pandora-kernel.git
17 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Tue, 17 Oct 2006 22:04:45 +0000 (15:04 -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: Use mmiowb after doorbell ring
  IB/ipath: Initialize diagpkt file on device init only
  RDMA/amso1100: Fix a NULL dereference in error path
  RDMA/amso1100: pci_module_init() conversion

17 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Tue, 17 Oct 2006 21:46:31 +0000 (14:46 -0700)]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] switch to new pci_get_bus_and_slot API
  [ARM] Update mach-types
  [ARM] Fix Zaurii keyboard/touchscreen drivers
  [ARM] Fix fallout from IRQ regs changes
  [ARM] 3890/1: [Jornada7xx] Addition of MCU commands into jornada720.h
  [ARM] 3889/1: [Jornada7xx] Addition of correct SDRAM params into cpu-sa1110.c
  [ARM] 3888/1: add pxa27x SSP FSRT register bit definition

17 years agosky2: GMAC pause frame
Stephen Hemminger [Tue, 17 Oct 2006 17:24:25 +0000 (10:24 -0700)]
sky2: GMAC pause frame

This reverts earlier change that attempted to fix flow control.
Device needs to discard pause frames, otherwise it hangs after a while.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: accept multicast pause frames
Stephen Hemminger [Tue, 17 Oct 2006 17:24:23 +0000 (10:24 -0700)]
sky2: accept multicast pause frames

When using flow control, the PHY needs to accept multicast pause frames.
Without this fix, these frames were getting discarded by the PHY before
doing any flow control.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: version 1.9
Stephen Hemminger [Tue, 17 Oct 2006 17:24:18 +0000 (10:24 -0700)]
sky2: version 1.9

Mark version, this has been a lot of patches.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: no message on rx fifo overflow
Stephen Hemminger [Tue, 17 Oct 2006 17:24:16 +0000 (10:24 -0700)]
sky2: no message on rx fifo overflow

Under high load it is possible to make the receiver FIFO get overloaded.
The driver/hardware recover properly, so there is no reason to fill the log
with lots of extra messages, just update counter.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: flow control setting fixes
Stephen Hemminger [Tue, 17 Oct 2006 17:24:13 +0000 (10:24 -0700)]
sky2: flow control setting fixes

The result of flow control negotiation should not limit the next
negotiatition. If board is plugged into an old half duplex 10Mbit port,
without pause, then replugged into a gigabit port, it should negotiate
what is desired, not inherit that last negotiation.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: don't reset PHY twice
Stephen Hemminger [Tue, 17 Oct 2006 17:24:10 +0000 (10:24 -0700)]
sky2: don't reset PHY twice

Don't need to reset PHY twice on startup.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: use duplex result bits
Stephen Hemminger [Tue, 17 Oct 2006 17:24:08 +0000 (10:24 -0700)]
sky2: use duplex result bits

The result of duplex negotiation is avaliable in the phy status
register, so use that to simplify code and avoid rereading the PHY.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: advertising register 16 bits
Stephen Hemminger [Tue, 17 Oct 2006 17:24:07 +0000 (10:24 -0700)]
sky2: advertising register 16 bits

The advertising bits (from ethtool.h) fit in 16 bits.

17 years agosky2: fiber pause bits
Stephen Hemminger [Tue, 17 Oct 2006 17:24:04 +0000 (10:24 -0700)]
sky2: fiber pause bits

The advertisement bits for flow control are located in
different location on fiber (1000baseX)

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: phy irq on shutdown
Stephen Hemminger [Tue, 17 Oct 2006 17:23:56 +0000 (10:23 -0700)]
sky2: phy irq on shutdown

When PHY is turned off on shutdown, it causes the IRQ to get stuck on.
Make sure and disable the IRQ first, and if IRQ occurs when device
is not running, don't access PHY because that will hang.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: turn of workaround timer
Stephen Hemminger [Tue, 17 Oct 2006 17:20:51 +0000 (10:20 -0700)]
sky2: turn of workaround timer

The workaround timer is not needed in most systems with proper IRQ
routing and by perodically waking up it adds to laptop power consumption.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years agosky2: MSI test is only a warning
Stephen Hemminger [Tue, 17 Oct 2006 17:17:18 +0000 (10:17 -0700)]
sky2: MSI test is only a warning

Some motherboards don't implement MSI correctly. The driver handles this
but the warning is too verbose and overly cautious.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
17 years ago[ARM] switch to new pci_get_bus_and_slot API
Alan Cox [Mon, 16 Oct 2006 15:49:50 +0000 (16:49 +0100)]
[ARM] switch to new pci_get_bus_and_slot API

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
17 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Tue, 17 Oct 2006 15:56:43 +0000 (08:56 -0700)]
Merge git://git./linux/kernel/git/dtor/input

* git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: fm801-gp - handle errors from pci_enable_device()
  Input: gameport core - handle errors returned by device_bind_driver()
  Input: serio core - handle errors returned by device_bind_driver()
  Lockdep: fix compile error in drivers/input/serio/serio.c
  Input: serio - add lockdep annotations
  Lockdep: add lockdep_set_class_and_subclass() and lockdep_set_subclass()
  Input: atkbd - supress "too many keys" error message
  Input: i8042 - supress ACK/NAKs when blinking during panic
  Input: add missing exports to fix modular build

17 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
Linus Torvalds [Tue, 17 Oct 2006 15:20:57 +0000 (08:20 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/paulus/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc:
  [POWERPC] Check for offline nodes in pci NUMA code
  [POWERPC] Better check in show_instructions
  [POWERPC] POWER6 has 6 PMCs
  [POWERPC] Never panic when taking altivec exceptions from userspace
  [POWERPC] Fix IO Window Updates on P2P bridges.
  [POWERPC] Add Makefile entry for MPC832x_mds support
  [POWERPC] Fix MPC8360EMDS PB board support
  [POWERPC] ppc: Add missing calls to set_irq_regs
  [POWERPC] Off-by-one in /arch/ppc/platforms/mpc8*
  [POWERPC] Add DOS partition table support to mpc834x_itx_defconfig
  [POWERPC] spufs: fix support for read/write on cntl
  [POWERPC] Don't crash on cell with 2 BEs when !CONFIG_NUMA

17 years ago[PATCH] ISDN: check for userspace copy faults
Jeff Garzik [Tue, 17 Oct 2006 07:10:40 +0000 (00:10 -0700)]
[PATCH] ISDN: check for userspace copy faults

Most of the ISDN ->readstat() implementations needed to check
copy_to_user() and put_user() return values.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Cc: Karsten Keil <kkeil@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ISDN: fix drivers, by handling errors thrown by ->readstat()
Jeff Garzik [Tue, 17 Oct 2006 07:10:39 +0000 (00:10 -0700)]
[PATCH] ISDN: fix drivers, by handling errors thrown by ->readstat()

This is a particularly ugly on-failure bug, possibly security, since the
lack of error handling here is covering up another class of bug: failure to
handle copy_to_user() return values.

The I4L API function ->readstat() returns an integer, and by looking at
several existing driver implementations, it is clear that a negative return
value was meant to indicate an error.

Given that several drivers already return a negative value indicating an
errno-style error, the current code would blindly accept that [negative]
value as a valid amount of bytes read.  Obvious damage ensues.

Correcting ->readstat() handling to properly notice errors fixes the
existing code to work correctly on error, and enables future patches to
more easily indicate errors during operation.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Cc: Karsten Keil <kkeil@suse.de>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] drivers/isdn/isdnloop: save_flags()/cli(), restore_flags() replaced appropriately
Amol Lad [Tue, 17 Oct 2006 07:10:37 +0000 (00:10 -0700)]
[PATCH] drivers/isdn/isdnloop: save_flags()/cli(), restore_flags() replaced appropriately

Signed-off-by: Amol Lad <amol@verismonetworks.com>
Acked-by: Karsten Keil <kkeil@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] drivers/isdn/hysdn: save_flags()/cli(), restore_flags() replaced appropriately
Amol Lad [Tue, 17 Oct 2006 07:10:36 +0000 (00:10 -0700)]
[PATCH] drivers/isdn/hysdn: save_flags()/cli(), restore_flags() replaced appropriately

With Karsten Keil <kkeil@suse.de>

save_flags()/cli() pair is replaced with spin_lock_irqsave() and
restore_flags() replaced with spin_unlock_irqrestore()

Tested compile only using allmodconfig

Signed-off-by: Amol Lad <amol@verismonetworks.com>
Acked-by: Karsten Keil <kkeil@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Convert cpu hotplug notifiers to use raw_notifier instead of blocking_notifier
Neil Brown [Tue, 17 Oct 2006 07:10:35 +0000 (00:10 -0700)]
[PATCH] Convert cpu hotplug notifiers to use raw_notifier instead of blocking_notifier

The use of blocking notifier by _cpu_up and _cpu_down in cpu.c has two
problem.

1/ An interaction with the workqueue notifier causes lockdep to spit a
   warning.

2/ A notifier could conceivable be added or removed while _cpu_up or
   _cpu_down are in process.  As each notifier is called twice (prepare
   then commit/abort) this could be unhealthy.

To fix to we simply take cpu_add_remove_lock while adding or removing
notifiers to/from the list.

This makes the 'blocking' usage unnecessary as all accesses to cpu_chain
are now protected by cpu_add_remove_lock.  So change "blocking" to "raw" in
all relevant places.  This fixes 1.

Credit: Andrew Morton
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Michal Piotrowski <michal.k.k.piotrowski@gmail.com> (reporter)
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rt-mutex: fixup rt-mutex debug code
Peter Zijlstra [Tue, 17 Oct 2006 07:10:33 +0000 (00:10 -0700)]
[PATCH] rt-mutex: fixup rt-mutex debug code

BUG: warning at kernel/rtmutex-debug.c:125/rt_mutex_debug_task_free() (Not tainted)
 [<c04051e3>] show_trace_log_lvl+0x58/0x16a
 [<c04057f0>] show_trace+0xd/0x10
 [<c0405900>] dump_stack+0x19/0x1b
 [<c043f03d>] rt_mutex_debug_task_free+0x35/0x6a
 [<c04224c0>] free_task+0x15/0x24
 [<c042378c>] copy_process+0x12bd/0x1324
 [<c0423835>] do_fork+0x42/0x113
 [<c04021dd>] sys_fork+0x19/0x1b
 [<c0403fb7>] syscall_call+0x7/0xb

In copy_process(), dup_task_struct() also duplicates the ->pi_lock,
->pi_waiters and ->pi_blocked_on members.  rt_mutex_debug_task_free()
called from free_task() validates these members.  However free_task() can
be invoked before these members are reset for the new task.

Move the initialization code before the first bail that can hit free_task().

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Add entry.S labels to tag file
Aneesh Kumar K.V [Tue, 17 Oct 2006 07:10:32 +0000 (00:10 -0700)]
[PATCH] Add entry.S labels to tag file

Add functions defined using ENTRY macro to the tags file.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@gmail.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] kbuild: allow multi-word $M in Makefile.modpost
Greg Banks [Tue, 17 Oct 2006 07:10:30 +0000 (00:10 -0700)]
[PATCH] kbuild: allow multi-word $M in Makefile.modpost

Some people want to do crazy things like pass multiple directories as the
value of $(SUBDIRS) or $M.  Mostly this kinda works, except that
Makefile.modpost constructs a modpost commandline which fails modpost's
argument parsing.  This patch fixes that little wrinkle.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] epca: prevent panic on tty_register_driver() failure
Akinobu Mita [Tue, 17 Oct 2006 07:10:28 +0000 (00:10 -0700)]
[PATCH] epca: prevent panic on tty_register_driver() failure

Make epca fail on initialization failure instead of panic.

Cc: "Digi International, Inc" <Eng.Linux@digi.com>
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Alan Cox <alan@redhat.com>
Acked-by: Scott Kilau <scottk@digi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rd: memory leak on rd_init() failure
Akinobu Mita [Tue, 17 Oct 2006 07:10:27 +0000 (00:10 -0700)]
[PATCH] rd: memory leak on rd_init() failure

If RAM disk driver initialization fails due to blk_alloc_queue() faulure, the
gendisk structs stored in rd_disks[] will not be freed completely.

This patch resolves that memory leak case by doing alloc_disk() and
blk_alloc_queue() at the same time.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockdep: annotate i386 apm
Peter Zijlstra [Tue, 17 Oct 2006 07:10:26 +0000 (00:10 -0700)]
[PATCH] lockdep: annotate i386 apm

Lockdep doesn't like to enable interrupts when they are enabled already.

BUG: warning at kernel/lockdep.c:1814/trace_hardirqs_on() (Not tainted)
 [<c04051ed>] show_trace_log_lvl+0x58/0x16a
 [<c04057fa>] show_trace+0xd/0x10
 [<c0405913>] dump_stack+0x19/0x1b
 [<c043abfb>] trace_hardirqs_on+0xa2/0x11e
 [<c041463c>] apm_bios_call_simple+0xcd/0xfd
 [<c0415242>] apm+0x92/0x5b1
 [<c0402005>] kernel_thread_helper+0x5/0xb
DWARF2 unwinder stuck at kernel_thread_helper+0x5/0xb
Leftover inexact backtrace:
 [<c04057fa>] show_trace+0xd/0x10
 [<c0405913>] dump_stack+0x19/0x1b
 [<c043abfb>] trace_hardirqs_on+0xa2/0x11e
 [<c041463c>] apm_bios_call_simple+0xcd/0xfd
 [<c0415242>] apm+0x92/0x5b1
 [<c0402005>] kernel_thread_helper+0x5/0xb

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rtc: fix printk of 64-bit res on 32-bit platform
Jeff Garzik [Tue, 17 Oct 2006 07:10:25 +0000 (00:10 -0700)]
[PATCH] rtc: fix printk of 64-bit res on 32-bit platform

With 64-bit resources on 32-bit platforms, the resource address might be
larger than a void*.  Fix printk to work regardless of resource size.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] fs/partitions/check: add sysfs error handling
Jeff Garzik [Tue, 17 Oct 2006 07:10:23 +0000 (00:10 -0700)]
[PATCH] fs/partitions/check: add sysfs error handling

Handle errors thrown in disk_sysfs_symlinks(), and propagate back to
caller.

The callers and associated functions don't do a real good job of handling
kobject errors anyway (add_partition, register_disk, rescan_partitions), so
this should do until something better comes along.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] I2O: handle a few sysfs errors
Jeff Garzik [Tue, 17 Oct 2006 07:10:22 +0000 (00:10 -0700)]
[PATCH] I2O: handle a few sysfs errors

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Cc: Markus Lidel <Markus.Lidel@shadowconnect.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] drivers/led: handle sysfs errors
Jeff Garzik [Tue, 17 Oct 2006 07:10:20 +0000 (00:10 -0700)]
[PATCH] drivers/led: handle sysfs errors

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Fix IO error reporting on fsync()
Jan Kara [Tue, 17 Oct 2006 07:10:19 +0000 (00:10 -0700)]
[PATCH] Fix IO error reporting on fsync()

When IO error happens on metadata buffer, buffer is freed from memory and
later fsync() is called, filesystems like ext2 fail to report EIO.  We

solve the problem by introducing a pointer to associated address space into
the buffer_head.  When a buffer is removed from a list of metadata buffers
associated with an address space, IO error is transferred from the buffer to
the address space, so that fsync can later report it.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: Allow lockd to drop replies as appropriate
NeilBrown [Tue, 17 Oct 2006 07:10:18 +0000 (00:10 -0700)]
[PATCH] knfsd: Allow lockd to drop replies as appropriate

It is possible for the ->fopen callback from lockd into nfsd to find that an
answer cannot be given straight away (an upcall is needed) and so the request
has to be 'dropped', to be retried later.  That error status is not currently
propagated back.

So:
  Change nlm_fopen to return nlm error codes (rather than a private
  protocol) and define a new nlm_drop_reply code.
  Cause nlm_drop_reply to cause the rpc request to get rpc_drop_reply
  when this error comes back.
  Cause svc_process to drop a request which returns a status of
  rpc_drop_reply.

[akpm@osdl.org: fix warning storm]
Cc: Marc Eshel <eshel@almaden.ibm.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: Fix bug in recent lockd patches that can cause reclaim to fail
NeilBrown [Tue, 17 Oct 2006 07:10:17 +0000 (00:10 -0700)]
[PATCH] knfsd: Fix bug in recent lockd patches that can cause reclaim to fail

When an nfs server shuts down, lockd needs to release all the locks even
though the client still holds them.

It should therefore not 'unmonitor' the clients, so that the files in nfs/sm
will still be there when the nfs server restarts, so that those clients will
be told to reclaim their locks.

However the hosts are fully unmonitored, so statd may well remove the files.

lockd has a test for 'sm_sticky' and avoid the unmonitor call if it is set,
but it is currently not set.

So set it when tearing down lockd.

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: Fix error handling in nfsd's callback client
J. Bruce Fields [Tue, 17 Oct 2006 07:10:16 +0000 (00:10 -0700)]
[PATCH] knfsd: nfsd4: Fix error handling in nfsd's callback client

Coverity noticed that the error handling code in the NFSv4 callback client
sets cb->cb_client to NULL, then calls rpc_shutdown_client with the NULL
pointer.

Coverity: #cid 1397

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: fix open permission checking
J. Bruce Fields [Tue, 17 Oct 2006 07:10:14 +0000 (00:10 -0700)]
[PATCH] knfsd: nfsd4: fix open permission checking

We weren't actually checking for SHARE_ACCESS_WRITE, with the result that the
owner could open a non-writeable file for write!

Continue to allow DENY_WRITE only with write access.

Thanks to Jim Rees for reporting the bug.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: nfsd4: fix owner-override on open
J. Bruce Fields [Tue, 17 Oct 2006 07:10:13 +0000 (00:10 -0700)]
[PATCH] knfsd: nfsd4: fix owner-override on open

If a client creates a file using an open which sets the mode to 000, or if a
chmod changes permissions after a file is opened, then situations may arise
where an NFS client knows that some IO is permitted (because a process holds
the file open), but the NFS server does not (because it doesn't know about the
open, and only sees that the IO conflicts with the current mode of the file).

As a hack to solve this problem, NFS servers normally allow the owner to
override permissions on IO.  The client can still enforce correct
permissions-checking on open by performing an explicit access check.

In NFSv4 the client can rely on the explicit on-the-wire open instead of an
access check.

Therefore we should not be allowing the owner to override permissions on an
over-the-wire open!

However, we should still allow the owner to override permissions in the case
where the client is claiming an open that it already made either before a
reboot, or while it was holding a delegation.

Thanks to Jim Rees for reporting the bug.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] fuse: fix dereferencing dentry parent
Miklos Szeredi [Tue, 17 Oct 2006 07:10:12 +0000 (00:10 -0700)]
[PATCH] fuse: fix dereferencing dentry parent

There's no locking for ->d_revalidate, so fuse_dentry_revalidate() should use
dget_parent() instead of simply dereferencing ->d_parent.

Due to topology changes in the directory tree the parent could become negative
or be destroyed while being used.  There hasn't been any reports about this
yet.

Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] fuse: fix handling of moved directory
Miklos Szeredi [Tue, 17 Oct 2006 07:10:11 +0000 (00:10 -0700)]
[PATCH] fuse: fix handling of moved directory

Fuse considered it an error (EIO) if lookup returned a directory inode, to
which a dentry already refered.  This is because directory aliases are not
allowed.

But in a network filesystem this could happen legitimately, if a directory is
moved on a remote client.  This patch attempts to relax the restriction by
trying to first evict the offending alias from the cache.  If this fails, it
still returns an error (EBUSY).

A rarer situation is if an mkdir races with an indenpendent lookup, which
finds the newly created directory already moved.  In this situation the mkdir
should return success, but that would be incorrect, since the dentry cannot be
instantiated, so return EBUSY.

Previously checking for a directory alias and instantiation of the dentry
weren't done atomically in lookup/mkdir, hence two such calls racing with each
other could create aliased directories.  To prevent this introduce a new
per-connection mutex: fuse_conn->inst_mutex, which is taken for instantiations
with a directory inode.

Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] fuse: fix spurious BUG
Miklos Szeredi [Tue, 17 Oct 2006 07:10:09 +0000 (00:10 -0700)]
[PATCH] fuse: fix spurious BUG

Fix a spurious BUG in an unlikely race, where at least three parallel lookups
return the same inode, but with different file type.  This has not yet been
observed in real life.

Allowing unlimited retries could delay fuse_iget() indefinitely, but this is
really for the broken userspace filesystem to worry about.

Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] fuse: locking fix for nlookup
Miklos Szeredi [Tue, 17 Oct 2006 07:10:08 +0000 (00:10 -0700)]
[PATCH] fuse: locking fix for nlookup

An inode could be returned by independent parallel lookups, in this case an
update of the lookup counter could be lost resulting in a memory leak in
userspace.

Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] document i_size_write locking rules
Miklos Szeredi [Tue, 17 Oct 2006 07:10:07 +0000 (00:10 -0700)]
[PATCH] document i_size_write locking rules

Unless someone reads the documentation for write_seqcount_{begin,end} it is
not obvious, that i_size_write() needs locking.  Especially, that lack of such
locking can result in a system hang.

Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] fuse: fix hang on SMP
Miklos Szeredi [Tue, 17 Oct 2006 07:10:06 +0000 (00:10 -0700)]
[PATCH] fuse: fix hang on SMP

Fuse didn't always call i_size_write() with i_mutex held which caused rare
hangs on SMP/32bit.  This bug has been present since fuse-2.2, well before
being merged into mainline.

The simplest solution is to protect i_size_write() with the per-connection
spinlock.  Using i_mutex for this purpose would require some restructuring of
the code and I'm not even sure it's always safe to acquire i_mutex in all
places i_size needs to be set.

Since most of vmtruncate is already duplicated for other reasons, duplicate
the remaining part as well, making all i_size_write() calls internal to fuse.

Using i_size_write() was unnecessary in fuse_init_inode(), since this function
is only called on a newly created locked inode.

Reported by a few people over the years, but special thanks to Dana Henriksen
who was persistent enough in helping me debug it.

Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] sx: fix user-visible typo (devic)
Alexey Dobriyan [Tue, 17 Oct 2006 07:10:05 +0000 (00:10 -0700)]
[PATCH] sx: fix user-visible typo (devic)

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] genirq: clean up irq-flow-type naming
Ingo Molnar [Tue, 17 Oct 2006 07:10:03 +0000 (00:10 -0700)]
[PATCH] genirq: clean up irq-flow-type naming

Introduce desc->name and eliminate the handle_irq_name() hack.  Add
set_irq_chip_and_handler_name() to set the flow type and name at once.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Matthew Wilcox <willy@debian.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] fix `make headers_install'
David Woodhouse [Tue, 17 Oct 2006 07:10:02 +0000 (00:10 -0700)]
[PATCH] fix `make headers_install'

Fix this:

make[3]: *** No rule to make target
`/mnt/md0/devel/linux-git/include/linux/version.h', needed by
`/mnt/md0/devel/linux-git-obj/usr/include/linux/version.h'.  Stop.
make[2]: *** [linux] Error 2
make[1]: *** [headers_install] Error 2

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] drivers/char/specialix.c: fix the baud conversion
Adrian Bunk [Tue, 17 Oct 2006 07:10:00 +0000 (00:10 -0700)]
[PATCH] drivers/char/specialix.c: fix the baud conversion

Correct the following bugs introduced by commit
67cc0161ecc9ebee6eba4af6cbfdba028090b1b9:

- remove one remaining and now incorrect baud_table[] usage
- "baud +=" is no longer correct

The former bug was spotted by the Coverity checker.

Rolf Eike Beer spotted a bug in the initial version of my patch.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] swsusp: fix memory leaks
Andrew Morton [Tue, 17 Oct 2006 07:09:59 +0000 (00:09 -0700)]
[PATCH] swsusp: fix memory leaks

My fancy new swsusp IO code had a big memory leak.  It's somewhat invisible
because the whole mem_map[] gets overwritten after resume, but it can cause us
to get low on memory during the actual suspend process.

Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] acpi_processor_latency_notifier(): UP warning fix
Andrew Morton [Tue, 17 Oct 2006 07:09:58 +0000 (00:09 -0700)]
[PATCH] acpi_processor_latency_notifier(): UP warning fix

drivers/acpi/processor_idle.c:1112: warning: 'smp_callback' defined but not used

Cc: Len Brown <lenb@kernel.org>
Cc: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] vmalloc(): don't pass __GFP_ZERO to slab
Andrew Morton [Tue, 17 Oct 2006 07:09:57 +0000 (00:09 -0700)]
[PATCH] vmalloc(): don't pass __GFP_ZERO to slab

A recent change to the vmalloc() code accidentally resulted in us passing
__GFP_ZERO into the slab allocator.  But we only wanted __GFP_ZERO for the
actual pages whcih are being vmalloc()ed, and passing __GFP_ZERO into slab is
not a rational thing to ask for.

Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rtc-max6902: month conversion fix
Francisco Larramendi [Tue, 17 Oct 2006 07:09:53 +0000 (00:09 -0700)]
[PATCH] rtc-max6902: month conversion fix

Fix October-only BCD-to-binary conversion bug:

0x08 -> 7
0x09 -> 8
0x10 -> 15 (!)
0x11 -> 19

Fixes http://bugzilla.kernel.org/show_bug.cgi?id=7361

Cc: Raphael Assenat <raph@raphnet.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] w1 kconfig fix
Evgeniy Polyakov [Tue, 17 Oct 2006 07:09:47 +0000 (00:09 -0700)]
[PATCH] w1 kconfig fix

Remove dependency of w1 subsytem from connector, only w1_con must depend on
it.  With attached patch applied to vanilla 2.6.19-git things works fine.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Cc: <dmb@pochta.ru>
Cc: Greg KH <greg@kroah.com>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Cc: Adrian Bunk <bunk@stusta.de>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] knfsd: add nfs-export support to tmpfs
David M. Grimes [Tue, 17 Oct 2006 07:09:45 +0000 (00:09 -0700)]
[PATCH] knfsd: add nfs-export support to tmpfs

We need to encode a decode the 'file' part of a handle.  We simply use the
inode number and generation number to construct the filehandle.

The generation number is the time when the file was created.  As inode numbers
cycle through the full 32 bits before being reused, there is no real chance of
the same inum being allocated to different files in the same second so this is
suitably unique.  Using time-of-day rather than e.g.  jiffies makes it less
likely that the same filehandle can be created after a reboot.

In order to be able to decode a filehandle we need to be able to lookup by
inum, which means that the inode needs to be added to the inode hash table
(tmpfs doesn't currently hash inodes as there is never a need to lookup by
inum).  To avoid overhead when not exporting, we only hash an inode when it is
first exported.  This requires a lock to ensure it isn't hashed twice.

This code is separate from the patch posted in June06 from Atal Shargorodsky
which provided the same functionality, but does borrow slightly from it.

Locking comment: Most filesystems that hash their inodes do so at the point
where the 'struct inode' is initialised, and that has suitable locking
(I_NEW).  Here in shmem, we are hashing the inode later, the first time we
need an NFS file handle for it.  We no longer have I_NEW to ensure only one
thread tries to add it to the hash table.

Cc: Atal Shargorodsky <atal@codefidence.com>
Cc: Gilad Ben-Yossef <gilad@codefidence.com>
Signed-off-by: David M. Grimes <dgrimes@navisite.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] remove carta_random32
Andrew Morton [Tue, 17 Oct 2006 07:09:44 +0000 (00:09 -0700)]
[PATCH] remove carta_random32

This library function should be in obj-y and not in lib-y.  But when we do
that it clashes unpleasantly with the assembly-language implementation in the
ia64 architecture.

Instead of trying to fix it all up, just remove the generic carta_random32 in
the expectation that the recently-made-generic random32() will suffice.

If/when perfmon is migrated to random32, ia64's private carta_random32
implementation can also be removed.

Cc: Stephane Eranian <eranian@hpl.hp.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rename net_random to random32
Stephen Hemminger [Tue, 17 Oct 2006 07:09:42 +0000 (00:09 -0700)]
[PATCH] rename net_random to random32

Make net_random() more widely available by calling it random32

akpm: hopefully this will permit the removal of carta_random32.  That needs
confirmation from Stephane - this code looks somewhat more computationally
expensive, and has a different (ie: callee-stateful) interface.

[akpm@osdl.org: lots of build fixes, cleanups]
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Stephane Eranian <eranian@hpl.hp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] PROC_NUMBUF is wrong
Andrew Morton [Tue, 17 Oct 2006 07:09:41 +0000 (00:09 -0700)]
[PATCH] PROC_NUMBUF is wrong

Actually, the decimal representation of a 32-bit signed number can take 12
bytes, including the \0.

And then some code adds a \n as well, so let's give it 13 bytes.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] posix-cpu-timers: prevent signal delivery starvation
Thomas Gleixner [Tue, 17 Oct 2006 07:09:39 +0000 (00:09 -0700)]
[PATCH] posix-cpu-timers: prevent signal delivery starvation

The integer divisions in the timer accounting code can round the result
down to 0.  Adding 0 is without effect and the signal delivery stops.

Clamp the division result to minimum 1 to avoid this.

Problem was reported by Seongbae Park <spark@google.com>, who provided
also an inital patch.

Roland sayeth:

  I have had some more time to think about the problem, and to reproduce it
  using Toyo's test case.  For the record, if my understanding of the problem
  is correct, this happens only in one very particular case.  First, the
  expiry time has to be so soon that in cputime_t units (usually 1s/HZ ticks)
  it's < nthreads so the division yields zero.  Second, it only affects each
  thread that is so new that its CPU time accumulation is zero so now+0 is
  still zero and ->it_*_expires winds up staying zero.  For the VIRT and PROF
  clocks when cputime_t is tick granularity (or the SCHED clock on
  configurations where sched_clock's value only advances on clock ticks), this
  is not hard to arrange with new threads starting up and blocking before they
  accumulate a whole tick of CPU time.  That's what happens in Toyo's test
  case.

  Note that in general it is fine for that division to round down to zero,
  and set each thread's expiry time to its "now" time.  The problem only
  arises with thread's whose "now" value is still zero, so that now+0 winds up
  0 and is interpreted as "not set" instead of ">= now".  So it would be a
  sufficient and more precise fix to just use max(ticks, 1) inside the loop
  when setting each it_*_expires value.

  But, it does no harm to round the division up to one and always advance
  every thread's expiry time.  If the thread didn't already fire timers for
  the expiry time of "now", there is no expectation that it will do so before
  the next tick anyway.  So I followed Thomas's patch in lifting the max out
  of the loops.

  This patch also covers the reload cases, which are harder to write a test
  for (and I didn't try).  I've tested it with Toyo's case and it fixes that.

[toyoa@mvista.com: fix: min_t -> max_t]
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Roland McGrath <roland@redhat.com>
Cc: Daniel Walker <dwalker@mvista.com>
Cc: Toyo Abe <toyoa@mvista.com>
Cc: john stultz <johnstul@us.ibm.com>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Seongbae Park <spark@google.com>
Cc: Peter Mattis <pmattis@google.com>
Cc: Rohit Seth <rohitseth@google.com>
Cc: Martin Bligh <mbligh@google.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] md: fix /proc/mdstat refcounting
Akinobu Mita [Tue, 17 Oct 2006 07:09:38 +0000 (00:09 -0700)]
[PATCH] md: fix /proc/mdstat refcounting

I have seen mdadm oops after successfully unloading md module.

This patch privents from unloading md module while
mdadm is polling /proc/mdstat.

Cc: Neil Brown <neilb@suse.de>
Signed-off-by: Akinbou Mita <akinobu.mita@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] invalidate: remove_mapping() fix
Andrew Morton [Tue, 17 Oct 2006 07:09:36 +0000 (00:09 -0700)]
[PATCH] invalidate: remove_mapping() fix

If remove_mapping() failed to remove the page from its mapping, don't go and
mark it not uptodate!  Makes kernel go dead.

(Actually, I don't think the ClearPageUptodate is needed there at all).

Says Nick Piggin:

   "Right, it isn't needed because at this point the page is guaranteed
    by remove_mapping to have no references (except us) and cannot pick
    up any new ones because it is removed from pagecache.

    We can delete it."

Signed-off-by: Andrew Morton <akpm@osdl.org>
Acked-by: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] null dereference in fs/jbd2/journal.c
Dave Kleikamp [Tue, 17 Oct 2006 07:09:35 +0000 (00:09 -0700)]
[PATCH] null dereference in fs/jbd2/journal.c

This is Eric Sesterhenn's jbd patch applied to jbd2.
Commit: 41716c7c21b15e7ecf14f0caf1eef3980707fb74

His words:

Since commit d1807793e1e7e502e3dc047115e9dbc3b50e4534 we dereference a NULL
pointer.  Coverity id #1432.  We set journal to NULL, and use it directly
afterwards.

Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
Cc: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] i386 Time: Avoid PIT SMP lockups
john stultz [Tue, 17 Oct 2006 07:09:32 +0000 (00:09 -0700)]
[PATCH] i386 Time: Avoid PIT SMP lockups

Avoid possible PIT livelock issues seen on SMP systems (and reported by
Andi), by not allowing it as a clocksource on SMP boxes.

However, since the PIT may no longer be present, we have to properly handle
the cases where SMP systems have TSC skew and fall back from the TSC.
Since the PIT isn't there, it would "fall back" to the TSC again.  So this
changes the jiffies rating to 1, and the TSC-bad rating value to 0.

Thus you will get the following behavior priority on i386 systems:

tsc [if present & stable]
hpet [if present]
cyclone [if present]
acpi_pm [if present]
pit [if UP]
jiffies

Rather then the current more complicated:
tsc [if present & stable]
hpet [if present]
cyclone [if present]
acpi_pm [if present]
pit [if cpus < 4]
tsc [if present & unstable]
jiffies

Signed-off-by: John Stultz <johnstul@us.ibm.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] New MMC maintainer
Pierre Ossman [Tue, 17 Oct 2006 07:09:30 +0000 (00:09 -0700)]
[PATCH] New MMC maintainer

I will be taking over after Russell King as the new maintainer of the
MMC layer.

Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] lockdep: increase max allowed recursion depth
Ingo Molnar [Tue, 17 Oct 2006 07:09:28 +0000 (00:09 -0700)]
[PATCH] lockdep: increase max allowed recursion depth

In general, lockdep warnings are intended to be non-fatal, so I have put in
various practical limits on internal data structure failure modes.  We haven't
had a /single/ lockdep-internal crash ever since lockdep went upstream [the
unwinder crashes are outside of lockdep], and that's largely due to the good
internal checks it does.

Recursion within the dependency graph is currently limited to 20, that's
probably not enough on some many-CPU boxes - this patch doubles it to 40.  I
have written the lockdep functions to have as small stackframes as possible,
so 40 should be OK too.  (The practical recursion limit should be somewhere
between 100 and 200 entries.  If we hit that then I'll change the algorithm to
be iteration-based.  Graph walking logic is so easy to program via recursion,
so i'd like to keep recursion as long as possible.)

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] synclink: remove PAGE_SIZE reference
Paul Fulghum [Tue, 17 Oct 2006 07:09:27 +0000 (00:09 -0700)]
[PATCH] synclink: remove PAGE_SIZE reference

Remove reference to PAGE_SIZE that causes errors if PAGE_SIZE != 4096

Signed-off-by: Paul Fulghum <paulkf@microgate.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ioc4: Enable build on non-SN2
Brent Casavant [Tue, 17 Oct 2006 07:09:25 +0000 (00:09 -0700)]
[PATCH] ioc4: Enable build on non-SN2

The SGI PCI-RT card, based on the SGI IOC4 chip, will be made available on
Altix XE (x86_64) platforms in the near future.  As such it is now a
misnomer for the IOC4 base device driver to live under drivers/sn, and
would complicate builds for non-SN2.

This patch moves the IOC4 base driver code from drivers/sn to drivers/misc,
and updates the associated Makefiles and Kconfig files to allow building on
non-SN2 configs.  Due to the resulting change in link order, it is now
necessary to use late_initcall() for IOC4 subdriver initialization.

[akpm@osdl.org: __udivdi3 fix]
[akpm@osdl.org: fix default in Kconfig]
Acked-by: Pat Gefre <pfg@sgi.com>
Acked-by: Jeremy Higdon <jeremy@sgi.com>
Signed-off-by: Brent Casavant <bcasavan@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ioc4: Remove SN2 feature and config dependencies
Brent Casavant [Tue, 17 Oct 2006 07:09:24 +0000 (00:09 -0700)]
[PATCH] ioc4: Remove SN2 feature and config dependencies

The SGI PCI-RT card, based on the SGI IOC4 chip, will be made available on
Altix XE (x86_64) platforms in the near future.  As such dependencies on
SN2-specific features and config dependencies need to be removed.

This patch updates the Kconfig files to remove the config dependency, and
updates the IOC4 bus speed detection routine to use universally available
time interfaces instead of mmtimer.

Signed-off-by: Brent Casavant <bcasavan@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoIB/mthca: Use mmiowb after doorbell ring
Arthur Kepner [Tue, 17 Oct 2006 03:22:35 +0000 (20:22 -0700)]
IB/mthca: Use mmiowb after doorbell ring

We discovered a problem when running IPoIB applications on multiple
CPUs on an Altix system. Many messages such as:

ib_mthca 0002:01:00.0: SQ 000014 full (19941644 head, 19941707 tail, 64 max, 0 nreq)

appear in syslog, and the driver wedges up.

Apparently this is because writes to the doorbells from different CPUs
reach the device out of order. The following patch adds mmiowb() calls
after doorbell rings to ensure the doorbell writes are ordered.

Signed-off-by: Arthur Kepner <akepner@sgi.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years ago[ARM] Update mach-types
Russell King [Mon, 16 Oct 2006 20:14:51 +0000 (21:14 +0100)]
[ARM] Update mach-types

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
17 years ago[PATCH] CONFIG_TELCLOCK depends on X86
Geert Uytterhoeven [Mon, 16 Oct 2006 17:59:43 +0000 (19:59 +0200)]
[PATCH] CONFIG_TELCLOCK depends on X86

The telecom clock driver for MPBL0010 ATCA SBC depends on X86

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Mark Gross <mark.gross@intel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoIB/ipath: Initialize diagpkt file on device init only
Robert Walsh [Tue, 10 Oct 2006 21:55:45 +0000 (14:55 -0700)]
IB/ipath: Initialize diagpkt file on device init only

Don't attempt to set up the diagpkt device in the module init code.
Instead, wait until a piece of hardware is initialized.  Fixes a
problem when loading the ib_ipath module when no InfiniPath hardware
is present: modprobe would go into the D state and stay there.

Signed-off-by: Robert Walsh <robert.walsh@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoRDMA/amso1100: Fix a NULL dereference in error path
Adrian Bunk [Tue, 10 Oct 2006 21:26:02 +0000 (14:26 -0700)]
RDMA/amso1100: Fix a NULL dereference in error path

This patch fixes a NULL dereference spotted by the Coverity checker.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Acked-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoRDMA/amso1100: pci_module_init() conversion
Henrik Kretzschmar [Tue, 10 Oct 2006 21:26:01 +0000 (14:26 -0700)]
RDMA/amso1100: pci_module_init() conversion

pci_module_init() convertion in amso1100 driver.

Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years ago[PATCH] fix epoll_pwait when EPOLL=n
Randy Dunlap [Mon, 16 Oct 2006 16:01:46 +0000 (09:01 -0700)]
[PATCH] fix epoll_pwait when EPOLL=n

Fixes http://bugzilla.kernel.org/show_bug.cgi?id=7371

sys_epoll_pwait needs to be listed as a conditional (weak)
entry point for CONFIG_EPOLL=n.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Mon, 16 Oct 2006 15:33:06 +0000 (08:33 -0700)]
Merge /pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6: (25 commits)
  [Bluetooth] Use work queue to trigger URB submission
  [Bluetooth] Add locking for bt_proto array manipulation
  [Bluetooth] Check if DLC is still attached to the TTY
  [Bluetooth] Fix reference count when connection lookup fails
  [Bluetooth] Disconnect HID interrupt channel first
  [Bluetooth] Support concurrent connect requests
  [Bluetooth] Make use of virtual devices tree
  [Bluetooth] Handle return values from driver core functions
  [Bluetooth] Fix compat ioctl for BNEP, CMTP and HIDP
  [IPV6] sit: Add missing MODULE_LICENSE
  [IPV6]: Remove bogus WARN_ON in Proxy-NA handling.
  [IPv6] rules: Use RT6_LOOKUP_F_HAS_SADDR and fix source based selectors
  [XFRM]: Fix xfrm_state_num going negative.
  [NET]: reduce sizeof(struct inet_peer), cleanup, change in peer_check_expire()
  NetLabel: the CIPSOv4 passthrough mapping does not pass categories correctly
  NetLabel: better error handling involving mls_export_cat()
  NetLabel: only deref the CIPSOv4 standard map fields when using standard mapping
  [BRIDGE]: flush forwarding table when device carrier off
  [NETFILTER]: ctnetlink: Remove debugging messages
  [NETFILTER]: Update MAINTAINERS entry
  ...

17 years ago[PATCH] FRV: Use the correct preemption primitives in kmap_atomic() and co
David Howells [Mon, 16 Oct 2006 13:10:49 +0000 (14:10 +0100)]
[PATCH] FRV: Use the correct preemption primitives in kmap_atomic() and co

Use inc/dec_preempt_count() rather than preempt_enable/disable() and manually
add in the compiler barriers that were provided by the latter.  This makes FRV
consistent with other archs.

Furthermore, the compiler barrier effects are now there unconditionally - at
least as far as preemption is concerned - because we don't want the compiler
moving memory accesses out of the section of code in which the mapping is in
force - in effect the kmap_atomic() must imply a LOCK-class barrier and the
kunmap_atomic() must imply an UNLOCK-class barrier to the compiler.

Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] scsi: megaraid_{mm,mbox}: 64-bit DMA capability fix
Andrey Mirkin [Mon, 16 Oct 2006 08:08:43 +0000 (12:08 +0400)]
[PATCH] scsi: megaraid_{mm,mbox}: 64-bit DMA capability fix

It is known that 2 LSI Logic MegaRAID SATA RAID Controllers (150-4 and
150-6) don't support 64-bit DMA.  Unfortunately currently this check is
wrong and driver sets 64-bit DMA mode for these devices.

Signed-off-by: Andrey Mirkin <amirkin@sw.ru>
Acked-by: Vasily Averin <vvs@sw.ru>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] ide: add sanity checking to ide taskfile ioctl
Alan Cox [Mon, 16 Oct 2006 15:31:02 +0000 (16:31 +0100)]
[PATCH] ide: add sanity checking to ide taskfile ioctl

Without this the user can feed in bogus values and get very bogus
results. Security impact is minimal as this ioctl isn't available to
unpriviledged processes anyway.

Reported to the l/k list and found with an auditing tool.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rio: fix array checking
Alan Cox [Mon, 16 Oct 2006 15:28:44 +0000 (16:28 +0100)]
[PATCH] rio: fix array checking

Found by an analysis tool and reported to the list. Fix is simple enough

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[Bluetooth] Use work queue to trigger URB submission
Marcel Holtmann [Sun, 15 Oct 2006 15:31:19 +0000 (17:31 +0200)]
[Bluetooth] Use work queue to trigger URB submission

The bcm203x firmware loading driver uses a timer to trigger the URB
submission. It is better to use a work queue instead.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
17 years ago[Bluetooth] Add locking for bt_proto array manipulation
Marcel Holtmann [Sun, 15 Oct 2006 15:31:14 +0000 (17:31 +0200)]
[Bluetooth] Add locking for bt_proto array manipulation

The bt_proto array needs to be protected by some kind of locking to
prevent a race condition between bt_sock_create and bt_sock_register.

And in addition all calls to sk_alloc need to be made GFP_ATOMIC now.

Signed-off-by: Masatake YAMATO <jet@gyve.org>
Signed-off-by: Frederik Deweerdt <frederik.deweerdt@gmail.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
17 years ago[Bluetooth] Check if DLC is still attached to the TTY
Marcel Holtmann [Sun, 15 Oct 2006 15:31:10 +0000 (17:31 +0200)]
[Bluetooth] Check if DLC is still attached to the TTY

If the DLC device is no longer attached to the TTY device, then it
makes no sense to go through with changing the termios settings.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
17 years ago[Bluetooth] Fix reference count when connection lookup fails
Marcel Holtmann [Sun, 15 Oct 2006 15:31:05 +0000 (17:31 +0200)]
[Bluetooth] Fix reference count when connection lookup fails

When the connection lookup for the device structure fails, the reference
count for the HCI device needs to be decremented.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
17 years ago[Bluetooth] Disconnect HID interrupt channel first
Marcel Holtmann [Sun, 15 Oct 2006 15:31:00 +0000 (17:31 +0200)]
[Bluetooth] Disconnect HID interrupt channel first

The Bluetooth HID specification demands that the interrupt channel
shall be disconnected first. This is needed to pass the qualification
tests.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
17 years ago[Bluetooth] Support concurrent connect requests
Marcel Holtmann [Sun, 15 Oct 2006 15:30:56 +0000 (17:30 +0200)]
[Bluetooth] Support concurrent connect requests

Most Bluetooth chips don't support concurrent connect requests, because
this would involve a multiple baseband page with only one radio. In the
case an upper layer like L2CAP requests a concurrent connect these chips
return the error "Command Disallowed" for the second request. If this
happens it the responsibility of the Bluetooth core to queue the request
and try again after the previous connect attempt has been completed.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
17 years ago[Bluetooth] Make use of virtual devices tree
Marcel Holtmann [Sun, 15 Oct 2006 15:30:50 +0000 (17:30 +0200)]
[Bluetooth] Make use of virtual devices tree

The Bluetooth subsystem currently uses a platform device for devices
with no parent. It is a better idea to use the new virtual devices
tree for these.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
17 years ago[Bluetooth] Handle return values from driver core functions
Marcel Holtmann [Sun, 15 Oct 2006 15:30:45 +0000 (17:30 +0200)]
[Bluetooth] Handle return values from driver core functions

Some return values of the driver core register and create functions
are not handled and so might cause unexpected problems.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
17 years ago[Bluetooth] Fix compat ioctl for BNEP, CMTP and HIDP
Marcel Holtmann [Sun, 15 Oct 2006 15:30:22 +0000 (17:30 +0200)]
[Bluetooth] Fix compat ioctl for BNEP, CMTP and HIDP

There exists no attempt do deal with the fact that a structure with
a uint32_t followed by a pointer is going to be different for 32-bit
and 64-bit userspace. Any 32-bit process trying to use it will be
failing with -EFAULT if it's lucky; suffering from having data dumped
at a random address if it's not.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
17 years ago[IPV6] sit: Add missing MODULE_LICENSE
Jan Dittmer [Fri, 13 Oct 2006 22:05:53 +0000 (15:05 -0700)]
[IPV6] sit: Add missing MODULE_LICENSE

This is missing the MODULE_LICENSE statements and taints the kernel
upon loading. License is obvious from the beginning of the file.

Signed-off-by: Jan Dittmer <jdi@l4x.org>
Signed-off-by: Joerg Roedel <joro-lkml@zlug.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IPV6]: Remove bogus WARN_ON in Proxy-NA handling.
YOSHIFUJI Hideaki [Fri, 13 Oct 2006 22:03:34 +0000 (15:03 -0700)]
[IPV6]: Remove bogus WARN_ON in Proxy-NA handling.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IPv6] rules: Use RT6_LOOKUP_F_HAS_SADDR and fix source based selectors
Thomas Graf [Fri, 13 Oct 2006 22:01:03 +0000 (15:01 -0700)]
[IPv6] rules: Use RT6_LOOKUP_F_HAS_SADDR and fix source based selectors

Fixes rt6_lookup() to provide the source address in the flow
and sets RT6_LOOKUP_F_HAS_SADDR whenever it is present in
the flow.

Avoids unnecessary prefix comparisons by checking for a prefix
length first.

Fixes the rule logic to not match packets if a source selector
has been specified but no source address is available.

Thanks to Kim Nordlund <kim.nordlund@nokia.com> for working
on this patch with me.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Acked-by: Ville Nuorvala <vnuorval@tcs.hut.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[XFRM]: Fix xfrm_state_num going negative.
David S. Miller [Fri, 13 Oct 2006 05:03:24 +0000 (22:03 -0700)]
[XFRM]: Fix xfrm_state_num going negative.

Missing counter bump when hashing in a new ACQ
xfrm_state.

Now that we have two spots to do the hash grow
check, break it out into a helper function.

Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NET]: reduce sizeof(struct inet_peer), cleanup, change in peer_check_expire()
Eric Dumazet [Fri, 13 Oct 2006 04:21:06 +0000 (21:21 -0700)]
[NET]: reduce sizeof(struct inet_peer), cleanup, change in peer_check_expire()

1) shrink struct inet_peer on 64 bits platforms.

17 years agoNetLabel: the CIPSOv4 passthrough mapping does not pass categories correctly
Paul Moore [Wed, 11 Oct 2006 23:10:49 +0000 (19:10 -0400)]
NetLabel: the CIPSOv4 passthrough mapping does not pass categories correctly

The CIPSO passthrough mapping had a problem when sending categories which
would cause no or incorrect categories to be sent on the wire with a packet.
This patch fixes the problem which was a simple off-by-one bug.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: James Morris <jmorris@namei.org>
17 years agoNetLabel: better error handling involving mls_export_cat()
Paul Moore [Wed, 11 Oct 2006 23:10:48 +0000 (19:10 -0400)]
NetLabel: better error handling involving mls_export_cat()

Upon inspection it looked like the error handling for mls_export_cat() was
rather poor.  This patch addresses this by NULL'ing out kfree()'d pointers
before returning and checking the return value of the function everywhere
it is called.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: James Morris <jmorris@namei.org>
17 years agoNetLabel: only deref the CIPSOv4 standard map fields when using standard mapping
Paul Moore [Wed, 11 Oct 2006 23:10:47 +0000 (19:10 -0400)]
NetLabel: only deref the CIPSOv4 standard map fields when using standard mapping

Fix several places in the CIPSO code where it was dereferencing fields which
did not have valid pointers by moving those pointer dereferences into code
blocks where the pointers are valid.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: James Morris <jmorris@namei.org>
17 years ago[BRIDGE]: flush forwarding table when device carrier off
Stephen Hemminger [Thu, 12 Oct 2006 21:45:38 +0000 (14:45 -0700)]
[BRIDGE]: flush forwarding table when device carrier off

Flush the forwarding table when carrier is lost. This helps for
availability because we don't want to forward to a downed device and
new packets may come in on other links.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: ctnetlink: Remove debugging messages
Pablo Neira Ayuso [Thu, 12 Oct 2006 21:09:16 +0000 (14:09 -0700)]
[NETFILTER]: ctnetlink: Remove debugging messages

Remove (compilation-breaking) debugging messages introduced at early
development stage.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: Update MAINTAINERS entry
Patrick McHardy [Thu, 12 Oct 2006 21:08:55 +0000 (14:08 -0700)]
[NETFILTER]: Update MAINTAINERS entry

Patches should go to myself CC netfilter-devel.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: xt_CONNSECMARK: fix Kconfig dependencies
Patrick McHardy [Thu, 12 Oct 2006 21:08:41 +0000 (14:08 -0700)]
[NETFILTER]: xt_CONNSECMARK: fix Kconfig dependencies

CONNSECMARK needs conntrack, add missing dependency to fix linking error
with CONNSECMARK=y and CONNTRACK=m.

Reported by Toralf Förster <toralf.foerster@gmx.de>.

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