RE: [PATCH] 2.4.x: update for PCI vendor id 0x12d4

RE: [PATCH] 2.4.x: update for PCI vendor id 0x12d4

[Khachaturov, Vassilii]

[Updated patch is in the end of this mail]
Thanks, Mark!

> submit patches as text in your message, since people want 
> to read them, and not waste time saving to a file, etc.
> also, explain patches: who is vid 0x12d4, how did you get 
> the information, what does it effect, etc.

BTW I noticed a funny thing: the file devlist.h I tried to 
patch doesn't always exist - as it gets built from the file 
pci.ids in the same directory. Noone complained on that :)
What I don't understand is why pci_ids.h doesn't get
generated from pci.ids as well.

So, here's the new patch, sent along also to the pci.ids 
maintainer, along with the required info.

12d4 was DGM&S. Then DGM&S was aquired by Comverse, and 
later the corresponding activities (and the same PCI boards 
manufacturing) went to Comverse spinoff Ulticom. So, in fact,
Ulticom is just DGM&S under a different name and controlled
by Comverse. I guess the PCI vendor registry should get 
updated at some point, but I thought it could be a great 
thing if Linux knew it ahead. (Currently reported by the 
PCI driver "DGM&S" is obsolete - such entity doesn't exist 
any more). Naturally, I have got this info being a 
Comverse employee.

This patch only has effect on kernel PCI driver reporting 
on the 12d4 vendor ID. There is no Linux kernel code which
supports the corresponding SS7 signalling boards.

Kind regards,
	Vassilii

--- /usr/src/linux-2.4.5/include/linux/pci_ids.h	Wed May 16 13:25:39
2001
+++ pci_ids.h	Wed May 30 14:55:01 2001
@@ -1199,6 +1199,8 @@
 #define PCI_VENDOR_ID_NVIDIA_SGS	0x12d2
 #define PCI_DEVICE_ID_NVIDIA_SGS_RIVA128 0x0018
 
+#define PCI_VENDOR_ID_ULTICOM	0x12d4 /* Formerly DGM&S */
+
 #define PCI_SUBVENDOR_ID_CHASE_PCIFAST		0x12E0
 #define PCI_SUBDEVICE_ID_CHASE_PCIFAST4		0x0031
 #define PCI_SUBDEVICE_ID_CHASE_PCIFAST8		0x0021
--- /usr/src/linux-2.4.5/drivers/pci/pci.ids	Sat May 19 20:49:14 2001
+++ pci.ids	Wed May 30 14:54:50 2001
@@ -3204,7 +3204,7 @@
 	002c  VTNT2
 	00a0  ITNT2
 12d3  Vingmed Sound A/S
-12d4  DGM&S
+12d4  Ulticom (Formerly DGM&S)
 12d5  Equator Technologies
 12d6  Analogic Corp
 12d7  Biotronic SRL

Re: [PATCH] 2.4.x: update for PCI vendor id 0x12d4

[Keith Owens]

On Wed, 30 May 2001 15:34:17 -0400, 
"Khachaturov, Vassilii" <Vassilii.Khachaturov@comverse.com> wrote:
>BTW I noticed a funny thing: the file devlist.h I tried to 
>patch doesn't always exist - as it gets built from the file 
>pci.ids in the same directory. Noone complained on that :)

drivers/pci/ devlist.h, classlist.h and gen-devlist are the generated
files.  Add them to your "don't diff" list when generating patches.

>What I don't understand is why pci_ids.h doesn't get
>generated from pci.ids as well.

Leave that question for Martin.

Re: [PATCH] 2.4.x: update for PCI vendor id 0x12d4

[Jeff Garzik]

> "Khachaturov, Vassilii" <Vassilii.Khachaturov@comverse.com> wrote:
> >What I don't understand is why pci_ids.h doesn't get
> >generated from pci.ids as well.

Although they are similar, you cannot reliably generate useable C
constants from english words in the PCI id list...

-- 
Jeff Garzik      | Disbelief, that's why you fail.
Building 1024    |
MandrakeSoft     |

RE: [PATCH] 2.4.x: update for PCI vendor id 0x12d4

[Khachaturov, Vassilii]

> From: Jeff Garzik [mailto:jgarzik@mandrakesoft.com]
> 
> > "Khachaturov, Vassilii" <Vassilii.Khachaturov@comverse.com> wrote:
> > >What I don't understand is why pci_ids.h doesn't get
> > >generated from pci.ids as well.
> 
> Although they are similar, you cannot reliably generate useable C
> constants from english words in the PCI id list...

I agree. Still, right now the update of the 2 is independant. Do you think
that 
merging the C constant names (with the common prefix stripped)
into the PCI id list so that there's a single source to control it would be
better? 

It's also possible to apply an heuristic to only store "exception" C names 
there - just take the 1st word and make it uppercase:

pci.ids:
1043  Asustek Computer, Inc.

would automatically map into what it is now:
pci_ids.h
#define PCI_VENDOR_ID_ASUSTEK		0x1043

whereas for 0x1042 one would have to change in the pci.ids: 
1042  Micron
	1000  FDC 37C665
	1001  37C922
	3000  Samurai_0
	3010  Samurai_1
	3020  Samurai_IDE

into
1042{PCTECH}  Micron
	1000{RZ1000}  FDC 37C665
	1001{RZ1001}  37C922
	3000  Samurai_0
	3010  Samurai_1
	3020  Samurai_IDE

to generate in the .h
#define PCI_VENDOR_ID_PCTECH		0x1042
#define PCI_DEVICE_ID_PCTECH_RZ1000	0x1000
#define PCI_DEVICE_ID_PCTECH_RZ1001	0x1001
#define PCI_DEVICE_ID_PCTECH_SAMURAI_0	0x3000
#define PCI_DEVICE_ID_PCTECH_SAMURAI_1	0x3010
#define PCI_DEVICE_ID_PCTECH_SAMURAI_IDE 0x3020

You get the idea... This way it is also very easy to look up the .ids file
and see instantly where the names mismatch severely.

Does this make sense? If you think it does (say so), I can easily make this
into the tree of your choice in a backward-compatible way. There'll be an
option to generate the older .ids, too, for easier merge back into the 2.2
series.

Vassilii
P.S. Even if you don't think the merge of the 2 files is worth it, don't
forget to apply the 12d4 patch...	

[PATCH] 2.4.x: update for PCI vendor id 0x12d4

[Khachaturov, Vassilii]

[-- Attachment #1: Type: text/plain, Size: 209 bytes --]

This is my 1st attempt to submit a patch here - flames welcome if I did smth
wrong...
It was done across 2.4.2, but it works safely against 2.4.4 as well.

 <<pci_vendor_12d4.patch>> 

Kind regards,
	Vassilii

[-- Attachment #2: pci_vendor_12d4.patch --]
[-- Type: application/octet-stream, Size: 753 bytes --]

--- /usr/src/linux-2.4.2/include/linux/pci_ids.h	Sat Feb  3 15:48:00 2001
+++ pci_ids.h	Fri May 25 13:15:51 2001
@@ -1110,6 +1110,8 @@
 #define PCI_VENDOR_ID_NVIDIA_SGS	0x12d2
 #define PCI_DEVICE_ID_NVIDIA_SGS_RIVA128 0x0018
 
+#define PCI_VENDOR_ID_ULTICOM	0x12d4 /* Formerly DGM&S */
+
 #define PCI_SUBVENDOR_ID_CHASE_PCIFAST		0x12E0
 #define PCI_SUBDEVICE_ID_CHASE_PCIFAST4		0x0031
 #define PCI_SUBDEVICE_ID_CHASE_PCIFAST8		0x0021
--- /usr/src/linux-2.4.2/drivers/pci/devlist.h	Tue Mar 27 16:10:22 2001
+++ devlist.h	Fri May 25 13:35:50 2001
@@ -3733,7 +3733,7 @@
 VENDOR(12d3,"Vingmed Sound A/S")
 ENDVENDOR()
 
-VENDOR(12d4,"DGM&S")
+VENDOR(12d4,"Ulticom (formerly DGM&S)")
 ENDVENDOR()
 
 VENDOR(12d5,"Equator Technologies")