Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 02 Mar 2006 00:46:14 +0100
From:      Volker <volker@vwsoft.com>
To:        Holger Kipp <hk@alogis.com>
Cc:        freebsd-stable@freebsd.org
Subject:   Re: ppp HDLC errors
Message-ID:  <44063246.1020000@vwsoft.com>
In-Reply-To: <20060301163513.GB70143@intserv.int1.b.intern>
References:  <4405BB89.204@vwsoft.com> <20060301163513.GB70143@intserv.int1.b.intern>

next in thread | previous in thread | raw e-mail | index | archive | help

On 2006-03-01 17:35, Holger Kipp wrote:
> On Wed, Mar 01, 2006 at 04:19:37PM +0100, Volker wrote:
>> Dear list members,
>>
>> I'm trying to get a 3g / UMTS connection to work on my RELENG_6
>> (i386, AMD Sempron) machine. I've installed a Novatel Merlin U630
>> card (pcmcia) which is providing a hayes serial modem to the system
>> (on sio4 in my case).
>>
>> Dialing works fine but most DNS queries are timing out, but also
>> most tcp sessions won't work. While in ppp (userland) I'm seeing
>> messages like:
> 
> I had a similar problem with PCI-800H card.
> 
>> Later I'm seeing message like that on the console screen:
>>
>> sio4: 152599 more interrupt-level buffer overflows (total 152599)
> 
> Have a look at http://www.freebsd.org/cgi/query-pr.cgi?pr=51982
> maybe it is related. You could change sio.c and increase cp4ticks
> by factor of 10, then recompile your kernel and see if it helps.

Thanks for that pointer - it DID help! :)

But multiplying cp4ticks by factor 10 didn't do it. I've set HZ=1000
in my kernel config and first tried the PR suggested value. Then I
tried the second PR suggested solution which was a bit better but
still produced error messages.

I've now set:
 cp4ticks = speed / 10 / 100 * 160
so I multiplied the default value by 40 (not just by 10) and the ppp
HDLC error messages and the interrupt kernel messages went away.
I've seen the error messages and bad behaviour even at 'low' baud
rates of 57600 kbit/s. With my mods I'm now running at 230400 kbit/s
and don't see the error messages anymore.


> 
> have you activated rtscts-flowcontrol (see /etc/rc.d/serial) for sio4?
> 

Now I did it but just after making sure it doesn't mask my true
problem. But on the other hand I've already told ppp to use ctsrts
before.

>> It's a RELENG_6 machine with a recent kernel (compiled yesterday
>> with kernel sources from Feb-26).
>>
>> sio4: unable to activate interrupt in fast mode - using normal mode
> 
> this might be if sio4 shares its irq with other resources...

Yes it does. I would have to move PCI cards to change irq
assignment. But as long as it works, I don't care about irq sharing
and now it seems to work.

Again, thank you for your hint! It saved me some more hours.

Greetings,

Volker



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?44063246.1020000>