Patch: Fixed patch for moving RSVD/SHVXD to vfs_default
urisimchoni at gmail.com
Fri Sep 18 08:41:23 UTC 2015
One difference that I see from ioctls, is that an ioctl says "I'm
providing this well-defined service. You don't like it? Go somewhere
else!". Whereas with samba we strive to serve the whims of all clients
and SMB versions.
A kernel ioctl handler doesn't have to ask "who's my client" - it has
for example one ioctl code for 32 bits and one for 64 bits, and the
ioctl code defines what to do. Do we have the same guarantees with
create contexts? Are they sufficiently stable and/or self-describing?
Having to ask "who's the client" from inside a VFS module, as the
protocols evolve and more implementations arrive, is one place where
the layer separation breaks.
On Fri, Sep 18, 2015 at 9:11 AM, Michael Adam <obnox at samba.org> wrote:
> On 2015-09-18 at 08:01 +0200, Volker Lendecke wrote:
>> On Fri, Sep 18, 2015 at 07:43:17AM +0200, Michael Adam wrote:
>> > :-)
>> > And we are not talking about ioctls but create calls,
>> > or I didn't get the parallel.
>> The parallel for me is that an ioctl is a wildcard mechanism
>> to pass arbitrary stuff from an application down to
>> wherever. Same for passing down create blobs.
> Ah! So then my argument about create contexts NOT being similar
> to ioctls in this respect... :-)
More information about the samba-technical