host->ops->host_stop() might access ports.  Free ports after
host_stop.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Commit 721449bf added support for using the
ADMA notifier bits to determine which commands to check for completion.
However there have been reports that this causes command timeouts in certain
cases. This is still being investigated. In addition, apparently the notifiers
won't work if ADMA is disabled on the other port as a result of an ATAPI device
being connected, and we don't handle this case properly.

For now, just restore the previous behavior of checking all active commands
to see if they are complete, without relying on the notifiers.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Recently I got my hands on nVidia's MCP61 PM-AM board, and
it contains IDE chip configured by BIOS with only primary
channel enabled.  This confuses code which probes for
device DMA capabilities - it gets 0x60 (happy duplex
device) from primary channel BMDMA, but 0xFF (nobody here)
from secondary channel BMDMA.  Due to this code then believes
that chip is simplex.  I do not address this problem in
my patch, as I'm not sure how to handle this.  Probably
ata_pci_init_one should have bitmap of enabled/possible
interfaces instead of their count, but it looks like
quite intrusive change, and maybe we do not care - for device
with only one channel simplex and regular DMA engines are
same.

But making device simplex pointed out that support for
DMA on simplex devices is currently broken - ata_dev_xfermask
tests whether device is simplex and if it is whether DMA
engine was assigned to this port.  If not then it strips
out DMA bits from device.  Problem is that code which assigns
DMA engine to port in ata_set_mode first detect device
mode and assigns DMA engine to channel only if some DMA
capable device was found.

And as xfermask stripped out DMA bits, host->simplex_claimed
is always NULL with current implementation.

By allowing DMA either if simplex_claimed is NULL or if it
points to current port DMA can be finally used - it gets
assigned to first port which contains any DMA capable
device.

Before:
pata_amd 0000:00:06.0: version 0.2.8
PCI: Setting latency timer of device 0000:00:06.0 to 64
ata5: PATA max UDMA/133 cmd 0x000101f0 ctl 0x000103f6 bmdma 0x0001f000 irq 14
ata6: PATA max UDMA/133 cmd 0x00010170 ctl 0x00010376 bmdma 0x0001f008 irq 15
scsi4 : pata_amd
ata5.00: ATAPI, max UDMA/66
ata5.00: simplex DMA is claimed by other device, disabling DMA
ata5.00: configured for PIO4
scsi5 : pata_amd
ata6: port disabled. ignoring.
ata6: reset failed, giving up
scsi 4:0:0:0: CD-ROM            ATAPI    DVD W  DH16W1P   LG12 PQ: 0 ANSI: 5

After:
pata_amd 0000:00:06.0: version 0.2.8
PCI: Setting latency timer of device 0000:00:06.0 to 64
ata5: PATA max UDMA/133 cmd 0x000101f0 ctl 0x000103f6 bmdma 0x0001f000 irq 14
ata6: PATA max UDMA/133 cmd 0x00010170 ctl 0x00010376 bmdma 0x0001f008 irq 15
scsi4 : pata_amd
ata5.00: ATAPI, max UDMA/66
ata5.00: configured for UDMA/33
scsi5 : pata_amd
ata6: port disabled. ignoring.
ata6: reset failed, giving up
scsi 4:0:0:0: CD-ROM            ATAPI    DVD W  DH16W1P   LG12 PQ: 0 ANSI: 5
Signed-off-by: NPetr Vandrovec <petr@vandrovec.name>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

A while ago I modified the libata code so that drivers can return -ENOENT
for unknown ports not fiddle with the EH flags and print stuff directly.
Somewhere along the line ata_piix didn't get fully converted.
Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

missing close bracket.
Signed-off-by: NDave Jones <davej@redhat.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

 - point to the sparse webpage
 - use git:// instead of rsync://
Signed-off-by: NDave Jones <davej@redhat.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

x86_64 nvidia_bugs() broke when we bailed out on not finding the HPET.
However, the quirk works by checking for _not_ finding the HPET...

Delete the nvidia_hpet_detected flag and simply test for
not finding the HPET, which is simple to do now that
acpi_table_parse returns 1 on failure.
Signed-off-by: NLen Brown <len.brown@intel.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband:
  IPoIB: Turn on interface's carrier after broadcast group is joined
  RDMA/ucma: Avoid sending reject if backlog is full
  RDMA/cxgb3: Fix MR permission problems
  RDMA/cxgb3: Don't reuse skbs that are non-linear or cloned
  RDMA/cxgb3: Squelch logging AE errors
  RDMA/cxgb3: Stop EP timer when MPA exchange is aborted by peer
  RDMA/cxgb3: Move QP to error on destroy if the state is IDLE
  RDMA/cxgb3: Fixes for "normal close" failures
  RDMA/cxgb3: Fix build on sparc64
  RDMA/cma: Initialize rdma_bind_list in cma_alloc_any_port()
  RDMA/cxgb3: Don't use mm after it's freed in iwch_mmap()
  RDMA/cxgb3: Start ep timer on a MPA reject
  IB/mthca: Fix error path in mthca_alloc_memfree()
  IB/ehca: Fix sync between completion handler and destroy cq
  IPoIB: Only handle async events for one port

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [SCTP] ipv6: inconsistent lock state ipv6_add_addr/sctp_v6_copy_addrlist

Need to provide {claim,release}_dma_lock() for this guy too.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Do netif_carrier_on() right after the IPv4 broadcast multicast group
is joined, rather than waiting for all of the initial set of multicast
group joins to finish.  This allows at least IPv4 traffic to limp
along on broken fabrics where not all multicast groups can be joined.
Signed-off-by: NShirley Ma <xma@us.ibm.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

lockdep found that dev->lock taken from softirq in ipv6_add_addr
is also taken in sctp_v6_copy_addrlist with softirqs enabled, so
lockup is possible.
Noticed-by: NSimon Arlott <simon@arlott.org>
Signed-off-by: NJarek Poplawski <jarkao2@o2.pl>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For 2.6.20 it mostly used to just not work, for 2.6.21-rc it crashes, this
seems to be down to luck (bad or good). The libata-acpi code needs to
avoid doing PCI work on non-PCI devices. This is one hack although it's
not pretty and perhaps there is a "right" way to check if a struct device
* is PCI ?
Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

ecryptfs uses a lock_parent() function, which I hope really locks the parents
and is not abused
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Michael Halcrow <mhalcrow@us.ibm.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The sn console driver was snagged by the use of CONFIG_DEBUG_SHIRQ!

The request_irq() immediate call to the interrupt handler caused
another attempt to lock the port lock - deadlock.

This is a patch to fix that.
Signed-off-by: NPatrick Gefre <pfg@sgi.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

In my previous x86_64 thread fix, I forgot to initialize thread.arch.fs in
arch_prctl.  A process calling arch_prctl to set %fs would lose it on the
next context switch.

It also turns out that you can switch to a process which is in the process
of exiting and which has lost its mm.  In this case, it's worse than
useless to try to call arch_prctl on the host process.
Signed-off-by: NJeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Fix inverted check introduced in 57881dd9 "Fix
check_partition routines".
Signed-off-by: NSuzuki K P <suzuki@in.ibm.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

IA64 and ARM-OABI are currently using their own version of epoll compat_
code.

An architecture needs epoll_event translation if alignof(u64) in 32 bit
mode is different from alignof(u64) in 64 bit mode.  If an architecture
needs epoll_event translation, it must define struct compat_epoll_event in
asm/compat.h and set CONFIG_HAVE_COMPAT_EPOLL_EVENT and use
compat_sys_epoll_ctl and compat_sys_epoll_wait.

All 64 bit architecture should use compat_sys_epoll_pwait.

[sfr: restructure and move to fs/compat.c, remove MIPS version
of compat_sys_epoll_pwait, use __put_user_unaligned]
Signed-off-by: NStephen Rothwell <sfr@canb.auug.org.au>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

[Bluetooth] Fix socket locking in hci_sock_dev_event()

hci_sock_dev_event() uses bh_lock_sock() to lock the socket lock.
This is not deadlock-safe against locking of the same socket lock in
l2cap_connect_cfm() from softirq context. In addition to that,
hci_sock_dev_event() doesn't seem to be called from softirq context,
so it is safe to use lock_sock()/release_sock() instead.

The lockdep warning can be triggered on my T42p simply by switching
the Bluetooth off by the keyboard button.

  =================================
  [ INFO: inconsistent lock state ]
  2.6.21-rc2 #4
  ---------------------------------
  inconsistent {in-softirq-W} -> {softirq-on-W} usage.
  khubd/156 [HC0[0]:SC0[0]:HE1:SE1] takes:
   (slock-AF_BLUETOOTH){-+..}, at: [<e0ca5520>] hci_sock_dev_event+0xa8/0xc5 [bluetooth]
  {in-softirq-W} state was registered at:
    [<c012d1db>] mark_lock+0x59/0x414
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<c012dfd7>] __lock_acquire+0x3e5/0xb99
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<c012e7f2>] lock_acquire+0x67/0x81
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<c036ee72>] _spin_lock+0x29/0x34
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<e0cef688>] l2cap_connect_cfm+0x4e/0x11f [l2cap]
    [<e0ca17c3>] hci_send_cmd+0x126/0x14f [bluetooth]
    [<e0ca4ce4>] hci_event_packet+0x729/0xebd [bluetooth]
    [<e0ca205b>] hci_rx_task+0x2a/0x20f [bluetooth]
    [<e0ca209d>] hci_rx_task+0x6c/0x20f [bluetooth]
    [<c012d7be>] trace_hardirqs_on+0x10d/0x14e
    [<c011ac85>] tasklet_action+0x3d/0x68
    [<c011abba>] __do_softirq+0x41/0x92
    [<c011ac32>] do_softirq+0x27/0x3d
    [<c0105134>] do_IRQ+0x7b/0x8f
    [<c0103dec>] common_interrupt+0x24/0x34
    [<c0103df6>] common_interrupt+0x2e/0x34
    [<c0248e65>] acpi_processor_idle+0x1b3/0x34a
    [<c0248e68>] acpi_processor_idle+0x1b6/0x34a
    [<c010232b>] cpu_idle+0x39/0x4e
    [<c04bab0c>] start_kernel+0x372/0x37a
    [<c04ba42b>] unknown_bootoption+0x0/0x202
    [<ffffffff>] 0xffffffff
Signed-off-by: NJiri Kosina <jkosina@suse.cz>
Acked-by: NMarcel Holtmann <marcel@holtmann.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Also PTRACE_OLDSETOPTIONS should be accepted, as done by kernel/ptrace.c
and forced by binary compatibility.  UML/32bit breaks because of this -
since it is wise enough to use PTRACE_OLDSETOPTIONS to be binary compatible
with 2.4 host kernels.

Until 2.6.17 (commit f0f2d653) we had:

       default:
                return sys_ptrace(request, pid, addr, data);

Instead here we have:
        case PTRACE_GET_THREAD_AREA:
	case ...:
                return sys_ptrace(request, pid, addr, data);

        default:
                return -EINVAL;

This change was a style change - when a case is added, it must be
explicitly tested this way.  In this case, not enough testing was done.

Cc: Andi Kleen <ak@suse.de>
Signed-off-by: NPaolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: <stable@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Avoid reusing userspace errno twice - it can be cleared by libc code
everywhere (in particular printk() does clear it in my setup).
Signed-off-by: NPaolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: NJeff Dike <jdike@addtoit.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Avoid returning ENOMEM in case of a duplicate IRQ - ENOMEM was saved into err
earlier.
Signed-off-by: NPaolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: NJeff Dike <jdike@addtoit.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Fix confusion about call context - comments and code are inconsistent and
plain wrong, my fault.
Signed-off-by: NPaolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: NJeff Dike <jdike@addtoit.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Since both UML consoles do not use percpu variables, they may be called when
the cpu is still offline, and they may be marked CON_ANYTIME (this is
documented in kernel/printk.c, grep for CON_ANYTIME to find mentions of this).

Works well in testing done with lock debug enabled, should be safe but is not
needed for next release.

This would probably help also stderr_console.c, but this is yet to test.
Signed-off-by: NPaolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: NJeff Dike <jdike@addtoit.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

os_usr1_signal() is totally unused, os_usr1_process() is used only by TT mode.
Signed-off-by: NPaolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: NJeff Dike <jdike@addtoit.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Memory allocated by mcast_user_init must be freed in the matching mcast_remove.
Signed-off-by: NPaolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: NJeff Dike <jdike@addtoit.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

When a given host directory is specified to be mounted both in hostfs=path1
and with mount option -o path2, we should give access to path1/path2, but this
does not happen.  Fix that in the simpler way.

Also, root_ino can be the empty string, since we use %s/%s as format.
Signed-off-by: NPaolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: NJeff Dike <jdike@addtoit.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Fix double free in the error path - when name is assigned into root_inode we
do not own it any more and we must not kfree() it - see patch for details.

Thanks to William Stearns for the initial report.

CC: William Stearns <wstearns@pobox.com>
Signed-off-by: NPaolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: NJeff Dike <jdike@addtoit.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/dtor/input:
  Input: i8042 - another attempt to fix AUX delivery checks

Do not assume that AUX_LOOP command is broken unless it
completes successfully but returns wrong (unexpected) data.
Signed-off-by: NDmitry Torokhov <dtor@mail.ru>

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [BRIDGE]: adding new device to bridge should enable if up
  [IPV6]: Do not set IF_READY if device is down
  [IPSEC]: xfrm audit hook misplaced in pfkey_delete and xfrm_del_sa
  [IPSEC]: Add xfrm policy change auditing to pfkey_spdget
  [IPSEC]: xfrm_policy delete security check misplaced
  [CONNECTOR]: Bugfix for cn_call_callback()
  [DCCP]: Revert patch which disables bidirectional mode
  [IPV6]: Handle np->opt being NULL in ipv6_getsockopt_sticky().
  [UDP]: Reread uh pointer after pskb_trim
  [NETFILTER]: nfnetlink_log: fix crash on bridged packet
  [NETFILTER]: nfnetlink_log: zero-terminate prefix
  [NETFILTER]: nf_conntrack_ipv6: fix incorrect classification of IPv6 fragments as ESTABLISHED

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] ARC: Fix several compiler warnings.
  [MIPS] ISA: Fix typo
  [CHAR] ds1286: Fix handling of seconds in RTC_ALM_SET ioctl.

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6:
  [IA64] fsys_getcpu for IA64
  [IA64] remove duplicate declaration of efi_initialize_iomem_resources
  [IA64] Pick highest possible saved_max_pfn for crash_dump
  [IA64] fix NULL pointer in ia64/irq_chip-mask/unmask function

  CC      arch/mips/arc/init.o
arch/mips/arc/init.c: In function 'prom_init':
arch/mips/arc/init.c:27: warning: ISO C90 forbids mixed declarations and code
arch/mips/arc/init.c:35: warning: format ‘%08lx’ expects type ‘long unsigned int’, but argument 2 has type 'ULONG'
arch/mips/arc/init.c:28: warning: unused variable 'c'
arch/mips/arc/init.c:27: warning: unused variable ‘cnt’
Signed-off-by: NRalf Baechle <ralf@linux-mips.org>

Lost CONFIG_ prefix discovered by Robert P. J. Day <rpjday@mindspring.com>.
Signed-off-by: NRalf Baechle <ralf@linux-mips.org>

 o Fix use of uninitialized variable sec.
 o Make the RTC_ALM_SET ioctl return -EINVAL for non-zero seconds - the
   DS1286 has no second field for the alarm time.
 o Replace the obscure BIN_TO_BCD macro with BIN2BCD.
Signed-off-by: NRalf Baechle <ralf@linux-mips.org>

On 1.6GHz Montectio Tiger4, the following performance data is measured with
kernel built with defconfig which has NUMA configured:

Fastest sys_getcpu: 502 itc counts.
Fastest fsys_getcpu: 28 itc counts.

fsys_getcpu performance is largly impacted by whether data (node_to_cpu_map
etc) is in cache. It can take fsys_getcpu up to ~150 itc counts in cold
cache case.
Signed-off-by: NFenghua Yu <fenghua.yu@intel.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

efi_initialize_iomem_resources() is declared in both include/linux/efi.h
and arch/ia64/kernel/setup.c. This patch removes the latter.
Signed-off-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NTony Luck <tony.luck@intel.com>

Berhhard Walle noted that on his HP rx8640 he ended up with saved_max_pfn
smaller than the highest address of system ram in /proc/iomem and proposed
a patch to base the address on the unrounded and unfiltered EFI memory
map address.  Simon Horman and Magnus Damm suggested that the whole test
be moved earlier in the function.  This is the combination of both of
these patches.
Signed-off-by: NTony Luck <tony.luck@intel.com>

One change introduced by the workqueue removal patch is that adding an
interface that is up to a bridge which is also up does not ever call
br_stp_enable_port(), leaving the port in DISABLED state until we do
ifconfig down and up or link events occur.

The following patch to the br_add_if function fixes it.
This is a regression introduced in 2.6.21.

Submitted-by: Aji_Srinivas@emc.com
Signed-off-by: NStephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>