[PATCH] 8xx: avoid "dcbst" misbehaviour with unpopulated TLB

The proposed _tlbie call at update_mmu_cache() is safe because:

Addresses for which update_mmu_cache() gets invocated are never inside the
static kernel virtual mapping, meaning that there is no risk for the
_tlbie() here to be thrashing the pinned entry, as Dan suspected.

The intermediate TLB state in which this bug can be triggered is not
visible by userspace or any other contexts, except the page fault handling
path.  So there is no need to worry about userspace dcbxxx users.

The other solution to this is to avoid dcbst misbehaviour in the first
place, which involves changing in-kernel "dcbst" callers to use 8xx
specific SPR's.

Summary:

On 8xx, cache control instructions (particularly "dcbst" from
flush_dcache_icache) fault as write operation if there is an unpopulated
TLB entry for the address in question.  To workaround that, we invalidate
the TLB here, thus avoiding dcbst misbehaviour.

Signed-off-by: Marcelo Tosatti <marcelo.tosatti@cyclades.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] mips: fixed try_to_freeze build error

arch/mips/kernel/signal.c: In function 'do_signal':
arch/mips/kernel/signal.c:460: error: too many arguments to function 'try_to_freeze'

Signed-off-by: Yoichi Yuasa <yuasa@hh.iij4u.or.jp>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fix WANT_PAGE_VIRTUAL in memmap_init

I spotted this issue while in memmap_init last week.  I can't say the
change has any test coverage by me.  start_pfn was formerly used in main
"for" loop.  The fix is replace start_pfn with pfn.

Signed-off-by: Bob Picco <bob.picco@hp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc32: Fix compiling of sandpoint platform

Lost a curly brace in translation.  Everything is better now.

Signed-off-by: Matt McClintock <msm@freescale.com>
Signed-off-by: Kumar Gala <kumar.gala@freescale.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge master.kernel.org:/home/rmk/linux-2.6-serial

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

Merge 'upstream' branch of /linux/kernel/git/jgarzik/netdev-2.6

Merge 'upstream' branch of /linux/kernel/git/jgarzik/libata-dev

Merge 'upstream' branch of /linux/kernel/git/jgarzik/misc-2.6

Merge /linux/kernel/git/dtor/input.git manually

Some manual fixups required due to clashes with the PF_FREEZE cleanups.

[PATCH] fix silly config option.

CONFIG_CONFIG_TUNER_MULTI_I2C probably isn't what the
author meant to create.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] CFQ io scheduler: scheduler switch oops

If cfq is managing a queue and a new scheduler is later selected, it is
possible for the cfqd unplug_work work to be queued after the kblockd
work struct has been flushed.  The problem is the ordering of
cfq_shutdown_timer_wq() and blk_put_queue() in cfq_put_cfqd().  The
latter may rearm the work, leaving cfq_kick_queue() with dead data.

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] CFQ io scheduler, add ioprio documentation

Add ioprio documentation

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] CFQ io scheduler updates

- Adjust slice values

- Instead of one async queue, one is defined per priority level. This
  prevents kernel threads (such as reiserfs/x and others) that run at
  higher io priority from conflicting with others. Previously, it was a
  coin toss what io prio the async queue got, it was defined by who
  first set up the queue.

- Let a time slice only begin, when the previous slice is completely
  done. Previously we could be somewhat unfair to a new sync slice, if
  the previous slice was async and had several ios queued. This might
  need a little tweaking if throughput suffers a little due to this,
  allowing perhaps an overlap of a single request or so.

- Optimize the calling of kblockd_schedule_work() by doing it only when
  it is strictly necessary (no requests in driver and work left to do).

- Correct sync vs async logic. A 'normal' process can be purely async as
  well, and a flusher can be purely sync as well. Sync or async is now a
  property of the class defined and requests pending. Previously writers
  could be considered sync, when they were really async.

- Get rid of the bit fields in cfqq and crq, use flags instead.

- Various other cleanups and fixes

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Fix cfq_find_next_crq()

In cfq_find_next_crq(), cfq tries to find the next request by choosing
one of two requests before and after the current one.  Currently, when
choosing the next request, if there's no next request, the next
candidate is NULL, resulting in selection of the previous request.  This
results in weird scheduling.  Once we reach the end, we always seek
backward.

The correct behavior is using the first request as the next candidate.
cfq_choose_req() already has logics for handling wrapped requests.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Update cfq io scheduler to time sliced design

This updates the CFQ io scheduler to the new time sliced design (cfq
v3).  It provides full process fairness, while giving excellent
aggregate system throughput even for many competing processes.  It
supports io priorities, either inherited from the cpu nice value or set
directly with the ioprio_get/set syscalls.  The latter closely mimic
set/getpriority.

This import is based on my latest from -mm.

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ARM: 2759/1: Fix IXP4xx debug code (again)

Patch from Deepak Saxena

Accidently swapped the order of movne and orrne. Bad.

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

[PATCH] ARM: 2758/1: Fix comment in file header to read "ARM" instead i386

Patch from Michael Burian

This does not look like an include file for "i386", so use "ARM" instead.

Signed-off-by: Michael Burian <dynmail1@gassner-waagen.at>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: Update mach-types

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

[PATCH] ARM: Fix speeling eroor in io.h

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

[PATCH] ARM: Move memmap freeing into init.c

It doesn't make sense for this to be in mm-armv.c now that 26-bit
ARM support is no longer integrated into arch/arm.

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

[PATCH] ARM: Move PGD kernel page table initialisation

It doesn't make sense to have the PGD kernel pointers initialisation
separate from the PGD user pointers, especially when we clean the
data cache over the whole range.

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

[PATCH] ARM: Add VST idle loop call

This call allows the dynamic tick support to reprogram the timer
immediately before the CPU idles.

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

[PATCH] ARM: Add missed AAEC2000 file

My scripts missed committing this file.

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

[PATCH] Serial: Split 8250 port table

Add separate files for the different 8250 ISA-based serial boards.

Looking across all the various architectures, it seems reasonable that
we can key the availability of the configuration options for these
beasts to the bus-related symbols (iow, CONFIG_ISA).  We also standardise
the base baud/uart clock rate for these boards - I'm sure that isn't
architecture specific, but is solely dependent on the crystal fitted
on the board (which should be the same no matter what type of machine
its fitted into.)

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

[PATCH] Serial: Remove {un,}register_serial from 68328serial.c

68328serial.c does not make use of register_serial/unregister_serial,
which is traditionally used to register 8250-compatible UARTs with
the 8250-compatible serial driver.

Acked-by: David McCullough
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] sound/oss/via82cxxx_audio: Use the DMA_32BIT_MASK constant

Use the DMA_32BIT_MASK constant from dma-mapping.h
when calling pci_set_dma_mask() or pci_set_consistent_dma_mask()
See http://marc.theaimsgroup.com/?t=108001993000001&r=1&w=2 for details

Signed-off-by: Tobias Klauser <tklauser@nuerscht.ch>
Signed-off-by: Domen Puncer <domen@coderock.org>

[PATCH] libata fix read capacity handling for more than 2TB

This is a multi-part message in MIME format.

[PATCH] b44: check link state during open

Check the link state during b44_open.  This closes a 1 HZ window
that existed after b44_open ran but before the b44_timer handler ran,
during which ethtool would report "Link detected: yes" no matter what
the link state actually was.

Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] fix tulip suspend/resume #2

This patch allows the tulip driver to suspend and resume properly.  It was
originally written by Karsten Keil and then modified by Adam Belay.

Signed-off-by: Karsten Keil <kkeil@suse.de>
Signed-off-by: Adam Belay <abelay@novell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[PATCH] uninitialized variable in prism54 isl38xx_trigger_device

drivers/net/wireless/prism54/isl_38xx.c:131: warning: 'current_time.tv_sec' is used uninitialized in this function
drivers/net/wireless/prism54/isl_38xx.c:131: warning: 'current_time.tv_usec' is used uninitialized in this function

Signed-off-by: Olaf Hering <olh@suse.de>

[PATCH] NETDEV: fix receiving multicast frames.

Some USB ethernet drivers did not accept multicast frames appropriately.
IPv6 did not work with those drivers without this patch.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>

[PATCH] NETDEV: Elecom (Laneed) LD-USBL/TX support.

Elecom (Laneed) LD-USBL/TX support.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>

[PATCH] 3c523: needs a license

Module needs a license to prevent kernel tainting.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
diffstat:=
 drivers/net/3c523.c |    1 +
 1 files changed, 1 insertion(+)

[PATCH] gianfar: Add support enhanced TSEC features on the MPC 8548

Jeff,

Just incase this got lost in the recent netdev mailing list transition
here is a nicer version of Andy's patch for gianfar.

- kumar

* TCP/IP/UDP checksumming and verification
* VLAN tag insertion/extraction
* Larger multicast hash-table
* Padding to align IP headers

Also added:
* msg lvl support
* Some whitespace cleanup

Signed-off-by: Andy Fleming <afleming@freescale.com>
Signed-off-by: Kumar Gala <kumar.gala@freescale.com>

[PATCH] smc91x: plug race between TX tasklet and driver reset

The race causes a kernel oops when smc_hardware_send_pkt() tries to
dereference pending_tx_skb which would have been freed from one of the
driver reset paths just after the tx_task tasklet has been scheduled.
This race is possible on SMP but was uncovered by the kernel RT work.

Signed-off-by: Nicolas Pitre <nico@cam.org>

[PATCH] via-rhine trivial whitespace patch

--Boundary-00=_F5lsC5eH1wGW5o9
Content-Type: text/plain;
  charset="koi8-r"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

Hi Jeff,

In some messages in via-rhine.c there is a leading space
for no apparent reason. This patch removes it.
--
vda

--Boundary-00=_F5lsC5eH1wGW5o9
Content-Type: text/x-diff;
  charset="koi8-r";
  name="via-rhine.c.diff"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
filename="via-rhine.c.diff"

[PATCH] e1000: Driver version, white space, comments & other

Driver version, white space, comments & other

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Included proposals to false late collisions due to H/W latencies

Included proposals to false late collisions due to H/W latencies

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Ethtool set speed/duplex validates parameters for consistency

Ethtool set speed/duplex validates parameters for consistency

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Enable ethtool phys_id feature for 82573 controllers

Enable ethtool phys_id feature for 82573 controllers

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Fixed VLAN tag processing error for big-endian architectures

Fixed VLAN tag processing error for big-endian architectures

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Ethtool cleanup patch from Stephen Hemminger

Ethtool cleanup patch from Stephen Hemminger
   * use ADVERTISED_xxx fields when setting advertised fields
   * don't hardcode constant for advertised field

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Fixed the loopback test failure for 82573 based adapters

Fixed the loopback test failure for 82573 based adapters

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Fixed register and loopback test failures with 82573 controllers

Fixed register and loopback test failures with 82573 controllers

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Cleanup debug message printed when Tx Unit hang is detected

Cleanup debug message printed when Tx Unit hang is detected

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Use netdev_priv() to get to netdev->priv

Use netdev_priv() to get to netdev->priv - from shemminger@osdl.org

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Use correct WOL settings for 82544 adapters

Use correct WOL settings for 82544 adapters

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e1000: Synchronize phy access between stats update and MII_IOCTL

Synchronize phy access between stats update and MII_IOCTL

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e100: NAPI performance enhancements

NAPI performance enhancements - Fixed issues with shared interrupts and NAPI resulting in bad performance.

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e100: e100_eeprom_load was called after e100_phy_init

e100_eeprom_load was called after e100_phy_init causing phy_init
not to use values set in EEPROM - from emann@mrv.com

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

[PATCH] e100: fixed e100 MDI/MDI-X issues

Added patch from Eran Mann to fix following e100 MDI/MDI-X issues
 * MDI/MDI-X autodetection should never be enabled for 82551ER/QM chips
 * enabling this feature based on eeprom settings

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>

Merge /spare/repo/netdev-2.6/ branch 'orinoco'

wireless/orinoco: remove changelog, bump version

[PATCH] orinoco: monitor mode support

Patch from Pavel Roskin

[PATCH] orinoco: always use 802.11 header for rx processing

If the frame has ToDS flag set, mark it by setting skb->pkt_type to
PACKET_OTHERHOST, so that applications unaware of promiscous mode won't get
uplink (STA->AP) packets for STA->STA transmissions relayed by the AP.
Thanks to John Denker and David Gibson for finding the problem and the
solution.

Patch from Pavel Roskin

[PATCH] orinoco: scanning support

Patch from Pavel Roskin

[PATCH] orinoco: manual roaming for Symbol and Intersilfirmware

Patch from Pavel Roskin

[PATCH] orinoco: basic ethtool support

I completely reimplemented this based on ethtool_ops, CVS has an
ioctl-based version.

[PATCH] orinoco: wireless API 15 support

(patch from Moustafa Youssef, updated by Jim Carter and Pavel Roskin).

[PATCH] orinoco: include <net/ieee80211.h>

We need constants from this header in the next patches.

[PATCH] bring over ieee80211.h from mainline

the prototypes and inlines aren't actually needed, but let's not diverge
from -mm too far.

[PATCH] cs89x0.c: support for Philips' pnx0105 network adapter

This patch is to provide support for cs89x0-based network device on
Philips' pnx0105 board.

Signed-off-by: dmitry pervushin <dpervushin@ru.mvista.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[PATCH] drivers/net/skfp/: cleanups

This patch contains the following cleanups:
- make needlessly global code static
- remove the completely unused smtparse.c
- remove the following unused global functions:
  - drvfbi.c: init_dma
  - drvfbi.c: dis_dma
  - drvfbi.c: get_rom_byte
  - drvfbi.c: mac_drv_vpd_read
  - drvfbi.c: mac_drv_pci_fix
  - fplustm.c: mac_set_func_addr
  - fplustm.c: mac_del_multicast
  - hwmtm.c: mac_drv_rx_frag
  - pcmplc.c: pcm_set_lct_short
  - smt.c: smt_please_reconnect
  - smt.c: smt_change_t_neg
  - smtdef.c: smt_set_defaults

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[netdrvr] tulip: add pci id

[PATCH] forcedeth: Add support for new device id

This is a multi-part message in MIME format.

[PATCH] forcedeth: Poll for link changes

This is a multi-part message in MIME format.

[PATCH] forcedeth: add two new pci ids

This is a multi-part message in MIME format.

[PATCH] drivers/block/sx8.c: Use the DMA_{64, 32}BIT_MASK constants

Use the DMA_{64,32}BIT_MASK constants from dma-mapping.h when calling
pci_set_dma_mask() or pci_set_consistent_dma_mask()
These patches include dma-mapping.h explicitly because it caused errors
on some architectures otherwise.
See http://marc.theaimsgroup.com/?t=108001993000001&r=1&w=2 for details

Signed-off-by: Tobias Klauser <tklauser@nuerscht.ch>
Signed-off-by: Domen Puncer <domen@coderock.org>

[PATCH] net/lanstreamer: replace schedule_timeout() with ssleep()/msleep_interruptible()

Use ssleep() / msleep_interruptible() [as appropriate]
instead of schedule_timeout() to guarantee the task delays as expected.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Maximilian Attems <janitor@sternwelten.at>
Signed-off-by: Domen Puncer <domen@coderock.org>

[PATCH] net/pcnet32: replace schedule_timeout() with msleep_interruptible()

Use msleep_interruptible() instead of schedule_timeout() to
guarantee the task delays as expected.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Maximilian Attems <janitor@sternwelten.at>
Signed-off-by: Domen Puncer <domen@coderock.org>

[PATCH] net/farsync: add set_current_state() before schedule_timeout()

Insert set_current_state() before schedule_timeout() so the
function delays as expected. Without the addition, schedule_timeout()
will return immediately.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Maximilian Attems <janitor@sternwelten.at>
Signed-off-by: Domen Puncer <domen@coderock.org>

[PATCH] net/sb1000: replace nicedelay() with ssleep()

Use ssleep() instead of nicedelay()
to guarantee the task delays as expected. Remove the prototype and
definition of nicedelay(). This is a very weird function, because it is
called to sleep in terms of usecs, but always sleeps for 1 second,
completely ignoring the parameter. I have gone ahead and followed suit,
just sleeping for a second in all cases, but maybe someone with the
hardware could tell me if perhaps the paramter *should* matter. Additionally,
nicedelay() is called in TASK_INTERRUPTIBLE state, but doesn't deal with signals
in case these longer delays do not complete, so I believe ssleep() is more
appropriate.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Domen Puncer <domen@coderock.org>

[PATCH] wireless: char* -> char[] conversion in airo.c

This conversion makes code from line 7101 right:
if (copy_to_user(com.data, swversion, sizeof(swversion)))

size output (before, after):
  55416    2228     160   57804    e1cc drivers/net/wireless/airo.o
  55412    2228     160   57800    e1c8 drivers/net/wireless/airo.o

more outputs from Alexey Dobriyan:
2.95.3:
           text    data     bss     dec     hex filename
before    51118    2156     160   53434    d0ba drivers/net/wireless/airo.o
after     51118    2156     160   53434    d0ba drivers/net/wireless/airo.o

3.3.5-20050130:
before    46999    2156     160   49315    c0a3 drivers/net/wireless/airo.o
after     46994    2156     160   49310    c09e drivers/net/wireless/airo.o

4.1.0-20050522:
before    45555    2220     160   47935    bb3f drivers/net/wireless/airo.o
after     45550    2220     160   47930    bb3a drivers/net/wireless/airo.o

Signed-off-by: Domen Puncer <domen@coderock.org>

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

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

[IPVS]: Fix for overflows

From: <pageexec@freemail.hu>

$subject was fixed in 2.4 already, 2.6 needs it as well.

The impact of the bugs is a kernel stack overflow and privilege escalation
from CAP_NET_ADMIN via the IP_VS_SO_SET_STARTDAEMON/IP_VS_SO_GET_DAEMON
ioctls.  People running with 'root=all caps' (i.e., most users) are not
really affected (there's nothing to escalate), but SELinux and similar
users should take it seriously if they grant CAP_NET_ADMIN to other users.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] DM9000 network driver bugfix

This patch fixes two bugs in the dm9000 network driver:

- Don't read one byte too much in 8bit mode.
- release correct resource

Signed-off-by: Jochen Karrer <j.karrer@lightmaze.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Jeff Garzik <jgarzik@pobox.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[PATCH] Documentation/networking/dmfe.txt: Make documentation nicer

Patch indents dmfe.txt to look like other docs.  It adds a tip about CNET
cards using Davicom chipsets.  Also it removes parts where it refers to how
to build driver out-of-kernel which seems to be cruft from times where the
driver was out of the kernel.

Signed-off-by: Ismail Donmez <ismail@kde.org.tr>
Signed-off-by: Domen Puncer <domen@coderock.org>
Cc: Jeff Garzik <jgarzik@pobox.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[PATCH] fealnx.c calls dev_kfree_skb from atomic context

Signed-off-by: Andrew Morton <akpm@osdl.org>

[PATCH] fix int vs. pm_message_t confusion in airo

Fix int vs. pm_message_t confusion in airo. Should change no code.

Signed-off-by: Pavel Machek <pavel@suse.cz>
Cc: Jeff Garzik <jgarzik@pobox.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[PATCH] defxx: Use irqreturn_t for the interrupt handler

This is a fix for the interrupt handler in the defxx driver to use
irqreturn_t.  Beside the obvious fix of returning a proper status at all,
it actually checks board registers as appropriate for determining if an
interrupt has been recorded in the bus-specific interface logic.

The patch also includes an obvious one-line fix for SET_NETDEV_DEV needed
for the EISA variation, for which I've decided there is no point in sending
separately.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: Jeff Garzik <jgarzik@pobox.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[PATCH] dmfe warning fix

drivers/net/tulip/dmfe.c: In function `dmfe_parse_srom':
drivers/net/tulip/dmfe.c:1805: warning: passing arg 1 of `__le16_to_cpup' from incompatible pointer type
drivers/net/tulip/dmfe.c:1817: warning: passing arg 1 of `__le32_to_cpup' from incompatible pointer type
drivers/net/tulip/dmfe.c:1817: warning: passing arg 1 of `__le32_to_cpup' from incompatible pointer type

This is basically a guess:

Cc: Jeff Garzik <jgarzik@pobox.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[PATCH] document that 8139TOO supports 8129/8130

The 8129/8130 support is a sub-option that is not visible if the user
hasn't enabled the 8139 support.

Let's make it a bit easier for users to find the driver for their nic.

Signed-off-by: Adrian Bunk <bunk@stusta.de>

[NETLINK]: Fix two socket hashing bugs.

1) netlink_release() should only decrement the hash entry
   count if the socket was actually hashed.

   This was causing hash->entries to underflow, which
   resulting in all kinds of troubles.

   On 64-bit systems, this would cause the following
   conditional to erroneously trigger:

err = -ENOMEM;
if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
goto err;

2) netlink_autobind() needs to propagate the error return from
   netlink_insert().  Otherwise, callers will not see the error
   as they should and thus try to operate on a socket with a zero pid,
   which is very bad.

   However, it should not propagate -EBUSY.  If two threads race
   to autobind the socket, that is fine.  This is consistent with the
   autobind behavior in other protocols.

   So bug #1 above, combined with this one, resulted in hangs
   on netlink_sendmsg() calls to the rtnetlink socket.  We'd try
   to do the user sendmsg() with the socket's pid set to zero,
   later we do a socket lookup using that pid (via the value we
   stashed away in NETLINK_CB(skb).pid), but that won't give us the
   user socket, it will give us the rtnetlink socket.  So when we
   try to wake up the receive queue, we dive back into rtnetlink_rcv()
   which tries to recursively take the rtnetlink semaphore.

Thanks to Jakub Jelink for providing backtraces.  Also, thanks to
Herbert Xu for supplying debugging patches to help track this down,
and also finding a mistake in an earlier version of this fix.

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

[PATCH] net/slip: replace schedule_timeout() with msleep_interruptible()

Use msleep_interruptible() instead of schedule_timeout() to guarantee
the task delays as expected.

[PATCH] drivers/net/ewrk3.c: remove dead code

This patch removes some obviously dead code found by the Coverity
checker.

Signed-off-by: Adrian Bunk <bunk@stusta.de>

[PATCH] drivers/net/skfp/: fix LITTLE_ENDIAN

This patch fixes the LITTLE_ENDIAN #define and a function prototype.

Signed-off-by: Adrian Bunk <bunk@stusta.de>

[PATCH] drivers/net/tokenring/: cleanups

This patch contains the follwing cleanups:
- make needlessly global code static
- remove obsolete Emacs settings

Signed-off-by: Adrian Bunk <bunk@stusta.de>

[ATALK]: Include asm/byteorder.h in linux/atalk.h

We're using __be16 in userland visible types, so we
have to include asm/byteorder.h so that works.

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

[PKTGEN]: Fix random packet sizes causing panic

Signed-off-by: Robert Olsson <robert.olsson@its.uu.se>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] pcnet_cs.c: IRQ handler optimization

During some performance diagnostics I stumbled on this slightly wasteful
code in pcnet_cs.c which I made the patch included at the bottom for (two
minor comment fixes included).

Improvement:
instead of *always* calculating
lea    0x2c0(%edx),%ebx
and then additionally doing the
mov    %edx,0xc0(%ebx)
addition *if we need it*,

we now do the *whole* calculation of
mov    %edx,0x380(%ebx)
*only* if we need it.
This even manages to save us a whole 16-byte alignment buffer loss
in this compilation case.

Result: slightly improves IRQ handler performance in both shared and
non-shared IRQ case, which should make my rusty P3/700 a slight bit happier.

Thank you for your support,

Andreas Mohr

old asm result (using gcc 3.3.5):

000015a0 <ei_irq_wrapper>:
    15a0:       55                      push   %ebp
    15a1:       89 e5                   mov    %esp,%ebp
    15a3:       53                      push   %ebx
    15a4:       8d 9a c0 02 00 00       lea    0x2c0(%edx),%ebx
    15aa:       e8 fc ff ff ff          call   15ab <ei_irq_wrapper+0xb>
    15af:       83 f8 01                cmp    $0x1,%eax
    15b2:       74 03                   je     15b7 <ei_irq_wrapper+0x17>
    15b4:       5b                      pop    %ebx
    15b5:       5d                      pop    %ebp
    15b6:       c3                      ret
    15b7:       31 d2                   xor    %edx,%edx
    15b9:       89 93 c0 00 00 00       mov    %edx,0xc0(%ebx)
    15bf:       eb f3                   jmp    15b4 <ei_irq_wrapper+0x14>
    15c1:       eb 0d                   jmp    15d0 <ei_watchdog>
    15c3:       90                      nop
    15c4:       90                      nop
    15c5:       90                      nop
    15c6:       90                      nop
    15c7:       90                      nop
    15c8:       90                      nop
    15c9:       90                      nop
    15ca:       90                      nop
    15cb:       90                      nop
    15cc:       90                      nop
    15cd:       90                      nop
    15ce:       90                      nop
    15cf:       90                      nop

000015d0 <ei_watchdog>:

new asm result:

000015a0 <ei_irq_wrapper>:
    15a0:       55                      push   %ebp
    15a1:       89 e5                   mov    %esp,%ebp
    15a3:       53                      push   %ebx
    15a4:       89 d3                   mov    %edx,%ebx
    15a6:       e8 fc ff ff ff          call   15a7 <ei_irq_wrapper+0x7>
    15ab:       83 f8 01                cmp    $0x1,%eax
    15ae:       74 03                   je     15b3 <ei_irq_wrapper+0x13>
    15b0:       5b                      pop    %ebx
    15b1:       5d                      pop    %ebp
    15b2:       c3                      ret
    15b3:       31 d2                   xor    %edx,%edx
    15b5:       89 93 80 03 00 00       mov    %edx,0x380(%ebx)
    15bb:       eb f3                   jmp    15b0 <ei_irq_wrapper+0x10>
    15bd:       8d 76 00                lea    0x0(%esi),%esi

000015c0 <ei_watchdog>:

Signed-off-by: Andrew Morton <akpm@osdl.org>

drivers/net/: Use the DMA_{64,32}BIT_MASK constants

Use the DMA_{64,32}BIT_MASK constants from dma-mapping.h when calling
pci_set_dma_mask() or pci_set_consistent_dma_mask()

This patch includes dma-mapping.h explicitly because it caused errors
on some architectures otherwise.

See http://marc.theaimsgroup.com/?t=108001993000001&r=1&w=2 for details

Signed-off-by: Tobias Klauser <tklauser@nuerscht.ch>
Signed-off-by: Domen Puncer <domen@coderock.org>

[TCP]: Let TCP_CONG_ADVANCED default to n

It doesn't seem to make much sense to let an "If unsure, say N." option
default to y.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Fix thinko in TCP_CONG_BIC default.

Since it is tristate when we offer it as a choice, we should
definte it also as tristate when forcing it as the default.
Otherwise kconfig warns.

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

[PATCH] arlan: module parameter fixes

Make sure the code compiles with and without ARLAN_ENTRY_EXIT_DEBUGGING.
Only provide parameter descriptions when parameters are defined.
Remove "arlan_"-prefix to shape up built-in parameter names:

arlan.arlan_debug -> arlan.debug
arlan.arlan_EEPROM_bad -> arlan.EEPROM_bad
arlan.arlan_entry_and_exit_debug -> arlan.entry_and_exit_debug
arlan.arlan_entry_debug -> arlan.entry_debug
arlan.arlan_exit_debug -> arlan.exit_debug

Signed-off-by: Magnus Damm <damm@opensource.se>

[PATCH] net/sis900: Use the DMA_32BIT_MASK constant

Use the DMA_32BIT_MASK constant from dma-mapping.h when calling
pci_set_dma_mask() or pci_set_consistent_dma_mask() instead of custom
macros.
This patch includes dma-mapping.h explicitly because it caused errors
on some architectures otherwise.
See http://marc.theaimsgroup.com/?t=108001993000001&r=1&w=2 for details

Signed-off-by: Tobias Klauser <tklauser@nuerscht.ch>

Merge /spare/repo/netdev-2.6/ branch 'register-netdev'

Merge /spare/repo/linux-2.6/

8139cp: safer spin loop for get_statistics

The spin loop in 8139cp is limited to 100 iterations when pulling hardware
stats. There is no allowance for processor speed so on a fast machine, the
stats may not be available that fast. Also, if the board doesn't return
soon enough make sure turn the address back off to prevent later updates
when memory has gone away.