openwrt/staging/blogic.git
15 years agomemcg: some modification to softlimit under hierarchical memory reclaim.
KAMEZAWA Hiroyuki [Thu, 1 Oct 2009 22:44:11 +0000 (15:44 -0700)]
memcg: some modification to softlimit under hierarchical memory reclaim.

This patch clean up/fixes for memcg's uncharge soft limit path.

Problems:
  Now, res_counter_charge()/uncharge() handles softlimit information at
  charge/uncharge and softlimit-check is done when event counter per memcg
  goes over limit. Now, event counter per memcg is updated only when
  memory usage is over soft limit. Here, considering hierarchical memcg
  management, ancesotors should be taken care of.

  Now, ancerstors(hierarchy) are handled in charge() but not in uncharge().
  This is not good.

  Prolems:
  1. memcg's event counter incremented only when softlimit hits. That's bad.
     It makes event counter hard to be reused for other purpose.

  2. At uncharge, only the lowest level rescounter is handled. This is bug.
     Because ancesotor's event counter is not incremented, children should
     take care of them.

  3. res_counter_uncharge()'s 3rd argument is NULL in most case.
     ops under res_counter->lock should be small. No "if" sentense is better.

Fixes:
  * Removed soft_limit_xx poitner and checks in charge and uncharge.
    Do-check-only-when-necessary scheme works enough well without them.

  * make event-counter of memcg incremented at every charge/uncharge.
    (per-cpu area will be accessed soon anyway)

  * All ancestors are checked at soft-limit-check. This is necessary because
    ancesotor's event counter may never be modified. Then, they should be
    checked at the same time.

Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocgroup: catch bad css refcnt at css_put
KAMEZAWA Hiroyuki [Thu, 1 Oct 2009 22:44:09 +0000 (15:44 -0700)]
cgroup: catch bad css refcnt at css_put

__css_put() doesn't check a bug as refcnt goes to minus.
I think it should be caught. This patch adds a check for it.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomemcg: fix refcnt going negative
KAMEZAWA Hiroyuki [Thu, 1 Oct 2009 22:44:08 +0000 (15:44 -0700)]
memcg: fix refcnt going negative

__mem_cgroup_largest_soft_limit_node() returns a mem_cgroup_per_zone "mz"
with incremnted mz->mem->css's refcnt.  Then, the caller of this function
has to call css_put(mz->mem->css).

But, mz can be !NULL even if "not found" i.e.  without css_get().  By
this, css->refcnt will go down to minus.

This may cause various things...one of results will be
initite-loop in css_tryget()  as this.

INFO: RCU detected CPU 0 stall (t=10000 jiffies)
sending NMI to all CPUs:
NMI backtrace for cpu 0
CPU 0:
<snip>

 <<EOE>>  <IRQ>  [<ffffffff810884bd>] trace_hardirqs_off+0xd/0x10
  [<ffffffff8102a940>] flat_send_IPI_mask+0x90/0xb0
  [<ffffffff8102a9c9>] flat_send_IPI_all+0x69/0x70
  [<ffffffff81027372>] arch_trigger_all_cpu_backtrace+0x62/0xa0
  [<ffffffff810bff8e>] __rcu_pending+0x7e/0x370
  [<ffffffff810c02c7>] rcu_check_callbacks+0x47/0x130
  [<ffffffff81063a26>] update_process_times+0x46/0x70
  [<ffffffff81085930>] tick_sched_timer+0x60/0x160
  [<ffffffff810858d0>] ? tick_sched_timer+0x0/0x160
  [<ffffffff8107a03a>] __run_hrtimer+0xba/0x150
  [<ffffffff8107a325>] hrtimer_interrupt+0xd5/0x1b0
  [<ffffffff81426dfe>] ? trace_hardirqs_off_thunk+0x3a/0x3c
  [<ffffffff8142cacd>] smp_apic_timer_interrupt+0x6d/0x9b
  [<ffffffff8100cb33>] apic_timer_interrupt+0x13/0x20
  <EOI>  [<ffffffff811317b6>] ? mem_cgroup_walk_tree+0x156/0x180
  [<ffffffff811316d3>] ? mem_cgroup_walk_tree+0x73/0x180
  [<ffffffff81131692>] ? mem_cgroup_walk_tree+0x32/0x180
  [<ffffffff81131a00>] ? mem_cgroup_get_local_stat+0x0/0x110
  [<ffffffff81131d5b>] ? mem_control_stat_show+0x14b/0x330
  [<ffffffff810a57fd>] ? cgroup_seqfile_show+0x3d/0x60

Above shows CPU0 caught in css_tryget()'s inifinite loop because
of bad refcnt.

This is a fix to set mz=NULL at the top of retry path.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Paul Menage <menage@google.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agouartlite: allow building for timberdale MFD
Richard Röjfors [Thu, 1 Oct 2009 22:44:07 +0000 (15:44 -0700)]
uartlite: allow building for timberdale MFD

Some configurations of the Timberdale FPGA has the uartlite
included.

Signed-off-by: Richard Röjfors <richard.rojfors@mocean-labs.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodocs: update patch size in SubmittingPatches
Randy Dunlap [Thu, 1 Oct 2009 22:44:06 +0000 (15:44 -0700)]
docs: update patch size in SubmittingPatches

This patch size comment is like so last millenium.  Update it to modern
times.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agosdio: pass whitelisted cis funce tuples to sdio drivers
Albert Herranz [Thu, 1 Oct 2009 22:44:05 +0000 (15:44 -0700)]
sdio: pass whitelisted cis funce tuples to sdio drivers

Some manufacturers provide vendor information in non-vendor specific CIS
tuples.  For example, Broadcom uses an Extended Function tuple to provide
the MAC address on some of their network cards, as in the case of the
Nintendo Wii WLAN daughter card.

This patch allows passing whitelisted FUNCE tuples unknown to the SDIO
core to a matching SDIO driver instead of rejecting them and failing.

Signed-off-by: Albert Herranz <albert_herranz@yahoo.es>
Cc: <linux-mmc@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm/rmap.c: fix comment
Huang Shijie [Thu, 1 Oct 2009 22:44:04 +0000 (15:44 -0700)]
mm/rmap.c: fix comment

The page_address_in_vma() is not only used in unuse_vma().

Signed-off-by: Huang Shijie <shijie8@gmail.com>
Acked-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agox86: fix csum_ipv6_magic asm memory clobber
Samuel Thibault [Thu, 1 Oct 2009 22:44:02 +0000 (15:44 -0700)]
x86: fix csum_ipv6_magic asm memory clobber

Just like ip_fast_csum, the assembly snippet in csum_ipv6_magic needs a
memory clobber, as it is only passed the address of the buffer, not a
memory reference to the buffer itself.

This caused failures in Hurd's pfinetv4 when we tried to compile it with
gcc-4.3 (bogus checksums).

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Acked-by: "David S. Miller" <davem@davemloft.net>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomn10300: fix kernel build failures when using gcc-4.x
Mark Salter [Thu, 1 Oct 2009 22:44:01 +0000 (15:44 -0700)]
mn10300: fix kernel build failures when using gcc-4.x

Fix some build failures when using gcc-4.x for MN10300.

Firstly, __get_user() fails to build because the pointer points to a const and
__gu_val ends up being read-only:

In file included from include/linux/mempolicy.h:62,
                 from init/main.c:50:
include/linux/pagemap.h: In function 'fault_in_pages_readable':
include/linux/pagemap.h:394: error: read-only variable '__gu_val' used as 'asm' output
include/linux/pagemap.h:394: error: read-only variable '__gu_val' used as 'asm' output
include/linux/pagemap.h:394: error: read-only variable '__gu_val' used as 'asm' output
include/linux/pagemap.h:400: error: read-only variable '__gu_val' used as 'asm' output
include/linux/pagemap.h:400: error: read-only variable '__gu_val' used as 'asm' output
include/linux/pagemap.h:400: error: read-only variable '__gu_val' used as 'asm' output
make[1]: *** [init/main.o] Error 1

Secondly, gcc-4 doesn't allow casts of lvalues:

  UPD     include/linux/compile.h
arch/mn10300/kernel/rtc.c: In function 'calibrate_clock':
arch/mn10300/kernel/rtc.c:170: error: lvalue required as left operand of assignment
arch/mn10300/kernel/rtc.c:172: error: lvalue required as left operand of assignment
make[1]: *** [arch/mn10300/kernel/rtc.o] Error 1

These are seen with gcc 4.2.1.

Signed-off-by: Mark Salter <msalter@redhat.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMAINTAINERS: ARM/Palm file patterns
Joe Perches [Thu, 1 Oct 2009 22:43:59 +0000 (15:43 -0700)]
MAINTAINERS: ARM/Palm file patterns

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Marek Vasut <marek.vasut@gmail.com>
Acked-by: Tomas Cech <sleep_walker@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoChar: vt_ioctl, fix BKL imbalance
Jiri Slaby [Thu, 1 Oct 2009 22:43:59 +0000 (15:43 -0700)]
Char: vt_ioctl, fix BKL imbalance

Stanse found (again) a BKL imbalance in vt_ioctl.

It's easily triggerable by ioctl(dev_tty_fd, VT_SETACTIVATE, NULL);

Introduced by

commit d3b5cffcf84a8bdc7073dce4745d67c72629af85                                 Author: Alan Cox <alan@linux.intel.com>
Date:   Sat Sep 19 13:13:26 2009 -0700

    vt: add an activate and lock

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Cc: Alan Cox <alan@linux.intel.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agorevert "m68k: convert to asm-generic/hardirq.h"
Andrew Morton [Thu, 1 Oct 2009 22:43:57 +0000 (15:43 -0700)]
revert "m68k: convert to asm-generic/hardirq.h"

Revert 45d80eea87c9f8292d2d33173d6866c0ec57238a ("m68k: convert to
asm-generic/hardirq.h") - it fails to compile due to an inclusion tangle:

In file included from include/linux/irq.h:12,
                 from include/asm-generic/hardirq.h:6,
                 from /usr/src/devel/arch/m68k/include/asm/hardirq_mm.h:6,
                 from /usr/src/devel/arch/m68k/include/asm/hardirq.h:4,
                 from include/linux/hardirq.h:10,
                 from /usr/src/devel/arch/m68k/include/asm/system_mm.h:69,
                 from /usr/src/devel/arch/m68k/include/asm/system.h:4,
                 from include/linux/list.h:7,
                 from include/linux/preempt.h:11,
                 from include/linux/spinlock.h:50,
                 from include/linux/seqlock.h:29,
                 from include/linux/time.h:8,
                 from include/linux/timex.h:56,
                 from include/linux/sched.h:56,
                 from arch/m68k/kernel/asm-offsets.c:14:
include/linux/smp.h:17: error: field 'list' has incomplete type

Cc: Christoph Hellwig <hch@lst.de>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoasm-generic/gpio.h: pull in linux/kernel.h for might_sleep()
Mike Frysinger [Thu, 1 Oct 2009 22:43:56 +0000 (15:43 -0700)]
asm-generic/gpio.h: pull in linux/kernel.h for might_sleep()

The asm-generic/gpio.h header uses the might_sleep() macro but doesn't
include the header for it, so any source code that might include
linux/gpio.h before linux/kernel.h can easily lead to a build failure.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoconst: constify remaining file_operations
Alexey Dobriyan [Thu, 1 Oct 2009 22:43:56 +0000 (15:43 -0700)]
const: constify remaining file_operations

[akpm@linux-foundation.org: fix KVM]
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodrivers/input/input.c: fix CONFIG_PM=n warning
Andrew Morton [Thu, 1 Oct 2009 22:43:55 +0000 (15:43 -0700)]
drivers/input/input.c: fix CONFIG_PM=n warning

drivers/input/input.c:1277: warning: 'input_dev_reset' defined but not used

Acked-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomodule: fix up CONFIG_KALLSYMS=n build.
Paul Mundt [Thu, 1 Oct 2009 22:43:54 +0000 (15:43 -0700)]
module: fix up CONFIG_KALLSYMS=n build.

Starting from commit 4a4962263f07d14660849ec134ee42b63e95ea9a "reduce
symbol table for loaded modules (v2)", the kernel/module.c build is broken
with CONFIG_KALLSYMS disabled.

  CC      kernel/module.o
kernel/module.c:1995: warning: type defaults to 'int' in declaration of 'Elf_Hdr'
kernel/module.c:1995: error: expected ';', ',' or ')' before '*' token
kernel/module.c: In function 'load_module':
kernel/module.c:2203: error: 'strmap' undeclared (first use in this function)
kernel/module.c:2203: error: (Each undeclared identifier is reported only once
kernel/module.c:2203: error: for each function it appears in.)
kernel/module.c:2239: error: 'symoffs' undeclared (first use in this function)
kernel/module.c:2239: error: implicit declaration of function 'layout_symtab'
kernel/module.c:2240: error: 'stroffs' undeclared (first use in this function)
make[1]: *** [kernel/module.o] Error 1
make: *** [kernel/module.o] Error 2

There are three different issues:

    - layout_symtab() takes a const Elf_Ehdr

    - layout_symtab() needs to return a value

    - symoffs/stroffs/strmap are referenced by the load_module() code
      despite being ifdefed out, which seems unnecessary given the noop
      behaviour of layout_symtab()/add_kallsyms() in the case of
      CONFIG_KALLSYMS=n.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Acked-by: Jan Beulich <jbeulich@novell.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Thu, 1 Oct 2009 00:36:45 +0000 (17:36 -0700)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  ax25: Fix possible oops in ax25_make_new
  net: restore tx timestamping for accelerated vlans
  Phonet: fix mutex imbalance
  sit: fix off-by-one in ipip6_tunnel_get_prl
  net: Fix sock_wfree() race
  net: Make setsockopt() optlen be unsigned.

15 years agoax25: Fix possible oops in ax25_make_new
Jarek Poplawski [Sun, 27 Sep 2009 10:57:02 +0000 (10:57 +0000)]
ax25: Fix possible oops in ax25_make_new

In ax25_make_new, if kmemdup of digipeat returns an error, there would
be an oops in sk_free while calling sk_destruct, because sk_protinfo
is NULL at the moment; move sk->sk_destruct initialization after this.

BTW of reported-by: Bernard Pidoux F6BVP <f6bvp@free.fr>

Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: restore tx timestamping for accelerated vlans
Eric Dumazet [Wed, 30 Sep 2009 23:42:42 +0000 (16:42 -0700)]
net: restore tx timestamping for accelerated vlans

Since commit 9b22ea560957de1484e6b3e8538f7eef202e3596
( net: fix packet socket delivery in rx irq handler )

We lost rx timestamping of packets received on accelerated vlans.

Effect is that tcpdump on real dev can show strange timings, since it gets rx timestamps
too late (ie at skb dequeueing time, not at skb queueing time)

14:47:26.986871 IP 192.168.20.110 > 192.168.20.141: icmp 64: echo request seq 1
14:47:26.986786 IP 192.168.20.141 > 192.168.20.110: icmp 64: echo reply seq 1

14:47:27.986888 IP 192.168.20.110 > 192.168.20.141: icmp 64: echo request seq 2
14:47:27.986781 IP 192.168.20.141 > 192.168.20.110: icmp 64: echo reply seq 2

14:47:28.986896 IP 192.168.20.110 > 192.168.20.141: icmp 64: echo request seq 3
14:47:28.986780 IP 192.168.20.141 > 192.168.20.110: icmp 64: echo reply seq 3

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoPhonet: fix mutex imbalance
Rémi Denis-Courmont [Wed, 30 Sep 2009 23:41:34 +0000 (16:41 -0700)]
Phonet: fix mutex imbalance

From: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>

port_mutex was unlocked twice.

Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosit: fix off-by-one in ipip6_tunnel_get_prl
Sascha Hlusiak [Tue, 29 Sep 2009 11:27:05 +0000 (11:27 +0000)]
sit: fix off-by-one in ipip6_tunnel_get_prl

When requesting all prl entries (kprl.addr == INADDR_ANY) and there are
more prl entries than there is space passed from userspace, the existing
code would always copy cmax+1 entries, which is more than can be handled.

This patch makes the kernel copy only exactly cmax entries.

Signed-off-by: Sascha Hlusiak <contact@saschahlusiak.de>
Acked-By: Fred L. Templin <Fred.L.Templin@boeing.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Fix sock_wfree() race
Eric Dumazet [Thu, 24 Sep 2009 10:49:24 +0000 (10:49 +0000)]
net: Fix sock_wfree() race

Commit 2b85a34e911bf483c27cfdd124aeb1605145dc80
(net: No more expensive sock_hold()/sock_put() on each tx)
opens a window in sock_wfree() where another cpu
might free the socket we are working on.

A fix is to call sk->sk_write_space(sk) while still
holding a reference on sk.

Reported-by: Jike Song <albcamus@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Make setsockopt() optlen be unsigned.
David S. Miller [Wed, 30 Sep 2009 23:12:20 +0000 (16:12 -0700)]
net: Make setsockopt() optlen be unsigned.

This provides safety against negative optlen at the type
level instead of depending upon (sometimes non-trivial)
checks against this sprinkled all over the the place, in
each and every implementation.

Based upon work done by Arjan van de Ven and feedback
from Linus Torvalds.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoMerge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 30 Sep 2009 22:10:40 +0000 (15:10 -0700)]
Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched_clock: Fix atomicity/continuity bug by using cmpxchg64()
  x86: Provide an alternative() based cmpxchg64()

15 years agosched_clock: Fix atomicity/continuity bug by using cmpxchg64()
Eric Dumazet [Wed, 30 Sep 2009 18:36:19 +0000 (20:36 +0200)]
sched_clock: Fix atomicity/continuity bug by using cmpxchg64()

Commit def0a9b2573 (sched_clock: Make it NMI safe) assumed
cmpxchg() of 64bit values was available on X86_32.

That is not so - and causes some subtle scheduler misbehavior due
to incorrect timestamps off to up by ~4 seconds.

Two symptoms are known right now:

 - interactivity problems seen by Arjan: up to 600 msecs
   latencies instead of the expected 20-40 msecs. These
   latencies are very visible on the desktop.

 - incorrect CPU stats: occasionally too high percentages in 'top',
   and crazy CPU usage stats.

Reported-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: John Stultz <johnstul@us.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20090930170754.0886ff2e@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: Provide an alternative() based cmpxchg64()
Arjan van de Ven [Wed, 30 Sep 2009 15:07:54 +0000 (17:07 +0200)]
x86: Provide an alternative() based cmpxchg64()

cmpxchg64() today generates, to quote Linus, "barf bag" code.

cmpxchg64() is about to get used in the scheduler to fix a bug there,
but it's a prerequisite that cmpxchg64() first be made non-sucking.

This patch turns cmpxchg64() into an efficient implementation that
uses the alternative() mechanism to just use the raw instruction on
all modern systems.

Note: the fallback is NOT smp safe, just like the current fallback
is not SMP safe. (Interested parties with i486 based SMP systems
are welcome to submit fix patches for that.)

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
[ fixed asm constraint bug ]
Fixed-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: John Stultz <johnstul@us.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20090930170754.0886ff2e@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Wed, 30 Sep 2009 20:46:56 +0000 (13:46 -0700)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  MIPS: Avoid spurious make includecheck message
  MIPS: VPE: Get rid of BKL.
  MIPS: VPE: Fix build after the credential changes a while ago.
  MIPS: Excite: Get rid of BKL.
  MIPS: Sibyte: Get rid of BKL.
  MIPS: BCM63xx: Add PCMCIA & Cardbus support.
  MIPS: MSP71xx: request_irq() failure ignored in msp_pcibios_config_access()
  MIPS: Decrease size of au1xxx_dbdma_pm_regs[][]
  MIPS: SMP: Inline arch_send_call_function_{single_ipi,ipi_mask}
  MIPS: SMP: Fix build.
  MIPS: MIPSxx SC: Avoid destructive invalidation on partial L2 cachelines.
  MIPS: Sibyte: Fix compilation error.
  MIPS: BCM1480: Re-apply patch lost due to bad resolution of merge conflict.
  MIPS: BCM63xx: Add serial driver for bcm63xx integrated UART.
  MIPS: Loongson2: Fix typo "enalbe" -> "enable"
  MIPS: SMTC: Remove duplicate structure field initialization
  MIPS: Remove duplicated #include
  MIPS: BCM63xx: Remove duplicated #include

15 years agoMIPS: Avoid spurious make includecheck message
Ralf Baechle [Tue, 29 Sep 2009 08:14:17 +0000 (10:14 +0200)]
MIPS: Avoid spurious make includecheck message

arch/mips/include/asm/unaligned.h: linux/unaligned/generic.h is included more than once.

Entirely legitimate but just noise.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: VPE: Get rid of BKL.
Ralf Baechle [Mon, 28 Sep 2009 23:52:27 +0000 (00:52 +0100)]
MIPS: VPE: Get rid of BKL.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: VPE: Fix build after the credential changes a while ago.
Ralf Baechle [Mon, 28 Sep 2009 23:19:20 +0000 (00:19 +0100)]
MIPS: VPE: Fix build after the credential changes a while ago.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Excite: Get rid of BKL.
Ralf Baechle [Mon, 28 Sep 2009 15:59:12 +0000 (16:59 +0100)]
MIPS: Excite: Get rid of BKL.

It's not obvious what good it was supposed to do here anyway.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Sibyte: Get rid of BKL.
Ralf Baechle [Mon, 28 Sep 2009 15:57:54 +0000 (16:57 +0100)]
MIPS: Sibyte: Get rid of BKL.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: BCM63xx: Add PCMCIA & Cardbus support.
Maxime Bizon [Mon, 28 Sep 2009 12:49:43 +0000 (14:49 +0200)]
MIPS: BCM63xx: Add PCMCIA & Cardbus support.

Signed-off-by: Maxime Bizon <mbizon@freebox.fr>
Reviewed-by: Wolfram Sang <w.sang@pengutronix.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: MSP71xx: request_irq() failure ignored in msp_pcibios_config_access()
Roel Kluin [Fri, 18 Sep 2009 19:50:11 +0000 (12:50 -0700)]
MIPS: MSP71xx: request_irq() failure ignored in msp_pcibios_config_access()

Produce an error if request_irq() fails.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Cc: "Ithamar R. Adema" <ithamar.adema@team-embedded.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Decrease size of au1xxx_dbdma_pm_regs[][]
Roel Kluin [Fri, 18 Sep 2009 19:50:10 +0000 (12:50 -0700)]
MIPS: Decrease size of au1xxx_dbdma_pm_regs[][]

There are 16 individual channels (NUM_DBDMA_CHANS) to save/restore plus the
global ddma block config (the +1).  The last register in a channel can be
skipped since it's read-only (at offset 0x18).

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Cc: Manuel Lauss <manuel.lauss@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: SMP: Inline arch_send_call_function_{single_ipi,ipi_mask}
Ralf Baechle [Fri, 25 Sep 2009 14:35:28 +0000 (15:35 +0100)]
MIPS: SMP: Inline arch_send_call_function_{single_ipi,ipi_mask}

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: SMP: Fix build.
Ralf Baechle [Fri, 25 Sep 2009 14:08:01 +0000 (15:08 +0100)]
MIPS: SMP: Fix build.

commit 48a048fed82a8e5fdd8618574f6d3de1a0d67a50
Author: Rusty Russell <rusty@rustcorp.com.au>
Date:   Thu Sep 24 09:34:44 2009 -0600

apparently only passed the "looks good" level of QA ;-)

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: MIPSxx SC: Avoid destructive invalidation on partial L2 cachelines.
Kevin Cernekee [Sat, 19 Sep 2009 02:12:45 +0000 (19:12 -0700)]
MIPS: MIPSxx SC: Avoid destructive invalidation on partial L2 cachelines.

This extends commit a8ca8b64e3fdfec17679cba0ca5ce6e3ffed092d to cover
MIPSxx-style board cache code.

Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Sibyte: Fix compilation error.
Mark Mason [Wed, 23 Sep 2009 20:35:09 +0000 (13:35 -0700)]
MIPS: Sibyte: Fix compilation error.

Build error introduced by d4f587c67fc39e0030ddd718675e252e208da4d7.

Signed-off-by: Mark Mason <mmason@upwardaccess.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: BCM1480: Re-apply patch lost due to bad resolution of merge conflict.
Ralf Baechle [Wed, 23 Sep 2009 20:37:08 +0000 (21:37 +0100)]
MIPS: BCM1480: Re-apply patch lost due to bad resolution of merge conflict.

Patch 14275ccdb1e4b487cca745aba994699c426a31ee and
d5dedd4507d307eb3f35f21b6e16f336fdc0d82a are conflicting and the
conflict was resolved badly in merge
92241940be501f798cb21db344bbb3d1ec3c4f1c resulting in the BCM1480 changes
of 14275ccdb1e4b487cca745aba994699c426a31ee getting lost.  Sort out the
damage.

Reported and initial patch by Mark Mason <mmason@upwardaccess.com>.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: BCM63xx: Add serial driver for bcm63xx integrated UART.
Maxime Bizon [Fri, 18 Sep 2009 11:04:58 +0000 (13:04 +0200)]
MIPS: BCM63xx: Add serial driver for bcm63xx integrated UART.

Signed-off-by: Maxime Bizon <mbizon@freebox.fr>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Loongson2: Fix typo "enalbe" -> "enable"
Uwe Kleine-König [Mon, 21 Sep 2009 08:40:37 +0000 (10:40 +0200)]
MIPS: Loongson2: Fix typo "enalbe" -> "enable"

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Cc: Yanhua <yanh@lemote.com>
Cc: Robert Richter <robert.richter@amd.com>
Acked-by: Wu Zhangjin <wuzj@lemote.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: SMTC: Remove duplicate structure field initialization
Julia Lawall [Mon, 21 Sep 2009 15:08:55 +0000 (17:08 +0200)]
MIPS: SMTC: Remove duplicate structure field initialization

The definition of the irq_ipi structure has two initializations of the
flags field.  This combines them.

[Ralf: The issue was originally introduced by commit
be4894196d79455f420dd7bb78be7dc73bec115c (linux-mips.org) rsp.
033890b084adfa367c544864451d7730552ce8bf (kernel.org).  The original
intention of the code was to initialize .flags with both flags ored together.
The broken C code as actually implemented will be compiled by an equally
broken gcc to use only the last initialization, that is IRQF_PERCPU
which means this turned into an SMTC bug for 2.6.23 and newer.]

The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@r@
identifier I, s, fld;
position p0,p;
expression E;
@@

struct I s =@p0 { ... .fld@p = E, ...};

@s@
identifier I, s, r.fld;
position r.p0,p;
expression E;
@@

struct I s =@p0 { ... .fld@p = E, ...};

@script:python@
p0 << r.p0;
fld << r.fld;
ps << s.p;
pr << r.p;
@@

if int(ps[0].line)!=int(pr[0].line) or int(ps[0].column)!=int(pr[0].column):
  cocci.print_main(fld,p0)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Remove duplicated #include
Huang Weiyi [Fri, 18 Sep 2009 07:14:35 +0000 (15:14 +0800)]
MIPS: Remove duplicated #include

Remove duplicated #include in arch/mips/kernel/smp.c.

Signed-off-by: Huang Weiyi <weiyi.huang@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: BCM63xx: Remove duplicated #include
Huang Weiyi [Fri, 18 Sep 2009 07:14:19 +0000 (15:14 +0800)]
MIPS: BCM63xx: Remove duplicated #include

Remove duplicated #include in arch/mips/bcm63xx/boards/board_bcm963xx.c.

Signed-off-by: Huang Weiyi <weiyi.huang@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke...
Linus Torvalds [Wed, 30 Sep 2009 16:42:24 +0000 (09:42 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/ryusuke/nilfs2

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2:
  nilfs2: fix missing initialization of i_dir_start_lookup member
  nilfs2: fix missing zero-fill initialization of btree node cache

15 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
Linus Torvalds [Wed, 30 Sep 2009 16:32:30 +0000 (09:32 -0700)]
Merge branch 'for_linus' of git://git./linux/kernel/git/tytso/ext4

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: Fix time encoding with extra epoch bits
  ext4: Add a stub for mpage_da_data in the trace header
  jbd2: Use tracepoints for history file
  ext4: Use tracepoints for mb_history trace file
  ext4, jbd2: Drop unneeded printks at mount and unmount time
  ext4: Handle nested ext4_journal_start/stop calls without a journal
  ext4: Make sure ext4_dirty_inode() updates the inode in no journal mode
  ext4: Avoid updating the inode table bh twice in no journal mode
  ext4: EXT4_IOC_MOVE_EXT: Check for different original and donor inodes first
  ext4: async direct IO for holes and fallocate support
  ext4: Use end_io callback to avoid direct I/O fallback to buffered I/O
  ext4: Split uninitialized extents for direct I/O
  ext4: release reserved quota when block reservation for delalloc retry
  ext4: Adjust ext4_da_writepages() to write out larger contiguous chunks
  ext4: Fix hueristic which avoids group preallocation for closed files
  ext4: Use ext4_msg() for ext4_da_writepage() errors
  ext4: Update documentation about quota mount options

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/hirofumi/fatfs-2.6
Linus Torvalds [Wed, 30 Sep 2009 16:31:14 +0000 (09:31 -0700)]
Merge git://git./linux/kernel/git/hirofumi/fatfs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/hirofumi/fatfs-2.6:
  fat: Check s_dirt in fat_sync_fs()
  vfat: change the default from shortname=lower to shortname=mixed
  fat/nls: Fix handling of utf8 invalid char

15 years agoMerge branch 'pm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/suspe...
Linus Torvalds [Wed, 30 Sep 2009 15:11:24 +0000 (08:11 -0700)]
Merge branch 'pm-fixes' of git://git./linux/kernel/git/rafael/suspend-2.6

* 'pm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/suspend-2.6:
  PM / yenta: Fix cardbus suspend/resume regression
  PM / PCMCIA: Drop second argument of pcmcia_socket_dev_suspend()

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Wed, 30 Sep 2009 15:07:12 +0000 (08:07 -0700)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (33 commits)
  sony-laptop: re-read the rfkill state when resuming from suspend
  sony-laptop: check for rfkill hard block at load time
  wext: add back wireless/ dir in sysfs for cfg80211 interfaces
  wext: Add bound checks for copy_from_user
  mac80211: improve/fix mlme messages
  cfg80211: always get BSS
  iwlwifi: fix 3945 ucode info retrieval after failure
  iwlwifi: fix memory leak in command queue handling
  iwlwifi: fix debugfs buffer handling
  cfg80211: don't set privacy w/o key
  cfg80211: wext: don't display BSSID unless associated
  net: Add explicit bound checks in net/socket.c
  bridge: Fix double-free in br_add_if.
  isdn: fix netjet/isdnhdlc build errors
  atm: dereference of he_dev->rbps_virt in he_init_group()
  ax25: Add missing dev_put in ax25_setsockopt
  Revert "sit: stateless autoconf for isatap"
  net: fix double skb free in dcbnl
  net: fix nlmsg len size for skb when error bit is set.
  net: fix vlan_get_size to include vlan_flags size
  ...

15 years agoMerge branch 'drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Wed, 30 Sep 2009 15:03:00 +0000 (08:03 -0700)]
Merge branch 'drm-next' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6: (25 commits)
  drm/radeon/kms: Convert R520 to new init path and associated cleanup
  drm/radeon/kms: Convert RV515 to new init path and associated cleanup
  drm: fix radeon DRM warnings when !CONFIG_DEBUG_FS
  drm: fix drm_fb_helper warning when !CONFIG_MAGIC_SYSRQ
  drm/r600: fix memory leak introduced with 64k malloc avoidance fix.
  drm/kms: make fb helper work for all drivers.
  drm/radeon/r600: fix offset handling in CS parser
  drm/radeon/kms/r600: fix forcing pci mode on agp cards
  drm/radeon/kms: fix for the extra pages copying.
  drm/radeon/kms/r600: add support for vline relocs
  drm/radeon/kms: fix some bugs in vline reloc
  drm/radeon/kms/r600: clamp vram to aperture size
  drm/kms: protect against fb helper not being created.
  drm/r600: get values from the passed in IB not the copy.
  drm: create gitignore file for radeon
  drm/radeon/kms: remove unneeded master create/destroy functions.
  drm/kms: start adding command line interface using fb.
  fb: change rules for global rules match.
  drm/radeon/kms: don't require up to 64k allocations. (v2)
  drm/radeon/kms: enable dac load detection by default.
  ...

Trivial conflicts in drivers/gpu/drm/radeon/radeon_asic.h due to adding
'->vga_set_state' function pointers.

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu
Linus Torvalds [Wed, 30 Sep 2009 14:58:46 +0000 (07:58 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tj/percpu

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu:
  percpu: make allocation failures more verbose
  percpu: make pcpu_setup_first_chunk() failures more verbose
  percpu: make embedding first chunk allocator check vmalloc space size
  sparc64: implement page mapping percpu first chunk allocator
  percpu: make pcpu_build_alloc_info() clear static buffers
  percpu: fix unit_map[] verification in pcpu_setup_first_chunk()

15 years agoMerge branch 'omap-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 30 Sep 2009 14:58:25 +0000 (07:58 -0700)]
Merge branch 'omap-fixes-for-linus' of git://git./linux/kernel/git/tmlind/linux-omap-2.6

* 'omap-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap-2.6:
  omap: Fix wrong condition check in while loop for mailbox and iommu2
  omap: rng: Use resource_size instead of manual calculation
  omap: Fix MMC gpio_wp for BeagleBoard C2 and above
  omap: Fix matrix_keymap_data usage
  omap: Fix a OMAP_MPUIO_VBASE typo for 850
  omap: Fix wrong jtag_id for 850
  omap: iovmm: Fix compiler warning
  omap: mailbox: Flush posted write when acking mailbox irq
  omap: mailbox: Execute softreset at startup
  omap: Add missing mux pin for EHCI phy reset line
  omap: Fix 44xx compile
  omap: Fix mcspi compile for 2420
  omap: Fix compile for arch/arm/mach-omap2

15 years agopty: reconnect the BSD TIOCSPTLCK handling to legacy ptys
Linus Torvalds [Wed, 30 Sep 2009 14:49:40 +0000 (07:49 -0700)]
pty: reconnect the BSD TIOCSPTLCK handling to legacy ptys

David Howells noticed (due to the compiler warning about an unused
'pty_ops_bsd' variable) that we haven't actually been using the code
that implements TIOCSPTLCK for legacy pty handling.  It's been that way
since 2.6.26, commit 3e8e88ca053150efdbecb45d8f481cf560ec808d to be
exact ("pty: prepare for tty->ops changes").

DavidH initially submitted a patch just removing the dead code entirely,
and since nobody has apparently ever complained, I'm not entirely sure
that wouldn't be the right thing to do.  But since the whole and only
point of the legacy pty code is to be compatible with legacy distros
that don't use the new unix98 pty model, let's just wire it up again.

And clean it up a bit while we're at it.

Acked-by: David Howells <dhowells@redhat.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoRevert "x86, mce: do not compile mcelog message on AMD"
Linus Torvalds [Wed, 30 Sep 2009 14:48:37 +0000 (07:48 -0700)]
Revert "x86, mce: do not compile mcelog message on AMD"

This reverts commit 22223c9b417be5fd0ab2cf9ad17eb7bd1e19f7b9, as
requested by Andi Kleen:

  "Obviously kernels compiled with AMD support can still run on non AMD
   systems, so messages like this can never be removed at compile time."

Requsted-by: Andi Kleen <andi@firstfloor.org>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoext4: Fix time encoding with extra epoch bits
Theodore Ts'o [Wed, 30 Sep 2009 05:13:55 +0000 (01:13 -0400)]
ext4: Fix time encoding with extra epoch bits

"Looking at ext4.h, I think the setting of extra time fields forgets to
mask the epoch bits so the epoch part overwrites nsec part. The second
change is only for coherency (2 -> EXT4_EPOCH_BITS)."

Thanks to Damien Guibouret for pointing out this problem.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Add a stub for mpage_da_data in the trace header
Josh Stone [Wed, 30 Sep 2009 04:51:22 +0000 (00:51 -0400)]
ext4: Add a stub for mpage_da_data in the trace header

The tracepoint ext4_da_write_pages has a struct mpage_da_data*
parameter, but that struct is only defined in fs/ext4/ext4.h.  This
patch adds a forward declaration for that struct, so this tracepoint
header can still be used by tools like SystemTap.

This is a continuation of the fix in commit 3661d286.

http://sourceware.org/bugzilla/show_bug.cgi?id=10703

Signed-off-by: Josh Stone <jistone@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agojbd2: Use tracepoints for history file
Theodore Ts'o [Wed, 30 Sep 2009 04:32:06 +0000 (00:32 -0400)]
jbd2: Use tracepoints for history file

The /proc/fs/jbd2/<dev>/history was maintained manually; by using
tracepoints, we can get all of the existing functionality of the /proc
file plus extra capabilities thanks to the ftrace infrastructure.  We
save memory as a bonus.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Use tracepoints for mb_history trace file
Theodore Ts'o [Wed, 30 Sep 2009 04:32:42 +0000 (00:32 -0400)]
ext4: Use tracepoints for mb_history trace file

The /proc/fs/ext4/<dev>/mb_history was maintained manually, and had a
number of problems: it required a largish amount of memory to be
allocated for each ext4 filesystem, and the s_mb_history_lock
introduced a CPU contention problem.

By ripping out the mb_history code and replacing it with ftrace
tracepoints, and we get more functionality: timestamps, event
filtering, the ability to correlate mballoc history with other ext4
tracepoints, etc.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4, jbd2: Drop unneeded printks at mount and unmount time
Theodore Ts'o [Tue, 29 Sep 2009 19:51:30 +0000 (15:51 -0400)]
ext4, jbd2: Drop unneeded printks at mount and unmount time

There are a number of kernel printk's which are printed when an ext4
filesystem is mounted and unmounted.  Disable them to economize space
in the system logs.  In addition, disabling the mballoc stats by
default saves a number of unneeded atomic operations for every block
allocation or deallocation.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Handle nested ext4_journal_start/stop calls without a journal
Curt Wohlgemuth [Tue, 29 Sep 2009 15:01:03 +0000 (11:01 -0400)]
ext4: Handle nested ext4_journal_start/stop calls without a journal

This patch fixes a problem with handling nested calls to
ext4_journal_start/ext4_journal_stop, when there is no journal present.

Signed-off-by: Curt Wohlgemuth <curtw@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Make sure ext4_dirty_inode() updates the inode in no journal mode
Curt Wohlgemuth [Tue, 29 Sep 2009 20:06:01 +0000 (16:06 -0400)]
ext4: Make sure ext4_dirty_inode() updates the inode in no journal mode

This patch a problem that ext4_dirty_inode() was not calling
ext4_mark_inode_dirty() if the current_handle is not valid, which it
is the case in no journal mode.

It also removes a test for non-matching transaction which can never
happen.

Signed-off-by: Curt Wohlgemuth <curtw@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Avoid updating the inode table bh twice in no journal mode
Frank Mayhar [Tue, 29 Sep 2009 14:07:47 +0000 (10:07 -0400)]
ext4: Avoid updating the inode table bh twice in no journal mode

This is a cleanup of commit 91ac6f4.  Since ext4_mark_inode_dirty()
has already called ext4_mark_iloc_dirty(), which in turn calls
ext4_do_update_inode(), it's not necessary to have ext4_write_inode()
call ext4_do_update_inode() in no journal mode.  Indeed, it would be
duplicated work.

Reviewed-by: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Frank Mayhar <fmayhar@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agonilfs2: fix missing initialization of i_dir_start_lookup member
Ryusuke Konishi [Mon, 28 Sep 2009 04:02:46 +0000 (13:02 +0900)]
nilfs2: fix missing initialization of i_dir_start_lookup member

The i_dir_start_lookup field in nilfs_inode_info objects should be
cleared when the objects are allocated, but the the initialization was
missing in case of reading from disk.  This adds the initialization.

Since the variable just gives a start page on directory lookups, the
bug was nonfatal until now.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
15 years agonilfs2: fix missing zero-fill initialization of btree node cache
Ryusuke Konishi [Sun, 27 Sep 2009 16:46:11 +0000 (01:46 +0900)]
nilfs2: fix missing zero-fill initialization of btree node cache

This will fix file system corruption which infrequently happens after
mount.  The problem was reported from users with the title "[NILFS
users] Fail to mount NILFS." (Message-ID:
<200908211918.34720.yuri@itinteg.net>), and so forth.  I've also
experienced the corruption multiple times on kernel 2.6.30 and 2.6.31.

The problem turned out to be caused due to discordance between
mapping->nrpages of a btree node cache and the actual number of pages
hung on the cache; if the mapping->nrpages becomes zero even as it has
pages, truncate_inode_pages() returns without doing anything.  Usually
this is harmless except it may cause page leak, but garbage collection
fairly infrequently sees a stale page remained in the btree node cache
of DAT (i.e. disk address translation file of nilfs), and induces the
corruption.

I identified a missing initialization in btree node caches was the
root cause.  This corrects the bug.

I've tested this for kernel 2.6.30 and 2.6.31.

Reported-by: Yuri Chislov <yuri@itinteg.net>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Cc: stable <stable@kernel.org>
15 years agodrm/radeon/kms: Convert R520 to new init path and associated cleanup
Jerome Glisse [Mon, 28 Sep 2009 18:39:19 +0000 (20:39 +0200)]
drm/radeon/kms: Convert R520 to new init path and associated cleanup

Convert the r520 asic support to new init path, change are smaller than
previous one as most of the architecture is now in place and more code
sharing can happen btw various asics.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm/radeon/kms: Convert RV515 to new init path and associated cleanup
Jerome Glisse [Mon, 28 Sep 2009 16:34:43 +0000 (18:34 +0200)]
drm/radeon/kms: Convert RV515 to new init path and associated cleanup

Convert the rv515 asic support to new init path also add an explanation
in radeon.h about the new init path. There is also few cleanups
associated with this change (others asic calling rv515 helper
functions).

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: fix radeon DRM warnings when !CONFIG_DEBUG_FS
Mikael Pettersson [Mon, 28 Sep 2009 16:27:23 +0000 (18:27 +0200)]
drm: fix radeon DRM warnings when !CONFIG_DEBUG_FS

Compiling the radeon DRM driver with !CONFIG_DEBUG_FS
throws the following warnings:

drivers/gpu/drm/radeon/radeon_ttm.c: In function 'radeon_ttm_debugfs_init':
drivers/gpu/drm/radeon/radeon_ttm.c:714: warning: unused variable 'i'
drivers/gpu/drm/radeon/radeon_ttm.c: At top level:
drivers/gpu/drm/radeon/radeon_ttm.c:692: warning: 'radeon_mem_types_list' defined but not used
drivers/gpu/drm/radeon/radeon_ttm.c:693: warning: 'radeon_mem_types_names' defined but not used

Fix: move these variables inside the #if defined(CONFIG_DEBUG_FS)
block in radeon_ttm_debugsfs_init(), which is the only place using them.

Signed-off-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: fix drm_fb_helper warning when !CONFIG_MAGIC_SYSRQ
Mikael Pettersson [Mon, 28 Sep 2009 16:26:25 +0000 (18:26 +0200)]
drm: fix drm_fb_helper warning when !CONFIG_MAGIC_SYSRQ

Compiling DRM throws the following warning if MAGIC_SYSRQ is disabled:

drivers/gpu/drm/drm_fb_helper.c:101: warning: 'sysrq_drm_fb_helper_restore_op' defined but not used

Fix: place sysrq_drm_fb_helper_restore_op and associated
definitions inside #ifdef CONFIG_MAGIC_SYSRQ.

Signed-off-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agopercpu: make allocation failures more verbose
Tejun Heo [Tue, 29 Sep 2009 00:17:58 +0000 (09:17 +0900)]
percpu: make allocation failures more verbose

Warn and dump stack when percpu allocation fails.  percpu allocator is
still young and unchecked NULL percpu pointer usage can result in
random memory corruption when combined with the pointer shifting in
access macros.  Allocation failures should be rare and the warning
message will be disabled after certain times.

Signed-off-by: Tejun Heo <tj@kernel.org>
15 years agopercpu: make pcpu_setup_first_chunk() failures more verbose
Tejun Heo [Thu, 24 Sep 2009 00:43:11 +0000 (09:43 +0900)]
percpu: make pcpu_setup_first_chunk() failures more verbose

The parameters to pcpu_setup_first_chunk() come from different sources
depending on architecture and can be quite complex.  The function runs
various sanity checks on the parameters and triggers BUG() if
something isn't right.  However, this is very early during the boot
and not reporting exactly what the problem is makes debugging even
harder.

Add PCPU_SETUP_BUG() macro which prints out enough information about
the parameters.  As the macro still puts separate BUG() for each
check, it won't lose any information even on the situations where only
the program counter can be retrieved.

While at it, also bump pcpu_dump_alloc_info() message to KERN_INFO so
that it's visible on the console if boot fails to complete.

Signed-off-by: Tejun Heo <tj@kernel.org>
15 years agopercpu: make embedding first chunk allocator check vmalloc space size
Tejun Heo [Thu, 24 Sep 2009 09:46:01 +0000 (18:46 +0900)]
percpu: make embedding first chunk allocator check vmalloc space size

Embedding first chunk allocator maintains the distances between units
in the vmalloc area and thus needs vmalloc space to be larger than the
maximum distances between units; otherwise, it wouldn't be able to
create any dynamic chunks.  This patch makes the embedding first chunk
allocator check vmalloc space size and if the maximum distance between
units is larger than 75% of it, print warning and, if page mapping
allocator is available, fail initialization so that the system falls
back onto it.

This should work around percpu allocation failure problems on certain
sparc64 configurations where distances between NUMA nodes are larger
than the vmalloc area and makes percpu allocator more robust for
future configurations.

Signed-off-by: Tejun Heo <tj@kernel.org>
15 years agosparc64: implement page mapping percpu first chunk allocator
Tejun Heo [Thu, 24 Sep 2009 09:18:55 +0000 (18:18 +0900)]
sparc64: implement page mapping percpu first chunk allocator

Implement page mapping percpu first chunk allocator as a fallback to
the embedding allocator.  The next patch will make the embedding
allocator check distances between units to determine whether it fits
within the vmalloc area so that this fallback can be used on such
cases.

sparc64 currently has relatively small vmalloc area which makes it
impossible to create any dynamic chunks on certain configurations
leading to percpu allocation failures.  This and the next patch should
allow those configurations to keep working until proper solution is
found.

While at it, mark pcpu_cpu_distance() with __init.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: David S. Miller <davem@davemloft.net>
15 years agopercpu: make pcpu_build_alloc_info() clear static buffers
Tejun Heo [Thu, 24 Sep 2009 09:50:34 +0000 (18:50 +0900)]
percpu: make pcpu_build_alloc_info() clear static buffers

pcpu_build_alloc_info() may be called multiple times when percpu is
falling back to different first chunk allocator.  Make it clear static
buffers so that they don't contain values from previous runs.

Signed-off-by: Tejun Heo <tj@kernel.org>
15 years agopercpu: fix unit_map[] verification in pcpu_setup_first_chunk()
Tejun Heo [Tue, 29 Sep 2009 00:17:56 +0000 (09:17 +0900)]
percpu: fix unit_map[] verification in pcpu_setup_first_chunk()

pcpu_setup_first_chunk() incorrectly used NR_CPUS as the impossible
unit number while unit number can equal and go over NR_CPUS with
sparse unit map.  This triggers BUG_ON() spuriously on machines which
have non-power-of-two number of cpus.  Use UINT_MAX instead.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-and-tested-by: Tony Vroon <tony@linx.net>
15 years agoPM / yenta: Fix cardbus suspend/resume regression
Rafael J. Wysocki [Mon, 28 Sep 2009 22:11:03 +0000 (00:11 +0200)]
PM / yenta: Fix cardbus suspend/resume regression

Since 2.6.29 the PCI PM core have been restoring the standard
configuration registers of PCI devices in the early phase of
resume.  In particular, PCI devices without drivers have been handled
this way since commit 355a72d75b3b4f4877db4c9070c798238028ecb5
(PCI: Rework default handling of suspend and resume).  Unfortunately,
this leads to post-resume problems with CardBus devices which cannot
be accessed in the early phase of resume, because the sockets they
are on have not been woken up yet at that point.

To solve this problem, move the yenta socket resume to the early
phase of resume and, analogously, move the suspend of it to the late
phase of suspend.  Additionally, remove some unnecessary PCI code
from the yenta socket's resume routine.

Fixes http://bugzilla.kernel.org/show_bug.cgi?id=13092, which is a
post-2.6.28 regression.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Reported-by: Florian <fs-kernelbugzilla@spline.de>
Cc: stable@kernel.org
15 years agoPM / PCMCIA: Drop second argument of pcmcia_socket_dev_suspend()
Rafael J. Wysocki [Mon, 28 Sep 2009 22:10:41 +0000 (00:10 +0200)]
PM / PCMCIA: Drop second argument of pcmcia_socket_dev_suspend()

pcmcia_socket_dev_suspend() doesn't use its second argument, so it
may be dropped safely.

This change is necessary for the subsequent yenta suspend/resume fix.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: stable@kernel.org
15 years agoMerge branch 'master' of ssh://master.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Mon, 28 Sep 2009 21:50:06 +0000 (14:50 -0700)]
Merge branch 'master' of ssh:///linux/kernel/git/linville/wireless-2.6

15 years agosony-laptop: re-read the rfkill state when resuming from suspend
Alan Jenkins [Fri, 25 Sep 2009 09:18:21 +0000 (10:18 +0100)]
sony-laptop: re-read the rfkill state when resuming from suspend

Without this, the hard-blocked state will be reported incorrectly if
the hardware switch is changed while the laptop is suspended.

Signed-off-by: Alan Jenkins <alan-jenkins@tuffmail.co.uk>
Tested-by: Norbert Preining <preining@logic.at>
Acked-by: Mattia Dongili <malattia@linux.it>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agosony-laptop: check for rfkill hard block at load time
Alan Jenkins [Thu, 24 Sep 2009 19:15:24 +0000 (20:15 +0100)]
sony-laptop: check for rfkill hard block at load time

"I recently (on a flight) I found out that when I boot with the hard-switch
activated, so turning off all wireless activity on my laptop, the state
is not correctly announced in /dev/rfkill (reading it with rfkill command,
or my own gnome applet)...

After turning off and on again the hard-switch the events were right."

We can fix this by querying the firmware at load time and calling
rfkill_set_hw_state().

Signed-off-by: Alan Jenkins <alan-jenkins@tuffmail.co.uk>
Tested-by: Norbert Preining <preining@logic.at>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Acked-by: Mattia Dongili <malattia@linux.it>
CC: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agowext: add back wireless/ dir in sysfs for cfg80211 interfaces
Johannes Berg [Mon, 28 Sep 2009 13:26:43 +0000 (15:26 +0200)]
wext: add back wireless/ dir in sysfs for cfg80211 interfaces

The move away from having drivers assign wireless handlers,
in favour of making cfg80211 assign them, broke the sysfs
registration (the wireless/ dir went missing) because the
handlers are now assigned only after registration, which is
too late.

Fix this by special-casing cfg80211-based devices, all
of which are required to have an ieee80211_ptr, in the
sysfs code, and also using get_wireless_stats() to have
the same values reported as in procfs.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Reported-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Tested-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agowext: Add bound checks for copy_from_user
Arjan van de Ven [Sat, 26 Sep 2009 18:51:14 +0000 (20:51 +0200)]
wext: Add bound checks for copy_from_user

The wireless extensions have a copy_from_user to a local stack
array "essid", but both me and gcc have failed to find where
the bounds for this copy are located in the code.

This patch adds some basic sanity checks for the copy length
to make sure that we don't overflow the stack buffer.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: linux-wireless@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: improve/fix mlme messages
Johannes Berg [Sat, 26 Sep 2009 12:45:41 +0000 (14:45 +0200)]
mac80211: improve/fix mlme messages

It's useful to know the MAC address when being
disassociated; fix a typo (missing colon) and
move some messages so we get them only when they
are actually taking effect.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agocfg80211: always get BSS
Johannes Berg [Sat, 26 Sep 2009 12:42:53 +0000 (14:42 +0200)]
cfg80211: always get BSS

Multiple problems were reported due to interaction
between wpa_supplicant and the wext compat code in
cfg80211, which appear to be due to it not getting
any bss pointer here when wpa_supplicant sets all
parameters -- do that now. We should still get the
bss after doing an extra scan, but that appears to
increase the time we need for connecting enough to
sometimes cause timeouts.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Tested-by: Hin-Tak Leung <hintak.leung@gmail.com>,
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: fix 3945 ucode info retrieval after failure
Reinette Chatre [Fri, 25 Sep 2009 21:24:23 +0000 (14:24 -0700)]
iwlwifi: fix 3945 ucode info retrieval after failure

When hardware or uCode problem occurs driver captures significant
information from device to enable debugging. The format of this information
is different between 3945 and 4965 and later devices, yet currently the
3945 uses the 4965 and later format. Fix this by adding a new library call
that is initialized to the correct formatting routine based on device.

This moves the iwlagn event and error log handling back to iwl-agn.c to
make it part of iwlagn module.

Also remove the 3945 sysfs file that triggers dump of event log - there is
already a debugfs file that can do it for all drivers.

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: fix memory leak in command queue handling
Reinette Chatre [Fri, 25 Sep 2009 21:24:22 +0000 (14:24 -0700)]
iwlwifi: fix memory leak in command queue handling

Also free the array of command pointers and meta data of each
command buffer when command queue is freed.

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: fix debugfs buffer handling
Reinette Chatre [Fri, 25 Sep 2009 21:24:21 +0000 (14:24 -0700)]
iwlwifi: fix debugfs buffer handling

We keep track of where to write into a buffer by keeping a count of how
much has been written so far. When writing to the buffer we thus take the
buffer pointer and adding the count of what has been written so far.
Keeping track of what has been written so far is done by incrementing
this number every time something is written to the buffer with how much has
been written at that time.

Currently this number is incremented incorrectly when using the
"hex_dump_to_buffer" call to add data to the buffer. Fix this by only
adding what has been added to the buffer in that call instead of what has
been added since beginning of buffer.

Issue was discovered and discussed during testing of
https://bugzilla.redhat.com/show_bug.cgi?id=464598 .

When a user views any of these files they will see something like:

[  179.355202] ------------[ cut here ]------------
[  179.355209] WARNING: at ../lib/vsprintf.c:989 vsnprintf+0x5ec/0x5f0()
[  179.355212] Hardware name: VGN-Z540N
[  179.355213] Modules linked in: i915 drm i2c_algo_bit i2c_core ipv6 acpi_cpufreq cpufreq_userspace cpufreq_powersave cpufreq_ondemand cpufreq_conservative cpufreq_stats freq_table container sbs sbshc arc4 ecb iwlagn iwlcore joydev led_class mac80211 af_packet pcmcia psmouse sony_laptop cfg80211 iTCO_wdt iTCO_vendor_support pcspkr serio_raw rfkill intel_agp video output tpm_infineon tpm tpm_bios button battery yenta_socket rsrc_nonstatic pcmcia_core processor ac evdev ext3 jbd mbcache sr_mod sg cdrom sd_mod ahci libata scsi_mod ehci_hcd uhci_hcd usbcore thermal fan thermal_sys
[  179.355262] Pid: 5449, comm: cat Not tainted 2.6.31-wl-54419-ge881071 #62
[  179.355264] Call Trace:
[  179.355267]  [<ffffffff811ad14c>] ? vsnprintf+0x5ec/0x5f0
[  179.355271]  [<ffffffff81041348>] warn_slowpath_common+0x78/0xd0
[  179.355275]  [<ffffffff810413af>] warn_slowpath_null+0xf/0x20
[  179.355277]  [<ffffffff811ad14c>] vsnprintf+0x5ec/0x5f0
[  179.355280]  [<ffffffff811ad23d>] ? scnprintf+0x5d/0x80
[  179.355283]  [<ffffffff811ad23d>] scnprintf+0x5d/0x80
[  179.355286]  [<ffffffff811aed29>] ? hex_dump_to_buffer+0x189/0x340
[  179.355290]  [<ffffffff810e91d7>] ? __kmalloc+0x207/0x260
[  179.355303]  [<ffffffffa02a02f8>] iwl_dbgfs_nvm_read+0xe8/0x220 [iwlcore]
[  179.355306]  [<ffffffff811a9b62>] ? __up_read+0x92/0xb0
[  179.355310]  [<ffffffff810f0988>] vfs_read+0xc8/0x1a0
[  179.355313]  [<ffffffff810f0b50>] sys_read+0x50/0x90
[  179.355316]  [<ffffffff8100bd6b>] system_call_fastpath+0x16/0x1b
[  179.355319] ---[ end trace 2383d0d5e0752ca0 ]---

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agocfg80211: don't set privacy w/o key
Johannes Berg [Thu, 24 Sep 2009 07:00:57 +0000 (09:00 +0200)]
cfg80211: don't set privacy w/o key

When wpa_supplicant is used to connect to open networks,
it causes the wdev->wext.keys to point to key memory, but
that key memory is all empty. Only use privacy when there
is a default key to be used.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Tested-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Tested-by: Kalle Valo <kalle.valo@iki.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agocfg80211: wext: don't display BSSID unless associated
Johannes Berg [Thu, 17 Sep 2009 15:35:54 +0000 (08:35 -0700)]
cfg80211: wext: don't display BSSID unless associated

Currently, cfg80211's SIOCGIWAP implementation returns
the BSSID that the user set, even if the connection has
since been dropped due to other changes. It only should
return the current BSSID when actually connected.

Also do a small code cleanup.

Reported-by: Thomas H. Guenther <thomas.h.guenther@intel.com>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Tested-by: Thomas H. Guenther <thomas.h.guenther@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoext4: EXT4_IOC_MOVE_EXT: Check for different original and donor inodes first
Theodore Ts'o [Mon, 28 Sep 2009 19:58:29 +0000 (15:58 -0400)]
ext4: EXT4_IOC_MOVE_EXT: Check for different original and donor inodes first

Move the check to make sure the original and donor inodes are
different earlier, to avoid a potential deadlock by trying to lock the
same inode twice.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agonet: Add explicit bound checks in net/socket.c
Arjan van de Ven [Mon, 28 Sep 2009 19:57:44 +0000 (12:57 -0700)]
net: Add explicit bound checks in net/socket.c

The sys_socketcall() function has a very clever system for the copy
size of its arguments. Unfortunately, gcc cannot deal with this in
terms of proving that the copy_from_user() is then always in bounds.
This is the last (well 9th of this series, but last in the kernel) such
case around.

With this patch, we can turn on code to make having the boundary provably
right for the whole kernel, and detect introduction of new security
accidents of this type early on.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobridge: Fix double-free in br_add_if.
Jeff Hansen [Mon, 28 Sep 2009 19:54:25 +0000 (12:54 -0700)]
bridge: Fix double-free in br_add_if.

There is a potential double-kfree in net/bridge/br_if.c.  If br_fdb_insert
fails, then the kobject is put back (which calls kfree due to the kobject
release), and then kfree is called again on the net_bridge_port.  This
patch fixes the crash.

Thanks to Stephen Hemminger for the one-line fix.

Signed-off-by: Jeff Hansen <x@jeffhansen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoext4: async direct IO for holes and fallocate support
Mingming Cao [Mon, 28 Sep 2009 19:48:29 +0000 (15:48 -0400)]
ext4: async direct IO for holes and fallocate support

For async direct IO that covers holes or fallocate, the end_io
callback function now queued the convertion work on workqueue but
don't flush the work rightaway as it might take too long to afford.

But when fsync is called after all the data is completed, user expects
the metadata also being updated before fsync returns.

Thus we need to flush the conversion work when fsync() is called.
This patch keep track of a listed of completed async direct io that
has a work queued on workqueue.  When fsync() is called, it will go
through the list and do the conversion.

Signed-off-by: Mingming Cao <cmm@us.ibm.com>
15 years agoext4: Use end_io callback to avoid direct I/O fallback to buffered I/O
Mingming Cao [Mon, 28 Sep 2009 19:48:41 +0000 (15:48 -0400)]
ext4: Use end_io callback to avoid direct I/O fallback to buffered I/O

Currently the DIO VFS code passes create = 0 when writing to the
middle of file.  It does this to avoid block allocation for holes, so
as not to expose stale data out when there is a parallel buffered read
(which does not hold the i_mutex lock).  Direct I/O writes into holes
falls back to buffered IO for this reason.

Since preallocated extents are treated as holes when doing a
get_block() look up (buffer is not mapped), direct IO over fallocate
also falls back to buffered IO.  Thus ext4 actually silently falls
back to buffered IO in above two cases, which is undesirable.

To fix this, this patch creates unitialized extents when a direct I/O
write into holes in sparse files, and registering an end_io callback which
converts the uninitialized extent to an initialized extent after the
I/O is completed.

Singed-Off-By: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Split uninitialized extents for direct I/O
Mingming Cao [Mon, 28 Sep 2009 19:49:08 +0000 (15:49 -0400)]
ext4: Split uninitialized extents for direct I/O

When writing into an unitialized extent via direct I/O, and the direct
I/O doesn't exactly cover the unitialized extent, split the extent
into uninitialized and initialized extents before submitting the I/O.
This avoids needing to deal with an ENOSPC error in the end_io
callback that gets used for direct I/O.

When the IO is complete, the written extent will be marked as initialized.

Singed-Off-By: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: release reserved quota when block reservation for delalloc retry
Mingming Cao [Mon, 28 Sep 2009 19:49:52 +0000 (15:49 -0400)]
ext4: release reserved quota when block reservation for delalloc retry

ext4_da_reserve_space() can reserve quota blocks multiple times if
ext4_claim_free_blocks() fail and we retry the allocation. We should
release the quota reservation before restarting.

Bug found by Jan Kara.

Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Adjust ext4_da_writepages() to write out larger contiguous chunks
Theodore Ts'o [Tue, 29 Sep 2009 17:31:31 +0000 (13:31 -0400)]
ext4: Adjust ext4_da_writepages() to write out larger contiguous chunks

Work around problems in the writeback code to force out writebacks in
larger chunks than just 4mb, which is just too small.  This also works
around limitations in the ext4 block allocator, which can't allocate
more than 2048 blocks at a time.  So we need to defeat the round-robin
characteristics of the writeback code and try to write out as many
blocks in one inode before allowing the writeback code to move on to
another inode.  We add a a new per-filesystem tunable,
max_writeback_mb_bump, which caps this to a default of 128mb per
inode.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoisdn: fix netjet/isdnhdlc build errors
Randy Dunlap [Mon, 28 Sep 2009 19:47:36 +0000 (12:47 -0700)]
isdn: fix netjet/isdnhdlc build errors

Commit cb3824bade2549d7ad059d5802da43312540fdee didn't fix this problem.

Fix build errors in netjet, using isdnhdlc module:

drivers/built-in.o: In function `mode_tiger':
netjet.c:(.text+0x1ca0c7): undefined reference to `isdnhdlc_rcv_init'
netjet.c:(.text+0x1ca0d4): undefined reference to `isdnhdlc_out_init'
drivers/built-in.o: In function `fill_dma':
netjet.c:(.text+0x1ca2bd): undefined reference to `isdnhdlc_encode'
drivers/built-in.o: In function `read_dma':
netjet.c:(.text+0x1ca614): undefined reference to `isdnhdlc_decode'
drivers/built-in.o: In function `nj_irq':
netjet.c:(.text+0x1cb07a): undefined reference to `isdnhdlc_encode'

drivers/built-in.o: In function `isdnhdlc_decode':
(.text+0x1c2088): undefined reference to `crc_ccitt_table'
drivers/built-in.o: In function `isdnhdlc_encode':
(.text+0x1c2339): undefined reference to `crc_ccitt_table'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoatm: dereference of he_dev->rbps_virt in he_init_group()
Juha Leppanen [Mon, 28 Sep 2009 19:45:29 +0000 (12:45 -0700)]
atm: dereference of he_dev->rbps_virt in he_init_group()

The prefix decrement causes a very long loop if pci_pool_alloc() failed
in the first iteration. Also I swapped rbps and rbpl arguments.

Reported-by: Juha Leppanen <juha_motorsportcom@luukku.com>
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoax25: Add missing dev_put in ax25_setsockopt
Ralf Baechle [Mon, 28 Sep 2009 19:26:28 +0000 (12:26 -0700)]
ax25: Add missing dev_put in ax25_setsockopt

ax25_setsockopt SO_BINDTODEVICE is missing a dev_put call in case of
success.  Re-order code to fix this bug.  While at it also reformat two
lines of code to comply with the Linux coding style.

Initial patch by Jarek Poplawski <jarkao2@gmail.com>.

Reported-by: Bernard Pidoux F6BVP <f6bvp@free.fr>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>