On Saturday 2015-11-07 14:01, Andrei Borzenkov wrote:
You added additional PCI bridge(s) which caused all subsequent numbers to be shifted. [...] I suspect the answer from upstream would be "we do not guarantee persistence if hardware configuration changes".
Mentioned it during #systemconf2015. Answer was that it is dependent on what the firmware does.
That response I am interpreting such that: the firmware ought to not hide PCI bridges even when no daughterboards are connected. Or do
But there are *no* PCI bridges *until* daughterboards are connected,
Indeed I cannot think of a suitable "cure" for bridges added by daughterboards. But I was concerned with a case where bridges preexisted, such as this one: -[0000:00]-+-00.0 Intel Corporation E7520 Memory Controller Hub +-02.0-[01-03]--+-00.0-[02]--+-08.0 LSI Logic / Symbios Logic 53c1030 PCI-X Fusion-MPT Dual Ultra320 SCSI | | \-08.1 LSI Logic / Symbios Logic 53c1030 PCI-X Fusion-MPT Dual Ultra320 SCSI | \-00.2-[03]-- +-04.0-[04]----00.0 Broadcom Corporation NetXtreme BCM5721 Gigabit Ethernet PCI Express +-05.0-[05]----00.0 Broadcom Corporation NetXtreme BCM5721 Gigabit Ethernet PCI Express +-06.0-[06-08]--+-00.0-[07]-- | \-00.2-[08]-- +-1d.0 Intel Corporation 82801EB/ER (ICH5/ICH5R) USB UHCI Controller #1 I theorized that, on some hardware, PCI bridges like these (p7 and p8 here) might not show up at all until a device is actually placed in them, in which case the blame would probably be on the firmware.
in this case it is plain bug because upstream promises persistent names when hardware is added or removed. That is simply not possible using current scheme (and I argue that it is even theoretically not possible to stuff persistent path into device name because this path could be arbitrary long).
Idea. We could use a hash function to compress the PCI path identifier. We do it all the time for file contents... -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org