[1647] in SIPB_Linux_Development
Re: Ted's patch to fix cache corruption
daemon@ATHENA.MIT.EDU (Derek Atkins)
Mon Apr 7 09:39:50 1997
To: amu@MIT.EDU (Aaron M. Ucko)
Cc: Erik Nygren <nygren@MIT.EDU>, linux-dev@MIT.EDU, linux-afs-bugs@MIT.EDU
From: Derek Atkins <warlord@MIT.EDU>
Date: 07 Apr 1997 09:39:27 -0400
In-Reply-To: amu@MIT.EDU's message of 06 Apr 1997 13:47:05 -0400
AFS is already writing out to the cache when the cache-write fails...
This could either be from a callback or from a user write. In either
case, it is already too late to report back to the user -- cache
writes are done asyncronously. There is NO WAY to stop corruption
from happening.
All I could do is remember that there was an error and warn the user
that the cache partition filled. But there is no way to stop
corruption from happening.
-derek
amu@MIT.EDU (Aaron M. Ucko) writes:
>
> Derek Atkins <warlord@MIT.EDU> writes:
>
> > It's just AFS's design to fail... It expects the space you tell it it
> > has to actually be there. By the time it finds out there isn't any
> > more space, its too late for the kernel to do anything. I guess I
> > could have the kernel panic in this case instead of corrupting files,
> > but I'm not convinced that would be a better solution.
>
> Here's an idea: instead of panicking or corrupting files, the AFS code
> should print an error message and set a flag causing all AFS
> operations to fail cleanly. Then nothing is corrupted and people can
> save their buffers to some other filesystem.
>
> --
> Aaron M. Ucko (amu@mit.edu) | For Geek Code, PGP public key, and other info,
> finger amu@monk.mit.edu. | "Kids! Bringing about Armageddon can be dangerous.
> Do not attempt it in your home." -- T. Pratchett & N. Gaiman, _Good Omens_
--
Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
Member, MIT Student Information Processing Board (SIPB)
URL: http://web.mit.edu/warlord/ PP-ASEL N1NWH
warlord@MIT.EDU PGP key available
