From: Steve S. <ste...@ho...> - 2002-11-13 06:57:13
|
<jd...@ka...> wrote: >ste...@ho... said: > > Attached is a patch that implements this feature, with the syntax > > described above. [...] >Other than these points, this looks basically right. I hope this looks even more right. :) Attached is my second attempt at the filemap feature patch, based on 2.4.19-29um. I broke it down into two parts: Patch #1 just contains cleanups to UML that try to ensure system functions necessary to filemap go through the os/file.c interface. I'm know this isn't your first choice, but until there is a better interface, I think these changes are benign and should help porting UML, and coding that interface, regardless of whether filemap ever makes it in. Patch #2 contains the actual filemap feature code. Note that hostfs does not go through the os/file.c interface, and thus does not go through filemap. This is intentional. This is also the best example I can think of why there should be an io interface on top of the os/file that filemap gets hooked into, and hostfs bypasses directly to os/file. Please note that this patch may cause problems and loss of data. Use at your own risk - I am. :) Steve Schmidtke _________________________________________________________________ Help STOP SPAM with the new MSN 8 and get 2 months FREE* http://join.msn.com/?page=features/junkmail |