Bruce Evans [Sun, 12 Mar 2000 14:23:21 +0000 (14:23 +0000)]
Try harder to make the lower 16 bits of fsids unique. The vfs type
number was packed very wastefully, giving perfect non-uniqeness in
the lower 16 bits of fsids for filesystems with the same vfs type.
This made linux_stat() return perfectly non-unique (broken) 16-bit
st_dev's for nfs mount points, and effectively reduced mntid_base to
8 bits so that the vfs_getnewfsid() looped endlessly when there are
already 256 mounted filesystems with the required vfs type.
Peter Dufault [Sun, 12 Mar 2000 13:25:14 +0000 (13:25 +0000)]
Disable RTPRIO and P1003.1B scheduling in ntpd. Only leave BSD_NICE
for a scheduling boost. This is a conservative change that should
make no difference in practice and eliminate concerns about this being
the source of some SMP hangs.
Configuration scripts should never auto-configure P1003.1B
without a second test. The behavior with respect to regular time
sharing, who can access it, etc., is not defined.
Yoshinobu Inoue [Sat, 11 Mar 2000 20:57:31 +0000 (20:57 +0000)]
Add missing return.
"ndp" command should deletes only neighbor cache entries, but the
program lacks necessary "return" after the neighbor cache
entry check, so it might deletes non neighbor cache entries.
(it seems that usually no problem happens.)
Yoshinobu Inoue [Sat, 11 Mar 2000 20:52:01 +0000 (20:52 +0000)]
Clear sin6_scope_id before passing routes into kernel, becaues kernel
don't support routes with sin6_scope_id set.
Without this fix, routes with IPv6 scoped addr won't work when it is
assigned by "route" command.
Yoshinobu Inoue [Sat, 11 Mar 2000 20:44:53 +0000 (20:44 +0000)]
Forbid include of netinet6/ip6.h from user-land, and if included,
print an error message which say, "include netinet/ip6.h".
This is postponed to apply to avoid tcpdump compile error.
Now apply this because tcpdump has been already fixed.
Warner Losh [Sat, 11 Mar 2000 20:22:09 +0000 (20:22 +0000)]
Fix crashes on card eject for pccard modems. We check for NULL when
we get the com address. If so, we go ahead and return. Bruce thinks
there's a bug in the pccard layer that it terminates devices with
extreme prejustice rather than letting them deside for themselves when
to terminate (and he's likely right). This fix doesn't change that,
but instead works around it by checking for NULL pointers at more
places than before.
The detach routine still calls functions at interrupt level that
aren't reentrant. In theory this could cause a problem, but none
showed up in practice. Future versions should correct this problem,
likely by making the detach process a thread/process at the pccard
level. NEWCARD will do this, and the current pccard layer should
likely be modified to that as well, should it live long enough.
A few style nits of the same form that were in my original patch sent
off to bde were also fixed as part of this process. Mostly use of
!ptr and return ENOPARENS.
This should prevent a crash on suspend with an active ppp link as
well, but that wasn't tested.
Yoshinobu Inoue [Sat, 11 Mar 2000 20:10:39 +0000 (20:10 +0000)]
Prevent multicast packets from being sent over stf0 interface,
to avoid error messages printed on /var/log/messages.
Recently added "stf" interface has IFF_MULTICAST bit set,
because IPv6 don't work on non IFF_MULTICAST set interfaces
in current implementation.
But "stf" interface is multi-dest IPv6 over IPv4 tunnel, so
can't treat multicast packet with no special configuration.
On the other hand, tools like "rtadvd", "pim6dd", "pim6sd"
send messages on IFF_MULTICAST set interfaces by default.
So there will be many bogus error message on /var/log/messages,
that multicast packets sent to "stf0" failed.
So,
-strip "stf0" from rtadvd sending interfaces in rc.network6
("rtadvd" is a daemon used only for local subnet, so it
will never need to send its packets over "stf0" interfaces.)
-Add default configuration files for "pim6dd" and "pim6sd",
and disable "stf0" in those files.
Yoshinobu Inoue [Sat, 11 Mar 2000 20:03:22 +0000 (20:03 +0000)]
Replace m_pkthdr.rcvif with oif when oif is not NULL, to count
icmp6 error statistics based on sending interface.
This also prevent kernel panic when rcvif is not initialized after M_PKTHDR().
(The initialization issue also need to be fixed in the future.)
Duncan Barclay [Sat, 11 Mar 2000 13:25:33 +0000 (13:25 +0000)]
Start to use ray_start_join_net (my old code removed on next checkin)
requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this.
Tidyed up softc due to above.
Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
Yoshinobu Inoue [Sat, 11 Mar 2000 11:25:00 +0000 (11:25 +0000)]
Prevent sockstat print out some empty entries related to IPv6.
Sockstat use netstat tcp/udp socket print, and fstat tcp/udp
socket print, but it just specify all of IPPROTO_IP and
IPPROTO_IPV6 sockets for fstat.
Now IPv6 socket also use raw socket, but only netstat don't print it,
so now they are printed as empty entries in sockstat output.
Yoshinobu Inoue [Sat, 11 Mar 2000 11:20:52 +0000 (11:20 +0000)]
This is null commit to add correct comment for the last fix to this file.
(Last commit message is not related to this file. This file is mistakenly
committed with last commit of other files.)
Prevent possible kernel panic at tcp_input().
When some IPv4 apps do setsockopt() to control message receipt,
then kernel will panic at control message creation processing.
(Currently, only bgp daemon in KAME package known to cause the problem.)
Yoshinobu Inoue [Sat, 11 Mar 2000 11:17:24 +0000 (11:17 +0000)]
IPv6 6to4 support.
Now most big problem of IPv6 is getting IPv6 address
assignment.
6to4 solve the problem. 6to4 addr is defined like below,
2002: 4byte v4 addr : 2byte SLA ID : 8byte interface ID
The most important point of the address format is that an IPv4 addr
is embeded in it. So any user who has IPv4 addr can get IPv6 address
block with 2byte subnet space. Also, the IPv4 addr is used for
semi-automatic IPv6 over IPv4 tunneling.
With 6to4, getting IPv6 addr become dramatically easy.
The attached patch enable 6to4 extension, and confirmed to work,
between "Richard Seaman, Jr." <dick@tar.com> and me.
Kazutaka YOKOTA [Sat, 11 Mar 2000 07:44:10 +0000 (07:44 +0000)]
One more patch for the atkbd driver. It will make sure that the
keyboard port and interrupt is enabled and the driver is attached even
when the keyboard itself is not present when the system is booting.
(This has been the behavior through out 2.X and 3.X, but is somehow
broken in 4.0.)
# I certainly don't recommend people to `hot-plug' the AT keyboard,
# because the interface isn't designed for hot-plugging and such act
# will often break the keyboard controller. But, so many people want to
# do that anyway...
Mike Smith [Sat, 11 Mar 2000 05:20:56 +0000 (05:20 +0000)]
Teach the 'dc' driver how to pick up settings left over by the
SRM on alpha systems. This is an expedient if not entirely
elegant solution to the problem.
Robert Watson [Sat, 11 Mar 2000 00:24:29 +0000 (00:24 +0000)]
The function arpintr() incorrectly checks m->m_len to detect incomplete
ARP packets. This can incorrectly reject complete frames since the frame
could be stored in more than one mbuf.
The following patches fix the length comparisson, and add several
diagnostic log messages to the interrupt handler for out-of-the-norm ARP
packets. This should make ARP problems easier to detect, diagnose and
fix.
Submitted by: C. Stephen Gunn <csg@waterspout.com>
Approved by: jkh
Reviewed by: rwatson
Nik Clayton [Fri, 10 Mar 2000 11:48:49 +0000 (11:48 +0000)]
- typos
- Add double spaces following full stops to improve typeset output
- mdoc-ification. (Though I'm uncertain whether option values and
contents should be .Dq or something else).
- Fix a missed /etc/ssh change
- Expand wording on RandomSeed and behaviour when X11 isn't forwarded.
- Change examples to literal mode.
- Trim trailing whitespace
PR: docs/17292
Submitted by: Peter Jeremy <peter.jeremy@alcatel.com.au>
Alan Cox [Fri, 10 Mar 2000 09:11:24 +0000 (09:11 +0000)]
shmat: If VM_PROT_READ_IS_EXEC is defined and prot includes VM_PROT_READ,
VM_PROT_EXECUTE must be added to prot before calling vm_map_find.
Without this change, an mprotect on a shmat'ed region fails (when
it shouldn't). This bug was reported Feb 28 by Brooks Davis
<brooks@one-eyed-alien.net> on -hackers.
Warner Losh [Fri, 10 Mar 2000 05:43:29 +0000 (05:43 +0000)]
Add support for PnP ISA cards, and some laptop PnP support:
o break out some of the probe routine the allocation of resources
into an attach routine
o Recognize PnP ids
o Allocate IRQ per card rather than per system
o Better polling reporting
o Remove unneeded include files in slot.h
o store a pseudo unit number on each device we find.
o Pass a unit number to interrupt/timeout routine and use it for polling
the hardware.
Tested on: My VAIO and with the Linksys pccard reader.
Approved by: jkh
Warner Losh [Fri, 10 Mar 2000 00:43:49 +0000 (00:43 +0000)]
Minor tweaks to Xircom entries. With this I can now attach to the
modem side of my Xircom CreditCard Ethernet+Modem 33.6. Eliminate bad
Xircom entry. Complete IBM Push/Pop Modem(14.4 -> Push/Pop
Modem(14.4K).
Fix a problem which caused i/o resource allocation conflict of this
particular ISDN card with the 2nd IDE controller and/or other cards
depending on configuration.
The amount of i/o ports allocated are reduced from 0x40 per range to
0x20 per range.
Submitted by: Andrew Gordon <arg@arg1.demon.co.uk>
Approved by: jkh
Matthew Dillon [Thu, 9 Mar 2000 18:54:59 +0000 (18:54 +0000)]
In the 'found' case for ufs_lookup() the underlying bp's data was
being accessed after the bp had been releaed. A simple move of the
brelse() solves the problem.
Approved by: jkh
Submitted by: Ian Dowse <iedowse@maths.tcd.ie>
Paul Richards [Thu, 9 Mar 2000 18:11:16 +0000 (18:11 +0000)]
Fix various unsigned vs signed errors that caused problems with uids
and gids bigger than 16 bits. Added checks for uids and gids that are
bigger than 32 bits.
Approved by: jkh (partly, this fix is bigger than I first intended)
Yoshinobu Inoue [Thu, 9 Mar 2000 15:07:38 +0000 (15:07 +0000)]
Fix addr length argument value passed to sendto().
Some inetd internal udp servers didn't worked with problem.
Also fix recvfrom() "fromlen" arg type from int * to socklen_t *.
Yoshinobu Inoue [Thu, 9 Mar 2000 14:57:16 +0000 (14:57 +0000)]
Initialize mbuf pointer at getting ipsec policy.
Without this, kernel will panic at getsockopt() of IPSEC_POLICY.
Also make compilable libipsec/test-policy.c which tries getsockopt() of
IPSEC_POLICY.
Yoshinobu Inoue [Thu, 9 Mar 2000 14:47:21 +0000 (14:47 +0000)]
Update icmp node info query message bit order of query types,
according to draft-ietf-ipngwg-icmp-name-lookups-04 to 05 change.
This is necessary before 4.0, because,
-This change is non backword compatible
-Other KAME derived platforms applied 05
-Author of the draft said he never do backword imcompatible changes
again.
Bruce Evans [Thu, 9 Mar 2000 05:21:10 +0000 (05:21 +0000)]
Don't forget to check for unsupported features when updating. It was
possible to defeat the check for rw incompatibilty by mounting ro and
updating to rw.
Mike Smith [Thu, 9 Mar 2000 01:27:55 +0000 (01:27 +0000)]
Remove the PCI device class and all the PCI-only devices. With the newbus
conversion this no longer works, and userconfig is not expected to ever
again be used for these devices.
Warner Losh [Thu, 9 Mar 2000 00:35:38 +0000 (00:35 +0000)]
More new/corrected devices:
o Correct Entry for Intel EtherExpress (jkoshy)
o Add Option International 56K modem (markm)
o Socket LP-E low power WinCE ethernet card (imp)