On Thu, Jan 16, 2025 at 5:41 PM Jingbo Xu <jefflexu@xxxxxxxxxxxxxxxxx> wrote: > > > > On 1/16/25 8:11 PM, Miklos Szeredi wrote: > > On Wed, 15 Jan 2025 at 20:41, Joanne Koong <joannelkoong@xxxxxxxxx> wrote: > > > >> Miklos, is this patchset acceptable for your tree? > > > > Looks good generally. > > > > I wonder why you chose to use a mount option instead of an FUSE_INIT param? > > IMO this timeout mechanism has no dependence on the implementation on > the server side, and it's self-contained by the kernel side. Thus it's > adequate to negotiate through the mount option instead of the INIT > feature negotiation. Although the FUSE mount instance is generally > mounted by the fuse server itself in the libfuse implementation. IOW > INIT feature negotiation is required only when the server side shall opt in. My understanding (maybe wrong) is that fuse servers only get mounted through the libfuse interface (eg fuse_session_mount()) which is directly called from the server code, so with either the mount option flag or INIT, both would require server-side changes. Am I wrong in this understanding? Is there a way to mount it through the command line? I know the fusermount3 utility exists, but AFAIU, that only supports unmounting. Thanks, Joanne > > -- > Thanks, > Jingbo
