Re: [PATCH] CIFS: make cifsd (more)

[ Trimmed Cc: list ]

On 6/30/07, Steve French [off-list ref] wrote:

The reason that cifs switched from wait_for_completion to the kthread
call to cifs_demultiplex_thread in the first place is because without
use of kthread it won't work with a linux-vserver.   See the thread:

http://marc.info/?l=linux-cifs-client&m=117552761703381&w=2

If we take out the kthread call, we break those guys.

I agree that using sk_callbacks is worth looking into - I only found
ocfs2 and SunRPC (NFS) though that used it.   Is there a better
example though?   The NFS socket handling code is huge
(net/sunrpc/xprtsck.c) - something seems wrong when replacing a few
lines of code with a new 1675 line file.  There must be a better
example of doing what you suggest...

You're correct. "Right" / "elegant" solutions are rarely (if ever?) complex
and involved. Simplicity _is_ good. I see no point in converting 5 good
lines of maintainable, readable, solid code with 1000 lines of kludge :-)
just to work-around this kthreads limitation. But then, of course, the call
is yours.

I am tempted to drop the socket timeout (which cifs sets to 7 seconds)
to a smaller number and not use signals at all rather than add that
much complexity

Timeout too low => CPU wastage => power wastage. [ Think laptop
batteries, with say 5 cifsd kthreads waking up once every second ... ]
Timeout too high => umount(2) hangs, annoys user, user takes
drastic actions ... so think of some good "magic number" :-)

I don't quite think of all these suggestions as solutions at all -- they
are workarounds at best, IMHO (for kthread's limitation in dealing with
kernel threads that want to block -- I still don't see any fundamental
reasoning / logic behind why kthreads should be banned from doing
blocking recv's -- if there is, please let me know too).

Don't have much else to say than what I already have on the two
threads discussing this.

Satyam