From: Roman S. <rv...@su...> - 2009-01-14 03:37:47
|
On Jan 12, 2009, at 3:01 AM, Miklos Szeredi wrote: > On Fri, 09 Jan 2009, Roman V. Shaposhnik wrote: >> True. Btw, is it possible to asynchronously notify in-kernel FUSE >> that a file has changed and the cache has to be invalidated? Suppose >> my user-level app gets notified asynchronously too. How am I supposed >> to pass these events up the chain? > > Not yet, but such an API is planned. Interesting. Are there any public discussions of what it might look like? >> This will work for me, except I don't really know how to implement >> #3. >> It looks like fuse_write_out only takes size. > > Right, fuse_write_out needs to be extended with an "__u64 new_offset" > field. Then the kernel needs to check if the new offset matches the > expected value and if not then invalidate the cache. > > Due to the size change of 'struct fuse_write_out' there needs to be > some compatibility tricks (similar to those for statfs). Good to know we are on the page on this one. Now for the shameless question ;-) Is it something that you would be interested in implementing in the short term, or is the only way to see this functionality in FUSE is a way of "the patch"? Thanks, Roman. |