Greg KH wrote:
> On Mon, Jan 25, 2010 at 09:57:45AM -0800, Gary Hade wrote:
>
>> On Mon, Jan 25, 2010 at 09:52:37AM +0100, Andreas Besse wrote:
>>
>>> Andreas Besse wrote:
>>>
>>>> Greg KH wrote:
>>>>
>>>>
>>>>> On Fri, Jan 22, 2010 at 06:34:51PM +0100, Andreas Besse wrote:
>>>>>
>>>>>
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> I observed that the PCI Bus id's of the onboard network cards changed
>>>>>> after I added a PCIe Dual DVB-Card to my system.
>>>>>>
>>>>>> The ID's changed from 02:00.0, 03:00.0 (see lspci_with_dvb.txt) to
>>>>>> 03:00.0, 04:00.0 (see lspci_with_dvb.txt)
>>>>>>
>>>>>> How is the ordering of PCI/PCIe devices determined in generally by the
>>>>>> linux PCI subsystem?
>>>>>>
>>>>>>
>>>>>>
>>>>> Totally random, depending on the phase of the moon, and the location of
>>>>> your computer in relation to the nearest large body of water.
>>>>>
>>>>> Seriously, this is expected, you can not rely on pci bus ids to remain
>>>>> constant, especially if you have changed the physical system (like
>>>>> adding a card which you did.)
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>> is the Bus ID for a specific PCI/PCIe Slot deterministic at every boot
>>>>>> if no hotplug (removing or adding PCI/PCie cards) is involved?
>>>>>>
>>>>>> can I assume that the Bus ID's don't change if I don't add or remove any
>>>>>> PCI/PCie cards?
>>>>>>
>>>>>>
>>>>>>
>>>>> Nope, I have a machine here that will reorder things every 4th boot or
>>>>> so. It's a "feature" of the BIOS according to the bug I filed with the
>>>>> manufacturer that they closed out as invalid.
>>>>>
>>>>>
>>>>>
>>>> do you have a link to the bug report?
>>>>
>>>>
>>>>> So never rely on these ids, they can not be guaranteed to be stable, and
>>>>> will not be.
>>>>>
>>>>>
>>>>>
>>>> If PCI bus IDs are not usable, what other mechanisms do exist for
>>>> knowing which particular physical hardware device (plugged into a
>>>> PCI/PCIe slot) is assigned to which particular Linux device ?
>>>>
>>>> I want to identify a DVB S2-Card which is plugged into a specific
>>>> PCIe-Slot by a udev rule and assigns a specific linux device (e.g.
>>>> /dev/dvb/adapter0).
>>>>
>>>>
>>> sorry my statement is a bit ambiguous. For a single card there is no
>>> issue, but for multiple cards of the same type.
>>>
>>> If multiple DVB cards of the same type, but with different
>>> signals/cables are used, the cards need to identified.
>>>
>>>> Does the behavior you described with your board exist with all boards?
>>>>
>>>> If yes, I wonder how Linux can at all be used for any reliable hardware appliance, which includes two identical PCI boards, which need to be distinguished due to different signals/cables being connected to each board.
>>>>
>> Andreas, Would the physical slot number to PCI ID mapping
>> provided in /sys/bus/pci/slots by the 'pci_slot' driver possibly
>> be of any use to you?
>>
>
> That mapping should be already there if it is provided by the BIOS.
>
> Also, Andreas, I see a symlink for all v4l devices already created in a
> persistant manner on the more recent udev releases. So you could just
> use that instead of the "raw" device node, and all should be good.
>
the v4l devices are created by the following udev rule in
/etc/udev/rules.d/50-udev-default.rules:
#default rule
SUBSYSTEM=="dvb", PROGRAM="/bin/sh -c 'K=%k; K=$${K#dvb}; printf dvb/adapter%%i/%%s $${K%%%%.*} $${K#*.}'", NAME="%c"
if i understand udev correctly, udev only uses information on top of /sys.
If there the information about the PCI slots is not deterministic after boot how can udev create the devices in the right order?
regards,
Andreas Besse
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
