Home » Mailing list archives » linux-kernel » 2010 » October » 20

Re: [PATCH 1/6] IMA: move read/write counters into struct inode

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
[view in full thread]
From: Linus Torvalds
Subject: Re: [PATCH 1/6] IMA: move read/write counters into struct inode
Date: Tuesday, October 19, 2010 - 5:25 pm

On Tue, Oct 19, 2010 at 3:58 PM, Eric Paris <eparis@redhat.com> wrote:

Hmm. The end result looks fine (adding four bytes to struct inode in
order to avoid all the complexity seems reasonable), but I do get the
feeling that this should likely be the last in the series, so that the
VFS level files would get minimal changes. IOW, do the cleanups inside
the IMA code first, and then do the switch-over to using counters in
the inode last.

Well, not last, since I think you need to do this before you can do
the "only allocate iint when needed" only after you've moved the
counters. But I think the logical order would be
 - switch to rbtree
 - drop opencount
 - switch counts to 'unsigned int'
 - drop iint->writecount and use i_writecount instead
 - move the remaining readcount to i_readcount
 - only allocate iint when necessary

That way you'd only have _one_ patch that touches <linux/fs.h>, rather
than four, and the remaining patches would all be to security/ima.

But maybe I missed some reason for this particular ordering.

Oh, and btw, due to alignment reasons it looks like the 4-byte
i_readcount would take 8 bytes due to bad structure packing. I don't
know if that is avoidable, but I do think it would make more sense to
put it next to i_writecount instead of in between two pointers. That
still doesn't help (we've got 3 32-bit values next to each other), but
it's at least -closer- to working out.

                         Linus
--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
Messages in current thread:
[PATCH 1/6] IMA: move read/write counters into struct inode, Eric Paris, (Tue Oct 19, 3:58 pm)
[PATCH 2/6] IMA: drop the inode opencount since it isn't n ..., Eric Paris, (Tue Oct 19, 3:58 pm)
[PATCH 3/6] IMA: use unsigned int instead of long for counters, Eric Paris, (Tue Oct 19, 3:58 pm)
[PATCH 4/6] IMA: only allocate iint when needed, Eric Paris, (Tue Oct 19, 3:58 pm)
[PATCH 5/6] IMA: use rbtree instead of radix tree for inod ..., Eric Paris, (Tue Oct 19, 3:58 pm)
[PATCH 6/6] IMA: use i_writecount rather than a private co ..., Eric Paris, (Tue Oct 19, 3:58 pm)
Re: [PATCH 5/6] IMA: use rbtree instead of radix tree for ..., Dave Chinner, (Tue Oct 19, 4:17 pm)
Re: [PATCH 1/6] IMA: move read/write counters into struct ..., Linus Torvalds, (Tue Oct 19, 5:25 pm)
Re: [PATCH 4/6] IMA: only allocate iint when needed, Al Viro, (Tue Oct 19, 8:53 pm)
Re: [PATCH 5/6] IMA: use rbtree instead of radix tree for ..., Peter Zijlstra, (Wed Oct 20, 4:31 am)
Re: [PATCH 5/6] IMA: use rbtree instead of radix tree for ..., Dave Chinner, (Wed Oct 20, 3:05 pm)
Re: [PATCH 5/6] IMA: use rbtree instead of radix tree for ..., Linus Torvalds, (Wed Oct 20, 3:22 pm)
Re: [PATCH 5/6] IMA: use rbtree instead of radix tree for ..., Trond Myklebust, (Wed Oct 20, 3:47 pm)
Re: [PATCH 5/6] IMA: use rbtree instead of radix tree for ..., Linus Torvalds, (Wed Oct 20, 5:58 pm)
Re: [PATCH 5/6] IMA: use rbtree instead of radix tree for ..., Dave Chinner, (Wed Oct 20, 7:17 pm)
Re: [PATCH 1/6] IMA: move read/write counters into struct ..., Eric Paris, (Fri Oct 22, 8:01 pm)
Re: [PATCH 1/6] IMA: move read/write counters into struct ..., Mimi Zohar, (Sat Oct 23, 11:52 pm)