| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
| |
Problem found by: Tomaz Borstnar <tomaz.borstnar@over.net>
|
|
|
|
| |
Submitted by: Adrian Penisoara <ady@freebsd.ady.ro>
|
| |
|
| |
|
|
|
|
| |
Obtained from: OpenBSD
|
| |
|
|
|
|
|
|
|
| |
queue. Doing ``set ifqueue 0'' and ``set urgent none'' will allow
full use of luigi's WF2Q code.
Requested by: luigi
|
|
|
|
|
|
| |
prioritisation.
Requested by: luigi
|
|
|
|
|
|
|
| |
when opening or closing an auto link due to the autoload
setting.
Spotted by: David Hedley <david@inty.co.uk>
|
|
|
|
| |
the mask
|
|
|
|
|
| |
drop the packet rather than just whinging about it in the log.
If the fsm header has a smaller length, continue to whinge.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
DATALINK_CARRIER and turn off scripting.
This should fix instances where ``term'' is used followed by ~.
and then ``dial''/``open'' (it currently just sits there looking
at you).
Reported by: Tim Vanderhoek <vanderh@ecf.utoronto.ca>
|
|
|
|
| |
Submitted by: Renaud Waldura <renaud@guppy.evolunet.com>
|
|
|
|
| |
future...
|
| |
|
|
|
|
|
|
|
|
|
| |
used together by creating a SOCK_DGRAM socketpair() between
the processes.
Be polite when closing !program links and send a HUP to the
process. This makes ssh tunnels over unreliable media (such
as via httptunnel) reconnect properly.
|
|
|
|
| |
controls a link.
|
|
|
|
|
| |
there's no resolv.conf.
Use a umask of 022 when creating resolv.conf, not 0644 !
|
|
|
|
|
| |
Submitted by: luoqi
Forgotten by: me
|
|
|
|
| |
report that the device name is of an invalid format.
|
|
|
|
| |
peer enddisc/authname is different.
|
| |
|
|
|
|
|
|
| |
the internal network when NAT is enabled.
Allow ``set target MYADDR'' to stop packets at the gateway.
|
|
|
|
| |
Submitted by: luoqi
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
when we're redialing/reconnecting.
While we're here, log redial, reconnect and phone number
announcements to LogCHAT, and reduce some other logging to
LogDEBUG.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When an NCP reaches TLF, *ONLY* datalink_Close() links that are
in DATALINK_OPEN.
When the last link reaches TLD, DOWN all NCPs (as we used to in the
links TLF (which was the wrong place anyway)), as the NCPs aren't
now going to datalink_Close() us unexpectedly, we get to continue
doing what we were told to do in the first place.
The result: When we lose a link, the IPCP layer goes down and
we actually call the stuff in ppp.linkdown !
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
It was not a good idea to remove csu_header from struct cspace, it had
ramifications which I didn't notice.
Restore src/usr.sbin/ppp/slcompress.h to the way it was, since MAX_HDR
was already defined as 128 there and it's a user program anyway.
In sys/net/slcompress.h make MAX_HDR 128 intead of MLEN to avoid
bloat.
My apologies for any inconvenience.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
tree. This considerably reduces unnecessary bloat in struct slcompress.
I'm running with this change right now and have seen no negative
side-effects.
On my sytem this reduced kernel BSS by about 25KB.
Submitted by: bde
Approved by: brian for user-ppp
|
|
|
|
| |
if it's not DEAD or ESTABLISH
|
|
|
|
|
| |
``nat target'', and suggest the use of ``nat target default'' as an
interesting possibility.
|
| |
|
| |
|
|
|
|
| |
Pointed out by: Maxim Sobolev <sobomax@altavista.net>
|
|
|
|
| |
``nat deny_incoming yes'') by libalias.
|
|
|
|
|
|
|
|
| |
numbers in all commands.
If people use hostnames and have dodgy resolvers or try to resolve
the hostname before the link is up, they get what they deserve....
Requested by: ru
|
| |
|
| |
|
| |
|
|
|
|
| |
process in -background mode (it should report failure if appropriate).
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
don't bother to re-initialise the NCPs. Instead wait for
bundle_LinkClosed() to be called - IFF it actually is called.
By initialising the NCPs at this point, ppp was recursing
back into the fsm_Down() routing for the link, and losing
track of the reason that the link was being brought down.
The end result was that ``set reconnect'' would never do
anything.
Patiently pointed out by: ru
|
|
|
|
|
| |
background/foreground/direct and zero'd the redial/reconnect
counts.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
if the childs exec() has succeeded or failed by taking advantage
of the fact that both processes share the same memory.
FWIW:
I tried to implement this by doing a pipe(), setting the
write desciptors close-on-exec flag in the child and writing
errno to the descriptor if the exec() fails. The parent can
then ``if (read()) got errno else exec worked''.
This didn't work though - the child could write() to fd[1] on
exec failure, but the parent got 0 trying to read() from fd[0] !
Is this a bug in execve() ?
|
|
|
|
|
|
|
|
| |
dropping out of background/foreground/direct mode.
This avoids either having to wait for the redial timer before
exiting or jaming up in select() waiting for something that'll
never happen.
|
|
|
|
| |
being dealt with correctly.
|
|
|
|
|
| |
Suggested by: sheldonh
Forgotten by: me
|