linux-2.6
18 years ago[GFS2] Make newly moved functions static
Steven Whitehouse [Thu, 18 May 2006 18:02:52 +0000 (14:02 -0400)] 
[GFS2] Make newly moved functions static

The functions moved from bits.c can now be made static.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
18 years ago[GFS2] Merge bits.[ch] into rgrp.c
Steven Whitehouse [Thu, 18 May 2006 17:52:39 +0000 (13:52 -0400)] 
[GFS2] Merge bits.[ch] into rgrp.c

Since they are small and will be inlined by the complier,
it makes sense to merge the contents of bits.[ch] into
rgrp.c

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
18 years ago[GFS2] Fix ref count bug that used to bite us on umount
Steven Whitehouse [Tue, 16 May 2006 17:37:11 +0000 (13:37 -0400)] 
[GFS2] Fix ref count bug that used to bite us on umount

The ref count of certain glock's got elevated too far during unlink
which caused umount to fail. This fixes it.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
18 years ago[GFS2] Fix attributes setting logic
Steven Whitehouse [Fri, 12 May 2006 21:07:56 +0000 (17:07 -0400)] 
[GFS2] Fix attributes setting logic

The attributes logic for immutable was wrong so that there was
not way to remove this attribute once set. This fixes the
bug.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
18 years ago[GFS2] Remove incorrect initialisation of gh_owner
Steven Whitehouse [Fri, 12 May 2006 18:06:02 +0000 (14:06 -0400)] 
[GFS2] Remove incorrect initialisation of gh_owner

The gh_owner field shouldn't be set or reset outside the glock code.
These were left over from when recursive locking was allowed. It
isn't any more, so they are not needed.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
18 years ago[GFS2] Reverse block order in build_height
Steven Whitehouse [Fri, 12 May 2006 16:09:15 +0000 (12:09 -0400)] 
[GFS2] Reverse block order in build_height

The original code ordered the blocks allocated in the build_height
routine backwards causing excessive disk seeks during a read of the
metadata. This patch reverses the order to try and reduce disk seeks.

Example: A five level metadata tree, I = Inode, P = Pointers, D = Data

You need to read the blocks in the order:

I P5 P4 P3 P2 P1 D

in order to read a single data block. The new code now orders the blocks
in this way. The old code used to order them as:

I P1 P2 P3 P4 P5 D

requiring two extra seeks on average. Note that for files which are
grown by gradual extension rather than by truncate or by llseek/write
at a large offset, this doesn't apply. In the case of writing to a
file linearly, this routine will only be called upon to extend the
height of the tree by one block at a time, so the ordering is
determined by when its called rather than by the internals of the
routine itself. Optimising that part of the ordering is a much
harder problem.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
18 years agoMerge branch 'master'
Steven Whitehouse [Fri, 12 May 2006 14:48:52 +0000 (10:48 -0400)] 
Merge branch 'master'

18 years agoLinux v2.6.17-rc4 v2.6.17-rc4
Linus Torvalds [Thu, 11 May 2006 23:31:53 +0000 (16:31 -0700)] 
Linux v2.6.17-rc4

18 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Thu, 11 May 2006 22:46:59 +0000 (15:46 -0700)] 
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] 3508/1: Update collie defconfig
  [ARM] Fix thread struct allocator for SMP case
  [ARM] Update mach-types
  [ARM] Update versatile_defconfig

18 years agoMerge master.kernel.org:/home/rmk/linux-2.6-serial
Linus Torvalds [Thu, 11 May 2006 22:46:42 +0000 (15:46 -0700)] 
Merge master.kernel.org:/home/rmk/linux-2.6-serial

* master.kernel.org:/home/rmk/linux-2.6-serial:
  [SERIAL] 8250: add locking to console write function
  [SERIAL] Remove unconditional enable of TX irq for console
  [SERIAL] 8250: set divisor register correctly for AMD Alchemy SoC uart
  [SERIAL] AMD Alchemy UART: claim memory range
  [SERIAL] Clean up serial locking when obtaining a reference to a port

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Thu, 11 May 2006 22:35:54 +0000 (15:35 -0700)] 
Merge /pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [NET_SCHED]: HFSC: fix thinko in hfsc_adjust_levels()
  [IPV6]: skb leakage in inet6_csk_xmit
  [BRIDGE]: Do sysfs registration inside rtnl.
  [NET]: Do sysfs registration as part of register_netdevice.
  [TG3]: Fix possible NULL deref in tg3_run_loopback().
  [NET] linkwatch: Handle jiffies wrap-around
  [IRDA]: Switching to a workqueue for the SIR work
  [IRDA]: smsc-ircc: Minimal hotplug support.
  [IRDA]: Removing unused EXPORT_SYMBOLs
  [IRDA]: New maintainer.
  [NET]: Make netdev_chain a raw notifier.
  [IPV4]: ip_options_fragment() has no effect on fragmentation
  [NET]: Add missing operstates documentation.

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6
Linus Torvalds [Thu, 11 May 2006 22:34:33 +0000 (15:34 -0700)] 
Merge /linux/kernel/git/jejb/scsi-rc-fixes-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6: (25 commits)
  [SCSI] mptfc: race between mptfc_register_dev and mptfc_target_alloc
  [SCSI] lpfc 8.1.6 : Fix Data Corruption in Bus Reset Path
  [SCSI] mptspi: revalidate negotiation parameters after host reset and resume
  [SCSI] srp.h: avoid padding of structs
  [SCSI] ibmvscsi: fix leak when failing to send srp event
  [SCSI] qla2xxx: Correct eh_abort recovery logic.
  [SCSI] megaraid_{mm,mbox}: fix a bug in reset handler
  [SCSI] fusion - bug fix stack overflow in mptbase
  [SCSI] scsi: Add IBM 2104-DU3 to blist
  [SCSI] Fix DVD burning issues.
  [SCSI] SCSI: aic7xxx_osm_pci resource leak fix.
  [SCSI] - fusion - mptfc bug fix's to prevent deadlock situations
  [SCSI] mptfusion: bug fix's for raid components adding/deleting
  [SCSI] aic7xxx: ahc_pci_write_config() fix
  [SCSI] megaraid: unused variable
  [SCSI] qla2xxx: only free_irq() after request_irq() succeeds
  [SCSI] Overrun in drivers/scsi/sim710.c
  [SCSI] lpfc 8.1.5 : Change version number to 8.1.5
  [SCSI] lpfc 8.1.5 : Misc small fixes
  [SCSI] lpfc 8.1.5 : Additional fixes to LOGO, PLOGI, and RSCN processing
  ...

18 years ago[BLOCK] limit request_fn recursion
Jens Axboe [Thu, 11 May 2006 06:20:16 +0000 (08:20 +0200)] 
[BLOCK] limit request_fn recursion

Don't recurse back into the driver even if the unplug threshold is met,
when the driver asks for a requeue. This is both silly from a logical
point of view (requeues typically happen due to driver/hardware
shortage), and also dangerous since we could hit an endless request_fn
-> requeue -> unplug -> request_fn loop and crash on stack overrun.

Also limit blk_run_queue() to one level of recursion, similar to how
blk_start_queue() works.

This patch fixed a real problem with SLES10 and lpfc, and it could hit
any SCSI lld that returns non-zero from it's ->queuecommand() handler.

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[NET_SCHED]: HFSC: fix thinko in hfsc_adjust_levels()
Patrick McHardy [Thu, 11 May 2006 19:22:03 +0000 (12:22 -0700)] 
[NET_SCHED]: HFSC: fix thinko in hfsc_adjust_levels()

When deleting the last child the level of a class should drop to zero.

Noticed by Andreas Mueller <andreas@stapelspeicher.org>

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years agoptrace_attach: fix possible deadlock schenario with irqs
Linus Torvalds [Thu, 11 May 2006 18:08:49 +0000 (11:08 -0700)] 
ptrace_attach: fix possible deadlock schenario with irqs

Eric Biederman points out that we can't take the task_lock while holding
tasklist_lock for writing, because another CPU that holds the task lock
might take an interrupt that then tries to take tasklist_lock for writing.

Which would be a nasty deadlock, with one CPU spinning forever in an
interrupt handler (although admittedly you need to really work at
triggering it ;)

Since the ptrace_attach() code is special and very unusual, just make it
be extra careful, and use trylock+repeat to avoid the possible deadlock.

Cc: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Wed, 10 May 2006 22:37:47 +0000 (15:37 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  IB/mthca: FMR ioremap fix
  IPoIB: Free child interfaces properly
  IB/mthca: Fix race in reference counting
  IB/srp: Fix tracking of pending requests during error handling
  IB: Fix display of 4-bit port counters in sysfs

18 years agoIB/mthca: FMR ioremap fix
Michael S. Tsirkin [Wed, 10 May 2006 14:58:41 +0000 (17:58 +0300)] 
IB/mthca: FMR ioremap fix

Addresses for ioremap must be calculated off of pci_resource_start;
we can't directly use the bus address as seen by the HCA.  Fix the
code that remaps device memory for FMR access.

Based on patch by Klaus Smolin.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
18 years agoMerge branch 'upstream' of master.kernel.org:/pub/scm/linux/kernel/git/shemminger...
Linus Torvalds [Wed, 10 May 2006 21:59:29 +0000 (14:59 -0700)] 
Merge branch 'upstream' of /linux/kernel/git/shemminger/netdev-2.6

* 'upstream' of master.kernel.org:/pub/scm/linux/kernel/git/shemminger/netdev-2.6:
  sis900: phy for FoxCon motherboard
  dl2k: use DMA_48BIT_MASK constant
  phy: mdiobus_register(): initialize all phy_map entries
  sky2: ifdown kills irq mask

18 years agosis900: phy for FoxCon motherboard
James Cameron [Wed, 10 May 2006 20:33:29 +0000 (13:33 -0700)] 
sis900: phy for FoxCon motherboard

661FX7MI-S motherboard which uses the SiS 661FX chipset.  The patch adds
an entry to mii_chip_info for the transceiver.

The PHY ids were found using the sis900_c_122.diff patch from
http://brownhat.org/sis900.html but that patch didn't solve the problem,
because the PHY at address 1 was already being chosen.

Without my patch, when bursts of packets arrive from other hosts on a
LAN, the interface dropped one roughly 10% of the time, causing
retransmits.  There were fifth second pauses in refresh of large xterms,
and it made Netrek suck.  I can provide further test data.

Workaround in lieu of patch is to use mii-tool to advertise
100baseTx-HD, then force renegotiation.

I wasn't able to identify the actual transceiver, so the description
field is a guess.

This patch is similar to Artur Skawina's patch:
http://marc.theaimsgroup.com/?l=linux-netdev&m=114297516729079&w=2

I'm not sure, but I wonder if it means the default behaviour should be
changed, so as to better handle future transceivers.

Diff is against 2.6.16.13.

Signed-off-by: James Cameron <james.cameron@hp.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agodl2k: use DMA_48BIT_MASK constant
Francois Romieu [Wed, 10 May 2006 19:48:57 +0000 (12:48 -0700)] 
dl2k: use DMA_48BIT_MASK constant

Typo will be harder with this one.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agophy: mdiobus_register(): initialize all phy_map entries
Herbert Valerio Riedel [Wed, 10 May 2006 16:12:57 +0000 (12:12 -0400)] 
phy: mdiobus_register(): initialize all phy_map entries

make sure phy_map entries whose PHY address is masked are initialized
to NULL, given that other code (such as mdiobus_unregister for
instance) assumes that non-NULL phy_map entries are allocated
phy_devices

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agosky2: ifdown kills irq mask
Stephen Hemminger [Tue, 9 May 2006 21:46:54 +0000 (14:46 -0700)] 
sky2: ifdown kills irq mask

Bringing down a port also masks off the status and other IRQ's
needed for device to function due to missing paren's.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years ago[Cardman 40x0] Fix udev device creation
Harald Welte [Wed, 10 May 2006 11:28:52 +0000 (13:28 +0200)] 
[Cardman 40x0] Fix udev device creation

This patch corrects the order of the calls to register_chrdev() and
pcmcia_register_driver().  Now udev correctly creates userspace device
files /dev/cmmN and /dev/cmxN respectively.

Based on an earlier patch by Jan Niehusmann <jan@gondor.com>.

Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[IPV6]: skb leakage in inet6_csk_xmit
Alexey Kuznetsov [Wed, 10 May 2006 20:24:38 +0000 (13:24 -0700)] 
[IPV6]: skb leakage in inet6_csk_xmit

inet6_csk_xit does not free skb when routing fails.

Signed-off-by: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[BRIDGE]: Do sysfs registration inside rtnl.
Stephen Hemminger [Wed, 10 May 2006 20:21:53 +0000 (13:21 -0700)] 
[BRIDGE]: Do sysfs registration inside rtnl.

Now that netdevice sysfs registration is done as part of
register_netdevice; bridge code no longer has to be tricky when adding
it's kobjects to bridges.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[NET]: Do sysfs registration as part of register_netdevice.
Stephen Hemminger [Wed, 10 May 2006 20:21:17 +0000 (13:21 -0700)] 
[NET]: Do sysfs registration as part of register_netdevice.

The last step of netdevice registration was being done by a delayed
call, but because it was delayed, it was impossible to return any error
code if the class_device registration failed.

Side effects:
 * one state in registration process is unnecessary.
 * register_netdevice can sleep inside class_device registration/hotplug
 * code in netdev_run_todo only does unregistration so it is simpler.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[SCSI] mptfc: race between mptfc_register_dev and mptfc_target_alloc
mdr@sgi.com [Mon, 1 May 2006 18:07:04 +0000 (13:07 -0500)] 
[SCSI] mptfc: race between mptfc_register_dev and mptfc_target_alloc

A race condition exists in mptfc between the thread registering a device
with the fc transport and the scan work generated by the transport.
This race existed prior to the application of the mptfc bug fix patch.

mptfc_register_dev() calls fc_remote_port_add() with the FC_RPORT_ROLE_TARGET
bit set in the rport ids passed to the function.  Having this bit set causes
fc_remote_port_add() to schedule a scan of the device.

This scan can execute before mptfc_register_dev() can fill in the dd_data
in the rport structure.  When this happens, mptfc_target_alloc() will fail
because dd_data is null.

Attached is a patch which fixes the problem.  The patch changes the rport ids
passed to fc_remote_port_add() to not have the TARGET bit set.  This prevents
the scan from being scheduled.  After mptfc_register_dev() fills in the rport
dd_data field, fc_remote_port_rolechg() is called, changing the role of the
rport to TARGET.  Thus, the scan is scheduled after dd_data is filled
in which prevents the failure in mptfc_target_alloc().

Signed-off-by: Michael Reed <mdr@sgi.com>
Signed-off-by: Eric Moore <Eric.Moore@lsil.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[TG3]: Fix possible NULL deref in tg3_run_loopback().
Jesper Juhl [Wed, 10 May 2006 06:14:35 +0000 (23:14 -0700)] 
[TG3]: Fix possible NULL deref in tg3_run_loopback().

tg3_run_loopback doesn't check that dev_alloc_skb() returns anything
useful.

Even if dev_alloc_skb() fails to return an skb to us we'll happily go
on and assume it did, so we risk dereferencing a NULL pointer.  Much
better to fail gracefully by returning -ENOMEM than crashing here.

Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years agoIPoIB: Free child interfaces properly
Roland Dreier [Wed, 10 May 2006 05:54:59 +0000 (22:54 -0700)] 
IPoIB: Free child interfaces properly

When deleting a child interface with a non-default P_Key via
/sys/class/net/ibX/delete_child, the interface must be freed with
free_netdev() (rather than kfree() on the private data).

Signed-off-by: Roland Dreier <rolandd@cisco.com>
18 years ago[NET] linkwatch: Handle jiffies wrap-around
Herbert Xu [Tue, 9 May 2006 22:27:54 +0000 (15:27 -0700)] 
[NET] linkwatch: Handle jiffies wrap-around

The test used in the linkwatch does not handle wrap-arounds correctly.
Since the intention of the code is to eliminate bursts of messages we
can afford to delay things up to a second.  Using that fact we can
easily handle wrap-arounds by making sure that we don't delay things
by more than one second.

This is based on diagnosis and a patch by Stefan Rompf.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Stefan Rompf <stefan@loplof.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[IRDA]: Switching to a workqueue for the SIR work
Christoph Hellwig [Tue, 9 May 2006 22:27:04 +0000 (15:27 -0700)] 
[IRDA]: Switching to a workqueue for the SIR work

Since sir_kthread.c pretty much duplicates the workqueue
functionality, we'd better switch.  The SIR fsm has been merged into
sir_dev.c and thus sir_kthread.c is deleted.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Samuel Ortiz <samuel.ortiz@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[IRDA]: smsc-ircc: Minimal hotplug support.
David Brownell [Tue, 9 May 2006 22:26:11 +0000 (15:26 -0700)] 
[IRDA]: smsc-ircc: Minimal hotplug support.

Minimal PNP hotplug support for the smsc-ircc2 driver.  A modular
driver will be modprobed via hotplug, but still bypasses driver model
probing.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Samuel Ortiz <samuel.ortiz@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[IRDA]: Removing unused EXPORT_SYMBOLs
Adrian Bunk [Tue, 9 May 2006 22:25:25 +0000 (15:25 -0700)] 
[IRDA]: Removing unused EXPORT_SYMBOLs

This patch removes the following unused EXPORT_SYMBOL's:
- irias_find_attrib
- irias_new_string_value
- irias_new_octseq_value

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Samuel Ortiz <samuel.ortiz@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[IRDA]: New maintainer.
Samuel Ortiz [Tue, 9 May 2006 22:24:49 +0000 (15:24 -0700)] 
[IRDA]: New maintainer.

As agreed with Jean Tourrilhes, I am taking over IrDA maintainership.

Signed-off-by: Samuel Ortiz <samuel.ortiz@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[NET]: Make netdev_chain a raw notifier.
Alan Stern [Tue, 9 May 2006 22:23:03 +0000 (15:23 -0700)] 
[NET]: Make netdev_chain a raw notifier.

From: Alan Stern <stern@rowland.harvard.edu>

This chain does it's own locking via the RTNL semaphore, and
can also run recursively so adding a new mutex here was causing
deadlocks.

Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[IPV4]: ip_options_fragment() has no effect on fragmentation
Wei Yongjun [Tue, 9 May 2006 22:18:50 +0000 (15:18 -0700)] 
[IPV4]: ip_options_fragment() has no effect on fragmentation

Fix error point to options in ip_options_fragment(). optptr get a
error pointer to the ipv4 header, correct is pointer to ipv4 options.

Signed-off-by: Wei Yongjun <weiyj@soft.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[NET]: Add missing operstates documentation.
Stefan Rompf [Tue, 9 May 2006 22:15:35 +0000 (15:15 -0700)] 
[NET]: Add missing operstates documentation.

Signed-off-by: Stefan Rompf <stefan@loplof.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[ARM] 3508/1: Update collie defconfig
Pavel Machek [Tue, 9 May 2006 21:27:51 +0000 (22:27 +0100)] 
[ARM] 3508/1: Update collie defconfig

Patch from Pavel Machek

Update collie defconfig to something that can bring closer-to-working
system to its user.

Signed-off-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[ARM] Fix thread struct allocator for SMP case
Russell King [Tue, 9 May 2006 21:14:28 +0000 (22:14 +0100)] 
[ARM] Fix thread struct allocator for SMP case

The ARM thread struct allocator is racy on SMP systems.  Fix it by
turning it into a per-cpu based allocator.  This also allows keeps
the cache cache warm for thread structs and kernel stacks.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years agoIB/mthca: Fix race in reference counting
Roland Dreier [Tue, 9 May 2006 17:50:29 +0000 (10:50 -0700)] 
IB/mthca: Fix race in reference counting

Fix races in in destroying various objects.  If a destroy routine
waits for an object to become free by doing

wait_event(&obj->wait, !atomic_read(&obj->refcount));
/* now clean up and destroy the object */

and another place drops a reference to the object by doing

if (atomic_dec_and_test(&obj->refcount))
wake_up(&obj->wait);

then this is susceptible to a race where the wait_event() and final
freeing of the object occur between the atomic_dec_and_test() and the
wake_up().  And this is a use-after-free, since wake_up() will be
called on part of the already-freed object.

Fix this in mthca by replacing the atomic_t refcounts with plain old
integers protected by a spinlock.  This makes it possible to do the
decrement of the reference count and the wake_up() so that it appears
as a single atomic operation to the code waiting on the wait queue.

While touching this code, also simplify mthca_cq_clean(): the CQ being
cleaned cannot go away, because it still has a QP attached to it.  So
there's no reason to be paranoid and look up the CQ by number; it's
perfectly safe to use the pointer that the callers already have.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
18 years agoIB/srp: Fix tracking of pending requests during error handling
Roland Dreier [Tue, 9 May 2006 17:50:28 +0000 (10:50 -0700)] 
IB/srp: Fix tracking of pending requests during error handling

If a SCSI abort completes, or the command completes successfully, then
the driver must remove the command from its queue of pending
commands.  Similarly, if a device reset succeeds, then all commands
queued for the given device must be removed from the queue.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
18 years agoIB: Fix display of 4-bit port counters in sysfs
Ralph Campbell [Tue, 9 May 2006 17:50:28 +0000 (10:50 -0700)] 
IB: Fix display of 4-bit port counters in sysfs

The code to display local_link_integrity_errors and
excessive_buffer_overrun_errors in
/sys/class/infiniband/<hca>/ports/<n>/counters/
uses the wrong shift to extract the 4 bit values.

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
18 years agoMerge branch 'upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/shemminger...
Linus Torvalds [Tue, 9 May 2006 17:18:35 +0000 (10:18 -0700)] 
Merge branch 'upstream' of git://git./linux/kernel/git/shemminger/netdev-2.6

* 'upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/shemminger/netdev-2.6:
  [PATCH] bcm43xx: Fix access to non-existent PHY registers
  [PATCH] bcm43xx: Fix array overrun in bcm43xx_geo_init
  [PATCH] bcm43xx: check for valid MAC address in SPROM
  [PATCH] ieee80211: Fix A band channel count (resent)
  [PATCH] bcm43xx: fix iwmode crash when down
  [PATCH] softmac: make non-operational after being stopped
  [PATCH] softmac: don't reassociate if user asked for deauthentication
  spidernet: enable support for bcm5461 ethernet phy
  spidernet: introduce new setting
  Fix RTL8019AS init for Toshiba RBTX49xx boards
  au1000_eth.c: use ether_crc() from <linux/crc32.h>
  sky2: version 1.3
  Add more support for the Yukon Ultra chip found in dual core centino laptops.
  sky2: synchronize irq on remove
  sky2: dont write status ring
  sky2: edge triggered workaround enhancement
  sky2: use mask instead of modulo operation
  sky2: tx ring index mask fix
  sky2: status irq hang fix
  sky2: backout NAPI reschedule

18 years ago[PATCH] ppc32/8xx: Fix r3 trashing due to 8MB TLB page instantiation
Marcelo Tosatti [Fri, 5 May 2006 20:09:29 +0000 (17:09 -0300)] 
[PATCH] ppc32/8xx: Fix r3 trashing due to 8MB TLB page instantiation

Instantiation of 8MB pages on the TLB cache for the kernel static
mapping trashes r3 register on !CONFIG_8xx_CPU6 configurations.
This ensures r3 gets saved and restored.

Signed-off-by: Marcelo Tosatti <marcelo@kvack.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years agopowerpc/32: Define an is_kernel_addr() to fix ARCH=ppc compilation
Paul Mackerras [Tue, 9 May 2006 06:00:59 +0000 (16:00 +1000)] 
powerpc/32: Define an is_kernel_addr() to fix ARCH=ppc compilation

My commit 6bfd93c32a5065d0e858780b3beb0b667081601c broke the ARCH=ppc
compilation by using the is_kernel_addr() macro in asm/uaccess.h.
This fixes it by defining a suitable is_kernel_addr() for ARCH=ppc.

Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years agoMerge git://oss.sgi.com:8090/xfs-2.6
Linus Torvalds [Tue, 9 May 2006 00:41:05 +0000 (17:41 -0700)] 
Merge git://oss.sgi.com:8090/xfs-2.6

* git://oss.sgi.com:8090/xfs-2.6:
  [XFS] Fix a possible metadata buffer (AGFL) refcount leak when fixing an
  [XFS] Fix a project quota space accounting leak on rename.
  [XFS] Fix a possible forced shutdown due to mishandling write barriers

18 years agoMerge branch 'upstream-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/linvil...
Stephen Hemminger [Mon, 8 May 2006 23:01:20 +0000 (16:01 -0700)] 
Merge branch 'upstream-fixes' of git://git./linux/kernel/git/linville/wireless-2.6

18 years agospidernet: enable support for bcm5461 ethernet phy
Jens Osterkamp [Thu, 4 May 2006 09:59:41 +0000 (05:59 -0400)] 
spidernet: enable support for bcm5461 ethernet phy

A newer board revision changed the type of ethernet phy.
Moreover, this generalizes the way that a phy gets switched
into fiber mode when autodetection is not available.

Signed-off-by: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agospidernet: introduce new setting
Jens Osterkamp [Thu, 4 May 2006 09:59:56 +0000 (05:59 -0400)] 
spidernet: introduce new setting

We found a new chip setting that we need in order
to make the driver work more reliable.

Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agoFix RTL8019AS init for Toshiba RBTX49xx boards
Sergei Shtylyov [Mon, 8 May 2006 20:58:28 +0000 (00:58 +0400)] 
Fix RTL8019AS init for Toshiba RBTX49xx boards

    Ensure that 8-bit mode is selected for the on-board Realtek RTL8019AS chip
on Toshiba RBHMA4x00, get rid of the duplicate #ifdef's when setting
ei_status.word16.
    The chip's datasheet says that the PSTOP register shouldn't exceed 0x60 in
8-bit mode -- ensure this too.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agoau1000_eth.c: use ether_crc() from <linux/crc32.h>
Herbert Valerio Riedel [Mon, 1 May 2006 13:37:09 +0000 (15:37 +0200)] 
au1000_eth.c: use ether_crc() from <linux/crc32.h>

since the au1000 driver already selects the CRC32 routines, simply replace
the internal ether_crc() implementation with the semantically equivalent
one from <linux/crc32.h>

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agosky2: version 1.3
Stephen Hemminger [Mon, 8 May 2006 22:11:34 +0000 (15:11 -0700)] 
sky2: version 1.3

Update version number, to track changes.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agoAdd more support for the Yukon Ultra chip found in dual core centino laptops.
Stephen Hemminger [Mon, 8 May 2006 22:11:33 +0000 (15:11 -0700)] 
Add more support for the Yukon Ultra chip found in dual core centino laptops.

The newest Yukon Ultra chipset's require more special tweaks.
They seem to be like the Yukon XL chipsets. This code is transliterated
from the latest SysKonnect driver; I don't have any Ultra hardware.

Signed-off-by: Stephe Hemminger <shemminger@osdl.org>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agosky2: synchronize irq on remove
Stephen Hemminger [Mon, 8 May 2006 22:11:32 +0000 (15:11 -0700)] 
sky2: synchronize irq on remove

Need to make sure interrupt is not racing with unregister of
network device.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agosky2: dont write status ring
Stephen Hemminger [Mon, 8 May 2006 22:11:31 +0000 (15:11 -0700)] 
sky2: dont write status ring

It is more efficient not to write the status ring from the
processor and just read the active portion.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agosky2: edge triggered workaround enhancement
Stephen Hemminger [Mon, 8 May 2006 22:11:30 +0000 (15:11 -0700)] 
sky2: edge triggered workaround enhancement

Need to make the edge-triggered workaround timer faster to get marginally
better peformance. The test_and_set_bit in schedule_prep() acts as a barrier
already. Make it a module parameter so that laptops who are concerned
about power can set it to 0; and user's stuck with broken BIOS's
can turn the driver into pure polling.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agosky2: use mask instead of modulo operation
Stephen Hemminger [Mon, 8 May 2006 22:11:29 +0000 (15:11 -0700)] 
sky2: use mask instead of modulo operation

Gcc isn't smart enough to know that it can do a modulo
operation with power of 2 constant by doing a mask.
So add macro to do it for us.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agosky2: tx ring index mask fix
Stephen Hemminger [Mon, 8 May 2006 22:11:28 +0000 (15:11 -0700)] 
sky2: tx ring index mask fix

Mask for transmit ring status was picking up bits from the
unused sync ring.  They were always zero, so far...
Also, make sure to remind self not to make tx ring too big.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agosky2: status irq hang fix
Stephen Hemminger [Mon, 8 May 2006 22:11:27 +0000 (15:11 -0700)] 
sky2: status irq hang fix

The status interrupt flag should be cleared before processing,
not afterwards to avoid race. Need to process in poll routine
even if no new interrupt status. This is a normal occurrence when
more than 64 frames (NAPI weight) are processed in one poll routine.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agosky2: backout NAPI reschedule
Stephen Hemminger [Mon, 8 May 2006 22:11:26 +0000 (15:11 -0700)] 
sky2: backout NAPI reschedule

This is a backout of earlier patch.

The whole rescheduling hack was a bad idea. It doesn't really solve
the problem and it makes the code more complicated for no good reason.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
18 years agoRevert "kbuild: fix modpost segfault for 64bit mipsel kernel"
Linus Torvalds [Mon, 8 May 2006 20:38:42 +0000 (13:38 -0700)] 
Revert "kbuild: fix modpost segfault for 64bit mipsel kernel"

This reverts commit c8d8b837ebe4b4f11e1b0c4a2bdc358c697692ed, which
caused problems for the x86 build. Quoth Sam:

  "It was discussed on mips list but apparently the fix was bogus.  I
   will not have time to look into it so mips can carry this local fix
   until we get a proper fix in mainline."

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[ARM] Update mach-types
Russell King [Mon, 8 May 2006 19:31:11 +0000 (20:31 +0100)] 
[ARM] Update mach-types

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[ARM] Update versatile_defconfig
Russell King [Mon, 8 May 2006 19:30:24 +0000 (20:30 +0100)] 
[ARM] Update versatile_defconfig

Update versatile default configuration, enabling the AACI sound driver,
VFP and Versatile AB support.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[PATCH] x86_64: Move ondemand timer into own work queue
Andi Kleen [Mon, 8 May 2006 13:17:31 +0000 (15:17 +0200)] 
[PATCH] x86_64: Move ondemand timer into own work queue

Taking the cpu hotplug semaphore in a normal events workqueue
is unsafe because other tasks can wait for any workqueues with
it hold. This results in a deadlock.

Move the DBS timer into its own work queue which is not
affected by other work queue flushes to avoid this.

Has been acked by Venkatesh.

Cc: venkatesh.pallipadi@intel.com
Cc: cpufreq@lists.linux.org.uk
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86_64: Avoid EBDA area in early boot allocator
Andi Kleen [Mon, 8 May 2006 13:17:28 +0000 (15:17 +0200)] 
[PATCH] x86_64: Avoid EBDA area in early boot allocator

Based on analysis&patch from Robert Hentosch

Observed on a Dell PE6850 with 16GB

The problem occurs very early on, when the kernel allocates space for the
temporary memory map called bootmap. The bootmap overlaps the EBDA region.
EBDA region is not historically reserved in the e820 mapping. When the
bootmap is freed it marks the EBDA region as usable.

If you notice in setup.c there is already code to work around the EBDA
in reserve_ebda_region(), this check however occurs after the bootmap
is allocated and doesn't prevent the bootmap from using this range.

AK: I redid the original patch. Thanks also to Jan Beulich for
spotting some mistakes.

Cc: Robert_Hentosch@dell.com
Cc: jbeulich@novell.com
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86_64: add nmi_exit to die_nmi
Corey Minyard [Mon, 8 May 2006 13:17:25 +0000 (15:17 +0200)] 
[PATCH] x86_64: add nmi_exit to die_nmi

Playing with NMI watchdog on x86_64, I discovered that it didn't
do what I expected.  It always panic-ed, even when it didn't
happen from interrupt context.  This patch solves that
problem for me.  Also, in this case, do_exit() will be called
with interrupts disabled, I believe.  Would it be wise to also
call local_irq_enable() after nmi_exit()?
[Yes I added it -AK]

Currently, on x86_64, any NMI watchdog timeout will cause a panic
because the irq count will always be set to be in an interrupt
when do_exit() is called from die_nmi().  If we add nmi_exit() to
the die_nmi() call (since the nmi will never exit "normally")
it seems to solve this problem.  The following small program
can be used to trigger the NMI watchdog to reproduce this:
  main ()
  {
        iopl(3);
        for (;;) asm("cli");
  }

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86_64: fix die_lock nesting
Corey Minyard [Mon, 8 May 2006 13:17:22 +0000 (15:17 +0200)] 
[PATCH] x86_64: fix die_lock nesting

I noticed this when poking around in this area.

The oops_begin() function in x86_64 would only conditionally claim
the die_lock if the call is nested, but oops_end() would always
release the spinlock. This patch adds a nest count for the die lock
so that the release of the lock is only done on the final oops_end().

Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86_64: Check for too many northbridges in IOMMU code
Andi Kleen [Mon, 8 May 2006 13:17:19 +0000 (15:17 +0200)] 
[PATCH] x86_64: Check for too many northbridges in IOMMU code

The IOMMU code can only deal with 8 northbridges. Error out when
more are found.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86_64: avoid IRQ0 ioapic pin collision
Kimball Murray [Mon, 8 May 2006 13:17:16 +0000 (15:17 +0200)] 
[PATCH] x86_64: avoid IRQ0 ioapic pin collision

The patch addresses a problem with ACPI SCI interrupt entry, which gets
re-used, and the IRQ is assigned to another unrelated device.  The patch
corrects the code such that SCI IRQ is skipped and duplicate entry is
avoided.  Second issue came up with VIA chipset, the problem was caused by
original patch assigning IRQs starting 16 and up.  The VIA chipset uses
4-bit IRQ register for internal interrupt routing, and therefore cannot
handle IRQ numbers assigned to its devices.  The patch corrects this
problem by allowing PCI IRQs below 16.

Cc: len.brown@intel.com
Signed-off by: Natalie Protasevich <Natalie.Protasevich@unisys.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6
Linus Torvalds [Mon, 8 May 2006 16:28:35 +0000 (09:28 -0700)] 
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] remove asm-ia64/bitops.h self-inclusion
  [IA64] strcpy returns NULL pointer and not destination pointer

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Mon, 8 May 2006 16:11:10 +0000 (09:11 -0700)] 
Merge /pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [IRDA] irda-usb: use NULL instead of 0
  [IPV4]: Remove likely in ip_rcv_finish()
  [NET]: Create netdev attribute_groups with class_device_add
  [CLASS DEVICE]: add attribute_group creation

18 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild
Linus Torvalds [Mon, 8 May 2006 16:10:44 +0000 (09:10 -0700)] 
Merge git://git./linux/kernel/git/sam/kbuild

* git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild:
  kbuild: Do not overwrite makefile as anohter user
  kbuild: drivers/video/logo/ - fix ident glitch
  kbuild: fix gen_initramfs_list.sh
  kbuild modpost - relax driver data name
  kbuild: removing .tmp_versions considered harmful
  kbuild: fix modpost segfault for 64bit mipsel kernel

18 years ago[PATCH] fs/locks.c: Fix lease_init
Trond Myklebust [Mon, 8 May 2006 03:02:42 +0000 (23:02 -0400)] 
[PATCH] fs/locks.c: Fix lease_init

It is insane to be giving lease_init() the task of freeing the lock it is
supposed to initialise, given that the lock is not guaranteed to be
allocated on the stack. This causes lockups in fcntl_setlease().
Problem diagnosed by Daniel Hokka Zakrisson <daniel@hozac.com>

Also fix a slab leak in __setlease() due to an uninitialised return value.
Problem diagnosed by Björn Steinbrink.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Tested-by: Daniel Hokka Zakrisson <daniel@hozac.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[XFS] Fix a possible metadata buffer (AGFL) refcount leak when fixing an
Nathan Scott [Mon, 8 May 2006 09:51:58 +0000 (19:51 +1000)] 
[XFS] Fix a possible metadata buffer (AGFL) refcount leak when fixing an
AG freelist.

SGI-PV: 952681
SGI-Modid: xfs-linux-melb:xfs-kern:25902a

Signed-off-by: Nathan Scott <nathans@sgi.com>
18 years ago[XFS] Fix a project quota space accounting leak on rename.
Nathan Scott [Mon, 8 May 2006 09:51:42 +0000 (19:51 +1000)] 
[XFS] Fix a project quota space accounting leak on rename.

SGI-PV: 951636
SGI-Modid: xfs-linux-melb:xfs-kern:25811a

Signed-off-by: Nathan Scott <nathans@sgi.com>
18 years ago[XFS] Fix a possible forced shutdown due to mishandling write barriers
Nathan Scott [Mon, 8 May 2006 09:51:28 +0000 (19:51 +1000)] 
[XFS] Fix a possible forced shutdown due to mishandling write barriers
with remount,ro.

SGI-PV: 951944
SGI-Modid: xfs-linux-melb:xfs-kern:25742a

Signed-off-by: Nathan Scott <nathans@sgi.com>
18 years agokbuild: Do not overwrite makefile as anohter user
Jan Beulich [Tue, 2 May 2006 10:33:20 +0000 (12:33 +0200)] 
kbuild: Do not overwrite makefile as anohter user

Change the conditional of the outputmakefile rule to be evaluated entirely
in make, and add a conditional to not touch the generated makefile when e.g.
running 'make install' as root while the build was done as non-root. Also
adjust the comment describing this, and move the message printing and
redirection to mkmakefile.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
18 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Mon, 8 May 2006 00:32:57 +0000 (17:32 -0700)] 
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] 3507/1: Replace map_desc.physical with map_desc.pfn: aaed2000
  [ARM] 3506/1: aaec2000: debug-macro.S needs hardware.h
  [ARM] 3505/1: aaec2000: entry-macro.S needs asm/arch/irqs.h
  [ARM] 3504/1: Fix clcd includes for aaec2000
  [ARM] 3503/1: Fix map_desc structure for aaec2000
  [ARM] 3501/1: i.MX: fix lowlevel debug macros
  [ARM] rtc-sa1100: fix compiler warnings and error cleanup
  [ARM] Allow SA1100 RTC alarm to be configured for wakeup

18 years ago[ARM] 3507/1: Replace map_desc.physical with map_desc.pfn: aaed2000
Bellido Nicolas [Sun, 7 May 2006 21:49:24 +0000 (22:49 +0100)] 
[ARM] 3507/1: Replace map_desc.physical with map_desc.pfn: aaed2000

Patch from Bellido Nicolas

aaed2000 map_desc.pfn conversion

Signed-off-by: Nicolas Bellido <ml@acolin.be>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[ARM] 3506/1: aaec2000: debug-macro.S needs hardware.h
Bellido Nicolas [Sun, 7 May 2006 21:49:23 +0000 (22:49 +0100)] 
[ARM] 3506/1: aaec2000: debug-macro.S needs hardware.h

Patch from Bellido Nicolas

Include hardware.h in debug-macro.S, otherwise io_p2v is undefined.

Signed-off-by: Nicolas Bellido <ml@acolin.be>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[ARM] 3505/1: aaec2000: entry-macro.S needs asm/arch/irqs.h
Bellido Nicolas [Sun, 7 May 2006 21:49:22 +0000 (22:49 +0100)] 
[ARM] 3505/1: aaec2000: entry-macro.S needs asm/arch/irqs.h

Patch from Bellido Nicolas

Since git commit 2b78838842346da390e8547cd37035184376d506, entry-macro.S needs to include asm/arch/irqs.h

Signed-off-by: Nicolas Bellido <ml@acolin.be>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[ARM] 3504/1: Fix clcd includes for aaec2000
Bellido Nicolas [Sun, 7 May 2006 21:49:21 +0000 (22:49 +0100)] 
[ARM] 3504/1: Fix clcd includes for aaec2000

Patch from Bellido Nicolas

Since this patch:
 [ARM] 3366/1: Allow the 16bpp mode configuration in the CLCD control register

linux/amba/bus.h needs to be included before linux/amba/clcd.h

Signed-off-by: Nicolas Bellido <ml@acolin.be>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[ARM] 3503/1: Fix map_desc structure for aaec2000
Bellido Nicolas [Sun, 7 May 2006 21:49:21 +0000 (22:49 +0100)] 
[ARM] 3503/1: Fix map_desc structure for aaec2000

Patch from Bellido Nicolas

Patch:
 [ARM] 2982/1: Replace map_desc.physical with map_desc.pfn: aaec2000
incorrectly expanded the struct map_desc for aaec2000.

Signed-off-by: Nicolas Bellido <ml@acolin.be>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[ARM] 3501/1: i.MX: fix lowlevel debug macros
Sascha Hauer [Sun, 7 May 2006 17:56:27 +0000 (18:56 +0100)] 
[ARM] 3501/1: i.MX: fix lowlevel debug macros

Patch from Sascha Hauer

This patch fixes the addruart macro to work with both mmu enabled and
disabled.

Signed-off-by: Sascha Hauer <s.hauer@pengutonix.de>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years agoFix ptrace_attach()/ptrace_traceme()/de_thread() race
Linus Torvalds [Sun, 7 May 2006 17:49:33 +0000 (10:49 -0700)] 
Fix ptrace_attach()/ptrace_traceme()/de_thread() race

This holds the task lock (and, for ptrace_attach, the tasklist_lock)
over the actual attach event, which closes a race between attacking to a
thread that is either doing a PTRACE_TRACEME or getting de-threaded.

Thanks to Oleg Nesterov for reminding me about this, and Chris Wright
for noticing a lost return value in my first version.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[IRDA] irda-usb: use NULL instead of 0
Randy Dunlap [Sun, 7 May 2006 01:34:10 +0000 (18:34 -0700)] 
[IRDA] irda-usb: use NULL instead of 0

Use NULL instead of 0 for a null pointer value (sparse warning):

drivers/net/irda/irda-usb.c:1781:30: warning: Using plain integer as NULL pointer

Also, correct timeout argument to use milliseconds instead of jiffies.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[IPV4]: Remove likely in ip_rcv_finish()
Hua Zhong [Sun, 7 May 2006 01:11:39 +0000 (18:11 -0700)] 
[IPV4]: Remove likely in ip_rcv_finish()

This is another result from my likely profiling tool
(dwalker@mvista.com just sent the patch of the profiling tool to
linux-kernel mailing list, which is similar to what I use).

On my system (not very busy, normal development machine within a
VMWare workstation), I see a 6/5 miss/hit ratio for this "likely".

Signed-off-by: Hua Zhong <hzhong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[NET]: Create netdev attribute_groups with class_device_add
Stephen Hemminger [Sun, 7 May 2006 00:56:03 +0000 (17:56 -0700)] 
[NET]: Create netdev attribute_groups with class_device_add

Atomically create attributes when class device is added. This avoids
the race between registering class_device (which generates hotplug
event), and the creation of attribute groups.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[CLASS DEVICE]: add attribute_group creation
Stephen Hemminger [Sun, 7 May 2006 00:55:11 +0000 (17:55 -0700)] 
[CLASS DEVICE]: add attribute_group creation

Extend the support of attribute groups in class_device's to allow
groups to be created as part of the registration process. This allows
network device's to avoid race between registration and creating
groups.

Note that unlike attributes that are a property of the class object,
the groups are a property of the class_device object. This is done
because there are different types of network devices (wireless for
example).

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[ARM] rtc-sa1100: fix compiler warnings and error cleanup
Russell King [Sat, 6 May 2006 10:29:21 +0000 (11:29 +0100)] 
[ARM] rtc-sa1100: fix compiler warnings and error cleanup

Fix:
drivers/rtc/rtc-sa1100.c: In function `sa1100_rtc_proc':
drivers/rtc/rtc-sa1100.c:298: warning: unsigned int format, long unsigned int arg (arg 3)

and arrange for sa1100_rtc_open() to pass the devid to free_irq()
rather than NULL.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[ARM] Allow SA1100 RTC alarm to be configured for wakeup
Russell King [Sat, 6 May 2006 10:26:30 +0000 (11:26 +0100)] 
[ARM] Allow SA1100 RTC alarm to be configured for wakeup

The SA1100 RTC alarm can be configured to wake up the CPU
from sleep mode, and the RTC driver has been using the
API to configure this mode.  Unfortunately, the code was
which sets the required bit in the hardware was missing.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
18 years ago[TCP]: Fix snd_cwnd adjustments in tcp_highspeed.c
John Heffner [Sat, 6 May 2006 00:41:44 +0000 (17:41 -0700)] 
[TCP]: Fix snd_cwnd adjustments in tcp_highspeed.c

Xiaoliang (David) Wei wrote:
> Hi gurus,
>
>    I am reading the code of tcp_highspeed.c in the kernel and have a
> question on the hstcp_cong_avoid function, specifically the following
> AI part (line 136~143 in net/ipv4/tcp_highspeed.c ):
>
>                /* Do additive increase */
>                if (tp->snd_cwnd < tp->snd_cwnd_clamp) {
>                        tp->snd_cwnd_cnt += ca->ai;
>                        if (tp->snd_cwnd_cnt >= tp->snd_cwnd) {
>                                tp->snd_cwnd++;
>                                tp->snd_cwnd_cnt -= tp->snd_cwnd;
>                        }
>                }
>
>    In this part, when (tp->snd_cwnd_cnt == tp->snd_cwnd),
> snd_cwnd_cnt will be -1... snd_cwnd_cnt is defined as u16, will this
> small chance of getting -1 becomes a problem?
> Shall we change it by reversing the order of the cwnd++ and cwnd_cnt -=
> cwnd?

Absolutely correct.  Thanks.

Signed-off-by: John Heffner <jheffner@psc.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[NETROM/ROSE]: Kill module init version kernel log messages.
Ralf Baechle [Sat, 6 May 2006 00:19:26 +0000 (17:19 -0700)] 
[NETROM/ROSE]: Kill module init version kernel log messages.

There are out of date and don't tell the user anything useful.
The similar messages which IPV4 and the core networking used
to output were killed a long time ago.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[DCCP]: Fix sock_orphan dead lock
Herbert Xu [Sat, 6 May 2006 00:09:13 +0000 (17:09 -0700)] 
[DCCP]: Fix sock_orphan dead lock

Calling sock_orphan inside bh_lock_sock in dccp_close can lead to dead
locks.  For example, the inet_diag code holds sk_callback_lock without
disabling BH.  If an inbound packet arrives during that admittedly tiny
window, it will cause a dead lock on bh_lock_sock.  Another possible
path would be through sock_wfree if the network device driver frees the
tx skb in process context with BH enabled.

We can fix this by moving sock_orphan out of bh_lock_sock.

The tricky bit is to work out when we need to destroy the socket
ourselves and when it has already been destroyed by someone else.

By moving sock_orphan before the release_sock we can solve this
problem.  This is because as long as we own the socket lock its
state cannot change.

So we simply record the socket state before the release_sock
and then check the state again after we regain the socket lock.
If the socket state has transitioned to DCCP_CLOSED in the time being,
we know that the socket has been destroyed.  Otherwise the socket is
still ours to keep.

This problem was discoverd by Ingo Molnar using his lock validator.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[BRIDGE]: keep track of received multicast packets
Stephen Hemminger [Sat, 6 May 2006 00:07:13 +0000 (17:07 -0700)] 
[BRIDGE]: keep track of received multicast packets

It makes sense to add this simple statistic to keep track of received
multicast packets.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[SCTP]: Fix state table entries for chunks received in CLOSED state.
Sridhar Samudrala [Sat, 6 May 2006 00:05:23 +0000 (17:05 -0700)] 
[SCTP]: Fix state table entries for chunks received in CLOSED state.

Discard an unexpected chunk in CLOSED state rather can calling BUG().

Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[SCTP]: Fix panic's when receiving fragmented SCTP control chunks.
Sridhar Samudrala [Sat, 6 May 2006 00:04:43 +0000 (17:04 -0700)] 
[SCTP]: Fix panic's when receiving fragmented SCTP control chunks.

Use pskb_pull() to handle incoming COOKIE_ECHO and HEARTBEAT chunks that
are received as skb's with fragment list.

Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[SCTP]: Prevent possible infinite recursion with multiple bundled DATA.
Vladislav Yasevich [Sat, 6 May 2006 00:03:49 +0000 (17:03 -0700)] 
[SCTP]: Prevent possible infinite recursion with multiple bundled DATA.

There is a rare situation that causes lksctp to go into infinite recursion
and crash the system.  The trigger is a packet that contains at least the
first two DATA fragments of a message bundled together. The recursion is
triggered when the user data buffer is smaller that the full data message.
The problem is that we clone the skb for every fragment in the message.
When reassembling the full message, we try to link skbs from the "first
fragment" clone using the frag_list. However, since the frag_list is shared
between two clones in this rare situation, we end up setting the frag_list
pointer of the second fragment to point to itself.  This causes
sctp_skb_pull() to potentially recurse indefinitely.

Proposed solution is to make a copy of the skb when attempting to link
things using frag_list.

Signed-off-by: Vladislav Yasevich <vladsilav.yasevich@hp.com>
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[SCTP]: Allow spillover of receive buffer to avoid deadlock.
Neil Horman [Sat, 6 May 2006 00:02:09 +0000 (17:02 -0700)] 
[SCTP]: Allow spillover of receive buffer to avoid deadlock.

This patch fixes a deadlock situation in the receive path by allowing
temporary spillover of the receive buffer.

- If the chunk we receive has a tsn that immediately follows the ctsn,
  accept it even if we run out of receive buffer space and renege data with
  higher TSNs.
- Once we accept one chunk in a packet, accept all the remaining chunks
  even if we run out of receive buffer space.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Mark Butler <butlerm@middle.net>
Acked-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years agoMerge master.kernel.org:/home/rmk/linux-2.6-mmc
Linus Torvalds [Fri, 5 May 2006 21:53:48 +0000 (14:53 -0700)] 
Merge master.kernel.org:/home/rmk/linux-2.6-mmc

* master.kernel.org:/home/rmk/linux-2.6-mmc:
  [BLOCK] Fix oops on removal of SD/MMC card