openwrt/staging/blogic.git
11 years ago[SCSI] sd: Reduce buffer size for vpd request
Bernd Schubert [Mon, 23 Sep 2013 12:47:32 +0000 (14:47 +0200)]
[SCSI] sd: Reduce buffer size for vpd request

Somehow older areca firmware versions have issues with
scsi_get_vpd_page() and a large buffer, the firmware
seems to crash and the scsi error-handler will start endless
recovery retries.
Limiting the buf-size to 64-bytes fixes this issue with older
firmware versions (<1.49 for my controller).

Fixes a regression with areca controllers and older firmware versions
introduced by commit: 66c28f97120e8a621afd5aa7a31c4b85c547d33d

Reported-by: Nix <nix@esperi.org.uk>
Tested-by: Nix <nix@esperi.org.uk>
Signed-off-by: Bernd Schubert <bernd.schubert@itwm.fraunhofer.de>
Cc: stable@vger.kernel.org # delay inclusion for 2 months for testing
Acked-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] scsi_debug: fix sparse warnings related to data integrity field
Akinobu Mita [Wed, 18 Sep 2013 12:27:28 +0000 (21:27 +0900)]
[SCSI] scsi_debug: fix sparse warnings related to data integrity field

Each member in data integrity field tuple is big-endian.  But the
endianness of the values being compared with these members are not
annotated.  So this fixes these sparse warnings.

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] scsi_debug: fix invalid value check for guard module parameter
Akinobu Mita [Wed, 18 Sep 2013 12:27:27 +0000 (21:27 +0900)]
[SCSI] scsi_debug: fix invalid value check for guard module parameter

In the module initialization, invalid value for guard module parameter
is detected by the following check:

        if (scsi_debug_guard > 1) {
                printk(KERN_ERR "scsi_debug_init: guard must be 0 or 1\n");
                return -EINVAL;
        }

But this check isn't enough, because the type of scsi_debug_guard is
'int' and scsi_debug_guard could be a negative value.

This fixes it by changing the type of scsi_debug_guard to 'unsigned int'
instead of adding extra check for a negative value.

Reported-by: Joe Perches <joe@perches.com>
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] scsi_debug: avoid partial copying PI from prot_sglist to dif_storep
Akinobu Mita [Wed, 18 Sep 2013 12:27:26 +0000 (21:27 +0900)]
[SCSI] scsi_debug: avoid partial copying PI from prot_sglist to dif_storep

If data integrity support is enabled, prot_verify_write() is called in
response to WRITE commands and it verifies protection info from
prot_sglist by comparing against data sglist, and copies protection info
to dif_storep.

When multiple blocks are transfered by a WRITE command, it verifies and
copies these blocks one by one.  So if it fails to verify protection
info in the middle of blocks, the actual data transfer to fake_storep
isn't proceeded at all although protection info for some blocks are
already copied to dif_storep.  Therefore, it breaks the data integrity
between fake_storep and dif_storep.

This fixes it by ensuring that copying protection info to dif_storep is
done after all blocks are successfully verified.  Reusing dif_copy_prot()
with supporting the opposite direction simplifies this fix.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] scsi_debug: factor out copying PI from dif_storep to prot_sglist
Akinobu Mita [Wed, 18 Sep 2013 12:27:25 +0000 (21:27 +0900)]
[SCSI] scsi_debug: factor out copying PI from dif_storep to prot_sglist

If data integrity support is enabled, prot_verify_read() is called in
response to READ commands and it verifies protection info from dif_storep
by comparing against fake_storep, and copies protection info to
prot_sglist.

This factors out the portion of copying protection info into a separate
function.  It will also be reused in the next change after supporting
the opposite direction (copying prot_sglist to dif_storep).

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] scsi_debug: fix buffer overrun when DIF/DIX is enabled and virtual_gb > 0
Akinobu Mita [Wed, 18 Sep 2013 12:27:24 +0000 (21:27 +0900)]
[SCSI] scsi_debug: fix buffer overrun when DIF/DIX is enabled and virtual_gb > 0

If the module parameter virtual_gb is greater than 0, the READ command
may request the blocks which exceed actual ramdisk storage (fake_storep).
prot_verify_read() should treat those blocks as wrap around the end of
fake_storep.  But it actually causes fake_storep and dif_storep buffer
overruns.

This fixes these buffer overruns.  In order to simplify the fix,
this also introduces fake_store() and dif_store() which return
corresponding wrap around addresses.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Bump driver version
Jayamohan Kallickal [Sat, 28 Sep 2013 22:36:00 +0000 (15:36 -0700)]
[SCSI] be2iscsi: Bump driver version

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix SGL posting for unaligned ICD values
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:59 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix SGL posting for unaligned ICD values

If certain configuration it is possible that ICD range is not page-aligned.
SGL posting failed in these configuration and driver load was not success.
This fix aligns ICD range values and SGL posting for IO is done.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix AER handling in driver
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:58 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix AER handling in driver

Signed-off-by: Minh Tran <minhduc.tran@emulex.com>
Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Invalidate WRB in Abort/Reset Path
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:57 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Invalidate WRB in Abort/Reset Path

When iSCSI stack invokes Abort or Reset handlers, the aborted tasks
Invalid Bit in WRB needs to be set. Else FW will not be aware of
the command invalidated which leads to BAD_WRB error posted by FW.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix Insufficient Buffer Error returned in MBX Completion
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:56 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix Insufficient Buffer Error returned in MBX Completion

When MBX_Cmd completion happens with error code Insufficient Buffer,
the MBX_Cmd is posted again with the new buffer size posted by FW.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix log level for protocol specific logs
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:55 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix log level for protocol specific logs

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix MSIx creation for SKH-R adapter
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:54 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix MSIx creation for SKH-R adapter

The MSIx to be created for SKH-R adapter should be based on
eq_count returned by get_fw_config.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Display Port Identifier for each iSCSI function
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:53 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Display Port Identifier for each iSCSI function

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Dispaly CID available for connection offload
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:52 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Dispaly CID available for connection offload

Display CID available on each iSCSI Fn which can be used to
offload a connection. The display is split across available CID
on each chute.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix chute cleanup during drivers unload.
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:51 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix chute cleanup during drivers unload.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix connection offload to support Dual Chute.
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:50 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix connection offload to support Dual Chute.

The connection is offload to each chute in a round-robin manner
if both the chute is loaded with iSCSI protocol

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix CID allocation/freeing to support Dual chute mode
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:49 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix CID allocation/freeing to support Dual chute mode

Configuration parameters returns the number of connection that
can be offloaded one each chute.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix WRB_Q posting to support Dual Chute mode
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:48 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix WRB_Q posting to support Dual Chute mode

Configuration parameters return number of CID each chute supports. The WRB_Q
is created for the passed CID count. If both the Chute has iSCSI Protocol then
WRB_Q creation is in a round robin mechanism.

For BE-X family iSCSI protocol is loaded only on single chute.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix SGL Initilization and posting Pages for Dual Chute
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:47 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix SGL Initilization and posting Pages for Dual Chute

Initialization of SGL and related PAGE posting is to be done for the chute.
Based on configuration value of each Chute,SGL initialization and page posting
is done.

For BE-X family iSCSI protocol is loaded only on single chute

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix Template HDR support for Dual Chute mode
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:46 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix Template HDR support for Dual Chute mode

Template HDR is created for each chute which has iSCSI Protocol loaded.
For BE-X family iSCSI protocol is loaded only on single chute.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix changes in ASYNC Path for SKH-R adapter
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:45 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix changes in ASYNC Path for SKH-R adapter

DEF_Q[HDR/DATA] is created on the chute on which iSCSI Protocol is loaded.
When a connection is offloaded, the DEF_Q HDR/Data ID needs to be passed.
FW posts ASYNC message received from target on the passed DEF_Q. Connection
can be offloaded on any of the chute so DEF_Q is created on each Chute.

Change in the ASYNC path initialization based on the configuration parameters
returned for each chute.

For BE-X family iSCSI protocol is loaded only on single chute.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Config parameters update for Dual Chute Support
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:44 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Config parameters update for Dual Chute Support

On the adapter iSCSI protocol can be loaded on either one or both
the CHUTE.Check on which CHUTE iSCSI Protocol is loaded and get
configuration parameters based on which driver initization is done.

For BE-X family iSCSI protocol is loaded only on single chute.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix soft lock up issue during UE or if FW taking time to respond
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:43 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix soft lock up issue during UE or if FW taking time to respond

The timeout set in MBX_CMD is 100sec and the ready bit checking in BMBX
mode is done for 4sec. After 4sec the task is scheduled out for 5 secs
to avoid kernel soft lockup stack trace. The loop of 4sec ready bit check
and then schedule out is done until the following conditon occur
- The Ready Bit is Set
- The timeout set in MBX_CMD expires

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix locking mechanism in Unsol Path
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:42 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix locking mechanism in Unsol Path

The default pdu is a common resource and needs to be protected
while manipulating it.

Signed-off-by: Minh Tran <minhduc.tran@emulex.com>
Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix negotiated parameters upload to FW
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:41 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix negotiated parameters upload to FW

- Removed the check of MaxXmitDSL == 0 as this is not a possible
  case.
- Update connection offload data structure for SKH-R adapters.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix repeated issue of MAC ADDR get IOCTL
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:40 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix repeated issue of MAC ADDR get IOCTL

Storing MAC ADDR of each function in it's priv structure to
avoid issuing MAC_ADDR get IOCTL. Based on a flag set/unset
it's decided if MAC_ADDR is stored in priv structure or IOCTL
needs to be issued.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix the MCCQ count leakage
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:39 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix the MCCQ count leakage

When MBX CMD is posted in MCCQ and if command times out,during
mccq resource cleanup for the timed out command  mccq->count
was not decremented. The led to BUG_ON being hit.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] be2iscsi: Fix Template HDR IOCTL
Jayamohan Kallickal [Sat, 28 Sep 2013 22:35:38 +0000 (15:35 -0700)]
[SCSI] be2iscsi: Fix Template HDR IOCTL

Allocating memory in the Host which will be used by the
TOE functionality during Session Offload. This fix will
allow performance improvement as adapter memory contention
will be reduced.

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] MAINTAINER: Updated maintainer info for bnx2fc
Eddie Wai [Wed, 18 Sep 2013 05:33:13 +0000 (22:33 -0700)]
[SCSI] MAINTAINER: Updated maintainer info for bnx2fc

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] bnx2fc: Bump version from 1.0.14 to 2.4.1
Eddie Wai [Wed, 18 Sep 2013 05:33:12 +0000 (22:33 -0700)]
[SCSI] bnx2fc: Bump version from 1.0.14 to 2.4.1

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] BNX2FC: hung task timeout warning observed when rmmod bnx2x with active FCoE...
Eddie Wai [Thu, 26 Sep 2013 05:01:20 +0000 (22:01 -0700)]
[SCSI] BNX2FC: hung task timeout warning observed when rmmod bnx2x with active FCoE targets

[v2] - removed the interface->enabled flag setting which prevented the
       fcoe ctlr link from being brought back up after a MTU change

A rtnl_lock deadlock was observed from the rmmod thread where it
tries to unregister the fcoe_ctlr device.  This unregistration
triggered a flush of the sysfs queue of the associated ctlr and led to
a call to the set_fcoe_ctlr_enabled routine.  This will eventually propagate
down to call the bnx2fc_disable routine and contented for the rtnl_lock
in the same context.

This patch creates a subset of the bnx2fc_enable/disable routine which
removes the unnecesary rtnl_lock and the bnx2fc_dev_lock acquisition from
the set_fcoe_ctlr_enabled path.

 kernel: INFO: task rmmod:7874 blocked for more than 120 seconds.
 kernel:      Tainted: G        W  ---------------    2.6.32-415.0.1.el6.x86_64 #1
 kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
 kernel: rmmod         D 000000000000000f     0  7874   6518 0x00000080
 kernel: ffff88022158f7d8 0000000000000086 0000000000000000 0000000000000000
 kernel: ffff88023fe72600 ffff88043c74d410 ffff88043c74d400 ffff88043c74d000
 kernel: ffff88021ecbe5f8 ffff88022158ffd8 000000000000fbc8 ffff88021ecbe5f8
 kernel: Call Trace:
 kernel: [<ffffffff81525985>] schedule_timeout+0x215/0x2e0
 kernel: [<ffffffff810680c0>] ? pick_next_task_fair+0xd0/0x130
 kernel: [<ffffffff81524858>] ? schedule+0x178/0x3b2
 kernel: [<ffffffff81525603>] wait_for_common+0x123/0x180
 kernel: [<ffffffff81066b40>] ? default_wake_function+0x0/0x20
 kernel: [<ffffffff811a486e>] ? ifind_fast+0x5e/0xb0
 kernel: [<ffffffff8152571d>] wait_for_completion+0x1d/0x20
 kernel: [<ffffffff81203868>] sysfs_addrm_finish+0x228/0x270
 kernel: [<ffffffff812014ab>] sysfs_hash_and_remove+0x5b/0x90
 kernel: [<ffffffff812056af>] sysfs_remove_group+0x5f/0x100
 kernel: [<ffffffff81367e8b>] device_remove_groups+0x3b/0x60
 kernel: [<ffffffff8136811d>] device_remove_attrs+0x3d/0x90
 kernel: [<ffffffff81368295>] device_del+0x125/0x1e0
 kernel: [<ffffffff81368372>] device_unregister+0x22/0x60
 kernel: [<ffffffffa038ead2>] fcoe_ctlr_device_delete+0xe2/0xf4 [libfcoe]
 kernel: [<ffffffffa03c43cb>] bnx2fc_interface_release+0x5b/0x90 [bnx2fc]
 kernel: [<ffffffffa03c4370>] ? bnx2fc_interface_release+0x0/0x90 [bnx2fc]
 kernel: [<ffffffff812835e7>] kref_put+0x37/0x70
 kernel: [<ffffffffa03c4192>] __bnx2fc_destroy+0x72/0xa0 [bnx2fc]
 kernel: [<ffffffffa03c5265>] bnx2fc_ulp_exit+0xf5/0x160 [bnx2fc]    <- got bnx2fc_dev_lock mutex_lock
 kernel: [<ffffffffa03b03c6>] cnic_ulp_exit+0xb6/0xc0 [cnic]
 kernel: [<ffffffffa03b5418>] cnic_netdev_event+0x368/0x370 [cnic]
 kernel: [<ffffffffa038c56c>] ? fcoe_del_netdev_mapping+0x8c/0xa0 [libfcoe]
 kernel: [<ffffffff8152a6e5>] notifier_call_chain+0x55/0x80
 kernel: [<ffffffff810a0a46>] raw_notifier_call_chain+0x16/0x20
 kernel: [<ffffffff81459beb>] call_netdevice_notifiers+0x1b/0x20
 kernel: [<ffffffff8145ab34>] rollback_registered_many+0x154/0x280
 kernel: [<ffffffff8145ad08>] rollback_registered+0x38/0x50
 kernel: [<ffffffff8145ad78>] unregister_netdevice_queue+0x58/0xa0
 kernel: [<ffffffff8145add0>] unregister_netdevice+0x10/0x20
 kernel: [<ffffffff8145adfe>] unregister_netdev+0x1e/0x30                  <- got rtnl_lock!!!!!!!!!
 kernel: [<ffffffffa0122278>] __bnx2x_remove+0x48/0x270 [bnx2x] <- got & rel rtnl_lock
 kernel: [<ffffffffa0122554>] bnx2x_remove_one+0x44/0x80 [bnx2x]
 kernel: [<ffffffff812a3af7>] pci_device_remove+0x37/0x70
 kernel: [<ffffffff8136b2ef>] __device_release_driver+0x6f/0xe0
 kernel: [<ffffffff8136b428>] driver_detach+0xc8/0xd0
 kernel: [<ffffffff8136a22e>] bus_remove_driver+0x8e/0x110
 kernel: [<ffffffff8136bc12>] driver_unregister+0x62/0xa0
 kernel: [<ffffffff812a3e04>] pci_unregister_driver+0x44/0xb0
 kernel: [<ffffffffa0191954>] bnx2x_cleanup+0x18/0x73 [bnx2x]
 kernel: [<ffffffff810b8be4>] sys_delete_module+0x194/0x260
 kernel: [<ffffffff810e1347>] ? audit_syscall_entry+0x1d7/0x200
 kernel: [<ffffffff8100b072>] system_call_fastpath+0x16/0x1b

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] bnx2fc: Fixed a SCSI CMD cmpl race condition between ABTS and CLEANUP
Eddie Wai [Wed, 18 Sep 2013 05:33:10 +0000 (22:33 -0700)]
[SCSI] bnx2fc: Fixed a SCSI CMD cmpl race condition between ABTS and CLEANUP

In the case when a SCSI_CMD times out, bnx2fc will initiate the sending of the
ABTS.  However, if the SCSI layer's SCSI command timer also times out, it'll
instantiate a task abort of the same xid.

The race condition this patch tries to fix is as follows:

SCSI_CMD timeout (20s)
thread 1                   thread 2
send ABTS
rx ABTS cmpl
                           task abort_eh
                           explicit LOGO since ABTS was engaged
                           CLEANUP cmpl
SCSI_CMD cmpl (ABTS cmpl)
instantiate RRQ
wait 10s
attempt to send RRQ (because of LOGO, it wouldn't continue)

Note that there is no call to scsi_done for this SCSI_CMD cmpletion
in this path.

The patch changes the path of execution to call scsi_done immediately
instead of instantiating the RRQ.

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] qla4xxx: Add support to get CHAP details for flash target session
Adheer Chandravanshi [Tue, 17 Sep 2013 11:54:50 +0000 (07:54 -0400)]
[SCSI] qla4xxx: Add support to get CHAP details for flash target session

Add support to get local CHAP - index, username and password,
sysfs params of iscsi session corresponding to flash target entry.

Signed-off-by: Adheer Chandravanshi <adheer.chandravanshi@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] qla4xxx: Add support to set CHAP entries
Adheer Chandravanshi [Tue, 17 Sep 2013 11:54:49 +0000 (07:54 -0400)]
[SCSI] qla4xxx: Add support to set CHAP entries

Provide support to add/update the CHAP entries in adapter's flash
using iscsi tools, like Open-iSCSI.

Signed-off-by: Adheer Chandravanshi <adheer.chandravanshi@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] scsi_transport_iscsi: Add support to set CHAP entries
Adheer Chandravanshi [Tue, 17 Sep 2013 11:54:48 +0000 (07:54 -0400)]
[SCSI] scsi_transport_iscsi: Add support to set CHAP entries

For offload iSCSI like qla4xxx, CHAP entries are stored in adapter's
flash.
This patch adds support to add/update CHAP entries in adapter's flash
using iscsi tools, like Open-iSCSI.

Signed-off-by: Adheer Chandravanshi <adheer.chandravanshi@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] qla4xxx: Use offset based on adapter type to set CHAP entry in flash
Adheer Chandravanshi [Tue, 17 Sep 2013 11:54:47 +0000 (07:54 -0400)]
[SCSI] qla4xxx: Use offset based on adapter type to set CHAP entry in flash

To write a CHAP entry in adapter's flash calculate the offset based
on the type of adapter.

Signed-off-by: Adheer Chandravanshi <adheer.chandravanshi@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] qla4xxx: Populate local CHAP credentials for flash target sessions
Adheer Chandravanshi [Tue, 17 Sep 2013 11:54:46 +0000 (07:54 -0400)]
[SCSI] qla4xxx: Populate local CHAP credentials for flash target sessions

If any flash target entry is using CHAP authentication then set
CHAP username and password sysfs params for the corresponding
iscsi sessions.

Signed-off-by: Adheer Chandravanshi <adheer.chandravanshi@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] qla4xxx: Support setting of local CHAP index for flash target entry
Adheer Chandravanshi [Tue, 17 Sep 2013 11:54:45 +0000 (07:54 -0400)]
[SCSI] qla4xxx: Support setting of local CHAP index for flash target entry

Support setting of CHAP_OUT_IDX param for the target entry in flash.
Setting of valid local CHAP index with enable CHAP AUTH for that
flash target entry and disabling CHAP AUTH will invalidate the CHAP
index for the flash target entry.

Signed-off-by: Adheer Chandravanshi <adheer.chandravanshi@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] qla4xxx: Correct the check for local CHAP entry type
Adheer Chandravanshi [Tue, 17 Sep 2013 11:54:44 +0000 (07:54 -0400)]
[SCSI] qla4xxx: Correct the check for local CHAP entry type

Signed-off-by: Adheer Chandravanshi <adheer.chandravanshi@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] qla4xxx: correctly update session discovery_parent_idx.
Manish Rangankar [Tue, 17 Sep 2013 11:30:02 +0000 (07:30 -0400)]
[SCSI] qla4xxx: correctly update session discovery_parent_idx.

Earlier logic for driver created iscsi_session->discovery_parent_idx
was to store ram index of a sendtarget entry, but driver frees
sendtarget ram index as soon as firmware is done with discovery,
which is available for further use. So changing the logic to point
iscsi_session->discovery_parent_idx to store sendtarget flashnode index.

Signed-off-by: Manish Rangankar <manish.rangankar@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] megaraid_sas: fixes for few endianess issues
Sumit.Saxena@lsi.com [Mon, 16 Sep 2013 09:48:06 +0000 (15:18 +0530)]
[SCSI] megaraid_sas: fixes for few endianess issues

Fixed two issues in this patch-
1) In function megasas_get_pd_list(), data read(pd_addr->deviceId) from DMAed memory is converted to CPU's endianess.
2) While register AEN, removed some endianness conversion on some fields, since their endianess is already converted.

Signed-off-by: Sumit Saxena <sumit.saxena@lsi.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] buslogic: Added check for DMA mapping errors
Khalid Aziz [Fri, 13 Sep 2013 19:44:06 +0000 (13:44 -0600)]
[SCSI] buslogic: Added check for DMA mapping errors

Added check for DMA mapping errors for request sense data
buffer. Checking for mapping error can avoid potential wild
writes. This patch was prompted by the warning from
dma_unmap when kernel is compiled with CONFIG_DMA_API_DEBUG.

Signed-off-by: Khalid Aziz <khalid.aziz@oracle.com>
Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] fnic: Incremented driver version
Hiral Patel [Fri, 13 Sep 2013 00:45:43 +0000 (17:45 -0700)]
[SCSI] fnic: Incremented driver version

Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] fnic: Fnic Statistics Collection
Hiral Patel [Fri, 13 Sep 2013 00:45:42 +0000 (17:45 -0700)]
[SCSI] fnic: Fnic Statistics Collection

This feature gathers active and cumulative per fnic stats for io,
abort, terminate, reset, vlan discovery path and it also includes
various important stats for debugging issues. It also provided
debugfs and ioctl interface for user to retrieve these stats.
It also provides functionality to reset cumulative stats through
user interface.

Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] fnic: host reset returns nonzero value(errno) on success
Narsimhulu Musini [Fri, 13 Sep 2013 00:45:41 +0000 (17:45 -0700)]
[SCSI] fnic: host reset returns nonzero value(errno) on success

Fixed appropriate error codes that returns negative error number on failure,
and 0 on success. fnic_reset() is used directly by the fc transport callback
issue_fc_host_lip which requires a negative error number on failure.

Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] export device_busy for sdev
Jack Wang [Thu, 12 Sep 2013 15:29:52 +0000 (17:29 +0200)]
[SCSI] export device_busy for sdev

If you mutiple devices connect to a host, we might be interested in
have an intensive I/O workload on one disk, and notice starvation on others.
This give the user more hint about current infight io for scsi device.

Signed-off-by: Jack Wang <jinpu.wang@profitbricks.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] scsi_dh_rdac: Add new IBM 1813 product id to rdac devlist
Stewart, Sean [Wed, 11 Sep 2013 22:54:07 +0000 (22:54 +0000)]
[SCSI] scsi_dh_rdac: Add new IBM 1813 product id to rdac devlist

Signed-off-by: Sean Stewart <Sean.Stewart@netapp.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] lpfc: remove unnecessary read of PCI_CAP_ID_EXP
Jon Mason [Wed, 11 Sep 2013 22:38:13 +0000 (15:38 -0700)]
[SCSI] lpfc: remove unnecessary read of PCI_CAP_ID_EXP

The PCIE capability offset is saved during PCI bus walking.  It will
remove an unnecessary search in the PCI configuration space if this
value is referenced instead of reacquiring it.  Also, pci_is_pcie is a
better way of determining if the device is PCIE or not (as it uses the
same saved PCIE capability offset).

Signed-off-by: Jon Mason <jdmason@kudzu.us>
Acked-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] hpsa: remove unused Smart Array ID
Mike Miller [Tue, 10 Sep 2013 18:18:23 +0000 (13:18 -0500)]
[SCSI] hpsa: remove unused Smart Array ID

This patch removes the PCI ID of a cancelled Smart Array.

Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] sd: call blk_pm_runtime_init before add_disk
Aaron Lu [Thu, 10 Oct 2013 05:22:36 +0000 (13:22 +0800)]
[SCSI] sd: call blk_pm_runtime_init before add_disk

Sujit has found a race condition that would make q->nr_pending
unbalanced, it occurs as Sujit explained:

"
sd_probe_async() ->
add_disk() ->
disk_add_event() ->
schedule(disk_events_workfn)
sd_revalidate_disk()
blk_pm_runtime_init()
return;

Let's say the disk_events_workfn() calls sd_check_events() which tries
to send test_unit_ready() and because of sd_revalidate_disk() trying to
send another commands the test_unit_ready() might be re-queued as the
tagged command queuing is disabled.

So the race condition is -

Thread 1    | Thread 2
sd_revalidate_disk()   | sd_check_events()
...nr_pending = 0 as q->dev = NULL| scsi_queue_insert()
blk_runtime_pm_init()   |  blk_pm_requeue_request() ->
  | nr_pending = -1 since
  | q->dev != NULL
"

The problem is, the test_unit_ready request doesn't get counted the
first time it is queued, so the later decrement of q->nr_pending in
blk_pm_requeue_request makes it unbalanced.

Fix this by calling blk_pm_runtime_init before add_disk so that all
requests initiated there will all be counted.

Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reported-and-tested-by: Sujit Reddy Thumma <sthumma@codeaurora.org>
Cc: stable@vger.kernel.org
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] qla2xxx: Fix request queue null dereference.
Chad Dupuis [Thu, 3 Oct 2013 07:21:13 +0000 (03:21 -0400)]
[SCSI] qla2xxx: Fix request queue null dereference.

If an invalid IOCB is returned on the response queue then the index into the
request queue map could be invalid and could return to us a bogus value. This
could cause us to try to deference an invalid pointer and cause an exception.

If we encounter this condition, simply return as no context can be established
for this response.

Signed-off-by: Chad Dupuis <chad.dupuis@qlogic.com>
Signed-off-by: Saurav Kashyap <saurav.kashyap@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] BusLogic: Fix an oops when intializing multimaster adapter
Khalid Aziz [Wed, 25 Sep 2013 17:45:11 +0000 (11:45 -0600)]
[SCSI] BusLogic: Fix an oops when intializing multimaster adapter

This fixes an oops caused by buslogic driver when initializing a BusLogic
MultiMaster adapter. Initialization code used scope of a variable
incorrectly which created a NULL pointer. Oops message is below:

BUG: unable to handle kernel NULL pointer dereference at 0000000c
IP: [<c150c137>] blogic_init_mm_probeinfo.isra.17+0x20a/0x583
*pde = 00000000
Oops: 002 [#1] PREEMPT SMP
Modules linked in:
CPU: 1 PID: 1 Comm: swapper/0 Not tainted 3.11.1.puz1 #1
Hardware name:    /Canterwood, BIOS 6.00 PG 05/16/2003
task: f7050000 ti: f7054000 task.ti: f7054000
EIP: 0060:[<c150c137>] EFLAGS: 00010246 CPU:1
EIP is at blogic_init_mm_probeinfo.isra.17+0x20a/0x583
EAX: 00000013 EBX: 00000000 ECX: 00000000 EDX: f8001000
ESI: f71cb800 EDI: f7388000 EBP: 00007800 ESP: f7055c84
 DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068
CR0: 8005003b CR2: 0000000c CR3: 0154f000 CR4: 000007d0
Stack:
 0000001c 00000000 c11a59f6 f7055c98 00008130 ffffffff ffffffff 00000000
 00000003 00000000 00000000 00000000 00000013 f8001000 00000001 000003d0
 00000000 00000000 00000000 c14e3f84 f78803c8 00000000 f738c000 000000e9
Call Trace:
 [<c11a59f6>] ? pci_get_subsys+0x33/0x38
 [<c150c4fb>] ? blogic_init_probeinfo_list+0x4b/0x19e
 [<c108d593>] ? __alloc_pages_nodemask+0xe3/0x623
 [<c108d593>] ? __alloc_pages_nodemask+0xe3/0x623
 [<c10fb99e>] ? sysfs_link_sibling+0x61/0x8d
 [<c10b0519>] ? kmem_cache_alloc+0x8b/0xb5
 [<c150cce5>] ? blogic_init+0xa1/0x10e8
 [<c10fc0a8>] ? sysfs_add_one+0x10/0x9d
 [<c10fc18a>] ? sysfs_addrm_finish+0x12/0x85
 [<c10fca37>] ? sysfs_do_create_link_sd+0x9d/0x1b4
 [<c117c272>] ? blk_register_queue+0x69/0xb3
 [<c10fcb68>] ? sysfs_create_link+0x1a/0x2c
 [<c1181a07>] ? add_disk+0x1a1/0x3c7
 [<c138737b>] ? klist_next+0x60/0xc3
 [<c122cc3a>] ? scsi_dh_detach+0x68/0x68
 [<c1213e36>] ? bus_for_each_dev+0x51/0x61
 [<c1000356>] ? do_one_initcall+0x22/0x12c
 [<c10f3688>] ? __proc_create+0x8c/0xba
 [<c150cc44>] ? blogic_setup+0x5f6/0x5f6
 [<c14e94aa>] ? repair_env_string+0xf/0x4d
 [<c14e949b>] ? do_early_param+0x71/0x71
 [<c103efaa>] ? parse_args+0x21f/0x33d
 [<c14e9a54>] ? kernel_init_freeable+0xdf/0x17d
 [<c14e949b>] ? do_early_param+0x71/0x71
 [<c1388b64>] ? kernel_init+0x8/0xc0
 [<c1392222>] ? ret_from_kernel_thread+0x6/0x28
 [<c1392227>] ? ret_from_kernel_thread+0x1b/0x28
 [<c1388b5c>] ? rest_init+0x6c/0x6c
Code: 89 44 24 10 0f b6 44 24 3d 89 44 24 0c c7 44 24 08 00 00 00 00 c7 44 24 04 38 62 46 c1 c7 04 24 02 00 00 00 e8 78 13 d2 ff 31 db <89> 6b 0c b0 20 89 ea ee
 c7 44 24 08 04 00 00 00 8d 44 24 4c 89
EIP: [<c150c137>] blogic_init_mm_probeinfo.isra.17+0x20a/0x583 SS:ESP 0068:f7055c84
CR2: 000000000000000c
---[ end trace 17f45f5196d40487 ]---
Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000009

Signed-off-by: Khalid Aziz <khalid.aziz@oracle.com>
Cc: <stable@vger.kernel.org> # 3.11.x
Reported-by: Pierre Uszynski <pierre@rahul.net>
Tested-by: Pierre Uszynski <pierre@rahul.net>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years agoLinux 3.12-rc5
Linus Torvalds [Sun, 13 Oct 2013 22:41:28 +0000 (15:41 -0700)]
Linux 3.12-rc5

11 years agoMerge git://www.linux-watchdog.org/linux-watchdog
Linus Torvalds [Sun, 13 Oct 2013 18:41:26 +0000 (11:41 -0700)]
Merge git://www.linux-watchdog.org/linux-watchdog

Pull watchdog fixes from Wim Van Sebroeck:
 "This will fix a deadlock on the ts72xx_wdt driver, fix bitmasks in the
  kempld_wdt driver and fix a section mismatch in the sunxi_wdt driver"

* git://www.linux-watchdog.org/linux-watchdog:
  watchdog: sunxi: Fix section mismatch
  watchdog: kempld_wdt: Fix bit mask definition
  watchdog: ts72xx_wdt: locking bug in ioctl

11 years agowatchdog: sunxi: Fix section mismatch
Maxime Ripard [Sat, 5 Oct 2013 14:20:17 +0000 (16:20 +0200)]
watchdog: sunxi: Fix section mismatch

This driver has a section mismatch, for probe and remove functions,
leading to the following warning during the compilation.

WARNING: drivers/watchdog/built-in.o(.data+0x24): Section mismatch in
reference from the variable sunxi_wdt_driver to the function
.init.text:sunxi_wdt_probe()
The variable sunxi_wdt_driver references
the function __init sunxi_wdt_probe()

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
11 years agowatchdog: kempld_wdt: Fix bit mask definition
Jingoo Han [Mon, 23 Sep 2013 10:16:57 +0000 (19:16 +0900)]
watchdog: kempld_wdt: Fix bit mask definition

STAGE_CFG bits are defined as [5:4] bits. However, '(((x) & 0x30) << 4)'
handles [9:8] bits. Thus, it should be fixed in order to handle
[5:4] bits.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
11 years agowatchdog: ts72xx_wdt: locking bug in ioctl
Dan Carpenter [Fri, 23 Aug 2013 08:40:59 +0000 (11:40 +0300)]
watchdog: ts72xx_wdt: locking bug in ioctl

Calling the WDIOC_GETSTATUS & WDIOC_GETBOOTSTATUS and twice will cause a
interruptible deadlock.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
11 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Sun, 13 Oct 2013 16:59:10 +0000 (09:59 -0700)]
Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "A small batch of fixes this week, mostly OMAP related.  Nothing stands
  out as particularly controversial.

  Also a fix for a 3.12-rc1 timer regression for Exynos platforms,
  including the Chromebooks"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: exynos: dts: Update 5250 arch timer node with clock frequency
  ARM: OMAP2: RX-51: Add missing max_current to rx51_lp5523_led_config
  ARM: mach-omap2: board-generic: fix undefined symbol
  ARM: dts: Fix pinctrl mask for omap3
  ARM: OMAP3: Fix hardware detection for omap3630 when booted with device tree
  ARM: OMAP2: gpmc-onenand: fix sync mode setup with DT

11 years agoARM: exynos: dts: Update 5250 arch timer node with clock frequency
Yuvaraj Kumar C D [Wed, 18 Sep 2013 10:11:53 +0000 (15:41 +0530)]
ARM: exynos: dts: Update 5250 arch timer node with clock frequency

Without the "clock-frequency" property in arch timer node, could able
to see the below crash dump.

[<c0014e28>] (unwind_backtrace+0x0/0xf4) from [<c0011808>] (show_stack+0x10/0x14)
[<c0011808>] (show_stack+0x10/0x14) from [<c036ac1c>] (dump_stack+0x7c/0xb0)
[<c036ac1c>] (dump_stack+0x7c/0xb0) from [<c01ab760>] (Ldiv0_64+0x8/0x18)
[<c01ab760>] (Ldiv0_64+0x8/0x18) from [<c0062f60>] (clockevents_config.part.2+0x1c/0x74)
[<c0062f60>] (clockevents_config.part.2+0x1c/0x74) from [<c0062fd8>] (clockevents_config_and_register+0x20/0x2c)
[<c0062fd8>] (clockevents_config_and_register+0x20/0x2c) from [<c02b8e8c>] (arch_timer_setup+0xa8/0x134)
[<c02b8e8c>] (arch_timer_setup+0xa8/0x134) from [<c04b47b4>] (arch_timer_init+0x1f4/0x24c)
[<c04b47b4>] (arch_timer_init+0x1f4/0x24c) from [<c04b40d8>] (clocksource_of_init+0x34/0x58)
[<c04b40d8>] (clocksource_of_init+0x34/0x58) from [<c049ed8c>] (time_init+0x20/0x2c)
[<c049ed8c>] (time_init+0x20/0x2c) from [<c049b95c>] (start_kernel+0x1e0/0x39c)

THis is because the Exynos u-boot, for example on the Chromebooks, doesn't set
up the CNTFRQ register as expected by arch_timer. Instead, we have to specify
the frequency in the device tree like this.

Signed-off-by: Yuvaraj Kumar C D <yuvaraj.cd@samsung.com>
[olof: Changed subject, added comment, elaborated on commit message]
Signed-off-by: Olof Johansson <olof@lixom.net>
11 years agoMerge tag 'fixes-against-v3.12-rc3-take2' of git://git.kernel.org/pub/scm/linux/kerne...
Olof Johansson [Sun, 13 Oct 2013 16:33:32 +0000 (09:33 -0700)]
Merge tag 'fixes-against-v3.12-rc3-take2' of git://git./linux/kernel/git/tmlind/linux-omap into fixes

From Tony Lindgren:

Few fixes for omap3 related hangs and errors that people have
noticed now that people are actually using the device tree
based booting for omap3.

Also one regression fix for timer compile for dra7xx when
omap5 is not selected, and a LED regression fix for n900.

* tag 'fixes-against-v3.12-rc3-take2' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap:
  ARM: OMAP2: RX-51: Add missing max_current to rx51_lp5523_led_config
  ARM: mach-omap2: board-generic: fix undefined symbol
  ARM: dts: Fix pinctrl mask for omap3
  ARM: OMAP3: Fix hardware detection for omap3630 when booted with device tree
  ARM: OMAP2: gpmc-onenand: fix sync mode setup with DT

Signed-off-by: Olof Johansson <olof@lixom.net>
11 years agoMerge branch 'parisc-3.12' of git://git.kernel.org/pub/scm/linux/kernel/git/deller...
Linus Torvalds [Sun, 13 Oct 2013 16:13:28 +0000 (09:13 -0700)]
Merge branch 'parisc-3.12' of git://git./linux/kernel/git/deller/parisc-linux

Pull parisc fixes from Helge Deller:
 "This patchset includes a bugfix to prevent a kernel crash when memory
  in page zero is accessed by the kernel itself, e.g.  via
  probe_kernel_read().

  Furthermore we now export flush_cache_page() which is needed
  (indirectly) by the lustre filesystem.  The other patches remove
  unused functions and optimizes the page fault handler to only evaluate
  variables if needed, which again protects against possible kernel
  crashes"

* 'parisc-3.12' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: let probe_kernel_read() capture access to page zero
  parisc: optimize variable initialization in do_page_fault
  parisc: fix interruption handler to respect pagefault_disable()
  parisc: mark parisc_terminate() noreturn and cold.
  parisc: remove unused syscall_ipi() function.
  parisc: kill SMP single function call interrupt
  parisc: Export flush_cache_page() (needed by lustre)

11 years agoMerge branch 'fixes' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Sun, 13 Oct 2013 16:02:03 +0000 (09:02 -0700)]
Merge branch 'fixes' of git://git.infradead.org/users/vkoul/slave-dma

Pull slave-dmaengine fixes from Vinod Koul:
 "Another week, time to send another fixes request taking time out of
  extended weekend for the festivities in this part of the world.

  We have two fixes from Sergei for rcar driver and one fixing memory
  leak of edma driver by Geyslan"

* 'fixes' of git://git.infradead.org/users/vkoul/slave-dma:
  dma: edma.c: remove edma_desc leakage
  rcar-hpbdma: add parameter to set_slave() method
  rcar-hpbdma: remove shdma_free_irq() calls

11 years agoparisc: let probe_kernel_read() capture access to page zero
Helge Deller [Wed, 9 Oct 2013 21:47:03 +0000 (23:47 +0200)]
parisc: let probe_kernel_read() capture access to page zero

Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: optimize variable initialization in do_page_fault
John David Anglin [Sat, 5 Oct 2013 14:55:36 +0000 (10:55 -0400)]
parisc: optimize variable initialization in do_page_fault

The attached change defers the initialization of the variables tsk, mm
and flags until they are needed. As a result, the code won't crash if a
kernel probe is done with a corrupt context and the code will be better
optimized.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: fix interruption handler to respect pagefault_disable()
Helge Deller [Tue, 1 Oct 2013 19:54:46 +0000 (21:54 +0200)]
parisc: fix interruption handler to respect pagefault_disable()

Running an "echo t > /proc/sysrq-trigger" crashes the parisc kernel.  The
problem is, that in print_worker_info() we try to read the workqueue info via
the probe_kernel_read() functions which use pagefault_disable() to avoid
crashes like this:
    probe_kernel_read(&pwq, &worker->current_pwq, sizeof(pwq));
    probe_kernel_read(&wq, &pwq->wq, sizeof(wq));
    probe_kernel_read(name, wq->name, sizeof(name) - 1);

The problem here is, that the first probe_kernel_read(&pwq) might return zero
in pwq and as such the following probe_kernel_reads() try to access contents of
the page zero which is read protected and generate a kernel segfault.

With this patch we fix the interruption handler to call parisc_terminate()
directly only if pagefault_disable() was not called (in which case
preempt_count()==0).  Otherwise we hand over to the pagefault handler which
will try to look up the faulting address in the fixup tables.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: <stable@vger.kernel.org> # v3.0+
Signed-off-by: John David Anglin <dave.anglin@bell.net>
Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: mark parisc_terminate() noreturn and cold.
Helge Deller [Wed, 9 Oct 2013 21:12:46 +0000 (23:12 +0200)]
parisc: mark parisc_terminate() noreturn and cold.

Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: remove unused syscall_ipi() function.
Helge Deller [Wed, 9 Oct 2013 21:10:14 +0000 (23:10 +0200)]
parisc: remove unused syscall_ipi() function.

Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: kill SMP single function call interrupt
Jiang Liu [Wed, 11 Sep 2013 16:07:18 +0000 (00:07 +0800)]
parisc: kill SMP single function call interrupt

Commit 9a46ad6d6df3b54 "smp: make smp_call_function_many() use logic
similar to smp_call_function_single()" has unified the way to handle
single and multiple cross-CPU function calls. Now only one interrupt
is needed for architecture specific code to support generic SMP function
call interfaces, so kill the redundant single function call interrupt.

Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Cc: Jiang Liu <liuj97@gmail.com>
Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: Export flush_cache_page() (needed by lustre)
Geert Uytterhoeven [Thu, 5 Sep 2013 09:33:15 +0000 (11:33 +0200)]
parisc: Export flush_cache_page() (needed by lustre)

ERROR: "flush_cache_page" [drivers/staging/lustre/lustre/libcfs/libcfs.ko] undefined!

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Helge Deller <deller@gmx.de>
11 years agovfs: allow O_PATH file descriptors for fstatfs()
Linus Torvalds [Mon, 30 Sep 2013 15:35:10 +0000 (08:35 -0700)]
vfs: allow O_PATH file descriptors for fstatfs()

Olga reported that file descriptors opened with O_PATH do not work with
fstatfs(), found during further development of ksh93's thread support.

There is no reason to not allow O_PATH file descriptors here (fstatfs is
very much a path operation), so use "fdget_raw()".  See commit
55815f70147d ("vfs: make O_PATH file descriptors usable for 'fstat()'")
for a very similar issue reported for fstat() by the same team.

Reported-and-tested-by: ольга крыжановская <olga.kryzhanovska@gmail.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@kernel.org # O_PATH introduced in 3.0+
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 12 Oct 2013 19:55:15 +0000 (12:55 -0700)]
Merge tag 'ext4_for_linus_stable' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 bugfixes from Ted Ts'o:
 "A bug fix and performance regression fix for ext4"

* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: fix memory leak in xattr
  ext4: fix performance regression in writeback of random writes

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Sat, 12 Oct 2013 19:54:24 +0000 (12:54 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "We've got more bug fixes in my for-linus branch:

  One of these fixes another corner of the compression oops from last
  time.  Miao nailed down some problems with concurrent snapshot
  deletion and drive balancing.

  I kept out one of his patches for more testing, but these are all
  stable"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: fix oops caused by the space balance and dead roots
  Btrfs: insert orphan roots into fs radix tree
  Btrfs: limit delalloc pages outside of find_delalloc_range
  Btrfs: use right root when checking for hash collision

11 years agoMerge tag 'sound-3.12' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
Linus Torvalds [Sat, 12 Oct 2013 18:53:43 +0000 (11:53 -0700)]
Merge tag 'sound-3.12' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "All stable fixes except for a trivial headset mic fixup: the removal
  of bogus frame checks in snd-usb-usx2y driver that have regressed in
  the recent kernel versions, the HD-audio HDMI channel map fix, and a
  few HD-audio device-specific fixes"

* tag 'sound-3.12' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda - Sony VAIO Pro 13 (haswell) now has a working headset jack
  ALSA: hda - Add a headset mic model for ALC269 and friends
  ALSA: hda - Fix microphone for Sony VAIO Pro 13 (Haswell model)
  ALSA: hda - Add fixup for ASUS N56VZ
  ALSA: hda - hdmi: Fix channel map switch not taking effect
  ALSA: hda - Fix mono speakers and headset mic on Dell Vostro 5470
  ALSA: snd-usb-usx2y: remove bogus frame checks

11 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sat, 12 Oct 2013 18:52:40 +0000 (11:52 -0700)]
Merge branch 'i2c/for-current' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "We had various reports of problems with deferred probing in the I2C
  subsystem, so this pull requst is a little bigger than usual.

  Most issues should be addressed now so devices will be found
  correctly.  A few ususal driver bugfixes are in here, too"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: i2c-mux-pinctrl: use deferred probe when adapter not found
  i2c: i2c-arb-gpio-challenge: use deferred probe when adapter not found
  i2c: i2c-mux-gpio: use deferred probing
  i2c: i2c-mux-gpio: don't ignore of_get_named_gpio errors
  i2c: omap: Clear ARDY bit twice
  i2c: Not all adapters have a parent
  i2c: i2c-stu300: replace platform_driver_probe to support deferred probing
  i2c: i2c-mxs: replace platform_driver_probe to support deferred probing
  i2c: i2c-imx: replace platform_driver_probe to support deferred probing
  i2c: i2c-designware-platdrv: replace platform_driver_probe to support deferred probing

11 years agoext4: fix memory leak in xattr
Dave Jones [Fri, 11 Oct 2013 00:05:35 +0000 (20:05 -0400)]
ext4: fix memory leak in xattr

If we take the 2nd retry path in ext4_expand_extra_isize_ea, we
potentionally return from the function without having freed these
allocations.  If we don't do the return, we over-write the previous
allocation pointers, so we leak either way.

Spotted with Coverity.

[ Fixed by tytso to set is and bs to NULL after freeing these
  pointers, in case in the retry loop we later end up triggering an
  error causing a jump to cleanup, at which point we could have a double
  free bug. -- Ted ]

Signed-off-by: Dave Jones <davej@fedoraproject.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Cc: stable@vger.kernel.org
11 years agoMerge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 12 Oct 2013 18:06:18 +0000 (11:06 -0700)]
Merge branch 'core-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull gcc "asm goto" miscompilation workaround from Ingo Molnar:
 "This is the fix for the GCC miscompilation discussed in the following
  lkml thread:

    [x86] BUG: unable to handle kernel paging request at 00740060

  The bug in GCC has been fixed by Jakub and the fix will be part of the
  GCC 4.8.2 release expected to be released next week - so the quirk's
  version test checks for <= 4.8.1.

  The quirk is only added to compiler-gcc4.h and not to the higher level
  compiler.h because all asm goto uses are behind a feature check"

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  compiler/gcc4: Add quirk for 'asm goto' miscompilation bug

11 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 12 Oct 2013 17:36:03 +0000 (10:36 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "A build fix and a reboot quirk"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/reboot: Add reboot quirk for Dell Latitude E5410
  x86, build, pci: Fix PCI_MSI build on !SMP

11 years agoMerge tag 'arc-fixes-for-3.12-part3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 12 Oct 2013 17:34:14 +0000 (10:34 -0700)]
Merge tag 'arc-fixes-for-3.12-part3' of git://git./linux/kernel/git/vgupta/arc

Pull ARC fix from Vineet Gupta:
 "Fix for broken gdb 'jump'"

* tag 'arc-fixes-for-3.12-part3' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  ARC: Ignore ptrace SETREGSET request for synthetic register "stop_pc"

11 years agoARC: Ignore ptrace SETREGSET request for synthetic register "stop_pc"
Vineet Gupta [Thu, 10 Oct 2013 14:03:57 +0000 (19:33 +0530)]
ARC: Ignore ptrace SETREGSET request for synthetic register "stop_pc"

ARCompact TRAP_S insn used for breakpoints, commits before exception is
taken (updating architectural PC). So ptregs->ret contains next-PC and
not the breakpoint PC itself. This is different from other restartable
exceptions such as TLB Miss where ptregs->ret has exact faulting PC.
gdb needs to know exact-PC hence ARC ptrace GETREGSET provides for
@stop_pc which returns ptregs->ret vs. EFA depending on the
situation.

However, writing stop_pc (SETREGSET request), which updates ptregs->ret
doesn't makes sense stop_pc doesn't always correspond to that reg as
described above.

This was not an issue so far since user_regs->ret / user_regs->stop_pc
had same value and both writing to ptregs->ret was OK, needless, but NOT
broken, hence not observed.

With gdb "jump", they diverge, and user_regs->ret updating ptregs is
overwritten immediately with stop_pc, which this patch fixes.

Reported-by: Anton Kolesov <akolesov@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
11 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Fri, 11 Oct 2013 18:24:58 +0000 (11:24 -0700)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS fix from Ralf Baechle:
 "Just one fix.  The stack protector was loading the value of the canary
  instead of its address"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: stack protector: Fix per-task canary switch

11 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 11 Oct 2013 17:41:21 +0000 (10:41 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "All over the map..

   - nouveau:
     disable MSI, needs more work, will try again next merge window
   - radeon:
      audio + uvd regression fixes, dpm fixes, reset fixes
   - i915:
     the dpms fix might fix your haswell

  And one pain in the ass revert, so we have VGA arbitration that when
  implemented 4-5 years ago really hoped that GPUs could remove
  themselves from arbitration completely once they had a kernel driver.

  It seems Intel hw designers decided that was too nice a facility to
  allow us to have so they removed it when they went on-die (so since
  Ironlake at least).  Now Alex Williamson added support for VGA
  arbitration for newer GPUs however this now exposes itself to
  userspace as requireing arbitration of GPU VGA regions and the X
  server gets involved and disables things that it can't handle when VGA
  access is possibly required around every operation.

  So in order to not break userspace we just reverted things back to the
  old known broken status so maybe we can try and design out way out.

  Ville also had a patch to use stop machine for the two times Intel
  needs to access VGA space, that might be acceptable with some rework,
  but for now myself and Daniel agreed to just go back"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux: (23 commits)
  Revert "i915: Update VGA arbiter support for newer devices"
  Revert "drm/i915: Delay disabling of VGA memory until vgacon->fbcon handoff is done"
  drm/radeon: re-enable sw ACR support on pre-DCE4
  drm/radeon/dpm: disable bapm on TN asics
  drm/radeon: improve soft reset on CIK
  drm/radeon: improve soft reset on SI
  drm/radeon/dpm: off by one in si_set_mc_special_registers()
  drm/radeon/dpm/btc: off by one in btc_set_mc_special_registers()
  drm/radeon: forever loop on error in radeon_do_test_moves()
  drm/radeon: fix hw contexts for SUMO2 asics
  drm/radeon: fix typo in CP DMA register headers
  drm/radeon/dpm: disable multiple UVD states
  drm/radeon: use hw generated CTS/N values for audio
  drm/radeon: fix N/CTS clock matching for audio
  drm/radeon: use 64-bit math to calculate CTS values for audio (v2)
  drm/edid: catch kmalloc failure in drm_edid_to_speaker_allocation
  Revert "drm/fb-helper: don't sleep for screen unblank when an oops is in progress"
  drm/gma500: fix things after get/put page helpers
  drm/nouveau/mc: disable msi support by default, it's busted in tons of places
  drm/i915: Only apply DPMS to the encoder if enabled
  ...

11 years agoALSA: hda - Sony VAIO Pro 13 (haswell) now has a working headset jack
David Henningsson [Fri, 11 Oct 2013 11:46:04 +0000 (13:46 +0200)]
ALSA: hda - Sony VAIO Pro 13 (haswell) now has a working headset jack

Just got the positive confirmation from a tester:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1227093/comments/28

Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
11 years agoALSA: hda - Add a headset mic model for ALC269 and friends
David Henningsson [Fri, 11 Oct 2013 08:18:46 +0000 (10:18 +0200)]
ALSA: hda - Add a headset mic model for ALC269 and friends

Using the headset mic model will cause the headset mic to be labeled
"headset mic" instead of just "mic".

Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
11 years agoALSA: hda - Fix microphone for Sony VAIO Pro 13 (Haswell model)
David Henningsson [Fri, 11 Oct 2013 08:18:45 +0000 (10:18 +0200)]
ALSA: hda - Fix microphone for Sony VAIO Pro 13 (Haswell model)

The external mic showed up with a precense detect of "always present",
essentially disabling the internal mic. Therefore turn off presence
detection for this pin.

Note: The external mic seems not yet working, but an internal mic is
certainly better than no mic at all.

Cc: stable@vger.kernel.org
BugLink: https://bugs.launchpad.net/bugs/1227093
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
11 years agocompiler/gcc4: Add quirk for 'asm goto' miscompilation bug
Ingo Molnar [Thu, 10 Oct 2013 08:16:30 +0000 (10:16 +0200)]
compiler/gcc4: Add quirk for 'asm goto' miscompilation bug

Fengguang Wu, Oleg Nesterov and Peter Zijlstra tracked down
a kernel crash to a GCC bug: GCC miscompiles certain 'asm goto'
constructs, as outlined here:

  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58670

Implement a workaround suggested by Jakub Jelinek.

Reported-and-tested-by: Fengguang Wu <fengguang.wu@intel.com>
Reported-by: Oleg Nesterov <oleg@redhat.com>
Reported-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Suggested-by: Jakub Jelinek <jakub@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: <stable@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agoRevert "i915: Update VGA arbiter support for newer devices"
Dave Airlie [Fri, 11 Oct 2013 05:12:04 +0000 (15:12 +1000)]
Revert "i915: Update VGA arbiter support for newer devices"

This reverts commit 81b5c7bc8de3e6f63419139c2fc91bf81dea8a7d.

Adding drm/i915 into the vga arbiter chain means that X (in a piece of
well-meant paranoia) will do a get/put on the vga decoding around
_every_ accel call down into the ddx. Which results in some nice
performance disasters [1]. This really breaks userspace, by disabling
DRI for everyone, and stops OpenGL from working, this isn't limited
to just the i915 but both the integrated and discrete GPUs on
multi-gpu systems, in other words this causes untold worlds of pain,

Ville tried to come up with a Great Hack to fiddle the required VGA
I/O ops behind everyone's back using stop_machine, but that didn't
really work out [2]. Given that we're fairly late in the -rc stage for
such games let's just revert this all.

One thing we might want to keep is to delay the disabling of the vga
decoding until the fbdev emulation and the fbcon screen is set up. If
we kill vga mem decoding beforehand fbcon can end up with a white
square in the top-left corner it tried to save from the vga memory for
a seamless transition. And we have bug reports on older platforms
which seem to match these symptoms.

But again that's something to play around with in -next.

References: [1] http://lists.x.org/archives/xorg-devel/2013-September/037763.html
References: [2] http://www.spinics.net/lists/intel-gfx/msg34062.html
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoRevert "drm/i915: Delay disabling of VGA memory until vgacon->fbcon handoff is done"
Dave Airlie [Fri, 11 Oct 2013 05:11:52 +0000 (15:11 +1000)]
Revert "drm/i915: Delay disabling of VGA memory until vgacon->fbcon handoff is done"

This reverts commit 6e1b4fdad5157bb9e88777d525704aba24389bee.

This is part of a revert due to a userspace breakage, better explained in the revert of 1a1a4cbf4906a13c0c377f708df5d94168e7b582.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoMerge branch 'drm-fixes-3.12' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Fri, 11 Oct 2013 03:07:15 +0000 (13:07 +1000)]
Merge branch 'drm-fixes-3.12' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

Regression fixes for audio and UVD, several hang fixes,
some DPM fixes.

* 'drm-fixes-3.12' of git://people.freedesktop.org/~agd5f/linux:
  drm/radeon: re-enable sw ACR support on pre-DCE4
  drm/radeon/dpm: disable bapm on TN asics
  drm/radeon: improve soft reset on CIK
  drm/radeon: improve soft reset on SI
  drm/radeon/dpm: off by one in si_set_mc_special_registers()
  drm/radeon/dpm/btc: off by one in btc_set_mc_special_registers()
  drm/radeon: forever loop on error in radeon_do_test_moves()
  drm/radeon: fix hw contexts for SUMO2 asics
  drm/radeon: fix typo in CP DMA register headers
  drm/radeon/dpm: disable multiple UVD states
  drm/radeon: use hw generated CTS/N values for audio
  drm/radeon: fix N/CTS clock matching for audio
  drm/radeon: use 64-bit math to calculate CTS values for audio (v2)
  drm/edid: catch kmalloc failure in drm_edid_to_speaker_allocation

11 years agodma: edma.c: remove edma_desc leakage
Geyslan G. Bem [Mon, 7 Oct 2013 22:19:58 +0000 (19:19 -0300)]
dma: edma.c: remove edma_desc leakage

Free memory allocated to edma_desc when failing to allocate slot.

Signed-off-by: Geyslan G. Bem <geyslan@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agorcar-hpbdma: add parameter to set_slave() method
Sergei Shtylyov [Wed, 25 Sep 2013 22:31:37 +0000 (02:31 +0400)]
rcar-hpbdma: add parameter to set_slave() method

Commit 4981c4dc194efb18f0e9a02f1b43e926f2f0d2bb (DMA: shdma: switch DT mode to
use configuration data from a match table) added a new parameter to set_slave()
method but unfortunately got merged later than commit c4f6c41ba790bbbfcebb4c47a
(dma: add driver for R-Car HPB-DMAC), so that the HPB-DMAC driver retained the
old prototype which caused this warning:

drivers/dma/sh/rcar-hpbdma.c:485: warning: initialization from incompatible
pointer type

The newly added parameter  is used to override DMA slave address from 'struct
hpb_dmae_slave_config', so we have to add the 'slave_addr' field to 'struct
hpb_dmae_chan', conditionally assign it in set_slave() method, and return in
slave_addr() method.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Tested-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agorcar-hpbdma: remove shdma_free_irq() calls
Sergei Shtylyov [Wed, 25 Sep 2013 22:28:37 +0000 (02:28 +0400)]
rcar-hpbdma: remove shdma_free_irq() calls

Commit c1c63a14f4f2419d093acd7164eccdff315baa86 (DMA: shdma: switch to managed
resource allocation) got rid of shdma_free_irq() but  unfortunately got merged
later than commit c4f6c41ba790bbbfcebb4c47a709ac8ff1fe1af9 (dma: add driver for
R-Car HPB-DMAC), so that the HPB-DMAC driver retained the calls and got broken:

drivers/dma/sh/rcar-hpbdma.c: In function `hpb_dmae_alloc_chan_resources':
drivers/dma/sh/rcar-hpbdma.c:435: error: implicit declaration of function
`shdma_free_irq'

Fix this compilation error by removing the remaining shdma_free_irq() calls.

Reported-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Tested-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agoBtrfs: fix oops caused by the space balance and dead roots
Miao Xie [Wed, 25 Sep 2013 13:47:44 +0000 (21:47 +0800)]
Btrfs: fix oops caused by the space balance and dead roots

When doing space balance and subvolume destroy at the same time, we met
the following oops:

kernel BUG at fs/btrfs/relocation.c:2247!
RIP: 0010: [<ffffffffa04cec16>] prepare_to_merge+0x154/0x1f0 [btrfs]
Call Trace:
 [<ffffffffa04b5ab7>] relocate_block_group+0x466/0x4e6 [btrfs]
 [<ffffffffa04b5c7a>] btrfs_relocate_block_group+0x143/0x275 [btrfs]
 [<ffffffffa0495c56>] btrfs_relocate_chunk.isra.27+0x5c/0x5a2 [btrfs]
 [<ffffffffa0459871>] ? btrfs_item_key_to_cpu+0x15/0x31 [btrfs]
 [<ffffffffa048b46a>] ? btrfs_get_token_64+0x7e/0xcd [btrfs]
 [<ffffffffa04a3467>] ? btrfs_tree_read_unlock_blocking+0xb2/0xb7 [btrfs]
 [<ffffffffa049907d>] btrfs_balance+0x9c7/0xb6f [btrfs]
 [<ffffffffa049ef84>] btrfs_ioctl_balance+0x234/0x2ac [btrfs]
 [<ffffffffa04a1e8e>] btrfs_ioctl+0xd87/0x1ef9 [btrfs]
 [<ffffffff81122f53>] ? path_openat+0x234/0x4db
 [<ffffffff813c3b78>] ? __do_page_fault+0x31d/0x391
 [<ffffffff810f8ab6>] ? vma_link+0x74/0x94
 [<ffffffff811250f5>] vfs_ioctl+0x1d/0x39
 [<ffffffff811258c8>] do_vfs_ioctl+0x32d/0x3e2
 [<ffffffff811259d4>] SyS_ioctl+0x57/0x83
 [<ffffffff813c3bfa>] ? do_page_fault+0xe/0x10
 [<ffffffff813c73c2>] system_call_fastpath+0x16/0x1b

It is because we returned the error number if the reference of the root was 0
when doing space relocation. It was not right here, because though the root
was dead(refs == 0), but the space it held still need be relocated, or we
could not remove the block group. So in this case, we should return the root
no matter it is dead or not.

Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
11 years agoBtrfs: insert orphan roots into fs radix tree
Miao Xie [Wed, 25 Sep 2013 13:47:43 +0000 (21:47 +0800)]
Btrfs: insert orphan roots into fs radix tree

Now we don't drop all the deleted snapshots/subvolumes before the space
balance. It means we have to relocate the space which is held by the dead
snapshots/subvolumes. So we must into them into fs radix tree, or we would
forget to commit the change of them when doing transaction commit, and it
would corrupt the metadata.

Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
11 years agoBtrfs: limit delalloc pages outside of find_delalloc_range
Josef Bacik [Tue, 8 Oct 2013 02:11:09 +0000 (22:11 -0400)]
Btrfs: limit delalloc pages outside of find_delalloc_range

Liu fixed part of this problem and unfortunately I steered him in slightly the
wrong direction and so didn't completely fix the problem.  The problem is we
limit the size of the delalloc range we are looking for to max bytes and then we
try to lock that range.  If we fail to lock the pages in that range we will
shrink the max bytes to a single page and re loop.  However if our first page is
inside of the delalloc range then we will end up limiting the end of the range
to a period before our first page.  This is illustrated below

[0 -------- delalloc range --------- 256mb]
                                  [page]

So find_delalloc_range will return with delalloc_start as 0 and end as 128mb,
and then we will notice that delalloc_start < *start and adjust it up, but not
adjust delalloc_end up, so things go sideways.  To fix this we need to not limit
the max bytes in find_delalloc_range, but in find_lock_delalloc_range and that
way we don't end up with this confusion.  Thanks,

Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
11 years agoBtrfs: use right root when checking for hash collision
Josef Bacik [Wed, 9 Oct 2013 16:24:04 +0000 (12:24 -0400)]
Btrfs: use right root when checking for hash collision

btrfs_rename was using the root of the old dir instead of the root of the new
dir when checking for a hash collision, so if you tried to move a file into a
subvol it would freak out because it would see the file you are trying to move
in its current root.  This fixes the bug where this would fail

btrfs subvol create test1
btrfs subvol create test2
mv test1 test2.

Thanks to Chris Murphy for catching this,

Cc: stable@vger.kernel.org
Reported-by: Chris Murphy <lists@colorremedies.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
11 years agobcache: Fix a null ptr deref regression
Kent Overstreet [Fri, 11 Oct 2013 00:31:15 +0000 (17:31 -0700)]
bcache: Fix a null ptr deref regression

Commit c0f04d88e46d ("bcache: Fix flushes in writeback mode") was fixing
a reported data corruption bug, but it seems some last minute
refactoring or rebasing introduced a null pointer deref.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
Reported-by: Gabriel de Perthuis <g2p.code@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck...
Linus Torvalds [Fri, 11 Oct 2013 01:16:02 +0000 (18:16 -0700)]
Merge tag 'hwmon-for-linus' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon fix from Guenter Roeck:
 "Fix root cause of crash/error seen in applesmc driver"

* tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (applesmc) Always read until end of data

11 years agoMerge branch 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild
Linus Torvalds [Fri, 11 Oct 2013 01:15:23 +0000 (18:15 -0700)]
Merge branch 'rc-fixes' of git://git./linux/kernel/git/mmarek/kbuild

Pull kbuild fix from Michal Marek:
 "Here is an ARM Makefile fix that you even acked.  After nobody wanted
  to take it, it ended up in the kbuild tree"

* 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild:
  arm, kbuild: make "make install" not depend on vmlinux

11 years agoMerge git://www.linux-watchdog.org/linux-watchdog
Linus Torvalds [Thu, 10 Oct 2013 20:57:10 +0000 (13:57 -0700)]
Merge git://www.linux-watchdog.org/linux-watchdog

Pull watchdog fix from Wim Van Sebroeck:
 "Make sure that the hpwdt driver will not load auxilary iLO devices"

* git://www.linux-watchdog.org/linux-watchdog:
  watchdog: hpwdt: Patch to ignore auxilary iLO devices