On 12 Mar 2001, Trond Myklebust wrote:
All that is needed is to unify all the disparate credential info we
already have at the VFS level. Just look at the ridiculous number of
different places in which we currently try to save the exact same
info: 10 fields in task structure + 2-3 fields
" " == Alexander Viro [EMAIL PROTECTED] writes:
Aye. It's certainly 2.5 stuff, but I'll include that into
namespace-patch for testing. We probably need to sort the API
issues out - I suspect that net/sunrpc will cause most of the
pain and you definitely know that code
" " == Alexander Viro [EMAIL PROTECTED] writes:
Umm... Let's make it pure copy-on-write cache a-la 4.4 and if
pthreads folks need this craziness with thread-group-wide
credentials change - let them pay upon changing credentials. I
really don't see any point in
Someone tell me if my chain of reasoning is wrong here...
(1) The only way to get a `struct file' is to call get_empty_filp()
(2) All callers of get_empty_filp() set -f_ops to a non-NULL value
(3) All checks for f_ops being NULL can be removed
--
Revolutions do not require corporate
On Thu, Mar 08, 2001 at 11:16:10AM -0500, Alexander Viro wrote:
I'm not sure on your #2. In principle, -i_fop can be NULL. It may be
a good thing to declare that it should never happens, but right now it's
not guaranteed.
Besides, revoke-like thing in proc/generic.c _does_ set f_op to NULL.
On Thu, 8 Mar 2001, Andi Kleen wrote:
There is code that creates private struct files without calling get_empty_filp()
NFS comes to mind, but there is other too. Now of course these could
be changed, but I think the change would cause much more work for debugging/
fixing than you think ;)
On Thu, 08 Mar 2001, Matthew Wilcox wrote:
On Thu, Mar 08, 2001 at 11:16:10AM -0500, Alexander Viro wrote:
I'm not sure on your #2. In principle, -i_fop can be NULL. It may be
a good thing to declare that it should never happens, but right now it's
not guaranteed.
Besides, revoke-like
On Thu, 8 Mar 2001, Daniel Phillips wrote:
I would like to see the (struct file *) in -readpage go away. I don't
think there is a technical obstacle to doing that, since the same
parameter was removed from -writepage without much fuss. The code
impact is in samba, coda and nfs. All block
can the same be done for d_ops in dentries?
- Chuck Lever
--
corporate: [EMAIL PROTECTED]
personal: [EMAIL PROTECTED]
-Original Message-
From: Matthew Wilcox [mailto:[EMAIL PROTECTED]]
Sent: Thursday, March 08, 2001 10:48 AM
To: [EMAIL PROTECTED]
Subject: NULL