Re: [PATCH v2 04/10] vfs: define getname_info struct and have getname() return it

[Jeff Layton <jlayton@redhat.com>]

On Thu, 20 Sep 2012 09:44:48 -0700
Arnaldo Carvalho de Melo <acme@ghostprotocols.net> wrote:

> Em Thu, Sep 20, 2012 at 10:23:13AM -0400, Jeff Layton escreveu:
> > getname() is intended to copy pathname strings from userspace into a
> > kernel buffer. The result is just a string in kernel space. It would
> > however be quite helpful to be able to attach some ancillary info to
> > the string.
> 
> Tried with:
> 
> [acme@mica linux]$ git remote add viro git://git.kernel.org/pub/scm/linux/kernel/git/viro/signal.git
> [acme@mica linux]$ git checkout -b getname-signal-git viro/experimental-kernel_thread
> 
> [acme@mica linux]$ grep Subject /wb/getname/4.patch 
> Subject: [PATCH v2 04/10] vfs: define getname_info struct and have getname()
> [acme@mica linux]$
> 
> [acme@mica linux]$ grep Subject /wb/getname/[1-3].patch 
> /wb/getname/1.patch:Subject: [PATCH v2 01/10] vfs: allocate page instead of names_cache buffer in
> /wb/getname/2.patch:Subject: [PATCH v2 02/10] vfs: make dir_name arg to do_mount a const char *
> /wb/getname/3.patch:Subject: [PATCH v2 03/10] acct: constify the name arg to acct_on
> [acme@mica linux]$ git log --oneline | head -3
> 9e7ab47 acct: constify the name arg to acct_on
> 2c1cd21 vfs: make dir_name arg to do_mount a const char *
> 99080c0 vfs: allocate page instead of names_cache buffer in mount_block_root
> [acme@mica linux]$ 
> 
> git am works for the first 3, then fails for the fourth, using patch I get:
> 
> [acme@mica linux]$ patch -p1 < /wb/getname/4.patch 
> <SNIP successfull part>
> patching file fs/namei.c
> Hunk #3 succeeded at 2124 (offset 4 lines).
> Hunk #4 succeeded at 2144 (offset 4 lines).
> Hunk #5 succeeded at 3070 (offset 4 lines).
> Hunk #6 succeeded at 3279 (offset 4 lines).
> Hunk #7 succeeded at 3374 (offset 4 lines).
> Hunk #8 succeeded at 3465 (offset 4 lines).
> Hunk #9 succeeded at 3478 (offset 4 lines).
> Hunk #10 succeeded at 3760 (offset 4 lines).
> Hunk #11 succeeded at 3998 (offset 4 lines).
> patching file fs/namespace.c
> patching file fs/open.c
> patching file fs/quota/quota.c
> patching file include/linux/audit.h
> Hunk #1 succeeded at 451 with fuzz 2 (offset -8 lines).
> Hunk #2 FAILED at 469.
> Hunk #3 succeeded at 495 (offset -10 lines).
> 1 out of 3 hunks FAILED -- saving rejects to file include/linux/audit.h.rej
> patching file include/linux/fs.h
> patching file ipc/mqueue.c
> Hunk #3 FAILED at 805.
> Hunk #4 FAILED at 825.
> 2 out of 6 hunks FAILED -- saving rejects to file ipc/mqueue.c.rej
> patching file kernel/acct.c
> patching file kernel/auditsc.c
> Hunk #1 FAILED at 103.
> Hunk #2 succeeded at 1006 (offset 2 lines).
> Hunk #3 succeeded at 1019 (offset 2 lines).
> Hunk #4 succeeded at 1531 (offset 2 lines).
> Hunk #5 succeeded at 1546 (offset 2 lines).
> Hunk #7 FAILED at 2048.
> Hunk #11 FAILED at 2166.
> Hunk #12 FAILED at 2189.
> Hunk #13 FAILED at 2230.
> Hunk #14 FAILED at 2243.
> Hunk #15 FAILED at 2277.
> 7 out of 15 hunks FAILED -- saving rejects to file kernel/auditsc.c.rej
> patching file mm/swapfile.c
> [acme@mica linux]$ 
> 

Sorry, I may not have made it clear in the cover letter here. This
patchset also depends on the audit patchset I sent on Sept. 7th. The
cover letter of that set is:

     [PATCH v6 00/10] audit: overhaul audit_names handling to allow for retrying on path-based syscalls

You may find it easier to just pull down the "estale" branch in my tree
and pop off the patches for the estale series itself.

-- 
Jeff Layton <jlayton@redhat.com>