On Sat, Jun 13, 2009 at 10:45 AM, Albrecht Dreß<albrecht.dr...@arcor.de> wrote: > Am 11.06.09 18:27 schrieb(en) Grant Likely: >>> >>> + *(u16 *)buf = *((volatile u16 *)(vdest - 1)); >>> + buf[1] = *((u8 *)src); >>> + *((volatile u16 *)(vdest - 1)) = *(u16 *)buf; >> >> what is the purpose of volatile here? If you need io barriers, then use >> the in_/out_be16 macros. > > Yes, you're right - should be completely superfluous here. A result of copy > & paste without thinking... :-( > >> Blech. ugly #ifdef and not really multiplatform safe (yeah, I know it >> shouldn't break non-5200 platforms, but it does have an undesirable impact). >> There's got to be a better way. > > Ouch, yes - I completely forgot the possibility of multi-platform builds. > > Am 11.06.09 19:28 schrieb(en) Grant Likely: >> >> So; the solution to me seems to be on an MPC5200 platform replace the >> offending hooks with MPC5200 specific variants at runtime. > > Will re-work the patch that way! BTW, a dumb question: what is the proper > way to determine which cpu the system is running on? Check the CPU node of > the of tree?
In this case I'd walk up the parent nodes in the tree looking for an "fsl,mpc5200-lpb" compatible value. That way you only apply the workaround if the device is on the locaplus bus, and not for example below the PCI bus. g. -- Grant Likely, B.Sc., P.Eng. Secret Lab Technologies Ltd. _______________________________________________ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
