Re: [draft] Automated choose of Windows drivers for PCI devices

[Pierre B. <pie...@ar...>]

Nils Olav Fossum a écrit :
> Onsdag 24 september 2008 18:44, skrev Pierre Bourgin:
> 
>>   So I would have just to search in .INF Windows files which mathc the
>> string "PCI\VEN_14e4&DEV_1677" in the whole (uncompressed) DriverPacks
>> collection for instance.
> 
>> - really quick: compute take 10s to parse the whole driverPacks.net driver
>>   collection
> 
> Make a index/reference file of the archive and parse that?
> There was some talk about making such a index on the driverpacks forum a while 
> back for inclusion in the "distribution" But I dont think it ever was made. 
> Probably because there was no clear use for it.
> 
> So, Im thinking: 
>   Check for index-file 
>   If its not there (or outdated) -- create it and save to c:\netinst
> 
> The administrator can then later move the new index-file to the archive.
> 
> This will give near instant driver installation (well, for the next computer).
> Even the more traditional drivers in $oem$\xx take quite a log time to copy 
> over when the installed computer-base has grown for a while.

when you talk about database, you think about a cache system in fact, 
now ? It can come later, just because we must take of that several 
drivers (of DriverPacks.net) can match a particular PCI_DEV,PCI_SUB.

To answer to the rest, I wrote this in a previous post (quoting myself, 
sorry):
"""
next step will be to develop the counterpart of this
mechanism under Windows:
- find PCI device(s) that are not handled by Windows drivers,
- parse unattended windows driver database to find drivers that match
then hints the list of folders to Windows driver installer ?
"""

>> Perhaps I've missed an important point that make it a nonsense ?
> 
> makes sense to me :-)

thanks ;-)

> We also have dmidecode in the linux-boot and its output can be used to 
> override the "auto-driverpacks-install" if needed or desired.

dmidecode output is difficult to parse, no ? Contrary to usgin 
/sys/bus/pci that provides the _exact_ PCI identifiers of device(s), 
that is a really valuable information.

I'm currently try to implement my draft, we will see if pratically it's 
still usable.
Currently, my primary goal is to provide the Ethernet and Video card 
driver and the chipset driver if possible.
If other devices have no driver, this is not a problem (at this time).

Pierre Bourgin
-- 
ARTERIS                          http://www.arteris.com
The Network-on-Chip Company