[stuff deleted]
> As two processes - the first mmaps a data area for read then writes it to
> another file. This takes page faults during the write (paging in the memory
> we mmap'd). The second process does an ftruncate which revokes the memory
> from under the first process mmap during a page fault: What you get therefore
> is ext2fs suddenely trying to write from non existant pages and "Oops..."
> messages galore as well as a mess that e2fsck moans about a fair bit
> but seems to fix.
Has this been fixed in .77 or is the mmap change Linus mentioned
something different ? (Linus: "- Oops. We didn't get exclusive file access
for mmap page writes")
Thank you,
--Craig