On Mon, Jan 18, 2010 at 12:58 PM, Andreas Besse <besse@xxxxxxxxxx> wrote: > Manu Abraham wrote: >> On Sat, Jan 16, 2010 at 3:00 AM, Oliver Endriss <o.endriss@xxxxxx> wrote: >> >>> Devin Heitmueller wrote: >>> >>>> On Thu, Jan 14, 2010 at 11:01 AM, Andreas Besse <besse@xxxxxxxxxx> wrote: >>>> >>>>> yes if there are different drivers I already observed the behaviour that >>>>> the ordering gets flipped after reboot. >>>>> >>>>> But if I assume, that there is only *one* driver that is loaded (e.g. >>>>> budget_av) for all dvb cards in the system, how is the ordering of these >>>>> devices determined? How does the driver "search" for available dvb cards? >>>>> >>> The driver does not 'search' for a card. The driver registers the ids of >>> all supported cards with the pci subsystem of the kernel. >>> >>> When the pci subsystem detects a new card, it calls the 'probe' routine >>> of the driver (for example saa7146_init_one for saa7146-based cards). >>> So the ordering is determined by the pci subsystem. >>> >>> >>>> I believe your assumption is incorrect. I believe the enumeration >>>> order is not deterministic even for multiple instances of the same >>>> driver. It is not uncommon to hear mythtv users complain that "I have >>>> two PVR-150 cards installed in my PC and the order sometimes get >>>> reversed on reboot". >>>> >>> Afaik the indeterministic behaviour is caused by udev, not by the >>> kernel. We never had these problems before udev was introduced. >>> >> >> >> True, the ordering is not exactly the same everytime. One will need to >> provide PCI Bus related info also to a practical udev configuration to >> get things sorted out in a sane way, rather than anything else. >> > with "PCI Bus related info" you mean the KERNELS parameter which is > reported by udevinfo? > > udevinfo -a -p $(udevinfo -q path -n /dev/dvb/adapter0/frontend0) > [...] > looking at parent device '/devices/pci0000:00/0000:00:1e.0/0000:08:00.0': > KERNELS=="0000:08:00.0" > SUBSYSTEMS=="pci" > > does this KERNELS parameter always match the Slot-Id of "lspci -vmm" ? > Slot: 08:00.0 > Class: Multimedia controller > Vendor: Philips Semiconductors > Device: SAA7146 > SVendor: Technotrend Systemtechnik GmbH > SDevice: S2-3200 > Rev: 01 > > is it right that the Slot-Id is deterministic for PCI/PCIe based systems? Slot can also change, since slots are behind a specific bridge which could be susceptible to events such as hotplug. Also things such as PCI reordering and things like that tend to muck up things even more.Things such as DVB_ADAPTER number are also pointless and useless. You can see an example how to handle it in a bit practical manner: http://www.wlug.org.nz/UDev Regards, Manu -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
