openwrt/staging/blogic.git
17 years ago[POWERPC] pasemi: Move cpus to hold loop before restart
Olof Johansson [Wed, 7 Nov 2007 15:31:06 +0000 (09:31 -0600)]
[POWERPC] pasemi: Move cpus to hold loop before restart

Use smp_send_stop() to move all cpus besides the one executing reboot
into a hold loop, to keep them from being in powersavings mode at the
time of reboot.

Signed-off-by: Olof Johansson <olof@lixom.net>
17 years ago[POWERPC] pasemi: Don't enter powersaving states from elevated astates
Olof Johansson [Wed, 7 Nov 2007 15:26:06 +0000 (09:26 -0600)]
[POWERPC] pasemi: Don't enter powersaving states from elevated astates

When the PWRficient cpus are entered into powersavings states, the
astate is automatically dropped down to 0. While we still restore it
when we come out of idle, it can still cause some weird effects with
respect to performance (especially since it takes a while to ramp up to
higher astates).

So, to avoid this, don't enter power savings when the cpufreq driver
(or user) has set higher astates than 0.

The restore is still required, since there's a chance the astate has
been raised from the other cpu while the idling one was asleep.

Signed-off-by: Olof Johansson <olof@lixom.net>
17 years ago[POWERPC] pasemi: Broaden specific references to 1682M
Olof Johansson [Mon, 5 Nov 2007 02:57:45 +0000 (20:57 -0600)]
[POWERPC] pasemi: Broaden specific references to 1682M

There will be more product numbers in the future than just PA6T-1682M,
but they will share much of the features. Remove some of the explicit
references and compatibility checks with 1682M, and replace most of them
with the more generic term "PWRficient".

Signed-off-by: Olof Johansson <olof@lixom.net>
Acked-by: Michael Buesch <mb@bu3sch.de>
Acked-by: Doug Thompson <dougthompson@xmission.com>
17 years ago[POWERPC] pasemi: clean up mdio_gpio a bit
Olof Johansson [Sun, 4 Nov 2007 21:44:15 +0000 (15:44 -0600)]
[POWERPC] pasemi: clean up mdio_gpio a bit

Misc cleanups of mdio_gpio:
* Better error handling/unrolling in case of init/alloc failures
* Go through child nodes and get their interrupts instead of using
  hardcoded values
* Remap the GPIO registers at module load/driver init instead of during probe
* Coding style and other misc cleanups

Signed-off-by: Olof Johansson <olof@lixom.net>
17 years ago[POWERPC] Fix 8xx build breakage due to _tlbie changes
Benjamin Herrenschmidt [Tue, 20 Nov 2007 07:32:12 +0000 (18:32 +1100)]
[POWERPC] Fix 8xx build breakage due to _tlbie changes

My changes to _tlbie to fix 4xx unfortunately broke 8xx build in a
couple of places.  This fixes it.

Spotted by Olof Johansson.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Vitaly Bordug <vitb@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years agoMerge branch 'for-2.6.24' of master.kernel.org:/pub/scm/linux/kernel/git/galak/powerp...
Paul Mackerras [Tue, 20 Nov 2007 07:40:23 +0000 (18:40 +1100)]
Merge branch 'for-2.6.24' of /linux/kernel/git/galak/powerpc into merge

17 years ago[POWERPC] Fix device tree interrupt map for Freescale ULI1575 boards
Kumar Gala [Tue, 20 Nov 2007 05:36:23 +0000 (23:36 -0600)]
[POWERPC] Fix device tree interrupt map for Freescale ULI1575 boards

The interrupt map for the PCI PHB that had the ULI1575 was not correct
on the boards that have it.

* 8544 DS:
   - Fix interrupt mask
   - Be explicit about use of INTA for on chip peripherals

* 8572 DS/8641 HPCN:
   - Fix interrupt mask
   - Expand interrupt map for PCI slots to cover all functions
   - Be explicit about use of INTA for on chip peripherals

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years ago[POWERPC] Fix possible division by zero in scaled time accounting
Michael Neuling [Tue, 20 Nov 2007 04:18:40 +0000 (15:18 +1100)]
[POWERPC] Fix possible division by zero in scaled time accounting

If we get no user time and no system time allocated since the last
account_system_vtime, the system to user time ratio estimate can end
up dividing by zero.

This was causing a problem noticed by Balbir Singh.

Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] spufs: Fix context destroy vs /spu readdir race
Jeremy Kerr [Fri, 16 Nov 2007 02:32:23 +0000 (13:32 +1100)]
[POWERPC] spufs: Fix context destroy vs /spu readdir race

We can currently cause an oops by repeatedly creating and destroying
contexts, while doing getdents() calls on the "/spu" directory.

This is due to the context's top-level dentry remaining hashed while
the context is being destroyed.

Fix this by unhashing the context's dentry with the
dentry->d_inode->i_mutex held. This way, we'll hit the check for
d_unhashed in dentry_readdir, and won't be included in the
list of subdirs for /spu.

test: spufs-testsuite:tests/01-spu_create/07-destroy-vs-readdir-race

Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] Fix RTAS os-term usage on kernel panic
Linas Vepstas [Tue, 20 Nov 2007 01:28:15 +0000 (12:28 +1100)]
[POWERPC] Fix RTAS os-term usage on kernel panic

The rtas_os_term() routine was being called at the wrong time.
The actual rtas call "os-term" will not ever return, and so
calling it from the panic notifier is too early.  Instead,
call it from the machine_reset() call.

This splits the rtas_os_term() routine into two: one part to capture
the kernel panic message, invoked during the panic notifier, and
another part that is invoked during machine_reset().

Prior to this patch, the os-term call was never being made,
because panic_timeout was always non-zero.  Calling os-term
helps keep the hypervisor happy!  We have to keep the hypervisor
happy to avoid service, dump and error reporting problems.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] 83xx: Handle mpc8360 rev. 2.1 RGMII timing erratum
Kim Phillips [Mon, 5 Nov 2007 18:15:51 +0000 (12:15 -0600)]
[POWERPC] 83xx: Handle mpc8360 rev. 2.1 RGMII timing erratum

If on a rev. 2.1, adjust UCC clock and data timing characteristics
as specified in the rev.2.1 erratum #2.

Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years ago[POWERPC] Document rgmii-rxid and rgmii-txid phy-connection-types
Kim Phillips [Mon, 5 Nov 2007 18:15:35 +0000 (12:15 -0600)]
[POWERPC] Document rgmii-rxid and rgmii-txid phy-connection-types

A h/w bug requires we program the PHY in RGMII mode for internal delay
on the receive or transmit side only; document the new property values.

Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years ago[POWERPC] 83xx: Fix 2nd UCC entry in mpc832x_mds.dts
Kim Phillips [Tue, 13 Nov 2007 23:26:31 +0000 (17:26 -0600)]
[POWERPC] 83xx: Fix 2nd UCC entry in mpc832x_mds.dts

correct the reg property, remove duplicate io port entry, whitespace fixes.

Thanks to Peter Van Ackeren for pointing this out.

Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years ago[POWERPC] 83xx: mpc832x mds: Fix board PHY reset code
Kim Phillips [Tue, 20 Nov 2007 01:04:43 +0000 (19:04 -0600)]
[POWERPC] 83xx: mpc832x mds: Fix board PHY reset code

currently the board-level PHY reset code for the mpc832x MDS messes with
reset configuration words source settings which is plain wrong (it
looks like this board code was cut-n-pasted from the mpc8360 mds code,
which has the PHY reset bits in a different BCSR); this patch points
the PHY reset code to the proper mpc832x mds PHY reset bits in the BCSR.

Signed-off-by: Peter Van Ackeren <peter.vanackeren@freescale.com>
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years ago[POWERPC] Fix potential NULL dereference
Cyrill Gorcunov [Thu, 15 Nov 2007 18:47:06 +0000 (21:47 +0300)]
[POWERPC] Fix potential NULL dereference

This patch does fix potential NULL pointer dereference that could take
place inside of strcmp() if of_get_property() call failed.

Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years ago[POWERPC] vdso: Fixes for cache block sizes
Olof Johansson [Tue, 20 Nov 2007 01:24:45 +0000 (12:24 +1100)]
[POWERPC] vdso: Fixes for cache block sizes

The current VDSO implementation is hardcoded to 128 byte cache blocks,
which are only used on IBM's 64-bit processors.

Convert it to get the cache block sizes out of vdso_data instead,
similar to how the ppc64 in-kernel cache flush does it.

Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years agoMerge branch 'for-2.6.24' of master.kernel.org:/pub/scm/linux/kernel/git/olof/pasemi...
Paul Mackerras [Tue, 20 Nov 2007 02:54:46 +0000 (13:54 +1100)]
Merge branch 'for-2.6.24' of /linux/kernel/git/olof/pasemi into merge

17 years ago[POWERPC] pasemi: Don't reset mpic at boot
Olof Johansson [Tue, 6 Nov 2007 02:35:33 +0000 (20:35 -0600)]
[POWERPC] pasemi: Don't reset mpic at boot

Due to an erratum, we don't want to reset the mpic at boot time. It can
sometimes cause problems with lost interrupts later on while running.

Signed-off-by: Olof Johansson <olof@lixom.net>
17 years agoMerge branch 'for-2.6.24' of master.kernel.org:/pub/scm/linux/kernel/git/jwboyer...
Paul Mackerras [Tue, 20 Nov 2007 00:39:26 +0000 (11:39 +1100)]
Merge branch 'for-2.6.24' of /linux/kernel/git/jwboyer/powerpc-4xx into merge

17 years ago[POWERPC] Fix kmalloc alignment on non-coherent DMA platforms
Benjamin Herrenschmidt [Mon, 19 Nov 2007 08:25:06 +0000 (19:25 +1100)]
[POWERPC] Fix kmalloc alignment on non-coherent DMA platforms

On platforms doing non-coherent DMA (4xx, 8xx, ...), it's important
that the kmalloc minimum alignment is set to the cache line size, to
avoid sharing cache lines between different objects, so that DMA to
one of the objects doesn't corrupt the other.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] Fix build failure on legacy iSeries
Kamalesh Babulal [Mon, 19 Nov 2007 06:44:05 +0000 (17:44 +1100)]
[POWERPC] Fix build failure on legacy iSeries

Include <asm/iseries/hv_call.h> in arch/powerpc/mm/stab.c to fix the
following compile error (found with randconfig):

  CC      arch/powerpc/mm/stab.o
arch/powerpc/mm/stab.c: In function "stab_initialize":
arch/powerpc/mm/stab.c:282: error: implicit declaration of function "HvCall1"
arch/powerpc/mm/stab.c:282: error: "HvCallBaseSetASR" undeclared (first use in this function)
arch/powerpc/mm/stab.c:282: error: (Each undeclared identifier is reported only once
arch/powerpc/mm/stab.c:282: error: for each function it appears in.)
make[1]: *** [arch/powerpc/mm/stab.o] Error 1
make: *** [arch/powerpc/mm] Error 2

Signed-off-by: Kamalesh Babulal <kamalesh@linux.vnet.ibm.com>
Acked-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] 4xx: Use virtual PVR value to init FPU on arch/ppc 440EP
Josh Boyer [Fri, 16 Nov 2007 17:29:56 +0000 (04:29 +1100)]
[POWERPC] 4xx: Use virtual PVR value to init FPU on arch/ppc 440EP

This fixes arch/ppc 440EP platforms to setup the FPU correctly.  A virtual
PVR is used, as 440GR platforms share an identical hardware PVR value and do
not have an FPU.

Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
17 years ago[POWERPC] 4xx: UIC add mask_ack callback
Valentine Barshak [Tue, 13 Nov 2007 20:25:21 +0000 (07:25 +1100)]
[POWERPC] 4xx: UIC add mask_ack callback

This adds uic_mask_ack_irq() callback to PowerPC 4xx uic code
to avoid kernel crash. It is used for edge-triggered interrupts
by handle_uic_irq().

Signed-off-by: Valentine Barshak <vbarshak@ru.mvista.com>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
17 years ago[POWERPC] 4xx: balance ioremap/ioumap calls for Yucca
Roel Kluin [Wed, 7 Nov 2007 22:22:48 +0000 (09:22 +1100)]
[POWERPC] 4xx: balance ioremap/ioumap calls for Yucca

Call iounmap before returning from an error case.

Signed-off-by: Roel Kluin <12o3l@tiscali.nl>
Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
17 years ago[POWERPC] Xilinx: Register AC97 Controller Reference with the platform bus
Joachim Foerster [Wed, 7 Nov 2007 16:14:15 +0000 (03:14 +1100)]
[POWERPC] Xilinx: Register AC97 Controller Reference with the platform bus

Add AC97 platform bus registration for Xilinx Virtex platforms.

Signed-off-by: Joachim Foerster <JOFT@gmx.de>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
17 years ago[POWERPC] 4xx: Replace #includes of asm/of_platform.h with linux/of_platform.h.
Jon Loeliger [Tue, 6 Nov 2007 18:13:04 +0000 (05:13 +1100)]
[POWERPC] 4xx: Replace #includes of asm/of_platform.h with linux/of_platform.h.

Signed-off-by: Jon Loeliger <jdl@freescale.com>
Acked-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
17 years ago[POWERPC] Fix declaration of pcibios_free_controller
Benjamin Herrenschmidt [Fri, 16 Nov 2007 07:42:18 +0000 (18:42 +1100)]
[POWERPC] Fix declaration of pcibios_free_controller

pcibios_free_controller() is now available for both 32 and 64 bits
but the header only declares it for 64 bits.  This moves the
declaration down next to the pcibios_alloc_controller() one.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] prpmc2800: Enable L2 cache
Mark A. Greer [Wed, 14 Nov 2007 19:04:02 +0000 (06:04 +1100)]
[POWERPC] prpmc2800: Enable L2 cache

Turn on the L2 cache on the prpmc2800 platform.

Signed-off-by: Mark A. Greer <mgreer@mvista.com
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] pSeries: make pseries_defconfig minus PCI build again
Stephen Rothwell [Wed, 14 Nov 2007 04:07:39 +0000 (15:07 +1100)]
[POWERPC] pSeries: make pseries_defconfig minus PCI build again

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] Fix multiple bugs in rtas_ibm_suspend_me code
Nathan Lynch [Tue, 13 Nov 2007 16:15:13 +0000 (03:15 +1100)]
[POWERPC] Fix multiple bugs in rtas_ibm_suspend_me code

There are several issues with the rtas_ibm_suspend_me code, which
enables platform-assisted suspension of an LPAR as covered in PAPR
2.2.

1.) rtas_ibm_suspend_me uses on_each_cpu() to invoke
rtas_percpu_suspend_me on all cpus via IPI:

if (on_each_cpu(rtas_percpu_suspend_me, &data, 1, 0))
...

'data' is on the calling task's stack, but rtas_ibm_suspend_me takes
no measures to ensure that all instances of rtas_percpu_suspend_me are
finished accessing 'data' before returning.  This can result in the
IPI'd cpus accessing random stack data and getting stuck in H_JOIN.

This is addressed by using an atomic count of workers and a completion
on the stack.

2.) rtas_percpu_suspend_me is needlessly calling H_JOIN in a loop.
The only event that can cause a cpu to return from H_JOIN is an H_PROD
from another cpu or a NMI/system reset.  Each cpu need call H_JOIN
only once per suspend operation.

Remove the loop and the now unnecessary 'waiting' state variable.

3.) H_JOIN must be called with MSR[EE] off, but lazy interrupt
disabling may cause the caller of rtas_ibm_suspend_me to call H_JOIN
with it on; the local_irq_disable() in on_each_cpu() is not
sufficient.

Fix this by explicitly saving the MSR and clearing the EE bit before
calling H_JOIN.

4.) H_PROD is being called with the Linux logical cpu number as the
parameter, not the platform interrupt server value.  (It's also being
called for all possible cpus, which is harmless, but unnecessary.)

This is fixed by calling H_PROD for each online cpu using
get_hard_smp_processor_id(cpu) for the argument.

Signed-off-by: Nathan Lynch <ntl@pobox.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux...
Linus Torvalds [Sat, 17 Nov 2007 16:36:10 +0000 (08:36 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/x86/linux-2.6-x86

* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-x86:
  x86: simplify "make ARCH=x86" and fix kconfig all.config
  x86: reboot fixup for wrap2c board
  x86: check boundary in count setup resource
  x86: fix reboot with no keyboard attached
  x86: add hpet sanity checks
  x86: on x86_64, correct reading of PC RTC when update in progress in time_64.c
  x86: fix freeze in x86_64 RTC update code in time_64.c
  ntp: fix typo that makes sync_cmos_clock erratic
  Remove x86 merge artifact from top Makefile
  x86: fixup cpu_info array conversion
  x86: show cpuinfo only for online CPUs
  x86: fix cpu-hotplug regression
  x86: ignore the sys_getcpu() tcache parameter
  x86: voyager use correct header file name
  x86: fix smp init sections
  x86: fix voyager_cat_init section
  x86: fix bogus memcpy in es7000_check_dsdt()

17 years agox86: simplify "make ARCH=x86" and fix kconfig all.config
Sam Ravnborg [Sat, 17 Nov 2007 14:37:31 +0000 (15:37 +0100)]
x86: simplify "make ARCH=x86" and fix kconfig all.config

Simplify "make ARCH=x86" and fix kconfig so we again can set 64BIT in
all.config.

For a fix the diffstat is nice:
 6 files changed, 3 insertions(+), 36 deletions(-)

The patch reverts these commits:
 - 0f855aa64b3f63d35a891510cf7db932a435c116 ("kconfig: add helper to set
   config symbol from environment variable")
 - 2a113281f5cd2febbab21a93c8943f8d3eece4d3 ("kconfig: use $K64BIT to
   set 64BIT with all*config targets")

Roman Zippel pointed out that kconfig supported string compares so
the additional complexity introduced by the above two patches were
not needed.

With this patch we have following behaviour:

  # make {allno,allyes,allmod,rand}config [ARCH=...]
  option \ host arch      | 32bit         | 64bit
  =====================================================
  ./.                     | 32bit         | 64bit
  ARCH=x86                | 32bit         | 32bit
  ARCH=i386               | 32bit         | 32bit
  ARCH=x86_64             | 64bit         | 64bit

The general rule are that ARCH= and native architecture takes
precedence over the configuration.

So make ARCH=i386 [whatever] will always build a 32-bit kernel
no matter what the configuration says.  The configuration will
be updated to 32-bit if it was configured to 64-bit and the
other way around.

This behaviour is consistent with previous behaviour so no
suprises here.

make ARCH=x86 will per default result in a 32-bit kernel but as
the only ARCH= value x86 allow the user to select between 32-bit
and 64-bit using menuconfig.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andreas Herrmann <aherrman@arcor.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86: simplify "make ARCH=x86" and fix kconfig all.config
Sam Ravnborg [Sat, 17 Nov 2007 14:37:31 +0000 (15:37 +0100)]
x86: simplify "make ARCH=x86" and fix kconfig all.config

Simplify "make ARCH=x86" and fix kconfig so we again
can set 64BIT in all.config.

For a fix the diffstat is nice:
 6 files changed, 3 insertions(+), 36 deletions(-)

The patch reverts these commits:
0f855aa64b3f63d35a891510cf7db932a435c116
-> kconfig: add helper to set config symbol from environment variable

2a113281f5cd2febbab21a93c8943f8d3eece4d3
-> kconfig: use $K64BIT to set 64BIT with all*config targets

Roman Zippel pointed out that kconfig supported string
compares so the additional complexity introduced by the
above two patches were not needed.

With this patch we have following behaviour:

# make {allno,allyes,allmod,rand}config [ARCH=...]
option \ host arch      | 32bit         | 64bit
=====================================================
./.                     | 32bit         | 64bit
ARCH=x86                | 32bit         | 32bit
ARCH=i386               | 32bit         | 32bit
ARCH=x86_64             | 64bit         | 64bit

The general rule are that ARCH= and native architecture
takes precedence over the configuration.
So make ARCH=i386 [whatever] will always build a 32-bit
kernel no matter what the configuration says.
The configuration will be updated to 32-bit if it was
configured to 64-bit and the other way around.

This behaviour is consistent with previous behaviour so
no suprises here.

make ARCH=x86 will per default result in a 32-bit kernel
but as the only ARCH= value x86 allow the user to select
between 32-bit and 64-bit using menuconfig.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andreas Herrmann <aherrman@arcor.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
17 years agox86: reboot fixup for wrap2c board
Denys [Sat, 17 Nov 2007 15:27:02 +0000 (16:27 +0100)]
x86: reboot fixup for wrap2c board

Needed to make the wireless board, WRAP2C reboot.

Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: check boundary in count setup resource
Yinghai Lu [Sat, 17 Nov 2007 15:27:01 +0000 (16:27 +0100)]
x86: check boundary in count setup resource

need to check info->res_num less than PCI_BUS_NUM_RESOURCES, so
info->bus->resource[info->res_num] = res will not beyond of bus resource
array when acpi returns too many resource entries.

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Gary Hade <gary.hade@us.ibm.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: fix reboot with no keyboard attached
Truxton Fulton [Sat, 17 Nov 2007 15:27:01 +0000 (16:27 +0100)]
x86: fix reboot with no keyboard attached

Attempt to fix http://bugzilla.kernel.org/show_bug.cgi?id=8378

Hiroto Shibuya wrote to tell me that he has a VIA EPIA-EK10000 which
suffers from the reboot problem when no keyboard is attached.  My first
patch works for him:

  http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=59f4e7d572980a521b7bdba74ab71b21f5995538

But the latest patch does not work for him :

  http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=8b93789808756bcc1e5c90c99f1b1ef52f839a51

We found that it was necessary to also set the "disable keyboard" flag in
the command byte, as the first patch was doing.  The second patch tries to
minimally modify the command byte, but it is not enough.

Please consider this simple one-line patch to help people with low end VIA
motherboards reboot when no keyboard is attached.  Hiroto Shibuya has
verified that this works for him (as I no longer have an afflicted
machine).

Additional discussion:

Note that original patch from Truxton DOES
disable keyboard and this has been in main tree since 2.6.14, thus it must have
quite a bit of air time already.

http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.14.y.git;a=commit;h=59f4e7d572980a521b7bdba74ab71b21f5995538

Note that he only mention "System flag" in the description and comment, but
in the code, "disable keyboard" flag is set.

  outb(0x14, 0x60);       /* set "System flag" */

In 2.6.23, he made a change to read the current byte and then mask the flags,
but along this change,  he only set the "System flag" and dropped the setting
of "disable keyboard" flag.

http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.23.y.git;a=commit;h=8b93789808756bcc1e5c90c99f1b1ef52f839a51

   outb(cmd | 0x04, 0x60); /* set "System flag" */

So my request is to restore the setting of disable keyboard flag which has been
there since 2.6.14 but disappeared in 2.6.23.

Cc: Lee Garrett <lee-in-berlin@web.de>
Cc: "Hiroto Shibuya" <hiroto.shibuya@gmail.com>
Cc: Natalie Protasevich <protasnb@gmail.com>
Cc: Dmitry Torokhov <dtor@mail.ru>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Aristeu Rozanski <aris@ruivo.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: add hpet sanity checks
Thomas Gleixner [Fri, 16 Nov 2007 02:41:50 +0000 (21:41 -0500)]
x86: add hpet sanity checks

Some BIOSes advertise HPET at 0x0. We really do no want to
allocate a resource there. Check for it and leave early.

Other BIOSes tell us the HPET is at 0xfed0000000000000
instead of 0xfed00000. Add a check and fix it up with a warning
on user request.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: on x86_64, correct reading of PC RTC when update in progress in time_64.c
David P. Reed [Thu, 15 Nov 2007 01:14:50 +0000 (20:14 -0500)]
x86: on x86_64, correct reading of PC RTC when update in progress in time_64.c

Correct potentially unstable PC RTC time register reading in time_64.c

Stop the use of an incorrect technique for reading the standard PC RTC
timer, which is documented to "disconnect" time registers from the bus
while updates are in progress.  The use of UIP flag while interrupts
are disabled to protect a 244 microsecond window is one of the
Motorola spec sheet's documented ways to read the RTC time registers
reliably.

tglx: removed locking changes from original patch, as they gain nothing
(read_persistent_clock is only called during boot, suspend, resume - so
no hot path affected) and conflict with the paravirt locking scheme
(see 32bit code), which we do not want to complicate for no benefit.

Signed-off-by: David P. Reed <dpreed@reed.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: fix freeze in x86_64 RTC update code in time_64.c
David P. Reed [Wed, 14 Nov 2007 22:47:35 +0000 (17:47 -0500)]
x86: fix freeze in x86_64 RTC update code in time_64.c

Fix hard freeze on x86_64 when the ntpd service calls
update_persistent_clock()

A repeatable but randomly timed freeze has been happening in Fedora 6
and 7 for the last year, whenever I run the ntpd service on my AMD64x2
HP Pavilion dv9000z laptop.  This freeze is due to the use of
spin_lock(&rtc_lock) under the assumption (per a bad comment) that
set_rtc_mmss is called only with interrupts disabled.  The call from
ntp.c to update_persistent_clock is made with interrupts enabled.

Signed-off-by: David P. Reed <dpreed@reed.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agontp: fix typo that makes sync_cmos_clock erratic
David P. Reed [Wed, 14 Nov 2007 22:49:21 +0000 (17:49 -0500)]
ntp: fix typo that makes sync_cmos_clock erratic

Fix a typo in ntp.c that has caused updating of the persistent (RTC)
clock when synced to NTP to behave erratically.

When debugging a freeze that arises on my AMD64 machines when I
run the ntpd service, I added a number of printk's to monitor the
sync_cmos_clock procedure.  I discovered that it was not syncing to
cmos RTC every 11 minutes as documented, but instead would keep trying
every second for hours at a time.  The reason turned out to be a typo
in sync_cmos_clock, where it attempts to ensure that
update_persistent_clock is called very close to 500 msec. after a 1
second boundary (required by the PC RTC's spec). That typo referred to
"xtime" in one spot, rather than "now", which is derived from "xtime"
but not equal to it.  This makes the test erratic, creating a
"coin-flip" that decides when update_persistent_clock is called - when
it is called, which is rarely, it may be at any time during the one
second period, rather than close to 500 msec, so the value written is
needlessly incorrect, too.

Signed-off-by: David P. Reed
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agoRemove x86 merge artifact from top Makefile
Thomas Gleixner [Thu, 15 Nov 2007 06:11:12 +0000 (07:11 +0100)]
Remove x86 merge artifact from top Makefile

The x86 merge modified the tags target to handle the two separate
source directories. Remove it now that i386/x86_64 are gone completely.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: fixup cpu_info array conversion
Thomas Gleixner [Sat, 17 Nov 2007 12:18:42 +0000 (13:18 +0100)]
x86: fixup cpu_info array conversion

92cb7612aee39642d109b8d935ad265e602c0563 sets cpu_info->cpu_index to zero
for no reason. Referencing cpu_info->cpu_index now points always to CPU#0,
which is apparently not what we want.

Remove it.

Spotted-by: Zou Nan hai <nanhai.zou@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: show cpuinfo only for online CPUs
Andreas Herrmann [Thu, 1 Nov 2007 18:32:17 +0000 (19:32 +0100)]
x86: show cpuinfo only for online CPUs

Fix regressions introduced with 92cb7612aee39642d109b8d935ad265e602c0563.

It can happen that cpuinfo is displayed for CPUs that are not online or
even worse for CPUs not present at all. As an example, following was
shown for a "second" CPU of a single core K8 variant:

    processor       : 0
    vendor_id       : unknown
    cpu family      : 0
    model           : 0
    model name      : unknown
    stepping        : 0
    cache size      : 0 KB
    fpu             : yes
    fpu_exception   : yes
    cpuid level     : 0
    wp              : yes
    flags           :
    bogomips        : 0.00
    clflush size    : 0
    cache_alignment : 0
    address sizes   : 0 bits physical, 0 bits virtual
    power management:

Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: fix cpu-hotplug regression
Andreas Herrmann [Wed, 7 Nov 2007 01:12:58 +0000 (02:12 +0100)]
x86: fix cpu-hotplug regression

Commit d435d862baca3e25e5eec236762a43251b1e7ffc
("cpu hotplug: mce: fix cpu hotplug error handling")
changed the error handling in mce_cpu_callback.

In cases where not all CPUs are brought up during
boot (e.g. using maxcpus and additional_cpus parameters)
mce_cpu_callback now returns NOTFIY_BAD because
for such CPUs cpu_data is not completely filled when
the notifier is called. Thus mce_create_device fails right
at its beginning:

        if (!mce_available(&cpu_data[cpu]))
                return -EIO;

As a quick fix I suggest to check boot_cpu_data for MCE.

To reproduce this regression:

(1) boot with maxcpus=2 addtional_cpus=2 on a 4 CPU x86-64 system
(2) # echo 1 >/sys/devices/system/cpu/cpu2/online
  -bash: echo: write error: Invalid argument

dmesg shows:

_cpu_up: attempt to bring up CPU 2 failed

Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: ignore the sys_getcpu() tcache parameter
Ingo Molnar [Wed, 7 Nov 2007 17:37:48 +0000 (18:37 +0100)]
x86: ignore the sys_getcpu() tcache parameter

dont use the vgetcpu tcache - it's causing problems with tasks
migrating, they'll see the old cache up to a jiffy after the
migration, further increasing the costs of the migration.

In the worst case they see a complete bogus information from
the tcache, when a sys_getcpu() call "invalidated" the cache
info by incrementing the jiffies _and_ the cpuid info in the
cache and the following vdso_getcpu() call happens after
vdso_jiffies have been incremented.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: voyager use correct header file name
Randy Dunlap [Mon, 12 Nov 2007 05:06:02 +0000 (21:06 -0800)]
x86: voyager use correct header file name

Fix header file name for Voyager build.

In file included from arch/x86/kernel/setup_32.c:61:
include/asm-x86/mach-voyager/setup_arch.h:2:26: error: asm/setup_32.h: No such file or directory
make[1]: *** [arch/x86/kernel/setup_32.o] Error 1

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: fix smp init sections
Randy Dunlap [Mon, 12 Nov 2007 05:06:45 +0000 (21:06 -0800)]
x86: fix smp init sections

Fix Voyager section mismatch due to using __devinit instead of __cpuinit.

WARNING: vmlinux.o(.text+0xd943): Section mismatch: reference to .init.text:init_gdt (between 'voyager_smp_prepare_boot_cpu' and 'smp_vic_cmn_interrupt')

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: fix voyager_cat_init section
Randy Dunlap [Mon, 12 Nov 2007 05:06:23 +0000 (21:06 -0800)]
x86: fix voyager_cat_init section

Fix Voyager section mismatches:  voyager_cat_init() should be __init.

WARNING: vmlinux.o(.text+0xee83): Section mismatch: reference to .init.data:eprom_buf (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xeea6): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xeeac): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xeeb2): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xef4c): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xef56): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xf10f): Section mismatch: reference to .init.data:eprom_buf (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xf13b): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xf14b): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xf159): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xf1b1): Section mismatch: reference to .init.data:eprom_buf (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xf1bb): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xf1c1): Section mismatch: reference to .init.data:eprom_buf (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xf1c7): Section mismatch: reference to .init.data:eprom_buf (between 'voyager_cat_init' and 'aes_enc_blk')
WARNING: vmlinux.o(.text+0xf1e6): Section mismatch: reference to .init.data: (between 'voyager_cat_init' and 'aes_enc_blk')

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agox86: fix bogus memcpy in es7000_check_dsdt()
Thomas Gleixner [Wed, 14 Nov 2007 12:20:48 +0000 (13:20 +0100)]
x86: fix bogus memcpy in es7000_check_dsdt()

es7000_check_dst() contains a memcpy from 0, which probably should have been
a memset. Remove it and check the retunr value from acpi_get_table_header.

Noticed by: Joe Perches <joe@perches.com>

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
17 years agoLinux 2.6.24-rc3
Linus Torvalds [Sat, 17 Nov 2007 05:16:36 +0000 (21:16 -0800)]
Linux 2.6.24-rc3

17 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Sat, 17 Nov 2007 02:32:28 +0000 (18:32 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jmorris/selinux-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/selinux-2.6:
  SELinux: return EOPNOTSUPP not ENOTSUPP

17 years agoMerge branch 'for-linus' of master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Sat, 17 Nov 2007 02:32:08 +0000 (18:32 -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] 4638/1: pxa: use PXA3xx specific macros to define clks
  [ARM] remove useless setting of VM_RESERVED

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davej/cpufreq
Linus Torvalds [Sat, 17 Nov 2007 02:30:26 +0000 (18:30 -0800)]
Merge /pub/scm/linux/kernel/git/davej/cpufreq

* master.kernel.org:/pub/scm/linux/kernel/git/davej/cpufreq:
  [CPUFREQ] Fix up whitespace in conservative governor.
  [CPUFREQ] Make cpufreq_conservative handle out-of-sync events properly
  [CPUFREQ] architectural pstate driver for powernow-k8

17 years agoSELinux: return EOPNOTSUPP not ENOTSUPP
Eric Paris [Fri, 16 Nov 2007 21:35:56 +0000 (16:35 -0500)]
SELinux: return EOPNOTSUPP not ENOTSUPP

ENOTSUPP is not a valid error code in the kernel (it is defined in some
NFS internal error codes and has been improperly used other places).  In
the !CONFIG_SECURITY_SELINUX case though it is possible that we could
return this from selinux_audit_rule_init().  This patch just returns the
userspace valid EOPNOTSUPP.

Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: James Morris <jmorris@namei.org>
17 years ago[ARM] 4638/1: pxa: use PXA3xx specific macros to define clks
eric miao [Tue, 30 Oct 2007 07:10:18 +0000 (08:10 +0100)]
[ARM] 4638/1: pxa: use PXA3xx specific macros to define clks

PXA3xx uses its own clk_pxa3xx_cken_ops, modify the code to use the
PXA3xx specific macros to define its clocks

Signed-off-by: eric miao <eric.miao@marvell.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
17 years agodirty page balancing: Get rid of broken unmapped_ratio logic
Linus Torvalds [Fri, 16 Nov 2007 00:41:52 +0000 (16:41 -0800)]
dirty page balancing: Get rid of broken unmapped_ratio logic

This code harks back to the days when we didn't count dirty mapped
pages, which led us to try to balance the number of dirty unmapped pages
by how much unmapped memory there was in the system.

That makes no sense any more, since now the dirty counts include the
mapped pages.  Not to mention that the math doesn't work with HIGHMEM
machines anyway, and causes the unmapped_ratio to potentially turn
negative (which we do catch thanks to clamping it at a minimum value,
but I mention that as an indication of how broken the code is).

The code also was written at a time when the default dirty ratio was
much larger, and the unmapped_ratio logic effectively capped that large
dirty ratio a bit.  Again, we've since lowered the dirty ratio rather
aggressively, further lessening the point of that code.

Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMerge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Fri, 16 Nov 2007 00:34:00 +0000 (16:34 -0800)]
Merge branch 'master' of /linux/kernel/git/davem/net-2.6

* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [NETFILTER]: Fix NULL pointer dereference in nf_nat_move_storage()
  [SUNHME]: VLAN support for sunhme
  [CHELSIO]: Fix skb->dev setting.
  [NETFILTER]: fix compat_nf_sockopt typo
  [INET]: Fix potential kfree on vmalloc-ed area of request_sock_queue
  [VIA_VELOCITY]: Don't oops on MTU change.
  iwl4965: fix not correctly dealing with hotunplug
  rt2x00: Fix chipset revision validation
  iwl3945: place CCK rates in front of OFDM for supported rates
  mac80211: Fix queuing of scan containing a SSID

17 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Fri, 16 Nov 2007 00:08:49 +0000 (16:08 -0800)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] N32 needs to use the compat version of sys_nfsservctl.
  [MIPS] irq_cpu: use handle_percpu_irq handler to avoid dropping interrupts.
  [MIPS] Sibyte: Fix name of clocksource.
  [MIPS] SNI: s/achknowledge/acknowledge/
  [MIPS] Makefile: Fix canonical system names
  [MIPS] vpe: handle halting TCs in an errata safe way.
  [MIPS] Sibyte: Stop timers before programming next even.
  [MIPS] Sibyte: Increase minimum oneshot timer interval to two ticks.
  [MIPS] Lasat: Fix overlap of interrupt number ranges.
  [MIPS] SNI PCIT CPLUS: workaround for b0rked irq wiring of onboard PCI bus 1
  [MIPS] Fix shadow register support.
  [MIPS] Change get_cycles to always return 0.
  [MIPS] Fix typo in R3000 TRACE_IRQFLAGS code
  [MIPS] Sibyte: Replace use of removed IO_SPACE_BASE with IOADDR.
  [MIPS] iounmap if in vr41xx_pciu_init() pci clock is over 33MHz
  [MIPS] BCM1480: Remove duplicate acknowledge of timer interrupt.
  [MIPS] Sibyte: pin timer interrupt to their cores.
  [MIPS] Qemu: Add early printk, your friend in a cold night.
  [MIPS] Convert reference to mem_map to pfn_to_page().
  [MIPS] Sibyte: resurrect old cache hack.

17 years ago[NETFILTER]: Fix NULL pointer dereference in nf_nat_move_storage()
Evgeniy Polyakov [Thu, 15 Nov 2007 23:52:32 +0000 (15:52 -0800)]
[NETFILTER]: Fix NULL pointer dereference in nf_nat_move_storage()

Reported by Chuck Ebbert as:

https://bugzilla.redhat.com/show_bug.cgi?id=259501#c14

This routine is called each time hash should be replaced, nf_conn has
extension list which contains pointers to connection tracking users
(like nat, which is right now the only such user), so when replace takes
place it should copy own extensions. Loop above checks for own
extension, but tries to move higer-layer one, which can lead to above
oops.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[SUNHME]: VLAN support for sunhme
Chris Poon [Thu, 15 Nov 2007 23:38:45 +0000 (15:38 -0800)]
[SUNHME]: VLAN support for sunhme

This patch enables VLAN support on sunhme by increasing BMAC_TXMAX/BMAC_RXMAX
and allocating extra space via skb_put for the VLAN header.

Signed-off-by: Chris Poon <dev-null@telus.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[MIPS] N32 needs to use the compat version of sys_nfsservctl.
Ralf Baechle [Thu, 15 Nov 2007 22:20:33 +0000 (22:20 +0000)]
[MIPS] N32 needs to use the compat version of sys_nfsservctl.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] irq_cpu: use handle_percpu_irq handler to avoid dropping interrupts.
Ralf Baechle [Thu, 15 Nov 2007 19:37:15 +0000 (19:37 +0000)]
[MIPS] irq_cpu: use handle_percpu_irq handler to avoid dropping interrupts.

This matters to any sort of device that is wired to one of the CPU
interrupt pins on an SMP system.  Typically the scenario is most easily
triggered with the count/compare timer interrupt where the same interrupt
number and thus irq_desc is used on each processor.

   CPU A CPU B

   do_IRQ()
   generic_handle_irq()
   handle_level_irq()
   spin_lock(desc_lock)
   set IRQ_INPROGRESS
   spin_unlock(desc_lock)
do_IRQ()
generic_handle_irq()
handle_level_irq()
spin_lock(desc_lock)
IRQ_INPROGRESS set => bail out
   spin_lock(desc_lock)
   clear IRQ_INPROGRESS
   spin_unlock(desc_lock)

In case of the cp0 compare interrupt this means the interrupt will be
acked and not handled or re-armed on CPU b, so there won't be any timer
interrupt until the count register wraps around.

With kernels 2.6.20 ... 2.6.23 we usually were lucky that things were just
working right on VSMP because the count registers are synchronized on
bootup so it takes something that disables interrupts for a long time on
one processor to trigger this one.

For scenarios where an interrupt is multicasted or broadcasted over several
CPUs the existing code was safe and the fix will break it.  There is no
way to know in the interrupt controller code because it is abstracted from
the platform code.  I think we do not have such a setup currently, so this
should be ok.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Sibyte: Fix name of clocksource.
Ralf Baechle [Tue, 13 Nov 2007 00:35:13 +0000 (00:35 +0000)]
[MIPS] Sibyte: Fix name of clocksource.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] SNI: s/achknowledge/acknowledge/
Maciej W. Rozycki [Mon, 12 Nov 2007 17:32:48 +0000 (17:32 +0000)]
[MIPS] SNI: s/achknowledge/acknowledge/

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Makefile: Fix canonical system names
Maciej W. Rozycki [Mon, 12 Nov 2007 17:30:52 +0000 (17:30 +0000)]
[MIPS] Makefile: Fix canonical system names

The GNU `config.guess' uses "linux-gnu" as the canonical system name.
Fix the list of compiler prefixes checked to spell it correctly.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] vpe: handle halting TCs in an errata safe way.
Nigel Stephens [Thu, 8 Nov 2007 13:25:51 +0000 (13:25 +0000)]
[MIPS] vpe: handle halting TCs in an errata safe way.

Adds a JR.HB after halting a TC, to ensure that the TC has really halted.
only modifies the TCSTATUS register when the TC is safely halted.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Sibyte: Stop timers before programming next even.
Ralf Baechle [Sun, 11 Nov 2007 17:33:55 +0000 (17:33 +0000)]
[MIPS] Sibyte: Stop timers before programming next even.

We have no guarantee by the generic time code that the timer is stopped
when the ->next_event method is called.  Modifying the Timer Initial Count
register while the timer is enabled has UNPREDICTABLE effect according to
the BCM1250/BCM1125/BCM1125H User Manual.  So stop the timer before
reprogramming.

This is a paranoia fix; no ill effects have been observed previously.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Sibyte: Increase minimum oneshot timer interval to two ticks.
Ralf Baechle [Sun, 11 Nov 2007 17:24:46 +0000 (17:24 +0000)]
[MIPS] Sibyte: Increase minimum oneshot timer interval to two ticks.

For the old minimum of a single tick a value of zero would be programmed
into the init value register which in the BCM1250/BCM1125/BCM1125H User
Manual in the Timer Special Cases section is documented to have
UNPREDICTABLE effect.

Observable sympthoms of this bug were hangs of several seconds on the
console during bootup and later if both dyntick and highres timer options
were activated.

In theory contiguous mode of the timers is also affected but in an act of
hopeless lack of realism I'll assume nobody will ever configure a KERNEL
for HZ > 500kHz but if so I leave that to evolution to sort out.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Lasat: Fix overlap of interrupt number ranges.
Yoichi Yuasa [Fri, 9 Nov 2007 09:42:35 +0000 (18:42 +0900)]
[MIPS] Lasat: Fix overlap of interrupt number ranges.

The range of MIPS_CPU IRQ and the range of LASAT IRQ overlap.

Signed-off-by: Yoichi Yuasa <yoichi_yuasa@tripeaks.co.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] SNI PCIT CPLUS: workaround for b0rked irq wiring of onboard PCI bus 1
Thomas Bogendoerfer [Thu, 8 Nov 2007 21:09:11 +0000 (22:09 +0100)]
[MIPS] SNI PCIT CPLUS: workaround for b0rked irq wiring of onboard PCI bus 1

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Fix shadow register support.
Ralf Baechle [Thu, 8 Nov 2007 18:02:29 +0000 (18:02 +0000)]
[MIPS] Fix shadow register support.

Shadow register support would not possibly have worked on multicore
systems.  The support code for it was also depending not on MIPS R2 but
VSMP or SMTC kernels even though it makes perfect sense with UP kernels.

SR sets are a scarce resource and the expected usage pattern is that
users actually hardcode the register set numbers in their code.  So fix
the allocator by ditching it.  Move the remaining CPU probe bits into
the generic CPU probe.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Change get_cycles to always return 0.
Ralf Baechle [Thu, 8 Nov 2007 11:37:47 +0000 (11:37 +0000)]
[MIPS] Change get_cycles to always return 0.

This avoids us executing an mfc0 c0_count instruction on processors which
don't have but also on certain R4000 and R4400 versions where reading from
the count register just in the very moment when its value equals
c0_compare will result in the timer interrupt getting lost.

There is still a number of users of get_cycles remaining outside the
arch code:

crypto/tcrypt.c:                start = get_cycles();
crypto/tcrypt.c:                end = get_cycles();
crypto/tcrypt.c:                start = get_cycles();
crypto/tcrypt.c:                end = get_cycles();
crypto/tcrypt.c:                start = get_cycles();
crypto/tcrypt.c:                end = get_cycles();
drivers/char/hangcheck-timer.c: return get_cycles();
drivers/char/hangcheck-timer.c: printk("Hangcheck: Using get_cycles().\n");
drivers/char/random.c:  sample.cycles = get_cycles();
drivers/input/joystick/analog.c:#define GET_TIME(x)     do { x = get_cycles(); }
include/linux/arcdevice.h:          _x = get_cycles(); \
include/linux/arcdevice.h:          _y = get_cycles(); \
mm/slub.c:      if (!s->defrag_ratio || get_cycles() % 1024 > s->defrag_ratio)
mm/slub.c:      p += 64 + (get_cycles() & 0xff) * sizeof(void *);

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Fix typo in R3000 TRACE_IRQFLAGS code
Atsushi Nemoto [Tue, 6 Nov 2007 16:08:48 +0000 (01:08 +0900)]
[MIPS] Fix typo in R3000 TRACE_IRQFLAGS code

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Sibyte: Replace use of removed IO_SPACE_BASE with IOADDR.
Ralf Baechle [Tue, 6 Nov 2007 00:43:51 +0000 (00:43 +0000)]
[MIPS] Sibyte: Replace use of removed IO_SPACE_BASE with IOADDR.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] iounmap if in vr41xx_pciu_init() pci clock is over 33MHz
Roel Kluin [Sun, 4 Nov 2007 12:00:06 +0000 (13:00 +0100)]
[MIPS] iounmap if in vr41xx_pciu_init() pci clock is over 33MHz

iounmap if pci clock is over 33MHz.  Cosmetic because the iomap() in this
case is just a bit of address magic.

Signed-off-by: Roel Kluin <12o3l@tiscali.nl>
Acked-by: Yoichi Yuasa <yoichi_yuasa@tripeaks.co.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] BCM1480: Remove duplicate acknowledge of timer interrupt.
Ralf Baechle [Mon, 5 Nov 2007 00:29:45 +0000 (00:29 +0000)]
[MIPS] BCM1480: Remove duplicate acknowledge of timer interrupt.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Sibyte: pin timer interrupt to their cores.
Ralf Baechle [Mon, 5 Nov 2007 00:18:05 +0000 (00:18 +0000)]
[MIPS] Sibyte: pin timer interrupt to their cores.

Or strange things will happen.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Qemu: Add early printk, your friend in a cold night.
Ralf Baechle [Sun, 4 Nov 2007 04:49:44 +0000 (04:49 +0000)]
[MIPS] Qemu: Add early printk, your friend in a cold night.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Convert reference to mem_map to pfn_to_page().
Ralf Baechle [Sun, 4 Nov 2007 04:42:03 +0000 (04:42 +0000)]
[MIPS] Convert reference to mem_map to pfn_to_page().

This was crashing the combination of highmem and sparsemem.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Sibyte: resurrect old cache hack.
Ralf Baechle [Sat, 3 Nov 2007 01:01:37 +0000 (01:01 +0000)]
[MIPS] Sibyte: resurrect old cache hack.

The recent switch of the Sibyte SOCs from the processor specific cache
managment code in c-sb1.c to c-r4k.c lost this old hack

    [MIPS] Hack for SB1 cache issues

    Removing flush_icache_page a while ago broke SB1 which was using an empty
    flush_data_cache_page function.  This glues things well enough so a more
    efficient but also more intrusive solution can be found later.

Signed-Off-By: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
in the hope it was no longer needed.  As it turns it still is so resurrect
it until there is a better solution.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[CHELSIO]: Fix skb->dev setting.
Divy Le Ray [Thu, 15 Nov 2007 23:06:32 +0000 (15:06 -0800)]
[CHELSIO]: Fix skb->dev setting.

eth_type_trans() now sets skb->dev.  Access skb->def after it gets
set.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: fix compat_nf_sockopt typo
Patrick McHardy [Thu, 15 Nov 2007 22:29:21 +0000 (14:29 -0800)]
[NETFILTER]: fix compat_nf_sockopt typo

It should pass opt to the ->get/->set functions, not ops.

Tested-by: Luca Tettamanti <kronos.it@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched
Linus Torvalds [Thu, 15 Nov 2007 20:14:52 +0000 (12:14 -0800)]
Merge git://git./linux/kernel/git/mingo/linux-2.6-sched

* git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched:
  sched: reorder SCHED_FEAT_ bits
  sched: make sched_nr_latency static
  sched: remove activate_idle_task()
  sched: fix __set_task_cpu() SMP race
  sched: fix SCHED_FIFO tasks & FAIR_GROUP_SCHED
  sched: fix accounting of interrupts during guest execution on s390

17 years agosched: reorder SCHED_FEAT_ bits
Ingo Molnar [Thu, 15 Nov 2007 19:57:40 +0000 (20:57 +0100)]
sched: reorder SCHED_FEAT_ bits

reorder SCHED_FEAT_ bits so that the used ones come first. Makes
tuning instructions easier.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years agosched: make sched_nr_latency static
Adrian Bunk [Thu, 15 Nov 2007 19:57:40 +0000 (20:57 +0100)]
sched: make sched_nr_latency static

sched_nr_latency can now become static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years agosched: remove activate_idle_task()
Dmitry Adamushko [Thu, 15 Nov 2007 19:57:40 +0000 (20:57 +0100)]
sched: remove activate_idle_task()

cpu_down() code is ok wrt sched_idle_next() placing the 'idle' task not
at the beginning of the queue.

So get rid of activate_idle_task() and make use of activate_task() instead.
It is the same as activate_task(), except for the update_rq_clock(rq) call
that is redundant.

Code size goes down:

   text    data     bss     dec     hex filename
  47853    3934     336   52123    cb9b sched.o.before
  47828    3934     336   52098    cb82 sched.o.after

Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years agosched: fix __set_task_cpu() SMP race
Dmitry Adamushko [Thu, 15 Nov 2007 19:57:40 +0000 (20:57 +0100)]
sched: fix __set_task_cpu() SMP race

Grant Wilson has reported rare SCHED_FAIR_USER crashes on his quad-core
system, which crashes can only be explained via runqueue corruption.

there is a narrow SMP race in __set_task_cpu(): after ->cpu is set up to
a new value, task_rq_lock(p, ...) can be successfuly executed on another
CPU. We must ensure that updates of per-task data have been completed by
this moment.

this bug has been hiding in the Linux scheduler for an eternity (we never
had any explicit barrier for task->cpu in set_task_cpu() - so the bug was
introduced in 2.5.1), but only became visible via set_task_cfs_rq() being
accidentally put after the task->cpu update. It also probably needs a
sufficiently out-of-order CPU to trigger.

Reported-by: Grant Wilson <grant.wilson@zen.co.uk>
Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years agosched: fix SCHED_FIFO tasks & FAIR_GROUP_SCHED
Oleg Nesterov [Thu, 15 Nov 2007 19:57:40 +0000 (20:57 +0100)]
sched: fix SCHED_FIFO tasks & FAIR_GROUP_SCHED

Suppose that the SCHED_FIFO task does

switch_uid(new_user);

Now, p->se.cfs_rq and p->se.parent both point into the old
user_struct->tg because sched_move_task() doesn't call set_task_cfs_rq()
for !fair_sched_class case.

Suppose that old user_struct/task_group is freed/reused, and the task
does

sched_setscheduler(SCHED_NORMAL);

__setscheduler() sets fair_sched_class, but doesn't update
->se.cfs_rq/parent which point to the freed memory.

This means that check_preempt_wakeup() doing

while (!is_same_group(se, pse)) {
se = parent_entity(se);
pse = parent_entity(pse);
}

may OOPS in a similar way if rq->curr or p did something like above.

Perhaps we need something like the patch below, note that
__setscheduler() can't do set_task_cfs_rq().

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years agosched: fix accounting of interrupts during guest execution on s390
Christian Borntraeger [Thu, 15 Nov 2007 19:57:39 +0000 (20:57 +0100)]
sched: fix accounting of interrupts during guest execution on s390

Currently the scheduler checks for PF_VCPU to decide if this timeslice
has to be accounted as guest time. On s390 host interrupts are not
disabled during guest execution. This causes theses interrupts to be
accounted as guest time if CONFIG_VIRT_CPU_ACCOUNTING is set. Solution
is to check if an interrupt triggered account_system_time. As the tick
is timer interrupt based, we have to subtract hardirq_offset.

I tested the patch on s390 with CONFIG_VIRT_CPU_ACCOUNTING and on
x86_64. Seems to work.

CC: Avi Kivity <avi@qumranet.com>
CC: Laurent Vivier <Laurent.Vivier@bull.net>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years agoMerge branch 'i2c-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6
Linus Torvalds [Thu, 15 Nov 2007 19:01:07 +0000 (11:01 -0800)]
Merge branch 'i2c-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6

* 'i2c-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6:
  i2c/eeprom: Recognize VGN as a valid Sony Vaio name prefix
  i2c/eeprom: Hide Sony Vaio serial numbers
  i2c-pasemi: Fix NACK detection
  i2c-pasemi: Replace obsolete "driverfs" reference with "sysfs"
  i2c: Make i2c_check_addr static
  i2c-dev: Unbound new-style i2c clients aren't busy
  i2c-dev: "how does it work" comments

17 years agoi2c/eeprom: Recognize VGN as a valid Sony Vaio name prefix
Jean Delvare [Thu, 15 Nov 2007 18:24:03 +0000 (19:24 +0100)]
i2c/eeprom: Recognize VGN as a valid Sony Vaio name prefix

Recent (i.e. 2005 and later) Sony Vaio laptops have names beginning
with VGN rather than PCG. Update the eeprom driver so that it
recognizes these.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
17 years agoi2c/eeprom: Hide Sony Vaio serial numbers
Jean Delvare [Thu, 15 Nov 2007 18:24:03 +0000 (19:24 +0100)]
i2c/eeprom: Hide Sony Vaio serial numbers

The sysfs interface to DMI data takes care to not make the system
serial number and UUID world-readable, presumably due to privacy
concerns. For consistency, we should not let the eeprom driver
export these same strings to the world on Sony Vaio laptops.
Instead, only make them readable by root, as we already do for BIOS
passwords.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
17 years agoi2c-pasemi: Fix NACK detection
Olof Johansson [Thu, 15 Nov 2007 18:24:02 +0000 (19:24 +0100)]
i2c-pasemi: Fix NACK detection

Turns out we don't actually check the status to see if there was a
device out there to talk to, just if we had a timeout when doing so.

Add the proper check, so we don't falsly think there are devices
on the bus that are not there, etc.

Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
17 years agoi2c-pasemi: Replace obsolete "driverfs" reference with "sysfs"
Robert P. J. Day [Thu, 15 Nov 2007 18:24:02 +0000 (19:24 +0100)]
i2c-pasemi: Replace obsolete "driverfs" reference with "sysfs"

Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
17 years agoi2c: Make i2c_check_addr static
Jean Delvare [Thu, 15 Nov 2007 18:24:02 +0000 (19:24 +0100)]
i2c: Make i2c_check_addr static

i2c_check_addr is only used inside i2c-core now, so we can make it
static and stop exporting it. Thanks to David Brownell for noticing.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
17 years agoi2c-dev: Unbound new-style i2c clients aren't busy
Jean Delvare [Thu, 15 Nov 2007 18:24:01 +0000 (19:24 +0100)]
i2c-dev: Unbound new-style i2c clients aren't busy

Let i2c-dev deal properly with new-style i2c clients. Instead of
considering them always busy, it needs to check wether a driver is
bound to them or not.

This is still not completely correct, as the client could become
busy later, but the same problem already existed before new-style
clients were introduced. We'll want to fix it someday.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
17 years agoi2c-dev: "how does it work" comments
David Brownell [Thu, 15 Nov 2007 18:24:01 +0000 (19:24 +0100)]
i2c-dev: "how does it work" comments

This adds some "how does this work" comments to the i2c-dev driver,
plus separators between the three main components:

  - The parallel list of i2c_adapters ("i2c_dev_list"), each of which
    gets a "struct i2c_dev" and a /dev/i2c-X character special file.

  - An i2cdev_driver gets adapter add/remove notifications, which are
    used to maintain that list of adapters.

  - Special file operations, which let userspace talk either directly to
    the adapter (for i2c_msg operations) or through cached addressing info
    using an anonymous i2c_client (never registered anywhere).

Plus there's the usual module load/unload record keeping.

After making sense of this code, I think that the anonymous i2c_client
is pretty shady.  But since it's never registered, using this code with
a system set up for "new style" I2C drivers is no more complicated than
always using the I2C_SLAVE_FORCE ioctl (instead of I2C_SLAVE).

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
17 years agoRevert "skge: fix ram buffer size calculation"
Linus Torvalds [Thu, 15 Nov 2007 16:44:36 +0000 (08:44 -0800)]
Revert "skge: fix ram buffer size calculation"

This reverts commit 7fb7ac241162dc51ec0f7644d4a97b2855213c32.

Heikki Orsila reports that it causes a regression:

  "Doing

nc host port < /dev/zero

   on a sending machine (not skge) to an skge machine that is receiving:

nc -l -p port >/dev/null

   with ~60 MiB/s speed, causes the interface go malfunct. A slow
   transfer doesn't cause a problem."

See

http://bugzilla.kernel.org/show_bug.cgi?id=9321

for some more information.

There is a workaround (also reported by Heikki):

  "After some fiddling, I noticed that not changing the register write
   order on patch:

   +       skge_write32(hw, RB_ADDR(q, RB_END), end);
           skge_write32(hw, RB_ADDR(q, RB_WP), start);
           skge_write32(hw, RB_ADDR(q, RB_RP), start);
   -       skge_write32(hw, RB_ADDR(q, RB_END), end);

   fixes the visible effect..  Possibly not the root cause of the
   problem, but changing the order back fixes networking here."

but that has yet to be ack'ed or tested more widely, so the whole
problem-causing commit gets reverted until this is resolved properly.

Bisected-and-requested-by: Heikki Orsila <shdl@zakalwe.fi>
Cc: Stephen Hemminger <shemminger@linux-foundation.org>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hskinnemoen...
Linus Torvalds [Thu, 15 Nov 2007 16:37:09 +0000 (08:37 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/hskinnemoen/avr32-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hskinnemoen/avr32-2.6:
  [AVR32] Export intc_get_pending symbol
  [AVR32] Add missing bit in PCCR sysreg
  [AVR32] Fix duplicate clock index in at32ap machine code
  [AVR32] remove UID16 option
  [AVR32] Turn off debugging in SMC driver
  Extend I/O resource for wdt0 for at32ap7000 devices
  [AVR32] pcmcia ioaddr_t should be 32 bits on AVR32

17 years agoslob: fix memory corruption
Nick Piggin [Thu, 15 Nov 2007 11:32:04 +0000 (12:32 +0100)]
slob: fix memory corruption

Previously, it would be possible for prev->next to point to
&free_slob_pages, and thus we would try to move a list onto itself, and
bad things would happen.

It seems a bit hairy to be doing list operations with the list marker as
an entry, rather than a head, but...

this resolves the following crash:

  http://bugzilla.kernel.org/show_bug.cgi?id=9379

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>