Hi Andrew.

Andrew Morton wrote:
> Here's a random hack for smbfs.  Does it help?
Unfortunately this (and the updated)
patch is very far from any help :(
I did some investigations myself.
At first I checked up the callers and
found that they do smb_rput() themselves
if smb_add_request() returns error. So
we have double-free. Then I found that
smb_rput() does list_del_init() itself,
so this is being done 3 times instead
of one. So all that piece of code you
were trying to correct, was actually
completely useless. I removed it (patch
attached). That made the smbfs to work
for me (for the first time in 2.6!), but
after about 5 minutes I've got the attached
Oops and Slab corruption again. A quick
glance to the code pointed by Oops, revealed
that there are plenty of use-after-rput
instances. Then I've found much more of
them all around the code.
I don't suppose this can ever work.
Now I am just scared to even load the smbfs
module.
I asked our admin to enable the NTLM auth
and mounted everything as cifs instead.
That works like a charm!
If there are any other patches to try -
I'll test them, but what is the current
official status of smbfs? Is it deprecated?
Or broken? Or undergoing some redesign?
Very strange...