safe/jmp/linux-2.6
15 years ago[SCSI] libosd: OSD2r05: OSD_CRYPTO_KEYID_SIZE will grow 20 => 32 bytes
Boaz Harrosh [Sun, 19 Apr 2009 16:13:39 +0000 (19:13 +0300)]
[SCSI] libosd: OSD2r05: OSD_CRYPTO_KEYID_SIZE will grow 20 => 32 bytes

In OSD2r04 draft, cryptographic key size changed to 32 bytes from
OSD1's 20 bytes. This causes a couple of on-the-wire structures
to change, including the CDB.

In this patch the OSD1/OSD2 handling is separated out in regard
to affected structures, but on-the-wire is still the same. All
on the wire changes will be submitted in one patch for bisect-ability.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] libosd: OSD2r05: Prepare for rev5 attribute list changes
Boaz Harrosh [Sun, 19 Apr 2009 16:11:42 +0000 (19:11 +0300)]
[SCSI] libosd: OSD2r05: Prepare for rev5 attribute list changes

In OSD2r05 draft each attribute list element header was changed
so attribute-value would be 8 bytes aligned. In OSD2r01-r04
it was aligned on 2 bytes. (This is because in OSD2r01 the complete
element was 8 bytes padded at end but the header was not adjusted
and caused permanent miss-alignment.)

OSD1 elements are not padded and might be or might not be aligned.
OSD1 is still supported.

In this code we do all the code re-factoring to separate OSD1/OSD2
differences but do not change actual wire format. All wire format
changes will happen in one patch later, for bisect-ability.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] libosd: fix potential ERR_PTR dereference in osd_initiator.c
Dan Carpenter [Sun, 19 Apr 2009 16:07:47 +0000 (19:07 +0300)]
[SCSI] libosd: fix potential ERR_PTR dereference in osd_initiator.c

bio_map_kern() returns an ERR_PTR() not NULL.

Found by smatch (http://repo.or.cz/w/smatch.git).  Compile tested.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas : bump driver version to 01.100.02.00
Eric Moore [Tue, 21 Apr 2009 21:44:27 +0000 (15:44 -0600)]
[SCSI] mpt2sas : bump driver version to 01.100.02.00

Bump driver version.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas: fix hotplug event processing
Eric Moore [Tue, 21 Apr 2009 21:43:33 +0000 (15:43 -0600)]
[SCSI] mpt2sas: fix hotplug event processing

Here's a fix for hotplug events.  The useage of queue_delayed_work seems
to broke the fifo for processing of firmware events.  After several iterations
of adding and removing cabling connected to jbods, the devices are not
getting added becuase kernel thread is activited out of order.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas : release diagnotic buffers prior host reset
Eric Moore [Tue, 21 Apr 2009 21:42:13 +0000 (15:42 -0600)]
[SCSI] mpt2sas : release diagnotic buffers prior host reset

Diagnostic buffer support is already there in the driver.  This support allows
applications to pull ring buffers from controller firmware for debugging
firmware related issues.

What this patch does is sends reqeust to firmware to release the buffers prior
to host reset.   This will allow what ever debug info is there prior to reset
to be dma'd to host memory. With out this fix, some of the debug data would
been lost.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas : Broadcast Primative AEN bug fix
Eric Moore [Tue, 21 Apr 2009 21:41:32 +0000 (15:41 -0600)]
[SCSI] mpt2sas : Broadcast Primative AEN bug fix

Bug fix in the broadcast primative async event code where the driver would
stop sending tm queries after the first queury was completed. This was due
driver not reseting the tm_cmds.status field back to MPT2_CMD_NOT_USED after
completing a task management request.

An addtional fix adding sanity check to insure sas_device->starget set to NULL.
During multipath testing fail over/fail back, the mid layer was holding onto
sdev longer than the fail back period, thus starget was getting set to NULL
for device being added.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas : Identify Dell series-7 adapters at driver load time
Eric Moore [Tue, 21 Apr 2009 21:40:48 +0000 (15:40 -0600)]
[SCSI] mpt2sas : Identify Dell series-7 adapters at driver load time

The Dell branding along with the VID, DID, SSVID, SSDID following the LSI
branding that contains the card firmware/chip/bios versions.  If the SSDID
is not known but it is a Dell HBA, the driver will print the SSDID instead
of the Dell branding string.  Nothing will be printed for non Dell HBAs

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas : driver name needs to be in the MPT2IOCINFO ioctl
Eric Moore [Tue, 21 Apr 2009 21:40:01 +0000 (15:40 -0600)]
[SCSI] mpt2sas : driver name needs to be in the MPT2IOCINFO ioctl

The driver name needs to be at the beginining of the driver_version string in
MPT2IOCINFO ioctl.  This is the same behaviour is there already in the mptsas
driver.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas : running out of message frames
Eric Moore [Tue, 21 Apr 2009 21:39:24 +0000 (15:39 -0600)]
[SCSI] mpt2sas : running out of message frames

The driver is not freeing message frame when returning failure from
_ctl_do_task_abort.   If you call this function 500 times when its unable
to find an active task mid, you end up with no message frames.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas : fix oops when firmware sends large sense buffer size
Eric Moore [Tue, 21 Apr 2009 21:38:43 +0000 (15:38 -0600)]
[SCSI] mpt2sas : fix oops when firmware sends large sense buffer size

There is a bug in firmware where the reply message frame says there is a
16kb sense buffer, when in reality its only 20 bytes.  This fix insures
the memcpy action doesn't corrupte the memory beyond the 90 bytes allocated in
the scsi command for sense buffer.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas : the sanity check in base_interrupt needs to be on dword boundary
Eric Moore [Tue, 21 Apr 2009 21:37:57 +0000 (15:37 -0600)]
[SCSI] mpt2sas : the sanity check in base_interrupt needs to be on dword boundary

The poison sanity check on the reply_post_free register needs to be by 32bit,
not 64bit. The poison check is there because its possible that the driver read
the 1st 32bit before the 2nd 32bit has been written to by firmware.  In other
words, this handles race between driver reading the 64 bit register, and it
being dma'd across pci memory from controller firmware as two 32bit pci writes.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] mpt2sas : unique ioctl magic number
Eric Moore [Tue, 21 Apr 2009 21:37:20 +0000 (15:37 -0600)]
[SCSI] mpt2sas : unique ioctl magic number

The current magic number is shared with mptsas driver. This to be unique to
fix issues with register_ioctls32_conversion in older kernels.  We are making
this change across all versions of the sas2.0 drivers.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] fix sign extension with 1.5TB usb-storage LBD=y
Dave Hansen [Tue, 21 Apr 2009 23:43:27 +0000 (16:43 -0700)]
[SCSI] fix sign extension with 1.5TB usb-storage LBD=y

Shifting an unsigned char implicitly casts it to a signed int.  This
caused 'lba' to sign-extend and Linux would then try READ CAPACITY 16
which was not supported by at least one drive.  Using the
get_unaligned_be*() helpers keeps us from having to worry about how the
extension might occur.

Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] ipr: Fix sleeping function called with interrupts disabled
Brian King [Wed, 22 Apr 2009 13:58:02 +0000 (08:58 -0500)]
[SCSI] ipr: Fix sleeping function called with interrupts disabled

The ata_sas_slave_configure was changed such that it now allocates
some memory for a drain buffer for ATAPI devices. Fixup the ipr
driver such that we no longer make this call with interrupts disabled.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] fcoe: fip: add multicast filter to receive FIP advertisements.
Joe Eykholt [Tue, 21 Apr 2009 23:27:46 +0000 (16:27 -0700)]
[SCSI] fcoe: fip: add multicast filter to receive FIP advertisements.

The FCoE forwarder (FCF) would be selected, but then would soon time
out after three advertisements were missed.  This would be 24 seconds
by default, or 3 times the keep-alive interval configured on the switch.

The cause was that the multicast address for all FIP E-nodes
was never added.

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] libfc: Fix compilation warnings with allmodconfig
Robert Love [Tue, 21 Apr 2009 23:27:41 +0000 (16:27 -0700)]
[SCSI] libfc: Fix compilation warnings with allmodconfig

When building with a .config generated from 'make allmodconfig'
some build warnings are generated. This patch corrects the warnings,
adds a FC_FID_NONE (= 0) enumeration for FC-IDs and cleans up one
variable naming to meet our variable naming conventions. For example,
fc_lport's should be named "lport," not "lp."

Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] fcoe: fix spelling typos and bad comments
Chris Leech [Tue, 21 Apr 2009 23:27:36 +0000 (16:27 -0700)]
[SCSI] fcoe: fix spelling typos and bad comments

Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] fcoe: don't export functions that are internal to fcoe
Chris Leech [Tue, 21 Apr 2009 23:27:30 +0000 (16:27 -0700)]
[SCSI] fcoe: don't export functions that are internal to fcoe

These probably never should have been exported.
If they were needed outside of the fcoe module, they
would have been moved to libfcoe.

Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] fcoe: kfree() -> kfree_skb()
Dan Carpenter [Tue, 21 Apr 2009 23:27:25 +0000 (16:27 -0700)]
[SCSI] fcoe: kfree() -> kfree_skb()

sk_buff pointers should use kfree_skb() instead of vanilla kfree().

Found by smatch (http://repo.or.cz/w/smatch.git).

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] libfc: whenever queueing delete ev for rport, set state to NONE
Abhijeet Joglekar [Tue, 21 Apr 2009 23:27:19 +0000 (16:27 -0700)]
[SCSI] libfc: whenever queueing delete ev for rport, set state to NONE

When a delete event is queued for an rport, set state to NONE so that no
other processing is done on the rport as it is being removed.

Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] libfc: Change state to NONE in fc_lport_destroy
Abhijeet Joglekar [Tue, 21 Apr 2009 23:27:14 +0000 (16:27 -0700)]
[SCSI] libfc: Change state to NONE in fc_lport_destroy

After lport_destroy, the local port should not be used again. Transition
to state NONE, any incoming frames or link up should not transition out
of this state since we are deleting exchange table and cleaning up the
local port. Also, mark link as down.

Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] libfc: During fabric logoff, flush the rport Q after logging off dns port
Abhijeet Joglekar [Tue, 21 Apr 2009 23:27:09 +0000 (16:27 -0700)]
[SCSI] libfc: During fabric logoff, flush the rport Q after logging off dns port

We want to generate the rport queue event (from the logoff)
before flushing the queue otherwise the event may still be
in the queue when we logoff.

Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] libfc: Track rogue remote ports
Abhijeet Joglekar [Tue, 21 Apr 2009 23:27:04 +0000 (16:27 -0700)]
[SCSI] libfc: Track rogue remote ports

Rogue ports are currently not tracked on any list. The only reference
to them is through any outstanding exchanges pending on the rogue ports.
If the module is removed while a retry is set on a rogue port
(say a Plogi retry for instance), this retry is not cancelled because there
is no reference to the rogue port in the discovery rports list. Thus the
local port can clean itself up, delete the exchange pool, and then the
rogue port timeout can fire and try to start up another exchange.

This patch tracks the rogue ports in a new list disc->rogue_rports. Creating
a new list instead of using the disc->rports list keeps remote port code
change to a minimum.

1)  Whenever a rogue port is created, it is immediately added to the
disc->rogue_rports list.

2) When the rogues port goes to ready, it is removed from the rogue list
and the real remote port is added to the disc->rports list

3) The removal of the rogue from the disc->rogue_rports list is done in
the context of the fc_rport_work() workQ thread in discovery callback.

4) Real rports are removed from the disc->rports list like before. Lookup
is done only in the real rports list. This avoids making large changes
to the remote port code.

5) In fc_disc_stop_rports, the rogues list is traversed in addition to the
real list to stop the rogue ports and issue logoffs on them. This way, rogue
ports get cleaned up when the local port goes away.

6) rogue remote ports are not removed from the list right away, but
removed late in fc_rport_work() context, multiple threads can find the same
remote port in the list and call rport_logoff(). Rport_logoff() only
continues with the logoff if port is not in NONE state, thus preventing
multiple logoffs and multiple list deletions.

7) Since the rport is removed from the disc list at a later stage
(in the disc callback), incoming frames can find the rport even if
rport_logoff() has been called on the rport. When rport_logoff() is called,
the rport state is set to NONE, and we are trying to cancel all exchanges
and retries on that port. While in this state, if an incoming
Plogi/Prli/Logo or other frames match the rport, we should not reply
because the rport is in the NONE state. Just drop the frame, since the
rport will be deleted soon in the disc callback (fc_rport_work)

8)  In fc_disc_single(), remove rport lookup and call to fc_disc_del_target.
fc_disc_single() is called from recv_rscn_req() where rport lookup
and rport_logoff is already done.

Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] libfc: Do not retry if the new state is not the same as old state
Abhijeet Joglekar [Tue, 21 Apr 2009 23:26:58 +0000 (16:26 -0700)]
[SCSI] libfc: Do not retry if the new state is not the same as old state

For instance, if there is a Plogi pending (remote port is in Plogi state),
and the state changes to say NONE (because the port is being logged off),
then when the Plogi resp times out, do not start a retry.

This patch partially reverts an earlier patch (libfc: check for err when
recv and state is incorrect), by moving the state check back to before
checking for error. However, if the state does not match, then there is
an additional check to see if its an error ptr or a real frame before
jumping to err or out respectively.

Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] libfc: Hold disc mutex while processing gpn ft resp
Abhijeet Joglekar [Tue, 21 Apr 2009 23:26:52 +0000 (16:26 -0700)]
[SCSI] libfc: Hold disc mutex while processing gpn ft resp

gpn_ft_resp processing currently does not hold the discovery lock.
disc_done() thus gets called from gpn_ft_resp or from gpn_ft_parse
without the lock held. This then sets disc->pending to zero or calls
gpn_ft_req() without disc_lock held.

- Hold disc mutex during gpn_ft resp processing
- In disc_done, release the disc mutex while calling lport callback

Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] cxgb3i: fix ddp map overrun
kxie@chelsio.com [Tue, 21 Apr 2009 20:32:35 +0000 (15:32 -0500)]
[SCSI] cxgb3i: fix ddp map overrun

(version 2)

Fixed a bug in calculating ddp map range when search for free entries:
it was going beyond the end by one, thus corrupting gl_skb[0].

Signed-off-by: Karen Xie <kxie@chelsio.com>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] cxgb3i: fix cpu use abuse during writes
Mike Christie [Tue, 21 Apr 2009 20:32:34 +0000 (15:32 -0500)]
[SCSI] cxgb3i: fix cpu use abuse during writes

When doing a lot (128) of large writes (256K) we can hit the cxgb3_snd_win
check pretty easily. The driver's xmit thread then takes 100% of the cpu.

The driver should not be returning -EAGAIN for this problem. It should
be returing -ENOBUFS, then when the window is opened again it should
queue the xmit thread (it already wakes the xmit thread).

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] cxgb3i: fix can_queue and cmd_per_lun initialization
Mike Christie [Tue, 21 Apr 2009 20:32:33 +0000 (15:32 -0500)]
[SCSI] cxgb3i: fix can_queue and cmd_per_lun initialization

cxgb3i was setting can_queue to only 128 commands, and was
setting the can_queue and cmd_per_lun to the same value.

This sets the can_queue to 1024 commands, and sets the cmd_per_lun
to a safer default of 32.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] cxgb3i, iser, iscsi_tcp: set target can queue
Mike Christie [Tue, 21 Apr 2009 20:32:32 +0000 (15:32 -0500)]
[SCSI] cxgb3i, iser, iscsi_tcp: set target can queue

Set target can queue limit to the number of preallocated
session tasks we have.

This along with the cxgb3i can_queue patch will fix a throughput
problem where it could only queue one LU worth of data at a time.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] iscsi_tcp: don't fire conn error if pdu init fails
Mike Christie [Tue, 21 Apr 2009 20:32:31 +0000 (15:32 -0500)]
[SCSI] iscsi_tcp: don't fire conn error if pdu init fails

If a command's scsi cmd pdu setup fails then we can just fail
the IO to the scsi layer. If a DATA_OUT for a R2T fails then
we will want to drop the session, because it means we got a
bad request from the target (iscsi protocol error).

This patch has us propogate the error upwards so libiscsi_tcp
or libiscsi can decide what the best action is to take. It
also fixes a bug where we could try to grab the session lock
while holding it, because if iscsi_tcp drops the session in the
pdu setup callout the session lock is held when setting up the
scsi cmd pdu.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] zfcp: Fix oops when port disappears
Christof Schmitt [Fri, 17 Apr 2009 13:08:15 +0000 (15:08 +0200)]
[SCSI] zfcp: Fix oops when port disappears

The zfcp_port might have been removed, while the FC fast_io_fail timer
is still running and could trigger the terminate_rport_io callback.
Set the pointer to the zfcp_port to NULL and check accordingly
before using it.

Reviewed-by: Martin Petermann <martin@linux.vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] zfcp: Reference counting for cfdc requests
Christof Schmitt [Fri, 17 Apr 2009 13:08:14 +0000 (15:08 +0200)]
[SCSI] zfcp: Reference counting for cfdc requests

Before dropping the reference count with zfcp_adapter_put, increase it
with zfcp_adapter_get when issuing cfdc requests.

Reviewed-by: Martin Petermann <martin@linux.vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] zfcp: Fix port reference counting
Martin Petermann [Fri, 17 Apr 2009 13:08:13 +0000 (15:08 +0200)]
[SCSI] zfcp: Fix port reference counting

If this problem appears zfcp ports cannot be de-queued since it is
checked for a zero refcount. The port reference counting is wrong for
existing zfcp ports when e.g. an adapter gets on-line again. During
port scanning the reference counting for existing ports should not be
changed.

Signed-off-by: Martin Petermann <martin.petermann@de.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] zfcp: revert previous patch for sbal counting
Martin Petermann [Fri, 17 Apr 2009 13:08:12 +0000 (15:08 +0200)]
[SCSI] zfcp: revert previous patch for sbal counting

The current sbal counting can be wrong if a fsf request is
waiting for free sbals and at the same time qdio request queue
is shutdown and re-opened. Revering a previous patch fixes this
issue.

Signed-off-by: Martin Petermann <martin.petermann@de.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] zfcp: Fix abort handler for completions in progress
Christof Schmitt [Fri, 17 Apr 2009 13:08:11 +0000 (15:08 +0200)]
[SCSI] zfcp: Fix abort handler for completions in progress

When the abort handler cannot find a pending FSF request, the request
completion could just be running. This means we cannot return SUCCESS,
since this would lead to call to scsi_done after exiting the SCSI
error handler which is not allowed.

Reviewed-by: Martin Petermann <martin@linux.vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] zfcp: no port recovery after ADISC request timeout
Swen Schillig [Fri, 17 Apr 2009 13:08:10 +0000 (15:08 +0200)]
[SCSI] zfcp: no port recovery after ADISC request timeout

A remote port remains in error state even if we receive a RSCN
stating that the connection is re-established. The port recovery
is not started due to a flag which is not reset.
The solution is to clear the flag in question before we trigger a 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>
15 years ago[SCSI] zfcp: Let actcli handle control file errors
Christof Schmitt [Fri, 17 Apr 2009 13:08:09 +0000 (15:08 +0200)]
[SCSI] zfcp: Let actcli handle control file errors

Error codes specific to the control file requests are evaluated by the
actcli tool, so don't report -ENXIO for those. Generic problems are
still checked for outside the command specific handler.

Reviewed-by: Martin Petermann <martin@linux.vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] zfcp: remove unit will fail if add unit is not finished
Martin Petermann [Fri, 17 Apr 2009 13:08:08 +0000 (15:08 +0200)]
[SCSI] zfcp: remove unit will fail if add unit is not finished

On some hardware it can take some time to add a unit. If
some remove this unit during this process the remove will
fail.

Signed-off-by: Martin Petermann <martin@linux.vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] zfcp: no port recovery after storage side error inject
Swen Schillig [Fri, 17 Apr 2009 13:08:07 +0000 (15:08 +0200)]
[SCSI] zfcp: no port recovery after storage side error inject

The remote port remains in error state even if the connection
is re-established. A wrong precondition check was performed on
the port status leading to a cancellation of the port reopen.
Remove the pre-req check because it's not required and better
handled within 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>
15 years ago[SCSI] zfcp: avoid false ERP complete due to sema race
Swen Schillig [Fri, 17 Apr 2009 13:08:06 +0000 (15:08 +0200)]
[SCSI] zfcp: avoid false ERP complete due to sema race

The ERP thread is performing a task before it is executing the
corresponding down on the semaphore. The response handler of the
just started exchange config should wait for the completion by
performing a down on this semaphore. Since this semaphore is still
positive from the ERP enqueue the handler won't wait and therefore
the exchange config will always fail leaving the adapter in error.
The problem can be solved by performing the down on the semaphore
before starting an ERP task. This is the logically correct order.
Only walk the ERP loop if there is a task to perform.

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>
15 years ago[SCSI] zfcp: Set WKA-port to offline on adapter deactivation
Swen Schillig [Fri, 17 Apr 2009 13:08:05 +0000 (15:08 +0200)]
[SCSI] zfcp: Set WKA-port to offline on adapter deactivation

The nameserver port might be in state online when the adapter is
offlined. On adapter reactivation the nameserver port is not
re-opened due to the PORT_ONLINE status. This results in an
unsuccessful recovery. In forcing the nameserver port status
to offline on all adapter offline events this issue is prevented.

Waiting for the reference count to drop to zero in
zfcp_wka_port_offline is not required, so remove it.

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>
15 years ago[SCSI] zfcp: Dont block zfcp_wq with scan
Swen Schillig [Fri, 17 Apr 2009 13:08:04 +0000 (15:08 +0200)]
[SCSI] zfcp: Dont block zfcp_wq with scan

When running the scsi_scan from the zfcp workqueue and the target
device does not respond, the zfcp workqueue can block until the
scsi_scan hits a timeout. Move the work to the scsi host workqueue,
since this one is also used for the scan from the SCSI midlayer.

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>
15 years ago[SCSI] zfcp: Dont call zfcp_fsf_req_free on NULL pointer
Christof Schmitt [Fri, 17 Apr 2009 13:08:03 +0000 (15:08 +0200)]
[SCSI] zfcp: Dont call zfcp_fsf_req_free on NULL pointer

Fix problem that zfcp_fsf_exchange_config_data_sync and
zfcp_fsf_exchange_config_data_sync could try to call zfcp_fsf_req_free
with a NULL pointer.

Reviewed-by: Martin Petermann <martin@linux.vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] zfcp: Enable auto-port discovery for NPIV.
Swen Schillig [Fri, 17 Apr 2009 13:08:02 +0000 (15:08 +0200)]
[SCSI] zfcp: Enable auto-port discovery for NPIV.

Since we're setting the host port type now to FC_PORTTYPE_NPIV
for adapters running in NPIV mode we should allow this port type
for auto-port scanning as well.

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>
15 years ago[SCSI] zfcp: Avoid referencing freed memory in req send
Martin Petermann [Fri, 17 Apr 2009 13:08:01 +0000 (15:08 +0200)]
[SCSI] zfcp: Avoid referencing freed memory in req send

Avoid referencing a fsf request after sending it in fcp_fsf_req_send,
it might have already completed and deallocated.

Signed-off-by: Martin Petermann <martin@linux.vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] fix q->lock not held warning when target is busy
Mike Christie [Thu, 9 Apr 2009 20:57:10 +0000 (15:57 -0500)]
[SCSI] fix q->lock not held warning when target is busy

We cannot call blk_plug_device from scsi_target_queue_ready
because the q lock is not held. And we do not need to call
it from there because when we return 0, the scsi_request_fn
not_ready handling will plug the queue for us if needed.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] lpfc 8.3.1 : Update version to 8.3.1
James Smart [Mon, 6 Apr 2009 22:48:17 +0000 (18:48 -0400)]
[SCSI] lpfc 8.3.1 : Update version to 8.3.1

Update driver version to 8.3.1
Also update copyright end year for driver.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] lpfc 8.3.1: misc fixes/changes
James Smart [Mon, 6 Apr 2009 22:48:10 +0000 (18:48 -0400)]
[SCSI] lpfc 8.3.1: misc fixes/changes

8.3.1 Fixes/Changes :

- Fix incorrect byte-swapping on word 4 of IOCB (data length) which
  caused LUNs to not be discovered on big-endian (e.g. PPC)

- Remove a bad cast of MBslimaddr which loses the __iomem (sparse)

- Make lpfc_debugfs_mask_disc_trc static (sparse)

- Correct misspelled word BlockGuard in lpfc_logmsg.h comment

- Replaced repeated code segment for canceling IOCBs from a list with
  a function call, lpfc_sli_cancel_iocbs().

- Increased HBQ buffers to support 40KB SSC sequences.

- Added sysfs interface to update speed and topology parameter without
  link bounce.

- Fixed bug with sysfs fc_host WWNs not being updated after changing
  the WWNs.

- Check if the active mailbox is NULL in the beginning of the mailbox
  timeout handler - fixes panic in the mailbox timeout handler while
  running IO stress test

- Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect
  reference to phba through vport

- Removed de-reference of scsi device after call to scsi_done() to fix
  panic in scsi completion path while accessing scsi device after
  scsi_done is called.

- Fixed "Nodelist not empty" message when unloading the driver after
  target reboot test

- Added LP2105 HBA model description

- Added code to print all 16 words of unrecognized ASYNC events

- Fixed memory leak in vport create + delete loop

- Added support for handling dual error bit from HBA

- Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb

- Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node

- Take NULL termintator into account when calculating available buffer space

Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] lpfc 8.3.1 : Fix up kernel-doc function comments
James Smart [Mon, 6 Apr 2009 22:47:14 +0000 (18:47 -0400)]
[SCSI] lpfc 8.3.1 : Fix up kernel-doc function comments

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] sg: return EFAULT for an invalid user address
FUJITA Tomonori [Fri, 3 Apr 2009 15:35:42 +0000 (00:35 +0900)]
[SCSI] sg: return EFAULT for an invalid user address

blk_rq_unmap_user() returns EFAULT if a program passes an invalid
address to kernel (the kernel fails to copy data to user space). sg
needs to pass the returned value to user space instead of ignoring
it. Before the block layer conversion, sg returns EFAULT
properly. This restores the old behavior.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Douglas Gilbert <dgilbert@interlog.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years ago[SCSI] sr: report more accurate drive status after closing the tray.
Peter Jones [Fri, 11 Jul 2008 20:17:05 +0000 (16:17 -0400)]
[SCSI] sr: report more accurate drive status after closing the tray.

So, what's happening here is that the drive is reporting a sense of
2/4/1 ("logical unit is becoming ready") from sr_test_unit_ready(), and
then we ask for the media event notification before checking that result
at all.  The check_media_event_descriptor() call isn't getting a check
condition, but it's also reporting that the tray is closed and that
there's no media.  In actuality it doesn't yet know if there's media or
not, but there's no way to express that in the media event status field.

My current thought is that if it told us the device isn't yet ready, we
should return that immediately, since there's nothing that'll tell us
any more data than that reliably:

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
15 years agosparc: convert to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:11:08 +0000 (22:11 -0400)]
sparc: convert to use __HEAD and HEAD_TEXT macros.

This has the consequence of changing the section name use for head
code from ".text.head" to ".head.text".  Since this commit changes all
users in the architecture, this change should be harmless.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: David S. Miller <davem@davemloft.net>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agosh: convert to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:11:07 +0000 (22:11 -0400)]
sh: convert to use __HEAD and HEAD_TEXT macros.

This has the consequence of changing the section name use for head
code from ".text.head" to ".head.text".  Since this commit changes all
users in the architecture, this change should be harmless.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: Paul Mundt <lethal@linux-sh.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agos390: convert to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:11:06 +0000 (22:11 -0400)]
s390: convert to use __HEAD and HEAD_TEXT macros.

This has the consequence of changing the section name use for head
code from ".text.head" to ".head.text".  Since this commit changes all
users in the architecture, this change should be harmless.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agopowerpc: convert to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:11:05 +0000 (22:11 -0400)]
powerpc: convert to use __HEAD and HEAD_TEXT macros.

This has the consequence of changing the section name use for head
code from ".text.head" to ".head.text".  Since this commit changes all
users in the architecture, this change should be harmless.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomn10300: convert to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:11:04 +0000 (22:11 -0400)]
mn10300: convert to use __HEAD and HEAD_TEXT macros.

This has the consequence of changing the section name use for head
code from ".text.head" to ".head.text".  Since this commit changes all
users in the architecture, this change should be harmless.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: David Howells <dhowells@redhat.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agom68k: convert to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:11:03 +0000 (22:11 -0400)]
m68k: convert to use __HEAD and HEAD_TEXT macros.

This has the consequence of changing the section name use for head
code from ".text.head" to ".head.text".  Since this commit changes all
users in the architecture, this change should be harmless.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agom32r: convert to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:11:02 +0000 (22:11 -0400)]
m32r: convert to use __HEAD and HEAD_TEXT macros.

This has the consequence of changing the section name use for head
code from ".text.head" to ".head.text".  Since this commit changes all
users in the architecture, this change should be harmless.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofrv: convert frv to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:10:59 +0000 (22:10 -0400)]
frv: convert frv to use __HEAD and HEAD_TEXT macros.

This has the consequence of changing the section name use for head
code from ".text.head" to ".head.text".  Since this commit changes all
users in the architecture, this change should be harmless.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: David Howells <dhowells@redhat.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoalpha: convert to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:10:58 +0000 (22:10 -0400)]
alpha: convert to use __HEAD and HEAD_TEXT macros.

This has the consequence of changing the section name use for head
code from ".text.head" to ".head.text".  Since this commit changes all
users in the architecture, this change should be harmless.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: Richard Henderson <rth@twiddle.net>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoxtensa: convert to use __HEAD and HEAD_TEXT macros.
Tim Abbott [Sun, 26 Apr 2009 02:10:57 +0000 (22:10 -0400)]
xtensa: convert to use __HEAD and HEAD_TEXT macros.

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Cc: Chris Zankel <chris@zankel.net>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoAdd new HEAD_TEXT_SECTION macro.
Tim Abbott [Sun, 26 Apr 2009 02:10:56 +0000 (22:10 -0400)]
Add new HEAD_TEXT_SECTION macro.

This patch is preparation for replacing all uses of ".head.text" or
".text.head" in the kernel with macros, so that the section name can
later be changed without having to touch a lot of the kernel.

Since some linker scripts do more complex things than referencing
HEAD_TEXT, we add a HEAD_TEXT_SECTION macro that just contains the
actual name.

I've defined HEAD_TEXT_SECTION in a new header,
include/linux/section-names.h, so that this section name only needs to
appear in one place.  I anticipate creating similar macro structures
for a number of other section names.

The long-term goal here is to be able to change the kernel's magic
section names to those that are compatible with -ffunction-sections
-fdata-sections.  This requires renaming all magic sections with names
of the form ".text.foo".

Signed-off-by: Tim Abbott <tabbott@mit.edu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoPM/Hibernate: Fix waiting for image device to appear on resume
Rafael J. Wysocki [Fri, 24 Apr 2009 22:16:06 +0000 (00:16 +0200)]
PM/Hibernate: Fix waiting for image device to appear on resume

Commit c751085943362143f84346d274e0011419c84202 ("PM/Hibernate: Wait for
SCSI devices scan to complete during resume") added a call to
scsi_complete_async_scans() to software_resume(), so that it waited for
the SCSI scanning to complete, but the call was added at a wrong place.

Namely, it should have been added after wait_for_device_probe(), which
is called only if the image partition hasn't been specified yet.  Also,
it's reasonable to check if the image partition is present and only wait
for the device probing and SCSI scanning to complete if it is not the
case.

Additionally, since noresume is checked right at the beginning of
software_resume() and the function returns immediately if it's set, it
doesn't make sense to check it once again later.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoROMFS: Advance destination buffer pointer when reading from a blockdev
David Howells [Thu, 23 Apr 2009 15:41:18 +0000 (16:41 +0100)]
ROMFS: Advance destination buffer pointer when reading from a blockdev

RomFS should advance the destination buffer pointer when reading data from a
blockdev source (the data may be split over multiple blocks, each requiring its
own sb_read() call).  Without this, all the data is copied to the beginning of
the output buffer.

Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Michal Simek <monstr@monstr.eu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoROMFS: romfs_lookup() shouldn't be doing a partial name comparison
David Howells [Thu, 23 Apr 2009 15:41:13 +0000 (16:41 +0100)]
ROMFS: romfs_lookup() shouldn't be doing a partial name comparison

romfs_lookup() should be using a routine akin to strcmp() on the backing store,
rather than one akin to strncmp().  If it uses the latter, it's liable to match
/bin/shutdown when looking up /bin/sh.

Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Michal Simek <monstr@monstr.eu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agolib: find_last_bit.o needed by a module only, move it from lib to obj
Fred Isaman [Thu, 23 Apr 2009 13:40:32 +0000 (16:40 +0300)]
lib: find_last_bit.o needed by a module only, move it from lib to obj

Currently, although find_last_bit is EXPORTed, it is statically linked
with the kernel and is referenced only under CONFIG_SMP.

When CONFIG_SMP is undefined and find_last_bit is referenced only by
modules, linking fails with:

  ERROR: "find_last_bit" [fs/nfs/nfs.ko] undefined!

Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Fred Isaman <iisaman@citi.umich.edu>
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoCacheFiles: Fix the documentation to use the correct credential pointer names
Marc Dionne [Thu, 23 Apr 2009 10:21:55 +0000 (11:21 +0100)]
CacheFiles: Fix the documentation to use the correct credential pointer names

Adjust the CacheFiles documentation to use the correct names of the credential
pointers in task_struct.

The documentation was using names from the old versions of the credentials
patches.

Signed-off-by: Marc Dionne <marc.c.dionne@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovirtio-rng: Remove false BUG for spurious callbacks
Christian Borntraeger [Thu, 23 Apr 2009 07:12:59 +0000 (16:42 +0930)]
virtio-rng: Remove false BUG for spurious callbacks

The virtio-rng drivers checks for spurious callbacks. Since
callbacks can be implemented via shared interrupts (e.g. PCI) this
could lead to guest kernel oopses with lots of virtio devices.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'fixes-for-linus' of git://git.monstr.eu/linux-2.6-microblaze
Linus Torvalds [Fri, 24 Apr 2009 15:54:30 +0000 (08:54 -0700)]
Merge branch 'fixes-for-linus' of git://git.monstr.eu/linux-2.6-microblaze

* 'fixes-for-linus' of git://git.monstr.eu/linux-2.6-microblaze:
  microblaze: add parameter to microblaze_read()
  microblaze: Use CFLAGS_KERNEL instead of CFLAGS
  microblaze: Add STATE_SAVE_ARG_SPACE for noMMU kernel too
  microblaze: Do not check use_dcache
  microblaze: Do not use PVR configuration for broken MB version
  microblaze: Fix USR1/2 pvr printing message
  microblaze: iowrite upon timeout
  microblaze: Correspond CONFIG...PCMP in Makefile/Kconfig
  microblaze: Remove redundant variable
  microblaze: Move start_thread to process.c
  microblaze: Add missing preadv and pwritev syscalls
  microblaze: Add missing declaration for die and _exception func
  microblaze: Remove sparse error in traps.c
  microblaze: Move task_pt_regs up
  microblaze: Rename kernel_mode to pt_mode in pt_regs
  microblaze: Remove uncache shadow condition
  microblaze: Remove while(1) loop from show_regs function
  microblaze: Remove unneded per cpu SYSCALL_SAVE variable

15 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux...
Linus Torvalds [Fri, 24 Apr 2009 15:48:17 +0000 (08:48 -0700)]
Merge branch 'release' of git://git./linux/kernel/git/lenb/linux-acpi-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6: (34 commits)
  ACPI, i915: Register ACPI video even when not modesetting
  Revert "ACPICA: delete check for AML access to port 0x81-83"
  I/O port protection: update for windows compatibility.
  sony-laptop: always try to unblock rfkill on load
  sony-laptop: fix bogus error message display on resume
  ACPI: EC: Fix ACPI EC resume non-query interrupt message
  sony-laptop: SNC input event 38 fix
  sony-laptop: SNC 127 Initialization Fix
  sony-laptop: Duplicate SNC 127 Event Fix
  ACPI: prevent processor.max_cstate=0 boot crash
  ACPI/hpet: prevent boot hang when hpet=force used on ICH-4M
  ACPI: delete obsolete "bus master activity" proc field
  ACPI: idle: mark_tsc_unstable() at init-time, not run-time
  ACPI: add /sys/firmware/acpi/interrupts/sci_not counter
  ACPI video: fix an error when the brightness levels on AC and on Battery are same
  acpi-cpufreq: Do not let get_measured perf depend on internal variable
  acpi-cpufreq: style-only: add parens to math expression
  acpi-cpufreq: Cleanup: Use printk_once
  x86, acpi_cpufreq: Fix the NULL pointer dereference in get_measured_perf
  thinkpad-acpi: bump up version to 0.23
  ...

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu
Linus Torvalds [Fri, 24 Apr 2009 15:45:53 +0000 (08:45 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/gerg/m68knommu

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu:
  m68knommu: update the default config for the ColdFire 5407C3 board
  m68knommu: update the default config for the ColdFire 5307C3 board
  m68knommu: update the default config for the ColdFire 5257EVB board
  m68knommu: update the default config for the ColdFire 5249EVB.
  m68knommu: add a defconfig for the ColdFire M5272C3 board
  m68knommu: update the defconfig for the ColdFire 5208evb board
  m68knommu: fix DMA support for ColdFire
  m68knommu: remove unused kernel stats offsets
  m68knommu: fix missing .data.cacheline_aligned section
  m68knommu: Fixed GPIO pin initialization for CONFIG_M5271 FEC.

15 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
Linus Torvalds [Fri, 24 Apr 2009 15:37:40 +0000 (08:37 -0700)]
Merge branch 'for_linus' of git://git./linux/kernel/git/tytso/ext4

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: Fix potential inode allocation soft lockup in Orlov allocator
  ext4: Make the extent validity check more paranoid
  jbd: use SWRITE_SYNC_PLUG when writing synchronous revoke records
  jbd2: use SWRITE_SYNC_PLUG when writing synchronous revoke records
  ext4: really print the find_group_flex fallback warning only once

15 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Fri, 24 Apr 2009 15:36:41 +0000 (08:36 -0700)]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] 5460/1: Orion: reduce namespace pollution
  [ARM] 5458/1: pcmcia: pxa2xx-sharpsl: check if we do have Scoop config
  [ARM] 5457/1: mach-imx gpio buildfix
  [ARM] 5456/1: add sys_preadv and sys_pwritev
  [ARM] pxa/pcm990: start external GPIOs immediately after built-in ones
  [ARM] pxa/palm27x: General fix for Palm27x aSoC driver
  [ARM] pxa/mioa701: use GPIO95 as AC97 reset line
  [ARM] pxa: merge AC97 platform data structures
  [ARM] pxa/magician: remove un-necessary #include of pxa-regs.h and hardware.h

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6
Linus Torvalds [Fri, 24 Apr 2009 15:36:06 +0000 (08:36 -0700)]
Merge git://git./linux/kernel/git/gregkh/usb-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6:
  USB: pwc : do not pass stack allocated buffers to USB core.
  USB: otg: Fix bug on remove path without transceiver
  USB: correct error handling in cdc-wdm
  USB: removal of tty->low_latency hack dating back to the old serial code
  USB: serial: sierra driver bug fix for composite interface
  USB: gadget: omap_udc uses platform_driver_probe()
  USB: ci13xxx_udc: fix build error
  USB: musb: Prevent multiple includes of musb.h
  USB: pass mem_flags to dma_alloc_coherent
  USB: g_file_storage: fix use-after-free bug when closing files
  USB: ehci-sched.c: EHCI SITD scheduling bugfix
  USB: fix mos7840 problem with minor numbers
  USB: mos7840: add new device id
  USB: musb: fix build when !CONFIG_PM
  USB: musb: Remove my email address from few musb related drivers
  USB: Gadget: MIPS CI13xxx UDC bugfixes
  USB: Unusual Device support for Gold MP3 Player Energy
  USB: serial: fix lifetime and locking problems

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ecryptfs...
Linus Torvalds [Fri, 24 Apr 2009 15:32:44 +0000 (08:32 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/ecryptfs/ecryptfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ecryptfs/ecryptfs-2.6:
  eCryptfs: Larger buffer for encrypted symlink targets
  eCryptfs: Lock lower directory inode mutex during lookup
  eCryptfs: Remove ecryptfs_unlink_sigs warnings
  eCryptfs: Fix data corruption when using ecryptfs_passthrough
  eCryptfs: Print FNEK sig properly in /proc/mounts
  eCryptfs: NULL pointer dereference in ecryptfs_send_miscdev()
  eCryptfs: Copy lower inode attrs before dentry instantiation

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux...
Linus Torvalds [Fri, 24 Apr 2009 15:32:23 +0000 (08:32 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/geert/linux-m68k

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k:
  m68k: Update defconfigs for 2.6.30-rc3
  m68k,m68knommu: Wire up preadv and pwritev
  scsi: a4000 - Correct driver unregistration in case of failure

15 years agoMerge branch 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6
Linus Torvalds [Fri, 24 Apr 2009 15:28:27 +0000 (08:28 -0700)]
Merge branch 'for-linus' of git://git390.marist.edu/linux-2.6

* 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6:
  [S390] update default configuration.
  [S390] omit frame pointers on s390 when possible
  [S390] Use tape_generic_offline directly.
  [S390] /proc/stat idle field for idle cpus
  [S390] appldata: avoid deadlock with appldata_mem
  [S390] ipl: fix compile breakage

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes
Linus Torvalds [Fri, 24 Apr 2009 15:27:02 +0000 (08:27 -0700)]
Merge git://git./linux/kernel/git/steve/gfs2-2.6-fixes

* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes:
  GFS2: Ensure that the inode goal block settings are updated
  GFS2: Fix bug in block allocation
  bitops: Add __ffs64 bitop

15 years agoMerge branch 'kvm-updates/2.6.30' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 24 Apr 2009 15:26:01 +0000 (08:26 -0700)]
Merge branch 'kvm-updates/2.6.30' of git://git./virt/kvm/kvm

* 'kvm-updates/2.6.30' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: Unregister cpufreq notifier on unload
  KVM: x86: release time_page on vcpu destruction
  KVM: Fix overlapping check for memory slots
  KVM: MMU: disable global page optimization
  KVM: ia64: fix locking order entering guest
  KVM: MMU: Fix off-by-one calculating large page count

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6
Linus Torvalds [Fri, 24 Apr 2009 15:16:05 +0000 (08:16 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/bart/ide-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6:
  MAINTAINERS: update IDE entry
  palm_bk3710: palm_bk3710_udmatimings[] CodingStyle fixup
  palm_bk3710: those registers/bitfields don't exist
  mediabay: fix build for CONFIG_BLOCK=n
  ide: Stop disks on reboot for laptop which cuts power
  ide-cd: fix kernel crash on hppa regression
  palm_bk3710: UDMA performance fix

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Fri, 24 Apr 2009 15:15:25 +0000 (08:15 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  ALSA: hda - Add quirk for Packard Bell RS65
  [ALSA] intel8x0: another attempt to fix ac97_clock measure routine
  [ALSA] ac97_codec: increase timeout for analog subsections
  ALSA: hda - Add quirks for Realtek codecs
  ALSA: hda - Fix alc662_init_verbs
  ALSA: keywest: Convert to new-style i2c driver
  ALSA: AOA: Convert onyx and tas codecs to new-style i2c drivers
  ALSA: Atiixp: Add SSID for mute_led quirk (unknown HP model)
  ALSA: us122l: add snd_us122l_free()
  ASoC: Fix warning in wm9705
  ASoC: OMAP: Update contact addresses
  ASoC: pxa-ssp: Don't use SSCR0_SerClkDiv and SSCR0_SCR
  ALSA: us122l: Fix signedness in comparisions

15 years agoradio_si470x: Fix free memory corruption
Alan Cox [Wed, 22 Apr 2009 14:03:15 +0000 (15:03 +0100)]
radio_si470x: Fix free memory corruption

The release path for a disconnected device frees the object then unlocks
the mutex in the freed object...

Found by Dan Carpenter using Smatch

Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoserial: remove contact data
Niels de Vos [Wed, 22 Apr 2009 14:02:44 +0000 (15:02 +0100)]
serial: remove contact data

Remove my name and emailaddress from note in the source. Wincor Nixdorf
only has some ITE-chips on their mainboards, other chips are not
available for me for testing.

Signed-off-by: Niels de Vos <niels.devos@wincor-nixdorf.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocafe_nand: Fix warning
Alan Cox [Wed, 22 Apr 2009 14:02:23 +0000 (15:02 +0100)]
cafe_nand: Fix warning

Wrong types on IRQ handler

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Fri, 24 Apr 2009 14:48:24 +0000 (07:48 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  cfq-iosched: cache prio_tree root in cfqq->p_root
  cfq-iosched: fix bug with aliased request and cooperation detection
  cfq-iosched: clear ->prio_trees[] on cfqd alloc
  block: fix intermittent dm timeout based oops
  umem: fix request_queue lock warning
  block: simplify I/O stat accounting
  pktcdvd.h should include mempool.h
  cfq-iosched: use the default seek distance when there aren't enough seek samples
  cfq-iosched: make seek_mean converge more quickly
  block: make blk_abort_queue() ignore non-request based devices
  block: include empty disks in /proc/diskstats
  bio: use bio_kmalloc() in copy/map functions
  bio: fix bio_kmalloc()
  block: fix queue bounce limit setting
  block: fix SG_IO vector request data length handling
  scatterlist: make sure sg_miter_next() doesn't return 0 sized mappings

15 years agoMN10300: Update the ASB2303 defconfig
David Howells [Wed, 22 Apr 2009 11:10:07 +0000 (12:10 +0100)]
MN10300: Update the ASB2303 defconfig

Update the defconfig for the ASB2303 evaluation board.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoDelete slow-work timers properly
Jonathan Corbet [Wed, 22 Apr 2009 11:01:49 +0000 (12:01 +0100)]
Delete slow-work timers properly

Slow-work appears to delete its timer as soon as the first user
unregisters, even though other users could be active.  At the same time, it
never seems to delete slow_work_oom_timer.  Arrange for both to happen in
the shutdown path.

Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Fri, 24 Apr 2009 14:46:51 +0000 (07:46 -0700)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (94 commits)
  netfilter: ctnetlink: fix gcc warning during compilation
  net/netrom: Fix socket locking
  netlabel: Always remove the correct address selector
  ucc_geth.c: Fix upsmr setting in RMII mode
  8139too: fix HW initial flow
  af_iucv: Fix race when queuing incoming iucv messages
  af_iucv: Test additional sk states in iucv_sock_shutdown
  af_iucv: Reject incoming msgs if RECV_SHUTDOWN is set
  af_iucv: fix oops in iucv_sock_recvmsg() for MSG_PEEK flag
  af_iucv: consider state IUCV_CLOSING when closing a socket
  iwlwifi: DMA fixes
  iwlwifi: add debugging for TX path
  mwl8: fix build warning.
  mac80211: fix alignment calculation bug
  mac80211: do not print WARN if config interface
  iwl3945: use cancel_delayed_work_sync to cancel rfkill_poll
  iwlwifi: fix EEPROM validation mask to include OTP only devices
  atmel: fix netdev ops conversion
  pcnet_cs: add cis(firmware) of the Allied Telesis LA-PCM
  mlx4_en: Fix cleanup if workqueue create in mlx4_en_add() fails
  ...

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6
Linus Torvalds [Fri, 24 Apr 2009 14:45:52 +0000 (07:45 -0700)]
Merge git://git./linux/kernel/git/davem/sparc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
  sparc: Fix bus type probing for ESP and LE devices.
  sparc32: Update defconfig.
  sparc64: Update defconfig.

15 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
Linus Torvalds [Fri, 24 Apr 2009 14:44:58 +0000 (07:44 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/paulus/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc:
  powerpc: Fix modular build of ide-pmac when mediabay is built in
  powerpc/pasemi: Fix build error on UP
  powerpc: Make macintosh/mediabay driver depend on CONFIG_BLOCK
  maintainers: Fix PS3 patterns
  powerpc/ps3: Fix CONFIG_PS3_FLASH=n build warning
  powerpc/32: Don't clobber personality flags on exec
  powerpc: Fix crash on CPU hotplug
  powerpc/85xx: Remove defconfigs that mpc85xx_{smp_}defconfig cover
  powerpc/85xx: Added SMP defconfig
  powerpc/85xx: Enabled a bunch of FSL specific drivers/options
  powerpc/85xx: Updated generic mpc85xx_defconfig
  powerpc: don't disable SATA interrupts on Freescale MPC8610 HPCD
  fsl_rio: Pass the proper device to dma mapping routines
  powerpc: Fix of_node_put() exit path in of_irq_map_one()
  powerpc/5200: defconfig updates
  powerpc/5200: Add FLASH nodes to lite5200 device tree
  powerpc/device-tree: Document MTD nodes with multiple "reg" tuples
  powerpc/of-device-tree: Factor MTD physmap bindings out of booting-without-of
  powerpc/5200: Bring the legacy fsl_spi_platform_data hooks back

15 years agoMerge branch 'bjorn.button' into release
Len Brown [Fri, 24 Apr 2009 14:42:29 +0000 (10:42 -0400)]
Merge branch 'bjorn.button' into release

15 years agoMerge branch 'bugzilla-13036' into release
Len Brown [Fri, 24 Apr 2009 14:42:21 +0000 (10:42 -0400)]
Merge branch 'bugzilla-13036' into release

15 years agoMerge branch 'bugzilla-13048' into release
Len Brown [Fri, 24 Apr 2009 14:42:11 +0000 (10:42 -0400)]
Merge branch 'bugzilla-13048' into release

15 years agoMerge branch 'bugzilla-13142' into release
Len Brown [Fri, 24 Apr 2009 14:42:03 +0000 (10:42 -0400)]
Merge branch 'bugzilla-13142' into release

15 years agoMerge branch 'hpet' into release
Len Brown [Fri, 24 Apr 2009 14:41:31 +0000 (10:41 -0400)]
Merge branch 'hpet' into release

15 years agocheck_unsafe_exec: s/lock_task_sighand/rcu_read_lock/
Oleg Nesterov [Thu, 23 Apr 2009 23:02:45 +0000 (01:02 +0200)]
check_unsafe_exec: s/lock_task_sighand/rcu_read_lock/

write_lock(&current->fs->lock) guarantees we can't wrongly miss
LSM_UNSAFE_SHARE, this is what we care about. Use rcu_read_lock()
instead of ->siglock to iterate over the sub-threads. We must see
all CLONE_THREAD|CLONE_FS threads which didn't pass exit_fs(), it
takes fs->lock too.

With or without this patch we can miss the freshly cloned thread
and set LSM_UNSAFE_SHARE, we don't care.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roland McGrath <roland@redhat.com>
[ Fixed lock/unlock typo  - Hugh ]
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodo_execve() must not clear fs->in_exec if it was set by another thread
Oleg Nesterov [Thu, 23 Apr 2009 23:01:56 +0000 (01:01 +0200)]
do_execve() must not clear fs->in_exec if it was set by another thread

If do_execve() fails after check_unsafe_exec(), it clears fs->in_exec
unconditionally. This is wrong if we race with our sub-thread which
also does do_execve:

Two threads T1 and T2 and another process P, all share the same
->fs.

T1 starts do_execve(BAD_FILE). It calls check_unsafe_exec(), since
->fs is shared, we set LSM_UNSAFE but not ->in_exec.

P exits and decrements fs->users.

T2 starts do_execve(), calls check_unsafe_exec(), now ->fs is not
shared, we set fs->in_exec.

T1 continues, open_exec(BAD_FILE) fails, we clear ->in_exec and
return to the user-space.

T1 does clone(CLONE_FS /* without CLONE_THREAD */).

T2 continues without LSM_UNSAFE_SHARE while ->fs is shared with
another process.

Change check_unsafe_exec() to return res = 1 if we set ->in_exec, and change
do_execve() to clear ->in_exec depending on res.

When do_execve() suceeds, it is safe to clear ->in_exec unconditionally.
It can be set only if we don't share ->fs with another process, and since
we already killed all sub-threads either ->in_exec == 0 or we are the
only user of this ->fs.

Also, we do not need fs->lock to clear fs->in_exec.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roland McGrath <roland@redhat.com>
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocfq-iosched: cache prio_tree root in cfqq->p_root
Jens Axboe [Thu, 23 Apr 2009 10:19:38 +0000 (12:19 +0200)]
cfq-iosched: cache prio_tree root in cfqq->p_root

Currently we look it up from ->ioprio, but ->ioprio can change if
either the process gets its IO priority changed explicitly, or if
cfq decides to temporarily boost it. So if we are unlucky, we can
end up attempting to remove a node from a different rbtree root than
where it was added.

Fix this by using ->org_ioprio as the prio_tree index, since that
will only change for explicit IO priority settings (not for a boost).
Additionally cache the rbtree root inside the cfqq, then we don't have
to add code to reinsert the cfqq in the prio_tree if IO priority changes.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
15 years agocfq-iosched: fix bug with aliased request and cooperation detection
Jens Axboe [Thu, 23 Apr 2009 10:14:56 +0000 (12:14 +0200)]
cfq-iosched: fix bug with aliased request and cooperation detection

cfq_prio_tree_lookup() should return the direct match, yet it always
returns zero. Fix that.

cfq_prio_tree_add() assumes that we don't get a direct match, while
it is very possible that we do. Using O_DIRECT, you can have different
cfqq with matching requests, since you don't have the page cache
to serialize things for you. Fix this bug by only adding the cfqq if
there isn't an existing match.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>