[PATCH] rcu_torture_lock deadlock fix

rcu_torture_lock is used in a softirq-unsafe manner, but it is also
taken by rcu_torture_cb(), which may execute in softirq-context,
resulting in potential deadlocks.

The fix is to acquire rcu_torture_lock in a softirq-safe manner.  With
this fix applied, the rcu-torture code passes validation.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Paul E. McKenney <paulmck@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fix deadlock in drivers/pci/msi.c

The lock validator caught another one: drivers/pci/msi.c is accessing
&irq_desc[i].lock with interrupts enabled (!).

The fix is to disable interrupts properly.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fix uidhash_lock <-> RCU deadlock

RCU task-struct freeing can call free_uid(), which is taking
uidhash_lock - while other users of uidhash_lock are softirq-unsafe.

The fix is to always take the uidhash_spinlock in a softirq-safe manner.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Paul E. McKenney <paulmck@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge branch 'for-linus' of git://brick.kernel.dk/data/git/linux-2.6-block

Merge branch 'upstream-fixes' of /linux/kernel/git/jgarzik/netdev-2.6

[PATCH] Fix boot-time slowdown for measure_migration_cost

This reduces the amount of time the migration cost calculations cost
during bootup. Based on numbers by Tony Luck <tony.luck@intel.com>.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge /pub/scm/linux/kernel/git/davem/net-2.6

Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

Don't try to "validate" a non-existing timeval.

settime() with a NULL timeval is silly but legal.

Noticed by Dave Jones <davej@redhat.com>

Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Clarify help text of SKGE/SK98LIN/SKY2

Some users have commented that it is unclear which driver they should be
using for their Marvell/SysKonnect network adapter, and which ones
are/aren't interchangable.

This patch attempts to reduce the confusion.

Signed-off-by: Daniel Drake <dsd@gentoo.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

Merge branch 'upstream-fixes' of git://git./linux/kernel/git/linville/wireless-2.6

[BLOCK] A few kerneldoc fixups

Signed-off-by: Jens Axboe <axboe@suse.de>

[TCP] H-TCP: Fix accounting

This fixes the accounting in H-TCP, the ccount variable is also
adjusted a few lines above this one.

This line was not supposed to be there and wasn't there in the patches
originally submitted, the four patches submitted were merged to one
and in that merge the bug was introduced.

Signed-Off-By: Baruch Even <baruch@ev-en.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4] igmp: remove pointless printk

This is easily triggerable by sending bogus packets,
allowing a malicious user to flood remote logs.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC]: Fix compile failures in math-emu.

Kill debugging default switch cases in do_one_mathemu().
That case is handled properly already and gcc hates
the empty statement that results when the debug code is
disabled.

Pointed out by kaffe.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: heartbeats exceed maximum retransmssion limit

The number of HEARTBEAT chunks that an association may transmit is
limited by Association.Max.Retrans count; however, the code allows
us to send one extra heartbeat.

This patch limits the number of heartbeats to the maximum count.

Signed-off-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>

[SCTP]: correct the number of INIT retransmissions

We currently count the initial INIT/COOKIE_ECHO chunk toward the
retransmit count and thus sends a total of sctp_max_retrans_init chunks.
The correct behavior is to retransmit the chunk sctp_max_retrans_init in
addition to sending the original.

Signed-off-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>

[PATCH] PCMCIA=m, HOSTAP_CS=y is not a legal configuration

CONFIG_PCMCIA=m, CONFIG_HOSTAP_CS=y doesn't compile.

Reported by "Gabriel C." <crazy@pimpmylinux.org>.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] Typo corrections for ieee80211

This patch, generated against 2.6.16-rc1-git4, corrects two typographical
errors in ieee80211_rx.c and adds the facility name to a bare printk.

Signed-Off-By: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] ipw2200: Fix a variable referenced after kfree() bug

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] ipw2200: Fix sw_reset doesn't clear the static essid problem

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] ipw2200: Fix "iwspy ethx off" causes kernel panic

Signed-off-by: Hong Liu <hong.liu@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] ipw2100: Fix setting txpower failed problem

The ipw2100 driver misunderstood the parameter of txpower.
Tx Power off means turn off the radio, but the driver interpret it as
"can't set txpower". So when getting the txpower, it sets disabled=1 to
the iwconifg tool in managed mode. And the tool will display "Tx Power off"
when disabled=1.

Now, in managed mode, iwconfig will not show "TX Power" if the radio is not
switched off. It will only display "Tx Power off" only if the radio is killed.

Signed-off-by: Hong Liu <hong.liu@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] ipw2100: Fix a gcc compile warning

drivers/net/wireless/ipw2100.c:2236: warning: `ipw2100_match_buf' defined
but not used

Cc: Yi Zhu <yi.zhu@intel.com>
Cc: James Ketrenos <jketreno@linux.intel.com>
Cc: Jeff Garzik <jgarzik@pobox.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[LIBATA] Blacklist certain Maxtor firmware revisions for FUA support

It looks like they are either discarding or corrupting data when the FUA
command is used, bad.

Signed-off-by: Jens Axboe <axboe@suse.de>

[SPARC64]: Kill compat_sys_clock_settime sign extension stub.

It's wrong and totally unneeded.

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'upstream-fixes' of git://git./linux/kernel/git/linville/wireless-2.6

[PATCH] ieee80211: Fix A band min and max channel definitions

Signed-off-by: Hong Liu <hong.liu@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] ieee80211: Fix iwlist scan can only show about 20 APs

Limit the amount of output given to iwlist scan.

Signed-off-by: Hong Liu <hong.liu@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] ieee80211: Fix problem with not decrypting broadcast packets

The code for pulling the key to use for decrypt was correctly using
the host_mc_decrypt flag.  The code that actually decrypted,
however, was based on host_decrypt.  This patch changes this
behavior.

Signed-off-by: Etay Bogner <etay.bogner@gmail.com>
Signed-off-by: James Ketrenos <jketreno@linux.intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] orinoco_cs: tweak Vcc debugging messages

The current orinoco_cs.c can issue the exact same error message for
2 different tests that can fail.  Alter them so we can tell which
one of the two failed.

Signed-off-by: Valdis Kletnieks <valdis.kletnieks@vt.edu>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[PATCH] mv643xx_eth: Fix for building as a module

Enable mv643xx_eth driver to work when built as a module on
mv64x60-based embedded systems.

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Whitespace cleanup

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Update dev->last_rx on packet receive

Update dev->last_rx on packet receive

This fix corrects errors seen during configuration of the bonding driver.

Signed-off-by: Paolo Galtieri <pgaltieri@mvista.com>
Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Fix spinlock recursion bug

This patch eliminates a spinlock recursion bug introduced recently.
Since eth_port_send() is always called with the lock held, we simply
remove the locking inside the function itself.

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] s2io: scatter-gather fix

There is a problem with fragmented skb in s2io driver version 2.0.9.4
available in 2.6.16-rc1 kernel. The adapter will fail to transmit if
any scatter-gather skb arrives.  This patch provides fix for the above
described problem.

Signed-off-by: Ananda Raju <ananda.raju@neterion.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] b44: fix laptop carrier detect

On my laptop, the b44 device is created and the carrier state defaults
to ON when created by alloc_etherdev. This means tools like NetworkManager
see the carrier as On and try and bring the device up.  The correct thing
to do is mark the carrier as Off when device is created.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] acenic: fix checking of read_eeprom_byte() return values

tmp in ace_init is u32 thus rendering read_eeprom_byte() return values
checks useless.

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] bonding: fix ->get_settings error checking

Since get_settings() returns a signed int and it gets checked
for < 0 to catch an error, res should be a signed int too.

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[SCSI] Remove host template ordered_flush variable

After the recent overhaul of the block layer the variable
"ordered_flush" is no longer used.

Signed-off-by: Tetsuo Takata <takatatt@intellilink.co.jp>
Signed-off-by: Jens Axboe <axboe@suse.de>

[IPV6] MLDv2: fix change records when transitioning to/from inactive

The following patch fixes these problems in MLDv2:

1) Add/remove "delete" records for sending change reports when
        addition of a filter results in that filter transitioning to/from
        inactive. [same as recent IPv4 IGMPv3 fix]
2) Remove 2 redundant "group_type" checks (can't be IPV6_ADDR_ANY
        within that loop, so checks are always true)
3) change an is_in() "return 0" to "return type == MLD2_MODE_IS_INCLUDE".
        It should always be "0" to get here, but it improves code locality
        to not assume it, and if some race allowed otherwise, doing
        the check would return the correct result.

Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[AF_KEY]: no message type set

When returning a message to userspace in reply to a SADB_FLUSH or
SADB_X_SPDFLUSH message, the type was not set for the returned PFKEY
message. The patch below corrects this problem.

Signed-off-by: Jerome Borsboom <j.borsboom@erasmusmc.nl>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BONDING]: Remove CAP_NET_ADMIN requirement for INFOQUERY ioctl

This information is already available via /proc/net/bonding/*
therefore it doesn't make sense to require CAP_NET_ADMIN
privileges.

Original patch by Laurent Deniel <laurent.deniel@free.fr>

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BLOCK] ll_rw_blk: fix setting of ->ordered on init

This makes XFS barrier mounts succeed on my SCSI system.

Signed-off-by: Tetsuo Takata <takatatt@intellilink.co.jp>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] elevator: allow default scheduler to potentially be modular

Jens has decided that allowing the default scheduler to be a module is
a bug, and should not be allowed under kconfig.  However, I find that
scenario useful for debugging, and wish for the kernel to be able to
handle this situation without OOPSing, if I enable such an option in
the .config directly.  This patch dynamically checks for the presence
of the compiled-in default, and falls back to no-op, emitting a
suitable error message, when the default is not available

Tested for a range of boot options on 2.6.16-rc1-mm2.

Signed-off-by: Nate Diller <nate.diller@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] elevator: default choice selection

My previous default iosched patch did a poor job dealing with the
'elevator=' boot-time option.  The old behavior falls back to the
compiled-in default if the requested one is not registered at boot
time.  This patch dynamically evaluates which default
to use, and emits a suitable error message when the requested scheduler
is not available.  It also does the 'as' -> 'anticipatory' conversion
before elevator registration, which along with a modified registration
function, allows it to correctly indicate which default scheduler is
in use.

Tested for a range of boot options on 2.6.16-rc1-mm2.

Signed-off-by: Nate Diller <nate.diller@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] ll_rw_blk: use preempt-disabling disk_stat_add() in completion

It can legally be called with interrupts/preemption enabled.

Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] ll_rw_blk: make max_sectors and max_hw_sectors unsigned ints

IDE lba48 can support full 64k request size, which overflows the
max_hw_sectors variable.

Signed-off-by: Jens Axboe <axboe@suse.de>

[SPARC64]: Implement __raw_read_trylock()

generic__raw_read_trylock() just does a raw_read_lock() so that
isn't very useful.

Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Fix skb fclone error path handling.

On the error path if we allocated an fclone then we will free it in
the wrong pool.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SUNGEM]: Make PM of PHYs more reliable (#2)

On my latest laptop, I've had occasional PHY dead on wakeup from
sleep... the PHY would be totally unresponsive even to toggling the hard
reset line until the machine is powered down... Looking closely at the
code, I found some possible issues in the way we setup the MDIO lines
during suspend along with slight divergences from what Darwin does when
resetting it that may explain the problem. That patch change these and
the problem appear to be gone for me at least... I also fixed an mdelay
-> msleep while I was at it to the pmac feature code that is called
when toggling the PHY reset line since sungem doesn't call it in an
atomic context anymore.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>b
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Fix some whitespace issues in af_packet.c

Signed-off-by: Kris Katterjohn <kjak@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: more whitespace issues in net/core/filter.c

This fixes some whitespace issues in net/core/filter.c

Signed-off-by: Kris Katterjohn <kjak@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC]: Increase NR_SYSCALLS to 299

To let new syscalls through.

Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Unbreak x-tables on x86.

x86 defines __alignof__(long long) as 8 yet it gives 4
for a struct containing a long long, ho hum... so my
simplified form doesn't work everywhere.

So use Harald Welte's original patch, which should work
on all platforms.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Use compat_sys_futimesat in 32-bit syscall table.

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge /pub/scm/linux/kernel/git/davem/net-2.6

Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

[PATCH] Fix regression added by ppoll/pselect code.

The compat layer timeout handling changes in:

9f72949f679df06021c9e43886c9191494fdb007

are busted.  This is most easily seen with an X application
that uses sub-second select/poll timeout such as emacs.  You
hit a key and it takes a second or so before the app responds.

The two ROUND_UP() calls upon entry are using {tv,ts}_sec where it
should instead be using {tv_usec,ts_nsec}, which perfectly explains
the observed incorrect behavior.

Another bug shot down with git bisect.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[NETFILTER] x_tables: Make XT_ALIGN align as strictly as necessary.

Or else we break on ppc32 and other 32-bit platforms.

Based upon a patch from Harald Welte.

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge /pub/scm/linux/kernel/git/sridhar/lksctp-2.6

[IA64] eliminate softlockup warning

Fix an unnecessary softlockup watchdog warning in the ia64
uncached_build_memmap() that occurs occasionally at 256p and always at
512p.  The problem occurs at boot time.

Signed-off-by: John Hawkes <hawkes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Tony Luck <tony.luck@intel.com>

[IA64] sem2mutex: arch/ia64/kernel/perfmon.c

Migrate perfmon from using an old semaphore to a completion handler.

Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Tony Luck <tony.luck@intel.com>

[IA64] sem2mutex: arch/ia64/ia32/sys_ia32.c

Migrate arch/ia64/ia32/sys_ia32 to using a mutex for mmap protection.

Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Tony Luck <tony.luck@intel.com>

[SPARC]: Add support for *at(), ppoll, and pselect syscalls.

This also includes by necessity _TIF_RESTORE_SIGMASK support,
which actually resulted in a lot of cleanups.

The sparc signal handling code is quite a mess and I should
clean it up some day.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC]: sparc32 needs PROMDEV_{I,O}RSC defines too.

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge /pub/scm/linux/kernel/git/sfrench/cifs-2.6

[PATCH] tlclk driver update

some driver clean ups, and a re-posting of changes that are needed

to match the updated TPS.

Signed-off-by: Mark Gross <mark.gross@intel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] EDAC: core EDAC support code

This is a subset of the bluesmoke project core code, stripped of the NMI work
which isn't ready to merge and some of the "interesting" proc functionality
that needs reworking or just has no place in kernel.  It requires no core
kernel changes except the added scrub functions already posted.

The goal is to merge further functionality only after the core code is
accepted and proven in the base kernel, and only at the point the upstream
extras are really ready to merge.

From: doug thompson <norsk5@xmission.com>

  This converts EDAC to sysfs and is the final chunk neccessary before EDAC
  has a stable user space API and can be considered for submission into the
  base kernel.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Signed-off-by: doug thompson <norsk5@xmission.com>
Signed-off-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] EDAC: drivers for Radisys 82600

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] EDAC: drivers for Intel i82860, i82875

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] EDAC: drivers for AMD 76x and Intel E750x, E752x

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] EDAC: atomic scrub operations

EDAC requires a way to scrub memory if an ECC error is found and the chipset
does not do the work automatically.  That means rewriting memory locations
atomically with respect to all CPUs _and_ bus masters.  That means we can't
use atomic_add(foo, 0) as it gets optimised for non-SMP

This adds a function to include/asm-foo/atomic.h for the platforms currently
supported which implements a scrub of a mapped block.

It also adjusts a few other files include order where atomic.h is included
before types.h as this now causes an error as atomic_scrub uses u32.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Add pselect/ppoll system calls on i386

Add the sys_pselect6() and sys_poll() calls to the i386 syscall table.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Add pselect/ppoll system call implementation

The following implementation of ppoll() and pselect() system calls
depends on the architecture providing a TIF_RESTORE_SIGMASK flag in the
thread_info.

These system calls have to change the signal mask during their
operation, and signal handlers must be invoked using the new, temporary
signal mask. The old signal mask must be restored either upon successful
exit from the system call, or upon returning from the invoked signal
handler if the system call is interrupted. We can't simply restore the
original signal mask and return to userspace, since the restored signal
mask may actually block the signal which interrupted the system call.

The TIF_RESTORE_SIGMASK flag deals with this by causing the syscall exit
path to trap into do_signal() just as TIF_SIGPENDING does, and by
causing do_signal() to use the saved signal mask instead of the current
signal mask when setting up the stack frame for the signal handler -- or
by causing do_signal() to simply restore the saved signal mask in the
case where there is no handler to be invoked.

The first patch implements the sys_pselect() and sys_ppoll() system
calls, which are present only if TIF_RESTORE_SIGMASK is defined. That
#ifdef should go away in time when all architectures have implemented
it. The second patch implements TIF_RESTORE_SIGMASK for the PowerPC
kernel (in the -mm tree), and the third patch then removes the
arch-specific implementations of sys_rt_sigsuspend() and replaces them
with generic versions using the same trick.

The fourth and fifth patches, provided by David Howells, implement
TIF_RESTORE_SIGMASK for FR-V and i386 respectively, and the sixth patch
adds the syscalls to the i386 syscall table.

This patch:

Add the pselect() and ppoll() system calls, providing core routines usable by
the original select() and poll() system calls and also the new calls (with
their semantics w.r.t timeouts).

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Michael Kerrisk <mtk-manpages@gmx.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml: use generic sys_rt_sigsuspend

Use the generic sys_rt_sigsuspend.

Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml: add TIF_RESTORE_SIGMASK support

Add support for TIF_RESTORE_SIGMASK.  I copy the i386 handling of the flag.
sys_sigsuspend is also changed to follow i386.
Also a bit of cleanup -
   turn an if into a switch
   get rid of a couple more emacs formatting comments

Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] TIF_RESTORE_SIGMASK support for arch/powerpc

Implement the TIF_RESTORE_SIGMASK flag in the new arch/powerpc kernel, for
both 32-bit and 64-bit system call paths.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Handle TIF_RESTORE_SIGMASK for i386

Handle TIF_RESTORE_SIGMASK as added by David Woodhouse's patch entitled:

        [PATCH] 2/3 Add TIF_RESTORE_SIGMASK support for arch/powerpc
        [PATCH] 3/3 Generic sys_rt_sigsuspend

It does the following:

 (1) Declares TIF_RESTORE_SIGMASK for i386.

 (2) Invokes it over to do_signal() when TIF_RESTORE_SIGMASK is set.

 (3) Makes do_signal() support TIF_RESTORE_SIGMASK, using the signal mask saved
     in current->saved_sigmask.

 (4) Discards sys_rt_sigsuspend() from the arch, using the generic one instead.

 (5) Makes sys_sigsuspend() save the signal mask and set TIF_RESTORE_SIGMASK
     rather than attempting to fudge the return registers.

 (6) Makes sys_sigsuspend() return -ERESTARTNOHAND rather than looping
     intrinsically.

 (7) Makes setup_frame(), setup_rt_frame() and handle_signal() return 0 or
     -EFAULT rather than true/false to be consistent with the rest of the
     kernel.

Due to the fact do_signal() is then only called from one place:

 (8) Makes do_signal() no longer have a return value is it was just being
     ignored; force_sig() takes care of this.

 (9) Discards the old sigmask argument to do_signal() as it's no longer
     necessary.

(10) Makes do_signal() static.

(11) Marks the second argument to do_notify_resume() as unused. The unused
     argument should remain in the middle as the arguments are passed in as
     registers, and the ordering is specific in entry.S

Given the way do_signal() is now no longer called from sys_{,rt_}sigsuspend(),
they no longer need access to the exception frame, and so can just take
arguments normally.

This patch depends on sys_rt_sigsuspend patch.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Handle TIF_RESTORE_SIGMASK for FRV

Handle TIF_RESTORE_SIGMASK as added by David Woodhouse's patch entitled:

        [PATCH] 2/3 Add TIF_RESTORE_SIGMASK support for arch/powerpc
        [PATCH] 3/3 Generic sys_rt_sigsuspend

It does the following:

 (1) Declares TIF_RESTORE_SIGMASK for FRV.

 (2) Invokes it over to do_signal() when TIF_RESTORE_SIGMASK is set.

 (3) Makes do_signal() support TIF_RESTORE_SIGMASK, using the signal mask saved
     in current->saved_sigmask.

 (4) Discards sys_rt_sigsuspend() from the arch, using the generic one instead.

 (5) Makes sys_sigsuspend() save the signal mask and set TIF_RESTORE_SIGMASK
     rather than attempting to fudge the return registers.

 (6) Makes sys_sigsuspend() return -ERESTARTNOHAND rather than looping
     intrinsically.

 (7) Makes setup_frame(), setup_rt_frame() and handle_signal() return 0 or
     -EFAULT rather than true/false to be consistent with the rest of the
      kernel.

Due to the fact do_signal() is then only called from one place:

 (8) Make do_signal() no longer have a return value is it was just being
     ignored; force_sig() takes care of this.

 (9) Discards the old sigmask argument to do_signal() as it's no longer
     necessary.

This patch depends on the FRV signalling patches as well as the
sys_rt_sigsuspend patch.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Generic sys_rt_sigsuspend()

The TIF_RESTORE_SIGMASK flag allows us to have a generic implementation of
sys_rt_sigsuspend() instead of duplicating it for each architecture.  This
provides such an implementation and makes arch/powerpc use it.

It also tidies up the ppc32 sys_sigsuspend() to use TIF_RESTORE_SIGMASK.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] vfs: *at functions: x86_64

Wire up the x86_64 syscalls.

Signed-off-by: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] vfs: *at functions: i386

Wire up the x86 syscalls

Signed-off-by: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] vfs: *at functions: core

Here is a series of patches which introduce in total 13 new system calls
which take a file descriptor/filename pair instead of a single file
name.  These functions, openat etc, have been discussed on numerous
occasions.  They are needed to implement race-free filesystem traversal,
they are necessary to implement a virtual per-thread current working
directory (think multi-threaded backup software), etc.

We have in glibc today implementations of the interfaces which use the
/proc/self/fd magic.  But this code is rather expensive.  Here are some
results (similar to what Jim Meyering posted before).

The test creates a deep directory hierarchy on a tmpfs filesystem.  Then
rm -fr is used to remove all directories.  Without syscall support I get
this:

real    0m31.921s
user    0m0.688s
sys     0m31.234s

With syscall support the results are much better:

real    0m20.699s
user    0m0.536s
sys     0m20.149s

The interfaces are for obvious reasons currently not much used.  But they'll
be used.  coreutils (and Jeff's posixutils) are already using them.
Furthermore, code like ftw/fts in libc (maybe even glob) will also start using
them.  I expect a patch to make follow soon.  Every program which is walking
the filesystem tree will benefit.

Signed-off-by: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@ftp.linux.org.uk>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: Michael Kerrisk <mtk-manpages@gmx.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] exportfs: add find_acceptable_alias helper

find_exported_dentry contains two duplicate loops to find an alias that the
acceptable callback likes.  Split this out to a new helper and switch from
list_for_each to list_for_each_entry to make it more readable.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: NeilBrown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] knfsd: Provide missing NFSv2 part of patch for checking vfs_getattr.

A recent patch which checked the return status of vfs_getattr in nfsd,
completely missed the nfsproc.c (NFSv2) part.  Here is it.

This patch moved the call to vfs_getattr from the xdr encoding (at which point
it is too late to return an error) to the call handling.  This means several
calls to vfs_getattr are needed in nfsproc.c.  Many are encapsulated in
nfsd_return_attrs and nfsd_return_dirop.

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] knfsd: Fix some more errno/nfserr confusion in vfs.c

nfsd_sync* return an errno, which usually needs to be converted to an errno,
sometimes immediately, sometimes a little later.

Also, nfsd_setattr returns an nfserr which SHOULDN'T be converted from
an errno (because it isn't one).

Also some tidyups of the form:
  err = XX
  err = nfserrno(err)
and
  err = XX
  if (err)
      err = nfserrno(err)
become
  err = nfserrno(XX)

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4_lock() returns bogus values to clients

missing nfserrno() in default case of a switch by return value of
posix_lock_file(); as the result we send negative host-endian to clients that
expect positive network-endian, preferably mentioned in RFC...  BTW, that case
is not impossible - posix_lock_file() can return -ENOLCK and we do not handle
that one explicitly.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] NFSERR_SERVERFAULT returned host-endian

->rp_status is network-endian and nobody byteswaps it before sending to
client; putting NFSERR_SERVERFAULT instead of nfserr_serverfault in there is
not nice...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4_truncate() bogus return value

-EINVAL (in host order, no less) is not a good thing to return to client.

nfsd4_truncate() returns it in one case and its callers expect nfs_....  from
it.  AFAICS, it should be nfserr_inval

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd/vfs.c: endianness fixes

Several failure exits return -E<something> instead of nfserr_<something> and
vice versa.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd4: clean up settattr code

Clean up some unnecessary special-casing in the setattr code..

Signed-off-by: Fred Isaman <iisaman@citi.umich.edu>
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>

[PATCH] nfsd4: Fix bug in rdattr_error return

Fix bug in rdattr_error return which causes correct error code to be
overwritten by nfserr_toosmall.

Signed-off-by: Fred Isaman <iisaman@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>

[PATCH] nfsd4: fix open_downgrade

Bad bookkeeping of the share reservations when handling open upgrades was
causing open downgrade to fail.

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>

[PATCH] nfsd4: don't create on open that fails due to ERR_GRACE

In an earlier patch (commit b648330a1d741d5df8a5076b2a0a2519c69c8f41) I noted
that a too-early grace-period check was preventing us from bumping the
sequence id on open.  Unfortunately in that patch I stupidly moved the
grace-period check back too far, so now an open for create can succesfully
create the file while still returning ERR_GRACE.

The correct place for that check is after we've set the open_owner and handled
any replays, but before we actually start mucking with the filesystem.

Thanks to Avishay Traeger 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>

[PATCH] nfsd4: simplify process-open1 logic

nfsd4_process_open1 is very highly nested; flatten it out a bit.

Also, the preceding comment, which just outlines the logic, seems redundant.

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>

[PATCH] nfsd4: nfs4state.c miscellaneous goto removals

Remove some goto's that made the logic here a little more tortuous than
necessary.

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>

[PATCH] nfsd4: no replays on unconfirmed owners

We shouldn't check for replays until after checking whether the open owner is
confirmed.  Clients are allowed to reuse openowners without bumping the seqid.

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>

[PATCH] nfsd4: handle replays of failed open reclaims

We need to make sure open reclaims are marked confirmed immediately so that we
can handle replays even if they fail (e.g.  with a seqid-incrementing error).
(See 8.1.8.)

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>