pandora-kernel.git
16 years ago[S390] arch/s390: Add missing "space"
Joe Perches [Sat, 26 Jan 2008 13:11:09 +0000 (14:11 +0100)]
[S390] arch/s390: Add missing "space"

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] dasd: fix return value of dasd_generic_probe()
Stefan Haberland [Sat, 26 Jan 2008 13:11:08 +0000 (14:11 +0100)]
[S390] dasd: fix return value of dasd_generic_probe()

Using the return value of ccw_device_set_online as return value for
dasd_generic_probe() causes the DASD to fail setting online

Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] crypto: move s390 Kconfig options.
Jan Glauber [Sat, 26 Jan 2008 13:11:07 +0000 (14:11 +0100)]
[S390] crypto: move s390 Kconfig options.

Move s390 crypto Kconfig options to drivers/crypto/Kconfig to have all
hardware crypto devices in one place.

This also makes messing up the kernel source tree easier for some people.

Signed-off-by: Jan Glauber <jan.glauber@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Remove appldata include from sysctl_check.c
Heiko Carstens [Sat, 26 Jan 2008 13:11:06 +0000 (14:11 +0100)]
[S390] Remove appldata include from sysctl_check.c

Forgot to remove this when removing the appldata binary sysctls.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Get rid of additional_cpus kernel parameter.
Heiko Carstens [Sat, 26 Jan 2008 13:11:05 +0000 (14:11 +0100)]
[S390] Get rid of additional_cpus kernel parameter.

It caused only a lot of confusion. From now on cpu hotplug of up to
NR_CPUS will work by default. If somebody wants to limit that then
the possible_cpus parameter can be used.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Use new style spinlock initializer in __RWSEM_INITIALIZER.
Heiko Carstens [Sat, 26 Jan 2008 13:11:04 +0000 (14:11 +0100)]
[S390] Use new style spinlock initializer in __RWSEM_INITIALIZER.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Remove owner_pc member from raw_spinlock_t.
Heiko Carstens [Sat, 26 Jan 2008 13:11:03 +0000 (14:11 +0100)]
[S390] Remove owner_pc member from raw_spinlock_t.

Used to contain the address of the holder of the lock. But since the
spinlock code is not inlined anymore all locks contain the same address
anyway. And since in addtition nobody complained about that for ages
its obviously unused. So remove it.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Get rid of HOLES_IN_ZONE requirement.
Heiko Carstens [Sat, 26 Jan 2008 13:11:02 +0000 (14:11 +0100)]
[S390] Get rid of HOLES_IN_ZONE requirement.

Align everything to MAX_ORDER so we can get rid of the extra checks.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Print kernel version in dump_stack() and show_regs().
Heiko Carstens [Sat, 26 Jan 2008 13:11:01 +0000 (14:11 +0100)]
[S390] Print kernel version in dump_stack() and show_regs().

Also print PREEMPT and/or SMP if the kernel was configured that way.
Makes s390 look a bit more like other architectures.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Change vmalloc defintions
Christian Borntraeger [Sat, 26 Jan 2008 13:11:00 +0000 (14:11 +0100)]
[S390] Change vmalloc defintions

Currently the vmalloc area starts at a dynamic address depending on
the memory size. There was also an 8MB security hole after the
physical memory to catch out-of-bounds accesses.
We can simplify the code by putting the vmalloc area explicitely at
the top of the kernel mapping and setting the vmalloc size to a fixed
value of 128MB/128GB for 31bit/64bit systems. Part of the vmalloc
area will be used for the vmem_map. This leaves an area of 96MB/1GB
for normal vmalloc allocations.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Avoid warnings in tlblush.h
Martin Schwidefsky [Sat, 26 Jan 2008 13:10:59 +0000 (14:10 +0100)]
[S390] Avoid warnings in tlblush.h

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Fix tlb flushing with idte.
Martin Schwidefsky [Sat, 26 Jan 2008 13:10:58 +0000 (14:10 +0100)]
[S390] Fix tlb flushing with idte.

The clear-by-asce operation of the idte instruction gets an asce
(address-space-control-element) as argument to specify which TLBs
need to get flushed. The current code passes a plain pointer to
the start of the pgd without the additional bits which would make
the pointer an asce. The current machines don't mind the difference
but a future model might want to use the designation type control
bits in the asce as a filter for the TLBs to flush.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] sclp: convert channel path configure code to use sync interface.
Heiko Carstens [Sat, 26 Jan 2008 13:10:57 +0000 (14:10 +0100)]
[S390] sclp: convert channel path configure code to use sync interface.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Standby cpu activation/deactivation.
Heiko Carstens [Sat, 26 Jan 2008 13:10:56 +0000 (14:10 +0100)]
[S390] Standby cpu activation/deactivation.

Add a new interface so that cpus can be put into standby state and
configured state.
Only offline cpus can be put into standby state or configured state.
For that the new percpu sysfs attribute "configure" must be used.
To put a cpu in standby state a "0" must be written to the attribute.
In order to switch it into configured state a "1" must be written to
the attribute.
Only cpus in configured state can be brought online.
In addition this patch introduces a static mapping of physical to
logical cpus. As a result only the sysfs directories of present cpus
will be created. To scan for new cpus the new sysfs attribute "rescan"
must be used.
Writing to /sys/devices/system/cpu/rescan will trigger a rescan of
cpus and will create directories for new cpus.

On IPL only configured cpus will be used. And on reboot/shutdown all
cpus will remain in their current state (configured/standby).

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] sclp: sysfs interface for SCLP cpi
Michael Ernst [Sat, 26 Jan 2008 13:10:55 +0000 (14:10 +0100)]
[S390] sclp: sysfs interface for SCLP cpi

Signed-off-by: Michael Ernst <mernst@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] qdio: set QDIO_ACTIVATE_TIMEOUT to 5s
Ursula Braun [Sat, 26 Jan 2008 13:10:54 +0000 (14:10 +0100)]
[S390] qdio: set QDIO_ACTIVATE_TIMEOUT to 5s

Current definition of QDIO_ACTIVATE_TIMEOUT results in value 0.
Thus it may cause endless wait in function qdio_activate().

Signed-off-by: Ursula Braun <braunu@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] qdio: Remove double checked value.
Sebastian Ott [Sat, 26 Jan 2008 13:10:53 +0000 (14:10 +0100)]
[S390] qdio: Remove double checked value.

Signed-off-by: Sebastian Ott <sebott@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: introduce timed recovery procedure
Peter Oberparleiter [Sat, 26 Jan 2008 13:10:52 +0000 (14:10 +0100)]
[S390] cio: introduce timed recovery procedure

Add a timed recovery procedure to reactivate ccw devices in cases
where HW/VM events are not sufficient to allow for proper recovery
of reappearing channel paths.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: memory leak in cio processing
Michael Ernst [Sat, 26 Jan 2008 13:10:51 +0000 (14:10 +0100)]
[S390] cio: memory leak in cio processing

Allocated kernel memory for locks is not freed in case of subchannel found
to be invalid.

Signed-off-by: Michael Ernst <mernst@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Fix ungroup vs. machine check race in ccwgroup.
Cornelia Huck [Sat, 26 Jan 2008 13:10:50 +0000 (14:10 +0100)]
[S390] cio: Fix ungroup vs. machine check race in ccwgroup.

There are two ways for a ccwgroup device to be unregistered:
Via the ungroup attribute, or when on of the slave devices is killed
by a machine check. As we have to unregister the ccwgroup device
via a callback, we'll now have to check whether it hasn't been
already unregistered because of a machine check.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Delete timer in ccw_device_verify_done().
Cornelia Huck [Sat, 26 Jan 2008 13:10:49 +0000 (14:10 +0100)]
[S390] cio: Delete timer in ccw_device_verify_done().

Make sure pending timers are always deleted, even if we don't go
through ccw_device_done().

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: reduce cpu utilization during device scan
Peter Oberparleiter [Sat, 26 Jan 2008 13:10:48 +0000 (14:10 +0100)]
[S390] cio: reduce cpu utilization during device scan

Minimize calls to cpu intensive function get_subchannel_by_schid()
by introducing function for_each_subchannel_staged() which
temporarily caches the information about registered subchannels
in a bitmap.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Set driver->owner on css, ccw and ccwgroup busses.
Cornelia Huck [Sat, 26 Jan 2008 13:10:47 +0000 (14:10 +0100)]
[S390] cio: Set driver->owner on css, ccw and ccwgroup busses.

Set the owner field in the embedded struct device_driver to the
value provided in the {css,ccw,ccwgroup}_driver.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Use dev_{g,s}et_drvdata().
Cornelia Huck [Sat, 26 Jan 2008 13:10:46 +0000 (14:10 +0100)]
[S390] cio: Use dev_{g,s}et_drvdata().

Also define helpers sch_{g,s}et_cdev() to make the intention more
clear.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: I/O subchannel specific fields.
Cornelia Huck [Sat, 26 Jan 2008 13:10:45 +0000 (14:10 +0100)]
[S390] cio: I/O subchannel specific fields.

Some fields may be !0 only for I/O subchannels. Add some checks
where required. Also adapt cio_enable_subchannel() to make the
caller specify the intparm, which makes it more generic.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Extend adapter interrupt interface.
Peter Oberparleiter [Sat, 26 Jan 2008 13:10:44 +0000 (14:10 +0100)]
[S390] cio: Extend adapter interrupt interface.

From: Cornelia Huck <cornelia.huck@de.ibm.com>

Change the adapter interrupt interface in order to allow multiple
adapter interrupt handlers to be registered. Indicators are now
allocated by cio instead of the device driver.

The qdio parts have been
Acked-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Introduce subchannel->private.
Cornelia Huck [Sat, 26 Jan 2008 13:10:43 +0000 (14:10 +0100)]
[S390] cio: Introduce subchannel->private.

Introduce a private pointer in struct subchannel to store
per-subchannel type data (cannot use dev->priv since this
is already used for something else).

Create a new header io_sch.h for I/O subchannel specific structures
and instructions.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Cleanup debug feature usage.
Cornelia Huck [Sat, 26 Jan 2008 13:10:42 +0000 (14:10 +0100)]
[S390] cio: Cleanup debug feature usage.

Cleanup cio_debug.h.
Also make CIO_DEBUG add the "cio:" prefix to the printk string
so that it isn't needed for the debug feature.
Fix outdated comments for cio_debug_init() and clean it up.
Enlarge cio_crw to the same size as cio_msg so we may actually
find some relevant information there.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Add css_driver_{register,unregister}.
Cornelia Huck [Sat, 26 Jan 2008 13:10:41 +0000 (14:10 +0100)]
[S390] cio: Add css_driver_{register,unregister}.

Add wrapper functions for driver_register and driver_unregister so
that css drivers don't need to muck with struct device_driver
directly.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Reset sch->driver.
Cornelia Huck [Sat, 26 Jan 2008 13:10:40 +0000 (14:10 +0100)]
[S390] cio: Reset sch->driver.

sch->driver needs to be reset to NULL on failed probe and after
remove. We also need to check for sch->driver on shutdown.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: css_driver: Use consistent parameters.
Cornelia Huck [Sat, 26 Jan 2008 13:10:39 +0000 (14:10 +0100)]
[S390] cio: css_driver: Use consistent parameters.

Make all callbacks in css_driver take a struct subchannel (and not
a struct device).

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Use helpers instead of container_of().
Cornelia Huck [Sat, 26 Jan 2008 13:10:38 +0000 (14:10 +0100)]
[S390] cio: Use helpers instead of container_of().

- Introduce to_cssdriver.
- Use to_xxx instead of container_of where possible.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] cio: Dump ccw device information in case of timeout.
Sebastian Ott [Sat, 26 Jan 2008 13:10:37 +0000 (14:10 +0100)]
[S390] cio: Dump ccw device information in case of timeout.

Information about a ccw device will be dumped in
case of a ccw timeout. This can be enabled with
the kernel parameter ccw_timeout_log.

Signed-off-by: Sebastian Ott <sebott@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Cleanup in Documentation/kernel-parameters.txt.
Sebastian Ott [Sat, 26 Jan 2008 13:10:36 +0000 (14:10 +0100)]
[S390] Cleanup in Documentation/kernel-parameters.txt.

Signed-off-by: Sebastian Ott <sebott@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6
Linus Torvalds [Sat, 26 Jan 2008 01:19:08 +0000 (17:19 -0800)]
Merge git://git./linux/kernel/git/jejb/scsi-misc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6: (200 commits)
  [SCSI] usbstorage: use last_sector_bug flag universally
  [SCSI] libsas: abstract STP task status into a function
  [SCSI] ultrastor: clean up inline asm warnings
  [SCSI] aic7xxx: fix firmware build
  [SCSI] aacraid: fib context lock for management ioctls
  [SCSI] ch: remove forward declarations
  [SCSI] ch: fix device minor number management bug
  [SCSI] ch: handle class_device_create failure properly
  [SCSI] NCR5380: fix section mismatch
  [SCSI] sg: fix /proc/scsi/sg/devices when no SCSI devices
  [SCSI] IB/iSER: add logical unit reset support
  [SCSI] don't use __GFP_DMA for sense buffers if not required
  [SCSI] use dynamically allocated sense buffer
  [SCSI] scsi.h: add macro for enclosure bit of inquiry data
  [SCSI] sd: add fix for devices with last sector access problems
  [SCSI] fix pcmcia compile problem
  [SCSI] aacraid: add Voodoo Lite class of cards.
  [SCSI] aacraid: add new driver features flags
  [SCSI] qla2xxx: Update version number to 8.02.00-k7.
  [SCSI] qla2xxx: Issue correct MBC_INITIALIZE_FIRMWARE command.
  ...

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Sat, 26 Jan 2008 01:15:23 +0000 (17:15 -0800)]
Merge git://git./linux/kernel/git/herbert/crypto-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  [CRYPTO] hifn795x: Disallow built-in hifn795x when HW_RANDOM is m

16 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfashe...
Linus Torvalds [Sat, 26 Jan 2008 01:11:13 +0000 (17:11 -0800)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/mfasheh/ocfs2

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2: (31 commits)
  ocfs2: clean up bh null checks
  ocfs2: document access rules for blocked_lock_list
  configfs: file.c fix possible recursive locking
  configfs: dir.c fix possible recursive locking
  configfs: Remove EXPERIMENTAL
  ocfs2: bump version number
  ocfs2/dlm: Clear joining_node on hearbeat node down
  ocfs2: convert byte order of constant instead of variable
  ocfs2: Update default cluster timeouts
  ocfs2: printf fixes
  ocfs2: Use generic_file_llseek
  ocfs2: Safer read_inline_data()
  ocfs2: Silence false lockdep warnings
  [PATCH 2/2] ocfs2: cluster aware flock()
  [PATCH 1/2] ocfs2: add flock lock type
  ocfs2: Local alloc window size changeable via mount option
  ocfs2: Support commit= mount option
  ocfs2: Add missing permission checks
  [PATCH 2/2] ocfs2: Implement group add for online resize
  [PATCH 1/2] ocfs2: Add group extend for online resize
  ...

16 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Sat, 26 Jan 2008 01:08:28 +0000 (17:08 -0800)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev: (67 commits)
  fix drivers/ata/sata_fsl.c double-decl
  [libata] Prefer SCSI_SENSE_BUFFERSIZE to sizeof()
  pata_legacy: Merge winbond support
  ata_generic: Cenatek support
  pata_winbond: error return
  pata_serverworks: Fix cable types and cosmetics
  pata_mpc52xx: remove un-needed assignment
  libata: fix off-by-one in error categorization
  ahci: factor out AHCI enabling and enable AHCI before reading CAP
  ata_piix: implement SIDPR SCR access
  ata_piix: convert to prepare - activate initialization
  libata: factor out ata_pci_activate_sff_host() from ata_pci_one()
  [libata] Prefer SCSI_SENSE_BUFFERSIZE to sizeof()
  pata_legacy: resychronize with upstream changes and resubmit
  [libata] pata_legacy: typo fix
  [libata] pata_winbond: update for new ->data_xfer hook
  pata_pcmcia: convert to new data_xfer prototype
  libata annotations and fixes
  libata: use dev_driver_string() instead of "libata" in libata-sff.c
  ata_piix: kill unused constants and flags
  ...

16 years agoocfs2: clean up bh null checks
Mark Fasheh [Thu, 24 Jan 2008 02:35:31 +0000 (18:35 -0800)]
ocfs2: clean up bh null checks

If we know a buffer_head is non-null, then brelse() is unnecessary and
put_bh() can be used instead. Also, an explicit check for NULL is
unnecessary when using brelse(). This patch only covers buffer_head_io.c and
resize.c, which have recently added code which exhibits this problem.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: document access rules for blocked_lock_list
Mark Fasheh [Thu, 24 Jan 2008 00:54:48 +0000 (16:54 -0800)]
ocfs2: document access rules for blocked_lock_list

ocfs2_super->blocked_lock_list and ocfs2_super->blocked_lock_count have some
usage restrictions which aren't immediately obvious to anyone reading the
code. It's a good idea to document this so that we avoid making costly
mistakes in the future.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoconfigfs: file.c fix possible recursive locking
Joonwoo Park [Wed, 26 Dec 2007 03:09:57 +0000 (12:09 +0900)]
configfs: file.c fix possible recursive locking

configfs_register_subsystem() with default_groups triggers recursive locking.
it seems that mutex_lock_nested is needed.

=============================================
[ INFO: possible recursive locking detected ]
2.6.24-rc6 #145
---------------------------------------------
swapper/1 is trying to acquire lock:
 (&sb->s_type->i_mutex_key#3){--..}, at: [<c40c9a9e>] configfs_add_file+0x2e/0x70

but task is already holding lock:
 (&sb->s_type->i_mutex_key#3){--..}, at: [<c40ca985>] configfs_register_subsystem+0x55/0x130

other info that might help us debug this:
1 lock held by swapper/1:
 #0:  (&sb->s_type->i_mutex_key#3){--..}, at: [<c40ca985>] configfs_register_subsystem+0x55/0x130

stack backtrace:
Pid: 1, comm: swapper Not tainted 2.6.24-rc6 #145
 [<c40053ba>] show_trace_log_lvl+0x1a/0x30
 [<c4005e82>] show_trace+0x12/0x20
 [<c400687e>] dump_stack+0x6e/0x80
 [<c404ec72>] __lock_acquire+0xe62/0x1120
 [<c404efb2>] lock_acquire+0x82/0xa0
 [<c43fda88>] mutex_lock_nested+0x98/0x2e0
 [<c40c9a9e>] configfs_add_file+0x2e/0x70
 [<c40c9b0c>] configfs_create_file+0x2c/0x40
 [<c40ca639>] configfs_attach_item+0x139/0x220
 [<c40ca734>] configfs_attach_group+0x14/0x140
 [<c40ca7e9>] configfs_attach_group+0xc9/0x140
 [<c40ca9f6>] configfs_register_subsystem+0xc6/0x130
 [<c45c8186>] init_netconsole+0x2b6/0x300
 [<c45a75f2>] kernel_init+0x142/0x320
 [<c4004fb3>] kernel_thread_helper+0x7/0x14
 =======================

Signed-off-by: Joonwoo Park <joonwpark81@gmail.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoconfigfs: dir.c fix possible recursive locking
Joonwoo Park [Wed, 26 Dec 2007 03:09:57 +0000 (12:09 +0900)]
configfs: dir.c fix possible recursive locking

configfs_register_subsystem() with default_groups triggers recursive locking.
it seems that mutex_lock_nested is needed.

=============================================
[ INFO: possible recursive locking detected ]
2.6.24-rc6 #141
---------------------------------------------
swapper/1 is trying to acquire lock:
 (&sb->s_type->i_mutex_key#3){--..}, at: [<c40ca76f>] configfs_attach_group+0x4f/0x190

but task is already holding lock:
 (&sb->s_type->i_mutex_key#3){--..}, at: [<c40ca9d5>] configfs_register_subsystem+0x55/0x130

other info that might help us debug this:
1 lock held by swapper/1:
 #0:  (&sb->s_type->i_mutex_key#3){--..}, at: [<c40ca9d5>] configfs_register_subsystem+0x55/0x130

stack backtrace:
Pid: 1, comm: swapper Not tainted 2.6.24-rc6 #141
 [<c40053ba>] show_trace_log_lvl+0x1a/0x30
 [<c4005e82>] show_trace+0x12/0x20
 [<c400687e>] dump_stack+0x6e/0x80
 [<c404ec72>] __lock_acquire+0xe62/0x1120
 [<c404efb2>] lock_acquire+0x82/0xa0
 [<c43fdad8>] mutex_lock_nested+0x98/0x2e0
 [<c40ca76f>] configfs_attach_group+0x4f/0x190
 [<c40caa46>] configfs_register_subsystem+0xc6/0x130
 [<c45c8186>] init_netconsole+0x2b6/0x300
 [<c45a75f2>] kernel_init+0x142/0x320
 [<c4004fb3>] kernel_thread_helper+0x7/0x14
 =======================

Signed-off-by: Joonwoo Park <joonwpark81@gmail.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoconfigfs: Remove EXPERIMENTAL
Joel Becker [Mon, 31 Dec 2007 21:56:47 +0000 (13:56 -0800)]
configfs: Remove EXPERIMENTAL

configfs has been alive and kicking for a while now.  It underpins some
non-EXPERIMENTAL subsystems, such as OCFS2's cluster stack.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: bump version number
Mark Fasheh [Tue, 6 Nov 2007 23:52:58 +0000 (15:52 -0800)]
ocfs2: bump version number

Bump the printed version to 1.5.0. This helps us quickly identify which
version of Ocfs2 a bug filer is running.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2/dlm: Clear joining_node on hearbeat node down
Tao Ma [Thu, 10 Jan 2008 07:20:55 +0000 (15:20 +0800)]
ocfs2/dlm: Clear joining_node on hearbeat node down

Currently the process of dlm join contains 2 steps: query join and assert join.
After query join, the joined node will set its joining_node. So if the joining
node happens to panic before the 2nd step, the joined node will fail to clear
its joining_node flag because that node isn't in the domain map. It at least
cause 2 problems.
1. All the new join request will fail. So no new node can mount the volume.
2. The joined node can't umount the volume since during the umount process it
   has to wait for the joining_node to be unknown. So the umount will be hanged.

The solution is to clear the joining_node before we check the domain map.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: convert byte order of constant instead of variable
Marcin Slusarz [Tue, 25 Dec 2007 14:52:59 +0000 (15:52 +0100)]
ocfs2: convert byte order of constant instead of variable

Convert byte order of constant instead of variable it will be done at
compile time vs run time. Remove unused le32_and_cpu.

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Update default cluster timeouts
Sunil Mushran [Wed, 7 Nov 2007 00:10:23 +0000 (16:10 -0800)]
ocfs2: Update default cluster timeouts

Lots of people are having trouble with the default timeouts, which are too
low. These new values are derived from an informal survey taken on
ocfs2-users, as well as data from bug reports. This should reduce the amount
of cluster disconnects and subsequent fencing seen during normal workloads.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: printf fixes
Jan Kara [Wed, 19 Dec 2007 14:25:42 +0000 (15:25 +0100)]
ocfs2: printf fixes

Explicitely convert loff_t to long long in printf. Just for sure...

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Use generic_file_llseek
Jan Kara [Wed, 19 Dec 2007 14:24:52 +0000 (15:24 +0100)]
ocfs2: Use generic_file_llseek

We should use generic_file_llseek() and not default_llseek() so that
s_maxbytes gets properly checked when seeking.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Safer read_inline_data()
Jan Kara [Wed, 19 Dec 2007 14:24:09 +0000 (15:24 +0100)]
ocfs2: Safer read_inline_data()

In ocfs2_read_inline_data() we should store file size in loff_t. Although
the file size should fit in 32 bits we cannot be sure in case filesystem is
corrupted.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Silence false lockdep warnings
Jan Kara [Thu, 10 Jan 2008 23:11:45 +0000 (00:11 +0100)]
ocfs2: Silence false lockdep warnings

Create separate lockdep lock classes for system file's i_mutexes. They are
used to guard allocations and similar things and thus rank differently
than i_mutex of a regular file or directory.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years ago[PATCH 2/2] ocfs2: cluster aware flock()
Mark Fasheh [Fri, 21 Dec 2007 00:49:04 +0000 (16:49 -0800)]
[PATCH 2/2] ocfs2: cluster aware flock()

Hook up ocfs2_flock(), using the new flock lock type in dlmglue.c. A new
mount option, "localflocks" is added so that users can revert to old
functionality as need be.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years ago[PATCH 1/2] ocfs2: add flock lock type
Mark Fasheh [Fri, 21 Dec 2007 00:43:10 +0000 (16:43 -0800)]
[PATCH 1/2] ocfs2: add flock lock type

This adds a new dlmglue lock type which is intended to back flock()
requests.

Since these locks are driven from userspace, usage rules are much more
liberal than the typical Ocfs2 internal cluster lock. As a result, we can't
make use of most dlmglue features - lock caching and lock level
optimizations in particular. Additionally, userspace is free to deadlock
itself, so we have to deal with that in the same way as the rest of the
kernel - by allowing a signal to abort a lock request.

In order to keep ocfs2_cluster_lock() complexity down, ocfs2_file_lock()
does it's own dlm coordination. We still use the same helper functions
though, so duplicated code is kept to a minimum.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Local alloc window size changeable via mount option
Sunil Mushran [Thu, 20 Dec 2007 22:58:11 +0000 (14:58 -0800)]
ocfs2: Local alloc window size changeable via mount option

Local alloc is a performance optimization in ocfs2 in which a node
takes a window of bits from the global bitmap and then uses that for
all small local allocations. This window size is fixed to 8MB currently.
This patch allows users to specify the window size in MB including
disabling it by passing in 0. If the number specified is too large,
the fs will use the default value of 8MB.

mount -o localalloc=X /dev/sdX /mntpoint

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Support commit= mount option
Mark Fasheh [Wed, 7 Nov 2007 22:40:36 +0000 (14:40 -0800)]
ocfs2: Support commit= mount option

Mostly taken from ext3. This allows the user to set the jbd commit interval,
in seconds. The default of 5 seconds stays the same, but now users can
easily increase the commit interval. Typically, this would be increased in
order to benefit performance at the expense of data-safety.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Add missing permission checks
Mark Fasheh [Wed, 19 Dec 2007 02:58:18 +0000 (18:58 -0800)]
ocfs2: Add missing permission checks

Check that an online resize is being driven by a user with permission to
change system resource limits.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years ago[PATCH 2/2] ocfs2: Implement group add for online resize
Tao Ma [Tue, 18 Dec 2007 07:47:25 +0000 (15:47 +0800)]
[PATCH 2/2] ocfs2: Implement group add for online resize

This patch adds the ability for a userspace program to request that a
properly formatted cluster group be added to the main allocation bitmap for
an Ocfs2 file system. The request is made via an ioctl, OCFS2_IOC_GROUP_ADD.
On a high level, this is similar to ext3, but we use a different ioctl as
the structure which has to be passed through is different.

During an online resize, tunefs.ocfs2 will format any new cluster groups
which must be added to complete the resize, and call OCFS2_IOC_GROUP_ADD on
each one. Kernel verifies that the core cluster group information is valid
and then does the work of linking it into the global allocation bitmap.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years ago[PATCH 1/2] ocfs2: Add group extend for online resize
Tao Ma [Tue, 18 Dec 2007 07:47:03 +0000 (15:47 +0800)]
[PATCH 1/2] ocfs2: Add group extend for online resize

This patch adds the ability for a userspace program to request an extend of
last cluster group on an Ocfs2 file system. The request is made via ioctl,
OCFS2_IOC_GROUP_EXTEND. This is derived from EXT3_IOC_GROUP_EXTEND, but is
obviously Ocfs2 specific.

tunefs.ocfs2 would call this for an online-resize operation if the last
cluster group isn't full.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years ago[SCSI] usbstorage: use last_sector_bug flag universally
Hans de Goede [Sun, 20 Jan 2008 10:27:29 +0000 (11:27 +0100)]
[SCSI] usbstorage: use last_sector_bug flag universally

This patch sets the last_sector_bug flag to 1 for all USB disks. This is
needed to makes the cardreader on various HP multifunction printers work.

Since the performance impact is negible we set this flag for all USB disks to
avoid an unusual_devs.h nightmare.

Signed-off-by: Hans de Goede <j.w.r.degoede@hhs.nl>
Acked-by: Matthew Dharm <mdharm-usb@one-eyed-alien.net>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years agoocfs2: Reserve ioctl range
Tao Ma [Tue, 18 Dec 2007 07:46:37 +0000 (15:46 +0800)]
ocfs2: Reserve ioctl range

We need to reserve at least two ioctls for online-resize. Reserve a small
range of ioctls for Ocfs2 use in Documentation/ioctl-number.txt. This should
give us enough room for future growth.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Initalize bitmap_cpg of ocfs2_super to be the maximum.
Tao Ma [Tue, 18 Dec 2007 07:46:10 +0000 (15:46 +0800)]
ocfs2: Initalize bitmap_cpg of ocfs2_super to be the maximum.

This value is initialized from global_bitmap->id2.i_chain.cl_cpg. If there
is only 1 group, it will be equal to the total clusters in the volume. So
as for online resize, it should change for all the nodes in the cluster.
It isn't easy and there is no corresponding lock for it.

bitmap_cpg is only used in 2 areas:
1. Check whether the suballoc is too large for us to allocate from the global
   bitmap, so it is little used. And now the suballoc size is 2048, it rarely
   meet this situation and the check is almost useless.
2. Calculate which group a cluster belongs to. We use it during truncate to
   figure out which cluster group an extent belongs too. But we should be OK
   if we increase it though as the cluster group calculated shouldn't change
   and we only ever have a small bitmap_cpg on file systems with a single
   cluster group.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years ago[CRYPTO] hifn795x: Disallow built-in hifn795x when HW_RANDOM is m
Herbert Xu [Fri, 25 Jan 2008 22:48:44 +0000 (09:48 +1100)]
[CRYPTO] hifn795x: Disallow built-in hifn795x when HW_RANDOM is m

Currently it is possible to select HW_RANDOM as a module and have
hifn795x built-in.  This causes a build problem because hifn795x
will then call hwrng_register which isn't built-in.

This patch introduces a new config option to control the hifn795x
RNG which lets us avoid this problem.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
16 years agoocfs2: Documentation update
Mark Fasheh [Tue, 30 Oct 2007 19:09:03 +0000 (12:09 -0700)]
ocfs2: Documentation update

Remove 'readpages' from the list in ocfs2.txt. Instead of having two
identical lists, I just removed the list in the OCFS2 section of fs/Kconfig
and added a pointer to Documentation/filesystems/ocfs2.txt.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Readpages support
Mark Fasheh [Tue, 30 Oct 2007 19:08:32 +0000 (12:08 -0700)]
ocfs2: Readpages support

Add ->readpages support to Ocfs2. This is rather trivial - all it required
is a small update to ocfs2_get_block (for mapping full extents via b_size)
and an ocfs2_readpages() function which partially mirrors ocfs2_readpage().

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agodlm: Split lock mode and flag constants into a sharable header.
Joel Becker [Fri, 5 Oct 2007 21:31:44 +0000 (14:31 -0700)]
dlm: Split lock mode and flag constants into a sharable header.

This allows others to use the DLM constants without being tied to the
function API of fs/dlm.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Rename ocfs2_meta_[un]lock
Mark Fasheh [Thu, 18 Oct 2007 22:30:42 +0000 (15:30 -0700)]
ocfs2: Rename ocfs2_meta_[un]lock

Call this the "inode_lock" now, since it covers both data and meta data.
This patch makes no functional changes.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Remove data locks
Mark Fasheh [Thu, 18 Oct 2007 22:23:46 +0000 (15:23 -0700)]
ocfs2: Remove data locks

The meta lock now covers both meta data and data, so this just removes the
now-redundant data lock.

Combining locks saves us a round of lock mastery per inode and one less lock
to ping between nodes during read/write.

We don't lose much - since meta locks were always held before a data lock
(and at the same level) ordered writeout mode (the default) ensured that
flushing for the meta data lock also pushed out data anyways.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Add data downconvert worker to inode lock
Mark Fasheh [Thu, 18 Oct 2007 22:13:59 +0000 (15:13 -0700)]
ocfs2: Add data downconvert worker to inode lock

In order to extend inode lock coverage to inode data, we use the same data
downconvert worker with only a small modification to only do work for
regular files.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2: Remove mount/unmount votes
Mark Fasheh [Mon, 24 Sep 2007 22:56:19 +0000 (15:56 -0700)]
ocfs2: Remove mount/unmount votes

The node maps that are set/unset by these votes are no longer relevant, thus
we can remove the mount and umount votes. Since those are the last two
remaining votes, we can also remove the entire vote infrastructure.

The vote thread has been renamed to the downconvert thread, and the small
amount of functionality related to managing it has been moved into
fs/ocfs2/dlmglue.c. All references to votes have been removed or updated.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Fri, 25 Jan 2008 22:41:24 +0000 (14:41 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (81 commits)
  RDMA/cxgb3: Fix the T3A workaround checks
  IB/ipath: Remove unnecessary cast
  IPoIB: Constify seq_operations function pointer tables
  RDMA/cxgb3: Mark QP as privileged based on user capabilities
  RDMA/cxgb3: Fix page shift calculation in build_phys_page_list()
  RDMA/cxgb3: Flush the receive queue when closing
  IB/ipath: Trivial simplification of ipath_make_ud_req()
  IB/mthca: Update latest "native Arbel" firmware revision
  IPoIB: Remove redundant check of netif_queue_stopped() in xmit handler
  IB/ipath: Add mappings from HW register to PortInfo port physical state
  IB/ipath: Changes to support PIO bandwidth check on IBA7220
  IB/ipath: Minor cleanup of unused fields and chip-specific errors
  IB/ipath: New sysfs entries to control 7220 features
  IB/ipath: Add new chip-specific functions to older chips, consistent init
  IB/ipath: Remove unused MDIO interface code
  IB/ehca: Prevent RDMA-related connection failures on some eHCA2 hardware
  IB/ehca: Add "port connection autodetect mode"
  IB/ehca: Define array to store SMI/GSI QPs
  IB/ehca: Remove CQ-QP-link before destroying QP in error path of create_qp()
  IB/iser: Add change_queue_depth method
  ...

16 years agoocfs2: Remove fs dependency on ocfs2_heartbeat module
Mark Fasheh [Mon, 24 Sep 2007 22:09:41 +0000 (15:09 -0700)]
ocfs2: Remove fs dependency on ocfs2_heartbeat module

Now that the dlm exposes domain information to us, we don't need generic
node up / node down callbacks. And since the DLM is only telling us when a
node goes down unexpectedly, we no longer need to optimize away node down
callbacks via the umount map.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoocfs2_dlm: Call node eviction callbacks from heartbeat handler
Mark Fasheh [Fri, 7 Sep 2007 18:11:10 +0000 (11:11 -0700)]
ocfs2_dlm: Call node eviction callbacks from heartbeat handler

With this, a dlm client can take advantage of the group protocol in the dlm
to get full notification whenever a node within the dlm domain leaves
unexpectedly.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
16 years agoRDMA/cxgb3: Fix the T3A workaround checks
Steve Wise [Thu, 24 Jan 2008 22:30:16 +0000 (16:30 -0600)]
RDMA/cxgb3: Fix the T3A workaround checks

Correctly work around T3A issues by checking "hwtype != T3A" instead of
"hwtype == T3B".  This will be needed for new hardware types.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ipath: Remove unnecessary cast
Jan Engelhardt [Tue, 22 Jan 2008 19:45:30 +0000 (20:45 +0100)]
IB/ipath: Remove unnecessary cast

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIPoIB: Constify seq_operations function pointer tables
Jan Engelhardt [Tue, 22 Jan 2008 19:45:30 +0000 (20:45 +0100)]
IPoIB: Constify seq_operations function pointer tables

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoRDMA/cxgb3: Mark QP as privileged based on user capabilities
Steve Wise [Mon, 21 Jan 2008 20:42:13 +0000 (14:42 -0600)]
RDMA/cxgb3: Mark QP as privileged based on user capabilities

This is needed to support zero-stag properly.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoRDMA/cxgb3: Fix page shift calculation in build_phys_page_list()
Steve Wise [Mon, 21 Jan 2008 20:42:11 +0000 (14:42 -0600)]
RDMA/cxgb3: Fix page shift calculation in build_phys_page_list()

The existing logic incorrectly maps this buffer list:

    0: addr 0x10001000, size 0x1000
    1: addr 0x10002000, size 0x1000

To this bogus page list:

    0: 0x10000000
    1: 0x10002000

The shift calculation must also take into account the address of the
first entry masked by the page_mask as well as the last address+size
rounded up to the next page size.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoRDMA/cxgb3: Flush the receive queue when closing
Steve Wise [Mon, 21 Jan 2008 20:42:09 +0000 (14:42 -0600)]
RDMA/cxgb3: Flush the receive queue when closing

- for kernel mode cqs, call event notification handler when flushing.
- flush QP when moving from RTS -> CLOSING.
- fix logic to identify a kernel mode qp.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ipath: Trivial simplification of ipath_make_ud_req()
Ralph Campbell [Fri, 25 Jan 2008 22:17:44 +0000 (14:17 -0800)]
IB/ipath: Trivial simplification of ipath_make_ud_req()

Move the increment of s_hdrwords into the existing if block that tests
if we're doing a send with immediate, to save one test of the opcode.

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/mthca: Update latest "native Arbel" firmware revision
Roland Dreier [Fri, 25 Jan 2008 22:17:44 +0000 (14:17 -0800)]
IB/mthca: Update latest "native Arbel" firmware revision

Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIPoIB: Remove redundant check of netif_queue_stopped() in xmit handler
Krishna Kumar [Thu, 15 Nov 2007 05:05:37 +0000 (10:35 +0530)]
IPoIB: Remove redundant check of netif_queue_stopped() in xmit handler

qdisc_run() now tests for queue_stopped() before calling
__qdisc_run(), and the same check is done in every iteration of
__qdisc_run(), so another check is not required in the driver xmit.
This means that ipoib_start_xmit() no longer needs to test
netif_queue_stopped(); the test was added to fix earlier kernels,
where the networking stack did not guarantee that the xmit method of
an LLTX driver would not be called after the queue was stopped, but
current kernels do provide this guarantee.

To validate, I put a debug in the TX_BUSY path which never hit with 64
threads running overnight exercising this code a few 100 million
times.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ipath: Add mappings from HW register to PortInfo port physical state
Ralph Campbell [Thu, 10 Jan 2008 08:50:41 +0000 (00:50 -0800)]
IB/ipath: Add mappings from HW register to PortInfo port physical state

Add new mappings from port physical state (a HW register value) to the
IB SubnGet(PortInfo) port physical state.

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ipath: Changes to support PIO bandwidth check on IBA7220
Dave Olson [Thu, 9 Aug 2007 10:11:38 +0000 (03:11 -0700)]
IB/ipath: Changes to support PIO bandwidth check on IBA7220

The IBA7220 uses a count-based triggering mechanism, and therefore
can't use the same bandwidth verification mechanism as older chips.

To support the 7220, allow enabling and disabling armlaunch errors on
application request.  Minor robustness improvements as well.

Signed-off-by: Dave Olson <dave.olson@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ipath: Minor cleanup of unused fields and chip-specific errors
Dave Olson [Tue, 8 Jan 2008 09:51:46 +0000 (01:51 -0800)]
IB/ipath: Minor cleanup of unused fields and chip-specific errors

Clean up some unused header fields, minor related cleanup.

Signed-off-by: Dave Olson <dave.olson@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ipath: New sysfs entries to control 7220 features
Michael Albaugh [Tue, 8 Jan 2008 08:37:34 +0000 (00:37 -0800)]
IB/ipath: New sysfs entries to control 7220 features

IBA7220 includes many more configurable IB settings. Getting/setting
these is now grouped into a pair of chip specific functions accessed via
function pointers.  Provide sysfs access to these settings.

Signed-off-by: Michael Albaugh <michael.albaugh@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ipath: Add new chip-specific functions to older chips, consistent init
Dave Olson [Tue, 8 Jan 2008 10:36:46 +0000 (02:36 -0800)]
IB/ipath: Add new chip-specific functions to older chips, consistent init

This adds the new (sometimes empty) chip-specific functions to the older
chips, and makes the initialization and related functions consistent across
all 3 chips.

Signed-off-by: Dave Olson <dave.olson@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ipath: Remove unused MDIO interface code
Dave Olson [Wed, 9 Jan 2008 07:16:17 +0000 (23:16 -0800)]
IB/ipath: Remove unused MDIO interface code

This code has been unused for some time, but still had leftovers
from when it was used.

Signed-off-by: Dave Olson <dave.olson@qlogic.com
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ehca: Prevent RDMA-related connection failures on some eHCA2 hardware
Joachim Fenkes [Thu, 17 Jan 2008 14:07:24 +0000 (15:07 +0100)]
IB/ehca: Prevent RDMA-related connection failures on some eHCA2 hardware

Some HW revisions of eHCA2 may cause an RC connection to break if they
received RDMA Reads over that connection before.  This can be
prevented by assuring that, after the first RDMA Read, the QP receives
a new RDMA Read every few million link packets.

Include code into the driver that inserts an empty (size 0) RDMA Read
into the message stream every now and then if the consumer doesn't
post them frequently enough.

Signed-off-by: Joachim Fenkes <fenkes@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ehca: Add "port connection autodetect mode"
Hoang-Nam Nguyen [Thu, 17 Jan 2008 14:05:45 +0000 (15:05 +0100)]
IB/ehca: Add "port connection autodetect mode"

This patch enhances ehca with a capability to "autodetect" the ports
being connected physically. In order to utilize that function the
module option nr_ports must be set to -1 (default is 2 - two
ports). This feature is experimental and will made the default later.

More detail:

If the user connects only one port to the switch, current code requires
  1) port one to be connected and
  2) module option nr_ports=1 to be given.

If autodetect is enabled, ehca will not wait at creation of the GSI QP
for the respective port to become active. Since firmware does not
accept modify_qp() while the port is down at initialization, we need
to cache all calls to modify_qp() for the SMI/GSI QP and just return a
good return code.

When a port is activated and we get a PORT_ACTIVE event, we replay the
cached modify-qp() parms and re-trigger any posted recv WRs. Only then
do we forward the PORT_ACTIVE event to registered clients.

The result of this autodetect patch is that all ports will be
accessible by the users. Depending on their respective cabling only
those ports that are connected properly will become operable. If a
user tries to modify a regular QP of a non-connected port, modify_qp()
will fail. Furthermore, ibv_devinfo should show the port state
accordingly.

Note that this patch primarily improves the loading behaviour of
ehca. If the cable is removed while the driver is operating and
plugged in again, firmware will handle that properly by sending an
appropriate async event.

Signed-off-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ehca: Define array to store SMI/GSI QPs
Hoang-Nam Nguyen [Thu, 17 Jan 2008 14:04:32 +0000 (15:04 +0100)]
IB/ehca: Define array to store SMI/GSI QPs

Signed-off-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ehca: Remove CQ-QP-link before destroying QP in error path of create_qp()
Hoang-Nam Nguyen [Thu, 17 Jan 2008 14:03:55 +0000 (15:03 +0100)]
IB/ehca: Remove CQ-QP-link before destroying QP in error path of create_qp()

Signed-off-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/iser: Add change_queue_depth method
Erez Zilber [Thu, 17 Jan 2008 09:53:17 +0000 (11:53 +0200)]
IB/iser: Add change_queue_depth method

Add a .change_queue_depth handler to the scsi_host_template in the
iSER driver.  iscsi_change_queue_depth was added to iscsi_tcp in order
to solve the problem of queue depth which was too high for some
targets.  It is also applicable for iSER.

Signed-off-by: Erez Zilber <erezz@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/iser: Print information about unhandled RDMA CM events
Erez Zilber [Thu, 17 Jan 2008 09:51:58 +0000 (11:51 +0200)]
IB/iser: Print information about unhandled RDMA CM events

Some RDMA CM events are not supported or not handled in iSER.
This patch adds some info (printk) for the user about them.

Signed-off-by: Erez Zilber <erezz@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/fmr_pool: ib_fmr_pool_flush() should flush all dirty FMRs
Olaf Kirch [Wed, 16 Jan 2008 17:37:16 +0000 (18:37 +0100)]
IB/fmr_pool: ib_fmr_pool_flush() should flush all dirty FMRs

When a FMR is released via ib_fmr_pool_unmap(), the FMR usually ends
up on the free_list rather than the dirty_list (because we allow a
certain number of remappings before actually requiring a flush).

However, ib_fmr_batch_release() only looks at dirty_list when flushing
out old mappings.  This means that when ib_fmr_pool_flush() is used to
force a flush of the FMR pool, some dirty FMRs that have not reached
their maximum remap count will not actually be flushed.

Fix this by flushing all FMRs that have been used at least once in
ib_fmr_batch_release().

Signed-off-by: Olaf Kirch <olaf.kirch@oracle.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/fmr_pool: Flush serial numbers can get out of sync
Olaf Kirch [Wed, 16 Jan 2008 17:36:27 +0000 (18:36 +0100)]
IB/fmr_pool: Flush serial numbers can get out of sync

Normally, the serial numbers for flush requests and flushes executed
for an FMR pool should be in sync.

However, if the FMR pool flushes dirty FMRs because the
dirty_watermark was reached, we wake up the cleanup thread and let it
do its stuff.  As a side effect, the cleanup thread increments
pool->flush_ser, which leaves it one higher than pool->req_ser.  The
next time the user calls ib_flush_fmr_pool(), the cleanup thread will
be woken up, but ib_flush_fmr_pool() won't wait for the flush to
complete because flush_ser is already past req_ser.  This means the
FMRs that the user expects to be flushed may not have all been flushed
when the function returns.

Fix this by telling the cleanup thread to do work exclusively by
incrementing req_ser, and by moving the comparison of dirty_len and
dirty_watermark into ib_fmr_pool_unmap().

Signed-off-by: Olaf Kirch <olaf.kirch@oracle.com>
16 years agoIB/umad: Simplify and fix locking
Roland Dreier [Fri, 25 Jan 2008 22:15:42 +0000 (14:15 -0800)]
IB/umad: Simplify and fix locking

In addition to being overly complex, the locking in user_mad.c is
broken: there were multiple reports of deadlocks and lockdep warnings.
In particular it seems that a single thread may end up trying to take
the same rwsem for reading more than once, which is explicitly
forbidden in the comments in <linux/rwsem.h>.

To solve this, we change the locking to use plain mutexes instead of
rwsems.  There is one mutex per open file, which protects the contents
of the struct ib_umad_file, including the array of agents and list of
queued packets; and there is one mutex per struct ib_umad_port, which
protects the contents, including the list of open files.  We never
hold the file mutex across calls to functions like ib_unregister_mad_agent(),
which can call back into other ib_umad code to queue a packet, and we
always hold the port mutex as long as we need to make sure that a
device is not hot-unplugged from under us.

This even makes things nicer for users of the -rt patch, since we
remove calls to downgrade_write() (which is not implemented in -rt).

Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ipath: Fix some sparse warnings about shadowed symbols
Roland Dreier [Fri, 25 Jan 2008 22:15:42 +0000 (14:15 -0800)]
IB/ipath: Fix some sparse warnings about shadowed symbols

There are a few places in the ipath driver where a variable is
re-declared within a block where it is already in scope.  Most of these
extra declarations can simply be removed, since the variable from the
outer scope is used in a way so that it does not need to keep its
variable across the block with the re-declaration.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoRDMA/cxgb3: Endianness annotation for irs field
Roland Dreier [Fri, 25 Jan 2008 22:15:42 +0000 (14:15 -0800)]
RDMA/cxgb3: Endianness annotation for irs field

t3_rdma_init_wr.irs is a big-endian field, so declare it as __be32.
This fixes one sparse warning.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoIB/ehca: Use round_jiffies() for EQ polling timer
Anton Blanchard [Mon, 15 Oct 2007 05:49:07 +0000 (00:49 -0500)]
IB/ehca: Use round_jiffies() for EQ polling timer

Use round_jiffies() to align ehca's 1-second timer with other timers
and potentially save power by sleeping cores for longer.

Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
16 years agoRDMA/cma: Override default responder_resources with user value
Sean Hefty [Fri, 4 Jan 2008 18:47:12 +0000 (10:47 -0800)]
RDMA/cma: Override default responder_resources with user value

By default, the responder_resources parameter is set to that received
in a connection request.  The passive side may override this value
when accepting the connection.  Use the value provided by the passive
side when transitioning the QP to RTR state, rather than the value
given in the connect request.  Without this change, the RTR transition
may fail if the passive side supports fewer responder_resources than
that in the request.

For code consistency and to protect against QP destruction, restructure
overriding initiator_depth to match how responder_resources is set.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>