Original patch and description from Neil Brown <neilb@suse.de>,
merged and adapted to splice branch by me. Neils text follows:

__generic_file_splice_read() currently samples the i_size at the start
and doesn't do so again unless it needs to call ->readpage to load
a page.  After ->readpage it has to re-sample i_size as a truncate
may have caused that page to be filled with zeros, and the read()
call should not see these.

However there are other activities that might cause ->readpage to be
called on a page between the time that __generic_file_splice_read()
samples i_size and when it finds that it has an uptodate page. These
include at least read-ahead and possibly another thread performing a
read

So we must sample i_size *after* it has an uptodate page.  Thus the
current sampling at the start and after a read can be replaced with a
sampling before page addition into spd.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

__generic_file_splice_read's partial page check, at eof after readpage,
not only got its calculations wrong, but also reused the loff variable:
causing data corruption when splicing from a non-0 offset in the file's
last page (revealed by ext2 -b 1024 testing on a loop of a tmpfs file).
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

I've seen inode related deadlocks, so move this call outside of the
actor itself, which may hold the inode lock.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

There's really no reason it's below the first use of the pointer
type, and it'll fail compilation for the network addition (for good
reason). So move it up a bit.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

It's only supposed to be used by do_sendfile(), which is never
modular. So kill the export.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

This bug was caught by LTP testcase fchmod06 on Blackfin platform.

In the manpage of fchmod, "EPERM: The effective UID does not match the
owner of the file, and the process is not privileged (Linux: it does not
have the CAP_FOWNER capability)."

But the ramfs nommu code missed the inode_change_ok POSIX UID/GID
verification. This patch fixed this.
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* git://git.linux-xtensa.org/kernel/xtensa-feed:
  Xtensa: use asm-generic/fcntl.h
  [XTENSA] Remove non-rt signal handling
  [XTENSA] Move common sections into bss sections
  [XTENSA] clean-up header files
  [XTENSA] Use generic 64-bit division
  [XTENSA] Remove multi-exported symbols from xtensa_ksyms.c
  [XTENSA] fix sources using deprecated assembler directive
  [XTENSA] Spelling fixes in arch/xtensa
  [XTENSA] fix bit operations in bitops.h

* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6:
  [SPARC64]: Fix SBUS IRQ regression caused by PCI-E driver.
  [SPARC64]: Fix 2 bugs in PCI Sabre bus scanning.

* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6: (24 commits)
  xfrm: Add security check before flushing SAD/SPD
  [NET_SCHED]: Fix filter double free
  [NET]: Avoid duplicate netlink notification when changing link state
  [UDP]: Revert 2-pass hashing changes.
  [AF_UNIX]: Fix stream recvmsg() race.
  [NETFILTER]: nf_conntrack_amanda: fix textsearch_prepare() error check
  [NETFILTER]: ip_tables: fix compat related crash
  [NETFILTER]: nf_conntrack: fix helper module unload races
  [RTNETLINK]: ifindex 0 does not exist
  [NETLINK]: Mark netlink policies const
  [TCP] tcp_probe: Attach printf attribute properly to printl().
  [TCP]: Use LIMIT_NETDEBUG in tcp_retransmit_timer().
  [NET]: Merge dst_discard_in and dst_discard_out.
  [RFKILL]: Make rfkill->name const
  [IPV4]: Restore old behaviour of default config values
  [IPV4]: Add default config support after inetdev_init
  [IPV4]: Convert IPv4 devconf to an array
  [IPV4]: Only panic if inetdev_init fails for loopback
  [TCP]: Honour sk_bound_dev_if in tcp_v4_send_ack
  [BNX2]: Update version and reldate.
  ...

This is a minor fix, but what is currently there is essentially wrong.
In do_page_fault, if the faulting address from user code happens to be
in kernel address space (int *p = (int*)-1; p = 0xbed;)  then the
do_page_fault handler will jump over the local_irq_enable with the

  goto bad_area_nosemaphore;

But the first line there sees this is user code and goes through the
process of sending a signal to send SIGSEGV to the user task. This whole
time interrupts are disabled and the task can not be preempted by a
higher priority task.

This patch always enables interrupts in the user path of the
bad_area_nosemaphore.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] pxa: fix pxa27x keyboard driver
  [ARM] Fix 4417/1: Serial: Fix AMBA drivers locking
  [ARM] 4421/1: AT91: Value of _KEY fields.
  [ARM] Solve buggy smp_processor_id() usage
  [ARM] 4422/1: Fix default value handling in gpio_direction_output (PXA)
  [ARM] 4419/1: AT91: SAM9 USB clocks check for suspending
  [ARM] 4418/1: AT91: Number of programmable clocks differs
  [ARM] 4392/2: Do not corrupt the SP register in compressed/head.S

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] Fix warning by moving do_default_vi into CONFIG_CPU_MIPSR2_SRS
  [MIPS] Fix some minor typoes in arch/mips/Kconfig.
  [MIPS] Remove prototype for deleted function qemu_handle_int
  [MIPS] Fix some system calls with long long arguments
  [MIPS] Make dma_map_sg handle sg elements which are longer than one page
  [MIPS] Drop __ARCH_WANT_SYS_FADVISE64
  [MIPS] Fix VGA corruption on RM300C
  [MIPS] RM300: Fix MMIO problems by marking the PCI INT ACK region busy
  [MIPS] EMMA2RH: remove dead KGDB code
  [MIPS] Remove duplicate fpu enable hazard code.
  [MIPS] Atlas, Malta, SEAD: Remove scroll from interrupt handler.

In file included from /usr/src/linux-2.6-2/net/ipv4/ip_input.c:118:

  include2/asm/system.h:245: error: parse error before "__cmpxchg_32"
  include2/asm/system.h:245: error: parse error before '*' token
  include2/asm/system.h:245: warning: type defaults to `int' in declaration of `__cmpxchg_32'
  include2/asm/system.h:245: warning: function declaration isn't a prototype
  include2/asm/system.h:245: warning: data definition has no type or storage class
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

We used to access the 64-bit IRQ IMAP and ICLR registers of bus
controllers 4-bytes in and as a 32-bit register word, since only the
low 32-bits were relevant.  This seemed like a good idea at the time.

But the PCI-E controller requires full 8-byte 64-bit access to
these registers, so we switched over to accessing them fully.

SBUS was not adjusted properly, which broke interrupts completely.
Signed-off-by: David S. Miller <davem@davemloft.net>

If we are on hummingbird, bus runs at 66MHZ.

pbm->pci_bus should be setup with the result of pci_scan_one_pbm()
or else we deref NULL pointers in the error interrupt handlers.
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently we check for permission before deleting entries from SAD and
SPD, (see security_xfrm_policy_delete() security_xfrm_state_delete())
However we are not checking for authorization when flushing the SPD and
the SAD completely. It was perhaps missed in the original security hooks
patch.

This patch adds a security check when flushing entries from the SAD and
SPD.  It runs the entire database and checks each entry for a denial.
If the process attempting the flush is unable to remove all of the
entries a denial is logged the the flush function returns an error
without removing anything.

This is particularly useful when a process may need to create or delete
its own xfrm entries used for things like labeled networking but that
same process should not be able to delete other entries or flush the
entire database.

Signed-off-by: Joy Latten<latten@austin.ibm.com>
Signed-off-by: Eric Paris <eparis@parisplace.org>
Signed-off-by: James Morris <jmorris@namei.org>

cbq and atm destroy their filters twice when destroying inner classes
during qdisc destruction.
Reported-and-tested-by: Strobl Anton <a.strobl@aws-it.at>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

When changing the link state from userspace not affecting any other
flags. Two duplicate notification are being sent, once as action
in the NETDEV_UP/NETDEV_DOWN notification chain and a second time
when comparing old and new device flags after the change has been
completed. Although harmless, the duplicates should be avoided.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

This reverts changesets:

6aaf47fa
b7b5f487
de34ed91
fc038410

There are still some correctness issues recently
discovered which do not have a known fix that doesn't
involve doing a full hash table scan on port bind.

So revert for now.
Signed-off-by: David S. Miller <davem@davemloft.net>

A recv() on an AF_UNIX, SOCK_STREAM socket can race with a
send()+close() on the peer, causing recv() to return zero, even though
the sent data should be received.

This happens if the send() and the close() is performed between
skb_dequeue() and checking sk->sk_shutdown in unix_stream_recvmsg():

process A  skb_dequeue() returns NULL, there's no data in the socket queue
process B  new data is inserted onto the queue by unix_stream_sendmsg()
process B  sk->sk_shutdown is set to SHUTDOWN_MASK by unix_release_sock()
process A  sk->sk_shutdown is checked, unix_release_sock() returns zero

I'm surprised nobody noticed this, it's not hard to trigger.  Maybe
it's just (un)luck with the timing.

It's possible to work around this bug in userspace, by retrying the
recv() once in case of a zero return value.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>

The return value from textsearch_prepare() needs to be checked
by IS_ERR(). Because it returns error code as a pointer.

Cc: "Brian J. Murrell" <netfilter@interlinx.bc.ca>
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

check_compat_entry_size_and_hooks iterates over the matches and calls
compat_check_calc_match, which loads the match and calculates the
compat offsets, but unlike the non-compat version, doesn't call
->checkentry yet. On error however it calls cleanup_matches, which in
turn calls ->destroy, which can result in crashes if the destroy
function (validly) expects to only get called after the checkentry
function.

Add a compat_release_match function that only drops the module reference
on error and rename compat_check_calc_match to compat_find_calc_match to
reflect the fact that it doesn't call the checkentry function.

Reported by Jan Engelhardt <jengelh@linux01.gwdg.de>
Signed-off-by: Dmitry Mishin <dim@openvz.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

When a helper module is unloaded all conntracks refering to it have their
helper pointer NULLed out, leading to lots of races. In most places this
can be fixed by proper use of RCU (they do already check for != NULL,
but in a racy way), additionally nf_conntrack_expect_related needs to
bail out when no helper is present.

Also remove two paranoid BUG_ONs in nf_conntrack_proto_gre that are racy
and not worth fixing.
Signed-off-by: Patrick McHarrdy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

ifindex == 0 does not exist and implies we should do a lookup by name if
one was given.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

GCC doesn't like the way Stephen initially did it:

net/ipv4/tcp_probe.c:83: warning: empty declaration
Signed-off-by: David S. Miller <davem@davemloft.net>

LIMIT_NETDEBUG allows the admin to disable some warning messages (echo 0
 >/proc/sys/net/core/warnings).

The "TCP: Treason uncloaked!" message can use this facility.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Denis Cheng <crquan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The rfkill name can be made const safely,
this makes the compiler happy when drivers make
it point to some const string used elsewhere.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Previously inet devices were only constructed when addresses are added
(or rarely in ipmr).  Therefore the default config values they get are
the ones at the time of these operations.

Now that we're creating inet devices earlier, this changes the
behaviour of default config values in an incompatible way (see bug
#8519).

This patch creates a compromise by setting the default values at the
same point as before but only for those that have not been explicitly
set by the user since the inet device's creation.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

Previously once inetdev_init has been called on a device any changes
made to ipv4_devconf_dflt would have no effect on that device's
configuration.

This creates a problem since we have moved the point where
inetdev_init is called from when an address is added to where the
device is registered.

This patch is the first half of a set that tries to mimic the old
behaviour while still calling inetdev_init.

It propagates any changes to ipv4_devconf_dflt to those devices that
have not had the corresponding attribute set.

The next patch will forcibly set all values at the point where
inetdev_init was previously called.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch converts the ipv4_devconf config members (everything except
sysctl) to an array.  This allows easier manipulation which will be
needed later on to provide better management of default config values.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

When I made the inetdev_init call work on all devices I incorrectly
left in the panic call as well.  It is obviously undesirable to
panic on an allocation failure for a normal network device.  This
patch moves the panic call under the loopback if clause.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

A time_wait socket inherits sk_bound_dev_if from the original socket,
but it is not used when sending ACK packets using ip_send_reply.

Fix by passing the oif to ip_send_reply in struct ip_reply_arg and
use it for output routing.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

Update to version 1.5.11.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

The statistics block DMA on 5708 can be messed up occasionally on the
average of about once per hour.  If the user is reading the counters
within one second after the corruption, the counters will be all
messed up.  One second later, the counters will be ok again until the
next corruption occurs.

The workaround is to disable the periodic statistics DMA.  Instead,
we manually trigger the DMA once a second in bnx2_timer().  This
manual trigger of the DMA avoids the problem.

As a consequence, we can only allow 0 or 1 second settings for
ethtool -C statistics block.

Thanks to Jean-Daniel Pauget <jd@disjunkt.com> and
CaT <cat@zip.com.au> for reporting this rare problem.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add missing code to enable DMA on 5709 A1.  The bit is a no-op on A0
and therefore can be set on all 5709 chips.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

For correctness, we need to wait for the MEM_INIT bit to be cleared
in the BNX2_CTX_COMMAND register before proceeding.

[Added return -EBUSY when the MEM_INIT bit doesn't clear, suggested
by Jeff Garzik.]
Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

There's a bug in the driver that only initializes half of the context
memory on the 5708.  Surprisingly, this works most of the time except
for some occasional netdev watchdogs when sending a lot of 64-byte
packets.  The fix is to add the missing code to initialize the 2nd
halves of all context memory.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>