Re: [PARPORT] System ressource pig??


Grant R. Guenther (grant@torque.net)
Fri, 6 Mar 1998 07:19:55 -0500 (EST)


In future, please send (suspected) bug reports to me directly - not to
the entire list.

> My motherboard is a P90 Premiere-II (Plato) with a SMC37C665 IO chip,
> which is supposed to be ECP and EPP compatible.

But ...
 
> Mar 5 22:21:06 skil kernel: pcd0: kbic 0.94, KingByte KBIC-951A at 0x3bc,
> using mode 2 (8-bit)

It would appear that you have the port setup incorrectly. Move it to
either 0x278 or 0x378 and set it to EPP mode (but not ECP).

> During the read from the pcd device I have experienced heavy system
> performance degradation: most of the cpu cycles seem to have been burnt in
> kernel mode (>80% system load).

Think about it for a moment: these devices do not have an interrupt
capability, so the drives must be actively polled. If you load the
driver with "nice=1" on the command line, it will use less CPU time, with
correspondingly lower transfer rates.

(I am looking at other polling strategies, but the coarse timer granularity
is still the major hurdle. I've heard some rumours that this might
eventually change ...)

> I have noticed after about 2 hours heavy PP-CDROM usage the (SW) clock
> (-> date) was 2 minutes behind the (CMOS) clock (-> clock) ! Both clocks
> are in synced at system boot time.

This is clearly a bug, I will look into it.

> When mounting the pcd device, I noticed some error messages:
>
> Mar 5 22:21:27 skil kernel: pcd0: read block completion: alt=0x51 stat=0x51
> err=0x20 loop=0 phase=3
> Mar 5 22:21:27 skil kernel: pcd0: read block data done: alt=0x51 stat=0x51
> err=0x20 loop=0 phase=3
> Mar 5 22:21:27 skil kernel: pcd0: read block completion: alt=0x51 stat=0x51
> err=0x68 loop=0 phase=3
> Mar 5 22:21:27 skil kernel: pcd0: read block data done: alt=0x51 stat=0x51
> err=0x68 loop=0 phase=3

This is a problem with some newer drives and the 0.94 driver. It has been
fixed in 1.01 (found in 2.1.8x ...).

> Do those err=0x20 err=0x68 have anything to do with the resource
> utilization and the responsiveness?

Probably not. The drive takes up to 15 seconds to get up to speed - pcd
didn't expect that. This may be where you are losing time, as well. If
you can investigate that and report to me, I'd appreciate it.

> Is the achieved data rate normal, or is there a way for higher
> throughput?

Fix your port settings.

--------------------------------------------------------------------------
Grant R. Guenther grant@torque.net
--------------------------------------------------------------------------

-- To unsubscribe, send mail to: linux-parport-request@torque.net --
-- with the single word "unsubscribe" in the body of the message. --



This archive was generated by hypermail 2.0b3 on Wed 30 Dec 1998 - 10:17:31 EST