]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
24 years agoimplement OSF_GET_MAX_UPROCS portion of osf1_getsysinfo() for AT&T ksh
Andrew Gallatin [Thu, 6 Jan 2000 02:37:45 +0000 (02:37 +0000)]
implement OSF_GET_MAX_UPROCS portion of osf1_getsysinfo() for AT&T ksh

24 years agoShutup complaints about unimplemented SVR4 ioctls
Andrew Gallatin [Thu, 6 Jan 2000 02:35:44 +0000 (02:35 +0000)]
Shutup complaints about unimplemented SVR4 ioctls

24 years agoMake example for handling "-##" work and comply with style(9). Still
Tim Vanderhoek [Thu, 6 Jan 2000 01:25:15 +0000 (01:25 +0000)]
Make example for handling "-##" work and comply with style(9).  Still
doesn't handle nastier corner cases such as "-j3 -33" correctly.  <shrug>

PR: docs/12994 (James Howard <howardjp@wam.umd.edu>)

24 years agomake_devices(): Create devices with permissions 640, not 740.
Greg Lehey [Wed, 5 Jan 2000 23:56:33 +0000 (23:56 +0000)]
make_devices(): Create devices with permissions 640, not 740.

Pointed-out-by: "Rodney W. Grimes" <freebsd@gndrsh.dnsmgr.net>

24 years agoUpdate copyright date for Y2K (did the other files but forgot this one).
Bill Paul [Wed, 5 Jan 2000 23:38:07 +0000 (23:38 +0000)]
Update copyright date for Y2K (did the other files but forgot this one).
Also update comments to credit to Rob Furr with donating the adapter
that I used for testing. No code changes.

24 years agoDon't tsleep with PCATCH, it's a wonderful way to hang the system.
Greg Lehey [Wed, 5 Jan 2000 22:59:36 +0000 (22:59 +0000)]
Don't tsleep with PCATCH, it's a wonderful way to hang the system.

Reported-by: Bernd Walter <ticso@cicely.de>

24 years agoAdd missing header file.
Greg Lehey [Wed, 5 Jan 2000 22:58:39 +0000 (22:58 +0000)]
Add missing header file.

Not-reported-by: peter

24 years agoAdd documentation for SVR4 options in LINT.
Mark Newton [Wed, 5 Jan 2000 21:41:16 +0000 (21:41 +0000)]
Add documentation for SVR4 options in LINT.

24 years agoUnbreak LINT -- typedefs from signal.h are needed to make prototypes in
Mark Newton [Wed, 5 Jan 2000 21:28:46 +0000 (21:28 +0000)]
Unbreak LINT -- typedefs from signal.h are needed to make prototypes in
this file make sense.

24 years agodon't panic if channel init fails, report and fail gracefully
Cameron Grant [Wed, 5 Jan 2000 20:44:41 +0000 (20:44 +0000)]
don't panic if channel init fails, report and fail gracefully

24 years agoFix a few obscure memory leaks.
Archie Cobbs [Wed, 5 Jan 2000 20:36:07 +0000 (20:36 +0000)]
Fix a few obscure memory leaks.

24 years agoCast myminor/mymajor results to u_int so
Matt Jacob [Wed, 5 Jan 2000 20:18:16 +0000 (20:18 +0000)]
Cast myminor/mymajor results to u_int so
comparisons to longs work correctly on
alpha.

24 years agoExport the nselcoll counter via the kern.nselcoll sysctl so we can see
Peter Wemm [Wed, 5 Jan 2000 19:40:17 +0000 (19:40 +0000)]
Export the nselcoll counter via the kern.nselcoll sysctl so we can see
just how bad it gets in various situations.

Reminded by:  adrian

24 years agoSupport filesystems with the not-so-new "filetype" feature. This
Bruce Evans [Wed, 5 Jan 2000 19:31:26 +0000 (19:31 +0000)]
Support filesystems with the not-so-new "filetype" feature.  This
feature gives the d_type field for struct dirent.  We used to panic
in ext2_readdir() for filesystems with this feature.

24 years agoActually remove unrequired SRCS.
Bruce Evans [Wed, 5 Jan 2000 19:22:10 +0000 (19:22 +0000)]
Actually remove unrequired SRCS.

24 years agoFixed the type of dllockinit() (const unpoisoning).
Bruce Evans [Wed, 5 Jan 2000 19:04:55 +0000 (19:04 +0000)]
Fixed the type of dllockinit() (const unpoisoning).

Use long lines instead of lines split with backslash-newline in synopsis.
My synopsis checker doesn't understand backslash-newline.

24 years agoFixed missing include in synopsis.
Bruce Evans [Wed, 5 Jan 2000 19:02:14 +0000 (19:02 +0000)]
Fixed missing include in synopsis.

Use long lines instead of lines split with backslash-newline in synopsis.
My synopsis checker doesn't understand backslash-newline.

24 years agoFixed missing include in synopsis.
Bruce Evans [Wed, 5 Jan 2000 18:54:37 +0000 (18:54 +0000)]
Fixed missing include in synopsis.

24 years agoCorrect the vendor ID for KLSI: it's actually 0x5e9. 0x3e8 is the
Bill Paul [Wed, 5 Jan 2000 18:25:53 +0000 (18:25 +0000)]
Correct the vendor ID for KLSI: it's actually 0x5e9. 0x3e8 is the
vendor ID for Aox Inc, which makes the controller engine and/or
firmware for the KLSI chip.

24 years agoMake the evil broken pnpbios compensation slightly less evil.
Peter Wemm [Wed, 5 Jan 2000 17:42:43 +0000 (17:42 +0000)]
Make the evil broken pnpbios compensation slightly less evil.
This is the hack that compensates for when bios vendors "forget" to
include the fdc control (0x3f7) port in their io port mappings.  Instead
of accessing ports outside of a range allocated to a handle, simply
allocate the port directly.  It even shows up in the probe..

24 years agoRefreshed the silly copy of <sys/types.h>. It was 3 years out of date.
Bruce Evans [Wed, 5 Jan 2000 17:40:25 +0000 (17:40 +0000)]
Refreshed the silly copy of <sys/types.h>.  It was 3 years out of date.

24 years agoRefreshed the silly copy of <sys/disklabel.h>. The old copy was
Bruce Evans [Wed, 5 Jan 2000 17:34:00 +0000 (17:34 +0000)]
Refreshed the silly copy of <sys/disklabel.h>.  The old copy was
seriously out of date.  It older than Lite1 (it has an undeprecated
DISKTAB but not _PATH_DISKTAB).

24 years agoDoh. Forgot to add the entry for the 3Com 3c19250 to the supported
Bill Paul [Wed, 5 Jan 2000 17:31:05 +0000 (17:31 +0000)]
Doh. Forgot to add the entry for the 3Com 3c19250 to the supported
devices table.

24 years agoDo not attempt to load the firmware a second time. If you reboot your
Bill Paul [Wed, 5 Jan 2000 17:13:50 +0000 (17:13 +0000)]
Do not attempt to load the firmware a second time. If you reboot your
machine but leave your KLSI adapter plugged into your USB port, it
may stay powered on and retain its firmware in memory. Trying to load
the firmware again in this case will wedge the chip. Try to detect this
in the kue_load_fw() routine and bail if the firmware is already
loaded and running.

Also, in the probe/match routine, force the revision code to the
hardware default and force a rescan of the quirk database. This is
necessary because the adapter will return a different revision code
if the firmware has been loaded. Without the firmware, the revision
code is 0x002. With the firmware, the revision code is 0x202. This
confuses the quirk mechanism, which won't match a quirk to a device
unless the revision code agrees with the quirk table entry.

This makes probe/attach of these devices somewhat more reliable.

Also add a few comments about the device's operation.

24 years agoRefreshed the silly copy of <sys/dirent.h>. _KERNEL was still spelled
Bruce Evans [Wed, 5 Jan 2000 17:04:36 +0000 (17:04 +0000)]
Refreshed the silly copy of <sys/dirent.h>.  _KERNEL was still spelled
KERNEL, and the changes to remove the <sys/types.h> prerequisite had
not reached here.

Removed unused/wrong include of <sys/types.h> in synopsis.

24 years agoRefreshed the silly literal copy of <sys/mtio.h>. _KERNEL was still
Bruce Evans [Wed, 5 Jan 2000 16:52:34 +0000 (16:52 +0000)]
Refreshed the silly literal copy of <sys/mtio.h>.  _KERNEL was still
spelled KERNEL...

24 years agoZap pci_map_dense() and pci_map_bwx() - they were for compatability but
Peter Wemm [Wed, 5 Jan 2000 16:40:10 +0000 (16:40 +0000)]
Zap pci_map_dense() and pci_map_bwx() - they were for compatability but
are not used.  All the drivers that use memory mapped IO on the Alpha have
been ported already.

24 years agoFixed missing includes in synopsis. <sys/file.h> went missing when KERNEL
Bruce Evans [Wed, 5 Jan 2000 16:38:58 +0000 (16:38 +0000)]
Fixed missing includes in synopsis.  <sys/file.h> went missing when KERNEL
was not updated to _KERNEL.  Actually including <sys/file.h> as specified
never actually worked, since a prerequisite was missing.

24 years agoPatch up some of the evilness left over from the early newbus porting.
Peter Wemm [Wed, 5 Jan 2000 16:31:27 +0000 (16:31 +0000)]
Patch up some of the evilness left over from the early newbus porting.
In particular:
 - Don't leave resources allocated in the probe routine.  Allocate them
   during probe and release them.  Probe's job is to identify devices only.
 - Don't abuse the ivars pointer.. (!).  Create real ivars and use the
   proper access system.  (the bus_read_ivar method)
 - Don't add the children until attach() has successfully grabbed the
   hardware, otherwise there are potential leaks if attach fails.

24 years agoFixed missing include in synopsis.
Bruce Evans [Wed, 5 Jan 2000 16:31:12 +0000 (16:31 +0000)]
Fixed missing include in synopsis.

Removed superfluous quoting of function name in .Fo macro.  My synopsis
checker doesn't understand it.

24 years agoKERNEL -> _KERNEL
Eivind Eklund [Wed, 5 Jan 2000 16:25:20 +0000 (16:25 +0000)]
KERNEL -> _KERNEL

24 years agomdoc(7)'fy
Alexey Zelkin [Wed, 5 Jan 2000 15:21:01 +0000 (15:21 +0000)]
mdoc(7)'fy
sort XREFs

24 years agomdoc(7)'fy
Alexey Zelkin [Wed, 5 Jan 2000 15:16:32 +0000 (15:16 +0000)]
mdoc(7)'fy

24 years agoDon't build with X support if DESTDIR is defined. This prevents
Marcel Moolenaar [Wed, 5 Jan 2000 12:59:31 +0000 (12:59 +0000)]
Don't build with X support if DESTDIR is defined. This prevents
breaking a cross-build caused by taking the X libraries on the
build machine. In general this means that we never compile with
X support. The user has to manually compile doscmd for that.

Suggested by: bde, imp (among others)

24 years agoSynced with sys/i386/isa/clock.c rev 1.149.
KATO Takenori [Wed, 5 Jan 2000 12:35:03 +0000 (12:35 +0000)]
Synced with sys/i386/isa/clock.c rev 1.149.

24 years agoSynced with sys/i386/i386/userconfig.c rev 1.166.
KATO Takenori [Wed, 5 Jan 2000 12:34:10 +0000 (12:34 +0000)]
Synced with sys/i386/i386/userconfig.c rev 1.166.

24 years agoSynced with sys/i386/conf/GENERIC rev 1.222.
KATO Takenori [Wed, 5 Jan 2000 12:33:10 +0000 (12:33 +0000)]
Synced with sys/i386/conf/GENERIC rev 1.222.

24 years agoMissed a filename variable quote in rc
Peter Wemm [Wed, 5 Jan 2000 09:19:27 +0000 (09:19 +0000)]
Missed a filename variable quote in rc

24 years agoQuote arguments in vi recover script.
Peter Wemm [Wed, 5 Jan 2000 08:17:18 +0000 (08:17 +0000)]
Quote arguments in vi recover script.

PR: 15901
Submitted by: KOJIMA Hajime <kjm@rins.ryukoku.ac.jp>

24 years agoExplicitly type large scalar parameters to avoid compilation warnings
Greg Lehey [Wed, 5 Jan 2000 06:13:01 +0000 (06:13 +0000)]
Explicitly type large scalar parameters to avoid compilation warnings
on alpha.

Submitted-by:       Bernd Walter <ticso@cicely.de>

struct sd: Add a field for the pid of the reviver when the subdisk is
reviving.

Replace block device macros with generalized device macros.

24 years agoWhen attaching a plex to a volume with 'up' subdisks, make the plexes
Greg Lehey [Wed, 5 Jan 2000 06:12:25 +0000 (06:12 +0000)]
When attaching a plex to a volume with 'up' subdisks, make the plexes
subdisks 'stale', not 'reviving'.

24 years agoCorrect printf format for pointers to avoid compilation warnings on
Greg Lehey [Wed, 5 Jan 2000 06:11:46 +0000 (06:11 +0000)]
Correct printf format for pointers to avoid compilation warnings on
alpha.

Explicitly type large scalar parameters to avoid compilation warnings
on alpha.

Submitted-by:       Bernd Walter <ticso@cicely.de>

Make better checks that the revive block size is valid, silently set
it to the defaults if not.

Replace block device macros with generalized device macros.

24 years agoCorrect printf format for pointers to avoid compilation warnings on
Greg Lehey [Wed, 5 Jan 2000 06:10:52 +0000 (06:10 +0000)]
Correct printf format for pointers to avoid compilation warnings on
alpha.

Modify the manner in which we lock RAID-5 plexes.  This appears to
solve some of the elusive panics we have seen with corrupted buffer
headers (specifically the zeroed-out b_iodone field).

Submitted-by:       Bernd Walter <ticso@cicely.de>

24 years agoModify the manner in which we lock RAID-5 plexes. This appears to
Greg Lehey [Wed, 5 Jan 2000 06:09:43 +0000 (06:09 +0000)]
Modify the manner in which we lock RAID-5 plexes.  This appears to
solve some of the elusive panics we have seen with corrupted buffer
headers (specifically the zeroed-out b_iodone field).

Submitted-by:       Bernd Walter <ticso@cicely.de>

24 years agoChange the name of the define 'SPACETAB' (originally 'isspace') to
Greg Lehey [Wed, 5 Jan 2000 06:08:55 +0000 (06:08 +0000)]
Change the name of the define 'SPACETAB' (originally 'isspace') to
'iswhite'.  The original change was required because of name
conflicts.

Add key pairs for the keywords 'mv' and 'move' (part of the move
command).

Add comments.

24 years agoInclude basename() function even if the system isn't i386.
Greg Lehey [Wed, 5 Jan 2000 06:08:23 +0000 (06:08 +0000)]
Include basename() function even if the system isn't i386.

Correct printf format for pointers to avoid compilation warnings on
alpha.

Submitted-by:       Bernd Walter <ticso@cicely.de>

24 years agoAdd the keywords 'mv' and 'move'.
Greg Lehey [Wed, 5 Jan 2000 06:07:59 +0000 (06:07 +0000)]
Add the keywords 'mv' and 'move'.

24 years agoAdd function moveobject, which currently moves subdisks to different
Greg Lehey [Wed, 5 Jan 2000 06:07:26 +0000 (06:07 +0000)]
Add function moveobject, which currently moves subdisks to different
drives.  This function just does the low-level configuration changes;
the resultant subdisk is stale if it previously had any contents,
otherwise it is empty (i.e. in need of initializing if it's RAID-5).
We still need to handle getting the contents moved over, but the
current version will suffice to migrate subdisks from a disk which has
failed.

Submitted-by: Marius Bendiksen <marius@marius.scancall.no>
24 years agoAdd VINUM_MOVE ioctl.
Greg Lehey [Wed, 5 Jan 2000 06:06:01 +0000 (06:06 +0000)]
Add VINUM_MOVE ioctl.

Fix transfer lengths for some ioctls.

Rearrange file: put structs first, then ioctl requests.

24 years agoExplicitly type large scalar parameters to avoid compilation warnings
Greg Lehey [Wed, 5 Jan 2000 06:05:33 +0000 (06:05 +0000)]
Explicitly type large scalar parameters to avoid compilation warnings
on alpha.

Submitted-by:       Bernd Walter <ticso@cicely.de>

Remove #include of vm/vm_zone.h.

Submitted-by: Someone, I'm sure, but I seem to have lost the
attribution.  Sorry.

Get the check for disk devices correct, and return an appropriate
message if the check fails.

24 years agoAdd declaration for give_sd_to_drive, needed for the 'move' command.
Greg Lehey [Wed, 5 Jan 2000 06:04:17 +0000 (06:04 +0000)]
Add declaration for give_sd_to_drive, needed for the 'move' command.

24 years agoSet P_SYSTEM in the daemon proc structure to alleviate delays on
Greg Lehey [Wed, 5 Jan 2000 06:03:56 +0000 (06:03 +0000)]
Set P_SYSTEM in the daemon proc structure to alleviate delays on
shutdown.

Submitted-by: Alfred Perlstein <bright@wintelcom.net>

Correct printf format for pointers to avoid compilation warnings on
alpha.

Submitted-by: Bernd Walter <ticso@cicely.de>

Identify daemon as 'vinum', not 'vinumd', in messages.  This
corresponds to the name in ps.

24 years agoExplicitly type large scalar parameters to avoid compilation warnings
Greg Lehey [Wed, 5 Jan 2000 06:02:57 +0000 (06:02 +0000)]
Explicitly type large scalar parameters to avoid compilation warnings
on alpha.

Submitted-by:       Bernd Walter <ticso@cicely.de>

Get parameters right for some error messages returned via
throw_rude_remark().

Fix typo in comment.

Remove the 'static' attribute from give_sd_to_drive.  This is needed
for the implementation of moveobject() in vinumioctl.c.

24 years agoDescribe 'move' function.
Greg Lehey [Wed, 5 Jan 2000 05:59:08 +0000 (05:59 +0000)]
Describe 'move' function.

Submitted-by: Marius Bendiksen <marius@marius.scancall.no>

Describe state transitions in more detail, in particular how to revive
subdisks.

24 years agoChange parameters of some declarations.
Greg Lehey [Wed, 5 Jan 2000 05:58:01 +0000 (05:58 +0000)]
Change parameters of some declarations.

Add declaration for vinum_mv.

24 years agomakedev: Give all devices permissions 740.
Greg Lehey [Wed, 5 Jan 2000 05:57:31 +0000 (05:57 +0000)]
makedev: Give all devices permissions 740.

Suggested-by: Bernd Walter (ticso@cicely.de)

Add key pairs for 'mv' and 'move' (a synonym for 'mv').  Required for
the move command code submitted by Marius Bendiksen
<marius@marius.scancall.no>

make_devices: Don't create symlinks for drives if they are only
referenced.  Previously, spurious symlinks appeared in the current
directory.

Problem-reported-by: Bernd Walter (ticso@cicely.de)

No longer create character devices, now that there is no difference.
Make the devices as character devices, not block devices.

24 years agoExplicitly type large scalar parameters to avoid compilation warnings
Greg Lehey [Wed, 5 Jan 2000 05:56:47 +0000 (05:56 +0000)]
Explicitly type large scalar parameters to avoid compilation warnings
on alpha.

Submitted-by: Bernd Walter <ticso@cicely.de>

Replace %q formats with %lld.

Desired-by: bde

Remove #ifdef RAID5

vinum_lsi: If a subdisk is in 'reviving' state, check the status of
the revive and report the status of the revive.  In verbose mode,
display the pid of the reviver.

24 years agoExplicitly type large scalar parameters to avoid compilation warnings
Greg Lehey [Wed, 5 Jan 2000 05:55:57 +0000 (05:55 +0000)]
Explicitly type large scalar parameters to avoid compilation warnings
on alpha.

Submitted-by: Bernd Walter <ticso@cicely.de>

Add function vinum_mv, which moves subdisks to different drives.  This
function just does the low-level configuration changes; the resultant
subdisk is stale if it previously had any contents, otherwise it is
empty (i.e. in need of initializing if it's RAID-5).  We still need to
handle getting the contents moved over, but the current version will
suffice to migrate subdisks from a disk which has failed.

Submitted-by: Marius Bendiksen <marius@marius.scancall.no>

vinum_start:  Get the revive block size right.

24 years ago Enhance reassignbuf(). When a buffer cannot be time-optimally inserted
Matthew Dillon [Wed, 5 Jan 2000 05:11:37 +0000 (05:11 +0000)]
Enhance reassignbuf().  When a buffer cannot be time-optimally inserted
    into vnode dirtyblkhd we append it to the list instead of prepend it to
    the list in order to maintain a 'forward' locality of reference, which
    is arguably better then 'reverse'.  The original algorithm did things this
    way to but at a huge time cost.

    Enhance the append interlock for NFS writes to handle intr/soft mounts
    better.

    Fix the hysteresis for NFS async daemon I/O requests to reduce the
    number of unnecessary context switches.

    Modify handling of NFS mount options.  Any given user option that is
    too high now defaults to the kernel maximum for that option rather then
    the kernel default for that option.

Reviewed by:  Alfred Perlstein <bright@wintelcom.net>

24 years agoMan pages for the VFS extended attribute and access control list vnops.
Robert Watson [Wed, 5 Jan 2000 04:59:02 +0000 (04:59 +0000)]
Man pages for the VFS extended attribute and access control list vnops.

Reviewed by: eivind

24 years agoAdd device driver support for USB ethernet adapters based on the
Bill Paul [Wed, 5 Jan 2000 04:27:24 +0000 (04:27 +0000)]
Add device driver support for USB ethernet adapters based on the
Kawasaki LSI KL5KUSB101B chip, including the LinkSys USB10T, the
Entrega NET-USB-E45, the Peracom USB Ethernet Adapter, the 3Com
3c19250 and the ADS Technologies USB-10BT. This device is 10mbs
half-duplex only, so there's miibus or ifmedia support. This device
also requires firmware to be loaded into it, however KLSI allows
redistribution of the firmware images (I specifically asked about
this; they said it was ok).

Special thanks to Annelise Anderson for getting me in touch with
KLSI (eventually) and thanks to KLSI for providing the necessary
programming info.

Highlights:
- Add driver files to /sys/dev/usb
- update usbdevs and regenerate attendate files
- update usb_quirks.c
- Update HARDWARE.TXT and RELNOTES.TXT for i386 and alpha
- Update LINT, GENERIC and others for i386, alpha and pc98
- Add man page
- Add module
- Update sysinstall and userconfig.c

24 years agoargh, forgot the bus_dma_tag_creates.
Cameron Grant [Wed, 5 Jan 2000 04:01:34 +0000 (04:01 +0000)]
argh, forgot the bus_dma_tag_creates.
also, panic if channel init fails instead of derefing null.

24 years agoAdd protocol to ``nat port'' examples - it's mandatory.
Brian Somers [Wed, 5 Jan 2000 02:59:28 +0000 (02:59 +0000)]
Add protocol to ``nat port'' examples - it's mandatory.

Suggested by: Mark Knight <markk@knigma.org>

24 years agoset up the pci regs properly for busmastering. this makes the card work on
Cameron Grant [Wed, 5 Jan 2000 02:03:55 +0000 (02:03 +0000)]
set up the pci regs properly for busmastering.  this makes the card work on
my smp box.

24 years agoRemove the unused 'func' arguments to the deregistration functions.
Mike Smith [Wed, 5 Jan 2000 01:09:34 +0000 (01:09 +0000)]
Remove the unused 'func' arguments to the deregistration functions.

Submitted by:  Darrell Anderson <anderson@cs.duke.edu>

24 years agoallocate isa bounce buffers of the right size for ess/mss cards, fixes
Cameron Grant [Wed, 5 Jan 2000 00:39:20 +0000 (00:39 +0000)]
allocate isa bounce buffers of the right size for ess/mss cards, fixes
panics reported

24 years ago Fix at least one source of the continued 'NFS append race'. close()
Matthew Dillon [Wed, 5 Jan 2000 00:32:18 +0000 (00:32 +0000)]
Fix at least one source of the continued 'NFS append race'.  close()
    was calling nfs_flush() and then clearing the NMODIFIED bit.  This is
    not legal since there might still be dirty buffers after the nfs_flush
    (for example, pending commits).  The clearing of this bit in turn prevented
    a necessary vinvalbuf() from occuring leaving left over dirty buffers
    even after truncating the file in a new operation.  The fix is to
    simply not clear NMODIFIED.

    Also added a sysctl vfs.nfs.nfsv3_commit_on_close which, if set to 1,
    will cause close() to do a stage 1 write AND a stage 2 commit
    synchronously.  By default only the stage 1 write is done synchronously.

Reviewed by: Alfred Perlstein <bright@wintelcom.net>

24 years agoRemove non-functional 'all:' target.
Greg Lehey [Tue, 4 Jan 2000 23:46:41 +0000 (23:46 +0000)]
Remove non-functional 'all:' target.
Remove unused 'state.h' and 'maketabs' targets.
Fix white space style bugs.

Submitted-by: bde

Sort module names in SRCS.

Suggested-by: bde

Correct breakage committed in revision 1.16.

24 years ago Update some of the network driver documentation in the LINT file, which
Matthew Dillon [Tue, 4 Jan 2000 23:09:42 +0000 (23:09 +0000)]
Update some of the network driver documentation in the LINT file, which
    is where most people look to match drivers up with cards.

Reviewed by: wpaul

24 years agoISA device drivers use the ISA source interrupt number in locations where
Tor Egge [Tue, 4 Jan 2000 22:24:59 +0000 (22:24 +0000)]
ISA device drivers use the ISA source interrupt number in locations where
the low level interrupt handler number should be used.  Change
setup_apic_irq_mapping() to allocate low level interrupt handler X (Xintr${X})
for any ISA interrupt X mentioned in the MP table.

Remove an assumption in the driver for the system clock (clock.c) that
interrupts mentioned in the MP table as delivered to IOAPIC #0 intpin Y
is handled by low level interrupt handler Y (Xintr${Y}) but don't assume
that low level interrupt handler 0 (Xintr0) is used.

Don't allocate two low level interrupt handlers for the system clock.
Reviewed by: NOKUBI Hirotaka <hnokubi@yyy.or.jp>

24 years agoFix race condition caused by missing splnet()'s.
Archie Cobbs [Tue, 4 Jan 2000 22:06:08 +0000 (22:06 +0000)]
Fix race condition caused by missing splnet()'s.

24 years agoAdd btxld to the list of cross-tools on machines that don't have it
Marcel Moolenaar [Tue, 4 Jan 2000 14:12:12 +0000 (14:12 +0000)]
Add btxld to the list of cross-tools on machines that don't have it
natively (ie non-i386 architectures).

24 years agoo Allow btxld to be compiled on 64-bit machines
Marcel Moolenaar [Tue, 4 Jan 2000 14:10:36 +0000 (14:10 +0000)]
o  Allow btxld to be compiled on 64-bit machines
o  s/unsigned/unsigned int/g
o  Add -Wall

btxld can now be built as a cross-tool for cross-building i386/pc98 on
platforms that don't have btxld (such as alpha).

24 years agoThe dumpon sysctl only accept cdevs thse days.
Poul-Henning Kamp [Tue, 4 Jan 2000 12:52:52 +0000 (12:52 +0000)]
The dumpon sysctl only accept cdevs thse days.

Submitted by: bde

24 years agoBe more careful about NOUDEV and NODEV.
Poul-Henning Kamp [Tue, 4 Jan 2000 12:51:50 +0000 (12:51 +0000)]
Be more careful about NOUDEV and NODEV.

Submitted by: bde

24 years agoCreate a separate pps_offset variable to use for applying the
Poul-Henning Kamp [Tue, 4 Jan 2000 12:04:39 +0000 (12:04 +0000)]
Create a separate pps_offset variable to use for applying the
hardpps() produced offset component.  This is tested and behaved
stable with frequency offsets from -338.05 to +499.91 PPM.

Interestingly the machine I tested this on would fail if the clock
were slower than 14.3132 MHz whereas it was perfectly happy to run
at 16.384 MHz, in other words [-340PPM ... +14.4%]

Make pps_shift tweakable with sysctl.

24 years agoPer BDE- make this more like the i386 code in that, as per the comment,
Matt Jacob [Tue, 4 Jan 2000 11:30:01 +0000 (11:30 +0000)]
Per BDE- make this more like the i386 code in that, as per the comment,
the simple calculation is good enough.
Submitted by: bde@freebsd.org

24 years agoadd wx0 driver
Matt Jacob [Tue, 4 Jan 2000 11:17:35 +0000 (11:17 +0000)]
add wx0 driver

24 years agoAdd first pass of the Intel Gigabit Ethernet (wiseman) driver. This
Matt Jacob [Tue, 4 Jan 2000 11:12:42 +0000 (11:12 +0000)]
Add first pass of the Intel Gigabit Ethernet (wiseman) driver. This
driver seems relatively functional, but could use some souping up,
particularly in the performance area. This has both NetBSD and FreeBSD
attachment code and a fair amount of effort has been put into making
it easy to port to different *BSD platforms.

The basic design is a one tfd per mbuf transmit (with no transmit
related interrupts- tfds are gc'd as needed). The receive ring
uses a 2K buffer per rfd with a +2 byte adjust for the ethernet
header (so the payload is aligned). There's support that *almost*
works for doing large packets- the rfd chaining code works, but there's
some problem with getting good checksums at the IP reassembly level
(ditto for doing short tfd's too).

The chip has support for TCP checksums insertion for transmit and
TCP checksum calculation on receive (for both you have to do some
appropriate backoff && twiddling), but this isn't in place.

This is nearly entirely reverse engineered from the released Intel
driver, so there's a lot of "We have to do this but do not know why"
stuff. There is somebody who has the chip specs who works in FreeBSD
but they're being a bit standoffish about even sharing hints which
is somewhat annoying. It's also apparent that all I had to work with
were the first rev boards.

This driver has been lightly tested on intel && alpha, but only
point-to-point. There may be some issues with switches- use of
boot time environment variables that override EEPROM settings
(e.g., 'set wx_ilos=1' which inverts the sense of optical signal
loss) may help with this.

I had this out for review for three weeks, and nobody said anything
negative or positive, ergo, this checkin has no 'reviewed by' field
which I would have preferred.

24 years agoadd wx (Intel Wiseman Gig Ethernet) driver
Matt Jacob [Tue, 4 Jan 2000 11:01:36 +0000 (11:01 +0000)]
add wx (Intel Wiseman Gig Ethernet) driver

24 years agoDecompression is supported via -Z, not -z.
Ruslan Ermilov [Tue, 4 Jan 2000 10:34:27 +0000 (10:34 +0000)]
Decompression is supported via -Z, not -z.

24 years ago* Make --null option work in (HAVE_LIBZ > 0) case.
Ruslan Ermilov [Tue, 4 Jan 2000 10:32:55 +0000 (10:32 +0000)]
* Make --null option work in (HAVE_LIBZ > 0) case.
* Remove "why we need this decl..." comment.  The `matcher' variable
  is defined in *grepmat.c files in the original distribution, which
  we did not import.

24 years agoReplace calendar with 2000 calendar.
Greg Lehey [Tue, 4 Jan 2000 06:18:13 +0000 (06:18 +0000)]
Replace calendar with 2000 calendar.

Supplied-by: Josef Grosch <jgrosch@MooseRiver.com>
PR: docs/15429

24 years agoAdd a whole bunch of example entries for pam. This should help get
Mark Murray [Tue, 4 Jan 2000 06:08:58 +0000 (06:08 +0000)]
Add a whole bunch of example entries for pam. This should help get
folk over various hurdles.

24 years agooop, use PRINTF not printf in MI code
Matt Jacob [Tue, 4 Jan 2000 05:17:48 +0000 (05:17 +0000)]
oop, use PRINTF not printf in MI code

24 years agoDeal with package filenames which contain spaces.
Jordan K. Hubbard [Tue, 4 Jan 2000 05:08:58 +0000 (05:08 +0000)]
Deal with package filenames which contain spaces.

Submitted by: Ming-I Hsieh <mihs@wm28.csie.ncu.edu.tw>
PR: 15667

24 years agoRemove references to now-obsolete XFree86 source collection (use the
Jordan K. Hubbard [Tue, 4 Jan 2000 05:06:22 +0000 (05:06 +0000)]
Remove references to now-obsolete XFree86 source collection (use the
port, it's far better).

24 years agoAdd new code for HTTP proxy support.
Jordan K. Hubbard [Tue, 4 Jan 2000 04:50:59 +0000 (04:50 +0000)]
Add new code for HTTP proxy support.

Submitted by: Philipp Mergenthaler <un1i@rz.uni-karlsruhe.de>
PR: 11316

24 years agoAdd support for FTP installation via HTTP proxies.
Jordan K. Hubbard [Tue, 4 Jan 2000 04:50:07 +0000 (04:50 +0000)]
Add support for FTP installation via HTTP proxies.

Submitted by: Philipp Mergenthaler <un1i@rz.uni-karlsruhe.de>
PR: 11316

24 years ago- Fixed warnings.
Yoshihiro Takahashi [Tue, 4 Jan 2000 04:46:50 +0000 (04:46 +0000)]
- Fixed warnings.
- Removed unnecessary include files.

24 years agoIgnore SIGPIPE by default.
Jordan K. Hubbard [Tue, 4 Jan 2000 04:31:29 +0000 (04:31 +0000)]
Ignore SIGPIPE by default.

Submitted by: Philipp Mergenthaler <un1i@rz.uni-karlsruhe.de>
PR: 13900

24 years agoUpdate the graph to include FreeBSD 3.4 and OpenBSD 2.6
Bill Fumerola [Tue, 4 Jan 2000 04:31:27 +0000 (04:31 +0000)]
Update the graph to include FreeBSD 3.4 and OpenBSD 2.6
Update the play-by-play history to include FreeBSD 3.[234] and OpenBSD 2.[56]

Dates obtained from: www.FreeBSD.org and www.OpenBSD.org

24 years agoTurn on a new /usr/bin/cpp that is a true binary rather than a shell script
David E. O'Brien [Tue, 4 Jan 2000 03:48:08 +0000 (03:48 +0000)]
Turn on a new /usr/bin/cpp that is a true binary rather than a shell script
wrapper.  /usr/bin/cpp knows about all the GCC predefined symbols and has
the functionality of the previous EGCS 1.1.2 /usr/libexec/cpp.

24 years agoClean up some debug printing. Find the correct lun when SCCLUN is
Matt Jacob [Tue, 4 Jan 2000 03:45:49 +0000 (03:45 +0000)]
Clean up some debug printing. Find the correct lun when SCCLUN is
defined. If we complete with a check condition but no sense data,
say we had an AUTOSENSE failure.

24 years agoMake Fibre Channel cards correctly note the presence/absence
Matt Jacob [Tue, 4 Jan 2000 03:44:21 +0000 (03:44 +0000)]
Make Fibre Channel cards correctly note the presence/absence
of ARQ data and punt the dealing with its presence/absence
to the platform layers.

24 years agoset default target mode debug to 0
Matt Jacob [Tue, 4 Jan 2000 03:43:03 +0000 (03:43 +0000)]
set default target mode debug to 0

24 years agoadd clarifying tag define for FC
Matt Jacob [Tue, 4 Jan 2000 03:42:19 +0000 (03:42 +0000)]
add clarifying tag define for FC

24 years agoAdd in an isp_tdebug environment variable. Clean up some debugging
Matt Jacob [Tue, 4 Jan 2000 03:41:40 +0000 (03:41 +0000)]
Add in an isp_tdebug environment variable. Clean up some debugging
printouts for clarity.

24 years agoUpdate for version 2.4.
David E. O'Brien [Tue, 4 Jan 2000 03:37:56 +0000 (03:37 +0000)]
Update for version 2.4.

24 years agoTurn off GNU Info docs until someone upgrades our texinfo system to one
David E. O'Brien [Tue, 4 Jan 2000 03:27:32 +0000 (03:27 +0000)]
Turn off GNU Info docs until someone upgrades our texinfo system to one
that understands "command{foo}".