[patch] PCI IRQ lockup solved

Dear parport maintainers,

This patch solves PCI lockups if the IRQ is in use by another driver.

  Note that the parport_EPP_sup test must disable IRQs.  Otherwise the 
  already installed handler is called at initialization ... for ever 
  ... until reboot.

It also allows for PCI interrupt use, specify irq=auto, detects a level 
coupled INTX# and disables interrupts if necessary.

Below are the debug messages of two systems, both with a motherboard IO 
controller and a PCI card.  Differences are accounted for in the interrupt 
handler via the  our  integer:

  our &   1 : we are the only interrupt source
  our &   2 : awaiting int and DSR nINT active -> signal good chipset
  our &   4 : awaiting int and not a good chipset -> take it (... false_alarm)
  our &   8 : nINT still active after reset, ints were disabled
  our &0x10 : ECP ecr int expected and found, unassert PCI INTX#, as in line 
              with ECP register specifications.

Also, the patch below allows to initialize selected PCI devices by specifying 
their bus/slot/function address, via 'io=-1 io_hi=0x<bus><slot><function>'.

Furthermore, the EPP Intel test is skipped except for a base address of 
0x378, 0x278 or 0x3bc.  This seems reasonable since most (all ?) onboard 
chipsets are only configurable at those addresses, and since Intel has not 
manufacturerd any stand-alone PCI controller AFAIK.

Re: Parport signals stability with SiS chipset

	I have the same problem here. My two computers have SiS chipsets
	and the signals at pins are not set up correctly (at least not
	as I would like to see them).

	When I write data to ports using outb(), the pins change for a
	while and return to the initial state. In the initial state some
	pins are high and others are low. Using a latch and one control
	pin I could notice that the pulse is maintained by at least 1us.

	The voltage at control pins are stable. I could set it once and
	measure the desired voltage after a long time (not one year but
	lots os seconds).
	
	I should use my circuit and I programmed it in MS-DOS
	where this problem doesn't exist while using the same computers.

--

-- 
Thiago A. M. Delatorre

Website questions

Hello from Gregg C Levine
A question for the audience.
I've noticed that Tim has moved his website dicussing his project over to
the current service, along with completing moving the list to its. However,
I decided to check to see if the Old Site, was still up. It is. What is
happening regarding that one?
Gregg C Levine yodathejediknight <at> att.net or
yodathejediknight <at> worldnet.att.net
"May the Force be with you!"

Re: Website questions

On Wed, Mar 03, 2004 at 01:02:17AM -0500, Gregg C Levine wrote:

> Hello from Gregg C Levine
> A question for the audience.
> I've noticed that Tim has moved his website dicussing his project over to
> the current service, along with completing moving the list to its. However,
> I decided to check to see if the Old Site, was still up. It is. What is
> happening regarding that one?

Grant set it up to forward to cyberelk.net.

Tim.
*/

_______________________________________________
Linux-parport mailing list
Linux-parport <at> lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-parport

ppdev or lp

Hi,

I am porting a Solaris parallel daemon used for printing onto linux 
kernel 2.4.21-4.
Should I use ppdev or lp?
What are pros and cons.
Also will the device node be created on /dev/lp0... OR /dev/parport0.

Thanks in advance.
-Sandya

Re: ppdev or lp

Sandya Menon,

The device nodes should already be in the file system.  If not they can be
created with "mknod".

As far as the pros and cons between modules...  That is a real open-ended
question.  That may depend on what you are trying to accomplish.
Communicating with something?  Printing?  Its like apples and oranges...

Nate

----- Original Message ----- 
From: "Sandya Menon" <Sandya.Menon <at> Sun.COM>
To: <linux-parport <at> lists.infradead.org>
Sent: Tuesday, March 02, 2004 1:24 PM
Subject: [Linux-parport] ppdev or lp

> Hi,
>
> I am porting a Solaris parallel daemon used for printing onto linux
> kernel 2.4.21-4.
> Should I use ppdev or lp?
> What are pros and cons.
> Also will the device node be created on /dev/lp0... OR /dev/parport0.
>
> Thanks in advance.
> -Sandya
>
>
> _______________________________________________

ppSCSI driver suite for linux 2.6

Hi, 

thanks to Jochen Eisinger for the submitted patch for the ppscsi
drivers on 2.6.x kernel. It could not works properly on mine. but with
the simple following modification I managed to get it working.

the message I got was : "Unknown symbol __udivdi3"

here's he patching patch.

--- linux-ppscsi-2.6.x.patch    2004-03-07 11:06:55.610596224 +0100
+++ linux-ppscsi-2.6.x.patched.patch    2004-03-07 11:11:26.118472776 +0100
 <at>  <at>  -2602,11 +2602,11  <at>  <at> 
 +{
 +      ip[0] = 0x40;
 +      ip[1] = 0x20;
-+      ip[2] = (capacity +1) / (ip[0] * ip[1]);
++      ip[2] = ((unsigned long) capacity +1) / (ip[0] * ip[1]);
 +      if (ip[2] > 1024) {
 +              ip[0] = 0xff;
 +              ip[1] = 0x3f;
-+              ip[2] = (capacity +1) / (ip[0] * ip[1]);
++              ip[2] = ((unsigned long) capacity +1) / (ip[0] * ip[1]);
 +              if (ip[2] > 1023)
 +                      ip[2] = 1023;
 +      }

ECR Register in PCI based ECP port

Hi All,

I just now connected a PCI card into my computer, to add an extra ECP port to my
computer.

But somehow I am not able to locate the ECR register the memory space (where my PCI card
is mapped) 

Here is the device header for the card:
-> pciHeaderShow(0,0,0)
vendor ID =                   0x9710
device ID =                   0x9805
command register =            0x0001
status register =             0x0280
revision ID =                 0x01
class code =                  0x07
sub class code =              0x80
programming interface =       0x00
cache line =                  0x08
latency time =                0x00
header type =                 0x00
BIST =                        0x00
base address 0 =              0x00000129
base address 1 =              0x00000131
base address 2 =              0x00000139
base address 3 =              0x00000169
base address 4 =              0x00000121
base address 5 =              0x00000101
cardBus CIS pointer =         0x00000000
sub system vendor ID =        0x1000

Re: AW: ECR Register in PCI based ECP port

Do I need to do anything before I can start seeing the parallel port registers (ECP), ie
do I need to set any bits?

Thanks,
Ravi

--- Michael Kastner <michael.kastner <at> gmx.at> wrote:
> > But somehow I am not able to locate the ECR register the 
> > memory space (where my PCI card is mapped) 
> > 
> > I tried reading the default values from these locations, and 
> > it all returns zeroes (which is not correct, it should be 
> > 0x01. Then I tried writing 0x34 at the ECR (guessed) 
> > location, it still does not work.
> > 
> > Any ideas whats wrong here. Do I need to do anything after 
> > connecting the PCI cards such that my registers become visible.
> 
> Hello Ravi!
> 
> You can determine the base addresses by typing
> cat /proc/pci
> in a shell. This will give you a listing of all installed PCI devices.
> Among them you will find the parallel port card.
> It should look something like this.
> 
>   ...
>   Bus  0, device   9, function  0:
>     Communication controller: PCI device 9710:9805 (NetMos Technology) (rev
> 1).

Re: AW: ECR Register in PCI based ECP port

Thanks Michael!

I am trying to locate ECR at baseaddress_1 + 0x2 (as pointed out by you). 

Just to make sure, to detect the existence of ECR, the default value at that location
should be 0x01, and on writing 0x34 to it, it shall return 0x35 (since the last two bits
are read only).

Please confirm if I am right,

Thanks,
Ravi

--- Michael Kastner <michael.kastner <at> gmx.at> wrote:
> > But somehow I am not able to locate the ECR register the 
> > memory space (where my PCI card is mapped) 
> > 
> > I tried reading the default values from these locations, and 
> > it all returns zeroes (which is not correct, it should be 
> > 0x01. Then I tried writing 0x34 at the ECR (guessed) 
> > location, it still does not work.
> > 
> > Any ideas whats wrong here. Do I need to do anything after 
> > connecting the PCI cards such that my registers become visible.
> 
> Hello Ravi!
> 
> You can determine the base addresses by typing
> cat /proc/pci
> in a shell. This will give you a listing of all installed PCI devices.