Matthew Dillon [Wed, 26 Jan 2000 20:51:29 +0000 (20:51 +0000)]
Fix catastrophic bug in NQNFS related to UDP mounts. The 'nqhost'
struct contains a major union for which lph_slp was being initialized
only for TCP connections, but accessed for all types of connections
leading to a crash. Also, a conditional controlling an nfs_slplock()
call contained an improper paren grouping, causing a second crash in
the UDP case.
The nqhost structure has been reorganized and lph_slp has been made a
normal structural field rather then a union field, and properly
initialized for all connection types.
Peter Wemm [Wed, 26 Jan 2000 17:38:51 +0000 (17:38 +0000)]
Sigh. Really fix it this time. It seems the first time through the
run it would modify the source tree, causing it to fail the second
time around. Sigh.
Peter Wemm [Wed, 26 Jan 2000 17:12:09 +0000 (17:12 +0000)]
Fix world breakage (kdump, truss) causes by rev 1.107 (adding dev/usb).
It was creating ${.OBJDIR}/${MACHINE}/usr/include/dev -> .../sys/dev
and mkioctls would descend that and saw *all* of src/sys/dev/*/*.h,
not just dev/usb/*.h. It then choked on the dpt includes.
Robert Watson [Wed, 26 Jan 2000 16:15:48 +0000 (16:15 +0000)]
A few more touchups:
- clean up unneeded AFS ID type
- Add Coda, NTFS, NWFS ACL types
- Add acl_dup() prototype
- Remove acl_calc_mask, which belongs in the editing library
- Introduce posix1e.3, a man page introducing POSIX.1e library calls
(more man pages to follow)
Peter Wemm [Wed, 26 Jan 2000 15:15:50 +0000 (15:15 +0000)]
Re-add vpo. I've just re-tested this in a 'make world' on a December
vintage system, well before the ppbus changes. When I called it an
"example" module, I meant as an example for the rest of the ppbus client
drivers, not that it was worthless. I'll mail my 5.8MB world.log to
anybody who doesn't believe me.
Yoshinobu Inoue [Wed, 26 Jan 2000 09:02:42 +0000 (09:02 +0000)]
bindresvport related changes
-changed bindresvport2 to bindresvport_sa
-merged the man into bindresvport.3
All discussion between Jean-Luc Richier <Jean-Luc.Richier@imag.fr>,
Theo de Raadt <deraadt@cvs.openbsd.org>, itojun, is reflected to
this code. (Actually Theo de Raadt write the code simultaneously as the
discussion change.)
Yoshinobu Inoue [Wed, 26 Jan 2000 08:37:29 +0000 (08:37 +0000)]
Fix getaddrinfo() behaviour to be more compliant with RFC2553. Patches are
obtained from itojun.
-don't filter address families which are not supported by system at
FQDN resolving.
-don't do reverse lookup
I think I checked all lib and tools which use getaddrinfo() if
this change affect them.
Robert Watson [Wed, 26 Jan 2000 04:19:38 +0000 (04:19 +0000)]
Minor fixes to library interface to improve POSIX.1e compliance. This
adds _np to a couple of function prototypes that provided more broad/useful
interfaces than POSIX.1e interfaces included.
Also, move from using a heuristic to identify POSIX.1e-semantic ACLs to
using different ACL types for non-POSIX.1e ACLs. This should clean up the
existing fuzzy logic that determined when acl_sort() should be applied
before kernel submission.
Mike Smith [Wed, 26 Jan 2000 01:32:51 +0000 (01:32 +0000)]
Correctly initialise the available IRQ numbers in the APIC_IO case.
IRQ 2 was being unilaterally disallowed, which is only appropriate if
the interrupt hardware is the traditional chained PIC arrangement.
Søren Schmidt [Tue, 25 Jan 2000 20:14:51 +0000 (20:14 +0000)]
Retry a bit more agressively on the atapi identify.
Try to support older systems reporting irq0 for the first channels.
Support sharing of the std interrupts (says peter :) )
Dont use READ_CD on normal data reads (2048 bytes), too many old drives
doesn't support this command even if the std says "shall" :(, but still
use READ_CD on all other blocksizes.
Add the geometry to the ad probe, its still usefull.
Argh, this is what I get for testing, making one last "harmless cosmetic
tweak" and then committing. /me got what he deserved, an embarassing build
failure.
Matthew Dillon [Tue, 25 Jan 2000 17:49:12 +0000 (17:49 +0000)]
The swapdev_vp changes made to rip out the swap specfs interaction
also broke diskless swapping. Moving the swapdev_vp initialization
to more commonly run code solves the problem.
PR: kern/16165
Additional testing by: David Gilbert <dgilbert@velocet.ca>
Matt Jacob [Tue, 25 Jan 2000 17:42:27 +0000 (17:42 +0000)]
Propagate sense data through from SIM (if there) and mark the CTIO
that there's sense to send with status (if the SIM does it), and
then clear any pending contingent allegiance state for this initiator
if the SIM actually did send the sense data.
Widen MAX_INITITATORS to 256- that's still not quite right, but will
accomodate the widest Fibre Channel support in FreeBSD now.
Matt Jacob [Tue, 25 Jan 2000 17:39:09 +0000 (17:39 +0000)]
Propagate sense data through from SIM (if there) and mark the CTIO
that there's sense to send with status (if the SIM does it).
Reviewed by: gibbs@freebsd.org
Peter Wemm [Tue, 25 Jan 2000 16:45:54 +0000 (16:45 +0000)]
Always leave SP_DCEN on (monitor DCD). Otherwise the firmware *really*
does ignore DCD. Even TIOCMGET cannot read DCD as the firmware doesn't
report it. This has pretty interesting effects for ppp(8) which runs
in clocal mode and polls carrier (!). (Specialix's linux driver does
this too)
Also update the firmware to 3.0.6 for the SX cards, as apparently there
was a problem with floating (disconnected) DCD pins causing stray carrier
transitions, especially at port open time.
It seems to work here, and carrier loss is detected nearly immediately
rather than having to wait for a LQR timeout (a few minutes) before ppp(8)
gives up.
Ruslan Ermilov [Tue, 25 Jan 2000 12:24:06 +0000 (12:24 +0000)]
Now that kernel is capable of notifying user processes about
the interface MTU change (src/sys/net/if_sl.c,v 1.83), track
interface MTU with -dynamic option as well.
David E. O'Brien [Tue, 25 Jan 2000 12:23:13 +0000 (12:23 +0000)]
Do not use "DEFAULT_VTABLE_THUNKS". The bugs that have existed since
EGCS 1.x have not been worked out. And with 4.0 RELEASE comming quickly
we need C++ to be stable and usable.
Bruce Evans [Tue, 25 Jan 2000 09:20:08 +0000 (09:20 +0000)]
Don't follow null pointers if we somehow have a null devswitch entry
despite having a non-null cn_tab entry. This case now works the same
as if there is no physical console, except i/o at the kernel printf
level may still work. This frees drivers of physical console drivers
from the responsibility of attaching the device no matter what.
Bruce Evans [Tue, 25 Jan 2000 09:01:55 +0000 (09:01 +0000)]
Fixed the profiling version ALTENTRY(). Again. The previous version
didn't set up the frame pointer before calling mcount, and then jumped
to the wrong place in ENTRY() to defeat the point of the jump.
David E. O'Brien [Tue, 25 Jan 2000 06:58:15 +0000 (06:58 +0000)]
Do not use "DEFAULT_VTABLE_THUNKS". The bugs that have existed since
EGCS 1.x have not been worked out. And with 4.0 RELEASE comming quickly
we need C++ to be stable and usable.
Matthew Dillon [Tue, 25 Jan 2000 01:51:21 +0000 (01:51 +0000)]
Type-o, change from[...] = 0 to fromb[...] = 0. The incorrect buffer
was having its last element zero'd. It turns out not to be a security
hole or to have any real effect on the code because 'from' was previously
pointing to a buffer of the same size as 'fromb', and the last
element in fromb is already 0 anyway due to the use of sizeof(fromb)-1
in the strncpy() call. But I'm not pressing my luck so only the type-o
is being fixed.
Yoshinobu Inoue [Tue, 25 Jan 2000 01:05:18 +0000 (01:05 +0000)]
Fix the bug that IPv4 ttl is not initialized when AF_INET6 socket is used
for IPv4 communication.(IPv4 mapped IPv6 addr.)
Also removed IPv6 hoplimit initialization because it is alway done at
tcp_output.
Confirmed by: Bernd Walter <ticso@cicely5.cicely.de>
Archie Cobbs [Mon, 24 Jan 2000 21:38:18 +0000 (21:38 +0000)]
Fix typo for the flag ``--ungzip'' which should have been ``--gunzip''
as is documented in the man page. Retain the older mistaken version
of the flag for backwards compatibility in case anybody is using it.
Add $FreeBSD$ tag as cvs requires it.
Søren Schmidt [Mon, 24 Jan 2000 20:45:24 +0000 (20:45 +0000)]
General cleanup.
Dont be so verbose in the probe, only ONE line printed now, to get more
info boot verbose. Centralise most printf's in ata-all & ata-dma to use
the ata_printf function, it saves alot of codelines.
Repeat the identify command if drive fails the first.
Protect the timeout functions with splbio.
Dont update the transfer details before we are sure the transfer
succeded, this way they are proberly retried on errors.
Move the handling of next_writeable to userland.
Use the READ_CD command to read CD's. That enables us to read _anything_
via the normal read/write interface. This kindof obsoletes the READAUDIO
ioctl, but we keep that for now.
Brian Somers [Mon, 24 Jan 2000 20:39:02 +0000 (20:39 +0000)]
Move the *intrq variables into net/intrq.c and unconditionally
include this in all kernels. Declare some const *intrq_present
variables that can be checked by a module prior to using *intrq
to queue data.
Make the if_tun module capable of processing atm, ip, ip6, ipx,
natm and netatalk packets when TUNSIFHEAD is ioctl()d on.
Mark Murray [Mon, 24 Jan 2000 20:16:41 +0000 (20:16 +0000)]
Add a (commented out) macro that will, when uncommented, cause Kerberos5
(AKA Heimdal) to be built. Suitable admonishments about the experimental
state of this code are included.
Peter Wemm [Mon, 24 Jan 2000 17:16:03 +0000 (17:16 +0000)]
Copy i386/isa/atapi-cd.[ch] to a new name so that it doesn't have the
same object file (atapi-cd.o) as the ata drivers. I'd have called it
wcd.[ch], but there's already one of those in the Attic that we can't
clobber - the good names are taken.
Fix building so that it can be compiled into LINT alongside ata.