[PATCH] IPMI: allow user to override the kernel IPMI daemon enable

After the previous patch to disable the kernel IPMI daemon if interrupts
were available, the issue of broken hardware was raised, and a reasonable
request to add an override was mode.  So here it is.

Allow the user to force the kernel ipmi daemon on or off.  This way,
hardware with broken interrupts or users that are not concerned with
performance can turn it on or off to their liking.

[akpm@osdl.org: save 4 bytes in vmlinux]
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Add Pekka to CREDITS

Whomever said...

  "When you meet someone now who is writing a compiler or hacking a Unix
   kernel, at least you know they're not just doing it to pick up chicks."

...has obviously never met a _Linux_ kernel hacker.

Anyway, sometimes people confuse my email addresses, which is why I
really should add the proper one to CREDITS ;-).

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] restore parport_pc probing on powermac

The last change for partport_pc did fix the common case for all PowerMacs,
but it broke the case for PCI multiport IO cards.  In fact, the config
option CONFIG_PARPORT_PC_SUPERIO=y lead to a hard crash when cups probed
the parport driver.  It enables the winbond and smsc probing.

Remove the PARPORT_BASE check again, parport_pc_find_nonpci_ports() will
take care of it.  All powerpc configs should have
CONFIG_PARPORT_PC_SUPERIO=n, the code did not find anything on the chrp
boards we tested it on.

Tested on a G4/466 with a PCI card:

0001:10:13.0 Serial controller: Timedia Technology Co Ltd PCI2S550 (Dual 16550 UART) (rev 01) (prog-if 02 [16550])
        Subsystem: Timedia Technology Co Ltd Unknown device 5079
        Control: I/O+ Mem- BusMaster- SpecCycle- MemWINV- VGASnoop- ParErr- Stepping+ SERR- FastB2B-
        Status: Cap- 66MHz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
        Interrupt: pin A routed to IRQ 53
        Region 0: I/O ports at f2000800 [size=32]
        Region 2: I/O ports at f2000870 [size=8]
        Region 3: I/O ports at f2000860 [size=8]

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Adam Belay <ambx1@neo.rr.com>
Cc: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ISDN warning fixes

Clean up warnings in drivers/isdn by using long not int for the values
where we pass void * and cast to integer types.  The code is ok (ok passing
the stuff this way isn't pretty but the code is valid).  In all the cases I
checked out the right thing happens anyway but this removes all the
warnings.

Signed-off-by: Alan Cox <alan@redhat.com>
Cc: Jeff Garzik <jeff@garzik.org>
Acked-by: Karsten Keil <kkeil@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] schedule ftape removal

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

[PATCH] drivers/char/ip2: kill unused code, label

Kill warning:

drivers/char/ip2/ip2main.c: In function â\80\98ip2_loadmainâ\80\99:
drivers/char/ip2/ip2main.c:782: warning: label â\80\98out_classâ\80\99 defined but not used

This driver's initialization (and cleanup of errors during init) is
extremely convoluted, and could stand to be transformed into the standard
unwinding-goto style of error cleanup.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Acked-by: Michael H. Warfield <mhw@wittsend.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ipmi: fix uninitialized data bug

gcc issues the following warning:

drivers/char/ipmi/ipmi_si_intf.c: In function â\80\98init_ipmi_siâ\80\99:
drivers/char/ipmi/ipmi_si_intf.c:1729: warning: â\80\98data.irqâ\80\99 may be used uninitialized in this function

This is indeed a bug.  data.irq is completely uninitialized in some code
paths.  Worse than that, data from a previous decode_dmi() run can easily
leak through successive calls.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Acked-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fs/eventpoll: error handling micro-cleanup

While reviewing the 'may be used uninitialized' bogus gcc warnings, I
noticed that an error code assignment was only needed if an error had
actually occured.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Cc: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] kernel-doc for kernel/resource.c

Add kernel-doc function headers in kernel/resource.c and use them in DocBook.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] kernel-doc for kernel/dma.c

Add kernel-doc function headers in kernel/dma.c and use it in DocBook.

Clean up kernel-doc in mca_dma.h (the colon (':') represents a
section header).

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] low performance of lib/sort.c

It is a non-standard heap-sort algorithm implementation because the index
of child node is wrong .  The sort function still outputs right result, but
the performance is O( n * ( log(n) + 1 ) ) , about 10% ~ 20% worse than
standard algorithm.

Signed-off-by: keios <keios.cn@gmail.com>
Acked-by: Matt Mackall <mpm@selenic.com>
Acked-by: Zou Nan hai <nanhai.zou@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Create kallsyms_lookup_size_offset()

Some uses of kallsyms_lookup() do not need to find out the name of a symbol
and its module's name it belongs.  This is specially true in arch specific
code, which needs to unwind the stack to show the back trace during oops
(mips is an example).  In this specific case, we just need to retreive the
function's size and the offset of the active intruction inside it.

Adds a new entry "kallsyms_lookup_size_offset()" This new entry does
exactly the same as kallsyms_lookup() but does not require any buffers to
store any names.

It returns 0 if it fails otherwise 1.

Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] FRV: Permit large kmalloc allocations [try #2]

Permit kmalloc() to make allocations of up to 32MB if so configured.  This
may be useful under NOMMU conditions where vmalloc() can't do this.

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

[PATCH] VFS: Make filldir_t and struct kstat deal in 64-bit inode numbers

These patches make the kernel pass 64-bit inode numbers internally when
communicating to userspace, even on a 32-bit system.  They are required
because some filesystems have intrinsic 64-bit inode numbers: NFS3+ and XFS
for example.  The 64-bit inode numbers are then propagated to userspace
automatically where the arch supports it.

Problems have been seen with userspace (eg: ld.so) using the 64-bit inode
number returned by stat64() or getdents64() to differentiate files, and
failing because the 64-bit inode number space was compressed to 32-bits, and
so overlaps occur.

This patch:

Make filldir_t take a 64-bit inode number and struct kstat carry a 64-bit
inode number so that 64-bit inode numbers can be passed back to userspace.

The stat functions then returns the full 64-bit inode number where
available and where possible.  If it is not possible to represent the inode
number supplied by the filesystem in the field provided by userspace, then
error EOVERFLOW will be issued.

Similarly, the getdents/readdir functions now pass the full 64-bit inode
number to userspace where possible, returning EOVERFLOW instead when a
directory entry is encountered that can't be properly represented.

Note that this means that some inodes will not be stat'able on a 32-bit
system with old libraries where they were before - but it does mean that
there will be no ambiguity over what a 32-bit inode number refers to.

Note similarly that directory scans may be cut short with an error on a
32-bit system with old libraries where the scan would work before for the
same reasons.

It is judged unlikely that this situation will occur because modern glibc
uses 64-bit capable versions of stat and getdents class functions
exclusively, and that older systems are unlikely to encounter
unrepresentable inode numbers anyway.

[akpm: alpha build fix]
Signed-off-by: David Howells <dhowells@redhat.com>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] pid.h cleanup

Make the pid.h macros look less revolting in an 80-col window.

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

[PATCH] Add unifdef to gitignore

This seems to have been missed when unifdef went in
via Sam's tree..

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] hp100: fix conditional compilation mess

The previous hp100 changeset attempted to kill warnings, but was only
tested on !CONFIG_ISA platforms.  The correct conditional compilation
setup involves tested CONFIG_ISA rather than just MODULE.

Fixes link on CONFIG_ISA platforms (i386) in current -git.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] revert "insert IOAPIC(s) and Local APIC into resource map"

Commit 54dbc0c9ebefb38840c6b07fa6eabaeb96c921f5 is causing various
people's machines to fail to map PCI resources.

Revert it in preparation for addressing the show-APICs-in-/proc/iomem
requirement in a different manner.

Cc: Aaron Durbin <adurbin@google.com>
Cc: Andi Kleen <ak@muc.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

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/ehca: Tweak trace message format
  IB/ehca: Fix device registration
  IB/ipath: Fix RDMA reads
  RDMA/cma: Optimize error handling
  RDMA/cma: Eliminate unnecessary remove_list
  RDMA/cma: Set status correctly on route resolution error
  RDMA/cma: Fix device removal race
  RDMA/cma: Fix leak of cm_ids in case of failures

IB/ehca: Tweak trace message format

Add an extra space to make things more readable.

Signed-off-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

IB/ehca: Fix device registration

Move the call to ib_register_device() later, since a device should not
be registered until it is completely read to be used.  This fixes
crashes that occur if an upper-layer driver such as IPoIB is loaded
before the ehca module.

Signed-off-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

IB/ipath: Fix RDMA reads

The PSN used to generate the request following a RDMA read was
incorrect and some state booking wasn't maintained correctly.  This
patch fixes that.

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

RDMA/cma: Optimize error handling

Reorganize code relating to cma_get_net_info() and rdam_create_id() to
optimize error case handling (no need to alloc memory/etc. as part of
rdma_create_id() if input parameters are wrong).

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

RDMA/cma: Eliminate unnecessary remove_list

Eliminate remove_list by using list_del_init() instead during device
removal handling.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

RDMA/cma: Set status correctly on route resolution error

On reporting a route error, also include the status for the error,
rather than indicating a status of 0 when an error has occurred.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

RDMA/cma: Fix device removal race

The race is as follows:

A process : cma_process_remove() calls cma_remove_id_dev(),
    which sets id state to CMA_DEVICE_REMOVAL and
    calls wait_event(dev_remove).

B process : cma_req_handler() had incremented dev_remove,
    and calls cma_acquire_ib_dev() and on failure
    calls cma_release_remove(), which does a
    wake_up of cma_process_remove(). Then
    cma_req_handler() calls rdma_destroy_id();

A Process : cma_remove_id_dev() gets woken and checks the
    state of id, and since it is still (wrongly)
    CMA_DEVICE_REMOVAL, it calls notify_user(id)
    and if that fails, the caller - cma_process_remove()
    calls rdma_destroy_id(id). Two processes can
    call rdma_destroy_id(), resulting in one
    de-referencing kfreed id_priv.

Fix is for process B to set CMA_DESTROYING in cma_req_handler()
so that process A will return instead of doing a rdma_destroy_id().

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

RDMA/cma: Fix leak of cm_ids in case of failures

cma_connect_ib() and cma_connect_iw() leak cm_id's in failure cases.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

Merge /linux/kernel/git/wim/linux-2.6-watchdog

* master.kernel.org:/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog:
  [WATCHDOG] improve machzwd detection
  [WATCHDOG] use ENOTTY instead of ENOIOCTLCMD in ioctl()
  [WATCHDOG] s3c24XX nowayout
  [WATCHDOG] pnx4008: add cpu_relax()
  [WATCHDOG] pnx4008_wdt.c - spinlock fixes.
  [WATCHDOG] pnx4008_wdt.c - remove patch
  [WATCHDOG] pnx4008_wdt.c - nowayout patch
  [WATCHDOG] pnx4008: add watchdog support
  [WATCHDOG] i8xx_tco remove pci_find_device.
  [WATCHDOG] alim remove pci_find_device

[WATCHDOG] improve machzwd detection

On a machine with no machzwd, loading the module prints out..

machzwd: MachZ ZF-Logic Watchdog driver initializing.
0xffff
machzwd: Watchdog using action = RESET

- the 0xffff printk is unnecessary
- 0xffff seems to be 'hardware not present'
- fix CodingStyle. (This driver could use some more work here)

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[WATCHDOG] use ENOTTY instead of ENOIOCTLCMD in ioctl()

Return ENOTTY instead of ENOIOCTLCMD in user-visible ioctl() results

The watchdog drivers used to return ENOIOCTLCMD for bad ioctl() commands.
ENOIOCTLCMD should not be visible by the user, so use ENOTTY instead.

Signed-off-by: Samuel Tardieu <sam@rfc1149.net>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[WATCHDOG] s3c24XX nowayout

If the driver is not configured for `no way out`,
then the open method should not automatically allow
the setting of allow_close to CLOSE_STATE_ALLOW.

The setting of allow_close nullifies the use of
the magic close via the write path. It means that
in the default state, the watchdog will shut-down
even if the magic close has not been issued.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>

[WATCHDOG] pnx4008: add cpu_relax()

Added cpu_relax as suggested by Alan Cox.

Signed-off-by: Vitaly Wool <vitalywool@gmail.com>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>

[WATCHDOG] pnx4008_wdt.c - spinlock fixes.

Add io spinlocks to prevent possible race
conditions between start and stop operations
that are issued from different child processes
where the master process opened /dev/watchdog.

Signed-off-by: Wim Van Sebroeck <wim@iguana.be>

Add prototype for sigset_from_compat()

Duh.  I screwed up editing David Howells patch in commit
3f2e05e90e0846c42626e3d272454f26be34a1bc, and the actual declaration for
the sigset_from_compat() function went missing. My bad.

Olaf Hering saved the day and noticed that I'm a moron.

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

[WATCHDOG] pnx4008_wdt.c - remove patch

Change remove code so that we first detach
the driver from userspace, then clean up the
clock and then clean up the memory we allocated.

Signed-off-by: Wim Van Sebroeck <wim@iguana.be>

[WATCHDOG] pnx4008_wdt.c - nowayout patch

Change nowayout to: WATCHDOG_NOWAYOUT as defined
in include/linux/watchdog.h .

Signed-off-by: Wim Van Sebroeck <wim@iguana.be>

[WATCHDOG] pnx4008: add watchdog support

Add watchdog support for Philips PNX4008 ARM board inlined.

Signed-off-by: Vitaly Wool <vitalywool@gmail.com>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>

[WATCHDOG] i8xx_tco remove pci_find_device.

Use refcounting for pci device obtaining.
Use PCI_DEVICE macro.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
Cc: Andrew Morton <akpm@osdl.org>

[WATCHDOG] alim remove pci_find_device

Convert pci_find_device to pci_get_device + pci_dev_put
in alim watchdog cards' drivers (refcounting).

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
Signed-off-by: Andrew Morton <akpm@osdl.org>

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

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6: (37 commits)
  [netdrvr] hp100: encapsulate all non-module code
  drivers/net/wireless/{airo,ipw2100}: fix error handling bugs
  [netdrvr] phy: Fix bugs in error handling
  [PATCH] spidernet: Use pci_dma_mapping_error()
  [PATCH] sky2: version 1.9
  [PATCH] sky2: fragmented receive for large MTU
  [PATCH] sky2: use netif_tx_lock instead of LLTX
  [PATCH] sky2: incremental transmit completion
  [PATCH] sky2: name irq after eth for irqbalance
  [PATCH] sky2: workarounds for some 88e806x chips
  [PATCH] sky2: use standard pci register capabilties for error register
  [PATCH] sky2: gigabit full duplex negotiation
  e100, e1000, ixgb: increment version numbers
  ixgb: convert to netdev_priv(netdev)
  ixgb: combine more rx descriptors to improve performance
  e1000: possible memory leak in e1000_set_ringparam
  e1000: Janitor: Use #defined values for literals
  e1000: don't strip vlan ID if 8021q claims it
  e1000: rework polarity, NVM, eeprom code and fixes.
  e1000: driver state fixes (race fix)
  ...

Merge branch 'for-linus' of git://git./linux/kernel/git/shaggy/jfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/shaggy/jfs-2.6:
  JFS: White space cleanup
  [PATCH] JFS: return correct error when i-node allocation failed
  JFS: Remove shadow variable from fs/jfs/jfs_txnmgr.c:xtLog()

Merge git://git.infradead.org/mtd-2.6

* git://git.infradead.org/mtd-2.6:
  [MTD] Cleanup of 'ioremap balanced with iounmap for drivers/mtd subsystem'
  [MTD] fix nftl_write warning
  [MTD] fix printk warning
  [MTD ONENAND] Check OneNAND lock scheme & all block unlock command support
  [MTD ONENAND] Remove unused MTD_ONENAND_SYNC_READ configuration
  [MTD ONENAND] Fix OneNAND probe
  [MTD NAND] Provide prototype for newly-exported nand_wait_ready()
  [MTD] Remove #ifndef __KERNEL__ hack in <mtd/mtd-abi.h>
  [MTD NAND] Allow override of page read and write functions.
  [MTD NAND] Allocate chip->buffers separately to allow it to be overridden
  [MTD NAND] Split nand_scan() into two parts; allow board driver to intervene
  [MTD NAND] Export nand_wait_ready() for use by board drivers

Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input: (35 commits)
  Input: wistron - add support for Acer TravelMate 2424NWXCi
  Input: wistron - fix setting up special buttons
  Input: add KEY_BLUETOOTH and KEY_WLAN definitions
  Input: add new BUS_VIRTUAL bus type
  Input: add driver for stowaway serial keyboards
  Input: make input_register_handler() return error codes
  Input: remove cruft that was needed for transition to sysfs
  Input: fix input module refcounting
  Input: constify input core
  Input: libps2 - rearrange exports
  Input: atkbd - support Microsoft Natural Elite Pro keyboards
  Input: i8042 - disable MUX mode on Toshiba Equium A110
  Input: i8042 - get rid of polling timer
  Input: send key up events at disconnect
  Input: constify psmouse driver
  Input: i8042 - add Amoi to the MUX blacklist
  Input: logips2pp - add sugnature 56 (Cordless MouseMan Wheel), cleanup
  Input: add driver for Touchwin serial touchscreens
  Input: add driver for Touchright serial touchscreens
  Input: add driver for Penmount serial touchscreens
  ...

Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] Remove unused galileo-boars header files
  [MIPS] Rename SERIAL_PORT_DEFNS for EV64120
  [MIPS] Add UART IRQ number for EV64120
  [MIPS] Remove excite_flash.c
  [MIPS] Update i8259 resources.
  [MIPS] Make unwind_stack() can dig into interrupted context
  [MIPS] Stacktrace build-fix and improvement
  [MIPS] QEMU: Add support for little endian mips
  [MIPS] Remove __flush_icache_page
  [MIPS] lockdep: update defconfigs
  [MIPS] lockdep: Add STACKTRACE_SUPPORT and enable LOCKDEP_SUPPORT
  [MIPS] lockdep: fix TRACE_IRQFLAGS_SUPPORT

[PATCH] BLOCK: Revert patch to hack around undeclared sigset_t in linux/compat.h

Revert Andrew Morton's patch to temporarily hack around the lack of a
declaration of sigset_t in linux/compat.h to make the block-disablement
patches build on IA64.  This got accidentally pushed to Linus and should
be fixed in a different manner.

Also make linux/compat.h #include asm/signal.h to gain a definition of
sigset_t so that it can externally declare sigset_from_compat().

This has been compile-tested for i386, x86_64, ia64, mips, mips64, frv, ppc and
ppc64 and run-tested on frv.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] replace cad_pid by a struct pid

There are a few places in the kernel where the init task is signaled.  The
ctrl+alt+del sequence is one them.  It kills a task, usually init, using a
cached pid (cad_pid).

This patch replaces the pid_t by a struct pid to avoid pid wrap around
problem.  The struct pid is initialized at boot time in init() and can be
modified through systctl with

/proc/sys/kernel/cad_pid

[ I haven't found any distro using it ? ]

It also introduces a small helper routine kill_cad_pid() which is used
where it seemed ok to use cad_pid instead of pid 1.

[akpm@osdl.org: cleanups, build fix]
Signed-off-by: Cedric Le Goater <clg@fr.ibm.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] introduce get_task_pid() to fix unsafe get_pid()

proc_pid_make_inode:

ei->pid = get_pid(task_pid(task));

I think this is not safe.  get_pid() can be preempted after checking "pid
!= NULL".  Then the task exits, does detach_pid(), and RCU frees the pid.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: comment what proc_fill_cache does

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: remove the useless SMP-safe comments from /proc

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: remove trailing blank entry from pid_entry arrays

It was pointed out that since I am taking ARRAY_SIZE anyway the trailing empty
entry is silly and just wastes space.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: properly compute TGID_OFFSET

The value doesn't change but this ensures I will have the proper value when
other files are added to proc_base_stuff.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: drop tasklist lock in task_state()

task_state() needs tasklist_lock to protect ->parent/->real_parent.  However
task->parent points to nowhere only when the actions below happen in order

1) release_task(task)
2) release_task(task->parent)
3) a grace period passed

But 3) implies that the memory ops from 1) should be finished, so pid_alive()
can't be true in such a case.

Otherwise, we don't care if ->parent/->real_parent changes under us.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: convert do_task_stat() to use lock_task_sighand()

Drop tasklist_lock. ->siglock protects almost all interesting data
(including sub-threads traversal) except:

->signal->tty
protected by tty_mutex

->real_parent
the task can't be unhashed while we are holding
->siglock, so ->real_parent can change from under us
but we can safely dereference it under rcu_read_lock()

->pgrp/->session
we can get inconsistent numbers if the task does
sys_setsid/daemonize at the same time. I hope this
is acceptable.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: convert task_sig() to use lock_task_sighand()

lock_task_sighand() can take ->siglock without holding tasklist_lock.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: Use pid_task instead of open coding it

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: Merge proc_tid_attr and proc_tgid_attr

The implementation is exactly the same and there is currently nothing to
distinguish proc_tid_attr, and proc_tgid_attr.  So it is pointless to have two
separate implementations.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: Remove the hard coded inode numbers

The hard coded inode numbers in proc currently limit its maintainability,
its flexibility, and what can be done with the rest of system.  /proc limits
pid-max to 32768 on 32 bit systems it limits fd-max to 32768 on all systems,
and placing the pid in the inode number really gets in the way of implementing
subdirectories of per process information.

Ever since people started adding to the middle of the file type enumeration we
haven't been maintaing the historical inode numbers, all we have really
succeeded in doing is keeping the pid in the proc inode number.  The pid is
already available in the directory name so no information is lost removing it
from the inode number.

So if something in user space cares if we remove the inode number from the
/proc inode it is almost certainly broken.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: Factor out an instantiate method from every lookup method

To remove the hard coded proc inode numbers it is necessary to be able to
create the proc inodes during readdir.  The instantiate methods are the subset
of lookup that is needed to accomplish that.

This first step just splits the lookup methods into 2 functions.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: Make the generation of the self symlink table driven

This patch generalizes the concept of files in /proc that are related to
processes but live in the root directory of /proc

Ideally this would reuse infrastructure from the rest of the process specific
parts of proc but unfortunately security_task_to_inode must not be called on
files that are not strictly per process.  security_task_to_inode really needs
to be reexamined as the security label can change in important places that we
are not currently catching, but I'm not certain that simplifies this problem.

By at least matching the structure of the rest of proc we get more idiom reuse
and it becomes easier to spot problems in the way things are put together.

Later things like /proc/mounts are likely to be moved into proc_base as well.
If union mounts are ever supported we may be able to make /proc a union mount,
and properly split it into 2 filesystems.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] AVR32: Implement kernel_execve

Move execve() into arch/avr32/kernel/sys_avr32.c, rename it to
kernel_execve() and return the syscall return value directly without
setting errno.

This also gets rid of the __KERNEL_SYSCALLS__ stuff from unistd.h and
expands #ifdef __KERNEL__ to cover everything in unistd.h except the
__NR_foo definitions.

Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] remove remaining errno and __KERNEL_SYSCALLS__ references

The last in-kernel user of errno is gone, so we should remove the definition
and everything referring to it.  This also removes the now-unused lib/execve.c
file that was introduced earlier.

Also remove every trace of __KERNEL_SYSCALLS__ that still remained in the
kernel.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Andi Kleen <ak@muc.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Ian Molton <spyro@f2s.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Hirokazu Takata <takata.hirokazu@renesas.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp>
Cc: Richard Curnow <rc@rc0.org.uk>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp>
Cc: Chris Zankel <chris@zankel.net>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] sh64: remove the use of kernel syscalls

sh64 is using system call macros to call some functions from the kernel.

The old debug code can simply be removed, since we don't really have that much
of a need for it anymore, it was mostly something that was handy during the
initial bringup.  This also brings us closer to something that looks like
readable code again..

I also added a sane kernel_thread() implementation that gets away from this,
so that should take care of sh64 at least.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Andi Kleen <ak@muc.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Ian Molton <spyro@f2s.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Hirokazu Takata <takata.hirokazu@renesas.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp>
Cc: Richard Curnow <rc@rc0.org.uk>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp>
Cc: Chris Zankel <chris@zankel.net>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Remove the use of _syscallX macros in UML

User mode linux uses _syscallX() to call into the host kernel.  The
recommended way to do this is to use the syscall() function from libc.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Andi Kleen <ak@muc.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Ian Molton <spyro@f2s.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Hirokazu Takata <takata.hirokazu@renesas.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp>
Cc: Richard Curnow <rc@rc0.org.uk>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp>
Cc: Chris Zankel <chris@zankel.net>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] provide kernel_execve on all architectures

This adds the new kernel_execve function on all architectures that were using
_syscall3() to implement execve.

The implementation uses code from the _syscall3 macros provided in the
unistd.h header file.  I don't have cross-compilers for any of these
architectures, so the patch is untested with the exception of i386.

Most architectures can probably implement this in a nicer way in assembly or
by combining it with the sys_execve implementation itself, but this should do
it for now.

[bunk@stusta.de: m68knommu build fix]
[markh@osdl.org: build fix]
[bero@arklinux.org: build fix]
[ralf@linux-mips.org: mips fix]
[schwidefsky@de.ibm.com: s390 fix]
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Andi Kleen <ak@muc.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Ian Molton <spyro@f2s.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Hirokazu Takata <takata.hirokazu@renesas.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp>
Cc: Richard Curnow <rc@rc0.org.uk>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp>
Cc: Chris Zankel <chris@zankel.net>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Bernhard Rosenkraenzer <bero@arklinux.org>
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] rename the provided execve functions to kernel_execve

Some architectures provide an execve function that does not set errno, but
instead returns the result code directly.  Rename these to kernel_execve to
get the right semantics there.  Moreover, there is no reasone for any of these
architectures to still provide __KERNEL_SYSCALLS__ or _syscallN macros, so
remove these right away.

[akpm@osdl.org: build fix]
[bunk@stusta.de: build fix]
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Andi Kleen <ak@muc.de>
Acked-by: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Ian Molton <spyro@f2s.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Hirokazu Takata <takata.hirokazu@renesas.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp>
Cc: Richard Curnow <rc@rc0.org.uk>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp>
Cc: Chris Zankel <chris@zankel.net>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] introduce kernel_execve

The use of execve() in the kernel is dubious, since it relies on the
__KERNEL_SYSCALLS__ mechanism that stores the result in a global errno
variable.  As a first step of getting rid of this, change all users to a
global kernel_execve function that returns a proper error code.

This function is a terrible hack, and a later patch removes it again after the
kernel syscalls are gone.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Andi Kleen <ak@muc.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Ian Molton <spyro@f2s.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Hirokazu Takata <takata.hirokazu@renesas.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp>
Cc: Richard Curnow <rc@rc0.org.uk>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp>
Cc: Chris Zankel <chris@zankel.net>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ipc: replace kmalloc and memset in get_undo_list with kzalloc

Simplify get_undo_list() by dropping the unnecessary cast, removing the
size variable, and switching to kzalloc() instead of a kmalloc() followed
by a memset().

This cleanup was split then modified from Jes Sorenson's Task Notifiers
patches.

Signed-off-by: Matt Helsley <matthltc@us.ibm.com>
Cc: Jes Sorensen <jes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nsproxy cloning error path fix

This patch fixes copy_namespaces()'s error path.

when new nsproxy (new_ns) is created pointers to namespaces (ipc, uts) are
copied from the old nsproxy.  Later in copy_utsname, copy_ipcs, etc.
according namespaces are get-ed.  On error path needed namespaces are
put-ed, so there's no need to put new nsproxy itelf as it woud cause
putting namespaces for the second time.

Found when incorporating namespaces into OpenVZ kernel.

Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IPC namespace - sysctls

Sysctl tweaks for IPC namespace

Signed-off-by: Pavel Emelianiov <xemul@openvz.org>
Signed-off-by: Kirill Korotaev <dev@openvz.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IPC namespace - shm

IPC namespace support for IPC shm code.

Signed-off-by: Pavel Emelianiov <xemul@openvz.org>
Signed-off-by: Kirill Korotaev <dev@openvz.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IPC namespace - sem

IPC namespace support for IPC sem code.

Signed-off-by: Pavel Emelianiov <xemul@openvz.org>
Signed-off-by: Kirill Korotaev <dev@openvz.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IPC namespace - msg

IPC namespace support for IPC msg code.

Signed-off-by: Pavel Emelianiov <xemul@openvz.org>
Signed-off-by: Kirill Korotaev <dev@openvz.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IPC namespace - utils

This patch adds basic IPC namespace functionality to
IPC utils:
- init_ipc_ns
- copy/clone/unshare/free IPC ns
- /proc preparations

Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Signed-off-by: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IPC namespace core

This patch set allows to unshare IPCs and have a private set of IPC objects
(sem, shm, msg) inside namespace.  Basically, it is another building block of
containers functionality.

This patch implements core IPC namespace changes:
- ipc_namespace structure
- new config option CONFIG_IPC_NS
- adds CLONE_NEWIPC flag
- unshare support

[clg@fr.ibm.com: small fix for unshare of ipc namespace]
[akpm@osdl.org: build fix]
Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Signed-off-by: Kirill Korotaev <dev@openvz.org>
Signed-off-by: Cedric Le Goater <clg@fr.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uts: copy nsproxy only when needed

The nsproxy was being copied in unshare() when anything was being unshared,
even if it was something not referenced from nsproxy.  This should end up
in some cases with far more memory usage than necessary.

Signed-off-by: Serge Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: utsname: implement CLONE_NEWUTS flag

Implement a CLONE_NEWUTS flag, and use it at clone and sys_unshare.

[clg@fr.ibm.com: IPC unshare fix]
[bunk@stusta.de: cleanup]
Signed-off-by: Serge Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Cedric Le Goater <clg@fr.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: utsname: remove system_utsname

The system_utsname isn't needed now that kernel/sysctl.c is fixed.
Nuke it.

Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: utsname: sysctl

Sysctl uts patch.  This will need to be done another way, but since sysctl
itself needs to be container aware, 'the right thing' is a separate patchset.

[akpm@osdl.org: ia64 build fix]
[sam.vilain@catalyst.net.nz: cleanup]
[sam.vilain@catalyst.net.nz: add proc_do_utsns_string]
Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: utsname: implement utsname namespaces

This patch defines the uts namespace and some manipulators.
Adds the uts namespace to task_struct, and initializes a
system-wide init namespace.

It leaves a #define for system_utsname so sysctl will compile.
This define will be removed in a separate patch.

[akpm@osdl.org: build fix, cleanup]
Signed-off-by: Serge Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: utsname: use init_utsname when appropriate

In some places, particularly drivers and __init code, the init utsns is the
appropriate one to use.  This patch replaces those with a the init_utsname
helper.

Changes: Removed several uses of init_utsname().  Hope I picked all the
right ones in net/ipv4/ipconfig.c.  These are now changed to
utsname() (the per-process namespace utsname) in the previous
patch (2/7)

[akpm@osdl.org: CIFS fix]
Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Cc: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: utsname: switch to using uts namespaces

Replace references to system_utsname to the per-process uts namespace
where appropriate.  This includes things like uname.

Changes: Per Eric Biederman's comments, use the per-process uts namespace
for ELF_PLATFORM, sunrpc, and parts of net/ipv4/ipconfig.c

[jdike@addtoit.com: UML fix]
[clg@fr.ibm.com: cleanup]
[akpm@osdl.org: build fix]
Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Cedric Le Goater <clg@fr.ibm.com>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: utsname: introduce temporary helpers

Define utsname() and init_utsname() which return &system_utsname.  Users of
system_utsname will be changed to use these helpers, after which
system_utsname will disappear.

Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: exit_task_namespaces() invalidates nsproxy

exit_task_namespaces() has replaced the former exit_namespace().  It
invalidates task->nsproxy and associated namespaces.  This is an issue for
the (futur) pid namespace which is required to be valid in exit_notify().

This patch moves exit_task_namespaces() after exit_notify() to keep nsproxy
valid.

Signed-off-by: Cedric Le Goater <clg@fr.ibm.com>
Cc: Serge E. Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: incorporate fs namespace into nsproxy

This moves the mount namespace into the nsproxy.  The mount namespace count
now refers to the number of nsproxies point to it, rather than the number of
tasks.  As a result, the unshare_namespace() function in kernel/fork.c no
longer checks whether it is being shared.

Signed-off-by: Serge Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nsproxy: move init_nsproxy into kernel/nsproxy.c

Move the init_nsproxy definition out of arch/ into kernel/nsproxy.c.  This
avoids all arches having to be updated.  Compiles and boots on s390.

Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] namespaces: add nsproxy

This patch adds a nsproxy structure to the task struct.  Later patches will
move the fs namespace pointer into this structure, and introduce a new utsname
namespace into the nsproxy.

The vserver and openvz functionality, then, would be implemented in large part
by virtualizing/isolating more and more resources into namespaces, each
contained in the nsproxy.

[akpm@osdl.org: build fix]
Signed-off-by: Serge Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] make kernel/sysctl.c:_proc_do_string() static

This patch makes the needlessly global _proc_do_string() static.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] proc: sysctl: add _proc_do_string helper

The logic in proc_do_string is worth re-using without passing in a
ctl_table structure (say, we want to calculate a pointer and pass that in
instead); pass in the two fields it uses from that structure as explicit
arguments.

Signed-off-by: Sam Vilain <sam.vilain@catalyst.net.nz>
Cc: Serge E. Hallyn <serue@us.ibm.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nfsd: lockdep annotation

while doing a kernel make modules_install install over an NFS mount.

  =============================================
  [ INFO: possible recursive locking detected ]
  ---------------------------------------------
  nfsd/9550 is trying to acquire lock:
   (&inode->i_mutex){--..}, at: [<c034c845>] mutex_lock+0x1c/0x1f

  but task is already holding lock:
   (&inode->i_mutex){--..}, at: [<c034c845>] mutex_lock+0x1c/0x1f

  other info that might help us debug this:
  2 locks held by nfsd/9550:
   #0:  (hash_sem){..--}, at: [<cc895223>] exp_readlock+0xd/0xf [nfsd]
   #1:  (&inode->i_mutex){--..}, at: [<c034c845>] mutex_lock+0x1c/0x1f

  stack backtrace:
   [<c0103508>] show_trace_log_lvl+0x58/0x152
   [<c0103b8b>] show_trace+0xd/0x10
   [<c0103c2f>] dump_stack+0x19/0x1b
   [<c012aa57>] __lock_acquire+0x77a/0x9a3
   [<c012af4a>] lock_acquire+0x60/0x80
   [<c034c6c2>] __mutex_lock_slowpath+0xa7/0x20e
   [<c034c845>] mutex_lock+0x1c/0x1f
   [<c0162edc>] vfs_unlink+0x34/0x8a
   [<cc891d98>] nfsd_unlink+0x18f/0x1e2 [nfsd]
   [<cc89884f>] nfsd3_proc_remove+0x95/0xa2 [nfsd]
   [<cc88f0d4>] nfsd_dispatch+0xc0/0x178 [nfsd]
   [<c033e84d>] svc_process+0x3a5/0x5ed
   [<cc88f5ba>] nfsd+0x1a7/0x305 [nfsd]
   [<c0101005>] kernel_thread_helper+0x5/0xb
  DWARF2 unwinder stuck at kernel_thread_helper+0x5/0xb
  Leftover inexact backtrace:
   [<c0103b8b>] show_trace+0xd/0x10
   [<c0103c2f>] dump_stack+0x19/0x1b
   [<c012aa57>] __lock_acquire+0x77a/0x9a3
   [<c012af4a>] lock_acquire+0x60/0x80
   [<c034c6c2>] __mutex_lock_slowpath+0xa7/0x20e
   [<c034c845>] mutex_lock+0x1c/0x1f
   [<c0162edc>] vfs_unlink+0x34/0x8a
   [<cc891d98>] nfsd_unlink+0x18f/0x1e2 [nfsd]
   [<cc89884f>] nfsd3_proc_remove+0x95/0xa2 [nfsd]
   [<cc88f0d4>] nfsd_dispatch+0xc0/0x178 [nfsd]
   [<c033e84d>] svc_process+0x3a5/0x5ed
   [<cc88f5ba>] nfsd+0x1a7/0x305 [nfsd]
   [<c0101005>] kernel_thread_helper+0x5/0xb

  =============================================
  [ INFO: possible recursive locking detected ]
  ---------------------------------------------
  nfsd/9580 is trying to acquire lock:
   (&inode->i_mutex){--..}, at: [<c034cc1d>] mutex_lock+0x1c/0x1f

  but task is already holding lock:
   (&inode->i_mutex){--..}, at: [<c034cc1d>] mutex_lock+0x1c/0x1f

  other info that might help us debug this:
  2 locks held by nfsd/9580:
   #0:  (hash_sem){..--}, at: [<cc89522b>] exp_readlock+0xd/0xf [nfsd]
   #1:  (&inode->i_mutex){--..}, at: [<c034cc1d>] mutex_lock+0x1c/0x1f

  stack backtrace:
   [<c0103508>] show_trace_log_lvl+0x58/0x152
   [<c0103b8b>] show_trace+0xd/0x10
   [<c0103c2f>] dump_stack+0x19/0x1b
   [<c012aa63>] __lock_acquire+0x77a/0x9a3
   [<c012af56>] lock_acquire+0x60/0x80
   [<c034ca9a>] __mutex_lock_slowpath+0xa7/0x20e
   [<c034cc1d>] mutex_lock+0x1c/0x1f
   [<cc892ad1>] nfsd_setattr+0x2c8/0x499 [nfsd]
   [<cc893ede>] nfsd_create_v3+0x31b/0x4ac [nfsd]
   [<cc8984a1>] nfsd3_proc_create+0x128/0x138 [nfsd]
   [<cc88f0d4>] nfsd_dispatch+0xc0/0x178 [nfsd]
   [<c033ec1d>] svc_process+0x3a5/0x5ed
   [<cc88f5ba>] nfsd+0x1a7/0x305 [nfsd]
   [<c0101005>] kernel_thread_helper+0x5/0xb
  DWARF2 unwinder stuck at kernel_thread_helper+0x5/0xb
  Leftover inexact backtrace:
   [<c0103b8b>] show_trace+0xd/0x10
   [<c0103c2f>] dump_stack+0x19/0x1b
   [<c012aa63>] __lock_acquire+0x77a/0x9a3
   [<c012af56>] lock_acquire+0x60/0x80
   [<c034ca9a>] __mutex_lock_slowpath+0xa7/0x20e
   [<c034cc1d>] mutex_lock+0x1c/0x1f
   [<cc892ad1>] nfsd_setattr+0x2c8/0x499 [nfsd]
   [<cc893ede>] nfsd_create_v3+0x31b/0x4ac [nfsd]
   [<cc8984a1>] nfsd3_proc_create+0x128/0x138 [nfsd]
   [<cc88f0d4>] nfsd_dispatch+0xc0/0x178 [nfsd]
   [<c033ec1d>] svc_process+0x3a5/0x5ed
   [<cc88f5ba>] nfsd+0x1a7/0x305 [nfsd]
   [<c0101005>] kernel_thread_helper+0x5/0xb

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Neil Brown <neilb@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] knfsd: allow admin to set nthreads per node

Add /proc/fs/nfsd/pool_threads which allows the sysadmin (or a userspace
daemon) to read and change the number of nfsd threads in each pool.  The
format is a list of space-separated integers, one per pool.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: make rpc threads pools numa aware

Actually implement multiple pools.  On NUMA machines, allocate a svc_pool per
NUMA node; on SMP a svc_pool per CPU; otherwise a single global pool.  Enqueue
sockets on the svc_pool corresponding to the CPU on which the socket bh is run
(i.e.  the NIC interrupt CPU).  Threads have their cpu mask set to limit them
to the CPUs in the svc_pool that owns them.

This is the patch that allows an Altix to scale NFS traffic linearly
beyond 4 CPUs and 4 NICs.

Incorporates changes and feedback from Neil Brown, Trond Myklebust, and
Christoph Hellwig.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: use svc_set_num_threads to manage threads in knfsd

Replace the existing list of all nfsd threads with new code using
svc_create_pooled().

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: add svc_set_num_threads

Currently knfsd keeps its own list of all nfsd threads in nfssvc.c; add a new
way of managing the list of all threads in a svc_serv.  Add
svc_create_pooled() to allow creation of a svc_serv whose threads are managed
by the sunrpc code.  Add svc_set_num_threads() to manage the number of threads
in a service, either per-pool or globally across the service.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: add svc_get

add svc_get() for those occasions when we need to temporarily bump up
svc_serv->sv_nrthreads as a pseudo refcount.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: split svc_serv into pools

Split out the list of idle threads and pending sockets from svc_serv into a
new svc_pool structure, and allocate a fixed number (in this patch, 1) of
pools per svc_serv.  The new structure contains a lock which takes over
several of the duties of svc_serv->sv_lock, which is now relegated to
protecting only sv_tempsocks, sv_permsocks, and sv_tmpcnt in svc_serv.

The point is to move the hottest fields out of svc_serv and into svc_pool,
allowing a following patch to arrange for a svc_pool per NUMA node or per CPU.
 This is a major step towards making the NFS server NUMA-friendly.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: test and set SK_BUSY atomically

The SK_BUSY bit in svc_sock->sk_flags ensures that we do not attempt to
enqueue a socket twice.  Currently, setting and clearing the bit is protected
by svc_serv->sv_lock.  As I intend to reduce the data that the lock protects
so it's not held when svc_sock_enqueue() tests and sets SK_BUSY, that test and
set needs to be atomic.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: convert sk_reserved to atomic_t

Convert the svc_sock->sk_reserved variable from an int protected by
svc_serv->sv_lock, to an atomic.  This reduces (by 1) the number of places we
need to take the (effectively global) svc_serv->sv_lock.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: use new lock for svc_sock deferred list

Protect the svc_sock->sk_deferred list with a new lock svc_sock->sk_defer_lock
instead of svc_serv->sv_lock.  Using the more fine-grained lock reduces the
number of places we need to take the svc_serv lock.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: convert sk_inuse to atomic_t

Convert the svc_sock->sk_inuse counter from an int protected by
svc_serv->sv_lock, to an atomic.  This reduces the number of places we need to
take the (effectively global) svc_serv->sv_lock.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>

[PATCH] knfsd: move tempsock aging to a timer

Following are 11 patches from Greg Banks which combine to make knfsd more
Numa-aware.  They reduce hitting on 'global' data structures, and create some
data-structures that can be node-local.

knfsd threads are bound to a particular node, and the thread to handle a new
request is chosen from the threads that are attach to the node that received
the interrupt.

The distribution of threads across nodes can be controlled by a new file in
the 'nfsd' filesystem, though the default approach of an even spread is
probably fine for most sites.

Some (old) numbers that show the efficacy of these patches: N == number of
NICs == number of CPUs == nmber of clients.  Number of NUMA nodes == N/2

N Throughput, MiB/s CPU usage, % (max=N*100)
Before After Before After
--- ------ ---- ----- -----
4 312 435 350 228
6 500 656 501 418
8 562 804 690 589

This patch:

Move the aging of RPC/TCP connection sockets from the main svc_recv() loop to
a timer which uses a mark-and-sweep algorithm every 6 minutes.  This reduces
the amount of work that needs to be done in the main RPC loop and the length
of time we need to hold the (effectively global) svc_serv->sv_lock.

[akpm@osdl.org: cleanup]
Signed-off-by: Greg Banks <gnb@melbourne.sgi.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>