Re: [PATCH v6 1/6] fuse: decouple fuse_ring creation from ent registration

On 7/22/26 14:00, Miklos Szeredi wrote:
> On Wed, 22 Jul 2026 at 01:10, Joanne Koong <[email protected]> wrote:
> 
>> We can fix up libfuse [1] to no longer send REGISTER before the
>> connection is initialized, so I don't think a kernel fix is strictly
>> needed here, but if you still prefer kernel code to coordinate this, I
>> can send that out.
> 
> With async INIT this is possible by waiting until the write(2) syscall
> for the INIT reply returns.  This is because the initialization is
> performed within the context of the write call.
> 
> For sync INIT the initialization is performed in the context of the
> mount/FSCONFIG_CMD_CREATE, libfuse can wait until the respective
> syscall returns.
> 
> So yes, it's possible to ensure this from userspace, no kernel change is need.


The patch here from Joanne needs a slight update:
https://github.com/libfuse/libfuse/pull/1561

I can amend myself next week (once I'm through with review here).

As I wrote there, I had run last into -EAGAIN last year and had added an
-EAGAIN handler. Pity that I didn't record in the commit message the
exact reason :/


Thanks,
Bernd