linux-2.6
15 years agousbnet: remove old compat_net_dev_ops code
Alexander Beregalov [Wed, 15 Apr 2009 07:58:28 +0000 (07:58 +0000)] 
usbnet: remove old compat_net_dev_ops code

Since all usb network drivers are already converted to net_device_ops
this code is useless.

Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoipw2x00: remove old compat_net_dev_ops code
Alexander Beregalov [Wed, 15 Apr 2009 07:58:27 +0000 (07:58 +0000)] 
ipw2x00: remove old compat_net_dev_ops code

Since both ipw2100 and ipw2200 are already converted to new
net_device_ops this code is useless.

Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agogro: New frags interface to avoid copying shinfo
Herbert Xu [Thu, 16 Apr 2009 09:02:07 +0000 (02:02 -0700)] 
gro: New frags interface to avoid copying shinfo

It turns out that copying a 16-byte area at ~800k times a second
can be really expensive :) This patch redesigns the frags GRO
interface to avoid copying that area twice.

The two disciples of the frags interface have been converted.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoNET/r8169: Rework suspend and resume
Rafael J. Wysocki [Sun, 5 Apr 2009 08:40:04 +0000 (08:40 +0000)] 
NET/r8169: Rework suspend and resume

The recent changes of the PCI PM core allow us to simplify the
suspend and resume handling in a number of device drivers, since they
don't need to carry out the general PCI PM operations, such as
changing the power state of the device, during suspend and resume any
more.

Simplify the suspend and resume callbacks of r8169 using the
observation that the PCI PM core can take care of some operations
carried out by the driver.

Additionally, make the shutdown callback of r8169 only put the device
into a low power state if the system is going to be powered off
(kexec is known to have problems with network adapters that are put
into low power states on shutdown).

This patch has been tested on MSI Wind U100.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Tested-by: Bruno Prémont <bonbons@linux-vserver.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agodrivers/net: replace BUG() with BUG_ON() if possible
Alexander Beregalov [Sat, 11 Apr 2009 14:50:23 +0000 (14:50 +0000)] 
drivers/net: replace BUG() with BUG_ON() if possible

Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agokaweth: Clean up code
Larry Finger [Thu, 9 Apr 2009 07:11:55 +0000 (07:11 +0000)] 
kaweth: Clean up code

The driver kaweth yields a -EBUSY error when starting, and a -ETIME
error when shutting down. These errors are avoided, and the RX status
is further checked for other potential errors.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoat76c50x-usb: remove pointless conditional before kfree_skb()
Wei Yongjun [Wed, 8 Apr 2009 22:16:38 +0000 (22:16 +0000)] 
at76c50x-usb: remove pointless conditional before kfree_skb()

Remove pointless conditional before kfree_skb().

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobe2net: changes for checksum and segmentation offload
Ajit Khaparde [Mon, 13 Apr 2009 22:41:22 +0000 (15:41 -0700)] 
be2net: changes for checksum and segmentation offload

- Enables Rx checksum feature by default.
- Disables support for ipv6 tso.
- Changes in Rx path to handle Rx completions with various checksum options.

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Be explict with what we are !'ing in ixgbe_sfp_config_module_task()
Tony Breeds [Thu, 9 Apr 2009 22:29:10 +0000 (22:29 +0000)] 
ixgbe: Be explict with what we are !'ing in ixgbe_sfp_config_module_task()

GCC warns:
drivers/net/ixgbe/ixgbe_main.c: In function 'ixgbe_sfp_config_module_task':
drivers/net/ixgbe/ixgbe_main.c:3920: warning: suggest parantheses around
operand of '!' or change '&' to '&&' or '!' to '~'

Which I think is right.  Bracket to remove ambiguity.

Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Update get_physical_layer() calls, plus a version bump
Peter P Waskiewicz Jr [Thu, 9 Apr 2009 22:28:50 +0000 (22:28 +0000)] 
ixgbe: Update get_physical_layer() calls, plus a version bump

Not all physical connection types are being correctly identified.  This
fixes that issue, and cleans up the logic to make it more maintainable.

Also clean up the code for device capabilities from the EEPROM to support
multiple SKUs of the same hardware.

Bump the version to reflect all the updates since the 82599 merge.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Update the usage of orig_autoc to be more consistent
PJ Waskiewicz [Thu, 9 Apr 2009 22:28:33 +0000 (22:28 +0000)] 
ixgbe: Update the usage of orig_autoc to be more consistent

The orig_autoc variable tracks the original setting of the autonegotiate
state prior to trying a new speed.  The usage is inconsistent and not very
maintainable.  This patch updates the usage to make it more consistent.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Remove unnecessary PHY reset, properly identify multispeed fiber modules
PJ Waskiewicz [Thu, 9 Apr 2009 22:28:15 +0000 (22:28 +0000)] 
ixgbe: Remove unnecessary PHY reset, properly identify multispeed fiber modules

This patch does two things:
1) On 82599, the PHY is emedded in the MAC.  On 82598, the SFP+ NIC has an external PHY.  The reset in the SFP+ setup patch for 82598 is unnecessary on 82599, and adds extra dead time to device initialization.  This removes that PHY reset for 82599 only.

2) On 82599, the SFP+ modules are multispeed fiber modules (10G/1G).  We need to make sure to identify them properly for the remaining init sections to properly set them up.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Move PHY ops initialization to centralize bus accesses
PJ Waskiewicz [Thu, 9 Apr 2009 22:27:57 +0000 (22:27 +0000)] 
ixgbe: Move PHY ops initialization to centralize bus accesses

When PHY operations are determined, the PHY must be identified.  This
identification causes bus access, and should be contained within its own
routines.  This also helps the 82599 PHY init paths for both SFP+ and
KX/KX4 devices to be easier to maintain.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Don't return error in flow control configuration if FC is off
PJ Waskiewicz [Thu, 9 Apr 2009 22:27:39 +0000 (22:27 +0000)] 
ixgbe: Don't return error in flow control configuration if FC is off

When flow control is disabled, an invalid low/high watermark configuration
should not matter.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: add define to support 82599 64 IVAR registers
Don Skidmore [Thu, 9 Apr 2009 22:27:19 +0000 (22:27 +0000)] 
ixgbe: add define to support 82599 64 IVAR registers

82599 supports 64 IVAR registers this patch adds a define to
allow us to access them.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: add support for addition VT_CTL defines
Don Skidmore [Thu, 9 Apr 2009 22:27:00 +0000 (22:27 +0000)] 
ixgbe: add support for addition VT_CTL defines

These defines allow for definition of target pool for packets that
pass L2 filtering but didn't pass any of the pool filters.  They are
needed to reset the default pool.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: fix IXGBE_MTQC values
Don Skidmore [Thu, 9 Apr 2009 22:26:40 +0000 (22:26 +0000)] 
ixgbe: fix IXGBE_MTQC values

IXGBE_MTQC_64VF was wrong and 32VF not defined at all.  This patch
corrects that.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Enable another bit for flow control operation
PJ Waskiewicz [Thu, 9 Apr 2009 22:26:21 +0000 (22:26 +0000)] 
ixgbe: Enable another bit for flow control operation

The discard pause frames bit was not enabled, so flow control frames could
be passed up to the host.  Enabled DPF to keep the frames off the stack.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoMAINTAINERS: rds list is moderated
Kyle McMartin [Thu, 9 Apr 2009 14:09:47 +0000 (14:09 +0000)] 
MAINTAINERS: rds list is moderated

Signed-off-by: Kyle McMartin <kyle@redhat.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoERR_PTR() dereference in net/rds/ib.c
Dan Carpenter [Thu, 9 Apr 2009 14:09:46 +0000 (14:09 +0000)] 
ERR_PTR() dereference in net/rds/ib.c

rdma_create_id() doesn't return NULL, only ERR_PTR().

Found by smatch (http://repo.or.cz/w/smatch.git).  Compile tested.

regards,
dan carpenter

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoERR_PTR() dereference in net/rds/iw.c
Dan Carpenter [Thu, 9 Apr 2009 14:09:45 +0000 (14:09 +0000)] 
ERR_PTR() dereference in net/rds/iw.c

rdma_create_id() returns ERR_PTR() not null.

Found by smatch (http://repo.or.cz/w/smatch.git).  Compile tested.

regards,
dan carpenter

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agords: use kmem_cache_zalloc instead of kmem_cache_alloc/memset
Wei Yongjun [Thu, 9 Apr 2009 14:09:44 +0000 (14:09 +0000)] 
rds: use kmem_cache_zalloc instead of kmem_cache_alloc/memset

Use kmem_cache_zalloc instead of kmem_cache_alloc/memset.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRDS: remove unused #include <version.h>
Huang Weiyi [Thu, 9 Apr 2009 14:09:43 +0000 (14:09 +0000)] 
RDS: remove unused #include <version.h>

Remove unused #include <version.h> in net/rds/af_rds.c.

Signed-off-by: Huang Weiyi <weiyi.huang@gmail.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRDS: use get_user_pages_fast()
Andy Grover [Thu, 9 Apr 2009 14:09:42 +0000 (14:09 +0000)] 
RDS: use get_user_pages_fast()

Use the new function that is simpler and faster.

Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRDS: Establish connection before parsing CMSGs
Andy Grover [Thu, 9 Apr 2009 14:09:41 +0000 (14:09 +0000)] 
RDS: Establish connection before parsing CMSGs

The first message to a remote node should prompt a new connection.
Even an RDMA op via CMSG. Therefore move CMSG parsing to after
connection establishment.

Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRDS: Fix ordering in a conditional
Andy Grover [Thu, 9 Apr 2009 14:09:40 +0000 (14:09 +0000)] 
RDS: Fix ordering in a conditional

Putting the constant first is a supposed "best practice" that actually makes
the code harder to read.

Thanks to Roland Dreier for finding a bug in this "simple, obviously correct"
patch.

Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRDS/IW+IB: Allow max credit advertise window.
Steve Wise [Thu, 9 Apr 2009 14:09:39 +0000 (14:09 +0000)] 
RDS/IW+IB: Allow max credit advertise window.

Fix hack that restricts the credit advertisement to 127.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRDS/IW+IB: Set the RDS_LL_SEND_FULL bit when we're throttled.
Steve Wise [Thu, 9 Apr 2009 14:09:38 +0000 (14:09 +0000)] 
RDS/IW+IB: Set the RDS_LL_SEND_FULL bit when we're throttled.

The RDS_LL_SEND_FULL bit should be set when we stop transmitted due to
flow control.  Otherwise the send worker will keep trying as opposed to
sleeping until we unthrottle.  Saves CPU.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRDS: Correct some iw references in rdma_transport.c
Andy Grover [Thu, 9 Apr 2009 14:09:37 +0000 (14:09 +0000)] 
RDS: Correct some iw references in rdma_transport.c

Had some lingering instances of _iw_ variable names from when
the listen code was centralized into rdma_transport.c

Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRDS/IW+IB: Set recv ring low water mark to 1/2 full.
Steve Wise [Thu, 9 Apr 2009 14:09:36 +0000 (14:09 +0000)] 
RDS/IW+IB: Set recv ring low water mark to 1/2 full.

Currently the recv ring low water mark is 1/4 the depth.  Performance
measurements show that this limits iWARP throughput by flow controlling
the rds-stress senders.  Setting it to 1/2 seems to max the T3
performance.  I tried even higher levels but that didn't help and it
started to increase the rds thread cpu utilization.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoISDN: update Documentation/isdn/00-INDEX
Tilman Schmidt [Wed, 8 Apr 2009 23:01:16 +0000 (16:01 -0700)] 
ISDN: update Documentation/isdn/00-INDEX

After the merging of mISDN, state which files refer only to the
old isdn4linux subsystem. Also add a few missing files and sort
alphabetically.

Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosfc: Don't specify unexistent IRQ
Jean Delvare [Wed, 8 Apr 2009 22:59:53 +0000 (15:59 -0700)] 
sfc: Don't specify unexistent IRQ

Neither the lm90 driver nor the lm87 driver do support interrupts, so
there is no point in specifying one when declaring the devices.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: cache align register map table
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:48 +0000 (22:50 +0000)] 
netxen: cache align register map table

Aligning register offset translation table imporves performance
on rx side.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: enable GRO support
Amit Kumar Salecha [Tue, 7 Apr 2009 22:50:47 +0000 (22:50 +0000)] 
netxen: enable GRO support

Signed-off-by: Amit Kumar Salecha <amit@netxen.com>
Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: enable rss for NX2031
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:46 +0000 (22:50 +0000)] 
netxen: enable rss for NX2031

Enable multiple rx rings for older NX2031 chip, firmware 3.4.336
or newer supports this feature.

Signed-off-by: Amit Kumar Salecha <amit@netxen.com>
Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: sanitize function names
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:45 +0000 (22:50 +0000)] 
netxen: sanitize function names

Replace superfluous wrapper functions with two macros:

NXWR32 replaces netxen_nic_reg_write, netxen_nic_write_w0,
netxen_nic_read_w1, netxen_crb_writelit_adapter.

NXRD32 replaces netxen_nic_reg_read, netxen_nic_read_w0,
netxen_nic_read_w1.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: annotate register access functions
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:44 +0000 (22:50 +0000)] 
netxen: annotate register access functions

o remove unnecessary length parameter since register access
  width is fixed 4 byte.
o remove superfluous pci_read_normalize and pci_write_normalize
  functions.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: allocate status rings dynamically
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:43 +0000 (22:50 +0000)] 
netxen: allocate status rings dynamically

This reduces netxen_adapter footprint when rss (msi-x) is disabled.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: async link event handling
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:42 +0000 (22:50 +0000)] 
netxen: async link event handling

Add support for asynchronous events from firmware,
received over one of the rx rings.

Add support for event based phy interrupts, enhanced links
status reporting from firmware.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: defer firmware handshake
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:41 +0000 (22:50 +0000)] 
netxen: defer firmware handshake

Removed duplicate firmware handshake, defer it until first
port (interface) is brought up.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: refactor transmit code
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:40 +0000 (22:50 +0000)] 
netxen: refactor transmit code

o move tx stuff into nx_host_tx_ring structure, this will
  help managing multiple tx rings in future.
o sanitize some variable names

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: refactor netxen_adapter
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:39 +0000 (22:50 +0000)] 
netxen: refactor netxen_adapter

Rearrange members to align them at right offset.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: code cleanup
Dhananjay Phadke [Tue, 7 Apr 2009 22:50:38 +0000 (22:50 +0000)] 
netxen: code cleanup

o remove unused structure defs.
o remove unnecessary includes.
o replace enums with specific #defines.
o reduce footprint of stats structure.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoLinux 2.6.30-rc1 v2.6.30-rc1
Linus Torvalds [Tue, 7 Apr 2009 21:25:01 +0000 (14:25 -0700)] 
Linux 2.6.30-rc1

15 years agoMerge branch 'core/softlockup' of git://git.kernel.org/pub/scm/linux/kernel/git/tip...
Linus Torvalds [Tue, 7 Apr 2009 21:11:07 +0000 (14:11 -0700)] 
Merge branch 'core/softlockup' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core/softlockup' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  softlockup: make DETECT_HUNG_TASK default depend on DETECT_SOFTLOCKUP
  softlockup: move 'one' to the softlockup section in sysctl.c
  softlockup: ensure the task has been switched out once
  softlockup: remove timestamp checking from hung_task
  softlockup: convert read_lock in hung_task to rcu_read_lock
  softlockup: check all tasks in hung_task
  softlockup: remove unused definition for spawn_softlockup_task
  softlockup: fix potential race in hung_task when resetting timeout
  softlockup: fix to allow compiling with !DETECT_HUNG_TASK
  softlockup: decouple hung tasks check from softlockup detection

15 years agoMerge branch 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 7 Apr 2009 21:10:10 +0000 (14:10 -0700)] 
Merge branch 'tracing-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  branch tracer, intel-iommu: fix build with CONFIG_BRANCH_TRACER=y
  branch tracer: Fix for enabling branch profiling makes sparse unusable
  ftrace: Correct a text align for event format output
  Update /debug/tracing/README
  tracing/ftrace: alloc the started cpumask for the trace file
  tracing, x86: remove duplicated #include
  ftrace: Add check of sched_stopped for probe_sched_wakeup
  function-graph: add proper initialization for init task
  tracing/ftrace: fix missing include string.h
  tracing: fix incorrect return type of ns2usecs()
  tracing: remove CALLER_ADDR2 from wakeup tracer
  blktrace: fix pdu_len when tracing packet command requests
  blktrace: small cleanup in blk_msg_write()
  blktrace: NUL-terminate user space messages
  tracing: move scripts/trace/power.pl to scripts/tracing/power.pl

15 years agoMerge branch 'irq/threaded' of git://git.kernel.org/pub/scm/linux/kernel/git/tip...
Linus Torvalds [Tue, 7 Apr 2009 21:07:52 +0000 (14:07 -0700)] 
Merge branch 'irq/threaded' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'irq/threaded' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  genirq: fix devres.o build for GENERIC_HARDIRQS=n
  genirq: provide old request_irq() for CONFIG_GENERIC_HARDIRQ=n
  genirq: threaded irq handlers review fixups
  genirq: add support for threaded interrupts to devres
  genirq: add threaded interrupt handler support

15 years agoNFS: Fix the return value in nfs_page_mkwrite()
Trond Myklebust [Tue, 7 Apr 2009 21:02:53 +0000 (14:02 -0700)] 
NFS: Fix the return value in nfs_page_mkwrite()

Commit c2ec175c39f62949438354f603f4aa170846aabb ("mm: page_mkwrite
change prototype to match fault") exposed a bug in the NFS
implementation of page_mkwrite.  We should be returning 0 on success...

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes...
Linus Torvalds [Tue, 7 Apr 2009 18:24:19 +0000 (11:24 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/jbarnes/pci-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
  PCI: pci_slot: grab refcount on slot's bus
  PCI Hotplug: acpiphp: grab refcount on p2p subordinate bus
  PCI: allow PCI core hotplug to remove PCI root bus
  PCI: Fix oops in pci_vpd_truncate
  PCI: don't corrupt enable_cnt when doing manual resource alignment
  PCI: annotate pci_rescan_bus as __ref, not __devinit
  PCI-IOV: fix missing kernel-doc
  PCI: Setup disabled bridges even if buses are added
  PCI: SR-IOV quirk for Intel 82576 NIC

15 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Tue, 7 Apr 2009 18:06:41 +0000 (11:06 -0700)] 
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  loop: mutex already unlocked in loop_clr_fd()
  cfq-iosched: don't let idling interfere with plugging
  block: remove unused REQ_UNPLUG
  cfq-iosched: kill two unused cfqq flags
  cfq-iosched: change dispatch logic to deal with single requests at the time
  mflash: initial support
  cciss: change to discover first memory BAR
  cciss: kernel scan thread for MSA2012
  cciss: fix residual count for block pc requests
  block: fix inconsistency in I/O stat accounting code
  block: elevator quiescing helpers

15 years agoFix build errors due to CONFIG_BRANCH_TRACER=y
Linus Torvalds [Tue, 7 Apr 2009 14:59:41 +0000 (07:59 -0700)] 
Fix build errors due to CONFIG_BRANCH_TRACER=y

The code that enables branch tracing for all (non-constant) branches
plays games with the preprocessor and #define's the C 'if ()' construct
to do tracing.

That's all fine, but it fails for some unusual but valid C code that is
sometimes used in macros, notably by the intel-iommu code:

if (i=drhd->iommu, drhd->ignored) ..

because now the preprocessor complains about multiple arguments to the
'if' macro.

So make the macro expansion of this particularly horrid trick use
varargs, and handle the case of comma-expressions in if-statements.  Use
another macro to do it cleanly in just one place.

This replaces a patch by David (and acked by Steven) that did this all
inside that one already-too-horrid macro.

Tested-by: Ingo Molnar <mingo@elte.hu>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'for-2.6.30' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie...
Linus Torvalds [Tue, 7 Apr 2009 15:54:43 +0000 (08:54 -0700)] 
Merge branch 'for-2.6.30' of git://git./linux/kernel/git/broonie/sound-2.6

* 'for-2.6.30' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound-2.6:
  ASoC: TWL4030: Compillation error fix

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Tue, 7 Apr 2009 15:53:38 +0000 (08:53 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6: (36 commits)
  ALSA: hda - Add VREF powerdown sequence for another board
  ALSA: oss - volume control for CSWITCH and CROUTE
  ALSA: hda - add missing comma in ad1884_slave_vols
  sound: usb-audio: allow period sizes less than 1 ms
  sound: usb-audio: save data packet interval in audioformat structure
  sound: usb-audio: remove check_hw_params_convention()
  sound: usb-audio: show sample format width in proc file
  ASoC: fsl_dma: Pass the proper device for dma mapping routines
  ASoC: Fix null dereference in ak4535_remove()
  ALSA: hda - enable SPDIF output for Intel DX58SO board
  ALSA: snd-atmel-abdac: increase periods_min to 6 instead of 4
  ALSA: snd-atmel-abdac: replace bus_id with dev_name()
  ALSA: snd-atmel-ac97c: replace bus_id with dev_name()
  ALSA: snd-atmel-ac97c: cleanup registers when removing driver
  ALSA: snd-atmel-ac97c: do a proper reset of the external codec
  ALSA: snd-atmel-ac97c: enable interrupts to catch events for error reporting
  ALSA: snd-atmel-ac97c: set correct size for buffer hardware parameter
  ALSA: snd-atmel-ac97c: do not overwrite OCA and ICA when assigning channels
  ALSA: snd-atmel-ac97c: remove dead break statements after return in switch case
  ALSA: snd-atmel-ac97c: cleanup register definitions
  ...

15 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Tue, 7 Apr 2009 15:53:02 +0000 (08:53 -0700)] 
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  sata_mv: shorten register names
  sata_mv: workaround errata SATA#13
  sata_mv: cosmetic renames
  sata_mv: workaround errata SATA#26
  sata_mv: workaround errata PCI#7
  sata_mv: replace 0x1f with ATA_PIO4 (v2)
  sata_mv: fix irq mask races
  sata_mv: revert SoC irq breakage
  libata: ahci enclosure management bios workaround
  ata: Add TRIM infrastructure
  ata_piix: VGN-BX297XP wants the controller power up on suspend
  libata: Remove some redundant casts from pata_octeon_cf.c
  pata_artop: typo

15 years agoMerge branch 'i2c-for-2630-v2' of git://aeryn.fluff.org.uk/bjdooks/linux
Linus Torvalds [Tue, 7 Apr 2009 15:45:12 +0000 (08:45 -0700)] 
Merge branch 'i2c-for-2630-v2' of git://aeryn.fluff.org.uk/bjdooks/linux

* 'i2c-for-2630-v2' of git://aeryn.fluff.org.uk/bjdooks/linux:
  i2c: imx: Make disable_delay a per-device variable
  i2c: xtensa s6000 i2c driver
  powerpc/85xx: i2c-mpc: use new I2C bindings for the Socates board
  i2c: i2c-mpc: make I2C bus speed configurable
  i2c: i2c-mpc: use dev based printout function
  i2c: i2c-mpc: various coding style fixes
  i2c: imx: Add missing request_mem_region in probe()
  i2c: i2c-s3c2410: Initialise Samsung I2C controller early
  i2c-s3c2410: Simplify bus frequency calculation
  i2c-s3c2410: sda_delay should be in ns, not clock ticks
  i2c: iMX/MXC support

15 years agoMerge branch 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6
Linus Torvalds [Tue, 7 Apr 2009 15:44:43 +0000 (08:44 -0700)] 
Merge branch 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6

* 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6:
  hwmon: Add Asus ATK0110 support
  hwmon: (lm95241) Convert to new-style i2c driver

15 years agoparport: Use the PCI IRQ if offered
Alan Cox [Tue, 7 Apr 2009 14:30:57 +0000 (15:30 +0100)] 
parport: Use the PCI IRQ if offered

PCI parallel port devices can IRQ share so we should stop them hogging
the line and making a mess on modern PC systems.  We know the sharing
side works as the PCMCIA driver has shared the parallel port IRQ for
some time.

Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotty: jsm cleanups
Breno Leitao [Tue, 7 Apr 2009 15:53:48 +0000 (16:53 +0100)] 
tty: jsm cleanups

Here are some cleanups, mainly removing unused variables and silly
declarations.

Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoAdjust path to gpio headers
Mike Frysinger [Tue, 7 Apr 2009 15:53:11 +0000 (16:53 +0100)] 
Adjust path to gpio headers

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoKGDB_SERIAL_CONSOLE check for module
Mike Frysinger [Tue, 7 Apr 2009 15:52:49 +0000 (16:52 +0100)] 
KGDB_SERIAL_CONSOLE check for module

Depend on KGDB_SERIAL_CONSOLE being set to N rather than !Y, since it can
be built as a module.

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoChange KCONFIG name
Mike Frysinger [Tue, 7 Apr 2009 15:52:39 +0000 (16:52 +0100)] 
Change KCONFIG name

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotty: Blackin CTS/RTS
Sonic Zhang [Tue, 7 Apr 2009 15:52:26 +0000 (16:52 +0100)] 
tty: Blackin CTS/RTS

Both software emulated and hardware based CTS and RTS are enabled in
serial driver.

The CTS RTS PIN connection on BF548 UART port is defined as a modem
device not as a host device.  In order to test it under Linux, please
nake a cross UART cable to exchange CTS and RTS signal.

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoChange hardware flow control from poll to interrupt driven
Sonic Zhang [Tue, 7 Apr 2009 15:51:15 +0000 (16:51 +0100)] 
Change hardware flow control from poll to interrupt driven

Only the CTS bit is affected.

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoAdd support for the MAX3100 SPI UART.
Christian Pellegrin [Tue, 7 Apr 2009 15:48:51 +0000 (16:48 +0100)] 
Add support for the MAX3100 SPI UART.

(akpm: queued pending confirmation of the new major number)

[randy.dunlap@oracle.com: select SERIAL_CORE]
Signed-off-by: Christian Pellegrin <chripell@fsfe.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agolanana: assign a device name and numbering for MAX3100
Alan Cox [Tue, 7 Apr 2009 15:48:35 +0000 (16:48 +0100)] 
lanana: assign a device name and numbering for MAX3100

This is a low density serial port so needs a real major/minor

Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoserqt: initial clean up pass for tty side
Alan Cox [Tue, 7 Apr 2009 15:48:27 +0000 (16:48 +0100)] 
serqt: initial clean up pass for tty side

Avoid using port->tty where possible (makes refcount fixing easier
later).

Remove unused code (the ioctl path is not used if the device has
mget/mset functions)

Remove various un-needed typecasts and long names so it could read it to
do the changes.

Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotty: Use the generic RS485 ioctl on CRIS
Claudio Scordino [Tue, 7 Apr 2009 15:48:19 +0000 (16:48 +0100)] 
tty: Use the generic RS485 ioctl on CRIS

Use the new general RS485 Linux data structure (introduced by Alan with
commit number c26c56c0f40e200e61d1390629c806f6adaffbcc) in the Cris
architecture too (currently, Cris still uses the old private data
structure instead of the new one).

Signed-off-by: Claudio Scordino <claudio@evidence.eu.com>
Tested-by: Hinko Kocevar <hinko.kocevar@cetrtapot.si>
Tested-by: Janez Cufer <janez.cufer@cetrtapot.si>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotty: Correct inline types for tty_driver_kref_get()
Adrian Bunk [Tue, 7 Apr 2009 15:48:07 +0000 (16:48 +0100)] 
tty: Correct inline types for tty_driver_kref_get()

tty_driver_kref_get() should be static inline and not extern inline
(the latter even changed it's semantics in gcc >= 4.3).

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agosplice: fix deadlock in splicing to file
Miklos Szeredi [Mon, 6 Apr 2009 15:41:00 +0000 (17:41 +0200)] 
splice: fix deadlock in splicing to file

There's a possible deadlock in generic_file_splice_write(),
splice_from_pipe() and ocfs2_file_splice_write():

 - task A calls generic_file_splice_write()
 - this calls inode_double_lock(), which locks i_mutex on both
   pipe->inode and target inode
 - ordering depends on inode pointers, can happen that pipe->inode is
   locked first
 - __splice_from_pipe() needs more data, calls pipe_wait()
 - this releases lock on pipe->inode, goes to interruptible sleep
 - task B calls generic_file_splice_write(), similarly to the first
 - this locks pipe->inode, then tries to lock inode, but that is
   already held by task A
 - task A is interrupted, it tries to lock pipe->inode, but fails, as
   it is already held by task B
 - ABBA deadlock

Fix this by explicitly ordering locks: the outer lock must be on
target inode and the inner lock (which is later unlocked and relocked)
must be on pipe->inode.  This is OK, pipe inodes and target inodes
form two nonoverlapping sets, generic_file_splice_write() and friends
are not called with a target which is a pipe.

Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Acked-by: Mark Fasheh <mfasheh@suse.com>
Acked-by: Jens Axboe <jens.axboe@oracle.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: support nanosecond timestamp
Ryusuke Konishi [Tue, 7 Apr 2009 02:02:00 +0000 (19:02 -0700)] 
nilfs2: support nanosecond timestamp

After a review of user's feedback for finding out other compatibility
issues, I found nilfs improperly initializes timestamps in inode;
CURRENT_TIME was used there instead of CURRENT_TIME_SEC even though nilfs
didn't have nanosecond timestamps on disk.  A few users gave us the report
that the tar program sometimes failed to expand symbolic links on nilfs,
and it turned out to be the cause.

Instead of applying the above displacement, I've decided to support
nanosecond timestamps on this occation.  Fortunetaly, a needless 64-bit
field was in the nilfs_inode struct, and I found it's available for this
purpose without impact for the users.

So, this will do the enhancement and resolve the tar problem.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: introduce secondary super block
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:59 +0000 (19:01 -0700)] 
nilfs2: introduce secondary super block

The former versions didn't have extra super blocks.  This improves the
weak point by introducing another super block at unused region in tail of
the partition.

This doesn't break disk format compatibility; older versions just ingore
the secondary super block, and new versions just recover it if it doesn't
exist.  The partition created by an old mkfs may not have unused region,
but in that case, the secondary super block will not be added.

This doesn't make more redundant copies of the super block; it is a future
work.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: simplify handling of active state of segments
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:58 +0000 (19:01 -0700)] 
nilfs2: simplify handling of active state of segments

will reduce some lines of segment constructor.  Previously, the state was
complexly controlled through a list of segments in order to keep
consistency in meta data of usage state of segments.  Instead, this
presents ``calculated'' active flags to userland cleaner program and stop
maintaining its real flag on disk.

Only by this fake flag, the cleaner cannot exactly know if each segment is
reclaimable or not.  However, the recent extension of nilfs_sustat ioctl
struct (nilfs2-extend-nilfs_sustat-ioctl-struct.patch) can prevent the
cleaner from reclaiming in-use segment wrongly.

So, now I can apply this for simplification.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: mark minor flag for checkpoint created by internal operation
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:57 +0000 (19:01 -0700)] 
nilfs2: mark minor flag for checkpoint created by internal operation

Nilfs creates checkpoints even for garbage collection or metadata updates
such as checkpoint mode change.  So, user often sees checkpoints created
only by such internal operations.

This is inconvenient in some situations.  For example, application that
monitors checkpoints and changes them to snapshots, will fall into an
infinite loop because it cannot distinguish internally created
checkpoints.

This patch solves this sort of problem by adding a flag to checkpoint for
identification.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: clean up sketch file
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:56 +0000 (19:01 -0700)] 
nilfs2: clean up sketch file

The sketch file is a file to mark checkpoints with user data.  It was
experimentally introduced in the original implementation, and now
obsolete.  The file was handled differently with regular files; the file
size got truncated when a checkpoint was created.

This stops the special treatment and will treat it as a regular file.
Most users are not affected because mkfs.nilfs2 no longer makes this file.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: super block operations fix endian bug
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:55 +0000 (19:01 -0700)] 
nilfs2: super block operations fix endian bug

This adds a missing endian conversion of checksum field in the super
block.  This fixes compatibility issue on big endian machines which will
come to surface after supporting recovery of super block.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: replace BUG_ON and BUG calls triggerable from ioctl
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:55 +0000 (19:01 -0700)] 
nilfs2: replace BUG_ON and BUG calls triggerable from ioctl

Pekka Enberg advised me:
> It would be nice if BUG(), BUG_ON(), and panic() calls would be
> converted to proper error handling using WARN_ON() calls. The BUG()
> call in nilfs_cpfile_delete_checkpoints(), for example, looks to be
> triggerable from user-space via the ioctl() system call.

This will follow the comment and keep them to a minimum.

Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: extend nilfs_sustat ioctl struct
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:54 +0000 (19:01 -0700)] 
nilfs2: extend nilfs_sustat ioctl struct

This adds a new argument to the nilfs_sustat structure.

The extended field allows to delete volatile active state of segments,
which was needed to protect freshly-created segments from garbage
collection but has confused code dealing with segments.  This
extension alleviates the mess and gives room for further
simplifications.

The volatile active flag is not persistent, so it's eliminable on this
occasion without affecting compatibility other than the ioctl change.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: use unlocked_ioctl
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:53 +0000 (19:01 -0700)] 
nilfs2: use unlocked_ioctl

Pekka Enberg suggested converting ->ioctl operations to use
->unlocked_ioctl to avoid BKL.

The conversion was verified to be safe, so I will take it on this
occasion.

Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: remove compat ioctl code
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:53 +0000 (19:01 -0700)] 
nilfs2: remove compat ioctl code

This removes compat code from the nilfs ioctls and applies the same
function for both .ioctl and .compat_ioctl file operations.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: use fixed sized types for ioctl structures
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:52 +0000 (19:01 -0700)] 
nilfs2: use fixed sized types for ioctl structures

Nilfs ioctl had structures not having fixed sized types such as:

  struct nilfs_argv {
         void *v_base;
         size_t v_nmembs;
         size_t v_size;
         int v_index;
         int v_flags;
  };

Further, some of them are wrongly aligned:

  e.g.

  struct nilfs_cpmode {
        __u64 cm_cno;
        int cm_mode;
  };

The size of wrongly aligned structures varies depending on
architectures, and it breaks the identity of ioctl commands, which
leads to arch dependent errors.

Previously, these are compensated by using compat_ioctl.

This fixes these problems and allows removal of compat ioctl.

Since this will change sizes of those structures, binary compatibility
for the past utilities will once break; new utilities have to be used
instead.  However, it would be helpful to avoid platform dependent
problems in the long term.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: remove timedwait ioctl command
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:51 +0000 (19:01 -0700)] 
nilfs2: remove timedwait ioctl command

This removes NILFS_IOCTL_TIMEDWAIT command from ioctl interface along
with the related flags and wait queue.

The command is terrible because it just sleeps in the ioctl.  I prefer
to avoid this by devising means of event polling in userland program.
By reconsidering the userland GC daemon, I found this is possible
without changing behaviour of the daemon and sacrificing efficiency.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: fix buggy behavior seen in enumerating checkpoints
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:50 +0000 (19:01 -0700)] 
nilfs2: fix buggy behavior seen in enumerating checkpoints

This will fix the weird behavior of lscp command in listing continuously
created checkpoints; the output of lscp is rewinded regularly for the
recent nilfs.  As a result of debugging, a defect was found in
nilfs_cpfile_do_get_cpinfo() function.

Though the function can be repeatedly called to enumerate checkpoints and
it can skip invalid checkpoint entries, the index value was not carried
between successive calls.

The bug has long been present, and came to surface after applying a bugfix
nilfs2-fix-problems-of-memory-allocation-in-ioctl.patch, which increased
frequency of calling the function.  The similar bugfix was already applied
for ``snapshots'' by
nilfs2-fix-gc-failure-on-volumes-keeping-numerous-snapshots.patch.

This fixes the problem by making the index argument bidirectional on the
function.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: clean up indirect function calling conventions
Pekka Enberg [Tue, 7 Apr 2009 02:01:49 +0000 (19:01 -0700)] 
nilfs2: clean up indirect function calling conventions

This cleans up the strange indirect function calling convention used in
nilfs to follow the normal kernel coding style.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: fix improper return values of nilfs_get_cpinfo ioctl
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:48 +0000 (19:01 -0700)] 
nilfs2: fix improper return values of nilfs_get_cpinfo ioctl

A few tool developers gave me requests for fixing inconvenient return
value of nilfs_get_cpinfo() ioctl; if the requested mode is NILFS_SNAPSHOT
and the specified start entry is not a snapshot, the ioctl unnaturally
returns one as the number of acquired snapshot item.

In addition, the ioctl function returns an ENOENT error for checkpoints
within blocks deleted by garbage collection.

These behaviors require corrections for programs which enumerate
snapshots.  This resolves the inconvenience by changing the return values
to zero for the above cases.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: fix gc failure on volumes keeping numerous snapshots
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:47 +0000 (19:01 -0700)] 
nilfs2: fix gc failure on volumes keeping numerous snapshots

This resolves the following failure of nilfs2 cleaner daemon:

 nilfs_cleanerd[20670]: cannot clean segments: No such file or directory
 nilfs_cleanerd[20670]: shutdown

When creating thousands of snapshots, the cleaner daemon had rarely died
as above due to an error returned from the kernel code.

After applying the recent patch which fixed memory allocation problems in
ioctl (Message-Id: <20081215.155840.105124170.ryusuke@osrg.net>), the
problem gets more frequent.

It turned out to be a bug of nilfs_ioctl_wrap_copy function and one of its
callback routines to read out information of snapshots; if the
nilfs_ioctl_wrap_copy function divided a large read request into multiple
requests, the second and later requests have failed since a restart
position on snapshot meta data was not properly set forward.

It's a deficiency of the callback interface that cannot pass the restart
position among multiple requests.  This patch fixes the issue by allowing
nilfs_ioctl_wrap_copy and snapshot read functions to exchange a position
argument.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: add maintainer
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:46 +0000 (19:01 -0700)] 
nilfs2: add maintainer

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: insert explanations in gcinode file
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:45 +0000 (19:01 -0700)] 
nilfs2: insert explanations in gcinode file

The file gcinode.c gives buffer cache functions for on-disk blocks
moved in garbage collection.  Joern Engel has suggested inserting its
explanations in the source file (Message-ID:
<20080917144146.GD8750@logfs.org> and
<20080917224953.GB14644@logfs.org>).

This follows the comment.

Cc: Joern Engel <joern@logfs.org>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: avoid double error caused by nilfs_transaction_end
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:45 +0000 (19:01 -0700)] 
nilfs2: avoid double error caused by nilfs_transaction_end

Pekka Enberg pointed out that double error handlings found after
nilfs_transaction_end() can be avoided by separating abort operation:

 OK, I don't understand this. The only way nilfs_transaction_end() can
 fail is if we have NILFS_TI_SYNC set and we fail to construct the
 segment. But why do we want to construct a segment if we don't commit?

 I guess what I'm asking is why don't we have a separate
 nilfs_transaction_abort() function that can't fail for the erroneous
 case to avoid this double error value tracking thing?

This does the separation and renames nilfs_transaction_end() to
nilfs_transaction_commit() for clarification.

Since, some calls of these functions were used just for exclusion control
against the segment constructor, they are replaced with semaphore
operations.

Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: cleanup nilfs_clear_inode
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:44 +0000 (19:01 -0700)] 
nilfs2: cleanup nilfs_clear_inode

This will remove the following unnecessary locks and cleanup code in
nilfs_clear_inode():

- unnecessary protection using nilfs_transaction_begin() and
  nilfs_transaction_end().

- cleanup code of i_dirty list field which is never chained
  when this function is called.

- spinlock used when releasing i_bh field.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: fix problems of memory allocation in ioctl
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:43 +0000 (19:01 -0700)] 
nilfs2: fix problems of memory allocation in ioctl

This is another patch for fixing the following problems of a memory
copy function in nilfs2 ioctl:

(1) It tries to allocate 128KB size of memory even for small objects.

(2) Though the function repeatedly tries large memory allocations
    while reducing the size, GFP_NOWAIT flag is not specified.
    This increases the possibility of system memory shortage.

(3) During the retries of (2), verbose warnings are printed
    because _GFP_NOWARN flag is not used for the kmalloc calls.

The first patch was still doing large allocations by kmalloc which are
repeatedly tried while reducing the size.

Andi Kleen told me that using copy_from_user for large memory is not
good from the viewpoint of preempt latency:

 On Fri, 12 Dec 2008 21:24:11 +0100, Andi Kleen <andi@firstfloor.org> wrote:
 > > In the current interface, each data item is copied twice: one is to
 > > the allocated memory from user space (via copy_from_user), and another
 >
 > For such large copies it is better to use multiple smaller (e.g. 4K)
 > copy user, that gives better real time preempt latencies. Each cfu has a
 > cond_resched(), but only one, not multiple times in the inner loop.

He also advised me that:

 On Sun, 14 Dec 2008 16:13:27 +0100, Andi Kleen <andi@firstfloor.org> wrote:
 > Better would be if you could go to PAGE_SIZE. order 0 allocations
 > are typically the fastest / least likely to stall.
 >
 > Also in this case it's a good idea to use __get_free_pages()
 > directly, kmalloc tends to be become less efficient at larger
 > sizes.

For the function in question, the size of buffer memory can be reduced
since the buffer is repeatedly used for a number of small objects.  On
the other hand, it may incur large preempt latencies for larger buffer
because a copy_from_user (and a copy_to_user) was applied only once
each cycle.

With that, this revision uses the order 0 allocations with
__get_free_pages() to fix the original problems.

Cc: Andi Kleen <andi@firstfloor.org>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: update makefile and Kconfig
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:41 +0000 (19:01 -0700)] 
nilfs2: update makefile and Kconfig

This adds a Makefile for the nilfs2 file system, and updates the
makefile and Kconfig file in the file system directory.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: ioctl operations
Koji Sato [Tue, 7 Apr 2009 02:01:41 +0000 (19:01 -0700)] 
nilfs2: ioctl operations

This adds userland interface implemented with ioctl.

Signed-off-by: Koji Sato <sato.koji@lab.ntt.co.jp>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: block cache for garbage collection
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:40 +0000 (19:01 -0700)] 
nilfs2: block cache for garbage collection

This adds the cache of on-disk blocks to be moved in garbage
collection.  The disk blocks are held with dummy inodes (called
gcinodes), and this file provides lookup function of the dummy inodes,
and their buffer read function.

Signed-off-by: Seiji Kihara <kihara.seiji@lab.ntt.co.jp>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Yoshiji Amagai <amagai.yoshiji@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: another dat for garbage collection
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:39 +0000 (19:01 -0700)] 
nilfs2: another dat for garbage collection

NILFS2 uses another DAT inode during garbage collection to ensure
atomicity and consistency of the DAT in the transient state.  This
twin inode is called GCDAT.

This adds functions to initialize the GCDAT and to switch page caches
and B-tree node caches between these two inodes.

Signed-off-by: Seiji Kihara <kihara.seiji@lab.ntt.co.jp>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Yoshiji Amagai <amagai.yoshiji@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: recovery functions
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:38 +0000 (19:01 -0700)] 
nilfs2: recovery functions

This adds recovery function on mount.

Usually the recovery is achieved by just finding the latest super
root.  When logs without checkpoints were appended for data sync
operations after the latest super root, the recovery function will
perform roll forwarding and reconstruct new log(s) with a super root.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: fix missed-sync issue for do_sync_mapping_range()
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:38 +0000 (19:01 -0700)] 
nilfs2: fix missed-sync issue for do_sync_mapping_range()

Chris Mason pointed out that there is a missed sync issue in
nilfs_writepages():

On Wed, 17 Dec 2008 21:52:55 -0500, Chris Mason wrote:
> It looks like nilfs_writepage ignores WB_SYNC_NONE, which is used by
> do_sync_mapping_range().

where WB_SYNC_NONE in do_sync_mapping_range() was replaced with
WB_SYNC_ALL by Nick's patch (commit:
ee53a891f47444c53318b98dac947ede963db400).

This fixes the problem by letting nilfs_writepages() write out the log of
file data within the range if sync_mode is WB_SYNC_ALL.

This involves removal of nilfs_file_aio_write() which was previously
needed to ensure O_SYNC sync writes.

Cc: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: segment constructor
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:37 +0000 (19:01 -0700)] 
nilfs2: segment constructor

This adds the segment constructor (also called log writer).

The segment constructor collects dirty buffers for every dirty inode,
makes summaries of the buffers, assigns disk block addresses to the
buffers, and then submits BIOs for the buffers.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: segment buffer
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:36 +0000 (19:01 -0700)] 
nilfs2: segment buffer

This adds the segment buffer which is used to constuct logs.

[akpm@linux-foundation.org: BIO_RW_SYNC got removed]
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: super block operations
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:35 +0000 (19:01 -0700)] 
nilfs2: super block operations

This adds super block operations for the nilfs2 file system.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonilfs2: operations for the_nilfs core object
Ryusuke Konishi [Tue, 7 Apr 2009 02:01:35 +0000 (19:01 -0700)] 
nilfs2: operations for the_nilfs core object

This adds functions on the_nilfs object, which keeps shared resources and
states among a read/write mount and snapshots mounts going individually.

the_nilfs is allocated per block device; it is created when user first
mount a snapshot or a read/write mount on the device, then it is reused
for successive mounts.  It will be freed when all mount instances on the
device are detached.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>