[PATCH] kbuild: When checking depmod version, redirect stderr

When running depmod to check for the correct version number, extra
output we don't need to see, such as "depmod: QM_MODULES: Function not
implemented" may show up.  Redirect stderr to /dev/null as the version
information that we do care about comes to stdout.

Signed-off-by: Tom Rini <trini@kernel.crashing.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

kbuild: fix make O=... build

It fixes the following error:

make[1]: *** No rule to make target `include/asm', needed by `arch/alpha/kernel/asm-offsets.s'.  Stop.

Reported by:
From: Jan Dittmer <j.dittmer@portrix.net>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

kbuild: Fix bug in make deb-pkg when using seperate source and output directories

From: Ryan Anderson <ryan@michonline.com>

When running "make O=something deb-pkg", I get a failure that claims I
haven't configured my kernel (I have).  Running it a second time tells
me to run "make mrproper"  (include/linux/version.h got built on the
first run)

Original patch from:
From: Ajay Patel <patela@gmail.com>

With modifications from:
Signed-off-By: Ryan Anderson <ryan@michonline.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

uml: Restore proper descriptions in make deb-pkg target

From: Ryan Anderson <ryan@michonline.com>

This pulls the description from the Debian user-mode-linux package, and
puts $version back in the appropriate places for both descriptions.

Signed-off-by: Ryan Anderson <ryan@michonline.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

uml: Make deb-pkg build target build a Debian-style user-mode-linux package

From: Ryan Anderson <ryan@michonline.com>

Make the deb-pkg build target understand the "um" arch and set up the
package and directory structure to match a mainline-Debian style
user-mode-linux package.

This is primarily so that it stops matching, exactly, the naming
convention used by normal, non-UML kernels generated by this command.

Installing "linux-2.6.11" and "linux-2.6.11", where one is a UML kernel
doesn't do the right thing.  This fixes that.

Signed-off-by: Ryan Anderson <ryan@michonline.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

kbuild: Don't fail if include/asm symlink exists

From: Andreas Gruenbacher <agruen@suse.de>

We're having the following situation: There are user-space applications
that include kernel headers directly. With a completely unconfigured
/usr/src/linux tree, including most headers fails because essential
files are not there:

include/asm
include/linux/autoconf.h
include/linux/version.h

So we create these files. On the other hand, we want to use
/usr/src/linux as read-only source for building kernels or additional
modules. Now when building a kernel with a separate output directory
(O=), there is a check in the main makefile for the include/asm symlink.
There is no real need for this check: if we ensure that
$(objdir)/include/asm is always created as the patch does,
$(srctree)/include/asm becomes irrelevant.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

kbuild: fix buildcheck

From: Randy Dunlap <rddunlap@osdl.org>

I should not have added init.text test here;
it's more than useless, it actually degrades the output.

Signed-off-by: Randy Dunlap <rddunlap@osdl.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

kbuild: Add target debug_kallsyms

From: Keith Owens <kaos@ocs.com.au>

Make it easier to generate maps for debugging kallsyms problems.
debug_kallsyms is only a debugging target so no help or silent mode.

Signed-off-by: Keith Owens <kaos@ocs.com.au>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

kbuild: "PREEMPT" in UTS_VERSION

From: Matt Mackall <mpm@selenic.com>

Add PREEMPT to UTS_VERSION where enabled as is done for SMP to make
preempt kernels easily identifiable.
Added SMP PREEMPT as comment in compile.h to force it to be
updated when they change (sam).

Signed-off-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

kbuild: Avoid inconsistent kallsyms data

Several reports on inconsistent kallsyms data has been caused by the aliased symbols
__sched_text_start and __down to shift places in the output of nm.
The root cause was that on second pass ld aligned __sched_text_start to a 4 byte boundary
which is the function alignment on i386.
sched.text and spinlock.text is now aligned to an 8 byte boundary to make sure they
are aligned to a function alignemnt on most (all?) archs.

Tested by: Paulo Marques <pmarques@grupopie.com>
Tested by: Alexander Stohr <Alexander.Stohr@gmx.de>

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

buildcheck: reduce DEBUG_INFO noise from reference* scripts

From: Randy Dunlap <rddunlap@osdl.org>

Reduce noise in 'make buildcheck' that is caused by CONFIG_DEBUG_INFO=y.

Signed-off-by: Randy Dunlap <rddunlap@osdl.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

kbuild: Fix build as root then user

From: Matthew Wilcox <matthew@wil.cx>
I inadvertently built a tree as root and then rebuilt it as a user.  I
got a lot of prompts ...

mv: overwrite `drivers/char/drm/drm_auth.o', overriding mode 0644?

Using mv -f fixes that.

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

[PATCH] scripts/kernel-doc: don't use uninitialized SRCTREE

Current kernel-doc (perl) script generates this warning:
Use of uninitialized value in concatenation (.) or string at scripts/kernel-doc line 1668.

So explicitly check for SRCTREE in the ENV before using it,
and then if it is set, append a '/' to the end of it, otherwise
the SRCTREE + filename can (will) be missing the intermediate '/'.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

[PATCH] Lindent: ignore .indent.pro

 When I recently submitted a Lindent patch, it turned out that my .indent.pro
 options were also applied to the tree. This patch directs indent(1) to ignore
 the .indent.pro directives and only use options specified on the command
 line.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

[PATCH] kbuild: restrain output of "make help" to 80 columns

This patch fixes the output of "make help" to fit in a 80 column
screen. Please push upstream as part of your other patches.

Signed-off-by: Yum Rayan <yum.rayan@gmail.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

[PATCH] kbuild: obey HOSTLOADLIBES_programname for single-file compilation

Single-file HOSTCC calls added the libraries from $(HOSTLOADLIBES),
but not from $(HOSTLOADLIBES_programname). Multi-file HOSTCC calls do
both.

This patch fixes that inconsistency.

Signed-Off-By: Matthias Urlichs <smurf@debian.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

[PATCH] kbuild: allow cscope to index multiple architectures

I have a single source tree which I cross compile for a couple of
different architectures using ARHC=foo O=blah etc.

The existing cscope target is very handy but only indexes the current
$(ARCH), which is a pain since inevitably I'm interested in the other
one at any given time ;-). This patch allows me to pass a list of
architectures for cscope to index. e.g.
make ALLSOURCE_ARCHS="i386 arm" cscope

This change also works for etags etc, and I presume it is just as useful
there.

Signed-off-by: Ian Campbell <ijc@hellion.org.uk>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

[PATCH] kbuild: make 'cscope -q' play well with cscope.el

I tried the Linux Makefile 'make cscope' target, and found that the
generated database is not compatible with 'cscope.el' under XEmacs.
The thing is that 'cscope.el' does not allow setting the command line
options to the 'cscope' commands it runs, and it errors with a message
about the options not matching the ones used to generate the index.

It turns out the cscope designers already thought of this.  The
options can be written into the "cscope.files".  The included patch
moves the "-q" and "-k" options from the 'cmd_cscope' to the
'cmd_cscope-file', echoing them into the top of the files listing.

Now the index is generated with the "-q" option, and when 'cscope.el'
performs it's search, it uses that argument as well.  Lookups are fast
and everyone is happy.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

[PATCH] kbuild: add ia64 support to rpm Makefile target

On ia64, only the EFI (fat) partition is available to boot from.  The rpm
needs to install the kernel under /boot/efi to be useable on ia64.

Signed-off-by: Greg Edwards <edwardsg@sgi.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

[PATCH] kbuild: modpost needs to cope with new glibc elf header on sparc

Recently a change in the glibc elf.h header has been introduced causing
modpost to spawn tons of warnings (like the one below) building the kernel
on sparc:

[SNIP]
*** Warning: "current_thread_info_reg" [net/sunrpc/auth_gss/auth_rpcgss.ko] undefined!
*** Warning: "" [net/sunrpc/auth_gss/auth_rpcgss.ko] undefined!
*** Warning: "" [net/sunrpc/auth_gss/auth_rpcgss.ko] undefined!
[SNIP]

Ben Collins discovered that the STT_REGISTERED definition in glibc did change
and that this change needs to be propagated to modpost.

glibc change:
-#define STT_REGISTER   13              /* Global register reserved to app. */
+#define STT_SPARC_REGISTER     13      /* Global register reserved to app. */

I did and tested this simple patch to maintain compatibility with newer (>= 2.3.4)
and older (<= 2.3.2) glibc.

Signed-off-by: Fabio M. Di Nitto <fabbione@fabbione.net>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

[PATCH] kbuild: create tarballs

It adds tarball packaging, which I prefer for distribution.
Also one of the two blanks after @echo is removed. One seems to be enough :)

Signed-off-by: Jan-Benedict Glaw <jbglaw@lug-owl.de>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

Merge /pub/scm/linux/kernel/git/aegl/linux-2.6

[SPARC64]: Add missing asm-sparc64/seccomp.h file.

Signed-off-by: David S. Miller <davem@davemloft.net>

[IA64] assign_irq_vector() should not panic

Current assign_irq_vector() will panic if interrupt vectors is running
out. But I think how to handle the case of lack of interrupt vectors
should be handled by the caller of this function. For example, some
PCI devices can raise the interrupt signal via both MSI and I/O
APIC. So even if the driver for these device fails to allocate a
vector for MSI, the driver still has a chance to use I/O APIC based
interrupt. But currently there is no chance for these driver to use
I/O APIC based interrupt because kernel will panic when
assign_irq_vector() fails to allocate interrupt vector.

The following patch changes assign_irq_vector() for ia64 to return
-ENOSPC on error instead of panic (as i386 and x86_64 versions do).

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

[IA64] use msleep_interruptible() instead of schedule_timeout

Description: Replace schedule_timeout() with msleep_interruptible() to
guarantee the task delays as expected.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

[PATCH] v850: Update mmu.h header to match implementation changes

Signed-off-by: Miles Bader <miles@gnu.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] v850: Update checksum.h to match changed function signatures

Signed-off-by: Miles Bader <miles@gnu.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge /pub/scm/linux/kernel/git/tglx/mtd-2.6

Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

[IA64] remove linux/version.h include from arch/ia64

changing CONFIG_LOCALVERSION rebuilds too much, for no appearent reason.

Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: Tony Luck <tony.luck@intel.com>

Auto merge with /home/aegl/GIT/linus

[SPARC64]: Add syscall auditing support.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Pass regs and entry/exit boolean to syscall_trace()

Also fix a bug in 32-bit syscall tracing.  We forgot to update
this code when we moved over to the convention that all 32-bit
syscall arguments are zero extended by default.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Add SECCOMP support.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Kill ancient and unused SYSCALL_TRACING debugging code.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Add __read_mostly support.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC]: Add ioprio system call support.

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge master.kernel.org:~rmk/linux-2.6-arm.git

[PATCH] remove asm-xtensa/ipc.h

Now that sys_ipc has been removed from xtensa, asm/ipc.h is no longer
needed for that architecture.  Not tested, but obviously correct.  This
file is included only from arch code and this patch also removes the only
inclusion.

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Sync up ieee-1394

Lots of this patch is trivial code cleanups (static vars were being
intialized to 0, etc).

There's also some fixes for ISO transmits (max buffer handling).
Aswell, we have a few fixes to disable IRM capabilites correctly.  We've
also disabled, by default some generally unused EXPORT symbols for the
sake of cleanliness in the kernel.  However, instead of removing them
completely, we felt it necessary to have a config option that allowed
them to be enabled for the many projects outside of the main kernel tree
that use our API for driver development.

The primary reason for this patch is to revert a MODE6->MODE10 RBC
conversion patch from the SCSI maintainers.  The new conversions handled
directly in the scsi layer do not seem to work for SBP2.  This patch
reverts to our old working code so that users can enjoy using Firewire
disks and dvd drives again.

We are working with the SCSI maintainers to resolve this issue outside
of the main kernel tree.  We'll merge the patch once the SCSI layer's
handling of the MODE10 conversion is working for us.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ARM: 2803/1: OMAP update 11/11: Add cpufreq support

Patch from Tony Lindgren

This patch adds minimal cpufreq support for OMAP
taking advantage of the clock framework.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2805/1: OMAP update 10/11: Update H2 defconfig

Patch from Tony Lindgren

This patch updates H2 defconfig.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2804/1: OMAP update 9/11: Update OMAP arch files

Patch from Tony Lindgren

This patch by various OMAP developers syncs the OMAP
specific arch files with the linux-omap tree.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2802/1: OMAP update 8/11: Update OMAP arch files

Patch from Tony Lindgren

This patch by various OMAP developers syncs the OMAP
specific arch files with the linux-omap tree.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2812/1: OMAP update 7c/11: Move arch-omap to plat-omap

Patch from Tony Lindgren

This patch move common OMAP code from arch-omap to plat-omap
directory.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2809/1: OMAP update 7b/11: Move arch-omap to plat-omap

Patch from Tony Lindgren

This patch move common OMAP code from arch-omap to plat-omap
directory.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2807/1: OMAP update 7a/11: Move arch-omap to plat-omap

Patch from Tony Lindgren

This patch move common OMAP code from arch-omap to plat-omap
directory.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2801/1: OMAP update 6/11: Split OMAP1 common code into id, io and serial

Patch from Tony Lindgren

This patch by Juha Yrjölä and other OMAP developers splits
OMAP1 specific common code into OMAP1 id, io, and serial
code in mach-omap1 directory.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2806/1: OMAP update 5/11: Move board files into mach-omap1 directory

Patch from Tony Lindgren

This patch by Paul Mundt and other OMAP developers
moves OMAP1 board files into mach-omap1 directory.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2799/1: OMAP update 4/11: Move OMAP1 LED code into mach-omap1 directory

Patch from Tony Lindgren

This patch by Paul Mundt and other OMAP developers
moves OMAP1 specific LED code into mach-omap1 directory.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2800/1: OMAP update 3/11: Move OMAP1 core code into mach-omap1 directory

Patch from Tony Lindgren

This patch by Paul Mundt and other OMAP developers
moves OMAP1 specific IRQ, time, and FPGA code into
mach-omap1 directory.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2798/1: OMAP update 2/11: Change ARM Kconfig to support omap1 and omap2

Patch from Tony Lindgren

This patch by Paul Mundt and other OMAP developers modifies
ARM specific Kconfig to allow sharing code between OMAP1 and
OMAP2 architectures.
In order to share code between OMAP1 and OMAP2, all OMAP1
specific code is moved into mach-omap1 directory in the
following patch. A new mach-omap2 directory will be added
later on.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2797/1: OMAP update 1/11: Update include files

Patch from Tony Lindgren

This patch by various OMAP developers syncs the OMAP
specific include files with the linux-omap tree.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2796/1: Fix ARMv5[TEJ] check in MMU initalization

Patch from Deepak Saxena

The code in mm-armv.c checks for the condition (cpu_architecture()<= ARMv5)
in a few places but should be checking for ARMv5TEJ as the MMU is shared
across all v5 variations.

Signed-off-by: Deepak Saxena <dsaxena@plexity.net>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2795/1: update ixp2000 defconfigs

Patch from Lennert Buytenhek

Update the ixp2000 defconfigs from 2.6.12-git6 to 2.6.13-rc2.

Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2793/1: platform serial support for ixp2000

Patch from Lennert Buytenhek

This patch converts the ixp2000 serial port over to a platform
serial device.

Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: Deepak Saxena <dsaxena@plexity.net>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

drm: fix stupid missing semicolon.

I fixed this in one git tree but that wasn't the one I pushed...

Signed-off-by: Dave Airlie <airlied@linux.ie>

Merge head 'drm-fixes' of /linux/kernel/git/airlied/drm-2.6

Merge head 'drm-3264' of /linux/kernel/git/airlied/drm-2.6

Merge head 'drm-via' of /linux/kernel/git/airlied/drm-2.6

Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

[SCTP]: Use struct list_head for chunk lists, not sk_buff_head.

Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Fix warning in ip6_mc_msfilter.

Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: fix IPv4 leave-group group matching

        This patch fixes the multicast group matching for
IP_DROP_MEMBERSHIP, similar to the IP_ADD_MEMBERSHIP fix in a prior
patch. Groups are identifiedby <group address,interface> and including
the interface address in the match will fail if a leave-group is done
by address when the join was done by index, or if different addresses
on the same interface are used in the join and leave.

Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: (INCLUDE,empty)/leave-group equivalence for full-state MSF APIs & errno fix

1) Adds (INCLUDE, empty)/leave-group equivalence to the full-state
   multicast source filter APIs (IPv4 and IPv6)

2) Fixes an incorrect errno in the IPv6 leave-group (ENOENT should be
   EADDRNOTAVAIL)

Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: multicast API "join" issues

1) In the full-state API when imsf_numsrc == 0
   errno should be "0", but returns EADDRNOTAVAIL

2) An illegal filter mode change
   errno should be EINVAL, but returns EADDRNOTAVAIL

3) Trying to do an any-source option without IP_ADD_MEMBERSHIP
   errno should be EINVAL, but returns EADDRNOTAVAIL

4) Adds comments for the less obvious error return values

Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: multicast API "join" issues

1) Changes IP_ADD_SOURCE_MEMBERSHIP and MCAST_JOIN_SOURCE_GROUP to ignore
   EADDRINUSE errors on a "courtesy join" -- prior membership or not
   is ok for these.

2) Adds "leave group" equivalence of (INCLUDE, empty) filters in the
   delta-based API. Without this, mixing delta-based API calls that
   end in an (INCLUDE, empty) filter would not allow a subsequent
   regular IP_ADD_MEMBERSHIP. It also frees socket buffer memory that
   isn't needed for both the multicast group record and source filter.

Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: multicast API "join" issues

        This patch corrects a few problems with the IP_ADD_MEMBERSHIP
socket option:

1) The existing code makes an attempt at reference counting joins when
   using the ip_mreqn/imr_ifindex interface. Joining the same group
   on the same socket is an error, whatever the API. This leads to
   unexpected results when mixing ip_mreqn by index with ip_mreqn by
   address, ip_mreq, or other API's. For example, ip_mreq followed by
   ip_mreqn of the same group will "work" while the same two reversed
   will not.
           Fixed to always return EADDRINUSE on a duplicate join and
   removed the (now unused) reference count in ip_mc_socklist.

2) The group-search list in ip_mc_join_group() is comparing a full
   ip_mreqn structure and all of it must match for it to find the
   group. This doesn't correctly match a group that was joined with
   ip_mreq or ip_mreqn with an address (with or without an index). It
   also doesn't match groups that are joined by different addresses on
   the same interface. All of these are the same multicast group,
   which is identified by group address and interface index.
           Fixed the check to correctly match groups so we don't get
   duplicate group entries on the ip_mc_socklist.

3) The old code allocates a multicast address before searching for
   duplicates requiring it to free in various error cases. This
   patch moves the allocate until after the search and
   igmp_max_memberships check, so never a need to allocate, then free
   an entry.

Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Apply sysctl_icmp_echo_ignore_broadcasts to ICMP_TIMESTAMP as well.

This was the full intention of the original code.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Support CONFIG_HZ

Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Fix sparse warnings

From: Victor Fusco <victor@cetuc.puc-rio.br>

Fix the sparse warning "implicit cast to nocast type"

Signed-off-by: Victor Fusco <victor@cetuc.puc-rio.br>
Signed-off-by: Domen Puncer <domen@coderock.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Transform skb_queue_len() binary tests into skb_queue_empty()

This is part of the grand scheme to eliminate the qlen
member of skb_queue_head, and subsequently remove the
'list' member of sk_buff.

Most users of skb_queue_len() want to know if the queue is
empty or not, and that's trivially done with skb_queue_empty()
which doesn't use the skb_queue_head->qlen member and instead
uses the queue list emptyness as the test.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Typo in dtlb_backend.S, _PAGE_SZ4M --> _PAGE_SZ4MB

Noticed by Eddie C. Dost

Signed-off-by: David S. Miller <davem@davemloft.net>

[IA64] Fix a typo in arch/ia64/kernel/entry.S

Both 2.4 and 2.6 kernels need this patch for the next binutils.

Signed-off-by: Tony Luck <tony.luck@intel.com>

Auto merge with /home/aegl/GIT/linus

[PATCH] Add MAINTAINERS entry for audit subsystem

I've been asked about this a couple times, and there's no info in
MAINTAINERS file.  Add MAINTAINERS entry for audit subsystem.

Signed-off-by: Chris Wright <chrisw@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] device-mapper: dm-raid1: Limit bios to size of mirror region

Set the target's split_io field when building a dm-mirror device so
incoming bios won't span the mirror's internal regions.  Without this,
regions can be accessed while not holding correct locks and data corruption
is possible.

Reported-By: "Zhao Qian" <zhaoqian@aaastor.com>
From: Kevin Corry <kevcorry@us.ibm.com>
Signed-Off-By: Alasdair G Kergon <agk@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] video doc: one more system where video works with S3

One more system where video works with S3.

Signed-off-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] m32r: framebuffer device support

This patch is for supporting Epson s1d13xxx framebuffer device for m32r.  #
Sorry, a little bigger.

The Epson s1d13806 is already supported by 2.6.12 kernel, and its driver is
placed as drivers/video/s1d13xxxfb.c.

For the m32r, a header file include/asm-m32r/s1d13806.h was prepared for
several m32r target platforms.  It was originally generated by an Epson
tool S1D13806CFG.EXE, and modified manually for the m32r platforms.

Signed-off-by: Hayato Fujiwara <fujiwara@linux-m32r.org>
Signed-off-by: Hirokazu Takata <takata@linux-m32r.org>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: fix fh_expire_type

After discussion at the recent NFSv4 bake-a-thon, I realized that my
assumption that NFS4_FH_PERSISTENT required filehandles to persist was a
misreading of the spec.  This also fixes an interoperability problem with the
Solaris client.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: check lock type against openmode.

We shouldn't be allowing, e.g., write locks on files not open for read.  To
enforce this, we add a pointer from the lock stateid back to the open stateid
it came from, so that the check will continue to be correct even after the
open is upgraded or downgraded.

Signed-off-by: Andy Adamson <andros@citi.umich.edu>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: clean up nfs4_preprocess_seqid_op

As long as we're here, do some miscellaneous cleanup.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: clarify close_lru handling

The handling of close_lru in preprocess_stateid_op was a source of some
confusion here recently.  Try to make the logic a little clearer, by renaming
find_openstateowner_id to make its purpose clearer and untangling some
unnecessarily complicated goto's.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: renew lease on seqid modifying operations

nfs4_preprocess_seqid_op is called by NFSv4 operations that imply an implicit
renewal of the client lease.

Signed-off-by: Andy Adamson <andros@citi.umich.edu>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: return better error on io incompatible with open mode

from RFC 3530:
"Share reservations are established by OPEN operations and by their
nature are mandatory in that when the OPEN denies READ or WRITE
operations, that denial results in such operations being rejected
with error NFS4ERR_LOCKED."

(Note that share_denied is really only a legal error for OPEN.)

Signed-off-by: Andy Adamson <andros@citi.umich.edu>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: always update stateid on open

An OPEN from the same client/open stateowner requires a stateid update because
of the share/deny access update.

Signed-off-by: Andy Adamson <andros@citi.umich.edu>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: relax new lock seqid check

We're insisting that the lock sequence id field passed in the
open_to_lockowner struct always be zero.  This is probably thanks to the
sentence in rfc3530: "The first request issued for any given lock_owner is
issued with a sequence number of zero."

But there doesn't seem to be any problem with allowing initial sequence
numbers other than zero.  And currently this is causing lock reclaims from the
Linux client to fail.

In the spirit of "be liberal in what you accept, conservative in what you
send", we'll relax the check (and patch the Linux client as well).

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: seqid comments

Add some comments on the use of so_seqid, in an attempt to avoid some of the
confusion outlined in the previous patch....

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: fix open_reclaim seqid

The sequence number we store in the sequence id is the last one we received
from the client.  So on the next operation we'll check that the client gives
us the next higher number.

We increment sequence id's at the last moment, in encode, so that we're sure
of knowing the right error return.  (The decision to increment the sequence id
depends on the exact error returned.)

However on the *first* use of a sequence number, if we set the sequence number
to the one received from the client and then let the increment happen on
encode, we'll be left with a sequence number one to high.

For that reason, ENCODE_SEQID_OP_TAIL only increments the sequence id on
*confirmed* stateowners.

This creates a problem for open reclaims, which are confirmed on first use.
Therefore the open reclaim code, as a special exception, *decrements* the
sequence id, cancelling out the undesired increment on encode.  But this
prevents the sequence id from ever being incremented in the case where
multiple reclaims are sent with the same openowner.  Yuch!

We could add another exception to the open reclaim code, decrementing the
sequence id only if this is the first use of the open owner.

But it's simpler by far to modify the meaning of the op_seqid field: instead
of representing the previous value sent by the client, we take op_seqid, after
encoding, to represent the *next* sequence id that we expect from the client.
This eliminates the need for special-case handling of the first use of a
stateowner.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: comment indentation

Yeah, it's trivial, but this drives me up the wall....

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: stop overusing RECLAIM_BAD

A misreading of the spec lead us to convert all errors on open and lock
reclaims to RECLAIM_BAD.  This causes problems--for example, a reboot within
the grace period could lead to reclaims with stale stateid's, and we'd like to
return STALE errors in those cases.

What rfc3530 actually says about RECLAIM_BAD: "The reclaim provided by the
client does not match any of the server's state consistency checks and is
bad." I'm assuming that "state consistency checks" refers to checks for
consistency with the state recorded to stable storage, and that the error
should be reserved for that case.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: ERR_GRACE should bump seqid on lock

A GRACE or NOGRACE response to a lock request should also bump the sequence
id.  So we delay the handling of grace period errors till after we've found
the relevant owner.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: ERR_GRACE should bump seqid on open

The GRACE and NOGRACE errors should bump the sequence id on open.  So we delay
the handling of these errors until nfsd4_process_open2, at which point we've
set the open owner, so the encode routine will be able to bump the sequence
id.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: fix release_lockowner

We oops in list_for_each_entry(), because release_stateowner frees something
on the list we're traversing.

Signed-off-by: Andy Adamson <andros@citi.umich.edu>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: prevent multiple unlinks of recovery directories

Make sure we don't try to delete client recovery directories multiple times;
fixes some spurious error messages.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: lookup_one_len takes i_sem

Oops, this lookup_one_len needs the i_sem.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: fix sync'ing of recovery directory

We need to fsync the recovery directory after writing to it, but we weren't
doing this correctly.  (For example, we weren't taking the i_sem when calling
->fsync().)

Just reuse the existing nfsd fsync code instead.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: reboot recovery fix

We need to remove the recovery directory here too.  (This chunk just got lost
somehow in the process of commuting the reboot recovery patches past the other
patches.)

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] yenta: don't depend on CardBus

As a follow-up, we can allow the yenta-driver to be limited to PCMCIA
operation.

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>