On Sun, Jul 26, 2015 at 3:03 AM, Helg <[email protected]> wrote: > Hi, > > Fuse does not handle the situation gracefully if a file system does not > respond in time while servicing a system call. Fuse should block indefinitely > until the call responds and the patch below does exactly this.
I do not agree with you on this point. With your patch when a fuse vnop is issued, if the client does not respond, the fuse client process gets stuck indefinitely....To reproduce the stuck try to "ls -R" in a big fuse tree and kill the fuse client at the same time... IMHO the code right now is good, the fuse client should always respond to fusebufs. if not, something is wrong so we close the fuse device.
