Date: Sat, 10 Oct 1998 11:40:02 +0200 From: Gary Jennejohn <garyj@peedub.muc.de> To: Christoph Weber-Fahr <listmail@helena.callcenter.systemhaus.net> Cc: freebsd-isdn@FreeBSD.ORG Subject: Re: what's wrong with a1 and irq 2 ? Message-ID: <199810100940.LAA10691@peedub.muc.de> In-Reply-To: Your message of "Fri, 09 Oct 1998 23:00:54 %2B0200." <199810092100.XAA10590@helena.otelo-call.de>
next in thread | previous in thread | raw e-mail | index | archive | help
Christoph Weber-Fahr writes: >This effect is still there. I would conclude that it is not related >to the use of IRQ 2/9 and probably caused by the high interrupt load. >Is Syscons particularily affected by high interrupt load ? >Are other people also experiencing this effect ? I seem to remember >from a web page that the i4l folks also had problems with too many >interrupts from an A1. > >> Another observation I made is that the card seems to put a heavy interrupt >> load on the machine. Top gives an interrupt load of around 10% while the >> card is under load, with occasional (rare) peaks up to 50 %. This is >> a 486/50 (ISA only) System. > As said above, this problem remains unchanged, which means that ... > >> Is this a general problem with this card or is the IRQ 2 issue involved >> here ? > >... this question can be answered with the claim that irq 2 >is not involved. > >I think from this one could conclude that > >- IRQ2/9 should be supported by the avm driver, maybe there should be > a dire warning about obscure VGA cards somewhere in the docs >- There is a problem in the AVM driver with a too high interrupt load. > >I don't think this interrupt stuff is 'normal'. This is eating 10% of the >computing power of a 486/50 with only one B channel used. >On the other hand I can put the same card into a lousy 386sx/16 and >blast 2 simultaneous IDTrans sesssions with full speed over both B >channels under good plain old DOS. (I've done that!). There >must be some problem in the driver somewhere, which AVM's own >DOS CAPI driver doesn't have. > I think this is more likely due to the chipset than anything else. Consider that you optimally get an interrupt every 32 bytes. At full speed, say 7.5 kB/s, this is roughly 230 interrupts/s. But normally there are more interrupts generated than that because a) you have many packets < 32 bytes, b) you're also transmitting at the same time and c) the D channel is also generating interrupts. I see about 300 interrupts/sec at full-bore. This should hold true for any of the passive cards which use the Siemens chipset. You could try running something like xsysinfo or xpermon++ (assuming you've got The X Window System running), or maybe vmstat to see just how many interrupts are being generated by the a1. My 200MHz Pentium has no problems with this interrupt load. I don't have a 486 box anymore to test this and can't remember anymore what the behavior was like when I had such a machine. Besides, that was with bisdn. I suspect that the DOS CAPI is just polling. This is OK under DOS, but not under UNIX. >> P.S. Should I try to submit a formal 'patch' for the avm driver >> to accept IRQ2 (it's utterly trivial) ? > sure --- Gary Jennejohn Home - garyj@muc.de Work - garyj@fkr.dec.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-isdn" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199810100940.LAA10691>