I cc this message to all other ISDN developers, I 'm more responsible for
the hardware driver part.
On Fri, Oct 22, 1999 at 05:31:42PM +0300, Ilpo Ruotsalainen wrote:
...
>
> >Yes, this message is not hide (but has only debug prio), because it
> >should be not happen in normal cases (I looked trough the log of my syncppp
> >machines and didn't found it).
> >It seems here is a problem with some places which allocate skb with not
> >enought headroom (old version was crashing in this case).
> >Now we reallocate the needed headroom and print this message to find out
> >the real reason.
>
> If this is any help in finding the reason, the message sequence is reliably
> triggered on small incoming TCP packets at least. Like this (nightrunner is
> the ISDN machine, gunbuster is a machine elsewhere in the Internet):
>
> nightrunner:$ nc -l -p 2000
>
> gunbuster:$ echo foo|nc nightrunner.whatever.fi 2000
> (nightrunner logs the sequence "under 0 2", "under 4 5", "under 0 2" here)
>
> Switching the machines (ie. listening on the outside and sending from the
> ISDN machine) no messages get logged. Card is Teles 16.0, HiSax driver,
This is independ from ISDN hardware.
> using one channel only. ICMP doesn't seem to trigger it (tried pings of
> 64bytes, 10000bytes and 50000bytes, nothing in log). I'll try more stuff if
> you think it's of any use, can provide more info about the configuration
> too if it seems necessary. Just ask if you want me to try something or give
> more info.
Interesting, because this function is called only in out direction (the
space in headroom is need for the syncppp header).
Maybe be some of the part which send a ACK for tcp don't alloc enougth
headroom.
The question is, is it possible to change this behaviour.
Again, no headroom isn't a error, only a bad thing, which cost a extra
copy of the frame to have enought place for the header.
If it is not possible that these parts know about our need of headroom, we
should remove the message.
What I don't understand is, that most frames seems to have enought headroom
and only some make trouble here.
>
> >> Could someone who understands the driver better look at isdn_ppp.c at line
> >> 1428 and say if this is just a stupid debug message left there to flood
> >> the kernel log?
> >
> >You can disable prio DEBUG in syslog.conf.
>
> Actually I can't since I want some other messages that come on that
> priority to really be logged, but perhaps I'll just comment this one and
> recompile my kernel (argh darn, another 402min compile ;)
>
OK. deleting the message is save and should only recompile this file, if you
don't did a make clean.
Karsten
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/