From: Chris Bajumpaa <cbajumpa@or8.net>

This patch fixes a problem with the TSC failing on via686a/KX133
motherboards either reverting to using the pit or deadlocking the machine
alltogether under heavy load.  (Specifically Abit KA7/KA7-100).

Message from the log:
Dec 18 18:20:37 grinder kernel: Losing too many ticks!
Dec 18 18:20:37 grinder kernel: TSC cannot be used as a timesource. (Are
you running with SpeedStep?)
Dec 18 18:20:37 grinder kernel: Falling back to a sane timesource.

The snippet of code that was missing from timer_tsc.c comes from
timer_pit.c.

From: "Maciej Soltysiak" <solt@dns.toxicfilms.tv>

make gconfig causes this:

scripts/kconfig/gconf.c: In function `on_treeview1_button_press_event':
scripts/kconfig/gconf.c:1175: warning: passing arg 1 of
`gtk_widget_grab_focus' from incompatible pointer type

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

The fatfs was ignoring the I/O error on two points. If I/O error
happen while checking a free block entries, this checks the all
entries, and reports an I/O error on each entry.

This problem became cause of the disk full by syslogd.

From: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>

From: Matthew Dobson <colpatch@us.ibm.com>

The code to parse the EBDA (Extended BIOS Data Area) for Summit boxen is
broken because it does not handle Lookout boxen (external boxes full of
additional PCI slots).  This patch cleans up some ugliness in
arch/i386/kernel/summit.c as well as fixing the code to handle Lookout boxen
and various other configurations of PCI buses.  Without this, Summit PCI bus
to node mappings are totally hosed with Lookout boxen attatched.  This patch
depends upon the GENERICARCH fix.

From: Matthew Dobson <colpatch@us.ibm.com>

Trying to build a kernel with both CONFIG_X86_GENERICARCH and
CONFIG_NUMA on results in a compile error.  This patch fixes that build 
problem by adding a config option for NUMA on Summit which is used to
correctly conditionally compile arch/i386/kernel/summit.c and properly
ifdef the function calls used in generic code.  Please apply.

Running make -j24 bzImage
arch/i386/mach-generic/built-in.o: In function `mps_oem_check':
arch/i386/mach-generic/built-in.o(.text+0x3ce): undefined reference to
`setup_summit'
arch/i386/mach-generic/built-in.o: In function `acpi_madt_oem_check':
arch/i386/mach-generic/built-in.o(.text+0x468): undefined reference to
`setup_summit'
make: *** [.tmp_vmlinux1] Error 1

From: Adrian Bunk <bunk@fs.tum.de>

  CC [M]  drivers/isdn/eicon/eicon_mod.o
drivers/isdn/eicon/eicon_mod.c: In function `eicon_exit':
drivers/isdn/eicon/eicon_mod.c:1362: warning: implicit declaration of
function `mca_mark_as_unused'

From: Stephen Smalley <sds@epoch.ncsc.mil>

This patch against 2.6.0 adds a control to the SELinux module over the
inheritance of signal-related state upon security context transitions in
order to protect the new security context.  If the permission is not
granted by the policy for a given pair of contexts, then transitions
between them will clear itimers, flush all pending signals, forcibly
flush signal handlers, and unblock all signals.  Roland McGrath provided
input and feedback on the patch. 
Please apply, or let James Morris and me know if you'd like this to be
resubmitted later.  Thanks.

From: Stephen Smalley <sds@epoch.ncsc.mil>

This patch removes the use of nameidata by selinux_inode_permission, as this
appears to be unsafe in certain cases (e.g.  path_walk call from
rpc_lookup_parent), leading to an Oops if d_path is subsequently called by
avc_audit on the (mnt,dentry) pair to generate a pathname for an audit
message.

The change does not affect the ability of SELinux to perform its permission
check (which only requires the inode), only the set of information that is
available for audit messages.  We'll investigate better approaches for the
SELinux audit generation in the future.

From: Stephen Smalley <sds@epoch.ncsc.mil>

This patch reduces the full capability check in the SELinux module for the
KDSKBENT/SENT ioctls to only check the corresponding SELinux permission,
avoiding a change to the Linux permissions model for these operations.

From: Stephen Smalley <sds@epoch.ncsc.mil>

This patch fixes the SELinux build for "make O=..." by removing the use of
-include and eliminating the global.h file, adding appropriate individual
#include's to the various files in the security/selinux/ss subdirectory.
The compilation error was reported by Sam Ravnborg and again by Adrian
Bunk.

From: Michael Hunold <hunold@convergence.de>

> Thomas Meyer <tantalus@gmx.ch> wrote:
>>there seems to be a problem with a DVB driver.
>>Dec 29 17:15:30 jupiter kernel: dvb_ttpci: Unknown symbol request_firmware

The dvb-ttpci/av7110 driver needs the firmware_class stuff to work now.

IMHO the appropriate fix is to let the driver automatically select the
firmware loader via Kconfig's SELECT facility.

The dvb-ttusb-dec driver should behave like this, too.  (It currently
depends on the FW_LOADER, which is not "the right thing" (tm))

From: Michael Hunold <hunold@linuxtv.org>

Add some valuable documentation about the DVB subsystem, the supported cards,
a faq, ...

From: Michael Hunold <hunold@linuxtv.org>

Use a kernel thread instead of schedule_work() when waiting for the firmware
upload to happen

From: Michael Hunold <hunold@linuxtv.org>

remove various LINUX_VERSION_CODE code paths

fix compile bug in new bt8xx/Makefile

From: Michael Hunold <hunold@linuxtv.org>

add support for the DEC3000-s (Alex Woods)

use the hotplug firmware loader for 2.6 kernels instead of compiling the
firmware into the module (Alex Woods)

From: Michael Hunold <hunold@linuxtv.org>

use new firmware_class firmware loading facilities in dvb-ttpci/av7110 driver

From: Michael Hunold <hunold@linuxtv.org>

av7110: follow changes in saa7146_core regarding saa7146_set_gpio() and
saa7146_wait_for_debi_done() function

av7110: increased I2C speed to 275 kHz, follow introduction of
SAA7146_I2C_SHORT_DELAY flag to speed up I2C access

budget: make budget-ci use this gpio function and the new wait_...()
function, this fixes
BORROWED_FROM_AV7110_H_BUT_REALLY_BELONGS_IN_SAA7146_DEFS_H remark

budget: use alternative values for BRS setup on budget cards (by Rober
Schlabbach)

budget: remote control table should be filled completely.  at least populate
the entries that come with the standard Hauppauge RC (Jamie Honan)

ttpci-eeprom: add proper MODULE_LICENSE("GPL") so we don't taint the kernel
anymore

From: Michael Hunold <hunold@linuxtv.org>

alps_tdmb7, cx24110: use correct delay values, don't divide by HZ when using
dvb_delay(), found by Artur Skawina

alps_tdmb7: set FE_HAS_LOCK only when all low-order bits are valid

mt312: patch for the mt312 module, targeting the VP310: reduced heat,
implement "auto" inversion mode, remove debugging verbosity, add module
parameter for debugging (Augusto Cardoso)

nxt6000: code review and beautification, use per i2c-adapater void pointer
for private data in nxt_attach() / nxt_detach, fix frontend private data
handling.  patch by Mikael Rosbacke <rosbacke at nada.kth.se>

sp887x: firmware loader implementation contributed by Martin Stubbs, C99
comile fixes by Wolfgang Thiel

stv0299: Added new module parameter to choose between BER and UCBLOCKs error
monitoring since the STV0299 can't do both at once, Added modifications based
on the recommended settings in the SU1278 datasheet.

tda1004x: remove FE_CAN_INVERSION_AUTO

ves1820:completed nokia board support, increased some delays to get constant
results,set default pwm value to 0x48 for boards which don't have an eeprom
(by Andreas Oberritter)

From: Michael Hunold <hunold@linuxtv.org>

add a parameter to dvb_filter_pes2ts function to specify whether the packet
is a payload unit start or not.

new section demux code by emard

change license GPL -> LGPL for dvb_ringbuffer, like all other DVB core files

fix rare crash on invalid packets, patch by Asier Aguirre

i2c: copy the data variable as well on register client so that detach sees it.

From: Michael Hunold <hunold@linuxtv.org>

Complete revamp of the original driver: code beautification + linux coding
sytle, full diseqc support, hardware filtering support, support for different
card revisions and lots of other stuff.

From: Michael Hunold <hunold@linuxtv.org>

new DVB driver for bt878 based "budget" DVB cards (Nebula, Pinnacle PCTV,
Twinhan DST)

From: Michael Hunold <hunold@linuxtv.org>

fix a bunch of race conditions and locking bugs in video and vbi capture
code on device closure

use vmalloc_32() instead of vmalloc() in saa7146_vmalloc_build_pgtable().
this makes sure that the pagetable is in lowmem kernel memory

i2c timeout fix by Gerd Knorr 

SAA7146_I2C_SHORT_DELAY flag to speed up I2C access by Oliver Endriss

move saa7146_set_gpio() from saa7146_vv to saa7146_core, it's needed by DVB
budget drivers

add "new" saa7146_wait_for_debi_done() function, remove other versions from
av7110 and budget.ci

make budget-ci use this gpio function and the new wait_...() function,

make saa7146_pgtable_build_single() deliver a return code, make sanity
checks of the arguments

sanitize enabling of video input pins and i2c pins, use some default
values, so the hardware is always in a sane state

remove SAA7146_EXT_SWAP_ODD_EVEN flag + handling, fix the hardware
initialization instead

change minimal picture size to 48x32 just like other drivers

set up arbitrition control for video dma3 correctly

remove unnecessary code for capture to framebuffer memory, it's handled in
the generic code

From: Michael Hunold <hunold@convergence.de>

Remove av7110 firmware

From: Jan Kara <jack@suse.cz>

Fix en error exit path so that we correctly unaccount for quota-related space
reservations on ENOSPC.

Fix a warning on 64-bit machines.

From: "Zhu, Yi" <yi.zhu@intel.com>

The "make rpm" rule in top Makefile isn't aware of the enviorment ARCH.  For
example, people issue "make ARCH=ia64" to compile the ia64 kernel on i386
platform for cross compilation.  This works pretty well now.  But if one uses
"make rpm ARCH=ia64", it will fail.  Because current rpm rule in Makefile and
mkspec are not aware of ARCH.

From: john stultz <johnstul@us.ibm.com>

Converts all uses of jiffies to jiffies_64 in x86 time sources to avoid
jiffies overflow problems.

(Contributions from Tim Schmielau <tim@physik3.uni-rostock.de>)

fgets puts a \n in the buffer before the terminating \0.

into nuts.ninka.net:/disk1/davem/BK/sparc-2.6

into home.osdl.org:/home/torvalds/v2.5/linux

The attached patch adds support for suspend/resume to the sis900 driver.
With this patch on resume the NIC is fully configured and operational,
before a module reload was needed because of the complete lack of
suspend/resume callbacks.

I added two functions, sis900_suspend and sis900_resume, with their
pointers in struct pci_driver. A vector of 16 u32 was then needed to the
to keep PCI data during suspend. I added it in struct sis900_private.
I updated the revision number to reflect my changes. 
Looking at the code I also killed three typos.

The patch doesn't touch any other code.

Since I don't know anything on ethernet drivers the rule 'works for me'
is fully valid.

into home.osdl.org:/home/torvalds/v2.5/linux

into home.osdl.org:/home/torvalds/v2.5/linux

From: Joe Thornber <thornber@sistina.com>

You can no longer call dm_table_event() from interrupt context.