Recently dm-multipath began calling blk_abort_queue. This causes all the
commands/request running on the path to have the timeout function called.
If a path does go down, and the LLD returns DID_*, dm-multpiath will eventually
get this error and begin to call the cmd timeout handler. qla4xxx currently
does not set a timed out handler and so the default one could return
BLK_EH_NOT_HANDLED and end up firing the scsi eh and stopping IO to all
paths on the host when only one path is affected.

For software and offload iscsi we have a timed out handler already.
This patch adds a driver specific one to qla4xxx because there
are some ddb->state and session->state and command completion races
that are better handled in the LLD.

This also handles the problem where if the session is down,
we do not need the scsi eh to run until the transport code
has tried to reconnect us.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

Fixed sense data errors occurring above the first 32 bytes,
as required by some third party applications.  Sense data
in the first 32 bytes has always been correct.

Patch updated to use srb data variables instead of scsi command
scratchpad data area, as scratchpad area is already used.

Also, corrected debug print alignment bug in dump_buffer routine.
Changed KERN_DEBUG to KERN_INFO in printk statements in this routine.

Changed version number to 5.01.00-k9
Signed-off-by: Karen Higgins <karen.higgins@qlogic.com>
[michaelc: fixed checkpath.pl errors]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

The session lock can be held in the scsi eh thread or the completion
paths run from the net softirq. This disables bhs in iscsi_eh_abort when
taking the session lock.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

The trace record for SCSI abort requests has a field for the request
id of the request to be aborted. Put the real request id instead of
zero.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

Under certain conditions it is possible that a WKA port ist not opened
within the expected timeframe of half a second. In this situation
the WKA port remains in the state OPENING preventing any succeding
request to open the port. This led to unrecoverable remote ports.
Fixing this by always setting an appropriate WKA port status before
leaving the function and removing the timeout value here since it's
not needed here because the general timeout processing would deal
with it if required.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

In a LOWMEM condition an ERP notification would have been sent twice
causing an unpredictable behaviour of the ERP.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

When calling fc_remote_port_add make sure to not call it again before
fc_remote_port_delete has been called. In other words, ensure to
create a new fc_rport, then delete it, then create a new one again.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

Depending on interruptions on some storage systems, the complete
channel can stall which looks like an outbound queue stall to Linux.
When trying to acquire a free SBAL for a non-SCSI command, zfcp waits
for 5 seconds for a free slot to appear. This is the right place to
detect a queue stall: If the wait times out, we assume a stalled queue
and try to recover this.

The overall strategy should be to trigger the erp from specific
events, and not try an overall escalation from one failed port to a
full-blown queue recovery. If we manage to send a command, the status
codes for this command or a timeout will trigger the right follow-on
actions.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

If an action fails, retry it until the erp count exceeds the
threshold. If there is something fundamentally wrong, the FSF layer
will trigger a more appropriate action depending on the FSF status
codes.

The followup for successful actions is a different followup than
retrying failed actions, so split the code two functions to make this
clear.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

After closing the port, we want it to be "not open" to consider the
action to be successful.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

-ENOMEM is for memory allocation problems, -EIO for queue/SBAL
allocation problems.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

The ELS ADISC and the GID_PN requests sent from zfcp fit into
unchained FSF requests. Change the FSF allocation logic to use
unchained requests whenever possible where everything fits in one
SBAL. This avoids acquiring more SBALs than necessary, especially
during zfcp recovery when things might be stalled.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

zfcp_erp_notify uses the ZFCP_ERP_STATUS_* flags, so it is
ZFCP_STATUS_ERP_LOWMEM instead of ZFCP_ERP_NOMEM. Signalling
ZFCP_ERP_FAILED is not necessary, the missing d_id will show that the
nameserver did not return the d_id.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

When a fsf_req or a qtcb cannot be allocated return -ENOMEM instead of
-EIO.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

req_q_util is not atomic, so the qdio_stat_lock must be held when
reading this variable.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

We should not modify the port status after triggering an ERP action
for the port. It is not guaranteed which status is finally active
when the ERP action is performed. This can lead to situations which
are unwanted and hard to debug in case of a failure.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

There's a hotplug problem in the way libsas allocates ports: it loops over the
available ports first trying to add to an existing for a wide port and
otherwise allocating the next free port.  This scheme only works if the port
array is packed from zero, which fails if a port gets hot unplugged and the
array becomes sparse.  In that case, a new port is formed even if there's a
wide port it should be part of.  Fix this by creating two loops over all the
ports:  the first to see if the phy should be part of a wide port and the
second to form a new port in an empty port slot.
Signed-off-by: Tom Peng <tom_peng@usish.com>
Signed-off-by: Jack Wang <jack_wang@usish.com>
Signed-off-by: Lindar Liu <lindar_liu@usish.com>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  ahci: add device ID for 82801JI sata controller
  drivers/ata: Move a dereference below a NULL test
  libata: implement and use HORKAGE_NOSETXFER, take#2
  libata: fix follow-up SRST failure path

Add device ID for Intel 82801JI SATA AHCI controller.
Signed-off-by: David Milburn <dmilburn@redhat.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

If the NULL test is necessary, then the dereference should be moved below
the NULL test.

The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@
type T;
expression E;
identifier i,fld;
statement S;
@@

- T i = E->fld;
+ T i;
  ... when != E
      when != i
  if (E == NULL) S
+ i = E->fld;
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

PIONEER DVD-RW DVRTD08 times out SETXFER if no media is present.  The
device is SATA and simply skipping SETXFER works around the problem.
Implement ATA_HORKAGE_NOSETXFER and apply it to the device.

Reported by Moritz Rigler in the following thread.

  http://thread.gmane.org/gmane.linux.ide/36790

and by Lars in bko#9540.

Updated to whine and ignore NOSETXFER if PATA component is detected as
suggested by Alan Cox.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Moritz Rigler <linux-ide@momail.e4ward.com>
Reported-by: Lars <lars21ce@gmx.de>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

ata_eh_reset() was missing error return handling after follow-up SRST
allowing EH to continue the normal probing path after reset failure.
This was discovered while testing new WD 2TB drives which take longer
than 10 secs to spin up and cause the first follow-up SRST to time
out.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

* 'i2c-for-2631-rc3' of git://aeryn.fluff.org.uk/bjdooks/linux:
  i2c: Use resource_size
  i2c-davinci: behave with i2cdetect
  i2c-davinci: convert clock usage after clkdev conversion

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/teigland/dlm:
  dlm: free socket in error exit path
  dlm: fix plock use-after-free
  dlm: Fix uninitialised variable warning in lock.c

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  mlx4_core: Add new ConnectX EN PCI ID 0x6764
  mlx4_core: Handle multi-physical function devices

Merge branch 'timers-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'timers-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  hrtimer: Fix migration expiry check
  hrtimer: migration: do not check expiry time on current CPU

Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: Fix warning in pvclock.c
  x86, apic: Fix false positive section mismatch in numaq_32.c
  x86: Fix false positive section mismatch in es7000_32.c
  x86: Remove spurious printk level from segfault message

Merge branch 'core-fixes-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'core-fixes-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  futexes: Fix infinite loop in get_futex_key() on huge page

Merge branch 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  tracing/function-profiler: do not free per cpu variable stat
  tracing/events: Move TRACE_SYSTEM outside of include guard

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  Revert "NET: Fix locking issues in PPP, 6pack, mkiss and strip line disciplines."
  skbuff.h: Fix comment for NET_IP_ALIGN
  drivers/net: using spin_lock_irqsave() in net_send_packet()
  NET: phy_device, fix lock imbalance
  gre: fix ToS/DiffServ inherit bug
  igb: gcc-3.4.6 fix
  atlx: duplicate testing of MCAST flag
  NET: Fix locking issues in PPP, 6pack, mkiss and strip line disciplines.
  netdev: restore MTU change operation
  netdev: restore MAC address set and validate operations
  sit: fix regression: do not release skb->dst before xmit
  net: ip_push_pending_frames() fix
  net: sk_prot_alloc() should not blindly overwrite memory

Use the function resource_size, which reduces the chance of introducing
off-by-one errors in calculating the resource size.

The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@
struct resource *res;
@@

- (res->end - res->start) + 1
+ resource_size(res)
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>

Make i2c-davinci cope properly with "i2cdetect":  don't spew
syslog spam on perfectly normal behaviors, or respond to any
address other than the one reserved for the SMBus host.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>

DaVinci core code has converted to the new clkdev API so
clock name strings are not needed.  Instead, just the a
'struct device' pointer is needed.
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>

This reverts commit adeab1af.

As Alan Cox explained, the TTY layer changes that went recently
to get rid of the tty->low_latency stuff fixes this already,
and even for -stable it's the ->low_latency changes that should
go in to fix this, rather than this patch.
Signed-off-by: David S. Miller <davem@davemloft.net>

Use the correct function call for skb_reserve in the comment for
NET_IP_ALIGN.
Signed-off-by: Tobias Klauser <klto@zhaw.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

spin_unlock_irq() will enable interrupt in net_send_packet(),
this patch changes it to spin_lock_irqsave/spin_lock_irqrestore,
so that it doesn't enable interrupts when already disabled,
and netconsole would work properly over cs89x0/isa-skeleton.

Call trace:
netconsole write_msg()
{
 ...
 -> spin_lock_irqsave();
        -> netpoll_send_udp()
          -> netpoll_send_skb()
            -> net_send_packet()
              ->...

 -> spin_unlock_irqrestore();
 ...
}
Signed-off-by: Dongdong Deng <dongdong.deng@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Don't forget to unlock a mutex in phy_scan_fixups on a fail path.
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In the tcp_connect_to_sock() error exit path, the socket
allocated at the top of the function was not being freed.
Signed-off-by: Casey Dahlin <cdahlin@redhat.com>
Signed-off-by: David Teigland <teigland@redhat.com>

Fixes two bugs:
- ToS/DiffServ inheritance was unintentionally activated when using impair fixed ToS values
- ECN bit was lost during ToS/DiffServ inheritance
Signed-off-by: Andreas Jaggi <aj@open.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

when building 32-bit, I see this ..
arch/x86/kernel/pvclock.c:63:7: warning: "__x86_64__" is not defined
Signed-off-by: Dave Jones <davej@redhat.com>
LKML-Reference: <20090713201437.GA12165@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>