Re: [PATCH 5/5 UPDATED] CUSE: implement CUSE - Character device in Userspace

[Tejun Heo <tj@kernel.org>]

Hello,

Miklos Szeredi wrote:
> Hi Tejun,
> 
> Apologies for the late review.

:-)

> On Sat, 22 Nov 2008, Tejun Heo wrote:
>> + * cuse_conn	: contains fuse_conn and serves as bonding structure
>> + * channel	: file handle connected to the userland CUSE server
>> + * cdev		: the implemented character device
>> + * mnt		: vfsmount which serves dentry and inode for cdev
> 
> Hmm, this mount thing seems rather pointless to me.  I understand the
> motivation: to let the unmodified fuse functions use
> get_fuse_conn(inode) and get_node_id(inode).  But I think we should
> really bite the bullet and reimplement these functions so that they
> take fuse_conn and node_id as arguments instead of the inode (and
> separate out those parts which do really need the inode for the
> memory mappings).
> 
> It's a bit more work on the fuse side, but overall it results in a
> much cleaner architecture, don't you think?

Yes, I agree.  I was mostly being lazy.  I'll try to rip off mnt
assumption from FUSE and make CUSE bypass the mnt thing.

>> +	worker = kthread_run(cuse_init_worker, cuse_conn_get(cc),
>> +			     "cuse-init-pid%d", current->pid);
> 
> This could be done without a kthread, by setting the ->end() callback
> for the INIT request, no?

Indeed.  Will switch to ->end().

Thanks.

-- 
tejun