* master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6: (25 commits)
  [SCSI] mptfc: race between mptfc_register_dev and mptfc_target_alloc
  [SCSI] lpfc 8.1.6 : Fix Data Corruption in Bus Reset Path
  [SCSI] mptspi: revalidate negotiation parameters after host reset and resume
  [SCSI] srp.h: avoid padding of structs
  [SCSI] ibmvscsi: fix leak when failing to send srp event
  [SCSI] qla2xxx: Correct eh_abort recovery logic.
  [SCSI] megaraid_{mm,mbox}: fix a bug in reset handler
  [SCSI] fusion - bug fix stack overflow in mptbase
  [SCSI] scsi: Add IBM 2104-DU3 to blist
  [SCSI] Fix DVD burning issues.
  [SCSI] SCSI: aic7xxx_osm_pci resource leak fix.
  [SCSI] - fusion - mptfc bug fix's to prevent deadlock situations
  [SCSI] mptfusion: bug fix's for raid components adding/deleting
  [SCSI] aic7xxx: ahc_pci_write_config() fix
  [SCSI] megaraid: unused variable
  [SCSI] qla2xxx: only free_irq() after request_irq() succeeds
  [SCSI] Overrun in drivers/scsi/sim710.c
  [SCSI] lpfc 8.1.5 : Change version number to 8.1.5
  [SCSI] lpfc 8.1.5 : Misc small fixes
  [SCSI] lpfc 8.1.5 : Additional fixes to LOGO, PLOGI, and RSCN processing
  ...

Don't recurse back into the driver even if the unplug threshold is met,
when the driver asks for a requeue. This is both silly from a logical
point of view (requeues typically happen due to driver/hardware
shortage), and also dangerous since we could hit an endless request_fn
-> requeue -> unplug -> request_fn loop and crash on stack overrun.

Also limit blk_run_queue() to one level of recursion, similar to how
blk_start_queue() works.

This patch fixed a real problem with SLES10 and lpfc, and it could hit
any SCSI lld that returns non-zero from it's ->queuecommand() handler.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Eric Biederman points out that we can't take the task_lock while holding
tasklist_lock for writing, because another CPU that holds the task lock
might take an interrupt that then tries to take tasklist_lock for writing.

Which would be a nasty deadlock, with one CPU spinning forever in an
interrupt handler (although admittedly you need to really work at
triggering it ;)

Since the ptrace_attach() code is special and very unusual, just make it
be extra careful, and use trylock+repeat to avoid the possible deadlock.

Cc: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  IB/mthca: FMR ioremap fix
  IPoIB: Free child interfaces properly
  IB/mthca: Fix race in reference counting
  IB/srp: Fix tracking of pending requests during error handling
  IB: Fix display of 4-bit port counters in sysfs

Addresses for ioremap must be calculated off of pci_resource_start;
we can't directly use the bus address as seen by the HCA.  Fix the
code that remaps device memory for FMR access.

Based on patch by Klaus Smolin.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

* 'upstream' of master.kernel.org:/pub/scm/linux/kernel/git/shemminger/netdev-2.6:
  sis900: phy for FoxCon motherboard
  dl2k: use DMA_48BIT_MASK constant
  phy: mdiobus_register(): initialize all phy_map entries
  sky2: ifdown kills irq mask

661FX7MI-S motherboard which uses the SiS 661FX chipset.  The patch adds
an entry to mii_chip_info for the transceiver.

The PHY ids were found using the sis900_c_122.diff patch from
http://brownhat.org/sis900.html but that patch didn't solve the problem,
because the PHY at address 1 was already being chosen.

Without my patch, when bursts of packets arrive from other hosts on a
LAN, the interface dropped one roughly 10% of the time, causing
retransmits.  There were fifth second pauses in refresh of large xterms,
and it made Netrek suck.  I can provide further test data.

Workaround in lieu of patch is to use mii-tool to advertise
100baseTx-HD, then force renegotiation.

I wasn't able to identify the actual transceiver, so the description
field is a guess.

This patch is similar to Artur Skawina's patch:
http://marc.theaimsgroup.com/?l=linux-netdev&m=114297516729079&w=2

I'm not sure, but I wonder if it means the default behaviour should be
changed, so as to better handle future transceivers.

Diff is against 2.6.16.13.
Signed-off-by: James Cameron <james.cameron@hp.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

Typo will be harder with this one.
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

make sure phy_map entries whose PHY address is masked are initialized
to NULL, given that other code (such as mdiobus_unregister for
instance) assumes that non-NULL phy_map entries are allocated
phy_devices
Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

Bringing down a port also masks off the status and other IRQ's
needed for device to function due to missing paren's.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

This patch corrects the order of the calls to register_chrdev() and
pcmcia_register_driver().  Now udev correctly creates userspace device
files /dev/cmmN and /dev/cmxN respectively.

Based on an earlier patch by Jan Niehusmann <jan@gondor.com>.
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

A race condition exists in mptfc between the thread registering a device
with the fc transport and the scan work generated by the transport.
This race existed prior to the application of the mptfc bug fix patch.

mptfc_register_dev() calls fc_remote_port_add() with the FC_RPORT_ROLE_TARGET
bit set in the rport ids passed to the function.  Having this bit set causes
fc_remote_port_add() to schedule a scan of the device.

This scan can execute before mptfc_register_dev() can fill in the dd_data
in the rport structure.  When this happens, mptfc_target_alloc() will fail
because dd_data is null.

Attached is a patch which fixes the problem.  The patch changes the rport ids
passed to fc_remote_port_add() to not have the TARGET bit set.  This prevents
the scan from being scheduled.  After mptfc_register_dev() fills in the rport
dd_data field, fc_remote_port_rolechg() is called, changing the role of the
rport to TARGET.  Thus, the scan is scheduled after dd_data is filled
in which prevents the failure in mptfc_target_alloc().
Signed-off-by: Michael Reed <mdr@sgi.com>
Signed-off-by: Eric Moore <Eric.Moore@lsil.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

When deleting a child interface with a non-default P_Key via
/sys/class/net/ibX/delete_child, the interface must be freed with
free_netdev() (rather than kfree() on the private data).
Signed-off-by: Roland Dreier <rolandd@cisco.com>

Fix races in in destroying various objects.  If a destroy routine
waits for an object to become free by doing

	wait_event(&obj->wait, !atomic_read(&obj->refcount));
	/* now clean up and destroy the object */

and another place drops a reference to the object by doing

	if (atomic_dec_and_test(&obj->refcount))
		wake_up(&obj->wait);

then this is susceptible to a race where the wait_event() and final
freeing of the object occur between the atomic_dec_and_test() and the
wake_up().  And this is a use-after-free, since wake_up() will be
called on part of the already-freed object.

Fix this in mthca by replacing the atomic_t refcounts with plain old
integers protected by a spinlock.  This makes it possible to do the
decrement of the reference count and the wake_up() so that it appears
as a single atomic operation to the code waiting on the wait queue.

While touching this code, also simplify mthca_cq_clean(): the CQ being
cleaned cannot go away, because it still has a QP attached to it.  So
there's no reason to be paranoid and look up the CQ by number; it's
perfectly safe to use the pointer that the callers already have.
Signed-off-by: Roland Dreier <rolandd@cisco.com>

If a SCSI abort completes, or the command completes successfully, then
the driver must remove the command from its queue of pending
commands.  Similarly, if a device reset succeeds, then all commands
queued for the given device must be removed from the queue.
Signed-off-by: Roland Dreier <rolandd@cisco.com>

The code to display local_link_integrity_errors and
excessive_buffer_overrun_errors in
/sys/class/infiniband/<hca>/ports/<n>/counters/
uses the wrong shift to extract the 4 bit values.
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

* 'upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/shemminger/netdev-2.6:
  [PATCH] bcm43xx: Fix access to non-existent PHY registers
  [PATCH] bcm43xx: Fix array overrun in bcm43xx_geo_init
  [PATCH] bcm43xx: check for valid MAC address in SPROM
  [PATCH] ieee80211: Fix A band channel count (resent)
  [PATCH] bcm43xx: fix iwmode crash when down
  [PATCH] softmac: make non-operational after being stopped
  [PATCH] softmac: don't reassociate if user asked for deauthentication
  spidernet: enable support for bcm5461 ethernet phy
  spidernet: introduce new setting
  Fix RTL8019AS init for Toshiba RBTX49xx boards
  au1000_eth.c: use ether_crc() from <linux/crc32.h>
  sky2: version 1.3
  Add more support for the Yukon Ultra chip found in dual core centino laptops.
  sky2: synchronize irq on remove
  sky2: dont write status ring
  sky2: edge triggered workaround enhancement
  sky2: use mask instead of modulo operation
  sky2: tx ring index mask fix
  sky2: status irq hang fix
  sky2: backout NAPI reschedule

Instantiation of 8MB pages on the TLB cache for the kernel static
mapping trashes r3 register on !CONFIG_8xx_CPU6 configurations.
This ensures r3 gets saved and restored.
Signed-off-by: Marcelo Tosatti <marcelo@kvack.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>

My commit 6bfd93c3 broke the ARCH=ppc
compilation by using the is_kernel_addr() macro in asm/uaccess.h.
This fixes it by defining a suitable is_kernel_addr() for ARCH=ppc.
Signed-off-by: Paul Mackerras <paulus@samba.org>

* git://oss.sgi.com:8090/xfs-2.6:
  [XFS] Fix a possible metadata buffer (AGFL) refcount leak when fixing an
  [XFS] Fix a project quota space accounting leak on rename.
  [XFS] Fix a possible forced shutdown due to mishandling write barriers

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

A newer board revision changed the type of ethernet phy.
Moreover, this generalizes the way that a phy gets switched
into fiber mode when autodetection is not available.
Signed-off-by: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

We found a new chip setting that we need in order
to make the driver work more reliable.
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

    Ensure that 8-bit mode is selected for the on-board Realtek RTL8019AS chip
on Toshiba RBHMA4x00, get rid of the duplicate #ifdef's when setting
ei_status.word16.
    The chip's datasheet says that the PSTOP register shouldn't exceed 0x60 in
8-bit mode -- ensure this too.
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

since the au1000 driver already selects the CRC32 routines, simply replace
the internal ether_crc() implementation with the semantically equivalent
one from <linux/crc32.h>
Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

Update version number, to track changes.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

The newest Yukon Ultra chipset's require more special tweaks.
They seem to be like the Yukon XL chipsets. This code is transliterated
from the latest SysKonnect driver; I don't have any Ultra hardware.
Signed-off-by: Stephe Hemminger <shemminger@osdl.org>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

Need to make sure interrupt is not racing with unregister of
network device.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

It is more efficient not to write the status ring from the
processor and just read the active portion.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

Need to make the edge-triggered workaround timer faster to get marginally
better peformance. The test_and_set_bit in schedule_prep() acts as a barrier
already. Make it a module parameter so that laptops who are concerned
about power can set it to 0; and user's stuck with broken BIOS's
can turn the driver into pure polling.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

Gcc isn't smart enough to know that it can do a modulo
operation with power of 2 constant by doing a mask.
So add macro to do it for us.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

Mask for transmit ring status was picking up bits from the
unused sync ring.  They were always zero, so far...
Also, make sure to remind self not to make tx ring too big.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

The status interrupt flag should be cleared before processing,
not afterwards to avoid race. Need to process in poll routine
even if no new interrupt status. This is a normal occurrence when
more than 64 frames (NAPI weight) are processed in one poll routine.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

This is a backout of earlier patch.

The whole rescheduling hack was a bad idea. It doesn't really solve
the problem and it makes the code more complicated for no good reason.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

This reverts commit c8d8b837, which
caused problems for the x86 build. Quoth Sam:

  "It was discussed on mips list but apparently the fix was bogus.  I
   will not have time to look into it so mips can carry this local fix
   until we get a proper fix in mainline."
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Taking the cpu hotplug semaphore in a normal events workqueue
is unsafe because other tasks can wait for any workqueues with
it hold. This results in a deadlock.

Move the DBS timer into its own work queue which is not
affected by other work queue flushes to avoid this.

Has been acked by Venkatesh.

Cc: venkatesh.pallipadi@intel.com
Cc: cpufreq@lists.linux.org.uk
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Based on analysis&patch from Robert Hentosch

Observed on a Dell PE6850 with 16GB

The problem occurs very early on, when the kernel allocates space for the
temporary memory map called bootmap. The bootmap overlaps the EBDA region.
EBDA region is not historically reserved in the e820 mapping. When the
bootmap is freed it marks the EBDA region as usable.

If you notice in setup.c there is already code to work around the EBDA
in reserve_ebda_region(), this check however occurs after the bootmap
is allocated and doesn't prevent the bootmap from using this range.

AK: I redid the original patch. Thanks also to Jan Beulich for
spotting some mistakes.

Cc: Robert_Hentosch@dell.com
Cc: jbeulich@novell.com
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Playing with NMI watchdog on x86_64, I discovered that it didn't
do what I expected.  It always panic-ed, even when it didn't
happen from interrupt context.  This patch solves that
problem for me.  Also, in this case, do_exit() will be called
with interrupts disabled, I believe.  Would it be wise to also
call local_irq_enable() after nmi_exit()?
[Yes I added it -AK]

Currently, on x86_64, any NMI watchdog timeout will cause a panic
because the irq count will always be set to be in an interrupt
when do_exit() is called from die_nmi().  If we add nmi_exit() to
the die_nmi() call (since the nmi will never exit "normally")
it seems to solve this problem.  The following small program
can be used to trigger the NMI watchdog to reproduce this:
  main ()
  {
        iopl(3);
        for (;;) asm("cli");
  }
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

I noticed this when poking around in this area.

The oops_begin() function in x86_64 would only conditionally claim
the die_lock if the call is nested, but oops_end() would always
release the spinlock. This patch adds a nest count for the die lock
so that the release of the lock is only done on the final oops_end().
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

The IOMMU code can only deal with 8 northbridges. Error out when
more are found.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>