Skip to contentSkip to conversation

HIL status

parisc-linux Matthew Wilcox 5 participants 11 messages Nov 9, 1999 – Nov 11, 1999

lore
Share & tools
lore
#0 lore raw

I've taken the hp300 HIL driver written by Phil Blundell and it `works' on my 715/33. It's producing the wrong keycodes, it's doing work that an ASP driver ought to be doing and it's producing a fair bit of extraneous crap on the screen. But I'm getting interrupts and keycodes so I'm happy for tonight :-)

The next obvious step is to have an ASP manager, like LASI. In fact, so much like LASI, it probably ought to be in the same driver. This seems to be how Mach does it, and there's a rumour this is also how HPUX does it. My question is: how? Mach hardcodes the address of ASP to 0xf0800000, but there's nothing even close in the driver list. I notice in the bootlogs which have previously been posted, LASI was at the address of `Core BA', but you can see that's wildly different on this machine. Should I invent my own and insert it into the structures we have?

Here's the driver list printed at boot on the 715 (is someone collecting these somewhere?)

1. Stinger Optional Graphics (10) at 0xf4000000, versions 0x6, 0x0, 0x77, 0x0, 0x0
2. Scorpio Jr Core BA (11) at 0xf082f000, versions 0xb, 0x0, 0x70, 0x0, 0x0
3. Scorpio Jr Core SCSI (10) at 0xf0825000, versions 0xb, 0x0, 0x71, 0x0, 0x0
4. Scorpio Jr Core LAN (802.3) (10) at 0xf0826000, versions 0xb, 0x0, 0x72, 0x0, 0x0
5. Scorpio Jr Core HIL (10) at 0xf0821000, versions 0xb, 0x0, 0x73, 0x0, 0x0
6. Scorpio Jr Core RS-232 (10) at 0xf0823000, versions 0xb, 0x0, 0x75, 0x0, 0x0
7. Scorpio Jr Core RS-232 (10) at 0xf0822000, versions 0xb, 0x0, 0x75, 0x0, 0x0
8. Scorpio Jr Core Centronics (10) at 0xf0824000, versions 0xb, 0x0, 0x74, 0x0, 0x0
9. Scorpio Jr Audio (10) at 0xf1000000, versions 0xb, 0x0, 0x76, 0x0, 0x0
10. Scorpio Jr (715/33) (0) at 0xfffbe000, versions 0x311, 0x0, 0x4, 0x0, 0x81
11. Scorpio Jr (1) at 0xfffbf000, versions 0x16, 0x0, 0x9, 0x0, 0x0

Signature · 4 lines
Matthew Wilcox <[email protected]>
"Windows and MacOS are products, contrived by engineers in the service of
specific companies. Unix, by contrast, is not so much a product as it is a
painstakingly compiled oral history of the hacker subculture." - N Stephenson
#1 lore raw
Quoted 5 lines

I've taken the hp300 HIL driver written by Phil Blundell and it `works' on my 715/33. It's producing the wrong keycodes, it's doing work that an ASP driver ought to be doing and it's producing a fair bit of extraneous crap on the screen. But I'm getting interrupts and keycodes so I'm happy for tonight :-)

Maybe this is due to the difference between keycodes for HP (e.g., C1429A) and PS/2 keyboards?

Signature · 2 lines
J. David Anglin                                  [email protected]
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)
#2 lore raw
Quoted 8 lines · John David Anglin
Quoted 5 lines

I've taken the hp300 HIL driver written by Phil Blundell and it `works' on my 715/33. It's producing the wrong keycodes, it's doing work that an ASP driver ought to be doing and it's producing a fair bit of extraneous crap on the screen. But I'm getting interrupts and keycodes so I'm happy for tonight :-)

Maybe this is due to the difference between keycodes for HP (e.g., C1429A) and PS/2 keyboards?

Sorry, I was rather unclear there. The HIL driver has its own built-in keymaps. What I'm seeing (for example, when I press the `g' key):

)
)
g
)
)

or sometimes

)
)
)
)
g

or some other permutation. I would prefer to not worry about this for the moment; there's lots of other things to tidy up and it'll probably sort itself out once something else is done properly.

I've now committed most of my changes. I'm _not_ committing this one as it'll definitely screw up every other type of machine:

ndex: irq.c
===================================================================
RCS file: /home/cvs/parisc/linux/arch/parisc/kernel/irq.c,v retrieving revision 1.36

11/10 00:05:28 renamed from 10/31 06:26:35 1.36 +7 -2
diff -u -r1.36 irq.c
@@ -225,12 +225,17 @@
                if(!(bit&mask))
                        continue;
 
+#if 1
+               if (irq == 3) {
+                       gsc_readl(0xf0800000);
+               }
+#else
                if(irq) {
-#if 0
                        printk("Interrupt %2d(%2d+%2d)\n", irq+region->data.irqbase, irq, region->data.irqbase);
+                       printk("ASP mask = %x", gsc_readl(0xf0800000));
 
-#endif
                }
+#endif
                mask_irq(irq);
                do_irq(&region->action[irq], region->data.irqbase + irq, regs);
                unmask_irq(irq);

(please excuse the copy & paste)

If you have a 715/33, you may wish to apply the patch and watch in amazement as your keyboard produces characters. Hopefully we'll thrash out how to get it working properly soon.

Signature · 4 lines
Matthew Wilcox <[email protected]>
"Windows and MacOS are products, contrived by engineers in the service of
specific companies. Unix, by contrast, is not so much a product as it is a
painstakingly compiled oral history of the hacker subculture." - N Stephenson
#3 lore raw
In reply to Matthew Wilcox
Quoted 5 lines

I've taken the hp300 HIL driver written by Phil Blundell and it `works' on my 715/33. It's producing the wrong keycodes, it's doing work that an ASP driver ought to be doing and it's producing a fair bit of extraneous crap on the screen. But I'm getting interrupts and keycodes so I'm happy for tonight :-)

congrats. The Lasi / PS/2 driver shouldn't be too difficult to write either except for one minor problems:

We do not know how the PS/2 keyboard (identical to AT keyboard I would hope) self-test procedure is. We do know that we can send a command to the keyboard to cause it to start a self-test, and we do know that the Intel keyboard con- troller has a separate bit in the status register indicating that the keyboard self-test succeeded.

What we can do (and the Intel keyboard controller cannot) is to manually control and read the PS/2 data and clock lines, so I suspect we'll have to make use of this to successfully do a keyboard self-test.

If you know of any documentation describing what happens during an AT keyboard self-test, please tell us.

Quoted 10 lines

My question is: how? Mach hardcodes the address of ASP to 0xf0800000, but there's nothing even close in the driver list. I notice in the

2. Scorpio Jr Core BA (11) at 0xf082f000, versions 0xb, 0x0, 0x70, 0x0, 0x0
3. Scorpio Jr Core SCSI (10) at 0xf0825000, versions 0xb, 0x0, 0x71, 0x0, 0x0
4. Scorpio Jr Core LAN (802.3) (10) at 0xf0826000, versions 0xb, 0x0, 0x72, 0x0, 0x0
5. Scorpio Jr Core HIL (10) at 0xf0821000, versions 0xb, 0x0, 0x73, 0x0, 0x0
6. Scorpio Jr Core RS-232 (10) at 0xf0823000, versions 0xb, 0x0, 0x75, 0x0, 0x0
7. Scorpio Jr Core RS-232 (10) at 0xf0822000, versions 0xb, 0x0, 0x75, 0x0, 0x0
8. Scorpio Jr Core Centronics (10) at 0xf0824000, versions 0xb, 0x0, 0x74, 0x0, 0x0

I am sorry, but when all this devices are within a 1 MB offset of the HPA we're looking for (0xf0800000 + 1 M = 0xf0900000, I hope), doesn't that look very nice indeed ? You are right that the interrupt management (which is at address 0xf0800000) is not specifically printed here, but I think we should be able to deduce it, shouldn't we ?

Philipp Rumpf

#4 lore raw
Quoted 11 lines · Philipp Rumpf
Quoted 10 lines

My question is: how? Mach hardcodes the address of ASP to 0xf0800000, but there's nothing even close in the driver list. I notice in the

2. Scorpio Jr Core BA (11) at 0xf082f000, versions 0xb, 0x0, 0x70, 0x0, 0x0
3. Scorpio Jr Core SCSI (10) at 0xf0825000, versions 0xb, 0x0, 0x71, 0x0, 0x0
4. Scorpio Jr Core LAN (802.3) (10) at 0xf0826000, versions 0xb, 0x0, 0x72, 0x0, 0x0
5. Scorpio Jr Core HIL (10) at 0xf0821000, versions 0xb, 0x0, 0x73, 0x0, 0x0
6. Scorpio Jr Core RS-232 (10) at 0xf0823000, versions 0xb, 0x0, 0x75, 0x0, 0x0
7. Scorpio Jr Core RS-232 (10) at 0xf0822000, versions 0xb, 0x0, 0x75, 0x0, 0x0
8. Scorpio Jr Core Centronics (10) at 0xf0824000, versions 0xb, 0x0, 0x74, 0x0, 0x0

Quoted 5 lines

I am sorry, but when all this devices are within a 1 MB offset of the HPA we're looking for (0xf0800000 + 1 M = 0xf0900000, I hope), doesn't that look very nice indeed ? You are right that the interrupt management (which is at address 0xf0800000) is not specifically printed here, but I think we should be able to deduce it, shouldn't we ?

Ummmm.. how exactly were you intending on going about this? In lasi.c, in lasi_init_chip(), round the address down to the nearest megabyte? (possibly with a check on the type of `lasi' we found?)

Signature · 4 lines
Matthew Wilcox <[email protected]>
"Windows and MacOS are products, contrived by engineers in the service of
specific companies. Unix, by contrast, is not so much a product as it is a
painstakingly compiled oral history of the hacker subculture." - N Stephenson
#5 lore raw
Quoted 3 lines · Matthew Wilcox

Ummmm.. how exactly were you intending on going about this? In lasi.c, in lasi_init_chip(), round the address down to the nearest megabyte? (possibly with a check on the type of `lasi' we found?)

We should probably have asp_init_chip() which gets triggered only for ASP components. Then you can adjust thhpa and go on with initialization.

- Alex

#6 lore raw
In reply to Philipp Rumpf
Quoted 9 lines · Philipp Rumpf

We do not know how the PS/2 keyboard (identical to AT keyboard I would hope) self-test procedure is. We do know that we can send a command to the keyboard to cause it to start a self-test, and we do know that the Intel keyboard con- troller has a separate bit in the status register indicating that the keyboard self-test succeeded.

What we can do (and the Intel keyboard controller cannot) is to manually control and read the PS/2 data and clock lines, so I suspect we'll have to make use of this to successfully do a keyboard self-test.

For the time being we can just ignore this test. I'm not entirely sure that there's an easy way to actually test it. Right now I coded it to test the controller, but not the keyboard.

Quoted 2 lines

If you know of any documentation describing what happens during an AT keyboard self-test, please tell us.

Note, this is NOT the 8042. It has it's own test sequence that the PS/2 keyboard itself does not handle. The next person to point me to 8042 docs will receive one (1) boot to the head.

- Alex

#7 lore raw
In reply to Matthew Wilcox
Quoted 5 lines · Matthew Wilcox

I've taken the hp300 HIL driver written by Phil Blundell and it `works' on my 715/33. It's producing the wrong keycodes, it's doing work that an ASP driver ought to be doing and it's producing a fair bit of extraneous crap on the screen. But I'm getting interrupts and keycodes so I'm happy for tonight :-)

Yay!

Having ASP interrupts working will enable a lot of different boxes to be supported, mostly 715s. I have a /50, /75 and probably /100 to test on.

The closest things we have to ASP docs are:
- take the portions of the Lasi docs that describe how Lasi is not like ASP, and read backwards
- openbsd
- mklinux

Quoted 2 lines

The next obvious step is to have an ASP manager, like LASI. In fact, so much like LASI, it probably ought to be in the same driver.

My best guess is that Lasi is really ASP but integrated onto one cell with a couple of bug fixes. So yes, it really should be the same driver if at all possible. I had actually started on this but hadn't had time to get it working. :)

Quoted 7 lines

This seems to
be how Mach does it, and there's a rumour this is also how HPUX does it. My question is: how? Mach hardcodes the address of ASP to 0xf0800000, but there's nothing even close in the driver list. I notice in the bootlogs which have previously been posted, LASI was at the address of `Core BA', but you can see that's wildly different on this machine. Should I invent my own and insert it into the structures we have?

Hrm. That's uncomfortable. So did it actually work at 0xf0800000?

The problem comes when we hit an ASP that's not at 0xf0800000, which I'd think might come up with ASP add on cards.

I don't think that firmware is lying, though. Perhaps we just need to subtract 0x2f000 for some reason.

Quoted 2 lines

Here's the driver list printed at boot on the 715 (is someone collecting these somewhere?)

Within a week, I'll send some mail about The Puffin PA RISC Hardware Guide. Details to come.

- Alex

#8 lore raw
Quoted 3 lines · Alex deVries

Having ASP interrupts working will enable a lot of different boxes to be supported, mostly 715s. I have a /50, /75 and probably /100 to test on.

The /50 and /75 are also Scorpio machines. The /100 is either a Mirage or a Turnip. As such, I don't know if it has ASP.

Quoted 3 lines

The closest things we have to ASP docs are:
- take the portions of the Lasi docs that describe how Lasi is not like ASP, and read backwards

umm.. 4MB download this evening then...

Quoted 1 line

Hrm. That's uncomfortable. So did it actually work at 0xf0800000?

Yep.

Quoted 2 lines

The problem comes when we hit an ASP that's not at 0xf0800000, which I'd think might come up with ASP add on cards.

Do ASP add-in cards exist? We could just refuse to support them until we know more :-)

Quoted 2 lines

I don't think that firmware is lying, though. Perhaps we just need to subtract 0x2f000 for some reason.

Found out what this is in the mach sources. It's called CORE_ASP:

/*
* Structures defining memory-mapped registers in the Core IOSS (ASP). */

        vu_char reset;          /* (WO) IOSS reset */
        vu_char version;        /* (RO) version byte */
        vu_char scsi_dsync;     /* (WO) SCSI DSYNC enable bit (1:enable) */
        vu_char error;          /* bus errors while ASP masters SGC/VSC */
        vu_char lan_oen;        /* (WO) output enable Ethernet/FDDI */

Shortly below this is:

/*
* Structures defining memory-mapped registers in the Core IOSS (LASI). */

struct core_lasi {
        vu_int  power;
        vu_int  error;
        vu_int  version;
        vu_int  reset;
        vu_int  arbmask;
};

which is defined to be at 0xF010C000.

Quoted 5 lines
Quoted 2 lines

Here's the driver list printed at boot on the 715 (is someone collecting these somewhere?)

Within a week, I'll send some mail about The Puffin PA RISC Hardware Guide. Details to come.

Oh good.

Signature · 4 lines
Matthew Wilcox <[email protected]>
"Windows and MacOS are products, contrived by engineers in the service of
specific companies. Unix, by contrast, is not so much a product as it is a
painstakingly compiled oral history of the hacker subculture." - N Stephenson
#9 lore raw
Quoted 7 lines · Matthew Wilcox
Quoted 3 lines · Alex deVries

Having ASP interrupts working will enable a lot of different boxes to be supported, mostly 715s. I have a /50, /75 and probably /100 to test on.

The /50 and /75 are also Scorpio machines. The /100 is either a Mirage or a Turnip. As such, I don't know if it has ASP.

Turnip is the 715/100XC. Has 1MB I/D cache vs the 128KB or 256KB that the regular 715/100. Alex is getting a 715/100 to test with and he can then tell you if it has ASP or LASI ( I think the latter but don't know for sure).

grant

Grant Grundler
Unix Developement Lab
+1.408.447.7253

#10 lore raw
In reply to Matthew Wilcox
Quoted 7 lines · Matthew Wilcox
Quoted 3 lines · Alex deVries

Having ASP interrupts working will enable a lot of different boxes to be supported, mostly 715s. I have a /50, /75 and probably /100 to test on.

The /50 and /75 are also Scorpio machines. The /100 is either a Mirage or a Turnip. As such, I don't know if it has ASP.

I'm pretty sure that all 715s have ASP in them.

This is easily checked though...

- alex