openwrt/staging/blogic.git
17 years ago[NETFILTER]: ip_tables: ipt and ipt_compat checks unification
Dmitry Mishin [Tue, 12 Dec 2006 08:29:26 +0000 (00:29 -0800)]
[NETFILTER]: ip_tables: ipt and ipt_compat checks unification

Matches and targets verification is duplicated in normal and compat processing
ways. This patch refactors code in order to remove this.

Signed-off-by: Dmitry Mishin <dim@openvz.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: x_tables: add missing try to load conntrack from match/targets
Yasuyuki Kozakai [Tue, 12 Dec 2006 08:29:02 +0000 (00:29 -0800)]
[NETFILTER]: x_tables: add missing try to load conntrack from match/targets

CLUSTERIP, CONNMARK, CONNSECMARK, and connbytes need ip_conntrack or
layer 3 protocol module of nf_conntrack.

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: x_tables: error if ip_conntrack is asked to handle IPv6 packets
Yasuyuki Kozakai [Tue, 12 Dec 2006 08:28:40 +0000 (00:28 -0800)]
[NETFILTER]: x_tables: error if ip_conntrack is asked to handle IPv6 packets

To do that, this makes nf_ct_l3proto_try_module_{get,put} compatible
functions. As a result we can remove '#ifdef' surrounds and direct call of
need_conntrack().

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: nf_nat: fix NF_NAT dependency
Yasuyuki Kozakai [Tue, 12 Dec 2006 08:28:09 +0000 (00:28 -0800)]
[NETFILTER]: nf_nat: fix NF_NAT dependency

NF_NAT depends on NF_CONNTRACK_IPV4, not NF_CONNTRACK.

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: Fix INET=n linking error
Patrick McHardy [Tue, 12 Dec 2006 08:27:46 +0000 (00:27 -0800)]
[NETFILTER]: Fix INET=n linking error

Building with INET=n results in

WARNING: "ip_route_output_key" [net/netfilter/nf_conntrack_h323.ko] undefined!

The entire code in net/netfilter is only used for IPv4/IPv6 currently, so
let it depend on INET.

Noticed by Toralf Förster <toralf.foerster@gmx.de>.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years agoMerge branch 'for-linus' of master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Wed, 13 Dec 2006 23:58:32 +0000 (15:58 -0800)]
Merge branch 'for-linus' of /home/rmk/linux-2.6-arm

* 'for-linus' of master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] 4017/1: [Jornada7xx] - Updating Jornada720.c
  [ARM] 3992/1: i.MX/MX1 CPU Frequency scaling support
  [ARM] Provide a method to alter the control register
  [ARM] 4016/1: prefetch macro is wrong wrt gcc's "delete-null-pointer-checks"
  [ARM] Remove empty fixup function
  [ARM] 4014/1: include drivers/hid/Kconfig
  [ARM] 4013/1: clocksource driver for netx
  [ARM] 4012/1: Clocksource for pxa
  [ARM] Clean up ioremap code
  [ARM] Unuse another Linux PTE bit
  [ARM] Clean up KERNEL_RAM_ADDR
  [ARM] Add sys_*at syscalls
  [ARM] 4004/1: S3C24XX: UDC remove implict addition of VA to regs
  [ARM] Formalise the ARMv6 processor name string
  [ARM] Handle HWCAP_VFP in VFP support code
  [ARM] 4011/1: AT91SAM9260: Fix compilation with NAND driver
  [ARM] 4010/1: AT91SAM9260-EK board: Prepare for MACB Ethernet support

17 years agoMerge branch 'release' of master.kernel.org:/home/ftp/pub/scm/linux/kernel/git/aegl...
Linus Torvalds [Wed, 13 Dec 2006 23:57:58 +0000 (15:57 -0800)]
Merge branch 'release' of /home/ftp/linux/kernel/git/aegl/linux-2.6

* 'release' of master.kernel.org:/home/ftp/pub/scm/linux/kernel/git/aegl/linux-2.6:
  [IA64] Move sg_dma_{len,address} from pci.h to scatterlist.h

17 years agoDriver core: deprecate PM_LEGACY, default it to N
David Brownell [Tue, 5 Dec 2006 11:07:38 +0000 (03:07 -0800)]
Driver core: deprecate PM_LEGACY, default it to N

Deprecate the old "legacy" PM API, and more importantly default it to "n".
Virtually nothing in-tree uses it any more.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoDriver core: Make platform_device_add_data accept a const pointer
Scott Wood [Mon, 4 Dec 2006 22:57:19 +0000 (14:57 -0800)]
Driver core: Make platform_device_add_data accept a const pointer

platform_device_add_data() makes a copy of the data that is given to it,
and thus the parameter can be const.  This removes a warning when data
from get_property() on powerpc is handed to platform_device_add_data(),
as get_property() returns a const pointer.

Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoDriver core: "platform_driver_probe() can save codespace": save codespace
Andrew Morton [Mon, 4 Dec 2006 22:56:36 +0000 (14:56 -0800)]
Driver core: "platform_driver_probe() can save codespace": save codespace

This function can be __init

Cc: David Brownell <dbrownell@users.sourceforge.net>
Cc: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoDebugFS : file/directory removal fix
Mathieu Desnoyers [Fri, 24 Nov 2006 18:51:14 +0000 (13:51 -0500)]
DebugFS : file/directory removal fix

Fix file and directory removal in debugfs. Add inotify support for file removal.

The following scenario :
create dir a
create dir a/b

cd a/b (some process goes in cwd a/b)

rmdir a/b
rmdir a

fails due to the fact that "a" appears to be non empty. It is because
the "b" dentry is not deleted from "a" and still in use. The same
problem happens if "b" is a file. d_delete is nice enough to know when
it needs to unhash and free the dentry if nothing else is using it or,
if someone is using it, to remove it from the hash queues and wait for
it to be deleted when it has no users.

The nice side-effect of this fix is that it calls the file removal
notification.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoDebugFS : more file/directory creation error handling
Mathieu Desnoyers [Fri, 24 Nov 2006 18:50:09 +0000 (13:50 -0500)]
DebugFS : more file/directory creation error handling

Correct dentry count to handle creation errors.

This patch puts a dput at the file creation instead of the file removal :
lookup_one_len already returns a dentry with reference count of 1. Then,
the dget() in simple_mknod increments it when the dentry is associated
with a file. In a scenario where simple_create or simple_mkdir returns
an error, this would lead to an unwanted increment of the reference
counter, therefore making file removal impossible.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoDebugFS : file/directory creation error handling
Mathieu Desnoyers [Fri, 24 Nov 2006 18:48:19 +0000 (13:48 -0500)]
DebugFS : file/directory creation error handling

Fix error handling of file and directory creation in DebugFS.

The error path should release the file system because no _remove will be called
for this erroneous creation.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoDebugFS : coding style fixes
Mathieu Desnoyers [Fri, 24 Nov 2006 18:46:30 +0000 (13:46 -0500)]
DebugFS : coding style fixes

Minor coding style fixes along the way : 80 cols and a white space.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoDebugFS : inotify create/mkdir support
Mathieu Desnoyers [Fri, 24 Nov 2006 18:45:37 +0000 (13:45 -0500)]
DebugFS : inotify create/mkdir support

Add inotify create and mkdir events to DebugFS.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agodriver core: delete virtual directory on class_unregister()
Akinobu Mita [Mon, 20 Nov 2006 19:53:18 +0000 (04:53 +0900)]
driver core: delete virtual directory on class_unregister()

Class virtual directory is created as the need arises.
But it is not deleted when the class is unregistered.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoDriver core: show "initstate" of module
Kay Sievers [Fri, 24 Nov 2006 11:15:25 +0000 (12:15 +0100)]
Driver core: show "initstate" of module

Show the initialization state(live, coming, going) of the module:
  $ cat /sys/module/usbcore/initstate
  live

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years ago[ARM] Merge AT91 and devel branches
Russell King [Wed, 13 Dec 2006 22:44:15 +0000 (22:44 +0000)]
[ARM] Merge AT91 and devel branches

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
17 years ago[ARM] 4017/1: [Jornada7xx] - Updating Jornada720.c
Kristoffer Ericson [Wed, 13 Dec 2006 20:32:08 +0000 (21:32 +0100)]
[ARM] 4017/1: [Jornada7xx] - Updating Jornada720.c

* HP Jornada 720 uses epson 1356 chip for graphics. This chip is compatible with s1d13xxxfb driver.

* HP Jornada 720 uses a Microprocessor Control Unit to talk to various
hardware. We add it as a platform device in jornada720_init()

* We provide pm_suspend() to avoid unresolved symbols in apm.o. We are
unable to truly suspend now, hence the stub.

* Speaker/microphone enabling got removed because it will be placed in the alsa driver.

Signed-off-by: Filip Zyzniewski <(address hidden)>
Signed-off-by: Kristoffer Ericson <(address hidden)>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
17 years ago[IA64] Move sg_dma_{len,address} from pci.h to scatterlist.h
Tony Luck [Wed, 13 Dec 2006 21:15:10 +0000 (13:15 -0800)]
[IA64] Move sg_dma_{len,address} from pci.h to scatterlist.h

IA64 is in a tiny minority providing these defines in pci.h.
Almost everyone else has them in scatterlist.h

Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years ago[ARM] 3992/1: i.MX/MX1 CPU Frequency scaling support
Pavel Pisa [Wed, 6 Dec 2006 16:25:04 +0000 (17:25 +0100)]
[ARM] 3992/1: i.MX/MX1 CPU Frequency scaling support

Support to change MX1 CPU frequency at runtime.
Tested on PiKRON's PiMX1 board and seems to be fully
stable up to 200 MHz end even as low as 8 MHz.

Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
17 years ago[ARM] Provide a method to alter the control register
Russell King [Wed, 13 Dec 2006 18:33:53 +0000 (18:33 +0000)]
[ARM] Provide a method to alter the control register

i.MX needs to tweak the control register to support CPU frequency
scaling.  Rather than have folk blindly try and change the control
register by writing to it and then wondering why it doesn't work,
provide a method (which is safe for UP only, and therefore only
available for UP) to achieve this.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
17 years ago[ARM] 4016/1: prefetch macro is wrong wrt gcc's "delete-null-pointer-checks"
Nicolas Pitre [Wed, 13 Dec 2006 17:39:26 +0000 (18:39 +0100)]
[ARM] 4016/1: prefetch macro is wrong wrt gcc's "delete-null-pointer-checks"
 optimization

The gcc manual says:

|`-fdelete-null-pointer-checks'
|     Use global dataflow analysis to identify and eliminate useless
|     checks for null pointers.  The compiler assumes that dereferencing
|     a null pointer would have halted the program.  If a pointer is
|     checked after it has already been dereferenced, it cannot be null.
|     Enabled at levels `-O2', `-O3', `-Os'.

Now the problem can be seen with this test case:

#include <linux/prefetch.h>
extern void bar(char *x);
void foo(char *x)
{
prefetch(x);
if (x)
bar(x);
}

Because the constraint to the inline asm used in the prefetch() macro is
a memory operand, gcc assumes that the asm code does dereference the
pointer and the delete-null-pointer-checks optimization kicks in.
Inspection of generated assembly for the above example shows that bar()
is indeed called unconditionally without any test on the value of x.

Of course in the prefetch case there is no real dereference and it
cannot be assumed that a null pointer would have been caught at that
point. This causes kernel oopses with constructs like
hlist_for_each_entry() where the list's 'next' content is prefetched
before the pointer is tested against NULL, and only when gcc feels like
applying this optimization which doesn't happen all the time with more
complex code.

It appears that the way to prevent delete-null-pointer-checks
optimization to occur in this case is to make prefetch() into a static
inline function instead of a macro. At least this is what is done on
x86_64 where a similar inline asm memory operand is used (I presume they
would have seen the same problem if it didn't work) and resulting code
for the above example confirms that.

An alternative would consist of replacing the memory operand by a
register operand containing the pointer, and use the addressing mode
explicitly in the asm template. But that would be less optimal than an
offsettable memory reference.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
17 years ago[PATCH] Add missing KORENIX PCI ID's
Russell King [Wed, 13 Dec 2006 17:56:11 +0000 (17:56 +0000)]
[PATCH] Add missing KORENIX PCI ID's

Oops, sorry about that.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Optimize D-cache alias handling on fork
Ralf Baechle [Tue, 12 Dec 2006 17:14:57 +0000 (17:14 +0000)]
[PATCH] Optimize D-cache alias handling on fork

Virtually index, physically tagged cache architectures can get away
without cache flushing when forking.  This patch adds a new cache
flushing function flush_cache_dup_mm(struct mm_struct *) which for the
moment I've implemented to do the same thing on all architectures
except on MIPS where it's a no-op.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] MIPS: Fix COW D-cache aliasing on fork
Atsushi Nemoto [Tue, 12 Dec 2006 17:14:56 +0000 (17:14 +0000)]
[PATCH] MIPS: Fix COW D-cache aliasing on fork

Provide a custom copy_user_highpage() to deal with aliasing issues on
MIPS.  It uses kmap_coherent() to map an user page for kernel with same
color.  Rewrite copy_to_user_page() and copy_from_user_page() with the
new interfaces to avoid extra cache flushing.

The main part of this patch was originally written by Ralf Baechle;
Atushi Nemoto did the the debugging.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Pass vma argument to copy_user_highpage().
Atsushi Nemoto [Tue, 12 Dec 2006 17:14:55 +0000 (17:14 +0000)]
[PATCH] Pass vma argument to copy_user_highpage().

To allow a more effective copy_user_highpage() on certain architectures,
a vma argument is added to the function and cow_user_page() allowing
the implementation of these functions to check for the VM_EXEC bit.

The main part of this patch was originally written by Ralf Baechle;
Atushi Nemoto did the the debugging.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Fix COW D-cache aliasing on fork
Atsushi Nemoto [Tue, 12 Dec 2006 17:14:54 +0000 (17:14 +0000)]
[PATCH] Fix COW D-cache aliasing on fork

Problem:

1. There is a process containing two thread (T1 and T2).  The
   thread T1 calls fork().  Then dup_mmap() function called on T1 context.

static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
...
flush_cache_mm(current->mm);
... /* A */
(write-protect all Copy-On-Write pages)
... /* B */
flush_tlb_mm(current->mm);
...

2. When preemption happens between A and B (or on SMP kernel), the
   thread T2 can run and modify data on COW pages without page fault
   (modified data will stay in cache).

3. Some time after fork() completed, the thread T2 may cause a page
   fault by write-protect on a COW page.

4. Then data of the COW page will be copied to newly allocated
   physical page (copy_cow_page()).  It reads data via kernel mapping.
   The kernel mapping can have different 'color' with user space
   mapping of the thread T2 (dcache aliasing).  Therefore
   copy_cow_page() will copy stale data.  Then the modified data in
   cache will be lost.

In order to allow architecture code to deal with this problem allow
architecture code to override copy_user_highpage() by defining
__HAVE_ARCH_COPY_USER_HIGHPAGE in <asm/page.h>.

The main part of this patch was originally written by Ralf Baechle;
Atushi Nemoto did the the debugging.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Add support for Korenix 16C950-based PCI cards
Russell King [Wed, 13 Dec 2006 14:45:46 +0000 (14:45 +0000)]
[PATCH] Add support for Korenix 16C950-based PCI cards

This adds initial support to 8250-pci for the Korenix Jetcard PCI serial
cards.  The JC12xx cards are standard RS232-based serial cards utilising
the Oxford 16C950 device.

The JC14xx are RS422/RS485-based cards, but in order for these to be
supported natively, we will need additional tweaks to the 8250 layers so
we can specify some values for the 950's registers.  Hence, these two
entries are commented out.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoMerge branch 'for-linus' of git://brick.kernel.dk/data/git/linux-2.6-block
Linus Torvalds [Wed, 13 Dec 2006 17:15:34 +0000 (09:15 -0800)]
Merge branch 'for-linus' of git://brick.kernel.dk/data/git/linux-2.6-block

* 'for-linus' of git://brick.kernel.dk/data/git/linux-2.6-block:
  [PATCH] Fixup cciss error handling
  [PATCH] Allow as-iosched to be unloaded
  [PATCH 2/2] cciss: remove calls to pci_disable_device
  [PATCH 1/2] cciss: map out more memory for config table
  [PATCH] Propagate down request sync flag

Resolve trivial whitespace conflict in drivers/block/cciss.c manually.

17 years agoMerge branch 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6
Linus Torvalds [Wed, 13 Dec 2006 17:13:19 +0000 (09:13 -0800)]
Merge branch 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6

* 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6:
  hwmon: Add MAINTAINERS entry for new ams driver
  hwmon: New AMS hardware monitoring driver
  hwmon/w83793: Add documentation and maintainer
  hwmon: New Winbond W83793 hardware monitoring driver
  hwmon: Update Rudolf Marek's e-mail address
  hwmon/f71805f: Fix the device address decoding
  hwmon/f71805f: Always create all fan inputs
  hwmon/f71805f: Add support for the Fintek F71872F/FG chip
  hwmon: New PC87427 hardware monitoring driver
  hwmon/it87: Remove the SMBus interface support
  hwmon/hdaps: Update the list of supported devices
  hwmon/hdaps: Move the DMI detection data to .data
  hwmon/pc87360: Autodetect the VRM version
  hwmon/f71805f: Document the fan control features
  hwmon/f71805f: Add support for "speed mode" fan speed control
  hwmon/f71805f: Support DC fan speed control mode
  hwmon/f71805f: Let the user adjust the PWM base frequency
  hwmon/f71805f: Add manual fan speed control
  hwmon/f71805f: Store the fan control registers

17 years ago[PATCH] getting rid of all casts of k[cmz]alloc() calls
Robert P. J. Day [Wed, 13 Dec 2006 08:35:56 +0000 (00:35 -0800)]
[PATCH] getting rid of all casts of k[cmz]alloc() calls

Run this:

#!/bin/sh
for f in $(grep -Erl "\([^\)]*\) *k[cmz]alloc" *) ; do
  echo "De-casting $f..."
  perl -pi -e "s/ ?= ?\([^\)]*\) *(k[cmz]alloc) *\(/ = \1\(/" $f
done

And then go through and reinstate those cases where code is casting pointers
to non-pointers.

And then drop a few hunks which conflicted with outstanding work.

Cc: Russell King <rmk@arm.linux.org.uk>, Ian Molton <spyro@f2s.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Greg KH <greg@kroah.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Paul Fulghum <paulkf@microgate.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Karsten Keil <kkeil@suse.de>
Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Cc: Ian Kent <raven@themaw.net>
Cc: Steven French <sfrench@us.ibm.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Neil Brown <neilb@cse.unsw.edu.au>
Cc: Jaroslav Kysela <perex@suse.cz>
Cc: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] sstfb: add sysfs interface
Helge Deller [Wed, 13 Dec 2006 08:35:55 +0000 (00:35 -0800)]
[PATCH] sstfb: add sysfs interface

Modify the sstfb (Voodoo1/2) driver:

- fix a memleak when removing the sstfb module

- fix sstfb to use the fbdev default videomode database

- add module option "mode_option" to set initial screen mode

- add sysfs-interface to turn VGA-passthrough on/off via
  /sys/class/graphics/fbX/vgapass

- remove old debug functions from ioctl interface

Signed-off-by: Helge Deller <deller@gmx.de>
Acked-By: James Simmons <jsimmons@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] fbdev: remove references to non-existent fbmon_valid_timings()
Geert Uytterhoeven [Wed, 13 Dec 2006 08:35:54 +0000 (00:35 -0800)]
[PATCH] fbdev: remove references to non-existent fbmon_valid_timings()

Remove references to non-existent fbmon_valid_timings()

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: James Simmons <jsimmons@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] HPT37x: read f_CNT saved by BIOS from port
Sergei Shtylyov [Wed, 13 Dec 2006 08:35:53 +0000 (00:35 -0800)]
[PATCH] HPT37x: read f_CNT saved by BIOS from port

The undocumented register BIOS uses for saving f_CNT seems to only be
mapped to I/O space while all the other HPT3xx regs are dual-mapped.  Looks
like another HighPoint's dirty trick.  With this patch, the deadly kernel
oops on the cards having the modern HighPoint BIOSes is now at last gone!

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ide: HPT3xx: fix PCI clock detection
Sergei Shtylyov [Wed, 13 Dec 2006 08:35:52 +0000 (00:35 -0800)]
[PATCH] ide: HPT3xx: fix PCI clock detection

Use the f_CNT value saved by the HighPoint BIOS if available as reading it
directly would give us a wrong PCI frequency after DPLL has already been
calibrated by BIOS.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ide: fix the case of multiple HPT3xx chips present
Sergei Shtylyov [Wed, 13 Dec 2006 08:35:51 +0000 (00:35 -0800)]
[PATCH] ide: fix the case of multiple HPT3xx chips present

init_chipset_hpt366() modifies some fields of the ide_pci_device_t structure
depending on the chip's revision, so pass it a copy of the structure to avoid
issues when multiple different chips are present.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ide: fix HPT3xx hotswap support
Sergei Shtylyov [Wed, 13 Dec 2006 08:35:50 +0000 (00:35 -0800)]
[PATCH] ide: fix HPT3xx hotswap support

Fix the broken hotswap code: on HPT37x it caused RESET- to glitch when
tristating the bus (the MISC control 3/6 and soft control 2 need to be written
to in the certain order), and for HPT36x the obsolete HDIO_TRISTATE_HWIF
ioctl() handler was called instead which treated the state argument wrong.
Also, get rid of the soft control reg.  1 wtite to enable IDE interrupt --
this is done in init_hpt37x() already...

Have been tested on HPT370 and 371N.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ide: optimize HPT37x timing tables
Sergei Shtylyov [Wed, 13 Dec 2006 08:35:49 +0000 (00:35 -0800)]
[PATCH] ide: optimize HPT37x timing tables

Save some space on the timing tables by introducing the separate transfer mode
table in which the mode lookup is done to get the index into the timing table
itself.  Get rid of the rest of the obsolete/duplicate tables and use one set
of tables for the whole HPT37x chip family like the HighPoint open-source
drivers do.  Documnent the different timing register layout for the HPT36x
chip family (this is my guesswork based on the timing values).

Have been tested and works fine on HPT370/302/371N.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ide: fix HPT37x timing tables
Sergei Shtylyov [Wed, 13 Dec 2006 08:35:49 +0000 (00:35 -0800)]
[PATCH] ide: fix HPT37x timing tables

Fix/remove bad/unused timing tables: HPT370/A 66 MHz tables weren't really
needed (the chips are not UltraATA/133 capable and shouldn't support 66 MHz
PCI) and had many modes over- and underclocked, HPT372 33 MHz table was in
fact for 66 MHz and 50 MHz table missed UltraDMA mode 6, HPT374 33 MHz table
was really for 50 MHz...  (Actually, HPT370/A 33 MHz tables also have issues.
e.g.  HPT370 has PIO modes 0/1 overlocked.)

There's also no need in the separate HPT374 tables because HPT372 timings
should be the same (and those tables has UltraDMA mode 6 which HPT374 supports
depending on HPT374_ALLOW_ATA133_6 #define)...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ide: HPT3xxN clocking fixes
Sergei Shtylyov [Wed, 13 Dec 2006 08:35:47 +0000 (00:35 -0800)]
[PATCH] ide: HPT3xxN clocking fixes

Fix serious problems with the HPT372N clock turnaround code:

- the wrong ports were written to when called for the secondary channel;

- it didn't serialize access to the channels;

- turnaround shou;dn't be done on 66 MHz PCI;

- caching the clock mode per-channel caused it to get out of sync with the
  actual register value.

Additionally, avoid calibrating PLL twice (for each channel) as the second try
results in a wrong PCI frequency and thus in the wrong timings.

Make the driver deal with HPT302N and HPT371N correctly -- the clocking and
(seemingly) a need for clock tunaround is the same as for HPT372N.  HPT371/N
chips have only one, secondary channel, so avoid touching their "pure virtual"
primary channel, and disable it if the BIOS haven't done this already.

Also, while at it, disable UltraATA/133 for HPT372 by default -- 50 MHz DPLL
clock don't allow for this speed anyway.  And remove the traces of the former
bad patch that wasn't even applicable to this version of driver.

Has been tested on HPT370/371N, unfortunately I don't have an instant access
to the other chips...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Optimize calc_load()
Eric Dumazet [Wed, 13 Dec 2006 08:35:45 +0000 (00:35 -0800)]
[PATCH] Optimize calc_load()

calc_load() is called by timer interrupt to update avenrun[].  It currently
calls nr_active() at each timer tick (HZ per second), while the update of
avenrun[] is done only once every 5 seconds.  (LOAD_FREQ=5 Hz)

nr_active() is quite expensive on SMP machines, since it has to sum up
nr_running and nr_uninterruptible of all online CPUS, bringing foreign
dirty cache lines.

This patch is an optimization of calc_load() so that nr_active() is called
only if we need it.

The use of unlikely() is welcome since the condition is true only once every
5*HZ time.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Cc: Ingo Molnar <mingo@elte.hu>
Acked-by: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: Fix up some bit-rot in exp_export
NeilBrown [Wed, 13 Dec 2006 08:35:45 +0000 (00:35 -0800)]
[PATCH] knfsd: Fix up some bit-rot in exp_export

The nfsservctl system call isn't used but recent nfs-utils releases for
exporting filesystems, and consequently the code that is uses - exp_export -
has suffered some bitrot.

Particular:
  - some newly added fields in 'struct svc_export' are being initialised
    properly.
  - the return value is now always -ENOMEM ...

This patch fixes both these problems.

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: Don't ignore kstrdup failure in rpc caches
NeilBrown [Wed, 13 Dec 2006 08:35:44 +0000 (00:35 -0800)]
[PATCH] knfsd: Don't ignore kstrdup failure in rpc caches

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: simplify filehandle check
J.Bruce Fields [Wed, 13 Dec 2006 08:35:43 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: simplify filehandle check

Kill another big "if" clause.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: simplify migration op check
J.Bruce Fields [Wed, 13 Dec 2006 08:35:39 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: simplify migration op check

I'm not too fond of these big if conditions.  Replace them by checks of a flag
in the operation descriptor.  To my eye this makes the code a bit more
self-documenting, and makes the complicated part of the code (proc_compound) a
little more compact.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: reorganize compound ops
J.Bruce Fields [Wed, 13 Dec 2006 08:35:38 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: reorganize compound ops

Define an op descriptor struct, use it to simplify nfsd4_proc_compound().

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: make verify and nverify wrappers
J.Bruce Fields [Wed, 13 Dec 2006 08:35:31 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: make verify and nverify wrappers

Make wrappers for verify and nverify, for consistency with other ops.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: don't inline nfsd4 compound op functions
J.Bruce Fields [Wed, 13 Dec 2006 08:35:29 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: don't inline nfsd4 compound op functions

The inlining contributes to bloating the stack of nfsd4_compound, and I want
to change the compound op functions to function pointers anyway.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: move replay_owner to cstate
J.Bruce Fields [Wed, 13 Dec 2006 08:35:28 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: move replay_owner to cstate

Tuck away the replay_owner in the cstate while we're at it.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: remove spurious replay_owner check
J.Bruce Fields [Wed, 13 Dec 2006 08:35:28 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: remove spurious replay_owner check

OK, this is embarassing--I've even looked back at the history, and cannot for
the life of me figure out why I added this check.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: pass saved and current fh together into nfsd4 operations
J.Bruce Fields [Wed, 13 Dec 2006 08:35:27 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: pass saved and current fh together into nfsd4 operations

Pass the saved and current filehandles together into all the nfsd4 compound
operations.

I want a unified interface to these operations so we can just call them by
pointer and throw out the huge switch statement.

Also I'll eventually want a structure like this--that holds the state used
during compound processing--for deferral.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: svcrpc: remove another silent drop from deferral code
J.Bruce Fields [Wed, 13 Dec 2006 08:35:26 +0000 (00:35 -0800)]
[PATCH] knfsd: svcrpc: remove another silent drop from deferral code

There's no point deferring something just to immediately fail the deferral,
especially now that we can do something more useful in the failure case by
returning an error.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd: don't drop silently on upcall deferral
J.Bruce Fields [Wed, 13 Dec 2006 08:35:25 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd: don't drop silently on upcall deferral

To avoid tying up server threads when nfsd makes an upcall (to mountd, to get
export options, to idmapd, for nfsv4 name<->id mapping, etc.), we temporarily
"drop" the request and save enough information so that we can revisit it
later.

Certain failures during the deferral process can cause us to really drop the
request and never revisit it.

This is often less than ideal, and is unacceptable in the NFSv4 case--rfc 3530
forbids the server from dropping a request without also closing the
connection.

As a first step, we modify the deferral code to return -ETIMEDOUT (which is
translated to nfserr_jukebox in the v3 and v4 cases, and remains a drop in the
v2 case).

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: handling more nfsd_cross_mnt errors in nfsd4 readdir
J.Bruce Fields [Wed, 13 Dec 2006 08:35:24 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: handling more nfsd_cross_mnt errors in nfsd4 readdir

This patch on its own causes no change in behavior, since nfsd_cross_mnt()
only returns -EAGAIN; but in the future I'd like it to also be able to return
-ETIMEDOUT, so we may as well handle any possible error here.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd: simplify exp_pseudoroot
J.Bruce Fields [Wed, 13 Dec 2006 08:35:23 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd: simplify exp_pseudoroot

Note there's no need for special handling of -EAGAIN here; nfserrno() does
what we want already.  So this is a pure cleanup with no change in
functionality.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd: make exp_rootfh handle exp_parent errors
J.Bruce Fields [Wed, 13 Dec 2006 08:35:21 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd: make exp_rootfh handle exp_parent errors

Since exp_parent can fail by returning an error (-EAGAIN) in addition to by
returning NULL, we should check for that case in exp_rootfh.

(TODO: we should check that userland handles these errors too.)

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: clarify units of COMPOUND_SLACK_SPACE
J.Bruce Fields [Wed, 13 Dec 2006 08:35:20 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: clarify units of COMPOUND_SLACK_SPACE

A comment here incorrectly states that "slack_space" is measured in words, not
bytes.  Remove the comment, and adjust a variable name and a few comments to
clarify the situation.

This is pure cleanup; there should be no change in functionality.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: svcrpc: fix gss krb5i memory leak
J.Bruce Fields [Wed, 13 Dec 2006 08:35:19 +0000 (00:35 -0800)]
[PATCH] knfsd: svcrpc: fix gss krb5i memory leak

The memory leak here is embarassingly obvious.

This fixes a problem that causes the kernel to leak a small amount of memory
every time it receives a integrity-protected request.

Thanks to Aim Le Rouzic for the bug report.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: remove a dprink from nfsd4_lock
J.Bruce Fields [Wed, 13 Dec 2006 08:35:18 +0000 (00:35 -0800)]
[PATCH] knfsd: nfsd4: remove a dprink from nfsd4_lock

This dprintk is printing the wrong error now, but it's probably an unnecessary
dprintk anyway; just remove it.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] remove the broken BLK_DEV_SWIM_IOP driver
Adrian Bunk [Wed, 13 Dec 2006 08:35:17 +0000 (00:35 -0800)]
[PATCH] remove the broken BLK_DEV_SWIM_IOP driver

The BLK_DEV_SWIM_IOP driver has:
- already been marked as BROKEN in 2.6.0 three years ago and
- is still marked as BROKEN.

Drivers that had been marked as BROKEN for such a long time seem to be
unlikely to be revived in the forseeable future.

But if anyone wants to ever revive this driver, the code is still
present in the older kernel releases.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: Jens Axboe <jens.axboe@oracle.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] one more EXPORT_UNUSED_SYMBOL removal
Adrian Bunk [Wed, 13 Dec 2006 08:35:16 +0000 (00:35 -0800)]
[PATCH] one more EXPORT_UNUSED_SYMBOL removal

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] update Tigran's email addresses
Tigran Aivazian [Wed, 13 Dec 2006 08:35:14 +0000 (00:35 -0800)]
[PATCH] update Tigran's email addresses

As Adrian pointed out recently, there were still a couple of places where
I should have fixed my email address.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ncpfs: ensure we free wdog_pid on parse_option or fill_inode failure
Eric W. Biederman [Wed, 13 Dec 2006 08:35:13 +0000 (00:35 -0800)]
[PATCH] ncpfs: ensure we free wdog_pid on parse_option or fill_inode failure

This took a little refactoring but now errors are handled cleanly.  When
this code used pid_t values this wasn't necessary because you can't
leak a pid_t.

Thanks to Peter Vandrovec for spotting this.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Peter Vandrovec <vandrove@vc.cvut.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ncpfs: Use struct pid to track the userspace watchdog process
Eric W. Biederman [Wed, 13 Dec 2006 08:35:11 +0000 (00:35 -0800)]
[PATCH] ncpfs: Use struct pid to track the userspace watchdog process

This patch converts the tracking of the user space watchdog process from using
a pid_t to use struct pid.  This makes us safe from pid wrap around issues and
prepares the way for the pid namespace.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Petr Vandrovec <VANDROVE@vc.cvut.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] smbfs: Make conn_pid a struct pid
Eric W. Biederman [Wed, 13 Dec 2006 08:35:10 +0000 (00:35 -0800)]
[PATCH] smbfs: Make conn_pid a struct pid

smbfs keeps track of the user space server process in conn_pid.  This converts
that track to use a struct pid instead of pid_t.  This keeps us safe from pid
wrap around issues and prepares the way for the pid namespace.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] n_r3964: Use struct pid to track user space clients
Eric W. Biederman [Wed, 13 Dec 2006 08:35:10 +0000 (00:35 -0800)]
[PATCH] n_r3964: Use struct pid to track user space clients

Currently this driver tracks user space clients it should send signals to.  In
the presenct of file descriptor passing this is appears susceptible to
confusion from pid wrap around issues.

Replacing this with a struct pid prevents us from getting confused, and
prepares for a pid namespace implementation.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] tty_io.c balance tty_ldisc_ref()
Dan Carpenter [Wed, 13 Dec 2006 08:35:09 +0000 (00:35 -0800)]
[PATCH] tty_io.c balance tty_ldisc_ref()

tty_ldisc_deref() should only be called when tty_ldisc_ref() succeeds
otherwise it triggers a BUG().  There's already a function
tty_ldisc_flush() that flushes properly.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rtc framewok: rtc_wkalrm.enabled reporting updates
David Brownell [Wed, 13 Dec 2006 08:35:08 +0000 (00:35 -0800)]
[PATCH] rtc framewok: rtc_wkalrm.enabled reporting updates

Fix a glitch in the procfs dumping of whether the alarm IRQ is enabled: use
the traditional name (from drivers/char/rtc.c and many other places) of
"alarm_IRQ", not "alrm_wakeup" (which didn't even match the efirtc code, which
originated that reporting API).

Also, update a few of the RTC drivers to stop providing that duplicate status,
and/or to expose it properly when reporting the alarm state.  We really don't
want every RTC driver doing their own thing here...

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Acked-by: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rtc: remove syslog spam on registration
David Brownell [Wed, 13 Dec 2006 08:35:06 +0000 (00:35 -0800)]
[PATCH] rtc: remove syslog spam on registration

This removes some syslog spam as RTC drivers register; debug messages
shouldn't come out at "info" level.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Acked-by: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] RTC driver init adjustment
Jan Beulich [Wed, 13 Dec 2006 08:35:05 +0000 (00:35 -0800)]
[PATCH] RTC driver init adjustment

- conditionalizes procfs code upon CONFIG_PROC_FS (to reduce code size when
  that option is not enabled)

- make initialization no longer fail when the procfs entry can't be
  allocated (namely would initialization always have failed when
  CONFIG_PROC_FS was not set)

- move the formerly file-scope static variable rtc_int_handler_ptr into
  the only function using it, and makes it automatic.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rtc: fx error case
Jan Beulich [Wed, 13 Dec 2006 08:35:04 +0000 (00:35 -0800)]
[PATCH] rtc: fx error case

Ensure RTC driver doesn't use its timer when it doesn't get to set it up
(as it cannot currently prevent other of its functions to be called from
outside when not built as a module - probably this should also be
addressed).

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockd endianness annotations
Al Viro [Wed, 13 Dec 2006 08:35:03 +0000 (00:35 -0800)]
[PATCH] lockd endianness annotations

Annotated, all places switched to keeping status net-endian.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] hci endianness annotations
Al Viro [Wed, 13 Dec 2006 08:35:01 +0000 (00:35 -0800)]
[PATCH] hci endianness annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] missing includes in hilkbd
Al Viro [Wed, 13 Dec 2006 08:35:01 +0000 (00:35 -0800)]
[PATCH] missing includes in hilkbd

Now that it's built on m68k too...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] uml problems with linux/io.h
Al Viro [Wed, 13 Dec 2006 08:35:00 +0000 (00:35 -0800)]
[PATCH] uml problems with linux/io.h

Remove useless includes of linux/io.h, don't even try to build iomap_copy
on uml (it doesn't have readb() et.al., so...)

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] appldata_mem dependes on vm counters
Al Viro [Wed, 13 Dec 2006 08:34:59 +0000 (00:34 -0800)]
[PATCH] appldata_mem dependes on vm counters

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] CONFIG_COMPUTONE should depend on ISA|EISA|PCI
Al Viro [Wed, 13 Dec 2006 08:34:58 +0000 (00:34 -0800)]
[PATCH] CONFIG_COMPUTONE should depend on ISA|EISA|PCI

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] tty: remove useless memory barrier
Ralf Baechle [Wed, 13 Dec 2006 08:34:57 +0000 (00:34 -0800)]
[PATCH] tty: remove useless memory barrier

I don't see why there is a memory barrier in copy_from_read_buf() at all.
Even if it was useful spin_unlock_irqrestore implies a barrier.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Fix numerous kcalloc() calls, convert to kzalloc()
Robert P. J. Day [Wed, 13 Dec 2006 08:34:52 +0000 (00:34 -0800)]
[PATCH] Fix numerous kcalloc() calls, convert to kzalloc()

All kcalloc() calls of the form "kcalloc(1,...)" are converted to the
equivalent kzalloc() calls, and a few kcalloc() calls with the incorrect
ordering of the first two arguments are fixed.

Signed-off-by: Robert P. J. Day <rpjday@mindspring.com>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Cc: Adam Belay <ambx1@neo.rr.com>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Cc: Greg KH <greg@kroah.com>
Cc: Mark Fasheh <mark.fasheh@oracle.com>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Use activate_mm() in fs/aio.c:use_mm()
Jeremy Fitzhardinge [Wed, 13 Dec 2006 08:34:49 +0000 (00:34 -0800)]
[PATCH] Use activate_mm() in fs/aio.c:use_mm()

activate_mm() is not the right thing to be using in use_mm().  It should be
switch_mm().

On normal x86, they're synonymous, but for the Xen patches I'm adding a
hook which assumes that activate_mm is only used the first time a new mm
is used after creation (I have another hook for dealing with dup_mm).  I
think this use of activate_mm() is the only place where it could be used
a second time on an mm.

>From a quick look at the other architectures I think this is OK (most
simply implement one in terms of the other), but some are doing some
subtly different stuff between the two.

Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockdep: fix possible races while disabling lock-debugging
Ingo Molnar [Wed, 13 Dec 2006 08:34:43 +0000 (00:34 -0800)]
[PATCH] lockdep: fix possible races while disabling lock-debugging

Jarek Poplawski noticed that lockdep global state could be accessed in a
racy way if one CPU did a lockdep assert (shutting lockdep down), while the
other CPU would try to do something that changes its global state.

This patch fixes those races and cleans up lockdep's internal locking by
adding a graph_lock()/graph_unlock()/debug_locks_off_graph_unlock helpers.

(Also note that as we all know the Linux kernel is, by definition, bug-free
and perfect, so this code never triggers, so these fixes are highly
theoretical.  I wrote this patch for aesthetic reasons alone.)

[akpm@osdl.org: build fix]
[jarkao2@o2.pl: build fix's refix]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Jarek Poplawski <jarkao2@o2.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockdep: print irq-trace info on asserts
Ingo Molnar [Wed, 13 Dec 2006 08:34:43 +0000 (00:34 -0800)]
[PATCH] lockdep: print irq-trace info on asserts

When we print an assert due to scheduling-in-atomic bugs, and if lockdep
is enabled, then the IRQ tracing information of lockdep can be printed
to pinpoint the code location that disabled interrupts. This saved me
quite a bit of debugging time in cases where the backtrace did not
identify the irq-disabling site well enough.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockdep: use chain hash on CONFIG_DEBUG_LOCKDEP too
Ingo Molnar [Wed, 13 Dec 2006 08:34:42 +0000 (00:34 -0800)]
[PATCH] lockdep: use chain hash on CONFIG_DEBUG_LOCKDEP too

CONFIG_DEBUG_LOCKDEP is unacceptably slow because it does not utilize
the chain-hash. Turn the chain-hash back on in this case too.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockdep: clean up VERY_VERBOSE define
Ingo Molnar [Wed, 13 Dec 2006 08:34:41 +0000 (00:34 -0800)]
[PATCH] lockdep: clean up VERY_VERBOSE define

Cleanup: the VERY_VERBOSE define was unnecessarily dependent on #ifdef VERBOSE
- while the VERBOSE switch is 0 or 1 (always defined).

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockdep: improve lockdep_reset()
Ingo Molnar [Wed, 13 Dec 2006 08:34:40 +0000 (00:34 -0800)]
[PATCH] lockdep: improve lockdep_reset()

Clear all the chains during lockdep_reset().  This fixes some locking-selftest
false positives i saw on -rt.  (never saw those on mainline though, but it
could happen.)

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockdep: improve verbose messages
Ingo Molnar [Wed, 13 Dec 2006 08:34:40 +0000 (00:34 -0800)]
[PATCH] lockdep: improve verbose messages

Make verbose lockdep messages (off by default) more informative by printing
out the hash chain key.  (this patch was what helped me catch the earlier
lockdep hash-collision bug)

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockdep: filter off by default
Ingo Molnar [Wed, 13 Dec 2006 08:34:39 +0000 (00:34 -0800)]
[PATCH] lockdep: filter off by default

Fix typo in the class_filter() function.  (filtering is not used by default so
this only affects lockdep-internal debugging cases)

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] debug: add sysrq_always_enabled boot option
Ingo Molnar [Wed, 13 Dec 2006 08:34:36 +0000 (00:34 -0800)]
[PATCH] debug: add sysrq_always_enabled boot option

Most distributions enable sysrq support but set it to 0 by default.  Add a
sysrq_always_enabled boot option to always-enable sysrq keys.  Useful for
debugging - without having to modify the disribution's config files (which
might not be possible if the kernel is on a live CD, etc.).

Also, while at it, clean up the sysrq interfaces.

[bunk@stusta.de: make sysrq_always_enabled_setup() static]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] optimize o_direct on block devices
Chen, Kenneth W [Wed, 13 Dec 2006 08:34:36 +0000 (00:34 -0800)]
[PATCH] optimize o_direct on block devices

Implement block device specific .direct_IO method instead of going through
generic direct_io_worker for block device.

direct_io_worker() is fairly complex because it needs to handle O_DIRECT on
file system, where it needs to perform block allocation, hole detection,
extents file on write, and tons of other corner cases.  The end result is
that it takes tons of CPU time to submit an I/O.

For block device, the block allocation is much simpler and a tight triple
loop can be written to iterate each iovec and each page within the iovec in
order to construct/prepare bio structure and then subsequently submit it to
the block layer.  This significantly speeds up O_D on block device.

[akpm@osdl.org: small speedup]
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Zach Brown <zach.brown@oracle.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ocfs2: relative atime support
Mark Fasheh [Wed, 13 Dec 2006 08:34:35 +0000 (00:34 -0800)]
[PATCH] ocfs2: relative atime support

Update ocfs2_should_update_atime() to understand the MNT_RELATIME flag and
to test against mtime / ctime accordingly.

[akpm@osdl.org: cleanups]
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Cc: Valerie Henson <val_henson@linux.intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] relative atime
Valerie Henson [Wed, 13 Dec 2006 08:34:34 +0000 (00:34 -0800)]
[PATCH] relative atime

Add "relatime" (relative atime) support.  Relative atime only updates the
atime if the previous atime is older than the mtime or ctime.  Like
noatime, but useful for applications like mutt that need to know when a
file has been read since it was last modified.

A corresponding patch against mount(8) is available at
http://userweb.kernel.org/~akpm/mount-relative-atime.txt

Signed-off-by: Valerie Henson <val_henson@linux.intel.com>
Cc: Mark Fasheh <mark.fasheh@oracle.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Karel Zak <kzak@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] touch_atime() cleanup
Andrew Morton [Wed, 13 Dec 2006 08:34:33 +0000 (00:34 -0800)]
[PATCH] touch_atime() cleanup

Simplify touch_atime() layout.

Cc: Valerie Henson <val_henson@linux.intel.com>
Cc: Mark Fasheh <mark.fasheh@oracle.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Xtensa: Add ktermios and minor filename fix
Chris Zankel [Wed, 13 Dec 2006 08:34:32 +0000 (00:34 -0800)]
[PATCH] Xtensa: Add ktermios and minor filename fix

The kernel termios (ktermios) changes were somehow missed for Xtensa.  This
patch adds the ktermios structure and also includes some minor file name
fix that was missed in the syscall patch.

Signed-off-by: Chris Zankel <chris@zankel.net>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] PM: Fix SMP races in the freezer
Rafael J. Wysocki [Wed, 13 Dec 2006 08:34:30 +0000 (00:34 -0800)]
[PATCH] PM: Fix SMP races in the freezer

Currently, to tell a task that it should go to the refrigerator, we set the
PF_FREEZE flag for it and send a fake signal to it.  Unfortunately there
are two SMP-related problems with this approach.  First, a task running on
another CPU may be updating its flags while the freezer attempts to set
PF_FREEZE for it and this may leave the task's flags in an inconsistent
state.  Second, there is a potential race between freeze_process() and
refrigerator() in which freeze_process() running on one CPU is reading a
task's PF_FREEZE flag while refrigerator() running on another CPU has just
set PF_FROZEN for the same task and attempts to reset PF_FREEZE for it.  If
the refrigerator wins the race, freeze_process() will state that PF_FREEZE
hasn't been set for the task and will set it unnecessarily, so the task
will go to the refrigerator once again after it's been thawed.

To solve first of these problems we need to stop using PF_FREEZE to tell
tasks that they should go to the refrigerator.  Instead, we can introduce a
special TIF_*** flag and use it for this purpose, since it is allowed to
change the other tasks' TIF_*** flags and there are special calls for it.

To avoid the freeze_process()-refrigerator() race we can make
freeze_process() to always check the task's PF_FROZEN flag after it's read
its "freeze" flag.  We should also make sure that refrigerator() will
always reset the task's "freeze" flag after it's set PF_FROZEN for it.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@ucw.cz>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: David Howells <dhowells@redhat.com>
Cc: Andi Kleen <ak@muc.de>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] PM: Fix freezing of stopped tasks
Rafael J. Wysocki [Wed, 13 Dec 2006 08:34:28 +0000 (00:34 -0800)]
[PATCH] PM: Fix freezing of stopped tasks

Currently, if a task is stopped (ie.  it's in the TASK_STOPPED state), it
is considered by the freezer as unfreezeable.  However, there may be a race
between the freezer and the delivery of the continuation signal to the task
resulting in the task running after we have finished freezing the other
tasks.  This, in turn, may lead to undesirable effects up to and including
data corruption.

To prevent this from happening we first need to make the freezer consider
stopped tasks as freezeable.  For this purpose we need to make freezeable()
stop returning 0 for these tasks and we need to force them to enter the
refrigerator.  However, if there's no continuation signal in the meantime,
the stopped tasks should remain stopped after all processes have been
thawed, so we need to send an additional SIGSTOP to each of them before
waking it up.

Also, a stopped task that has just been woken up should first check if
there's a freezing request for it and go to the refrigerator if that's the
case.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] SLAB: use a multiply instead of a divide in obj_to_index()
Eric Dumazet [Wed, 13 Dec 2006 08:34:27 +0000 (00:34 -0800)]
[PATCH] SLAB: use a multiply instead of a divide in obj_to_index()

When some objects are allocated by one CPU but freed by another CPU we can
consume lot of cycles doing divides in obj_to_index().

(Typical load on a dual processor machine where network interrupts are
handled by one particular CPU (allocating skbufs), and the other CPU is
running the application (consuming and freeing skbufs))

Here on one production server (dual-core AMD Opteron 285), I noticed this
divide took 1.20 % of CPU_CLK_UNHALTED events in kernel.  But Opteron are
quite modern cpus and the divide is much more expensive on oldest
architectures :

On a 200 MHz sparcv9 machine, the division takes 64 cycles instead of 1
cycle for a multiply.

Doing some math, we can use a reciprocal multiplication instead of a divide.

If we want to compute V = (A / B)  (A and B being u32 quantities)
we can instead use :

V = ((u64)A * RECIPROCAL(B)) >> 32 ;

where RECIPROCAL(B) is precalculated to ((1LL << 32) + (B - 1)) / B

Note :

I wrote pure C code for clarity. gcc output for i386 is not optimal but
acceptable :

mull   0x14(%ebx)
mov    %edx,%eax // part of the >> 32
xor     %edx,%edx // useless
mov    %eax,(%esp) // could be avoided
mov    %edx,0x4(%esp) // useless
mov    (%esp),%ebx

[akpm@osdl.org: small cleanups]
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] cpuset: rework cpuset_zone_allowed api
Paul Jackson [Wed, 13 Dec 2006 08:34:25 +0000 (00:34 -0800)]
[PATCH] cpuset: rework cpuset_zone_allowed api

Elaborate the API for calling cpuset_zone_allowed(), so that users have to
explicitly choose between the two variants:

  cpuset_zone_allowed_hardwall()
  cpuset_zone_allowed_softwall()

Until now, whether or not you got the hardwall flavor depended solely on
whether or not you or'd in the __GFP_HARDWALL gfp flag to the gfp_mask
argument.

If you didn't specify __GFP_HARDWALL, you implicitly got the softwall
version.

Unfortunately, this meant that users would end up with the softwall version
without thinking about it.  Since only the softwall version might sleep,
this led to bugs with possible sleeping in interrupt context on more than
one occassion.

The hardwall version requires that the current tasks mems_allowed allows
the node of the specified zone (or that you're in interrupt or that
__GFP_THISNODE is set or that you're on a one cpuset system.)

The softwall version, depending on the gfp_mask, might allow a node if it
was allowed in the nearest enclusing cpuset marked mem_exclusive (which
requires taking the cpuset lock 'callback_mutex' to evaluate.)

This patch removes the cpuset_zone_allowed() call, and forces the caller to
explicitly choose between the hardwall and the softwall case.

If the caller wants the gfp_mask to determine this choice, they should (1)
be sure they can sleep or that __GFP_HARDWALL is set, and (2) invoke the
cpuset_zone_allowed_softwall() routine.

This adds another 100 or 200 bytes to the kernel text space, due to the few
lines of nearly duplicate code at the top of both cpuset_zone_allowed_*
routines.  It should save a few instructions executed for the calls that
turned into calls of cpuset_zone_allowed_hardwall, thanks to not having to
set (before the call) then check (within the call) the __GFP_HARDWALL flag.

For the most critical call, from get_page_from_freelist(), the same
instructions are executed as before -- the old cpuset_zone_allowed()
routine it used to call is the same code as the
cpuset_zone_allowed_softwall() routine that it calls now.

Not a perfect win, but seems worth it, to reduce this chance of hitting a
sleeping with irq off complaint again.

Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] More slab.h cleanups
Christoph Lameter [Wed, 13 Dec 2006 08:34:24 +0000 (00:34 -0800)]
[PATCH] More slab.h cleanups

More cleanups for slab.h

1. Remove tabs from weird locations as suggested by Pekka

2. Drop the check for NUMA and SLAB_DEBUG from the fallback section
   as suggested by Pekka.

3. Uses static inline for the fallback defs as also suggested by Pekka.

4. Make kmem_ptr_valid take a const * argument.

5. Separate the NUMA fallback definitions from the kmalloc_track fallback
   definitions.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Cleanup slab headers / API to allow easy addition of new slab allocators
Christoph Lameter [Wed, 13 Dec 2006 08:34:23 +0000 (00:34 -0800)]
[PATCH] Cleanup slab headers / API to allow easy addition of new slab allocators

This is a response to an earlier discussion on linux-mm about splitting
slab.h components per allocator.  Patch is against 2.6.19-git11.  See
http://marc.theaimsgroup.com/?l=linux-mm&m=116469577431008&w=2

This patch cleans up the slab header definitions.  We define the common
functions of slob and slab in slab.h and put the extra definitions needed
for slab's kmalloc implementations in <linux/slab_def.h>.  In order to get
a greater set of common functions we add several empty functions to slob.c
and also rename slob's kmalloc to __kmalloc.

Slob does not need any special definitions since we introduce a fallback
case.  If there is no need for a slab implementation to provide its own
kmalloc mess^H^H^Hacros then we simply fall back to __kmalloc functions.
That is sufficient for SLOB.

Sort the function in slab.h according to their functionality.  First the
functions operating on struct kmem_cache * then the kmalloc related
functions followed by special debug and fallback definitions.

Also redo a lot of comments.

Signed-off-by: Christoph Lameter <clameter@sgi.com>?
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>