🚀 go-pugleaf

Hi all,

I have just noticed, after running Linux for about a year and a half,
that my serial throughput is absolutely dreadful, yet when running any
other operating system, the throughput is fine.

It seems that under Linux, there are errors in many of the characters
received, such that when a PPP link is run, an annoyingly high rate of 
packets are dropped:

ppp0      Link encap:Point-Point Protocol  
          inet addr:10.0.2.15  P-t-P:128.250.80.9  Mask:255.0.0.0
          UP POINTOPOINT RUNNING  MTU:552  Metric:1
          RX packets:5351 errors:126 dropped:126 overruns:0
          TX packets:1340 errors:0 dropped:0 overruns:0

The above, at 2.3% bad packets, is actually one of my *better* rates. Normally
it sits around 5-7% dropped packets, which seems to give an overall throughput
of about 1.1kbytes per sec. as reported by ncftp or Netscape when used on a 
28.8kbps link. Under any other operating system, I get around 3.1 or so kbytes 
per second.

Obviously, given the success in non-Linux OS's, I can rule out the modem,
phone line, serial card as being the problem (I'd already ruled out the
first two, anyway, by swapping them with other modems/phone lines).

That leaves it to be my configuration of Linux (running kernal 2.0.27, libc
5.4.17).

I've noticed that the errors tend to occur when there is hard drive activity:
the following is an output of 'ls -la' on a remote Unix box over a straight
serial link - as you can see, several characters were lost in transit
because my computer accessed the HDD for various reasons whilst this was being
done. (Given that the serial ports are on the same card as the IDE controller,
it seems logical, but since other OS's don't seem to have problems with this,
I don't see why Linux should).

-rwx------  1 pbd      staff        2247 Dec 16 10:02 .cshrc.old*
-rw-r--r--  1 pbd      staff          66 Jan 11 19:28 .display
drwxr-xr-x 11 pbd      staff         512 Dec 12 10:26 .dt/
-rwxr-xr-x  1 pbd      staff        5400 Dec  4 16:49 .dtprofile*
w---  pb   staff      3786 Nov 28 12:17 .emacs/                                 
-rw-r--r--  1 pbd      staff       49450 Oct 12 091 .error                      
rw-r--r--  1 pbd      staff          63 Dec  6 10:30 .forwardtml                
-rw-r--r--  1 pbd    staff       13878 Dec  9 11:10 .fvwmrc                     
-rw-r--r--  1 pbd      staff       10572 Feb  2  1996 .fvwmrc-ugrad             
-rw-------  1 pbd      staff           0 Dec 13 08:55 .ircmotd                  

Has anyone else had this problem before? I'd much appreciate any advice on
this, because the stalls in data transfer are becoming very annoying...

Cheers,

Paul.


-- 
Paul Dwerryhouse                              | p.dwerryhouse@ee.mu.oz.au
Computer Systems & Networks Manager           |
Dept. of Electrical & Electronic Engineering  |
University of Melbourne, Australia            |