Messages in this thread | | | Date | Fri, 9 Jan 2009 12:07:30 +0000 | From | Alan Cox <> | Subject | Re: [PATCH] libata: use WARN_ON_ONCE on hot paths |
| |
> If I had to guess, I would say that Arjan's patches assume it is OK to > treat two ports on a single IDE controller as completely independent, > when that is not really the reality of the hardware.
The PCI probe paths and the setup paths of several of the PATA controller drivers know they are single threaded in a few spots so it wouldn't suprise me in the slightest. Really this async stuff needs to spend 2 or 3 months in -next getting the kinks worked out so that driver assumptions can be pinned down and fixed, and someone can give it an API with names that anyone but Arjan understands ;)
Alan
| |