[HECnet] DECnet for Linux

Johnny Billquist bqt at softjar.se
Tue Jan 23 14:34:35 PST 2018


Erik, thanks for all the analyzing.
This both (to me) shows the issues with a constantly changing internal 
API of the Linux kernel, and that maintaining DECnet for Linux is a 
constant effort and annoyance. If someone is willing to do it, great, 
but I suspect most people will tire after some time.
Also, we still have the issues with DECnet under Linux in general, where 
it seems it don't work right at all in some situations.

   Johnny

On 2018-01-23 13:34, Erik Olofsen wrote:
> One more note:
> 
> In the last longterm 3.x.y kernel, 3.18.92, sock_alloc_send_skb()
> sets the errcode to ENOBUFS even when the sk_buff is allocated
> successfully. This causes sendmsg() to stop with "No buffer space
> available". That is easy to fix in dn_alloc_send_pskb().
> 
> 
> On Mon, Jan 15, 2018 at 09:43:53PM +0100, Erik Olofsen wrote:
>> Some more notes on this:
>>
>> The problem of accessing a Linux machine seems to be related to
>> the standard mtu of 1500 of a network interface.
>>
>> Reducing the maximum packet size to 576 helps;
>>
>> https://sourceforge.net/p/linux-decnet/mailman/message/2395733/
>>
>> This can be done using ifconfig or using 576 instead of dst_mtu()
>> in dn_current_mss() af_decnet.c of the kernel module source.
>>
>> tcpdump shows that a large data packet is one byte too long.
>> Perhaps this occurs because the pad byte in the header is not counted
>> in dn_mss_from_pmtu() in af_decnet.c; with an additional mtu--
>> this seems to be fixed.
>>
>> Kernel warnings related to dst.h;
>>
>> https://www.spinics.net/lists/netdev/msg225996.html
>>
>> indeed disappear by using dst_metric_raw at two places.
>>
>> The above was tested with kernel 3.10.17, and can be tested
>> with http://mim.update.uu.se/hecnet and nodename RULLFL.
>>
>> dntype mim::nodenames.dat works very well on a real machine; why
>> it doesn't work very well with the virtual machine is not really
>> solved yet...
>>
>>
>> On Wed, Jan 03, 2018 at 04:44:54PM +0100, Johnny Billquist wrote:
>>> On 2018-01-03 16:02, Erik Olofsen wrote:
>>>> On a virtual machine with lubuntu 12.04, kernel 3.2.0, I have
>>>> DECnet for Linux V.2.5.68s working, with dnprogs version 2.65.
>>>> It is node RULLFL.
>>>>
>>>> The daemons and utilities use decnet.conf with node information,
>>>> so something useful would be to create it from mim::nodenames.dat.
>>>>
>>>> On a VAX, TYPEing it works well, but on the Linux machine, dntype
>>>> hangs after giving parts of the file; dndir mim:: works well.
>>>>
>>>> Does anyone have dntype mim::nodenames.dat working properly (with
>>>> perhaps different versions of the above)?
>>>
>>> I can't help, but I can report that trying to read a file from a Linux
>>> machine on MIM also hangs after a while.
>>> A big problem with the DECnet/Linux implementation is that the developers
>>> only ever tested against VMS, and nothing else.
>>>
>>> I've known for ages that it does not play well with RSX. But I do not want
>>> to dig into the Linux implementation.
>>>
>>> (The Linux DAP claims to be a VMS system, by the way, which is probably also
>>> not a good idea. Better if it had claimed to be some Unix derivative.)
>>>
>>>    Johnny
>>>
>>> -- 
>>> Johnny Billquist                  || "I'm on a bus
>>>                                    ||  on a psychedelic trip
>>> email: bqt at softjar.se             ||  Reading murder books
>>> pdp is alive!                     ||  tryin' to stay hip" - B. Idol


-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol


More information about the Hecnet-list mailing list