Bill C. Riemers wrote:
> Does this problem also effect the 2.4 kernels?
> I'm just wondering, because it sounds like a possible explination for
> strange errors occurring with /dev/cobd? in coLinux, especially when used as
> swap devices. Basically /dev/cobd? are simmular to ramdisks, but they
> refere to files mmap'ed under Windows. They work fairly reliable, until
> someone does something like "swapoff -a;swapon -a". However, coLinux is
> only used with 2.4 kernels, so if the problem does not effect 2.4 kernels,
> then this is not the cause.
Yes, Matt Porter previously determined that this behavior changed
sometime after 2.4.2 and on or before 2.4.17, when rd.c was modified to
perform a truncate_inode_pages() for BLKFLSBUF (could track down the
date/revision more accurately if you're interested).