]> err.no Git - linux-2.6/log
linux-2.6
17 years ago[SCSI] qla2xxx: Add MSI-X support.
Andrew Vasquez [Mon, 29 Jan 2007 18:22:19 +0000 (10:22 -0800)]
[SCSI] qla2xxx: Add MSI-X support.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla2xxx: Correct sector-erase issues while writing flash.
Andrew Vasquez [Mon, 29 Jan 2007 18:22:18 +0000 (10:22 -0800)]
[SCSI] qla2xxx: Correct sector-erase issues while writing flash.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: fix task attribute
FUJITA Tomonori [Sat, 27 Jan 2007 16:19:41 +0000 (01:19 +0900)]
[SCSI] libsas: fix task attribute

Why TASK_ATTR_HOQ?

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: remove duplicate device id
Greg Kroah-Hartman [Fri, 26 Jan 2007 02:15:03 +0000 (18:15 -0800)]
[SCSI] ipr: remove duplicate device id

This patch removes a duplicate device id from the IPR driver.  Based on
the ipr.h file, I'm not so sure this was intended to be a duplicate, and
if so, the .h file should be modified to use the proper sub-device id
instead.

This was pointed out to me by Kay Sievers <kay.sievers@suse.de>

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Enable automatic spin-up of SAS disks
Darrick J. Wong [Fri, 26 Jan 2007 22:08:58 +0000 (14:08 -0800)]
[SCSI] libsas: Enable automatic spin-up of SAS disks

Set allow_restart=1 for all SAS disks so that they are spun up when needed.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aic94xx: Register eh_device_reset_handler
Darrick J. Wong [Fri, 26 Jan 2007 22:08:55 +0000 (14:08 -0800)]
[SCSI] aic94xx: Register eh_device_reset_handler

Register libsas's default device reset code with the scsi.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Handle SCSI commands that complete with failure codes
Darrick J. Wong [Fri, 26 Jan 2007 22:08:52 +0000 (14:08 -0800)]
[SCSI] libsas: Handle SCSI commands that complete with failure codes

This patch moves the code that handles SAS failures out of the main EH
function and into a separate function.  It also detects commands that have
no sas_task (i.e. they completed, but with error data) and sends them into
scsi_error for processing.  This allows us to handle SCSI errors (and
enables auto-spinup as a side effect) instead of dropping them on the
floor and falling into an infinite loop.  It also requires the
implementation of a device reset function, which the SAS failure code has
been modified to employ for REQ_DEVICE_RESET.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] scsi_error.c: Export some scsi_eh_* functions
Darrick J. Wong [Fri, 26 Jan 2007 22:08:49 +0000 (14:08 -0800)]
[SCSI] scsi_error.c: Export some scsi_eh_* functions

Export a couple of functions from scsi_error that are needed to handle
failed SCSI commands from the SAS EH.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
make exports GPL and
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Check return values of sysfs_create_link
Darrick J. Wong [Fri, 26 Jan 2007 22:08:46 +0000 (14:08 -0800)]
[SCSI] libsas: Check return values of sysfs_create_link

Get rid of: "warning: ignoring return value of sysfs_create_link..."

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Clean up discovery failure handler code
Darrick J. Wong [Fri, 26 Jan 2007 22:08:43 +0000 (14:08 -0800)]
[SCSI] libsas: Clean up discovery failure handler code

sas_rphy_delete does two things: it removes the sas_rphy from the transport
layer and frees the sas_rphy.  This can be broken down into two functions,
sas_rphy_remove and sas_rphy_free; sas_rphy_remove is of interest to
sas_discover_root_expander because it calls functions that require
sas_rphy_add as a prerequisite and can fail (namely sas_discover_expander).
In that case, sas_discover_root_expander needs to be able to undo the effects
of sas_rphy_add yet leave the job of freeing the sas_rphy to the caller of
sas_discover_root_expander.

This patch also removes some unnecessary code from sas_discover_end_dev
to eliminate an unnecessary cycle of sas_notify_lldd_gone/found for SAS
devices, thus eliminating a sas_rphy_remove call (and fixing a race condition
where a SCSI target scan can come in between the gone and found call).
It also moves the sas_rphy_free calls into sas_discover_domain and
sas_ex_discover_end_dev to complement the sas_rphy_allocation via
sas_get_port_device.

This patch does not change the semantics of sas_rphy_delete.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Fix incorrect sas_port deformation in sas_form_port
Darrick J. Wong [Fri, 26 Jan 2007 22:08:41 +0000 (14:08 -0800)]
[SCSI] libsas: Fix incorrect sas_port deformation in sas_form_port

Currently, sas_form_port checks the given asd_sas_phy's sas_phy to see if
there's already a port attached.  If so, the SAS addresses of the port and
the phy are compared to determine if we need to detach from the port
because the addresses don't match or if we can stop; the SAS address stored
in the sas_port reflects whatever device _was_ attached to the port/phy, and
the SAS address stored in the sas_port reflects whatever device we just
discovered.  As written, the code detaches from the port if the addresses
_do_ match, and prints an error if they do _not_ match.  I believe this to
be incorrect, as it seems more logical to keep the port if the addresses
match (i.e. the phy was reset but the device didn't change), and detach it
they do not (i.e. the device changed).

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] NCR_D700: fix compile error
Mark Haverkamp [Wed, 24 Jan 2007 17:31:30 +0000 (09:31 -0800)]
[SCSI] NCR_D700: fix compile error

Fix a typo in NCR_D700

Signed-off-by Mark Haverkamp <markh@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aacraid: expanded expose physical device code (new)
Mark Haverkamp [Fri, 26 Jan 2007 17:23:32 +0000 (09:23 -0800)]
[SCSI] aacraid: expanded expose physical device code (new)

Received from Mark Salyzyn,

Take the expose_physicals flag and allow the user to select default (physicals
available via /dev/sg), exposed (physicals available via /dev/sd for
experimental reasons) and hidden (physicals blocked from all access). This
expands the functionality of the previous expose_physicals insmod parameter
which was added to support some experimental configurations.

Signed-off-by Mark Haverkamp <markh@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aacraid: rework packet support code
Mark Haverkamp [Tue, 23 Jan 2007 23:00:30 +0000 (15:00 -0800)]
[SCSI] aacraid: rework packet support code

Received from Mark Salyzyn,

Replace all if/else packet formations with platform function calls. This is in
recognition of the proliferation of read and write packet types, and in the
need to migrate to up-and-coming packets for new products.

Signed-off-by Mark Haverkamp <markh@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aacraid: Begin adding support for new adapter type
Mark Haverkamp [Tue, 23 Jan 2007 23:00:13 +0000 (15:00 -0800)]
[SCSI] aacraid: Begin adding support for new adapter type

Received from Mark Salyzyn,

Add in the NEMER/ARK physical register mapping, represented in up and coming
products currently under test at Adaptec.

Signed-off-by Mark Haverkamp <markh@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aacraid: rework communication support code
Mark Haverkamp [Tue, 23 Jan 2007 22:59:20 +0000 (14:59 -0800)]
[SCSI] aacraid: rework communication support code

Received from Mark Salyzyn,

Replace all if/else communication transports with a platform function call.
This is in recognition of the need to migrate to up-and-coming transports.
Currently the Linux driver does not support two available communication
transports provided by our products, these will be added in future patches, and
will expand the platform function set.

Signed-off-by Mark Haverkamp <markh@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Driver version 2.3.1
Brian King [Tue, 23 Jan 2007 17:25:43 +0000 (11:25 -0600)]
[SCSI] ipr: Driver version 2.3.1

Bump driver version to 2.3.1.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: PCI error recovery fix
Brian King [Tue, 23 Jan 2007 17:25:37 +0000 (11:25 -0600)]
[SCSI] ipr: PCI error recovery fix

Since the pci_block_user_cfg_access API was modified to track
block/unblocks, it was discovered that the ipr driver had a
path through its code (in PCI error recovery) which would unblock
when not previously blocked.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Tolerate not finding PCI-X registers
Brian King [Tue, 23 Jan 2007 17:25:30 +0000 (11:25 -0600)]
[SCSI] ipr: Tolerate not finding PCI-X registers

Don't fail initialization of an adapter if the PCI-X registers
cannot be found since it may be a PCI-E adapter.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Remove usage of pci driver data
Brian King [Tue, 23 Jan 2007 17:25:23 +0000 (11:25 -0600)]
[SCSI] ipr: Remove usage of pci driver data

Since ipr handles dynamic ids, it must handle driver_data
not being set, so remove the current usage of driver_data
so it can be used for other things in future patches.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] fusion: mpi header update - version 1.05.14
Eric Moore [Sat, 20 Jan 2007 01:46:17 +0000 (18:46 -0700)]
[SCSI] fusion: mpi header update - version 1.05.14

Here are the lastest mpi headers for mpt fusion driver, which defines
the firmware to driver interface.

Signed-off-by: Eric Moore <Eric.Moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aic94xx: fix typos and update verison number
Alexis Bruemmer [Tue, 16 Jan 2007 23:36:12 +0000 (15:36 -0800)]
[SCSI] aic94xx: fix typos and update verison number

fix typos and bump version number

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Acked-by: Alexis Bruemmer <alexisb@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Start I_T recovery if ABORT TASK fails
Darrick J. Wong [Thu, 11 Jan 2007 22:15:46 +0000 (14:15 -0800)]
[SCSI] libsas: Start I_T recovery if ABORT TASK fails

The EH should fall into I_T recovery (and potentially stronger
remedies) if ABORT TASK fails.

Signed-off-by: Alexis Bruemmer <alexisb@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Add SAS_HA state flags to avoid queueing events while unloading
Darrick J. Wong [Thu, 11 Jan 2007 22:15:43 +0000 (14:15 -0800)]
[SCSI] libsas: Add SAS_HA state flags to avoid queueing events while unloading

Track sas_ha_struct state so that we ignore events that come in while
we're shutting things down.

Signed-off-by: Malahal Naineni <malahal@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: phy port lock needs irq spinlocks
Darrick J. Wong [Thu, 11 Jan 2007 22:15:40 +0000 (14:15 -0800)]
[SCSI] libsas: phy port lock needs irq spinlocks

Convert the phy port locks to use irq spinlocks.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aic94xx: Scan SAS devices asynchronously
Darrick J. Wong [Thu, 11 Jan 2007 22:15:38 +0000 (14:15 -0800)]
[SCSI] aic94xx: Scan SAS devices asynchronously

Add the necessary hooks to the aic94xx driver to support the asynchronous SCSI
device scan infrastructure.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: sysfs phy control attributes should not be S_IWUGO
Darrick J. Wong [Thu, 11 Jan 2007 22:15:35 +0000 (14:15 -0800)]
[SCSI] libsas: sysfs phy control attributes should not be S_IWUGO

Allowing the phy reset controls to be world-triggerable does not seem like
a terribly good idea because SAS devices can be disrupted (and ATA devices
are really disrupted) by a phy reset.  By default only root should be able
to do things like that.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aic94xx: Lock DDB read/write accesses
Darrick J. Wong [Thu, 11 Jan 2007 22:15:32 +0000 (14:15 -0800)]
[SCSI] aic94xx: Lock DDB read/write accesses

Extend the use of the DDB lock to include all DDB accesses, because
DDB updates now occur from multiple threads.  This fixes the SMP timeout
problems that we were occasionally seeing with a x260, because the
controller got confused when the DDBs got corrupted.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aic94xx: Fix DDB and SCB initialization
Darrick J. Wong [Thu, 11 Jan 2007 22:15:29 +0000 (14:15 -0800)]
[SCSI] aic94xx: Fix DDB and SCB initialization

Ed Chim of Adaptec informs us that the DDB registers need to be zeroed at
initialization time and that some SCB initializations need to happen even if
we don't use the SCB.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aic94xx: Match request_firmware with release_firmware
Darrick J. Wong [Thu, 11 Jan 2007 22:15:26 +0000 (14:15 -0800)]
[SCSI] aic94xx: Match request_firmware with release_firmware

The vmalloc() blob holding the sequencer firmware wasn't being released at
module unload time, which resulted in a memory leak.

Signed-off-by: Alexis Bruemmer <alexisb@us.ibm.com>
Acked-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aic94xx: Remove workqueue code from REQ_TASK_ABORT/REQ_DEVICE_RESET code
Darrick J. Wong [Thu, 11 Jan 2007 22:15:23 +0000 (14:15 -0800)]
[SCSI] aic94xx: Remove workqueue code from REQ_TASK_ABORT/REQ_DEVICE_RESET code

Now that task aborts and device port resets are done by the EH, we can
remove all the code that set up workqueues and such and simply call
sas_task_abort and let libsas figure things out.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Delay issuing ABORT TASK TMF until the error handler
Darrick J. Wong [Thu, 11 Jan 2007 22:15:20 +0000 (14:15 -0800)]
[SCSI] libsas: Delay issuing ABORT TASK TMF until the error handler

sas_task_abort() should simply abort the upper-level SCSI command and wait
until the error handler to send the actual ABORT TASK command.  By
deferring things to the EH we simplify the concurrency coordination and
eliminate some race conditions.  Note that sas_task_abort has a few hooks
to handle libsas internal commands properly too.

Also rename do_sas_task_abort to __sas_task_abort just in case we really
want to abort the task *right now* and we don't have a scsi_cmnd attached
to the command.  This is a hook for libata internal commands to abort.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Enable the EH strategy handler to reset a phy after a command
Darrick J. Wong [Thu, 11 Jan 2007 22:15:17 +0000 (14:15 -0800)]
[SCSI] libsas: Enable the EH strategy handler to reset a phy after a command

When a SAS LLDD needs to request a device port reset, it needs to have all
commands aborted before it can reset the port.  Since commands are put on
the EH's list in the order that they were queued, the LLDD can set a "need
reset" flag in the last task to be aborted so that the EH can reset the
port after all commands are aborted.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Remove SAS_TASK_INITIATOR_ABORTED flag
Darrick J. Wong [Thu, 11 Jan 2007 22:15:15 +0000 (14:15 -0800)]
[SCSI] libsas: Remove SAS_TASK_INITIATOR_ABORTED flag

This flag is no longer necessary because we push tasks to be aborted into
the EH as soon as we possibly can, and let the SCSI EH code take care of
the coordination for which this flag was used.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aic94xx: Don't eat TMF_QUERY_TASK results
Darrick J. Wong [Thu, 11 Jan 2007 22:15:12 +0000 (14:15 -0800)]
[SCSI] aic94xx: Don't eat TMF_QUERY_TASK results

In this driver, TMF_QUERY_TASK translates to QUERY_SSP_TASK.  The
sequencer, it seems, is perfectly happy sending us a SSP response, which
this function promptly "converts" into TMF_RESP_FUNC_FAILED.  This leads to
the SAS EH making bad decisions based on bad data, so we should not perform
the conversion in this case.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Use SCAN_WILD_CARD instead of ~0
Darrick J. Wong [Thu, 11 Jan 2007 22:15:09 +0000 (14:15 -0800)]
[SCSI] libsas: Use SCAN_WILD_CARD instead of ~0

Magic number cleanup.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aic94xx: Set lldd_max_execute_num in sas_ha
Darrick J. Wong [Thu, 11 Jan 2007 22:15:06 +0000 (14:15 -0800)]
[SCSI] aic94xx: Set lldd_max_execute_num in sas_ha

The aic94xx module has a parameter that looks like it should set
lldd_max_execute_num in the sas_ha, but it never sets this value.  Either
we should set it or remove the parameter.  This allows us to enable task
collector mode for this driver, though it is still off by default.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Destroy the task collector thread after releasing ports
Darrick J. Wong [Thu, 11 Jan 2007 22:15:03 +0000 (14:15 -0800)]
[SCSI] libsas: Destroy the task collector thread after releasing ports

If we use task collector mode, we can end up destroying the task collector
thread before we release the ports, which is bad if a port release causes
a disk I/O (such as cache flushing).

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Reset timer on taskless scsi_cmnds in sas_scsi_timed_out
Darrick J. Wong [Thu, 11 Jan 2007 22:15:00 +0000 (14:15 -0800)]
[SCSI] libsas: Reset timer on taskless scsi_cmnds in sas_scsi_timed_out

Every so often, a scsi_cmnd will time out, and the libsas timeout handler
will discover that the scsi_cmnd does not have a sas_task attached to it.
This can happen in two cases: (1) the scsi_cmnd actually made it through
libsas to the HBA and is now going through scsi_done, or (2) the
scsi_cmnd has been held up (host lock, slab alloc, etc) and libsas has
not yet attached a sas_task.  In both cases, it is safe to ask SCSI for
more time to process the command via EH_RESET_TIMER; we cannot blindly
return EH_HANDLED because if (2) happens, we could end up calling
scsi_done while another CPU is heading towards sas_queuecommand, which
causes slab corruption when sas_task_done updates the freed scsi_cmnd.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Add a sysfs knob to enable/disable a phy
Darrick J. Wong [Thu, 11 Jan 2007 22:14:57 +0000 (14:14 -0800)]
[SCSI] libsas: Add a sysfs knob to enable/disable a phy

This patch lets a user arbitrarily enable or disable a phy via sysfs.
Potential applications include shutting down a phy to replace one
lane of wide port, and (more importantly) providing a method for the
libata SATL to control the phy.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Don't give scsi_cmnds to the EH if they never made it to the SAS LLDD...
Darrick J. Wong [Thu, 11 Jan 2007 22:14:55 +0000 (14:14 -0800)]
[SCSI] libsas: Don't give scsi_cmnds to the EH if they never made it to the SAS LLDD or have already returned

On a system with many SAS targets, it appears possible that a scsi_cmnd
can time out without ever making it to the SAS LLDD or at the same time
that a completion is occurring.  In both of these cases, telling the
LLDD to abort the sas_task makes no sense because the LLDD won't know
about the sas_task; what we really want to do is to increase the timer.
Note that this involves creating another sas_task bit to indicate
whether or not the task has been sent to the LLDD; I could have
implemented this by slightly redefining SAS_TASK_STATE_PENDING, but
this way seems cleaner.

This second version amends the aic94xx portion to set the
TASK_AT_INITIATOR flag for all sas_tasks that were passed to
lldd_execute_task.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Clean up rphys/port dev list after a discovery error on an expander
Darrick J. Wong [Thu, 11 Jan 2007 22:14:52 +0000 (14:14 -0800)]
[SCSI] libsas: Clean up rphys/port dev list after a discovery error on an expander

sas_get_port_device assigns a rphy to a domain device in anticipation
of finding a disk.  When a discovery error occurs in
sas_discover_{sata,sas,expander}*, however, we need to clean up that
rphy and the port device list so that we don't GPF.  In addition, we
need to check the result of the second sas_notify_lldd_dev_found.
This patch seems ok on a x260, x366 and x206m.

This patch fixes up sas_expander.c separately because jejb has some
cleanup patches of his own that are a prerequisite.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] libsas: Clean up rphys/port dev list after a discovery error.
Darrick J. Wong [Thu, 11 Jan 2007 22:14:49 +0000 (14:14 -0800)]
[SCSI] libsas: Clean up rphys/port dev list after a discovery error.

sas_get_port_device assigns a rphy to a domain device in anticipation
of finding a disk.  When a discovery error occurs in
sas_discover_{sata,sas,expander}*, however, we need to clean up that
rphy and the port device list so that we don't GPF.  In addition, we
need to check the result of the second sas_notify_lldd_dev_found.
This patch seems ok on a x260, x366 and x206m.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla2xxx: correct locking while call starget_for_each_device()
Seokmann Ju [Tue, 9 Jan 2007 19:37:52 +0000 (11:37 -0800)]
[SCSI] qla2xxx: correct locking while call starget_for_each_device()

Removed spin_unlock_irq()/spin_lock_irq() pairs surrounding
starget_for_each_device() calls.
As Matthew W. pointed out, starget_for_each_device() can be called under
a spinlock being held.
The change has been tested and verified on qla2xxx.ko module.
Thanks Matthew W. and Hisashi H. for help.

Signed-off-by: Andrew Vasquez <Andrew.vasquez@qlogic.com>
Signed-off-by: Seokmann Ju <Seokmann.ju@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] scsi_scan message cosmetic error
Kurt Garloff [Tue, 9 Jan 2007 01:28:54 +0000 (02:28 +0100)]
[SCSI] scsi_scan message cosmetic error

Hi,

Minor typo ...
In my first iteration of patches (that got merged), the
BLIST_ATTACH_PQ3 actually had the value 0x800000, but that
got changed later to avoid conflicts. This piece must have
been overlooked.
You could obviously do something like %x and then add the
bitflags, but that looks overkill for something that does
not tend to change.

Please merge.
(Patch applied against latest 2.6.20rc version that I tested.)

From: Kurt Garloff <kurt@garloff.de>
Subject: [SCSI SCAN] Fix logging message for PQ3 devices

The blacklist flags BLIST_ATTACH_PQ3 has value 0x1000000,
not 0x800000.

Signed-off-by: Kurt Garloff <garloff@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] megaraid: more kernel-doc fixes
Randy Dunlap [Wed, 10 Jan 2007 05:40:52 +0000 (21:40 -0800)]
[SCSI] megaraid: more kernel-doc fixes

More megaraid kernel-doc fixes.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Sumant Patro <sumantp@lsil.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] megaraid: fix kernel-doc
Randy Dunlap [Sat, 6 Jan 2007 06:41:48 +0000 (22:41 -0800)]
[SCSI] megaraid: fix kernel-doc

kernel-doc modifications:
- change "@param var" notation to @var;
- change function/description separator from ':' to '-';
- change var/description separator from '-' to ':';
- fix a few doc. typos;
- don't use kernel-doc /** lead-in when the doc. block is not kernel-doc;
- use Linux common */ ending comment format instead of **/;
- use correct function parameter names;
- place function parameters immediately after the function short description;
- place kernel-doc immediately before its function or macro;

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Sumant Patro <sumantp@lsil.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] megaraid_{mm,mbox}: init fix for kdump
Sumant Patro [Fri, 5 Jan 2007 15:10:09 +0000 (07:10 -0800)]
[SCSI] megaraid_{mm,mbox}: init fix for kdump

1.      Changes in Initialization to fix kdump failure.
        Send SYNC command on loading.
        This command clears the pending commands in the adapter
        and re-initialize its internal RAID structure.
        Without this change, megaraid driver either panics or fails to
        initialize the adapter during kdump's second kernel boot
        if there are pending commands or interrupts from other devices
        sharing the same IRQ.
2.      Authors email-id domain name changed from lsil.com to lsi.com.
        Also modified the MODULE_AUTHOR to megaraidlinux@lsi.com

Signed-off-by: Sumant Patro <sumant.patro@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] scsi_debug: error processing
Douglas Gilbert [Fri, 5 Jan 2007 05:05:25 +0000 (00:05 -0500)]
[SCSI] scsi_debug: error processing

After discussions in the thread titled:
    [PATCH] scsi_debug: illegal blocking memory allocation
here is a patch containing the discussed fix and some other
fixes and additions. The patch is against lk 2.6.20-rc3 .
The version is bumped to 1.81 .

ChangeLog:
  - Change several GFP_KERNEL allocations to GFP_ATOMIC
    as they can be called from queuecommand() context
  - check above allocation returns and if out of memory
    report DID_REQUEUE in two cases, DID_NO_CONNECT in
    another, and fail slave configure() in another
  - add support for WRITE BUFFER command
  - add aborted_command error injection support
    (opts mask 0x10), similar mechanism to
    recovered_error injection.

Signed-off-by: Douglas Gilbert <dougg@torque.net>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] mptctl for mptsas
Douglas Gilbert [Fri, 5 Jan 2007 04:48:54 +0000 (23:48 -0500)]
[SCSI] mptctl for mptsas

This patch makes the mptctl pass through available if
the mptsas driver is selected. Without this patch
if mptsas is the only fusion driver chosen, then
the mptctl is not presented as an option.

smp_utils uses the mptctl driver to pass SAS SMP
functions through a MPT SAS HBA.

Signed-off-by: Douglas Gilbert <dougg@torque.net>
Acked-by: "Moore, Eric" <Eric.Moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] kill scsi_rety_command
Christoph Hellwig [Tue, 2 Jan 2007 11:56:00 +0000 (12:56 +0100)]
[SCSI] kill scsi_rety_command

scsi_retry_command only has a single caller, so there is no point
in having this function.  Additionally the memset of the sense
buffer it does is entirely superflous as scsi_request_fn already
calls scsi_init_cmd_errh to perform this memset before the command
is reissued.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] SNI RM 53c710 driver
Thomas Bogendoerfer [Mon, 25 Dec 2006 20:32:04 +0000 (21:32 +0100)]
[SCSI] SNI RM 53c710 driver

This patch adds a SCSI driver for the onboard 53c710 chip of some
SNI RM machines.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] NCR_D700: needs burst length setting to 8
James Bottomley [Sat, 13 Jan 2007 19:46:51 +0000 (13:46 -0600)]
[SCSI] NCR_D700: needs burst length setting to 8

The D700 needs the burst length setting to the previous 53c700 default
of 8 otherwise it will be effectively disabled.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] 53c700: Allow setting burst length
Thomas Bogendoerfer [Mon, 25 Dec 2006 20:30:08 +0000 (21:30 +0100)]
[SCSI] 53c700: Allow setting burst length

This is a patch, which allows not only disabling bursting but to specify
different burst lenghts. This feature is needed to get the 53c700 driver
working for the onboard SCSI controller of SNI RM machines, which only
work reliably with a 4 word burst length.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years agoLinux v2.6.20-rc5 v2.6.20-rc5
Linus Torvalds [Fri, 12 Jan 2007 18:54:26 +0000 (10:54 -0800)]
Linux v2.6.20-rc5

17 years ago[PATCH] blktrace: only add a bounce trace when we really bounce
Jens Axboe [Fri, 12 Jan 2007 11:20:26 +0000 (12:20 +0100)]
[PATCH] blktrace: only add a bounce trace when we really bounce

Currently we issue a bounce trace when __blk_queue_bounce() is called,
but that merely means that the device has a lower dma mask than the
higher pages in the system. The bio itself may still be lower pages. So
move the bounce trace into __blk_queue_bounce(), when we know there will
actually be page bouncing.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoMerge branch 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Fri, 12 Jan 2007 02:26:28 +0000 (18:26 -0800)]
Merge branch 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jikos/hid

* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jikos/hid:
  HID: Fix DRIVER_DESC macro
  HID: mousepoll parameter makes no sense for generic HID
  HID: tiny patch to remove a kmalloc cast
  HID: fix mappings for DiNovo Edge Keyboard - Logitech USB BT receiver

17 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux...
Linus Torvalds [Fri, 12 Jan 2007 02:25:44 +0000 (18:25 -0800)]
Merge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6:
  Revert "ACPI: ibm-acpi: make non-generic bay support optional"
  ACPI: update MAINTAINERS
  ACPI: schedule obsolete features for deletion
  ACPI: delete two spurious ACPI messages
  ACPI: rename cstate_entry_s to cstate_entry
  ACPI: ec: enable printk on cmdline use
  ACPI: Altix: ACPI _PRT support

17 years agoMerge branch 'drm-patches' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Fri, 12 Jan 2007 02:25:20 +0000 (18:25 -0800)]
Merge branch 'drm-patches' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6

* 'drm-patches' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  i915: Fix a DRM_ERROR that should be DRM_DEBUG.

17 years ago[PATCH] PCMCIA: fix drivers broken by recent cleanup
Daniel Ritz [Fri, 8 Dec 2006 17:07:01 +0000 (18:07 +0100)]
[PATCH] PCMCIA: fix drivers broken by recent cleanup

Setting .ConfigBase and .Present is now done at the pcmcia core.

The driver cleanup missed a few places where the driver did set .Present
to PRESENT_OPTION and later to the values from the CIS.  Setting to
PRESENT_OPTION now overrides the values from the CIS.  So just remove
those lines.

Signed-off-by: Daniel Ritz <daniel.ritz@gmx.ch>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] FD_ZERO build fix
Andrew Morton [Thu, 11 Jan 2007 07:15:45 +0000 (23:15 -0800)]
[PATCH] FD_ZERO build fix

unionfs managed to hit this on s390.  Some architectures use __ptr_t in their
FD_ZERO implementation.  We don't have a __ptr_t.  Switch them over to plain
old void*.

Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Josef 'Jeff' Sipek <jsipek@cs.sunysb.edu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] fix typo in geode_configre()@cyrix.c
takada [Thu, 11 Jan 2007 07:15:44 +0000 (23:15 -0800)]
[PATCH] fix typo in geode_configre()@cyrix.c

We write back the wrong register when configuring the Geode processor.
Instead of storing to CCR4, it stores to CCR3.

Cc: Jordan Crouse <jordan.crouse@amd.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Fix HWRNG built-in initcalls priority
Michael Buesch [Thu, 11 Jan 2007 07:15:43 +0000 (23:15 -0800)]
[PATCH] Fix HWRNG built-in initcalls priority

This changes all HWRNG driver initcalls to module_init().  We must probe
the RNGs after the major kernel subsystems are already up and running (like
PCI).

This fixes Bug 7730.
http://bugzilla.kernel.org/show_bug.cgi?id=7730

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Cc: Jan Beulich <jbeulich@novell.com>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] intel-rng workarounds
Jan Beulich [Thu, 11 Jan 2007 07:15:41 +0000 (23:15 -0800)]
[PATCH] intel-rng workarounds

Add a load option to intel-rng to allow skipping the FWH detection,
necessary in case the BIOS has locked read-only the firmware hub space.
Also prevent any attempt to write to firmware space if it cannot be write
enabled (apparently caused hangs on some systems not having an FWH and thus
also not having a respective RNG).

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Michael Buesch <mb@bu3sch.de>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Revert bd_mount_mutex back to a semaphore
David Chinner [Thu, 11 Jan 2007 07:15:41 +0000 (23:15 -0800)]
[PATCH] Revert bd_mount_mutex back to a semaphore

Revert bd_mount_mutex back to a semaphore so that xfs_freeze -f /mnt/newtest;
xfs_freeze -u /mnt/newtest works safely and doesn't produce lockdep warnings.

(XFS unlocks the semaphore from a different task, by design.  The mutex
code warns about this)

Signed-off-by: Dave Chinner <dgc@sgi.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] really fix funsoft driver
David Miller [Thu, 11 Jan 2007 07:15:40 +0000 (23:15 -0800)]
[PATCH] really fix funsoft driver

Fix void cast and re-enable on sparc.

Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] NFS: Fix race in nfs_release_page()
Trond Myklebust [Thu, 11 Jan 2007 07:15:39 +0000 (23:15 -0800)]
[PATCH] NFS: Fix race in nfs_release_page()

    NFS: Fix race in nfs_release_page()

    invalidate_inode_pages2() may find the dirty bit has been set on a page
    owing to the fact that the page may still be mapped after it was locked.
    Only after the call to unmap_mapping_range() are we sure that the page
    can no longer be dirtied.
    In order to fix this, NFS has hooked the releasepage() method and tries
    to write the page out between the call to unmap_mapping_range() and the
    call to remove_mapping(). This, however leads to deadlocks in the page
    reclaim code, where the page may be locked without holding a reference
    to the inode or dentry.

    Fix is to add a new address_space_operation, launder_page(), which will
    attempt to write out a dirty page without releasing the page lock.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
    Also, the bare SetPageDirty() can skew all sort of accounting leading to
    other nasties.

[akpm@osdl.org: cleanup]
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] KVM: add VM-exit profiling
Ingo Molnar [Thu, 11 Jan 2007 07:15:38 +0000 (23:15 -0800)]
[PATCH] KVM: add VM-exit profiling

This adds the profile=kvm boot option, which enables KVM to profile VM
exits.

Use: "readprofile -m ./System.map | sort -n" to see the resulting
output:

   [...]
   18246 serial_out                               148.3415
   18945 native_flush_tlb                         378.9000
   23618 serial_in                                212.7748
   29279 __spin_unlock_irq                        622.9574
   43447 native_apic_write                        2068.9048
   52702 enable_8259A_irq                         742.2817
   54250 vgacon_scroll                             89.3740
   67394 ide_inb                                  6126.7273
   79514 copy_page_range                           98.1654
   84868 do_wp_page                                86.6000
  140266 pit_read                                 783.6089
  151436 ide_outb                                 25239.3333
  152668 native_io_delay                          21809.7143
  174783 mask_and_ack_8259A                       783.7803
  362404 native_set_pte_at                        36240.4000
 1688747 total                                      0.5009

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Avi Kivity <avi@qumranet.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] md: pass down BIO_RW_SYNC in raid{1,10}
Lars Ellenberg [Thu, 11 Jan 2007 07:15:37 +0000 (23:15 -0800)]
[PATCH] md: pass down BIO_RW_SYNC in raid{1,10}

md raidX make_request functions strip off the BIO_RW_SYNC flag, thus
introducing additional latency.

Fixing this in raid1 and raid10 seems to be straightforward enough.

For our particular usage case in DRBD, passing this flag improved some
initialization time from ~5 minutes to ~5 seconds.

Acked-by: NeilBrown <neilb@suse.de>
Signed-off-by: Lars Ellenberg <lars@linbit.com>
Acked-by: Jens Axboe <jens.axboe@oracle.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] i386: sched_clock using init data tsc_disable fix
Vivek Goyal [Thu, 11 Jan 2007 07:15:36 +0000 (23:15 -0800)]
[PATCH] i386: sched_clock using init data tsc_disable fix

o sched_clock() a non-init function is using init data tsc_disable. This
  is flagged by MODPOST on i386 if CONFIG_RELOCATABLE=y

WARNING: vmlinux - Section mismatch: reference to .init.data:tsc_disable from .text between 'sched_clock' (at offset 0xc0109d58) and 'tsc_update_callback'

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Kdump documentation update
Vivek Goyal [Thu, 11 Jan 2007 07:15:35 +0000 (23:15 -0800)]
[PATCH] Kdump documentation update

o Kdump documentation update.
- Update details for using relocatable kernel.
- Start using kexec-tools-testing release as it is latest and old
  kexec-tools can't load relocatable bzImage file.
- Also add kdump on ia64 specific details.

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Cc: Horms <horms@verge.net.au>
Cc: Mohan Kumar M <mohan@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Change cpu_up and co from __devinit to __cpuinit
Gautham R Shenoy [Thu, 11 Jan 2007 07:15:34 +0000 (23:15 -0800)]
[PATCH] Change cpu_up and co from __devinit to __cpuinit

Compiling the kernel with CONFIG_HOTPLUG = y and CONFIG_HOTPLUG_CPU = n
with CONFIG_RELOCATABLE = y generates the following modpost warnings

WARNING: vmlinux - Section mismatch: reference to .init.data: from
.text between '_cpu_up' (at offset 0xc0141b7d) and 'cpu_up'
WARNING: vmlinux - Section mismatch: reference to .init.data: from
.text between '_cpu_up' (at offset 0xc0141b9c) and 'cpu_up'
WARNING: vmlinux - Section mismatch: reference to .init.text:__cpu_up
from .text between '_cpu_up' (at offset 0xc0141bd8) and 'cpu_up'
WARNING: vmlinux - Section mismatch: reference to .init.data: from
.text between '_cpu_up' (at offset 0xc0141c05) and 'cpu_up'
WARNING: vmlinux - Section mismatch: reference to .init.data: from
.text between '_cpu_up' (at offset 0xc0141c26) and 'cpu_up'
WARNING: vmlinux - Section mismatch: reference to .init.data: from
.text between '_cpu_up' (at offset 0xc0141c37) and 'cpu_up'

This is because cpu_up, _cpu_up and __cpu_up (in some architectures) are
defined as __devinit
AND
__cpu_up calls some __cpuinit functions.

Since __cpuinit would map to __init with this kind of a configuration,
we get a .text refering .init.data warning.

This patch solves the problem by converting all of __cpu_up, _cpu_up
and cpu_up from __devinit to __cpuinit. The approach is justified since
the callers of cpu_up are either dependent on CONFIG_HOTPLUG_CPU or
are of __init type.

Thus when CONFIG_HOTPLUG_CPU=y, all these cpu up functions would land up
in .text section, and when CONFIG_HOTPLUG_CPU=n, all these functions would
land up in .init section.

Tested on a i386 SMP machine running linux-2.6.20-rc3-mm1.

Signed-off-by: Gautham R Shenoy <ego@in.ibm.com>
Cc: Vivek Goyal <vgoyal@in.ibm.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] rtc-sh: correctly report rtc_wkalrm.enabled
David Brownell [Thu, 11 Jan 2007 07:15:32 +0000 (23:15 -0800)]
[PATCH] rtc-sh: correctly report rtc_wkalrm.enabled

This fixes the SH rtc driver to
  (a) correctly report 'enabled' status with other alarm status;
  (b) not duplicate that status in its procfs dump

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Acked-by: Paul Mundt <lethal@linux-sh.org>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] qconf: (re)fix SIGSEGV on empty menu items
Roman Zippel [Thu, 11 Jan 2007 07:15:31 +0000 (23:15 -0800)]
[PATCH] qconf: (re)fix SIGSEGV on empty menu items

Back out the recent fix for this bug, fix it by correctly initialising
ConfigInfoView.sym.

Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Cc: "Cyrill V. Gorcunov" <gorcunov@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] Fix sparsemem on Cell
Dave Hansen [Thu, 11 Jan 2007 07:15:30 +0000 (23:15 -0800)]
[PATCH] Fix sparsemem on Cell

Fix an oops experienced on the Cell architecture when init-time functions,
early_*(), are called at runtime.  It alters the call paths to make sure
that the callers explicitly say whether the call is being made on behalf of
a hotplug even, or happening at boot-time.

It has been compile tested on ppc64, ia64, s390, i386 and x86_64.

Acked-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] increment pos before looking for the next cap in __pci_find_next_ht_cap
Brice Goglin [Thu, 11 Jan 2007 07:15:29 +0000 (23:15 -0800)]
[PATCH] increment pos before looking for the next cap in __pci_find_next_ht_cap

While testing 2.6.20-rc3 on a machine with some CK804 chipsets, we noticed
that quirk_nvidia_ck804_msi_ht_cap() was not detecting HT MSI capabilities
anymore.  It is actually caused by the MSI mapping on the root chipset
being the 2nd HT capability in the chain.  pci_find_ht_capability() does
not seem to find anything but the first HT cap correctly, because it
forgets to increment the position before looking for the next cap.  The
following patch seems to fix it.

At least, this proves that having a ttl is good idea since the machine
would have been stucked in an infinite loop if we didn't have a ttl :)

We have to pass pos + PCI_CAP_LIST_NEXT to __pci_find_next_cap_ttl to
get the next HT cap instead of the same one again.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Andrew J. Gallatin <gallatin@myri.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[PATCH] sched: tasks cannot run on cpus onlined after boot
Nathan Lynch [Thu, 11 Jan 2007 07:15:28 +0000 (23:15 -0800)]
[PATCH] sched: tasks cannot run on cpus onlined after boot

Commit 5c1e176781f43bc902a51e5832f789756bff911b ("sched: force /sbin/init
off isolated cpus") sets init's cpus_allowed to a subset of cpu_online_map
at boot time, which means that tasks won't be scheduled on cpus that are
added to the system later.

Make init's cpus_allowed a subset of cpu_possible_map instead.  This should
still preserve the behavior that Nick's change intended.

Thanks to Giuliano Pochini for reporting this and testing the fix:

http://ozlabs.org/pipermail/linuxppc-dev/2006-December/029397.html

Signed-off-by: Nathan Lynch <ntl@pobox.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoDon't put "linux_banner" in the .init section
Linus Torvalds [Fri, 12 Jan 2007 02:18:04 +0000 (18:18 -0800)]
Don't put "linux_banner" in the .init section

It might save a few bytes after bootup, but it causes the string to be
linked in at the end of the final vmlinux image, which defeats the whole
point of doing all this, namely allowing some broken user-space binaries
to search for the kernel version string in the kernel binary.

So just remove the __init specifier.

Cc: Olaf Hering <olaf@aepfle.de>
Cc: Jean Delvare <khali@linux-fr.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andrey Borzenkov <arvidjaar@mail.ru>
Cc: Andrew Morton <akpm@osdl.org>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoRevert "ACPI: ibm-acpi: make non-generic bay support optional"
Henrique de Moraes Holschuh [Thu, 11 Jan 2007 07:58:15 +0000 (02:58 -0500)]
Revert "ACPI: ibm-acpi: make non-generic bay support optional"

This reverts commit 2df910b4c3edcce9a0c12394db6f5f4a6e69c712.

ACPI_BAY has not been merged into mainline yet, so the changes to ibm-acpi
related Kconfig entries that depend on ACPI_BAY were permanently disabling
ibm-acpi bay support.  This is a serious regression for ThinkPad users.

Signed-off-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoPull sgi into release branch
Len Brown [Thu, 11 Jan 2007 06:55:34 +0000 (01:55 -0500)]
Pull sgi into release branch

17 years agoPull trivial into release branch
Len Brown [Thu, 11 Jan 2007 06:55:25 +0000 (01:55 -0500)]
Pull trivial into release branch

17 years agoACPI: update MAINTAINERS
Len Brown [Tue, 9 Jan 2007 00:03:28 +0000 (19:03 -0500)]
ACPI: update MAINTAINERS

s/Maintained/Supported/
and document some sub-maintainers for ACPI drivers.

Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoACPI: schedule obsolete features for deletion
Len Brown [Thu, 11 Jan 2007 06:49:44 +0000 (01:49 -0500)]
ACPI: schedule obsolete features for deletion

Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoACPI: delete two spurious ACPI messages
Venkatesh Pallipadi [Thu, 11 Jan 2007 04:16:36 +0000 (23:16 -0500)]
ACPI: delete two spurious ACPI messages

ACPI: Getting cpuindex for acpiid 0x4

acpi_processor-0742 [00] processor_preregister_: Error while parsing _PSD domain information. Assuming no coordination

http://bugzilla.kernel.org/show_bug.cgi?id=7286

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoACPI: rename cstate_entry_s to cstate_entry
Venkatesh Pallipadi [Thu, 11 Jan 2007 04:08:38 +0000 (23:08 -0500)]
ACPI: rename cstate_entry_s to cstate_entry

style change only.

Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoMerge branch 'for-linus' of git://one.firstfloor.org/home/andi/git/linux-2.6
Linus Torvalds [Thu, 11 Jan 2007 02:06:14 +0000 (18:06 -0800)]
Merge branch 'for-linus' of git://one.firstfloor.org/home/andi/git/linux-2.6

* 'for-linus' of git://one.firstfloor.org/home/andi/git/linux-2.6:
  [PATCH] x86-64: Fix warnings in ia32_aout.c
  [PATCH] i386: Convert some functions to __init to avoid MODPOST warnings
  [PATCH] i386: Fix memory hotplug related MODPOST generated warning
  [PATCH] x86-64: tighten up printks
  [PATCH] x86-64: - Ignore long SMI interrupts in clock calibration
  [PATCH] x86-64: pci quirks MODPOST warning fix
  [PATCH] x86-64: Modpost whitelist reference to more symbols (pattern 3)
  [PATCH] x86-64: modpost add more symbols to whitelist pattern2
  [PATCH] i386: make apic probe function non-init
  [PATCH] i386: cpu hotplug/smpboot misc MODPOST warning fixes
  [PATCH] x86-64: Use different constraint for gcc < 4.1 in bitops.h
  [PATCH] x86-64: Make noirqdebug_setup function non init to fix modpost warning
  [PATCH] i386: Update defconfig
  [PATCH] x86-64: Update defconfig

17 years ago[PATCH] x86-64: Fix warnings in ia32_aout.c
Andi Kleen [Thu, 11 Jan 2007 00:52:45 +0000 (01:52 +0100)]
[PATCH] x86-64: Fix warnings in ia32_aout.c

Fix

linux/arch/x86_64/ia32/ia32_aout.c: In function ‘create_aout_tables’:
linux/arch/x86_64/ia32/ia32_aout.c:244: warning: cast from pointer to integer of different size
linux/arch/x86_64/ia32/ia32_aout.c:253: warning: cast from pointer to integer of different size

with gcc 4.3
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Convert some functions to __init to avoid MODPOST warnings
Vivek Goyal [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] i386: Convert some functions to __init to avoid MODPOST warnings

o Some functions which should have been in init sections as they are called
  only once. Put them in init sections. Otherwise MODPOST generates warning
  as these functions are placed in .text and they end up accessing something
  in init sections.

WARNING: vmlinux - Section mismatch: reference to .init.text:migration_init
from .text between 'do_pre_smp_initcalls' (at offset 0xc01000d1) and
'run_init_process'

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Fix memory hotplug related MODPOST generated warning
Vivek Goyal [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] i386: Fix memory hotplug related MODPOST generated warning

o Fix modpost generated warning.

WARNING: vmlinux - Section mismatch: reference to .init.text: from .text
between 'add_one_highpage_hotplug' (at offset 0xc0113d3f) and 'online_page'

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: tighten up printks
Muli Ben-Yehuda [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] x86-64: tighten up printks

Signed-off-by: Muli Ben-Yehuda <muli@il.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] x86-64: - Ignore long SMI interrupts in clock calibration
Jack Steiner [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] x86-64: - Ignore long SMI interrupts in clock calibration

Ensure that no SMI interrupts occur between the read of the HPET & TSC
in the clock calibration loop.

I noticed that a 2.66GHz system incorrectly detected the processor
clock speed about 1/7 of the time:

time.c: Detected 2660.005 MHz processor. (most of the time)
time.c: Detected 2988.203 MHz processor. (sometime)

The problem is caused by an SMI interrupt occuring in hpet_calibrate_tsc()
between the read of the HPET & TSC. Prior to switching the BIOS into
ACPI mode, it appears that every 27msec an SMI interrupt occurs. The
SMI interrupt takes 4.8 msec to process.

Note: On my test system, TICK_MIN had to be >380. I picked 5000
to minimize risk of having a value that is too small for other
platforms.

Signed-off-by: Jack Steiner <steiner@sgi.com>
Signed-off-by: Andi Kleen <ak@suse.de>
 arch/x86_64/kernel/time.c |   25 +++++++++++++++++++++----
 1 file changed, 21 insertions(+), 4 deletions(-)

17 years ago[PATCH] x86-64: pci quirks MODPOST warning fix
Vivek Goyal [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] x86-64: pci quirks MODPOST warning fix

o MODPOST generates warnings for i386 if kernel is compiled with
  CONFIG_RELOCATABLE=y

WARNING: vmlinux - Section mismatch: reference to .init.data: from .text between 'asus_hides_smbus_lpc_ich6' (at offset 0xc0217d58) and 'quirk_cardbus_legacy'
WARNING: vmlinux - Section mismatch: reference to .init.data: from .text between 'asus_hides_smbus_lpc' (at offset 0xc0217fd9) and 'pci_match_id'

o Two quirk functions which are non __init, are accessing data which is
  of type __init.

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Modpost whitelist reference to more symbols (pattern 3)
Vivek Goyal [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] x86-64: Modpost whitelist reference to more symbols (pattern 3)

o MODPOST generates warning on i386 if kernel is compiled with
  CONFIG_RELOCATABLE=y.

WARNING: vmlinux - Section mismatch: reference to .init.text:__init_begin from .text between 'free_initmem' (at offset 0xc0114fd3) and 'do_test_wp_bit'
WARNING: vmlinux - Section mismatch: reference to .init.text:_sinittext from .text between 'core_kernel_text' (at offset 0xc012aeae) and 'kernel_text_address'
WARNING: vmlinux - Section mismatch: reference to .init.text:_einittext from .text between 'core_kernel_text' (at offset 0xc012aeb7) and 'kernel_text_address'
WARNING: vmlinux - Section mismatch: reference to .init.text:_sinittext from .text between 'get_symbol_pos' (at offset 0xc0135776) and 'reset_iter'
WARNING: vmlinux - Section mismatch: reference to .init.text:_einittext from .text between 'get_symbol_pos' (at offset 0xc013577d) and 'reset_iter'

o These symbols (__init_begin, _sinittext, _einittext) belong to init
  section and generally represent a section boundary. These are special
  symbols in the sense that their size is zero and no memory is allocated
  for them in init section. Their addr and value are same. So even if
  we free the init section, it is ok to reference them.

o Whitelist access to such select symbols in MODPOST.

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] x86-64: modpost add more symbols to whitelist pattern2
Vivek Goyal [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] x86-64: modpost add more symbols to whitelist pattern2

o MODPOST generates warning for i386 if compiled with CONFIG_RELOCATABLE=y
  and serial console support is enabled.

o Serial console setup function, serial8250_console_setup(), is a non __init
  function and it calls functions which are of type __init().
  (uart_parse_options() and uart_set_options()). Assuming, setup will
  be called during init time, changing serial8250_console_setup() to __init.

o Adding one more pattern to modpost whitelist. Console drivers might
  have *_console structures containing references to setup functions which
  can be of __init type. Don't generate warnings for those.

WARNING: vmlinux - Section mismatch: reference to .init.text: from .data between 'serial8250_console' (at offset 0xc05a33d8) and 'serial8250_reg'

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: make apic probe function non-init
Vivek Goyal [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] i386: make apic probe function non-init

o struct genapic contains pointer to probe() function which is of type
  __init. Hence MODPOST generates warning if kernel is compiled with
  CONFIG_RELOCATABLE=y for i386.

WARNING: vmlinux - Section mismatch: reference to .init.text: from .data between 'apic_summit' (at offset 0xc058b504) and 'apic_bigsmp'
WARNING: vmlinux - Section mismatch: reference to .init.text: from .data between 'apic_bigsmp' (at offset 0xc058b5a4) and 'cpu.4471'
WARNING: vmlinux - Section mismatch: reference to .init.text: from .data between 'apic_es7000' (at offset 0xc058b644) and 'apic_default'
WARNING: vmlinux - Section mismatch: reference to .init.text: from .data between 'apic_default' (at offset 0xc058b6e4) and 'interrupt'

o One of the possible options is to put special case check in MODPOST to
  not emit warnings for this case but I think it is not a very good option
  in terms of maintenance.

o Another option is to make probe() function non __init. Anyway this function
  is really small so not freeing this memory after init is not a big deal.
  Secondly, from a programming perspective, probably genapic should not
  provide pointers to functions which have been freed as genapic is non
  __init and is used even after initialization is complete.

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: cpu hotplug/smpboot misc MODPOST warning fixes
Vivek Goyal [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] i386: cpu hotplug/smpboot misc MODPOST warning fixes

o Misc smpboot/cpu hotplug path cleanups. I did those to supress the
  warnings generated by MODPOST. These warnings are visible only
  if CONFIG_RELOCATABLE=y.

o CONFIG_RELOCATABLE compiles the kernel with --emit-relocs option. This
  option retains relocation information in vmlinux file and MODPOST
  is quick to spit out "Section mismatch" warnings.

o This patch fixes some of those warnings. Many of the functions in
  smpboot case are __devinit type and they in turn accesses text/data which
  if of type __cpuinit. Now if CONFIG_HOTPLUG=y and CONFIG_HOTPLUG_CPU=n
  then we end up in cases where a function in .text segment is calling
  another function in .init.text segment and MODPOST emits warning.

WARNING: vmlinux - Section mismatch: reference to .init.text:identify_cpu from .text between 'smp_store_cpu_info' (at offset 0xc011020d) and 'do_boot_cpu'
WARNING: vmlinux - Section mismatch: reference to .init.text:init_gdt from .text between 'do_boot_cpu' (at offset 0xc01102ca) and '__cpu_up'
WARNING: vmlinux - Section mismatch: reference to .init.text:print_cpu_info from .text between 'do_boot_cpu' (at offset 0xc01105d0) and '__cpu_up'

o It also fixes the issues where CONFIG_HOTPLUG_CPU=y and start_secondary()
  is calling smp_callin() which in-turn calls synchronize_tsc_ap() which is
  of type __init. This should have meant broken CPU hotplug.

WARNING: vmlinux - Section mismatch: reference to .init.data: from .text between 'start_secondary' (at offset 0xc011603f) and 'initialize_secondary'
WARNING: vmlinux - Section mismatch: reference to .init.data: from .text between 'MP_processor_info' (at offset 0xc0116a4f) and 'mp_register_lapic'
WARNING: vmlinux - Section mismatch: reference to .init.data: from .text between 'MP_processor_info' (at offset 0xc0116a4f) and 'mp_register_lapic'

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Use different constraint for gcc < 4.1 in bitops.h
Andi Kleen [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] x86-64: Use different constraint for gcc < 4.1 in bitops.h

+m is really correct for a RMW instruction, but some older gccs
error out. I finally gave in and ifdefed it.

This fixes compilation errors with some compiler version.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Make noirqdebug_setup function non init to fix modpost warning
Vivek Goyal [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] x86-64: Make noirqdebug_setup function non init to fix modpost warning

o noirqdebug_setup() is __init but it is being called by
  quirk_intel_irqbalance() which if of type __devinit. If CONFIG_HOTPLUG=y,
  quirk_intel_irqbalance() is put into text section and it is wrong to
  call a function in __init section.

o MODPOST flags this on i386 if CONFIG_RELOCATABLE=y

WARNING: vmlinux - Section mismatch: reference to .init.text:noirqdebug_setup from .text between 'quirk_intel_irqbalance' (at offset 0xc010969e) and 'i8237A_suspend'

o Make noirqdebug_setup() non-init.

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Update defconfig
Andi Kleen [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] i386: Update defconfig

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Update defconfig
Andi Kleen [Thu, 11 Jan 2007 00:52:44 +0000 (01:52 +0100)]
[PATCH] x86-64: Update defconfig

Signed-off-by: Andi Kleen <ak@suse.de>