kernel/kernel-generic.git
17 years agoUSB: ratelimit debounce error messages
Oliver Neukum [Tue, 27 Feb 2007 09:25:00 +0000 (10:25 +0100)]
USB: ratelimit debounce error messages

flaky hardware can cause a lot of debounce failed messages. To limit
the performance impact, a ratelimit should be used.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoUSB: pxa2xx_udc: fix hardcoded irq number
Milan Svoboda [Tue, 27 Feb 2007 09:20:09 +0000 (09:20 +0000)]
USB: pxa2xx_udc: fix hardcoded irq number

This patch changes last use of hardcoded number of irq to
use platfrom_get_irq.

Signed-off-by: Milan Svoboda <msvoboda@ra.rockwell.com>
Acked-by: David Brownell <david-b@pacbell.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoUHCI: fix port resume problem
Alan Stern [Mon, 26 Feb 2007 22:16:06 +0000 (17:16 -0500)]
UHCI: fix port resume problem

This patch (as863) fixes a problem encountered sometimes when resuming
a port on a UHCI controller.  The hardware may turn off the
Resume-Detect bit before turning off the Suspend bit, leading usbcore
to think that the port is still suspended and the resume has failed.
The patch makes uhci_finish_suspend() wait until both bits are safely
off.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoUSB: set the correct interval for interrupt URBs
Alan Stern [Mon, 26 Feb 2007 19:56:14 +0000 (14:56 -0500)]
USB: set the correct interval for interrupt URBs

This patch (as862) fixes a couple of bugs in the way usbcore handles
intervals for interrupt URBs.  usb_interrupt_msg (and usb_bulk_msg for
backward compatibility) don't set the interval correctly for
high-speed devices.  proc_do_submiturb() doesn't set it correctly when
a bulk URB is submitted to an interrupt endpoint.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoUSB: goku_udc: Remove crude cache coherency code
Ralf Baechle [Mon, 26 Feb 2007 18:10:15 +0000 (18:10 +0000)]
USB: goku_udc: Remove crude cache coherency code

This is deep architecture specific magic and does should not to exist
in a driver.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Cc: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoUSB storage: Nokia 6288 unusual_devs entry
Andrew Nayenko [Sun, 25 Feb 2007 08:40:20 +0000 (00:40 -0800)]
USB storage: Nokia 6288 unusual_devs entry

This patch adds an usual_devs entry for the Nokia 6288. Originally from
Andrew with a re-diff by Phil.

From: Andrew Nayenko <relan@bk.ru>
Signed-off-by: Phil Dibowitz <phil@ipom.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoUSB storage: Removed duplicate supertop unusual_dev entry
Phil Dibowitz [Sun, 25 Feb 2007 03:23:41 +0000 (19:23 -0800)]
USB storage: Removed duplicate supertop unusual_dev entry

This patch removes the duplicate supertop entries that made it into the
.21 rc kernels.

Signed-off-by: Phil Dibowitz <phil@ipom.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoUSB: at91_udc, fix more modpost bogosity (rename driver struct)
David Brownell [Sat, 24 Feb 2007 21:54:56 +0000 (13:54 -0800)]
USB: at91_udc, fix more modpost bogosity (rename driver struct)

Rename the driver struct used with at91_udc to prevent yet another
bogus warning from "modpost".

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
17 years agoMerge branch 'merge' of master.kernel.org:/pub/scm/linux/kernel/git/paulus/powerpc
Linus Torvalds [Fri, 9 Mar 2007 22:22:35 +0000 (14:22 -0800)]
Merge branch 'merge' of /linux/kernel/git/paulus/powerpc

* 'merge' of master.kernel.org:/pub/scm/linux/kernel/git/paulus/powerpc:
  [POWERPC] 85xx: Enable CONFIG_SERIAL_8250_SHARE_IRQ
  [POWERPC] Select u-image as default image for Linkstation
  [POWERPC] 83xx: Minor fixes for 834x_mds USB setup code
  [POWERPC] Fix warning in powermac pci.c
  [POWERPC] Fix warning in powermac feature.c
  [POWERPC] Fix warning in prom_parse.c of_irq_map_oldworld()
  [POWERPC] Celleb: bug fix caused by not casting pointer types
  [POWERPC] Add missing newline in xmon help output
  [POWERPC] No DEEPNAP on 970MP 1.0

17 years ago[MIPS] Sibyte: Fix ZBbus profiler
Ralf Baechle [Fri, 9 Mar 2007 15:59:56 +0000 (15:59 +0000)]
[MIPS] Sibyte: Fix ZBbus profiler

 o Fix build error.
 o Handle error returns.
 o Deal with signals received while sleeping.
 o Don't allow to be selected when we're not building the directory with
   the driver anyway.
 o Coding style cleanups.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Sibyte: Do not allow enabling LDT support if PCI is disabled.
Ralf Baechle [Fri, 9 Mar 2007 12:17:32 +0000 (12:17 +0000)]
[MIPS] Sibyte: Do not allow enabling LDT support if PCI is disabled.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Fri, 9 Mar 2007 18:38:09 +0000 (10:38 -0800)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] DEC: Remove call to register_prom_console.

17 years agoMerge git://git.infradead.org/mtd-2.6
Linus Torvalds [Fri, 9 Mar 2007 18:34:55 +0000 (10:34 -0800)]
Merge git://git.infradead.org/mtd-2.6

* git://git.infradead.org/mtd-2.6:
  [JFFS2] print a message when marking bad block
  [JFFS2] Check for all-zero node headers
  [MTD] [OneNAND] Classify the page data and oob buffer
  [MTD] [OneNAND] Exit the loop when transferring/filling of the oob is finished
  [MTD] [OneNAND] add Nokia Copyright and a credit
  [MTD] [OneNAND] Fix typo & wrong comments
  [MTD] [OneNAND] Use oob buffer instead of main one in oob functions
  [MTD] Correct partition failed erase address
  [JFFS2] Use yield() between GC passes in background thread.
  [MTD] [NAND] Correct misspelled preprocessor variable.
  [MTD] [MAPS] dilnetpc: Fix printk warning
  [MTD] [NOR] Fix oops in cfi_amdstd_sync
  [MTD] ESB2 check for closed ROM window
  [JFFS2] Fix writebuffer recovery in the first page of a block
  [MTD] [NAND] make oobavail public

17 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Fri, 9 Mar 2007 18:12:46 +0000 (10:12 -0800)]
Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/netdev-2.6

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6:
  ibmtr: Drain rich supply of modpost warnings.
  3c59x: Fix several modpost warnings
  z85230: Fix FIFO handling
  r8169: fix a race between PCI probe and dev_open
  r8169: revert bogus BMCR reset
  sky2: turn off Rx checksum on bad hardware
  mv643xx: Clear pending interrupts before calling request_irq
  myri10ge: add a wc_enabled flag to myri10ge_priv
  myri10ge: prevent 4k rdma on SGI TIOCE chipset
  myri10ge: use pci_map_page to prepare the dmatest buffer
  myri10ge: fix error checking and return value in myri10ge_allocate_rings
  netxen: fix warnings
  via-rhine: set avoid_D3 for broken BIOSes
  [PATCH] bcm43xx: Fix assertion failures in interrupt handler
  [PATCH] bcm43xx: Fix bug in frequency to channel conversion
  [PATCH] bcm43xx: Fix errors in specs to code translation in B6PHY init

17 years agoMerge branch 'upstream-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/linvil...
Jeff Garzik [Fri, 9 Mar 2007 16:58:11 +0000 (11:58 -0500)]
Merge branch 'upstream-fixes' of git://git./linux/kernel/git/linville/wireless-2.6 into upstream-fixes

17 years agoibmtr: Drain rich supply of modpost warnings.
Ralf Baechle [Thu, 8 Mar 2007 12:53:07 +0000 (12:53 +0000)]
ibmtr: Drain rich supply of modpost warnings.

Building ibmtr as a module produces a spectacular pile of modpost warnings:

WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x450) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x454) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x458) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x45c) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x4e8) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x4ec) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x4f0) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x4f4) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:find_turbo_adapters from .text between 'ibmtr_probe_card' (at offset 0x500) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:find_turbo_adapters from .text between 'ibmtr_probe_card' (at offset 0x504) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:find_turbo_adapters from .text between 'ibmtr_probe_card' (at offset 0x508) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:find_turbo_adapters from .text between 'ibmtr_probe_card' (at offset 0x50c) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x520) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x524) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x528) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x534) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data: from .text between 'ibmtr_probe_card' (at offset 0x540) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data: from .text between 'ibmtr_probe_card' (at offset 0x544) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data: from .text between 'ibmtr_probe_card' (at offset 0x548) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data: from .text between 'ibmtr_probe_card' (at offset 0x54c) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x558) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x55c) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x560) and 'ibmtr_reset_timer'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x564) and 'ibmtr_reset_timer'

Fix by making ibmtr_probe an __init function.

While at it, move move ibmtr_probe_card below ibmtr_probe so the protoype
for ibmtr_probe can be deleted.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
 drivers/net/tokenring/ibmtr.c |   25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago3c59x: Fix several modpost warnings
Ralf Baechle [Thu, 8 Mar 2007 15:33:16 +0000 (15:33 +0000)]
3c59x: Fix several modpost warnings

WARNING: drivers/net/3c59x.o - Section mismatch: reference to .init.text: from .text between 'vortex_eisa_probe' (at offset 0x4580) and 'vortex_eisa_remove'
WARNING: drivers/net/3c59x.o - Section mismatch: reference to .init.text: from .text between 'vortex_eisa_probe' (at offset 0x4584) and 'vortex_eisa_remove'
WARNING: drivers/net/3c59x.o - Section mismatch: reference to .init.text: from .text between 'vortex_eisa_probe' (at offset 0x4588) and 'vortex_eisa_remove'
WARNING: drivers/net/3c59x.o - Section mismatch: reference to .init.text: from .text between 'vortex_eisa_probe' (at offset 0x458c) and 'vortex_eisa_remove'

Fixed by:

 o move definition of vortex_eisa_driver below the functions it references.
 o remove now unnecessary prototypes for vortex_eisa_probe and
   vortex_eisa_remove.
 o Make vortex_eisa_probe an __init function.
 o Make vortex_eisa_remove a __devexit function.
 o Wrap vortex_eisa_driver reference to vortex_eisa_remove with
   __devexit_p().

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
 drivers/net/3c59x.c |   28 +++++++++++++---------------
 1 file changed, 13 insertions(+), 15 deletions(-)
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoz85230: Fix FIFO handling
Alan Cox [Wed, 7 Mar 2007 16:37:38 +0000 (16:37 +0000)]
z85230: Fix FIFO handling

We must exit immediately on a FIFO fill not take the end of packet path
otherwise each underrun in PIO transmit mode causes a runt packet and the
data is lost.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agor8169: fix a race between PCI probe and dev_open
Francois Romieu [Thu, 8 Mar 2007 23:00:05 +0000 (00:00 +0100)]
r8169: fix a race between PCI probe and dev_open

Initialize the timer with the rest of the private-struct.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agor8169: revert bogus BMCR reset
Francois Romieu [Thu, 8 Mar 2007 22:59:54 +0000 (23:59 +0100)]
r8169: revert bogus BMCR reset

Added during bf793295e1090af84972750898bf8470df5e5419

The current code requests a reset but prohibits autoneg, 1000 Mb/s,
100 Mb/s and full duplex. The 8168 does not like it at all.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agosky2: turn off Rx checksum on bad hardware
Stephen Hemminger [Thu, 8 Mar 2007 20:42:30 +0000 (12:42 -0800)]
sky2: turn off Rx checksum on bad hardware

On Yukon FE, occasional hardware receive checksum errors are seen.
An early indication of the problem is single bit differences in the two
checksum engines.  Use this as a detection mechanism to turn off Rx
checksumming.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agomv643xx: Clear pending interrupts before calling request_irq
Dale Farnsworth [Thu, 8 Mar 2007 05:21:23 +0000 (22:21 -0700)]
mv643xx: Clear pending interrupts before calling request_irq

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
Acked-by: Giridhar Pemmasani <pgiri@yahoo.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agomyri10ge: add a wc_enabled flag to myri10ge_priv
Brice Goglin [Wed, 7 Mar 2007 19:02:32 +0000 (20:02 +0100)]
myri10ge: add a wc_enabled flag to myri10ge_priv

Add a wc_enabled flag in the myri10ge_priv instead of relying
on mtrr >= 0.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agomyri10ge: prevent 4k rdma on SGI TIOCE chipset
Brice Goglin [Wed, 7 Mar 2007 19:01:53 +0000 (20:01 +0100)]
myri10ge: prevent 4k rdma on SGI TIOCE chipset

Do not use 4k rdma request on SGI TIOCE chipset since this
bridge does not support it.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agomyri10ge: use pci_map_page to prepare the dmatest buffer
Brice Goglin [Wed, 7 Mar 2007 19:00:45 +0000 (20:00 +0100)]
myri10ge: use pci_map_page to prepare the dmatest buffer

Allocate a specific page and use pci_map_page for dma test instead
of relying on another existing buffer.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agomyri10ge: fix error checking and return value in myri10ge_allocate_rings
Brice Goglin [Wed, 7 Mar 2007 18:59:52 +0000 (19:59 +0100)]
myri10ge: fix error checking and return value in myri10ge_allocate_rings

Fix a missing error check in myri10ge_allocate_rings() and set status
to -ENOMEM before all actual allocations so that the error path returns
what it should.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agonetxen: fix warnings
Ralf Baechle [Tue, 6 Mar 2007 20:35:37 +0000 (20:35 +0000)]
netxen: fix warnings

  CC [M]  drivers/net/netxen/netxen_nic_hw.o
drivers/net/netxen/netxen_nic_hw.c: In function 'netxen_nic_hw_resources':
drivers/net/netxen/netxen_nic_hw.c:231: warning: format '%llx' expects type 'long long unsigned int', but argument 2 has type 'dma_addr_t'
drivers/net/netxen/netxen_nic_hw.c:250: warning: format '%llx' expects type 'long long unsigned int', but argument 2 has type 'dma_addr_t'

u64 is unsigned long so the cast to u64 will result in a warning on the
printf arguments for 64-bit builds.  So cast to unsigned long long instead.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agovia-rhine: set avoid_D3 for broken BIOSes
Roger Luethi [Tue, 6 Mar 2007 18:57:37 +0000 (19:57 +0100)]
via-rhine: set avoid_D3 for broken BIOSes

This patch sets avoid_D3 for BIOSes known to be broken. Said BIOSes fail
at PXE boot if the chip is in power state D3.

Signed-off-by: Roger Luethi <rl@hellgate.ch>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agodevres: release resources on device_del()
Tejun Heo [Fri, 9 Mar 2007 10:34:19 +0000 (19:34 +0900)]
devres: release resources on device_del()

Some platform devices are driven without driver attached, so managed
resources can be acquired without driver attached.  Make sure such
resources are released by calling devres_release_all() in
device_del().

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agolibata: fix ata_host_release() free order
Tejun Heo [Fri, 9 Mar 2007 10:36:12 +0000 (19:36 +0900)]
libata: fix ata_host_release() free order

host->ops->host_stop() might access ports.  Free ports after
host_stop.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agosata_nv: revert use of notifiers for now
Robert Hancock [Fri, 9 Mar 2007 00:02:18 +0000 (18:02 -0600)]
sata_nv: revert use of notifiers for now

Commit 721449bf0d51213fe3abf0ac3e3561ef9ea7827a added support for using the
ADMA notifier bits to determine which commands to check for completion.
However there have been reports that this causes command timeouts in certain
cases. This is still being investigated. In addition, apparently the notifiers
won't work if ADMA is disabled on the other port as a result of an ATAPI device
being connected, and we don't handle this case properly.

For now, just restore the previous behavior of checking all active commands
to see if they are complete, without relying on the notifiers.

Signed-off-by: Robert Hancock <hancockr@shaw.ca>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoFix simplex adapters with libata
Petr Vandrovec [Thu, 8 Mar 2007 09:12:12 +0000 (10:12 +0100)]
Fix simplex adapters with libata

Recently I got my hands on nVidia's MCP61 PM-AM board, and
it contains IDE chip configured by BIOS with only primary
channel enabled.  This confuses code which probes for
device DMA capabilities - it gets 0x60 (happy duplex
device) from primary channel BMDMA, but 0xFF (nobody here)
from secondary channel BMDMA.  Due to this code then believes
that chip is simplex.  I do not address this problem in
my patch, as I'm not sure how to handle this.  Probably
ata_pci_init_one should have bitmap of enabled/possible
interfaces instead of their count, but it looks like
quite intrusive change, and maybe we do not care - for device
with only one channel simplex and regular DMA engines are
same.

But making device simplex pointed out that support for
DMA on simplex devices is currently broken - ata_dev_xfermask
tests whether device is simplex and if it is whether DMA
engine was assigned to this port.  If not then it strips
out DMA bits from device.  Problem is that code which assigns
DMA engine to port in ata_set_mode first detect device
mode and assigns DMA engine to channel only if some DMA
capable device was found.

And as xfermask stripped out DMA bits, host->simplex_claimed
is always NULL with current implementation.

By allowing DMA either if simplex_claimed is NULL or if it
points to current port DMA can be finally used - it gets
assigned to first port which contains any DMA capable
device.

Before:
pata_amd 0000:00:06.0: version 0.2.8
PCI: Setting latency timer of device 0000:00:06.0 to 64
ata5: PATA max UDMA/133 cmd 0x000101f0 ctl 0x000103f6 bmdma 0x0001f000 irq 14
ata6: PATA max UDMA/133 cmd 0x00010170 ctl 0x00010376 bmdma 0x0001f008 irq 15
scsi4 : pata_amd
ata5.00: ATAPI, max UDMA/66
ata5.00: simplex DMA is claimed by other device, disabling DMA
ata5.00: configured for PIO4
scsi5 : pata_amd
ata6: port disabled. ignoring.
ata6: reset failed, giving up
scsi 4:0:0:0: CD-ROM            ATAPI    DVD W  DH16W1P   LG12 PQ: 0 ANSI: 5

After:
pata_amd 0000:00:06.0: version 0.2.8
PCI: Setting latency timer of device 0000:00:06.0 to 64
ata5: PATA max UDMA/133 cmd 0x000101f0 ctl 0x000103f6 bmdma 0x0001f000 irq 14
ata6: PATA max UDMA/133 cmd 0x00010170 ctl 0x00010376 bmdma 0x0001f008 irq 15
scsi4 : pata_amd
ata5.00: ATAPI, max UDMA/66
ata5.00: configured for UDMA/33
scsi5 : pata_amd
ata6: port disabled. ignoring.
ata6: reset failed, giving up
scsi 4:0:0:0: CD-ROM            ATAPI    DVD W  DH16W1P   LG12 PQ: 0 ANSI: 5

Signed-off-by: Petr Vandrovec <petr@vandrovec.name>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoata_piix: Remove ugly layering violation
Alan Cox [Wed, 7 Mar 2007 16:13:07 +0000 (16:13 +0000)]
ata_piix: Remove ugly layering violation

A while ago I modified the libata code so that drivers can return -ENOENT
for unknown ports not fiddle with the EH flags and print stuff directly.
Somewhere along the line ata_piix didn't get fully converted.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[JFFS2] print a message when marking bad block
Artem Bityutskiy [Thu, 8 Mar 2007 08:35:10 +0000 (10:35 +0200)]
[JFFS2] print a message when marking bad block

New bad eraseblock is an event which is important enough to be printed
about.

Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MIPS] DEC: Remove call to register_prom_console.
Ralf Baechle [Thu, 8 Mar 2007 17:36:13 +0000 (17:36 +0000)]
[MIPS] DEC: Remove call to register_prom_console.

Register_prom_console was removed when mips was converted to early printk.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[JFFS2] Check for all-zero node headers
David Woodhouse [Fri, 9 Mar 2007 11:44:00 +0000 (11:44 +0000)]
[JFFS2] Check for all-zero node headers

Due to a poor choice of CRC32 seed, a node header which is all zeroes
would pass the CRC32 check. Explicitly check for this case, and treat it
as we do a CRC failure.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [OneNAND] Classify the page data and oob buffer
Kyungmin Park [Fri, 9 Mar 2007 01:08:11 +0000 (10:08 +0900)]
[MTD] [OneNAND] Classify the page data and oob buffer

Classify the page data and oob buffer
and it prevents the memory fragementation (writesize + oobsize)

Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [OneNAND] Exit the loop when transferring/filling of the oob is finished
Kyungmin Park [Fri, 9 Mar 2007 00:41:07 +0000 (09:41 +0900)]
[MTD] [OneNAND] Exit the loop when transferring/filling of the oob is finished

When transferring/filling of the oob is finished in OOB_AUTO, we exit the loop

Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [OneNAND] add Nokia Copyright and a credit
Adrian Hunter [Thu, 15 Feb 2007 00:47:29 +0000 (09:47 +0900)]
[MTD] [OneNAND] add Nokia Copyright and a credit

add Nokia Copyright and a credit

Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [OneNAND] Fix typo & wrong comments
Kyungmin Park [Thu, 15 Feb 2007 00:36:39 +0000 (09:36 +0900)]
[MTD] [OneNAND] Fix typo & wrong comments

Fix typo & wrong comments

Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [OneNAND] Use oob buffer instead of main one in oob functions
Kyungmin Park [Mon, 12 Feb 2007 01:34:39 +0000 (10:34 +0900)]
[MTD] [OneNAND] Use oob buffer instead of main one in oob functions

In oob functions, it is used main buffer instead of oob one. So fix it.

Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[PATCH] build fix for i386 earlyquirk.c
Dave Jones [Fri, 9 Mar 2007 04:47:35 +0000 (23:47 -0500)]
[PATCH] build fix for i386 earlyquirk.c

missing close bracket.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] update 'getting sparse' info.
Dave Jones [Fri, 9 Mar 2007 00:45:26 +0000 (19:45 -0500)]
[PATCH] update 'getting sparse' info.

 - point to the sparse webpage
 - use git:// instead of rsync://

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] ACPI: repair nvidia early quirk breakage on x86_64
Len Brown [Thu, 8 Mar 2007 23:28:32 +0000 (18:28 -0500)]
[PATCH] ACPI: repair nvidia early quirk breakage on x86_64

x86_64 nvidia_bugs() broke when we bailed out on not finding the HPET.
However, the quirk works by checking for _not_ finding the HPET...

Delete the nvidia_hpet_detected flag and simply test for
not finding the HPET, which is simple to do now that
acpi_table_parse returns 1 on failure.

Signed-off-by: Len Brown <len.brown@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMerge branch 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband
Linus Torvalds [Fri, 9 Mar 2007 00:04:11 +0000 (16:04 -0800)]
Merge branch 'for-linus' of /linux/kernel/git/roland/infiniband

* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband:
  IPoIB: Turn on interface's carrier after broadcast group is joined
  RDMA/ucma: Avoid sending reject if backlog is full
  RDMA/cxgb3: Fix MR permission problems
  RDMA/cxgb3: Don't reuse skbs that are non-linear or cloned
  RDMA/cxgb3: Squelch logging AE errors
  RDMA/cxgb3: Stop EP timer when MPA exchange is aborted by peer
  RDMA/cxgb3: Move QP to error on destroy if the state is IDLE
  RDMA/cxgb3: Fixes for "normal close" failures
  RDMA/cxgb3: Fix build on sparc64
  RDMA/cma: Initialize rdma_bind_list in cma_alloc_any_port()
  RDMA/cxgb3: Don't use mm after it's freed in iwch_mmap()
  RDMA/cxgb3: Start ep timer on a MPA reject
  IB/mthca: Fix error path in mthca_alloc_memfree()
  IB/ehca: Fix sync between completion handler and destroy cq
  IPoIB: Only handle async events for one port

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Thu, 8 Mar 2007 23:09:57 +0000 (15:09 -0800)]
Merge /pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [SCTP] ipv6: inconsistent lock state ipv6_add_addr/sctp_v6_copy_addrlist

17 years ago[SPARC64]: Fix PARPORT build (again).
David S. Miller [Thu, 8 Mar 2007 22:48:36 +0000 (14:48 -0800)]
[SPARC64]: Fix PARPORT build (again).

Need to provide {claim,release}_dma_lock() for this guy too.

Signed-off-by: David S. Miller <davem@davemloft.net>
17 years agoIPoIB: Turn on interface's carrier after broadcast group is joined
Shirley Ma [Thu, 8 Mar 2007 22:59:30 +0000 (14:59 -0800)]
IPoIB: Turn on interface's carrier after broadcast group is joined

Do netif_carrier_on() right after the IPv4 broadcast multicast group
is joined, rather than waiting for all of the initial set of multicast
group joins to finish.  This allows at least IPv4 traffic to limp
along on broken fabrics where not all multicast groups can be joined.

Signed-off-by: Shirley Ma <xma@us.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years ago[SCTP] ipv6: inconsistent lock state ipv6_add_addr/sctp_v6_copy_addrlist
Jarek Poplawski [Thu, 8 Mar 2007 22:43:41 +0000 (14:43 -0800)]
[SCTP] ipv6: inconsistent lock state ipv6_add_addr/sctp_v6_copy_addrlist

lockdep found that dev->lock taken from softirq in ipv6_add_addr
is also taken in sctp_v6_copy_addrlist with softirqs enabled, so
lockup is possible.

Noticed-by: Simon Arlott <simon@arlott.org>
Signed-off-by: Jarek Poplawski <jarkao2@o2.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[PATCH] libata-acpi: Try and stop all the non PCI devices crashing
Alan Cox [Thu, 8 Mar 2007 23:13:50 +0000 (23:13 +0000)]
[PATCH] libata-acpi: Try and stop all the non PCI devices crashing

For 2.6.20 it mostly used to just not work, for 2.6.21-rc it crashes, this
seems to be down to luck (bad or good). The libata-acpi code needs to
avoid doing PCI work on non-PCI devices. This is one hack although it's
not pretty and perhaps there is a "right" way to check if a struct device
* is PCI ?

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] ecryptfs: nested locking annotation
Peter Zijlstra [Thu, 8 Mar 2007 04:41:30 +0000 (20:41 -0800)]
[PATCH] ecryptfs: nested locking annotation

ecryptfs uses a lock_parent() function, which I hope really locks the parents
and is not abused

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Michael Halcrow <mhalcrow@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] 2.6 Altix: console fix for CONFIG_DEBUG_SHIRQ usage
Pat Gefre [Thu, 8 Mar 2007 04:41:29 +0000 (20:41 -0800)]
[PATCH] 2.6 Altix: console fix for CONFIG_DEBUG_SHIRQ usage

The sn console driver was snagged by the use of CONFIG_DEBUG_SHIRQ!

The request_irq() immediate call to the interrupt handler caused
another attempt to lock the port lock - deadlock.

This is a patch to fix that.

Signed-off-by: Patrick Gefre <pfg@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] uml: arch_prctl should set thread fs
Jeff Dike [Thu, 8 Mar 2007 04:41:26 +0000 (20:41 -0800)]
[PATCH] uml: arch_prctl should set thread fs

In my previous x86_64 thread fix, I forgot to initialize thread.arch.fs in
arch_prctl.  A process calling arch_prctl to set %fs would lose it on the
next context switch.

It also turns out that you can switch to a process which is in the process
of exiting and which has lost its mm.  In this case, it's worse than
useless to try to call arch_prctl on the host process.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] check_partition(): fix error check
suzuki [Thu, 8 Mar 2007 04:41:24 +0000 (20:41 -0800)]
[PATCH] check_partition(): fix error check

Fix inverted check introduced in 57881dd9df40b76dc7fc6a0d13fd75f337accb32 "Fix
check_partition routines".

Signed-off-by: Suzuki K P <suzuki@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] Add epoll compat_ code to fs/compat.c
Davide Libenzi [Thu, 8 Mar 2007 04:41:21 +0000 (20:41 -0800)]
[PATCH] Add epoll compat_ code to fs/compat.c

IA64 and ARM-OABI are currently using their own version of epoll compat_
code.

An architecture needs epoll_event translation if alignof(u64) in 32 bit
mode is different from alignof(u64) in 64 bit mode.  If an architecture
needs epoll_event translation, it must define struct compat_epoll_event in
asm/compat.h and set CONFIG_HAVE_COMPAT_EPOLL_EVENT and use
compat_sys_epoll_ctl and compat_sys_epoll_wait.

All 64 bit architecture should use compat_sys_epoll_pwait.

[sfr: restructure and move to fs/compat.c, remove MIPS version
of compat_sys_epoll_pwait, use __put_user_unaligned]

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] bluetooth: fix socket locking in hci_sock_dev_event()
Jiri Kosina [Thu, 8 Mar 2007 04:41:18 +0000 (20:41 -0800)]
[PATCH] bluetooth: fix socket locking in hci_sock_dev_event()

[Bluetooth] Fix socket locking in hci_sock_dev_event()

hci_sock_dev_event() uses bh_lock_sock() to lock the socket lock.
This is not deadlock-safe against locking of the same socket lock in
l2cap_connect_cfm() from softirq context. In addition to that,
hci_sock_dev_event() doesn't seem to be called from softirq context,
so it is safe to use lock_sock()/release_sock() instead.

The lockdep warning can be triggered on my T42p simply by switching
the Bluetooth off by the keyboard button.

  =================================
  [ INFO: inconsistent lock state ]
  2.6.21-rc2 #4
  ---------------------------------
  inconsistent {in-softirq-W} -> {softirq-on-W} usage.
  khubd/156 [HC0[0]:SC0[0]:HE1:SE1] takes:
   (slock-AF_BLUETOOTH){-+..}, at: [<e0ca5520>] hci_sock_dev_event+0xa8/0xc5 [bluetooth]
  {in-softirq-W} state was registered at:
    [<c012d1db>] mark_lock+0x59/0x414
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<c012dfd7>] __lock_acquire+0x3e5/0xb99
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<c012e7f2>] lock_acquire+0x67/0x81
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<c036ee72>] _spin_lock+0x29/0x34
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<e0ca17c3>] hci_send_cmd+0x126/0x14f [bluetooth]
    [<e0ca4ce4>] hci_event_packet+0x729/0xebd [bluetooth]
    [<e0ca205b>] hci_rx_task+0x2a/0x20f [bluetooth]
    [<e0ca209d>] hci_rx_task+0x6c/0x20f [bluetooth]
    [<c012d7be>] trace_hardirqs_on+0x10d/0x14e
    [<c011ac85>] tasklet_action+0x3d/0x68
    [<c011abba>] __do_softirq+0x41/0x92
    [<c011ac32>] do_softirq+0x27/0x3d
    [<c0105134>] do_IRQ+0x7b/0x8f
    [<c0103dec>] common_interrupt+0x24/0x34
    [<c0103df6>] common_interrupt+0x2e/0x34
    [<c0248e65>] acpi_processor_idle+0x1b3/0x34a
    [<c0248e68>] acpi_processor_idle+0x1b6/0x34a
    [<c010232b>] cpu_idle+0x39/0x4e
    [<c04bab0c>] start_kernel+0x372/0x37a
    [<c04ba42b>] unknown_bootoption+0x0/0x202
    [<ffffffff>] 0xffffffff

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] x86_64: fix 2.6.18 regression - PTRACE_OLDSETOPTIONS should be accepted
Paolo 'Blaisorblade' Giarrusso [Thu, 8 Mar 2007 04:41:17 +0000 (20:41 -0800)]
[PATCH] x86_64: fix 2.6.18 regression - PTRACE_OLDSETOPTIONS should be accepted

Also PTRACE_OLDSETOPTIONS should be accepted, as done by kernel/ptrace.c
and forced by binary compatibility.  UML/32bit breaks because of this -
since it is wise enough to use PTRACE_OLDSETOPTIONS to be binary compatible
with 2.4 host kernels.

Until 2.6.17 (commit f0f2d6536e3515b5b1b7ae97dc8f176860c8c2ce) we had:

       default:
                return sys_ptrace(request, pid, addr, data);

Instead here we have:
        case PTRACE_GET_THREAD_AREA:
case ...:
                return sys_ptrace(request, pid, addr, data);

        default:
                return -EINVAL;

This change was a style change - when a case is added, it must be
explicitly tested this way.  In this case, not enough testing was done.

Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] um: fix errno usage
Paolo 'Blaisorblade' Giarrusso [Thu, 8 Mar 2007 04:41:14 +0000 (20:41 -0800)]
[PATCH] um: fix errno usage

Avoid reusing userspace errno twice - it can be cleared by libc code
everywhere (in particular printk() does clear it in my setup).

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] uml: activate_fd: return ENOMEM only when appropriate
Paolo 'Blaisorblade' Giarrusso [Thu, 8 Mar 2007 04:41:13 +0000 (20:41 -0800)]
[PATCH] uml: activate_fd: return ENOMEM only when appropriate

Avoid returning ENOMEM in case of a duplicate IRQ - ENOMEM was saved into err
earlier.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] um: fix confusion irq early reenabling
Paolo 'Blaisorblade' Giarrusso [Thu, 8 Mar 2007 04:41:12 +0000 (20:41 -0800)]
[PATCH] um: fix confusion irq early reenabling

Fix confusion about call context - comments and code are inconsistent and
plain wrong, my fault.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] um: mark both consoles as CON_ANYTIME
Paolo 'Blaisorblade' Giarrusso [Thu, 8 Mar 2007 04:41:11 +0000 (20:41 -0800)]
[PATCH] um: mark both consoles as CON_ANYTIME

Since both UML consoles do not use percpu variables, they may be called when
the cpu is still offline, and they may be marked CON_ANYTIME (this is
documented in kernel/printk.c, grep for CON_ANYTIME to find mentions of this).

Works well in testing done with lock debug enabled, should be safe but is not
needed for next release.

This would probably help also stderr_console.c, but this is yet to test.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] um: remove dead code about os_usr1_signal() and os_usr1_process()
Paolo 'Blaisorblade' Giarrusso [Thu, 8 Mar 2007 04:41:10 +0000 (20:41 -0800)]
[PATCH] um: remove dead code about os_usr1_signal() and os_usr1_process()

os_usr1_signal() is totally unused, os_usr1_process() is used only by TT mode.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] um: fix a memory leak in the multicast driver
Paolo 'Blaisorblade' Giarrusso [Thu, 8 Mar 2007 04:41:09 +0000 (20:41 -0800)]
[PATCH] um: fix a memory leak in the multicast driver

Memory allocated by mcast_user_init must be freed in the matching mcast_remove.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] uml: hostfs: make hostfs= option work as a jail, as intended.
Paolo 'Blaisorblade' Giarrusso [Thu, 8 Mar 2007 04:41:08 +0000 (20:41 -0800)]
[PATCH] uml: hostfs: make hostfs= option work as a jail, as intended.

When a given host directory is specified to be mounted both in hostfs=path1
and with mount option -o path2, we should give access to path1/path2, but this
does not happen.  Fix that in the simpler way.

Also, root_ino can be the empty string, since we use %s/%s as format.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] uml: hostfs: fix double free
Paolo 'Blaisorblade' Giarrusso [Thu, 8 Mar 2007 04:41:07 +0000 (20:41 -0800)]
[PATCH] uml: hostfs: fix double free

Fix double free in the error path - when name is assigned into root_inode we
do not own it any more and we must not kfree() it - see patch for details.

Thanks to William Stearns for the initial report.

CC: William Stearns <wstearns@pobox.com>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMerge branch 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Thu, 8 Mar 2007 15:28:30 +0000 (07:28 -0800)]
Merge branch 'for-linus' of /linux/kernel/git/dtor/input

* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/dtor/input:
  Input: i8042 - another attempt to fix AUX delivery checks

17 years ago[MTD] Correct partition failed erase address
Adrian Hunter [Thu, 8 Mar 2007 10:20:12 +0000 (12:20 +0200)]
[MTD] Correct partition failed erase address

If an erase operation fails, the address at which the
failure occurred is returned by the driver.  The MTD
partition must adjust this address (by subtracting the
partition offset) before returning to the caller.
This was not happening, which caused JFFS2 to mark
the wrong block bad!

Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[JFFS2] Use yield() between GC passes in background thread.
David Woodhouse [Thu, 8 Mar 2007 10:28:30 +0000 (10:28 +0000)]
[JFFS2] Use yield() between GC passes in background thread.

The garbage collection thread is strictly an optimisation. Everything it
does would also be done just-in-time in the context of something in
userspace trying to access the file system.

Sometimes, however, it's a pessimisation. Especially during early boot
when it's checksumming nodes and scanning inodes which are shortly going
to be pulled in by read_inode anyway. We end up building the rbtree of
node coverage twice for the same inode.

By switching to yield() instead of cond_resched() in the main loop, we
observe boot times on the OLPC system going down from about 100 seconds to
60.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [NAND] Correct misspelled preprocessor variable.
Robert P. J. Day [Wed, 7 Mar 2007 23:33:25 +0000 (18:33 -0500)]
[MTD] [NAND] Correct misspelled preprocessor variable.

Replace the apparently misspelled preprocessor variable
"MTD_NAND_DISKONCHIP_BBTWRITE" with the correct form
"CONFIG_MTD_NAND_DISKONCHIP_BBTWRITE".

Signed-off-by: Robert P. J. Day <rpjday@mindspring.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [MAPS] dilnetpc: Fix printk warning
Alan Cox [Wed, 7 Mar 2007 16:36:01 +0000 (16:36 +0000)]
[MTD] [MAPS] dilnetpc: Fix printk warning

The type of a resource could be 32 or 64bit depending upon platform or
option so cast it explicitly.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [NOR] Fix oops in cfi_amdstd_sync
Vijay Sampath [Tue, 6 Mar 2007 10:39:44 +0000 (02:39 -0800)]
[MTD] [NOR] Fix oops in cfi_amdstd_sync

The files cfi_cmdset_0002.c and cfi_cmdset_0020.c do not initialize their
wait queues like is done in cfi_cmdset_0001.c.  This causes an oops when
the wait queue is accessed.  I have copied the code from cfi_cmdset_0001.c
that is pertinent to initialization of the wait queue.

Signed-off-by: Vijay Sampath <vsampath@gmail.com>
Acked-by: Joern Engel <joern@lazybastard.org>
Acked-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] ESB2 check for closed ROM window
Cyrill Gorcunov [Tue, 6 Mar 2007 10:39:45 +0000 (02:39 -0800)]
[MTD] ESB2 check for closed ROM window

Add checking for closed ROM window on Intel ESB2 Southbridge.

Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[JFFS2] Fix writebuffer recovery in the first page of a block
Vitaly Wool [Tue, 6 Mar 2007 14:01:04 +0000 (17:01 +0300)]
[JFFS2] Fix writebuffer recovery in the first page of a block

For the case when nand_write_page fail with -EIO for the first page in an
eraseblock, jffs2_wbuf_recover ends up producing a BUG in jffs2_block_refile
as jeb->first_node is not yet set up (it's set up later in jffs2_wbuf_recover).
This BUG is not really a bug; it's just jffs2_wbuf_recover calling
jffs2_block_refile with the wrong second parameter.
This patch takes care of this situation.

Signed-off-by: Vitaly Wool <vwool@ru.mvista.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [NAND] make oobavail public
Vitaly Wool [Tue, 6 Mar 2007 13:56:34 +0000 (16:56 +0300)]
[MTD] [NAND] make oobavail public

During the MTD rework the oobavail parameter of mtd_info structure has become
private. This is not quite correct in terms of integrity and logic. If we have
means to write to OOB area, then we'd like to know upfront how many bytes out
of OOB are spare per page to be able to adapt to specific cases.
The patch inlined adds the public oobavail parameter.

Signed-off-by: Vitaly Wool <vwool@ru.mvista.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[POWERPC] 85xx: Enable CONFIG_SERIAL_8250_SHARE_IRQ
Kumar Gala [Thu, 8 Mar 2007 05:14:57 +0000 (23:14 -0600)]
[POWERPC] 85xx: Enable CONFIG_SERIAL_8250_SHARE_IRQ

On chip 8250 UARTs share an interrupts on existing 85xx processors.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years ago[POWERPC] Select u-image as default image for Linkstation
Segher Boessenkool [Thu, 22 Feb 2007 22:52:02 +0000 (23:52 +0100)]
[POWERPC] Select u-image as default image for Linkstation

Fixes image build error:

ln: accessing `arch/powerpc/boot/zImage': No such file or directory
make[1]: *** [arch/powerpc/boot/zImage] Error 1
make: *** [zImage] Error 2

Signed-off-by: Segher Boessenkool <segher@kernel.crashing.org>
Acked-by: G. Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years ago[POWERPC] 83xx: Minor fixes for 834x_mds USB setup code
Li Yang [Tue, 6 Mar 2007 09:11:06 +0000 (17:11 +0800)]
[POWERPC] 83xx: Minor fixes for 834x_mds USB setup code

Fix broken node manipulating code, and clarify inaccurate comment.

Signed-off-by: Li Yang <leoli@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years agoInput: i8042 - another attempt to fix AUX delivery checks
Dmitry Torokhov [Thu, 8 Mar 2007 04:20:55 +0000 (23:20 -0500)]
Input: i8042 - another attempt to fix AUX delivery checks

Do not assume that AUX_LOOP command is broken unless it
completes successfully but returns wrong (unexpected) data.

Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
17 years ago[PATCH] bcm43xx: Fix assertion failures in interrupt handler
Pavel Roskin [Tue, 6 Mar 2007 01:28:00 +0000 (19:28 -0600)]
[PATCH] bcm43xx: Fix assertion failures in interrupt handler

In the bcm43xx interrupt handler, sanity checks are wrongly done before the
verification that the interrupt is for the bcm43xx.

Signed-off-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] bcm43xx: Fix bug in frequency to channel conversion
Joerg Sommer [Fri, 2 Mar 2007 18:32:06 +0000 (12:32 -0600)]
[PATCH] bcm43xx: Fix bug in frequency to channel conversion

The frequency to channel routine in bcm43xx requires that the frequency
be in MHz, but that condition is not always met. This patch does the
necessary conversion.

Signed-off-by: Joerg Sommer <joerg@alea.gnuu.de>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] bcm43xx: Fix errors in specs to code translation in B6PHY init
Larry Finger [Wed, 28 Feb 2007 20:54:39 +0000 (14:54 -0600)]
[PATCH] bcm43xx: Fix errors in specs to code translation in B6PHY init

There are three errors in the transcription of the latest revision to the
B6PHY init specifications.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Thu, 8 Mar 2007 03:05:34 +0000 (19:05 -0800)]
Merge /pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [BRIDGE]: adding new device to bridge should enable if up
  [IPV6]: Do not set IF_READY if device is down
  [IPSEC]: xfrm audit hook misplaced in pfkey_delete and xfrm_del_sa
  [IPSEC]: Add xfrm policy change auditing to pfkey_spdget
  [IPSEC]: xfrm_policy delete security check misplaced
  [CONNECTOR]: Bugfix for cn_call_callback()
  [DCCP]: Revert patch which disables bidirectional mode
  [IPV6]: Handle np->opt being NULL in ipv6_getsockopt_sticky().
  [UDP]: Reread uh pointer after pskb_trim
  [NETFILTER]: nfnetlink_log: fix crash on bridged packet
  [NETFILTER]: nfnetlink_log: zero-terminate prefix
  [NETFILTER]: nf_conntrack_ipv6: fix incorrect classification of IPv6 fragments as ESTABLISHED

17 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Thu, 8 Mar 2007 03:04:21 +0000 (19:04 -0800)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] ARC: Fix several compiler warnings.
  [MIPS] ISA: Fix typo
  [CHAR] ds1286: Fix handling of seconds in RTC_ALM_SET ioctl.

17 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6
Linus Torvalds [Thu, 8 Mar 2007 03:03:51 +0000 (19:03 -0800)]
Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6:
  [IA64] fsys_getcpu for IA64
  [IA64] remove duplicate declaration of efi_initialize_iomem_resources
  [IA64] Pick highest possible saved_max_pfn for crash_dump
  [IA64] fix NULL pointer in ia64/irq_chip-mask/unmask function

17 years ago[MIPS] ARC: Fix several compiler warnings.
Ralf Baechle [Thu, 8 Mar 2007 00:47:11 +0000 (00:47 +0000)]
[MIPS] ARC: Fix several compiler warnings.

  CC      arch/mips/arc/init.o
arch/mips/arc/init.c: In function 'prom_init':
arch/mips/arc/init.c:27: warning: ISO C90 forbids mixed declarations and code
arch/mips/arc/init.c:35: warning: format ‘%08lx’ expects type ‘long unsigned int’, but argument 2 has type 'ULONG'
arch/mips/arc/init.c:28: warning: unused variable 'c'
arch/mips/arc/init.c:27: warning: unused variable ‘cnt’

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] ISA: Fix typo
Ralf Baechle [Thu, 8 Mar 2007 00:45:26 +0000 (00:45 +0000)]
[MIPS] ISA: Fix typo

Lost CONFIG_ prefix discovered by Robert P. J. Day <rpjday@mindspring.com>.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[CHAR] ds1286: Fix handling of seconds in RTC_ALM_SET ioctl.
Ralf Baechle [Thu, 8 Mar 2007 00:14:25 +0000 (00:14 +0000)]
[CHAR] ds1286: Fix handling of seconds in RTC_ALM_SET ioctl.

 o Fix use of uninitialized variable sec.
 o Make the RTC_ALM_SET ioctl return -EINVAL for non-zero seconds - the
   DS1286 has no second field for the alarm time.
 o Replace the obscure BIN_TO_BCD macro with BIN2BCD.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[IA64] fsys_getcpu for IA64
Fenghua Yu [Tue, 13 Feb 2007 00:27:10 +0000 (16:27 -0800)]
[IA64] fsys_getcpu for IA64

On 1.6GHz Montectio Tiger4, the following performance data is measured with
kernel built with defconfig which has NUMA configured:

Fastest sys_getcpu: 502 itc counts.
Fastest fsys_getcpu: 28 itc counts.

fsys_getcpu performance is largly impacted by whether data (node_to_cpu_map
etc) is in cache. It can take fsys_getcpu up to ~150 itc counts in cold
cache case.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years ago[IA64] remove duplicate declaration of efi_initialize_iomem_resources
Horms [Tue, 6 Mar 2007 07:20:10 +0000 (16:20 +0900)]
[IA64] remove duplicate declaration of efi_initialize_iomem_resources

efi_initialize_iomem_resources() is declared in both include/linux/efi.h
and arch/ia64/kernel/setup.c. This patch removes the latter.

Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years ago[IA64] Pick highest possible saved_max_pfn for crash_dump
Tony Luck [Thu, 8 Mar 2007 00:13:25 +0000 (16:13 -0800)]
[IA64] Pick highest possible saved_max_pfn for crash_dump

Berhhard Walle noted that on his HP rx8640 he ended up with saved_max_pfn
smaller than the highest address of system ram in /proc/iomem and proposed
a patch to base the address on the unrounded and unfiltered EFI memory
map address.  Simon Horman and Magnus Damm suggested that the whole test
be moved earlier in the function.  This is the combination of both of
these patches.

Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years ago[BRIDGE]: adding new device to bridge should enable if up
Aji Srinivas [Thu, 8 Mar 2007 00:10:53 +0000 (16:10 -0800)]
[BRIDGE]: adding new device to bridge should enable if up

One change introduced by the workqueue removal patch is that adding an
interface that is up to a bridge which is also up does not ever call
br_stp_enable_port(), leaving the port in DISABLED state until we do
ifconfig down and up or link events occur.

The following patch to the br_add_if function fixes it.
This is a regression introduced in 2.6.21.

Submitted-by: Aji_Srinivas@emc.com
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IPV6]: Do not set IF_READY if device is down
Herbert Xu [Thu, 8 Mar 2007 00:02:40 +0000 (16:02 -0800)]
[IPV6]: Do not set IF_READY if device is down

Now that we add the IPv6 device at registration time we don't need
to set IF_READY in ipv6_add_dev anymore because we will always get
a NETDEV_UP event later on should the device ever become ready.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IPSEC]: xfrm audit hook misplaced in pfkey_delete and xfrm_del_sa
Eric Paris [Thu, 8 Mar 2007 00:02:16 +0000 (16:02 -0800)]
[IPSEC]: xfrm audit hook misplaced in pfkey_delete and xfrm_del_sa

Inside pfkey_delete and xfrm_del_sa the audit hooks were not called if
there was any permission/security failures in attempting to do the del
operation (such as permission denied from security_xfrm_state_delete).
This patch moves the audit hook to the exit path such that all failures
(and successes) will actually get audited.

Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Venkat Yekkirala <vyekkirala@trustedcs.com>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IPSEC]: Add xfrm policy change auditing to pfkey_spdget
Eric Paris [Thu, 8 Mar 2007 00:01:45 +0000 (16:01 -0800)]
[IPSEC]: Add xfrm policy change auditing to pfkey_spdget

pfkey_spdget neither had an LSM security hook nor auditing for the
removal of xfrm_policy structs.  The security hook was added when it was
moved into xfrm_policy_byid instead of the callers to that function by
my earlier patch and this patch adds the auditing hooks as well.

Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Venkat Yekkirala <vyekkirala@trustedcs.com>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IPSEC]: xfrm_policy delete security check misplaced
Eric Paris [Wed, 7 Mar 2007 23:37:58 +0000 (15:37 -0800)]
[IPSEC]: xfrm_policy delete security check misplaced

The security hooks to check permissions to remove an xfrm_policy were
actually done after the policy was removed.  Since the unlinking and
deletion are done in xfrm_policy_by* functions this moves the hooks
inside those 2 functions.  There we have all the information needed to
do the security check and it can be done before the deletion.  Since
auditing requires the result of that security check err has to be passed
back and forth from the xfrm_policy_by* functions.

This patch also fixes a bug where a deletion that failed the security
check could cause improper accounting on the xfrm_policy
(xfrm_get_policy didn't have a put on the exit path for the hold taken
by xfrm_policy_by*)

It also fixes the return code when no policy is found in
xfrm_add_pol_expire.  In old code (at least back in the 2.6.18 days) err
wasn't used before the return when no policy is found and so the
initialization would cause err to be ENOENT.  But since err has since
been used above when we don't get a policy back from the xfrm_policy_by*
function we would always return 0 instead of the intended ENOENT.  Also
fixed some white space damage in the same area.

Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Venkat Yekkirala <vyekkirala@trustedcs.com>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[CONNECTOR]: Bugfix for cn_call_callback()
Philipp Reisner [Wed, 7 Mar 2007 20:55:39 +0000 (12:55 -0800)]
[CONNECTOR]: Bugfix for cn_call_callback()

When system under heavy stress and must allocate new work
instead of reusing old one, new work must use correct
completion callback.

Patch is based on Philipp's and Lars' work.
I only cleaned small stuff (and removed spaces instead of tabs).

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[DCCP]: Revert patch which disables bidirectional mode
Gerrit Renker [Wed, 7 Mar 2007 20:53:48 +0000 (12:53 -0800)]
[DCCP]: Revert patch which disables bidirectional mode

This reverts an earlier patch which disabled bidirectional mode, meaning that
a listening (passive) socket was not allowed to write to the other (active)
end of the connection.

This mode had been disabled when there were problems with CCID3, but it
imposes a constraint on socket programming and thus hinders deployment.

A change is included to ignore RX feedback received by the TX CCID3 module.

Many thanks to Andre Noll for pointing out this issue.

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IPV6]: Handle np->opt being NULL in ipv6_getsockopt_sticky().
David S. Miller [Wed, 7 Mar 2007 20:50:46 +0000 (12:50 -0800)]
[IPV6]: Handle np->opt being NULL in ipv6_getsockopt_sticky().

Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[UDP]: Reread uh pointer after pskb_trim
Herbert Xu [Wed, 7 Mar 2007 04:26:48 +0000 (20:26 -0800)]
[UDP]: Reread uh pointer after pskb_trim

The header may have moved when trimming.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: nfnetlink_log: fix crash on bridged packet
Patrick McHardy [Wed, 7 Mar 2007 04:24:53 +0000 (20:24 -0800)]
[NETFILTER]: nfnetlink_log: fix crash on bridged packet

physoutdev is only set on purely bridged packet, when nfnetlink_log is used
in the OUTPUT/FORWARD/POSTROUTING hooks on packets forwarded from or to a
bridge it crashes when trying to dereference skb->nf_bridge->physoutdev.

Reported by Holger Eitzenberger <heitzenberger@astaro.com>

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>