]> err.no Git - linux-2.6/log
linux-2.6
17 years ago[SCSI] zfcp: clear adapter failed flag if an fsf request times out.
Heiko Carstens [Tue, 8 May 2007 09:18:50 +0000 (11:18 +0200)]
[SCSI] zfcp: clear adapter failed flag if an fsf request times out.

Must clear adapter failed flag if an fsf request times out. This is
necessary because on link down situations the failed flags gets set
but the QDIO queues are still up. Since an adapter reopen will be
skipped if the failed flag is set an adapter_reopen that is issued
on fsf request timeout has no effect if the local link is down.
Might lead to locked up system if the SCSI stack is waiting for abort
completion.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] zfcp: rework request ID management.
Heiko Carstens [Tue, 8 May 2007 09:17:54 +0000 (11:17 +0200)]
[SCSI] zfcp: rework request ID management.

Simplify request ID management and make sure that frequently used
functions are inlined. Also fix a memory leak in zfcp_adapter_enqueue()
which only gets hit in error handling.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] zfcp: Fix deadlock between zfcp ERP and SCSI
Christof Schmitt [Tue, 8 May 2007 09:16:52 +0000 (11:16 +0200)]
[SCSI] zfcp: Fix deadlock between zfcp ERP and SCSI

The SCSI stack requires low level drivers to register and
unregister devices. For zfcp this leads to the situation where
zfcp calls the SCSI stack, the SCSI tries to scan the new device
and the scan SCSI command fails. This would require the zfcp erp,
but the erp thread is already blocked in the register call.

The fix is to make sure that the calls from the ERP thread to
the SCSI stack do not block the ERP thread. In detail:
1) Use a workqueue to avoid blocking of the scsi_scan_target calls.
2) When removing a unit make sure that no scsi_scan_target call is
   pending.
3) Replace scsi_flush_work with scsi_target_unblock. This avoids
   blocking and has the same result.

Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] zfcp: Locking for req_no and req_seq_no
Christof Schmitt [Tue, 8 May 2007 09:15:48 +0000 (11:15 +0200)]
[SCSI] zfcp: Locking for req_no and req_seq_no

There is a possible race condition while generating the unique
request ids and sequence numbers. Both might be read at the
same time and have the same value. Fix this by serializing the
access through the queue lock of the adapter: First call
zfcp_fsf_req_sbal_get that acquires the lock, then read and
increment the unique ids.

Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] zfcp: print S_ID and D_ID with 3 bytes
Christof Schmitt [Tue, 8 May 2007 09:14:41 +0000 (11:14 +0200)]
[SCSI] zfcp: print S_ID and D_ID with 3 bytes

S_ID and D_ID are defined in the FCP spec as 3 byte fields.
Change the output in zfcp print statements accordingly to print
them with only 3 bytes.

Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Use PCI-E reset API for new ipr adapter
Brian King [Mon, 7 May 2007 22:09:05 +0000 (17:09 -0500)]
[SCSI] ipr: Use PCI-E reset API for new ipr adapter

Use a newly added PCI API to issue a PCI Fundamental reset
(warm reset) to a new ipr PCI-E adapter. Typically, the
ipr adapter uses the start BIST bit in config space to reset
an adapter. Issuing start BIST on this particular adapter
results in the PCI-E logic on the card losing sync, which
causes PCI-E errors, making the card unusable. The only reset
mechanism that exists on this hardware that does not have this
problem is PCI Fundamental reset (warm reset).

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla2xxx: Update version number to 8.01.07-k7.
Andrew Vasquez [Mon, 7 May 2007 14:43:03 +0000 (07:43 -0700)]
[SCSI] qla2xxx: Update version number to 8.01.07-k7.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla2xxx: Add MSI support.
Andrew Vasquez [Mon, 7 May 2007 14:43:02 +0000 (07:43 -0700)]
[SCSI] qla2xxx: Add MSI support.

Supported ISP types include ISP2422 and revision-2 type
ISP2432 chips.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla2xxx: Correct pci_set_msi() usage semantics.
Andrew Vasquez [Mon, 7 May 2007 14:43:01 +0000 (07:43 -0700)]
[SCSI] qla2xxx: Correct pci_set_msi() usage semantics.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla2xxx: Attempt to stop firmware only if it had been previously executed.
Andrew Vasquez [Mon, 7 May 2007 14:43:00 +0000 (07:43 -0700)]
[SCSI] qla2xxx: Attempt to stop firmware only if it had been previously executed.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla2xxx: Honor NVRAM port-down-retry-count settings.
Andrew Vasquez [Mon, 7 May 2007 14:42:59 +0000 (07:42 -0700)]
[SCSI] qla2xxx: Honor NVRAM port-down-retry-count settings.

Hardcoding the qlport_down_retry module-parameter
effectively disallowed any user-defined NVRAM setting to go
into effect.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla2xxx: Error-out during probe() if we're unable to complete HBA initialization.
Andrew Vasquez [Mon, 7 May 2007 14:42:58 +0000 (07:42 -0700)]
[SCSI] qla2xxx: Error-out during probe() if we're unable to complete HBA initialization.

Remove a stale check against ha->device_flags
(DFLG_NO_CABLE) as topology scanning is performed within the
DPC-thread context.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] zfcp: Stop system after memory corruption
Christof Schmitt [Mon, 7 May 2007 14:35:04 +0000 (16:35 +0200)]
[SCSI] zfcp: Stop system after memory corruption

For each request that is sent to the FCP adapter, zfcp allocates
memory. Status information and data that is being read from the
device is written to this memory by the hardware. After that,
the hardware signals this via the response queue and zfcp
continues processing.

Now, if zfcp detects that there is a signal for an incoming
response from the hardware, but there is no outstanding request
for that request id, then some memory that can be in use anywhere
in the system has just been overwritten. This should never happen,
but if it does, stop the system with a panic.

Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] mesh: cleanup variable usage in interrupt handler
Olaf Hering [Thu, 26 Apr 2007 07:35:19 +0000 (00:35 -0700)]
[SCSI] mesh: cleanup variable usage in interrupt handler

_ convert void* to struct mesh_state*

- remove unused irq argument from mesh_interrupt()

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] megaraid: replace yield() with cond_resched()
Amol Lad [Thu, 26 Apr 2007 07:35:13 +0000 (00:35 -0700)]
[SCSI] megaraid: replace yield() with cond_resched()

For this driver cond_resched() seems to be a better alternative

Signed-off-by: Amol Lad <amol@verismonetworks.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] megaraid: fix warnings when CONFIG_PROC_FS=n
walter harms [Thu, 26 Apr 2007 07:35:09 +0000 (00:35 -0700)]
[SCSI] megaraid: fix warnings when CONFIG_PROC_FS=n

drivers/scsi/megaraid.c: In function 'megaraid_probe_one':
drivers/scsi/megaraid.c:4893: warning: implicit declaration of function 'mega_create_proc_entry'
drivers/scsi/megaraid.c: In function 'megaraid_remove_one':
drivers/scsi/megaraid.c:4968: warning: unused variable 'buf'

Fix by adding #defines

Signed-off-by: walter harms <wharms@bfs.de>
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aacraid: correct SUN products to README
Salyzyn, Mark [Thu, 3 May 2007 20:08:39 +0000 (16:08 -0400)]
[SCSI] aacraid: correct SUN products to README

Correct SUN products in aacraid documentation, preliminary names were
changed from internal project to customer product prior to release.

Signed-off-by: Mark Salyzyn <aacraid@adaptec.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aacraid: superfluous adapter reset for IBM 8 series ServeRAID controllers
Salyzyn, Mark [Tue, 1 May 2007 15:43:06 +0000 (11:43 -0400)]
[SCSI] aacraid: superfluous adapter reset for IBM 8 series ServeRAID controllers

The kexec patch introduced a superfluous (and otherwise inert) reset of
some adapters. The register can have a hardware default value that has
zeros for the undefined interrupts. This patch refines the test of the
interrupt enable register to focus on only the interrupts that affect
the driver in order to detect if an incomplete shutdown of the Adapter
had occurred (kdump).

Signed-off-by: Mark Salyzyn <aacraid@adaptec.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aacraid: kexec fix (reset interrupt handler)
Salyzyn, Mark [Mon, 30 Apr 2007 17:22:24 +0000 (13:22 -0400)]
[SCSI] aacraid: kexec fix (reset interrupt handler)

Another layer on this onion also discovered by Duane, the
interrupt enable handler also needed to be set ... The interrupt enable
was called from within the synchronous command handler.

Signed-off-by: Mark Salyzyn <aacraid@adaptec.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] tmscsim: Remove the last bus_to_virt()
Guennadi Liakhovetski [Fri, 4 May 2007 21:00:49 +0000 (23:00 +0200)]
[SCSI] tmscsim: Remove the last bus_to_virt()

Dynamically map the buffer for PIO for the residue byte.

Signed-off-by: G. Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] tmscsim: remove bogus endianness conversions
Guennadi Liakhovetski [Fri, 4 May 2007 20:59:40 +0000 (22:59 +0200)]
[SCSI] tmscsim: remove bogus endianness conversions

cpu_to_le32 endianness conversions in tmscsim.c, followed by
arithmetic operations don't look correct. Besides, {in,out}[wl]
already perform the necessary conversions. Further, bus addresses
of request buffers are guaranteed to be (mapped) under 4G by
current scsi- and block-layer defaults. This could be explicitly
enforced by using blk_queue_bounce_limit(), which, however,
doesn't seem to be the common practice among SCSI drivers.

Signed-off-by: G. Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] tmscsim: remove long dead DMA_INT
Guennadi Liakhovetski [Fri, 4 May 2007 20:57:37 +0000 (22:57 +0200)]
[SCSI] tmscsim: remove long dead DMA_INT

DMA_INT code is disabled since 1998, remove it to prepare
for further cleanup.

Signed-off-by: G. Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Better handle adapter boot time errors
Brian King [Wed, 2 May 2007 15:44:11 +0000 (10:44 -0500)]
[SCSI] ipr: Better handle adapter boot time errors

If an ipr adapter encounters an adapter error requiring an
adapter reset to recover from prior to driver load time, the
error will be ignored and recovery will not happen until the
initial timeout occurs waiting for the firmware to come ready,
which means a five minute timeout. Fix is to read the interrupt
register before clearing any of the interrupts at probe time.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] deprecate the old NCR53C9x driver
Christoph Hellwig [Sun, 29 Apr 2007 22:05:48 +0000 (00:05 +0200)]
[SCSI] deprecate the old NCR53C9x driver

Now that we have the much better esp_scsi driver and low level drivers
are easy to port over deprecate the old NCR53C9x driver.

I've Cc'ed the m68k and mips lists because all but one bus glues are
for these platforms.  Chances stand bad for the remaining driver,
mca_53c9x which hasn't gotten any non-trivial update since it was
merge in late 2.1.x and whos maintainers mail address bounces.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] fc_transport: make all rports wait dev_loss_tmo before removing them
James Smart [Fri, 27 Apr 2007 15:53:17 +0000 (11:53 -0400)]
[SCSI] fc_transport: make all rports wait dev_loss_tmo before removing them

Per the comment in the change - it's not always prudent to immediately
remove the rport upon first notice of a disconnect. Make all rports
wait dev_loss_tmo before being deleted (and each could have a separate
dev_loss_tmo value).

The original post was:
http://marc.info/?l=linux-scsi&m=117392196006703&w=2

The repost contains the following changes:
 - Bug fix in fc_starget_delete(). Dev_loss_tmo_callbk() was called prior to
   tearing down the target. The callback is to be the last thing called, as
   it tells the LLDD that the rport is completely finished and can be torn
   down.  Rework so that terminate_rport_io() is called to terminate the
   outstanding io. Isolated work so it's is simply "starget" work.
 - Fix holes in original patch. There were code paths that did not expect
   the dev_loss_tmo timer to be running for the non-fcp rports.
 - Bug Fix: the transport wasn't protecting against a LLDD calling
   fc_remote_port_delete() back-to-back. Thus, the dev_loss_tmo timer
   could be restarted such that it fires after the rport had been deleted.
   Validate rport state before starting the timer.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Enable multi-initator RAID support
Brian King [Thu, 26 Apr 2007 21:00:16 +0000 (16:00 -0500)]
[SCSI] ipr: Enable multi-initator RAID support

Enables multi-initiator support on ipr RAID adapters that support it.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Improved dual adapter errors
Brian King [Thu, 26 Apr 2007 21:00:14 +0000 (16:00 -0500)]
[SCSI] ipr: Improved dual adapter errors

Formats ipr dual adapter errors so that they are more compact.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Increase adapter operational timeout
Brian King [Thu, 26 Apr 2007 21:00:13 +0000 (16:00 -0500)]
[SCSI] ipr: Increase adapter operational timeout

Increases the adapter operational timeout for some adapters that support
dual controller configurations, since they may take longer to come ready.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Handle IOA reset request
Brian King [Thu, 26 Apr 2007 21:00:12 +0000 (16:00 -0500)]
[SCSI] ipr: Handle IOA reset request

In ipr dual adapter configurations, the ipr adapter firmware
may require an adapter reset for various reasons. The reset
is requested by the adapter firmware logging an error with
an IOASC of 0x02048000. Add support to log this error, and
reset the 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: Handle check condition status from disk array device
Brian King [Thu, 26 Apr 2007 21:00:10 +0000 (16:00 -0500)]
[SCSI] ipr: Handle check condition status from disk array device

On newer levels of microcode for ipr RAID adapters supporting
multi-initiator configurations, the disk array, or VSET, resources
are capable of generating a check condition. This patch prevents
ipr from generating sense data in this scenario and retrieving it
from the logical device instead.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Handle UA on disk array following an adapter reset
Brian King [Thu, 26 Apr 2007 21:00:08 +0000 (16:00 -0500)]
[SCSI] ipr: Handle UA on disk array following an adapter reset

On certain ipr RAID adapters, which are capable of multi-initiator
configurations, the disk array, or VSET, resources will be in a POR
Unit Attention state following an adapter reset. In order for the
midlayer to handle the UA, I must report a bus reset to the logical
disk array bus at the end of an adapter reset.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Allow driver_data to be passed for dynamic ids
Brian King [Thu, 26 Apr 2007 21:00:07 +0000 (16:00 -0500)]
[SCSI] ipr: Allow driver_data to be passed for dynamic ids

Since driver_data for pci ids in the ipr driver is now
just flags, we can allow these to be passed in from userspace
for dynamic ids.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ipr: Prevent overlapped adapter resets
Brian King [Thu, 26 Apr 2007 21:00:06 +0000 (16:00 -0500)]
[SCSI] ipr: Prevent overlapped adapter resets

This patch fixes some scenarios where an ipr adapter
could get reset overlapped, which could cause very
long timeouts to occur, or PCI bus errors.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] sas_scsi_host: Convert to use the kthread API
Christoph Hellwig [Thu, 26 Apr 2007 13:38:01 +0000 (09:38 -0400)]
[SCSI] sas_scsi_host: Convert to use the kthread API

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla4xxx: possible cleanups
Adrian Bunk [Thu, 26 Apr 2007 07:35:16 +0000 (00:35 -0700)]
[SCSI] qla4xxx: possible cleanups

- make needlessly global code static
- #if 0 unused functions

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: Ravi Anand <ravi.anand@qlogic.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Change version number to 8.1.12
James Smart [Wed, 25 Apr 2007 13:53:42 +0000 (09:53 -0400)]
[SCSI] lpfc 8.1.12 : Change version number to 8.1.12

Change version number to 8.1.12

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Update copyright year to 2007
James Smart [Wed, 25 Apr 2007 13:53:35 +0000 (09:53 -0400)]
[SCSI] lpfc 8.1.12 : Update copyright year to 2007

Update copyright year to 2007

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Added support for 8G speed and new HBAs
James Smart [Wed, 25 Apr 2007 13:53:28 +0000 (09:53 -0400)]
[SCSI] lpfc 8.1.12 : Added support for 8G speed and new HBAs

Added support for 8G speed and new HBAs.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Add support for async scanning
James Smart [Wed, 25 Apr 2007 13:53:22 +0000 (09:53 -0400)]
[SCSI] lpfc 8.1.12 : Add support for async scanning

Add support for async scanning

Notes: This is the async scan patch to our driver from Matthew Wilcox.
  The async scan logic is still subject to errors in insmod/rmmod, as
  the async scan threads don't get shutdown when the module unloads
  underneath them. See http://marc.info/?l=linux-scsi&m=117551999925582&w=2

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Don't process ERATT interrupts when issuing KILL_BOARD mbx command
James Smart [Wed, 25 Apr 2007 13:53:15 +0000 (09:53 -0400)]
[SCSI] lpfc 8.1.12 : Don't process ERATT interrupts when issuing KILL_BOARD mbx command

Don't process ERATT interrupts when issuing KILL_BOARD mbx command

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Collapse discovery lists to a single node list
James Smart [Wed, 25 Apr 2007 13:53:08 +0000 (09:53 -0400)]
[SCSI] lpfc 8.1.12 : Collapse discovery lists to a single node list

Collapse discovery lists to a single node list.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Reference count node structures for node lifetime management
James Smart [Wed, 25 Apr 2007 13:53:01 +0000 (09:53 -0400)]
[SCSI] lpfc 8.1.12 : Reference count node structures for node lifetime management

Reference count node structures for node lifetime management.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Improve handling of failed ELS aborts
James Smart [Wed, 25 Apr 2007 13:52:55 +0000 (09:52 -0400)]
[SCSI] lpfc 8.1.12 : Improve handling of failed ELS aborts

Improve handling of failed ELS aborts.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Improve diagnostic messages and change local loopback message...
James Smart [Wed, 25 Apr 2007 13:52:49 +0000 (09:52 -0400)]
[SCSI] lpfc 8.1.12 : Improve diagnostic messages and change local loopback message to KERN_INFO

Improve diagnostic messages and change local loopback message to KERN_INFO

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Fixed recovery of rport after race with dev_loss_tmo
James Smart [Wed, 25 Apr 2007 13:52:41 +0000 (09:52 -0400)]
[SCSI] lpfc 8.1.12 : Fixed recovery of rport after race with dev_loss_tmo

Fixed recovery of rport after race with dev_loss_tmo

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Round 2 of Miscellaneous fixes
James Smart [Wed, 25 Apr 2007 13:52:34 +0000 (09:52 -0400)]
[SCSI] lpfc 8.1.12 : Round 2 of Miscellaneous fixes

Round 2 of Miscellaneous fixes:
 - Ensure we don't prematurely re-enable IRQs in lpfc_sli_abort_fcp_cmpl().
 - Prevent freeing of iocb after IOCB_TIMEDOUT error.
 - Added code to cleanup REG_LOGIN mailbox command when a LOGO is received.
 - Fix offline window where more work can sneak in after clearing work_ha
 - Use target reset instead of LU reset in bus_device_reset_handler
 - Fixed system hangs due to leaked host_lock.
 - Fixed NULL pointer dereference during I/O with LIP.
 - Fixed false iocb timeout.
 - Fixed name server query response handling.
 - Change rport dev_loss_tmo value when user change lpfc HBA's dev_loss_tmo.
 - Fixed a memory leak in lpfc_sli_wake_mbox_wait.
 - Fixed check for dropped frames.
 - Removed hba queue depth calculation based on device PCI IDs
 - Change min cr_count value specified in comment to agree with setting

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Reorganize lpfc_nlp_list() and callers to prepare for nodelist...
James Smart [Wed, 25 Apr 2007 13:52:27 +0000 (09:52 -0400)]
[SCSI] lpfc 8.1.12 : Reorganize lpfc_nlp_list() and callers to prepare for nodelist simplification

Reorganize lpfc_nlp_list() and callers to prepare for nodelist simplification.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Fix unlock inside list traversal
James Smart [Wed, 25 Apr 2007 13:52:20 +0000 (09:52 -0400)]
[SCSI] lpfc 8.1.12 : Fix unlock inside list traversal

Fix unlock inside list traversal.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Rework offline path to solve HBA reset issues
James Smart [Wed, 25 Apr 2007 13:51:45 +0000 (09:51 -0400)]
[SCSI] lpfc 8.1.12 : Rework offline path to solve HBA reset issues

Rework offline path to solve HBA reset issues

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Modify ELS abort handling to prevent double completion
James Smart [Wed, 25 Apr 2007 13:51:38 +0000 (09:51 -0400)]
[SCSI] lpfc 8.1.12 : Modify ELS abort handling to prevent double completion

Modify ELS abort handling to prevent double completion

Rework portions of ELS abort handling to prevent double completion
 - Rework ELS iotags and correct abort routine
 - Move the (badly wrong) ELS completion logic from the initial ELS
   abort request function to the ELS completion function.
 - Fixup the iocb completion handling to account for the ELS abort
   completions.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] lpfc 8.1.12 : Misc bug fixes and code cleanup
James Smart [Wed, 25 Apr 2007 13:51:30 +0000 (09:51 -0400)]
[SCSI] lpfc 8.1.12 : Misc bug fixes and code cleanup

Misc bug fixes and code cleanup:
 - Fix system hang while running on systems with IOMMU
 - Fix use after free issues with rports
 - Don't free mailbox structure if it's still on the mboxq list
 - Decrement txq_cnt rather than txcmplq_cnt when parsing the txq list
 - Use msleep for long delays to prevent soft lockup bug check
 - Don't remove node during dev_loss_tmo if discovery is active
 - Fix memory leaks in get/reset statistics and link attention paths
 - Fixed lpfc_ns_rsp to handle entire GID_FT response.
 - mbox interface should use MAILBOX_CMD_SIZE rather than sizeof(MAILBOX_t)
 - Fixed bug check in add_timer.
 - Fixup messages 0116, 0117, and 0128 to report ELS I/O tag.
 - Remove unused parameter to lpfc_cleanup.
 - Change mailbox timeout handling.
 - Remove unused buflist. Code cleanup.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] use sysfs configured timeout for EH Start Unit timeout
Brian King [Thu, 19 Apr 2007 18:59:59 +0000 (13:59 -0500)]
[SCSI] use sysfs configured timeout for EH Start Unit timeout

Use the sysfs configurable timeout when issuing a START_UNIT
command from the scsi error handler. This is needed for devices which
take longer than thirty seconds to respond to the start
unit. The problem was observed when sending a start unit
to a disk array device in an ipr RAID adapter, which results
in the adapter firmware sending potentially multiple commands
to physical devices as a result of this command, which ended
up timing out sometimes. This patch does not change the default
value used for this command.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] megaraid: update version reported by MEGAIOC_QDRVRVER
David Milburn [Thu, 19 Apr 2007 17:10:24 +0000 (12:10 -0500)]
[SCSI] megaraid: update version reported by MEGAIOC_QDRVRVER

Update the driver version reported by MEGAIOC_QDRVRVER to
match LSI_COMMON_MOD_VERSION.

Signed-off-by: David Milburn <dmilburn@redhat.com>
Acked-by: "Patro, Sumant" <Sumant.Patro@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] qla1280: use DMA_64BIT_MASK instead of ~ 0ULL
Matthias Kaehlcke [Tue, 17 Apr 2007 07:58:36 +0000 (09:58 +0200)]
[SCSI] qla1280: use DMA_64BIT_MASK instead of ~ 0ULL

QLA1280: call pci_set_dma_mask with DMA_64BIT_MASK instead of  ~ 0ULL

Signed-off-by: Matthias Kaehlcke <matthias.kaehlcke@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] aacraid: kmalloc/memset->kzalloc
Salyzyn, Mark [Mon, 16 Apr 2007 15:21:50 +0000 (11:21 -0400)]
[SCSI] aacraid: kmalloc/memset->kzalloc

Inspired somewhat by Vignesh Babu <vignesh.babu@wipro.com> patch to
dpt_i2o.c to replace kmalloc/memset sequences with kzalloc, doing the
same for the aacraid driver.

Signed-off-by: Mark Salyzyn <aacraid@adaptec.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] dpt_i2o: kmalloc/memset->kzalloc
vignesh.babu@wipro.com [Mon, 16 Apr 2007 06:05:38 +0000 (11:35 +0530)]
[SCSI] dpt_i2o: kmalloc/memset->kzalloc

Replacing kmalloc/memset combination with kzalloc.

Signed-off-by: vignesh babu <vignesh.babu@wipro.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years ago[SCSI] ch: kmalloc/memset->kzalloc
vignesh.babu@wipro.com [Mon, 16 Apr 2007 06:05:33 +0000 (11:35 +0530)]
[SCSI] ch: kmalloc/memset->kzalloc

Replacing kmalloc/memset combination with kzalloc.

Signed-off-by: vignesh babu <vignesh.babu@wipro.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
17 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Sat, 5 May 2007 22:30:53 +0000 (15:30 -0700)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  [CIFS] Fix typo in cifs readme from previous commit
  [CIFS] Make sec=none force an anonymous mount
  [CIFS] Change semaphore to mutex for cifs lock_sem
  [CIFS] Fix oops in reset_cifs_unix_caps on reconnect
  [CIFS] UID/GID override on CIFS mounts to Samba
  [CIFS] prefixpath mounts to servers supporting posix paths used wrong slash
  [CIFS] Update cifs version to 1.49
  [CIFS] Replace kmalloc/memset combination with kzalloc
  [CIFS]  Add IPv6 support
  [CIFS] New CIFS POSIX mkdir performance improvement (part 2)
  [CIFS] New CIFS POSIX mkdir performance improvement
  [CIFS] Add write perm for usr to file on windows should remove r/o dos attr
  [CIFS] Remove unnecessary parm to cifs_reopen_file
  [CIFS] Switch cifsd to kthread_run from kernel_thread
  [CIFS] Remove unnecessary checks

17 years ago[CIFS] Fix typo in cifs readme from previous commit
Steve French [Sat, 5 May 2007 22:08:06 +0000 (22:08 +0000)]
[CIFS] Fix typo in cifs readme from previous commit

Signed-off-by: Steve French <sfrench@us.ibm.com>
17 years agoMerge branch 'for-linus' of git://one.firstfloor.org/home/andi/git/linux-2.6
Linus Torvalds [Sat, 5 May 2007 21:55:20 +0000 (14:55 -0700)]
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: (231 commits)
  [PATCH] i386: Don't delete cpu_devs data to identify different x86 types in late_initcall
  [PATCH] i386: type may be unused
  [PATCH] i386: Some additional chipset register values validation.
  [PATCH] i386: Add missing !X86_PAE dependincy to the 2G/2G split.
  [PATCH] x86-64: Don't exclude asm-offsets.c in Documentation/dontdiff
  [PATCH] i386: avoid redundant preempt_disable in __unlazy_fpu
  [PATCH] i386: white space fixes in i387.h
  [PATCH] i386: Drop noisy e820 debugging printks
  [PATCH] x86-64: Fix allnoconfig error in genapic_flat.c
  [PATCH] x86-64: Shut up warnings for vfat compat ioctls on other file systems
  [PATCH] x86-64: Share identical video.S between i386 and x86-64
  [PATCH] x86-64: Remove CONFIG_REORDER
  [PATCH] x86-64: Print type and size correctly for unknown compat ioctls
  [PATCH] i386: Remove copy_*_user BUG_ONs for (size < 0)
  [PATCH] i386: Little cleanups in smpboot.c
  [PATCH] x86-64: Don't enable NUMA for a single node in K8 NUMA scanning
  [PATCH] x86: Use RDTSCP for synchronous get_cycles if possible
  [PATCH] i386: Add X86_FEATURE_RDTSCP
  [PATCH] i386: Implement X86_FEATURE_SYNC_RDTSC on i386
  [PATCH] i386: Implement alternative_io for i386
  ...

Fix up trivial conflict in include/linux/highmem.h manually.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoFix compile of tmscsim SCSI driver
Linus Torvalds [Sat, 5 May 2007 21:23:40 +0000 (14:23 -0700)]
Fix compile of tmscsim SCSI driver

It still used the long-deprecated "pci_module_init()" interface, rather
than the proper "pci_register_driver()" one.

[ I don't have the hardware, and I doubt many do, but the fix is
  trivial and obvious, and can't be worse than not compiling ]

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoFix nfsroot build
Ralf Baechle [Sat, 5 May 2007 21:05:11 +0000 (22:05 +0100)]
Fix nfsroot build

  CC      fs/nfs/nfsroot.o
fs/nfs/nfsroot.c:131: error: tokens causes a section type conflict
make[2]: *** [fs/nfs/nfsroot.o] Error 1

This is due to mixing const and non-const content in the same section
which halfway recent gccs absolutely hate.  Fixed by dropping the const.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Sat, 5 May 2007 21:13:36 +0000 (14:13 -0700)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [TG3]: Add TG3_FLAG_SUPPORT_MSI flag.
  [TG3]: Eliminate the TG3_FLAG_5701_REG_WRITE_BUG flag.
  [TG3]: Eliminate the TG3_FLAG_GOT_SERDES_FLOWCTL flag.
  [TG3]: Remove reset during MAC address changes.
  [TG3]: WoL fixes.
  [TG3]: Clear GPIO mask before storing.
  [TG3]: Improve NVRAM sizing.
  [TG3]: Fix TSO bugs.
  [MAC80211]: Add maintainers entry for mac80211.
  [MAC80211]: Add debugfs attributes.
  [MAC80211]: Add mac80211 wireless stack.
  [MAC80211]: Add generic include/linux/ieee80211.h
  [NETLINK]: Remove references to process ID
  [AF_IUCV]: Compile fix - adopt to skbuff changes.

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6
Linus Torvalds [Sat, 5 May 2007 20:30:44 +0000 (13:30 -0700)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6: (87 commits)
  [SCSI] fusion: fix domain validation loops
  [SCSI] qla2xxx: fix regression on sparc64
  [SCSI] modalias for scsi devices
  [SCSI] sg: cap reserved_size values at max_sectors
  [SCSI] BusLogic: stop using check_region
  [SCSI] tgt: fix rdma transfer bugs
  [SCSI] aacraid: fix aacraid not finding device
  [SCSI] aacraid: Correct SMC products in aacraid.txt
  [SCSI] scsi_error.c: Add EH Start Unit retry
  [SCSI] aacraid: [Fastboot] Panics for AACRAID driver during 'insmod' for kexec test.
  [SCSI] ipr: Driver version to 2.3.2
  [SCSI] ipr: Faster sg list fetch
  [SCSI] ipr: Return better qc_issue errors
  [SCSI] ipr: Disrupt device error
  [SCSI] ipr: Improve async error logging level control
  [SCSI] ipr: PCI unblock config access fix
  [SCSI] ipr: Fix for oops following SATA request sense
  [SCSI] ipr: Log error for SAS dual path switch
  [SCSI] ipr: Enable logging of debug error data for all devices
  [SCSI] ipr: Add new PCI-E IDs to device table
  ...

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/jejb/voyager-2.6
Linus Torvalds [Sat, 5 May 2007 20:30:23 +0000 (13:30 -0700)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/jejb/voyager-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/jejb/voyager-2.6:
  [VOYAGER] add smp alternatives
  [VOYAGER] Use modern techniques to setup and teardown low identiy mappings.
  [VOYAGER] Convert the monitor thread to use the kthread API
  [VOYAGER] clockevents driver: bring voyager in to line
  [VOYAGER] clockevents: correct boot cpu is zero assumption
  [VOYAGER] add smp_call_function_single

17 years ago[TG3]: Add TG3_FLAG_SUPPORT_MSI flag.
Michael Chan [Sat, 5 May 2007 20:08:32 +0000 (13:08 -0700)]
[TG3]: Add TG3_FLAG_SUPPORT_MSI flag.

And fix up the code to always allow MSI on 5714 A2.

Call tg3_find_peer() earlier because we need that information before
we can determine whether we can set TG3_FLAG_SUPPORT_MSI or not.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years agoide-cs: recognize 2GB CompactFlash from Transcend
Fabrice Aeschbacher [Sat, 5 May 2007 20:03:51 +0000 (22:03 +0200)]
ide-cs: recognize 2GB CompactFlash from Transcend

Without the following patch, the kernel does not automatically detect
2GB CompactFlash cards from Transcend.

Signed-off-by: Fabrice Aeschbacher <fabrice.aeschbacher@siemens.com>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Acked-by: Peter Stuge <peter@stuge.se>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agohpt366: don't check enablebits for HPT36x
Sergei Shtylyov [Sat, 5 May 2007 20:03:51 +0000 (22:03 +0200)]
hpt366: don't check enablebits for HPT36x

HPT36x chip don't seem to have the channel enable bits, so prevent the IDE core
from checking them...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Michal Kepien <michal.kepien@poczta.onet.pl>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-cris: fix ->speedproc and wrong ->swdma_mask
Bartlomiej Zolnierkiewicz [Sat, 5 May 2007 20:03:51 +0000 (22:03 +0200)]
ide-cris: fix ->speedproc and wrong ->swdma_mask

* fix ->speedproc to set the drive speed

* this driver doesn't support SWDMA so use the correct ->swdma_mask

* BUG() if an unsupported mode is passed to ->speedproc

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agosiimage: fix wrong ->swdma_mask
Bartlomiej Zolnierkiewicz [Sat, 5 May 2007 20:03:51 +0000 (22:03 +0200)]
siimage: fix wrong ->swdma_mask

This driver doesn't support SWDMA so use the correct ->swdma_mask.

While at it:

* no need to call config_chipset_for_pio() in config_chipset_for_dma(),
  if DMA is not available config_chipset_for_pio() will be called
  by siimage_config_drive_for_dma() and if DMA is available
  config_siimage_chipset_for_pio() will be called by siimage_tune_chipset()

* remove needless config_chipset_for_pio() wrapper

* bump driver version

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoit821x: PIO mode setup fixes
Bartlomiej Zolnierkiewicz [Sat, 5 May 2007 20:03:50 +0000 (22:03 +0200)]
it821x: PIO mode setup fixes

* limit max PIO mode to PIO4, this driver doesn't support PIO5 and attempt
  to setup PIO5 by it821x_tuneproc() could result in incorrect PIO timings
  + incorrect base clock being set for controller in the passthrough mode

* move code limiting max PIO according to the pair device capabilities from
  config_it821x_chipset_for_pio() to it821x_tuneproc() so the check is also
  applied for mode change requests coming through ->tuneproc and ->speedproc
  interfaces

* set device speed in it821x_tuneproc()

* in it821x_tune_chipset() call it821x_tuneproc() also if the controller is
  in the smart mode (so the check for pair device max PIO is done)

* rename it821x_tuneproc() to it821x_tune_pio(), then add it821x_tuneproc()
  wrapper which does the max PIO mode check;  it worked by the pure luck
  previously, pio[4] and pio_want[4] arrays were used with index == 255
  so random PIO timings and base clock were set for the controller in the
  passthrough mode, thankfully PIO timings and base clock were corrected
  later by config_it821x_chipset_for_pio() call (but it was not called for
  PIO-only devices during resume and for user requested PIO autotuning)

* remove config_it821x_chipset_for_pio() call from config_chipset_for_dma()
  as the driver sets ->autotune to 1 and ->tuneproc does the proper job now

* convert the last user of config_it821x_chipset_for_pio() to use
  it821x_tuneproc(drive, 255) and remove no longer needed function

While at it:

* fix few comments

* bump driver version

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agopdc202xx_new: enable DMA for all ATAPI devices
Bartlomiej Zolnierkiewicz [Sat, 5 May 2007 20:03:50 +0000 (22:03 +0200)]
pdc202xx_new: enable DMA for all ATAPI devices

There is no reason to limit DMA to ide_cdrom type devices.

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoalim15x3: PIO fallback fix
Bartlomiej Zolnierkiewicz [Sat, 5 May 2007 20:03:50 +0000 (22:03 +0200)]
alim15x3: PIO fallback fix

If DMA tuning fails always set the best PIO mode.

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoaec62xx: fix PIO/DMA setup issues
Sergei Shtylyov [Sat, 5 May 2007 20:03:50 +0000 (22:03 +0200)]
aec62xx: fix PIO/DMA setup issues

Teach the driver's tuneproc() method to do PIO auto-runing properly since it
treated 5 instead of 255 as auto-tune request, and also passed the mode limit
of PIO5 to ide_get_best_pio_mode() despite supporting up to PIO4 only.

While at it, also:

- remove the driver's wrong claim about supporting SWDMA modes;

- stop hooking ide_dma_timeout() method as the handler clearly doesn't fit for
  the task...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agocmd64x: use interrupt status from MRDMODE register (take 2)
Sergei Shtylyov [Sat, 5 May 2007 20:03:50 +0000 (22:03 +0200)]
cmd64x: use interrupt status from MRDMODE register (take 2)

Fold the parts of the ide_dma_end() methods identical to __ide_dma_end() into a
mere call to it.
Start using faster versions of the ide_dma_end() and ide_dma_test_irq() methods
for the PCI0646U and newer chips that have the duplicate interrupt status bits
in the I/O mapped MRDMODE register, determing what methods to use at the driver
load time. Do some cleanup/renaming in the "old" ide_dma_test_irq() method too.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agocmd64x: procfs code fixes/cleanups (take 2)
Sergei Shtylyov [Sat, 5 May 2007 20:03:50 +0000 (22:03 +0200)]
cmd64x: procfs code fixes/cleanups (take 2)

Fix several issues with the driver's procfs output:

- when testing if channel is enabled, the code looks at the "simplex" bits, not
  at the real enable bits -- add #define for the primary channel enable bit;

- UltraDMA modes 0, 1, 3 for slave drive reported incorrectly due to using the
  master drive's clock cycle resolution bit.

While at it, also perform the following cleanups:

- don't print extra newline before the first controller's dump;

- correct the chipset names (from CMDxxx to PCI-xxx)

- don't read from the registers which aren't used for dump;

- better align the table column sizes;

- rework UltraDMA mode dump code;

- remove PIO mode dump code that has never been finished;

- remove the duplicate interrupt status (the MRDMODE register bits mirror those
  those in the CFR and ARTTIM23 registers) and fold the dump into single line;

- correct the style of the ?: operators...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agocmd64x: add/fix enablebits (take 2)
Sergei Shtylyov [Sat, 5 May 2007 20:03:49 +0000 (22:03 +0200)]
cmd64x: add/fix enablebits (take 2)

The IDE core looks at the wrong bit when checking if the secondary channel is
enabled on PCI0646 -- CNTRL register bit 7 is read-ahead disable, bit 3 is the
correct one.
Starting with PCI0646U chip, the primary channel can also be enabled/disabled --
so, add 'enablebits' initializers to each 'ide_pci_device_t' structure, handling
the original PCI0646 via adding the init_setup() method and clearing the 'reg'
field there if necessary...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agocmd64x: interrupt status fixes (take 2)
Sergei Shtylyov [Sat, 5 May 2007 20:03:49 +0000 (22:03 +0200)]
cmd64x: interrupt status fixes (take 2)

The driver's ide_dma_test_irq() method was reading the MRDMODE register even on
PCI0643/6 where it was write-only -- fix this by always reading the "backward-
compatible" interrupt bits, renaming dma_alt_stat to irq_stat as the interrupt
status bits are not coupled to DMA.
In addition, wrong interrupt bit was tested/cleared for the primary channel --
it's bit 2 in all the chip specs and the driver used bit 1... :-/

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agocmd64x: fix multiword and remove single-word DMA support
Sergei Shtylyov [Sat, 5 May 2007 20:03:49 +0000 (22:03 +0200)]
cmd64x: fix multiword and remove single-word DMA support

Fix the multiword DMA and drop the single-word DMA support (which nobody will
miss, I think).  In order to do it, a number of changes was necessary:

- rename program_drive_counts() to program_cycle_times(), pass to it cycle's
  total/active times instead of the clock counts, and convert them into the
  active/recovery clocks there instead of cmd64x_tune_pio() -- this causes
  quantize_timing() to also move;

- contrarywise, move all the code handling the address setup timing into
  cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup;

- remove from the speedproc() method the  bogus code pretending to set the DMA
  timings by twiddling bits in the BMIDE status register, handle setting MWDMA
  by just calling program_cycle_times(); while at it, improve the style of that
  whole switch statement;

- stop fiddling with the DMA capable bits in the speedproc() method -- they do
  not enable DMA, and are properly dealt with by the dma_host_{on,off} methods;

- don't set hwif->swdma_mask in the init_hwif() method anymore.

In addition to those changes, do the following:

- in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt
  status bit, eliminate local_irq_{save|restore}() around this code as there's
  *no* actual race with the interrupt handler, and move cmdprintk() to a more
  fitting place -- after ide_get_best_pio_mode() call;

- make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn;

- rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[];

- in  the speedproc() method, get rid of the duplicate reads/writes from/to the
  UDIDETCRx registers and of the extra variable used to store the transfer mode
  value after filtering,  use another method of determining master/slave drive,
  and cleanup useless parens;

- beautify cmdprintk() output here and there.

While at it, remove meaningless comment about the driver being used only on
UltraSPARC and long non-relevant RCS tag. :-)

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agosl82c105: DMA support code cleanup (take 4)
Sergei Shtylyov [Sat, 5 May 2007 20:03:49 +0000 (22:03 +0200)]
sl82c105: DMA support code cleanup (take 4)

Fold the now equivalent code in the ide_dma_check() method into a mere call to
ide_use_dma().  Make config_for_dma() return non-zero if DMA mode has been set
and call it from the ide_dma_check() method instead of ide_dma_on().
Defer writing the DMA timings to the chip registers until DMA is really turned
on (and do not enable IORDY for DMA).
Remove unneeded code from the init_hwif() method, improve its overall looks.
Rename the dma_start(), ide_dma_check(), and ide_dma_lostirq() methods, and
also use more proper hwif->dma_command, fix printk() and comment in the latter
one as well.  While at it, cleanup style in several places.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agosl82c105: rework PIO support (take 2)
Sergei Shtylyov [Sat, 5 May 2007 20:03:49 +0000 (22:03 +0200)]
sl82c105: rework PIO support (take 2)

Get rid of the 'pio_speed' member of 'ide_drive_t' that was only used by this
driver by storing the PIO mode timings in the 'drive_data' instead -- this
allows us to greatly  simplify the process of "reloading" of the chip's timing
register and do it right in sl82c150_dma_off_quietly() and to get rid of two
extra arguments to config_for_pio() -- which got renamed to sl82c105_tune_pio()
and now returns a PIO mode selected, with ide_config_drive_speed() call moved
into the tuneproc() method, now called sl82c105_tune_drive() with the code to
set drive's 'io_32bit' and 'unmask' flags in its turn moved to its proper place
in the init_hwif() method.
Also, while at it, rename get_timing_sl82c105() into get_pio_timings() and get
rid of the code in it clamping cycle counts to 32 which was both incorrect and
never executed anyway...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years ago[TG3]: Eliminate the TG3_FLAG_5701_REG_WRITE_BUG flag.
Matt Carlson [Sat, 5 May 2007 19:47:25 +0000 (12:47 -0700)]
[TG3]: Eliminate the TG3_FLAG_5701_REG_WRITE_BUG flag.

This patch removes the use of the TG3_FLAG_5701_REG_WRITE_BUG flag.
It's logic is only used to set a function pointer and thus the
logic can be collapsed and the flag removed.

[ Comment tidy by Christoph Hellwig. -DaveM ]

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
17 years ago[TG3]: Eliminate the TG3_FLAG_GOT_SERDES_FLOWCTL flag.
Michael Chan [Sat, 5 May 2007 19:11:21 +0000 (12:11 -0700)]
[TG3]: Eliminate the TG3_FLAG_GOT_SERDES_FLOWCTL flag.

This flag does not do anything useful.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[TG3]: Remove reset during MAC address changes.
Michael Chan [Sat, 5 May 2007 19:10:20 +0000 (12:10 -0700)]
[TG3]: Remove reset during MAC address changes.

The reset was added a while back so that ASF could re-init whatever
MAC address it wanted to use after the MAC address was changed.
Instead of resetting, we can just keep MAC address 1 unchanged during
MAC address changes if MAC address 1 is different from MAC address 0.

This fixes 2 problems:

1. Bonding calls set_mac_address in contexts that cannot sleep.
It no longer sleeps with the chip reset removed.

2. When ASF shares the same MAC address as the NIC, it needs to
always do that even when the MAC address is changed.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[TG3]: WoL fixes.
Gary Zambrano [Sat, 5 May 2007 18:52:19 +0000 (11:52 -0700)]
[TG3]: WoL fixes.

Change TG3_FLAG_SERDES_WOL_CAP to TG3_FLAG_WOL_CAP to make it easier
to manage WoL.  This flag is now used consistently during ethtool WoL
setup and power setting changes.

Signed-off-by: Gary Zambrano <zambrano@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[TG3]: Clear GPIO mask before storing.
Gary Zambrano [Sat, 5 May 2007 18:51:45 +0000 (11:51 -0700)]
[TG3]: Clear GPIO mask before storing.

The GPIO settings may change during reset and so the stored values in
tp->grc_local_ctrl should be cleared first.

Signed-off-by: Gary Zambrano <zambrano@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[TG3]: Improve NVRAM sizing.
Matt Carlson [Sat, 5 May 2007 18:51:05 +0000 (11:51 -0700)]
[TG3]: Improve NVRAM sizing.

This patch changes the NVRAM sizing procedure so that the driver can
take advantage of devices with 1:1 NVRAM strapping configurations.  This
is useful in cases where the traditional NVRAM sizing method fails.  In
the event that the flash size cannot be determined, the largest known
NVRAM size is used.  The patch also removes support for 5755 NVRAM
devices that are not supported by Broadcom and adds explicit sizing for
this device.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[TG3]: Fix TSO bugs.
Matt Carlson [Sat, 5 May 2007 18:50:04 +0000 (11:50 -0700)]
[TG3]: Fix TSO bugs.

1. Remove the check for skb->len greater than MTU when doing TSO.
When the destination has a smaller MSS than the source, a TSO packet
may be smaller than the MTU and we still need to process it as a TSO
packet.

2. On 5705A3 devices with TSO enabled, the DMA engine can hang due to a
hardware bug.  This patch avoids the hanging condition by reducing the
DMA burst size.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[MAC80211]: Add maintainers entry for mac80211.
Jiri Benc [Sat, 5 May 2007 18:47:08 +0000 (11:47 -0700)]
[MAC80211]: Add maintainers entry for mac80211.

Add MAINTAINERS entry for mac80211.

Signed-off-by: Jiri Benc <jbenc@suse.cz>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[MAC80211]: Add debugfs attributes.
Jiri Benc [Sat, 5 May 2007 18:46:38 +0000 (11:46 -0700)]
[MAC80211]: Add debugfs attributes.

Export various mac80211 internal variables through debugfs.

Signed-off-by: Jiri Benc <jbenc@suse.cz>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[MAC80211]: Add mac80211 wireless stack.
Jiri Benc [Sat, 5 May 2007 18:45:53 +0000 (11:45 -0700)]
[MAC80211]: Add mac80211 wireless stack.

Add mac80211, the IEEE 802.11 software MAC layer.

Signed-off-by: Jiri Benc <jbenc@suse.cz>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[MAC80211]: Add generic include/linux/ieee80211.h
Jiri Benc [Sat, 5 May 2007 18:43:04 +0000 (11:43 -0700)]
[MAC80211]: Add generic include/linux/ieee80211.h

Add generic IEEE 802.11 definitions.

Signed-off-by: Jiri Benc <jbenc@suse.cz>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETLINK]: Remove references to process ID
Herbert Xu [Sat, 5 May 2007 18:42:03 +0000 (11:42 -0700)]
[NETLINK]: Remove references to process ID

People treating the *_pid fields in netlink as a process ID has caused
endless confusion over the years.  The fact that our own netlink.h
does this only adds to the confusion.

So here is a patch to change the comments to refer to it as the port
ID which hopefully will make it clear what the purpose of the fields
really is.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[AF_IUCV]: Compile fix - adopt to skbuff changes.
Heiko Carstens [Sat, 5 May 2007 18:41:18 +0000 (11:41 -0700)]
[AF_IUCV]: Compile fix - adopt to skbuff changes.

From: Heiko Carstens <heiko.carstens@de.ibm.com>

  CC [M]  net/iucv/af_iucv.o
net/iucv/af_iucv.c: In function `iucv_fragment_skb':
net/iucv/af_iucv.c:984: error: structure has no member named `h'
net/iucv/af_iucv.c:985: error: structure has no member named `nh'
net/iucv/af_iucv.c:988: error: incompatible type for argument 1 of
`skb_queue_tail'

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc
Linus Torvalds [Sat, 5 May 2007 04:44:34 +0000 (21:44 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc: (46 commits)
  mmc-omap: Clean up omap set_ios and make MMC_POWER_ON work
  mmc-omap: Fix omap to use MMC_POWER_ON
  mmc-omap: add missing '\n'
  mmc: make tifm_sd_set_dma_data() static
  mmc: remove old card states
  mmc: support unsafe resume of cards
  mmc: separate out reading EXT_CSD
  mmc: break apart switch function
  MMC: Fix handling of low-voltage cards
  MMC: Consolidate voltage definitions
  mmc: add bus handler
  wbsd: check for data opcode earlier
  mmc: Separate out protocol ops
  mmc: Move core functions to subdir
  mmc: deprecate mmc bus topology
  mmc: remove card upon suspend
  mmc: allow suspended block driver to be removed
  mmc: Flush pending detects on host removal
  mmc: Move host and card drivers to subdirs
  mmc: Move queue functions to mmc_block
  ...

17 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfashe...
Linus Torvalds [Sat, 5 May 2007 03:44:54 +0000 (20:44 -0700)]
Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2:
  ocfs2: Force use of GFP_NOFS in ocfs2_write()
  ocfs2: fix sparse warnings in fs/ocfs2/cluster
  ocfs2: fix sparse warnings in fs/ocfs2/dlm
  ocfs2: fix sparse warnings in fs/ocfs2
  [PATCH] Copy i_flags to ocfs2 inode flags on write
  [PATCH] ocfs2: use __set_current_state()
  ocfs2: Wrap access of directory allocations with ip_alloc_sem.
  [PATCH] fs/ocfs2/: make 3 functions static
  ocfs2: Implement compat_ioctl()

17 years agoiomap: make the default iomap functions fail softer
Linus Torvalds [Sat, 5 May 2007 03:44:23 +0000 (20:44 -0700)]
iomap: make the default iomap functions fail softer

We used to BUG_ON() for a badly mapped IO port, which is certainly
correct, but actually made it harder to debug the case where the ATA
drivers had incorrectly mapped a nonconnected ATA port.

So make badly mapped ports trigger a WARN_ON(), and throw the IO away
instead (and return all ones for reads).  For things like broken driver
initialization - which is the most likely cause anyway - that should
mean that the machine comes up and is usable (at least that was the case
for the ATA breakage that triggered this patch).

It tends to be a whole lot easier to do a "dmesg" on a working machine
than to try to capture logs off a dead one.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[CIFS] Make sec=none force an anonymous mount
Jeff Layton [Sat, 5 May 2007 03:27:49 +0000 (03:27 +0000)]
[CIFS] Make sec=none force an anonymous mount

We had a customer report that attempting to make CIFS mount with a null
username (i.e. doing an anonymous mount) doesn't work. Looking through the
code, it looks like CIFS expects a NULL username from userspace in order
to trigger an anonymous mount. The mount.cifs code doesn't seem to ever
pass a null username to the kernel, however.

It looks also like the kernel can take a sec=none option, but it only seems
to look at it if the username is already NULL. This seems redundant and
effectively makes sec=none useless.

The following patch makes sec=none force an anonymous mount.

Signed-off-by: Steve French <sfrench@us.ibm.com>
17 years agoMerge git://git.linux-nfs.org/pub/linux/nfs-2.6
Linus Torvalds [Sat, 5 May 2007 02:55:11 +0000 (19:55 -0700)]
Merge git://git.linux-nfs.org/pub/linux/nfs-2.6

* git://git.linux-nfs.org/pub/linux/nfs-2.6: (28 commits)
  NFS: Fix a compile glitch on 64-bit systems
  NFS: Clean up nfs_create_request comments
  spkm3: initialize hash
  spkm3: remove bad kfree, unnecessary export
  spkm3: fix spkm3's use of hmac
  NFS4: invalidate cached acl on setacl
  NFS: Fix directory caching problem - with test case and patch.
  NFS: Set meaningful value for fattr->time_start in readdirplus results.
  NFS: Added support to turn off the NFSv3 READDIRPLUS RPC.
  SUNRPC: RPC client should retry with different versions of rpcbind
  SUNRPC: remove old portmapper
  NFS: switch NFSROOT to use new rpcbind client
  SUNRPC: switch the RPC server to use the new rpcbind registration API
  SUNRPC: switch socket-based RPC transports to use rpcbind
  SUNRPC: introduce rpcbind: replacement for in-kernel portmapper
  SUNRPC: Eliminate side effects from rpc_malloc
  SUNRPC: RPC buffer size estimates are too large
  NLM: Shrink the maximum request size of NLM4 requests
  NFS: Use pgoff_t in structures and functions that pass page cache offsets
  NFS: Clean up nfs_sync_mapping_wait()
  ...

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Sat, 5 May 2007 02:36:58 +0000 (19:36 -0700)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6: (49 commits)
  [SCTP]: Set assoc_id correctly during INIT collision.
  [SCTP]: Re-order SCTP initializations to avoid race with sctp_rcv()
  [SCTP]: Fix the SO_REUSEADDR handling to be similar to TCP.
  [SCTP]: Verify all destination ports in sctp_connectx.
  [XFRM] SPD info TLV aggregation
  [XFRM] SAD info TLV aggregationx
  [AF_RXRPC]: Sort out MTU handling.
  [AF_IUCV/IUCV] : Add missing section annotations
  [AF_IUCV]: Implementation of a skb backlog queue
  [NETLINK]: Remove bogus BUG_ON
  [IPV6]: Some cleanups in include/net/ipv6.h
  [TCP]: zero out rx_opt in tcp_disconnect()
  [BNX2]: Fix TSO problem with small MSS.
  [NET]: Rework dev_base via list_head (v3)
  [TCP] Highspeed: Limited slow-start is nowadays in tcp_slow_start
  [BNX2]: Update version and reldate.
  [BNX2]: Print bus information for PCIE devices.
  [BNX2]: Add 1-shot MSI handler for 5709.
  [BNX2]: Restructure PHY event handling.
  [BNX2]: Add indirect spinlock.
  ...