From: Miklos S. <mi...@sz...> - 2007-10-15 22:00:47
|
> > This is actually a bug in the fuse kernel module. Here's a patch > > against 2.6.23-mm1. > > Wow! Is there any chance of getting this fixed in the fuse kernel > module itself at some point in the future? Sure, I'll submit this patch for 2.6.24. > > I guess you don't really want to compile a new kernel for this, > > Would not I just have to recompile the fuse kernel module? (fuse is a > module on Kubuntu 7.04.) > > Or is there no patch for Linux 2.6.20? The patch probably doesn't apply to anything other than 2.6.23-mm1, due to other pending changes to the fuse module. > > so there are some workarounds as well: > > > - in a shell on the client do > > while true; do stat file_being_read > /dev/null; sleep 10; done > > Yeah - I thought about something like that. > > > - start sshfs with the '-odirect_io' option > > "-odirect_io" sounds like the best choice for my needs. What is the > downside (if any)? I'm not sure what exactly "direct_io" means in > this context. It means, that read() and write() calls go directly to the filesystem, bypassing the cache in the kernel. It has some side effects, like prevenging mmap() from working, but that is only used by very few applications. Miklos |