar_t if we are a C++ compiler.

PR: 31864, 40084

22 years agoPreliminary NEWCARD docs.
imp [Tue, 9 Jul 2002 05:08:49 +0000 (05:08 +0000)]
Preliminary NEWCARD docs.

22 years agoUpdate dates.
imp [Tue, 9 Jul 2002 04:46:30 +0000 (04:46 +0000)]
Update dates.

22 years agoDocument what the pcic driver in newcard will support.
imp [Tue, 9 Jul 2002 04:43:30 +0000 (04:43 +0000)]
Document what the pcic driver in newcard will support.

22 years agoFinish the repo copy of pcic.4 -> oldcard.4. oldcard.4 will document
imp [Tue, 9 Jul 2002 03:50:29 +0000 (03:50 +0000)]
Finish the repo copy of pcic.4 -> oldcard.4.  oldcard.4 will document
the legacy pccard mode stuff called 'oldcard.'  pcic.4 will be
hijacked for newcard purposes.

22 years agoPrint out commands with NTOFD/NFROMFD redirections that close the
tjr [Tue, 9 Jul 2002 03:26:47 +0000 (03:26 +0000)]
Print out commands with NTOFD/NFROMFD redirections that close the
descriptors (">&-" or similar) correctly in the jobs(1) command.

22 years agoAdd C code that counts the bits in a 32 bit word, right next to the C code
peter [Tue, 9 Jul 2002 02:56:10 +0000 (02:56 +0000)]
Add C code that counts the bits in a 32 bit word, right next to the C code
that reverses the bits in a 32 bit word.  Who thinks up this stuff?

22 years agoAdd ability to only beep when mail arrives.
johan [Tue, 9 Jul 2002 02:16:49 +0000 (02:16 +0000)]
Add ability to only beep when mail arrives.

comsat:
        only send two bell charecters if S_IXGRP is set and S_IXUSR is not.

biff:
        add new option 'b' to set S_IXGRP.

PR:             10931
Submitted by:   Andrew J. Korty <ajk@purdue.edu>
Approved by:    sheldonh (mentor)
MFC after:      1 month

22 years agoAdd DIAGNOSTICS section to describe exit values.
johan [Tue, 9 Jul 2002 02:13:26 +0000 (02:13 +0000)]
Add DIAGNOSTICS section to describe exit values.

Approved by:    sheldonh (mentor)
MFC after:      1 week

22 years agoRemove an unused type.
mike [Tue, 9 Jul 2002 01:20:51 +0000 (01:20 +0000)]
Remove an unused type.

22 years agoMove the type definition of ointhand2_t from i386/include/types.h to
mike [Tue, 9 Jul 2002 01:16:18 +0000 (01:16 +0000)]
Move the type definition of ointhand2_t from i386/include/types.h to
i386/isa/isa_device.h.  This is a more appropriate location and
helps restrict <machine/types.h> to only types that exist on all
platforms.

22 years agoDelay the creation of the ugenX device node until we're certain
joe [Tue, 9 Jul 2002 01:01:25 +0000 (01:01 +0000)]
Delay the creation of the ugenX device node until we're certain
that the attach succeeded.  (Fixes a potential panic for devices
that fail to attach properly and are subsquently unplugged and then
plugged back in again.)

22 years agoOops, the previous revision (1.22) introduced a potential alignment
iedowse [Mon, 8 Jul 2002 23:53:21 +0000 (23:53 +0000)]
Oops, the previous revision (1.22) introduced a potential alignment
issue, since the MAXBSIZE-sized buffers are accessed as arrays of
block pointers, but were declared as char[] arrays. Use a union to
avoid this, which also makes a number of casts unnecessary.

Pointed out by: bde
Reviewed by: bde

22 years agovm_page_queue_free_mtx is a spin mutex, not a normal sleep mutex.
peter [Mon, 8 Jul 2002 23:12:37 +0000 (23:12 +0000)]
vm_page_queue_free_mtx is a spin mutex, not a normal sleep mutex.
I do not know why this didn't panic my box, but I have most certainly
been using it:
peter@overcee[3:14pm]~src/sys/i386/i386-110> sysctl -a | grep zero
vm.stats.misc.zero_page_count: 2235
vm.stats.misc.cnt_prezero: 638951
vm.idlezero_enable: 1
vm.idlezero_maxrun: 16

Submitted by: Tor.Egge@cvsup.no.freebsd.org
Approved by: Tor's patches are never wrong. :-)

22 years agoTurn the zeroidle process off for SMP systems, there is still a possible
peter [Mon, 8 Jul 2002 23:09:11 +0000 (23:09 +0000)]
Turn the zeroidle process off for SMP systems, there is still a possible
TLB problem when bouncing from one cpu to another (the original cpu will
not have purged its TLB if the it simply went idle).

Pointed out by:  Tor.Egge@cvsup.no.freebsd.org
Approved by: Tor is never wrong. :-)

22 years agoFix a bug caused by dereferencing an invalid pointer when
luigi [Mon, 8 Jul 2002 22:57:35 +0000 (22:57 +0000)]
Fix a bug caused by dereferencing an invalid pointer when
no punch_fw was used.
Fix another couple of bugs which prevented rules from being
installed properly.

On passing, use IPFW2 instead of NEW_IPFW to compile the new code,
and slightly simplify the instruction generation code.

22 years agoNo functional changes, but:
luigi [Mon, 8 Jul 2002 22:46:01 +0000 (22:46 +0000)]
No functional changes, but:

Following Darren's suggestion, make Dijkstra happy and rewrite the
ipfw_chk() main loop removing a lot of goto's and using instead a
variable to store match status.

Add a lot of comments to explain what instructions are supposed to
do and how -- this should ease auditing of the code and make people
more confident with it.

In terms of code size: the entire file takes about 12700 bytes of text,
about 3K of which are for the main function, ipfw_chk(), and 2K (ouch!)
for ipfw_log().

22 years agoRemove one unused command name.
luigi [Mon, 8 Jul 2002 22:39:19 +0000 (22:39 +0000)]
Remove one unused command name.

22 years agoForgot to update one field name in one of the latest commits.
luigi [Mon, 8 Jul 2002 22:37:55 +0000 (22:37 +0000)]
Forgot to update one field name in one of the latest commits.

22 years agoBack out previous TCBHASHSIZE change. This should not be a kernel option.
mp [Mon, 8 Jul 2002 22:00:43 +0000 (22:00 +0000)]
Back out previous TCBHASHSIZE change. This should not be a kernel option.

Pointed out by: bde

22 years agoAdd my interest in etc/rc.* to help keep etc/rc.d in sync as much as possible.
gordon [Mon, 8 Jul 2002 21:07:42 +0000 (21:07 +0000)]
Add my interest in etc/rc.* to help keep etc/rc.d in sync as much as possible.

22 years agoRemove debugging nit
gordon [Mon, 8 Jul 2002 21:04:16 +0000 (21:04 +0000)]
Remove debugging nit

Submitted by: bmah

22 years agoGrammar fix
blackend [Mon, 8 Jul 2002 20:57:19 +0000 (20:57 +0000)]
Grammar fix

PR: docs/40210
Submitted by: Munish Chopra <mchopra@engmail.uwaterloo.ca>
Reviewed by: keramida
Approved by: orion
MFC after: 7 days

22 years agoSimplify isp(4) line; add sa(4) for pre-commit review. Remove myself from
mjacob [Mon, 8 Jul 2002 20:42:44 +0000 (20:42 +0000)]
Simplify isp(4) line; add sa(4) for pre-commit review. Remove myself from
CAM review list.

22 years agoRegister an interest in pseudofs, procfs and linprocfs.
des [Mon, 8 Jul 2002 20:29:24 +0000 (20:29 +0000)]
Register an interest in pseudofs, procfs and linprocfs.

22 years agoTypo fix: Setlogin() -> setlogin().
mux [Mon, 8 Jul 2002 20:16:15 +0000 (20:16 +0000)]
Typo fix: Setlogin() -> setlogin().

Submitted by: Olivier Houchard <cognet@ci0.org>

22 years agoFixed some world breakage caused by not updating clients when <timeconv.h>
bde [Mon, 8 Jul 2002 20:13:07 +0000 (20:13 +0000)]
Fixed some world breakage caused by not updating clients when <timeconv.h>
was split off from <time.h>.  This became fatal here when -Werror was
reenabled.

22 years agoFixed some world breakage caused by not updating clients when <timeconv.h>
bde [Mon, 8 Jul 2002 19:49:52 +0000 (19:49 +0000)]
Fixed some world breakage caused by not updating clients when <timeconv.h>
was split off from <time.h>.  This became fatal here when -Werror was
reenabled.

22 years agoAdd a VFS_START() call in vfs_mountroot_try() for the sake
mux [Mon, 8 Jul 2002 19:10:15 +0000 (19:10 +0000)]
Add a VFS_START() call in vfs_mountroot_try() for the sake
of being correct.  None of the root mountable filesystems
do something at VFS_START().

Shorten a comment to fix a style bug while I'm here.

PR: kern/18505

22 years agolink in the man page for thread_exit
julian [Mon, 8 Jul 2002 17:52:00 +0000 (17:52 +0000)]
link in the man page for thread_exit
Also change some line breaks.

22 years agoAdd 2002 to copyright.
mjacob [Mon, 8 Jul 2002 17:48:39 +0000 (17:48 +0000)]
Add 2002 to copyright.

Oops; I forgot for previous delta... If we're and FC or ULTRA2 or better
card, we can have a 1024 element request queue instead of 256.

MFC after: 1 week

22 years agoAdd get/set param ioctl support.
mjacob [Mon, 8 Jul 2002 17:42:47 +0000 (17:42 +0000)]
Add get/set param ioctl support.

Remove sim queue freezes for resource shortages. I've had too many
strange race conditions where I freeze on a resource shortage but
never get unfrozen.

Consolidate the remaining sim queue freeze condition (for loopdown)
into an inline with debug messages that allows us to track problems
at ISP_LOGDEBUG0 level easier. Change a bunch of debug messages about
loop down/up conditions to ISP_LOGDEBUG0 level.

Remove dead isp_relsim code.

Change some internal flag stuff for efficiency.

Complain vociferously if we try and use our FC scratch area while it's
busy being used already (I mean, if we don't have solaris' ability
to sleep as an interrupt thread which would allow us to just use
a p/v semaphore, at least *say* when you've just borked yourself).

Add infrastructure to allow overrides of hard loopid && initiator
id from boot variables.

Fix the usual quota of silly bugs:

 + 'ktmature' needs to be per-instance. Argh.
 + When entering isp_watchdog, set intsok to zero, preserving
   old value to restore later. It's not nice to try and sleep
   from splsoftclock.
 + Fix tick overflow buglet in checking timeout value.

MFC after: 1 week

22 years agoAdd get/set param ioctls.
mjacob [Mon, 8 Jul 2002 17:34:56 +0000 (17:34 +0000)]
Add get/set param ioctls.

MFC after: 1 week

22 years agoAdd override so that we can force set our hard loopdid.
mjacob [Mon, 8 Jul 2002 17:34:32 +0000 (17:34 +0000)]
Add override so that we can force set our hard loopdid.

MFC after: 1 week

22 years agoRemove the 'bogus registrant' hack for fabric searches. It really
mjacob [Mon, 8 Jul 2002 17:33:37 +0000 (17:33 +0000)]
Remove the 'bogus registrant' hack for fabric searches. It really
turns out that there's something of a hole in our new fabric name
server stuff.  We ask the name server for entities that have
registered as a specific type.  That type is FC-SCSI. If the entity
hasn't performed a REGISTER FC4 TYPES, the fabric nameserver won't
return it.

This brings this driver to a bit of a fork in the road as to what
the right thing to do is. For servicing the needs of accessing
FC-SCSI devices, this method is fine, and to be preferred. It is
extremely unlikely we're interested in fabric devices that *don't*
register correctly. If I ever get around to adding an FC-IP stack,
then asking for devices that have registers as FC-IP types is also
the right thing to do.

So- asking the fabric nameserver for a specific type is fine, *as
long as you are only interested in specific types*. If, on the other
hand, you want to create (as for management tool support) a picture
of everything on the fabric, this is *not* so fine. There are a
large class of FC-SCSI *initiators* who *don't* correctly register,
so we never will *see* them.

Is this a problem? Yes, but only a little one. If we want to do such
management tool support, we should probably run a *different* fabric
nameserver query algorithm. Better yet, we should talk to the management
nameserver in Brocade switches instead of the standard FC-GS-2 fabric
nameserver (which can be unwieldy).

Other changes: if we've overrrides marked, don't set some default
values from reading NVRAM. This allows us to override things like
EXEC throttle without having to ignore NVRAM entirely.

MFC after: 1 week

22 years agoMove __offsetof() macro from <machine/ansi.h> to <sys/cdefs.h>. It's
mike [Mon, 8 Jul 2002 16:43:35 +0000 (16:43 +0000)]
Move __offsetof() macro from <machine/ansi.h> to <sys/cdefs.h>.  It's
hardly MD, since all our platforms share the same macro.  It's not
really compiler dependent either, but this helps in reducing
<machine/ansi.h> to only type definitions.

22 years agoBring poll.h up to conformance with POSIX.1-2001 by adding some
mike [Mon, 8 Jul 2002 16:37:35 +0000 (16:37 +0000)]
Bring poll.h up to conformance with POSIX.1-2001 by adding some
visibility conditionals, adding the nfds_t type, and changing the
poll() prototype a little.  Update the manual to match.

22 years agoBring back WFORMAT=0.
jmallett [Mon, 8 Jul 2002 15:48:39 +0000 (15:48 +0000)]
Bring back WFORMAT=0.

22 years agoFixed some printf format errors (4 new ones reported by gcc and 5 nearby
bde [Mon, 8 Jul 2002 12:42:29 +0000 (12:42 +0000)]
Fixed some printf format errors (4 new ones reported by gcc and 5 nearby
old ones not reported by gcc).  This helps unbreak LINT.

22 years agoFixed some printf format errors (one new one reported by gcc and 3 nearby
bde [Mon, 8 Jul 2002 12:21:11 +0000 (12:21 +0000)]
Fixed some printf format errors (one new one reported by gcc and 3 nearby
old ones not reported by gcc).  This helps unbreak LINT.

22 years agoComment out apm; ACPI is the modern replacement, and folks who really
markm [Mon, 8 Jul 2002 11:56:44 +0000 (11:56 +0000)]
Comment out apm; ACPI is the modern replacement, and folks who really
need it can uncomment it. This may buy us some kernel space.

Discussed with: imp & msmith (quite a while ago)

22 years agoReenable the m3 interrupts in m3_pci_resume() and do not
robert [Mon, 8 Jul 2002 11:41:53 +0000 (11:41 +0000)]
Reenable the m3 interrupts in m3_pci_resume() and do not
enable them in m3_pci_attach().

22 years agoBandaid for sparc64. Do not build libstdc++.so for now on sparc64.
peter [Mon, 8 Jul 2002 10:53:35 +0000 (10:53 +0000)]
Bandaid for sparc64.  Do not build libstdc++.so for now on sparc64.
Binaries compiled with it segfault.

22 years agoAlways upload new files, even if the timestamps match. This is a workaround
fenner [Mon, 8 Jul 2002 10:05:26 +0000 (10:05 +0000)]
Always upload new files, even if the timestamps match.  This is a workaround
for the trouble that DES and I had with MFCs: when "cvs update -jfoo -jbar"
creates a new file, it sets the version to 0 ("new") but sets the timestamp
in the Entries file to the timestamp of the file that's new on the branch.
The CVS client doesn't upload files whose timestamps match with the Entries
file, so these new files don't get uploaded to the server and the server
fails when trying to check them in.

PR: bin/40227
Approved by: peter
MFC after: 2 weeks

22 years agoRemove #define APIC_INTR_HIGHPRI_CLOCK - this is no longer used
peter [Mon, 8 Jul 2002 09:14:01 +0000 (09:14 +0000)]
Remove #define APIC_INTR_HIGHPRI_CLOCK - this is no longer used
(see i386/isa/intr_machdep.c 1.72)

22 years agoThe clock is already allocated as 'fast' - no need to try and intercept a
peter [Mon, 8 Jul 2002 09:12:22 +0000 (09:12 +0000)]
The clock is already allocated as 'fast' - no need to try and intercept a
'slow' interrupt registration and convert it into 'fast'.

22 years ago - Use (MAXLOGNAME - 1) where UT_NAMESIZE was used to be able to
robert [Mon, 8 Jul 2002 09:08:51 +0000 (09:08 +0000)]
 - Use (MAXLOGNAME - 1) where UT_NAMESIZE was used to be able to
(-)remove the inclusions of <utmp.h>.

22 years ago - Do not include <utmp.h> for no reason.
robert [Mon, 8 Jul 2002 09:02:02 +0000 (09:02 +0000)]
 - Do not include <utmp.h> for no reason.
 - Remove unneeded utmp path constant.

22 years agoCosmetic. Remove #if 0 definition of vtophys() - it predates 4MB pages.
peter [Mon, 8 Jul 2002 08:14:28 +0000 (08:14 +0000)]
Cosmetic.  Remove #if 0 definition of vtophys() - it predates 4MB pages.
Remove avtophys(), it isn't referenced anywhere.

22 years agoAdd a man page. Style comments welcome. I have a bunch-o-new-pages
julian [Mon, 8 Jul 2002 07:51:20 +0000 (07:51 +0000)]
Add a man page. Style comments welcome. I have a bunch-o-new-pages
to add so I might as we find out what I did wrong now :-)

22 years agoSlight changes to reflect some of the changes in -current.
julian [Mon, 8 Jul 2002 07:34:46 +0000 (07:34 +0000)]
Slight changes to reflect some of the changes in -current.
mi_switch(9) is still wildly innacurate. I suggest that every kernel
developer takes 20 minutes a day for the next few days and updates one or
two of his favourite chapter 9 man pages as they are now WAY out of date
in general. I will add a couple of KSE related pages soon.

22 years agoMake cspace()'s second argument const. Make `escapes' in lputs() const.
tjr [Mon, 8 Jul 2002 06:00:55 +0000 (06:00 +0000)]
Make cspace()'s second argument const. Make `escapes' in lputs() const.

22 years agoSilence WARNS=4 by passing a non-const buffer to regerror(). Since it
tjr [Mon, 8 Jul 2002 05:36:24 +0000 (05:36 +0000)]
Silence WARNS=4 by passing a non-const buffer to regerror(). Since it
doesn't actually use the buffer when errbuf_size == 0, pass NULL.

22 years agoAdd a special page zero entry point intended to be called via the single
peter [Mon, 8 Jul 2002 04:24:26 +0000 (04:24 +0000)]
Add a special page zero entry point intended to be called via the single
threaded VM pagezero kthread outside of Giant.  For some platforms, this
is really easy since it can just use the direct mapped region.  For others,
IPI sending is involved or there are other issues, so grab Giant when
needed.

We still have preemption issues to deal with, but Alan Cox has an
interesting suggestion on how to minimize the problem on x86.

Use Luigi's hack for preserving the (lack of) priority.

Turn the idle zeroing back on since it can now actually do something useful
outside of Giant in many cases.

22 years agoGRRR! rc.* cannot do an "exit 0" or it aborts the entire /etc/rc process.
peter [Mon, 8 Jul 2002 04:03:21 +0000 (04:03 +0000)]
GRRR!  rc.* cannot do an "exit 0" or it aborts the entire /etc/rc process.
We do a '. /etc/rc.syscons' - not run it in a seperate shell.

22 years agoDocument TCBHASHSIZE in NOTES and add it to the allowable kernel options.
mp [Mon, 8 Jul 2002 02:53:59 +0000 (02:53 +0000)]
Document TCBHASHSIZE in NOTES and add it to the allowable kernel options.

PR: 32912
Submitted by: Carl Schmidt <carl@slackerbsd.org>
MFC after: 3 days

22 years ago - VT_PSEUDOFS and VT_PROCFS support locking now
jeff [Mon, 8 Jul 2002 01:53:14 +0000 (01:53 +0000)]
 - VT_PSEUDOFS and VT_PROCFS support locking now
 - Remove VBLK from the list of vtypes that are ignored for locking ops.

22 years agos/procrunnable/kserunnable/ in a comment
peter [Mon, 8 Jul 2002 01:52:55 +0000 (01:52 +0000)]
s/procrunnable/kserunnable/ in a comment

22 years agoLock down pseudofs:
jeff [Mon, 8 Jul 2002 01:50:14 +0000 (01:50 +0000)]
Lock down pseudofs:
 - Initialize lock structure in vncache_alloc
 - Return locked vnodes from vncache_alloc
 - Setup vnode op vectors to use default lock, unlock, and islocked
 - Implement simple locking scheme required for lookup

22 years agoRemove OBE prototype for procrunnable()
peter [Mon, 8 Jul 2002 01:49:49 +0000 (01:49 +0000)]
Remove OBE prototype for procrunnable()

22 years agoUse a fixed MAXBSIZE-size auto array instead of a static pointer
iedowse [Mon, 8 Jul 2002 01:25:54 +0000 (01:25 +0000)]
Use a fixed MAXBSIZE-size auto array instead of a static pointer
to a malloc'd buffer in dmpindir() and dirindir(). These functions
recursively call themselves to handle deeper levels of indirect
blocks, so a single static buffer was not suitable.

Bug tracked down by: Don Lewis <dl-freebsd@catspoiler.org>
Approach suggested by: bde

22 years agoAvoid vm_page_lookup() [grabs a spinlock] and just process the upage
peter [Mon, 8 Jul 2002 01:11:10 +0000 (01:11 +0000)]
Avoid vm_page_lookup() [grabs a spinlock] and just process the upage
object memq instead.

Suggested by: alc

22 years agoReplace the use of %qd with intmax_t/%jd and fix a number of -Wall
iedowse [Mon, 8 Jul 2002 00:29:23 +0000 (00:29 +0000)]
Replace the use of %qd with intmax_t/%jd and fix a number of -Wall
and -Wformat warnings:
 o Include timeconv.h for the time conversion functions.
 o Remove unused variables.
 o Correct a few cases where %d was used when printing longs.

22 years agoFix a hideous TLB bug. pmap_unmapdev neglected to remove the device
peter [Mon, 8 Jul 2002 00:17:43 +0000 (00:17 +0000)]
Fix a hideous TLB bug.  pmap_unmapdev neglected to remove the device
mappings from the page tables, which were mapped with PG_G!  We could
reuse the page table entry for another mapping (pmap_mapdev) but it
would never have cleared any remaining PG_G TLB entries.

22 years agoEnable netboot support by default, since it can now coexist with disk and
jake [Sun, 7 Jul 2002 23:08:22 +0000 (23:08 +0000)]
Enable netboot support by default, since it can now coexist with disk and
cdrom support.  This avoids having to distribute separate loaders.

22 years agoCollect all the (now equivalent) pmap_new_proc/pmap_dispose_proc/
peter [Sun, 7 Jul 2002 23:05:27 +0000 (23:05 +0000)]
Collect all the (now equivalent) pmap_new_proc/pmap_dispose_proc/
pmap_swapin_proc/pmap_swapout_proc functions from the MD pmap code
and use a single equivalent MI version.  There are other cleanups
needed still.

While here, use the UMA zone hooks to keep a cache of preinitialized
proc structures handy, just like the thread system does.  This eliminates
one dependency on 'struct proc' being persistent even after being freed.
There are some comments about things that can be factored out into
ctor/dtor functions if it is worth it.  For now they are mostly just
doing statistics to get a feel of how it is working.

22 years agoAdd a hack (kludge?) to avoid trying to access files backed by disk
jake [Sun, 7 Jul 2002 23:01:36 +0000 (23:01 +0000)]
Add a hack (kludge?) to avoid trying to access files backed by disk
devices as though they were backed by network devices.

22 years agoUpdate for post-kse3 pmap kthread allocation changes
peter [Sun, 7 Jul 2002 22:56:31 +0000 (22:56 +0000)]
Update for post-kse3 pmap kthread allocation changes

22 years agoFix (s/proc/thread/) some typos in two panic messages.
peter [Sun, 7 Jul 2002 22:50:41 +0000 (22:50 +0000)]
Fix (s/proc/thread/) some typos in two panic messages.

22 years agoRemove 0 initializers for global/static variables, so they end up in
luigi [Sun, 7 Jul 2002 22:42:57 +0000 (22:42 +0000)]
Remove 0 initializers for global/static variables, so they end up in
BSS instead of DATA. This marginally reduces the kernel image size, though
the difference is almost irrelevant for compressed kernels.

22 years ago - Require locks for getattr. At some point this could only require shared
jeff [Sun, 7 Jul 2002 22:37:45 +0000 (22:37 +0000)]
 - Require locks for getattr.  At some point this could only require shared
   locks.