GitHub/exynos8895/android_kernel_samsung_universal8895.git
18 years agoieee1394: sbp2: don't prefer MODE SENSE 10
Stefan Richter [Wed, 6 Sep 2006 23:06:00 +0000 (01:06 +0200)]
ieee1394: sbp2: don't prefer MODE SENSE 10

In the old days, sbp2 used to coerce all MODE SENSE commands into the
10 bytes version.  When all command set conversions were removed from
sbp2 several months ago, sdev->use_10_for_ms = 1 was added.  Meaning,
higher SCSI layers preferred the 10 bytes version but would try the 6
bytes version if the former failed.

Recently, a problem with the 10 bytes version was discovered.  An Initio
INIC-1530 firmware accepted the 10 bytes version but replied with bogus
data, showing the HDD incorrectly as write-protected.  Since RBC
actually mandates MODE SENSE (6), I checked which version was sent by
Windows XP and Mac OS X 10.3 to an SBP-2 target hosted by Linux --- it
was the 6 bytes version.  (Exception: OS X sent the 10 bytes version to
an MMC target.  RBC and SBC got MODE SENSE (6).)

Therefore, drop the use_10_for_ms flag from sbp2.  Now the upper layers
will try MODE SENSE (6) before MODE SENSE (10) on all SBP-2 devices.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: nodemgr: grab class.subsys.rwsem in nodemgr_resume_ne
Stefan Richter [Wed, 6 Sep 2006 17:04:00 +0000 (19:04 +0200)]
ieee1394: nodemgr: grab class.subsys.rwsem in nodemgr_resume_ne

nodemgr_resume_ne was iterating over nodemgr_ud_class.children without
protection by nodemgr_ud_class.subsys.rwsem.

FIXME:
Shouldn't we rather use class->sem there, not class->subsys.rwsem?

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: nodemgr: fix rwsem recursion
Stefan Richter [Wed, 6 Sep 2006 17:04:00 +0000 (19:04 +0200)]
ieee1394: nodemgr: fix rwsem recursion

nodemgr_update_pdrv grabbed an rw semaphore (as reader) which was
already taken by its caller's caller, nodemgr_probe_ne (as reader too).
Reported by Miles Lane, call path pointed out by Arjan van de Ven.

FIXME:
Shouldn't we rather use class->sem there, not class->subsys.rwsem?

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: more help in Kconfig
Stefan Richter [Wed, 30 Aug 2006 16:22:00 +0000 (18:22 +0200)]
ieee1394: sbp2: more help in Kconfig

Add some pointers to SCSI to the configuration menu item of sbp2.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: prevent rare deadlock in shutdown
Stefan Richter [Mon, 14 Aug 2006 16:59:00 +0000 (18:59 +0200)]
ieee1394: sbp2: prevent rare deadlock in shutdown

Scsi_remove_device() may go into uninterruptible sleep if blocked.
Therefore sbp2_remove() unblocks the Scsi_Host before the device is
requested to be removed.  But there could be another 1394 bus reset
after that which would block the host again.  The 1394 subsystem won't
call sbp2_update() concurrently to sbp2_remove(), which is why there is
no chance for sbp2_remove() to be unblocked by sbp2_update().

The fix is to tell sbp2's bus reset handler when a device is to be shut
down so that it skips scsi_block_requests() on that host.  As before,
any new commands after a reset without reconnect will be failed quickly
by sbp2scsi_queuecommand().

In the long term, means to go without scsi_block_requests() should be
found.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: update includes
Stefan Richter [Mon, 14 Aug 2006 16:56:00 +0000 (18:56 +0200)]
ieee1394: sbp2: update includes

Remove unused includes.  Add missing includes, i.e. explicitly include
all used headers.  Sort includes alphabetically.  Replace one call to
signal_pending(current) to avoid to include headers just for this line.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: better handling of transport errors
Stefan Richter [Mon, 14 Aug 2006 16:51:00 +0000 (18:51 +0200)]
ieee1394: sbp2: better handling of transport errors

If the target signals a transport failure via status block, complete the
request with DID_BUSY to indicate to the SCSI subsystem that the command
may succeed when retried.

Also log diagnostic information if the status block shows a transport
related problem.  It may point to hardware faults.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: recheck node generation in sbp2_update
Stefan Richter [Mon, 14 Aug 2006 16:46:00 +0000 (18:46 +0200)]
ieee1394: sbp2: recheck node generation in sbp2_update

While sbp2_update() is doing its duties after a bus reset, another reset
could happen.  Don't accept new requests until the next undisturbed
sbp2_update() or until sbp2_remove().

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: safer agent reset in error handlers
Stefan Richter [Mon, 14 Aug 2006 16:44:00 +0000 (18:44 +0200)]
ieee1394: sbp2: safer agent reset in error handlers

The scsi_host_template's eh_abort_handler and eh_device_reset_handler
are allowed to sleep.  Use this to run sbp2_agent_reset in the more
reliable mode which returns _after_ its write transaction was finished.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: handle "sbp2util_node_write_no_wait failed"
Stefan Richter [Mon, 14 Aug 2006 16:43:00 +0000 (18:43 +0200)]
ieee1394: sbp2: handle "sbp2util_node_write_no_wait failed"

Fix for http://bugzilla.kernel.org/show_bug.cgi?id=6948

Because sbp2 writes to the target's fetch agent's registers from within
atomic context, it cannot sleep to guaranteedly get a free transaction
label.  This may repeatedly lead to "sbp2util_node_write_no_wait failed"
and consequently to SCSI command abortion after timeout.  A likely cause
is that many queue_command softirqs may occur before khpsbpkt (the
ieee1394 driver's thread which cleans up after finished transactions) is
woken up to recycle tlabels.

Sbp2 now schedules a workqueue job whenever sbp2_link_orb_command fails
in sbp2util_node_write_no_wait.  The job will reliably get a transaction
label because it can sleep.

We use the kernel-wide shared workqueue because it is unlikely that the
job itself actually needs to sleep.  In the improbable case that it has
to sleep, it doesn't need to sleep long since the standard transaction
timeout is 100ms.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoCONFIG_PM=n slim: drivers/ieee1394/ohci1394.c
Alexey Dobriyan [Fri, 11 Aug 2006 23:46:14 +0000 (16:46 -0700)]
CONFIG_PM=n slim: drivers/ieee1394/ohci1394.c

Remove some code which is unneeded if CONFIG_PM=n.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: safer definition of empty macros
Stefan Richter [Wed, 2 Aug 2006 16:44:00 +0000 (18:44 +0200)]
ieee1394: safer definition of empty macros

A deactivated macro, defined as "#define foo(bar)", will result in
silent corruption if somebody forgets a semicolon after a call to foo.
Replace it by "#define foo(bar) do {} while (0)" which will reveal any
respective syntax errors.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agovideo1394: add poll file operation support
David Moore [Tue, 1 Aug 2006 23:00:00 +0000 (01:00 +0200)]
video1394: add poll file operation support

This patch adds support for the poll file operation to the video1394
driver.

Signed-off-by: David Moore <dcm@acm.org>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agothe scheduled removal of drivers/ieee1394/sbp2.c:force_inquiry_hack
Adrian Bunk [Sat, 29 Jul 2006 17:48:28 +0000 (19:48 +0200)]
the scheduled removal of drivers/ieee1394/sbp2.c:force_inquiry_hack

This patch contains the scheduled removal of the force_inquiry_hack
module parameter.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoInitialize ieee1394 early when built in
Andi Kleen [Thu, 27 Jul 2006 19:54:00 +0000 (21:54 +0200)]
Initialize ieee1394 early when built in

This makes debugging with firescope easier.

Signed-off-by: Andi Kleen <ak@suse.de> (original patch)
Update:
 - no need for #ifdef MODULE
 - add comment in ieee1394_core, more verbose comment in ohci1394

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> (update)
18 years agoieee1394: sbp2: convert sbp2util_down_timeout to waitqueue
Stefan Richter [Sun, 23 Jul 2006 20:19:00 +0000 (22:19 +0200)]
ieee1394: sbp2: convert sbp2util_down_timeout to waitqueue

The waitqueue API is used to replace a custom wait mechanism.  Only one
global waitqueue (instead of per-device waitqueues or completions) is
added because there is usually just one waiter.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: more checks of status block
Stefan Richter [Sun, 23 Jul 2006 20:18:00 +0000 (22:18 +0200)]
ieee1394: sbp2: more checks of status block

 - Add checks for the (very unlikely) cases that the target writes too
   little or too much status data or writes unsolicited status.
 - Indicate that these and similar conditions are unlikely().
 - Check the 'resp' and 'sbp_status' fields for possible failure status.
 - Slightly optimize access macros for the status block bitfields.
 - Unify a few related log messages.

TODO:  Check if 'src'==1, then withhold the respective ORB from reuse
until status for any subsequent ORB was received.  This is an old bug
whose fix requires more complex command queue handling.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: safer initialization of status fifo
Stefan Richter [Sun, 23 Jul 2006 20:16:00 +0000 (22:16 +0200)]
ieee1394: sbp2: safer initialization of status fifo

Sbp2's copy of the status fifo was cleared when management ORBs or new
command ORBs were prepared.  The latter had potential for a race
condition if the block layer's soft IRQ and the 1394 LLD's interrupt
handler ran on different CPUs.  It would also yield wrong status if a
command was completed with non-zero completion status before other
commands that had zero completion status, and no new command was
enqueued in the meantime.

Now, the status buffer is cleared right before it is written.  Thus it
ends up in the following simpler and safer access pattern:
 - sbp2_alloc_device: allocates and implicitly clears once,
 - sbp2_handle_status_write: clears, writes, and reads,
 - sbp2_query_logins, sbp2_login_device, sbp2_reconnect_device: read.
The latter three do not race with sbp2_handle_status_write because of
how the protocol works.

As a tiny optimization, the first two quadlets of the status never need
to be cleared.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: optimize DMA direction of command ORBs
Stefan Richter [Sun, 23 Jul 2006 20:57:00 +0000 (22:57 +0200)]
ieee1394: sbp2: optimize DMA direction of command ORBs

Only the driver writes ORBs, the device just reads them.  Therefore
PCI_DMA_BIDIRECTIONAL can be replaced by PCI_DMA_TODEVICE which may be
cheaper on some architectures.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: discard return value of sbp2_link_orb_command
Stefan Richter [Sun, 23 Jul 2006 20:10:00 +0000 (22:10 +0200)]
ieee1394: sbp2: discard return value of sbp2_link_orb_command

Since sbp2 is at the moment unable to do anything with the return value
of sbp2_link_orb_command, just discard it.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: sbp2: safer last_orb and next_ORB handling
Stefan Richter [Sun, 23 Jul 2006 20:12:00 +0000 (22:12 +0200)]
ieee1394: sbp2: safer last_orb and next_ORB handling

The sbp2 initiator has two ways to tell a target's fetch agent about new
command ORBs:
 - Write the ORB's address to the ORB_POINTER register.  This must not
   be done while the fetch agent is active.
 - Put the ORB's address into the previously submitted ORB's next_ORB
   field and write to the DOORBELL register.  This may be done while the
   fetch agent is active or suspended.  It must not be done while the
   fetch agent is in reset state.
Sbp2 has a last_orb pointer which indicates in what way a new command
should be announced.  That pointer is concurrently accessed at various
occasions.  Furthermore, initiator and target are accessing the next_ORB
field of ORBs concurrently and asynchronously.

This patch does:
 - Protect all initiator accesses to last_orb by sbp2_command_orb_lock.
 - Add pci_dma_sync_single_for_device before a previously submitted
   ORB's next_ORB field is overwritten.
 - Insert a memory barrier between when next_ORB_lo and next_ORB_hi are
   overwritten.  Next_ORB_hi must not be updated before next_ORB_lo.
 - Remove the rather unspecific and now superfluous qualifier "volatile"
   from the next_ORB fields.
 - Add comments on how last_orb is connected with what is known about
   the target's fetch agent's state.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: remove #include <asm/semaphore.h>
Stefan Richter [Sun, 23 Jul 2006 20:01:00 +0000 (22:01 +0200)]
ieee1394: remove #include <asm/semaphore.h>

These includes in ieee1394_core and eth1394 are obsolete.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: shrink tlabel pools, remove tpool semaphores
Stefan Richter [Sun, 2 Jul 2006 12:17:00 +0000 (14:17 +0200)]
ieee1394: shrink tlabel pools, remove tpool semaphores

This patch reduces the size of struct hpsb_host and also removes
semaphores from ieee1394_transactions.c.  On i386, struct hpsb_host
shrinks from 10656 bytes to 6688 bytes.  This is accomplished by
 - using a single wait_queue for hpsb_get_tlabel instead of many
   instances of semaphores,
 - using a single lock to serialize access to all tlabel pools (the
   protected code regions are small, i.e. lock contention very low),
 - omitting the sysfs attribute tlabels_allocations.

Drawback:  In the rare case that a process needs to sleep because all
transaction labels for the node are temporarily exhausted, it is also
woken up if a tlabel for a different node became free, checks for an
available tlabel, and is put to sleep again.  The check is not costly
and the situation occurs extremely rarely.  (Tlabels are typically
only exhausted if there was no context switch to the khpsbpkt thread
which recycles tlables.)  Therefore the benefit of reduced tpool size
outweighs this drawback.

The sysfs attributes tlabels_free and tlabels_mask are not compiled
anymore unless CONFIG_IEEE1394_VERBOSEDEBUG is set.

The by far biggest member of struct hpsb_host, the struct csr_control
csr (5272 bytes on i386), is now placed at the end of struct hpsb_host.

Note, hpsb_get_tlabel calls the macro wait_event_interruptible with a
condition argument which has a side effect (allocation of a tlabel and
manipulation of the packet).  This side effect happens only if the
condition is true.  The patch relies on wait_event_interruptible not
evaluating the condition again after it became true.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoieee1394: merge from Linus
Stefan Richter [Sun, 17 Sep 2006 16:17:19 +0000 (18:17 +0200)]
ieee1394: merge from Linus

Conflicts: drivers/ieee1394/hosts.c
Patch "lockdep: annotate ieee1394 skb-queue-head locking" was meddling
with patch "ieee1394: fix kerneldoc of hpsb_alloc_host".

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
18 years agoMerge branch 'master' into upstream-fixes
Jeff Garzik [Sun, 17 Sep 2006 05:05:43 +0000 (01:05 -0400)]
Merge branch 'master' into upstream-fixes

18 years ago[PATCH] s390: qeth driver fixes [6/6]
Frank Pavlic [Fri, 15 Sep 2006 14:27:02 +0000 (16:27 +0200)]
[PATCH] s390: qeth driver fixes [6/6]

[PATCH 9/9] s390: qeth driver fixes [6/6]

From: Frank Pavlic <fpavlic@de.ibm.com>
- Hipersockets has no IPV6 support, thus prevent issueing
  SETRTG_IPV6 control commands on Hipersockets devices.
- fixed error handling in qeth_sysfs_(un)register

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] s390: qeth driver fixes [5/6]
Frank Pavlic [Fri, 15 Sep 2006 14:26:52 +0000 (16:26 +0200)]
[PATCH] s390: qeth driver fixes [5/6]

[PATCH 8/9] s390: qeth driver fixes [5/6]

From: Frank Pavlic <fpavlic@de.ibm.com>
fix kernel panic in qdio queue handling.
qeth_qdio_clear_card() could be invoked by 2 CPUs
simultaneously (for example reboot event and recovery).

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] s390: qeth driver fixes [4/6]
Frank Pavlic [Fri, 15 Sep 2006 14:26:34 +0000 (16:26 +0200)]
[PATCH] s390: qeth driver fixes [4/6]

[PATCH 7/9] s390: qeth driver fixes [4/6]

From: Frank Pavlic <fpavlic@de.ibm.com>
- fix kernel crash due to race,
  set card->state to SOFTSETUP after
  card and card->dev are initialized properly.
- remove CONFIG_QETH_PERF_STATS, use sysfs attribute instead,
  as we want to have the ability to turn on/off the
  statistics at runtime.

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] s390: qeth driver fixes [3/6]
Frank Pavlic [Fri, 15 Sep 2006 14:26:19 +0000 (16:26 +0200)]
[PATCH] s390: qeth driver fixes [3/6]

[PATCH 6/9] s390: qeth driver fixes [3/6]

From: Frank Pavlic <fpavlic@de.ibm.com>
        fixed kernel panic caused by qeth driver:
        Using a bonding device qeth driver will realloc
        headroom for every skb coming from the bond device.
        Once this happens qeth frees the original skb and
        set the skb pointer to the new realloced skb.
        Under heavy transmit workload (e.g.UDP streams) through bond
        network device the qdio output queue might get full.
        In this case we return with EBUSY from qeth_send_packet.
        Returning to qeth_hard_start_xmit routine
        the skb address on the stack still points to the old address,
        which has been freed before.
        Returning from qeth_hard_start_xmit with EBUSY results in
        requeuing the skb. In this case it corrupts the qdisc queue
        and results in kernel panic.

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] s390: qeth driver fixes [2/6]
Frank Pavlic [Fri, 15 Sep 2006 14:26:07 +0000 (16:26 +0200)]
[PATCH] s390: qeth driver fixes [2/6]

[PATCH 5/9] s390: qeth driver fixes [2/6]

From: Frank Pavlic <fpavlic@de.ibm.com>
- fixed error handling in create_device_attributes
- fixed some minor bugs in IPv4
  and IPv6 address checking

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] s390: qeth driver fixes [1/6]
Frank Pavlic [Fri, 15 Sep 2006 14:25:56 +0000 (16:25 +0200)]
[PATCH] s390: qeth driver fixes [1/6]

[PATCH 4/9] s390: qeth driver fixes [1/6]

From: Frank Pavlic <fpavlic@de.ibm.com>
- Drop incoming packets with vlan_tag set
          if card->vlangrp is not set.
        - use always vlan_hwaccel_rx to pass
  vlan frames to the stack.
        - fix recovery problem. Device was recovered
  properly but still not working.
  netif_carrier_on call right before
          recovery start fixes it.

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] s390: Makefile cleanup
Frank Pavlic [Fri, 15 Sep 2006 14:25:39 +0000 (16:25 +0200)]
[PATCH] s390: Makefile cleanup

[PATCH 3/9] s390: Makefile cleanup

From: Frank Pavlic <fpavlic@de.ibm.com>
  remove CONFIG_MPC from Makefile which was
introduced accidently in the past.

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] s390: netiucv driver fixes
Frank Pavlic [Fri, 15 Sep 2006 14:25:19 +0000 (16:25 +0200)]
[PATCH] s390: netiucv driver fixes

[PATCH 2/9] s390: netiucv driver fixes

From: Frank Pavlic <fpavlic@de.ibm.com>
- missing lock initialization added
        - avoid duplicate iucv-interfaces to the same peer
        - rw-lock added for manipulating the list of
          defined iucv connections

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] s390: minor s390 network driver fixes
Frank Pavlic [Fri, 15 Sep 2006 14:25:03 +0000 (16:25 +0200)]
[PATCH] s390: minor s390 network driver fixes

Hi Jeff,
this is a RESEND of the nine s390 network driver patches.
I finally found  that my kmail corrupted almost every patch
I sent the last time. Please apply these 9 patches and forget
about my first attempt! Sorry for the delay, I had some fights
with sendmail, IMAP and mutt configuration.

Frank

[RESEND PATCH 1/9] s390: minor s390 network driver fixes

From: Frank Pavlic <fpavlic@de.ibm.com>
- iucv driver:
          use do { } while (0) constructs
  instead of empty defines to avoid compile bugs.
        - ctc driver:
          missing lock initialization added
        - lcs driver:
          BUG_ON usage was removed accidently
  with the last lcs patch.
          Put them back in place.

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] i386/x86-64: Add core 2 to oprofile
Benjamin LaHaise [Sat, 16 Sep 2006 23:35:46 +0000 (16:35 -0700)]
[PATCH] i386/x86-64: Add core 2 to oprofile

Add the CPU identification needed by oprofile for Intel (r) Core (tm) 2
CPUs.

Signed-off-by: Benjamin LaHaise <benjamin.c.lahaise@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: "Arun Sharma" <aruns@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years agoMerge branch 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband
Linus Torvalds [Sun, 17 Sep 2006 04:35:15 +0000 (21:35 -0700)]
Merge branch 'for-linus' of /linux/kernel/git/roland/infiniband

* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband:
  RDMA/cma: Increase the IB CM retry count in CMA
  IPoIB: Retry failed send-only multicast group joins
  IB/srp: Don't schedule reconnect from srp

18 years ago[PATCH] JFFS2: SUMMARY: fix a summary collecting bug
Zoltan Sogor [Sat, 16 Sep 2006 19:15:59 +0000 (12:15 -0700)]
[PATCH] JFFS2: SUMMARY: fix a summary collecting bug

In some special case (padding because of sync or umount) it can be possible
that summary information is not fit to the end of the erase block.  In
these cases the collecting of summary is disabled for this erase block.

The problem was that this was not respected by jffs2_sum_add_kvec().  This
patch fix this bug.

Signed-off-by: Ferenc Havasi <havasi@inf.u-szeged.hu>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] MTD NAND: OOB buffer offset fixups
Vitaly Wool [Sat, 16 Sep 2006 19:15:58 +0000 (12:15 -0700)]
[PATCH] MTD NAND: OOB buffer offset fixups

In the case of data-pad-ecc-pad-data...  layout the oob start position has
to be sizeof(data) in nand_write_oob_syndrom().

In nand_fill_oob() we need to copy to buf + buffer offset instead of buf +
write offset.

Signed-off-by: Vitaly Wool <vwool@ru.mvista.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] ext3 sequential read regression fix
Suparna Bhattacharya [Sat, 16 Sep 2006 19:15:58 +0000 (12:15 -0700)]
[PATCH] ext3 sequential read regression fix

ext3-get-blocks support caused ~20% degrade in Sequential read
performance (tiobench). Problem is with marking the buffer boundary
so IO can be submitted right away. Here is the patch to fix it.

  2.6.18-rc6:
  -----------
  # ./iotest
  1048576+0 records in
  1048576+0 records out
  4294967296 bytes (4.3 GB) copied, 75.2726 seconds, 57.1 MB/s

  real    1m15.285s
  user    0m0.276s
  sys     0m3.884s

  2.6.18-rc6 + fix:
  -----------------
  [root@elm3a241 ~]# ./iotest
  1048576+0 records in
  1048576+0 records out
  4294967296 bytes (4.3 GB) copied, 62.9356 seconds, 68.2 MB/s

The boundary block check in ext3_get_blocks_handle needs to be adjusted
against the count of blocks mapped in this call, now that it can map
more than one block.

Signed-off-by: Suparna Bhattacharya <suparna@in.ibm.com>
Tested-by: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] bug fix in kernel/kmod.c
Kenneth Lee [Sat, 16 Sep 2006 19:15:55 +0000 (12:15 -0700)]
[PATCH] bug fix in kernel/kmod.c

I think there is a bug in kmod.c: In __call_usermodehelper(), when
kernel_thread(wait_for_helper, ...) return success, since wait_for_helper()
might call complete() at any time, the sub_info should not be used any
more.

Normally wait_for_helper() take a long time to finish, you may not get
problem for most of the case.  But if you remove /sbin/modprobe, it may
become easier for you to get a oop in khelper.

Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] headers_check: use a different default directory
David Woodhouse [Sat, 16 Sep 2006 19:15:55 +0000 (12:15 -0700)]
[PATCH] headers_check: use a different default directory

`make headers_check' wants to go and write stuff in /lib/modules, which
requires root, whic is unfortunate.

In fact, there's no _particular_ reason for headers_install to put it there
either -- it can go into a subdirectory of the build tree in both cases.
It's not intended to go directly into /usr/include, which is why we didn't
put it there -- and we certainly don't want people screwing around with
symlinking to it.  It's for distributors to take away and do stuff with, so
leaving it in $(objtree) is fine, even in the headers_install case.

I picked $(objtree)/usr/include but I have no _particular_ preference
for that; it just seemed reasonable.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Kconfig: move CONFIG_EMBEDDED options to submenu
Chuck Ebbert [Sat, 16 Sep 2006 19:15:53 +0000 (12:15 -0700)]
[PATCH] Kconfig: move CONFIG_EMBEDDED options to submenu

Fix two problems with the CONFIG_EMBEDDED submenu:

(1) The menu was split in two by the rt_mutex patch, which moved
    half the items into the "General setup" menu.

(2) CONFIG_SYSCTL and CONFIG_UID16 were added to the main menu
    instead of the submenu.

Signed-off-by: Chuck Ebbert <76306.1226@compuserve.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Fix 'make headers_check' for Alpha
David Woodhouse [Sat, 16 Sep 2006 19:15:49 +0000 (12:15 -0700)]
[PATCH] Fix 'make headers_check' for Alpha

Alpha currently fails 'make headers_check' in the 2.6.18-rc kernels.  This
patch fixes it, by moving the existing #ifdef __KERNEL__ in asm/page.h so that
it covers everything that userspace shouldn't so, and by adding asm/compiler.h
to the list of exported files so that its use within asm/byteorder.h is
successful.

[ Note that at least with GCC 4, <linux/compiler.h> doesn't do the forced
  inlining about which there are nasty comments (and a workaround) in
  <asm/compiler.h>, unless you set CONFIG_FORCED_INLINING.  Rather than keep
  the mess you have in <asm/compiler.h> you could perhaps just make sure
  CONFIG_FORCED_INLINING=n is also honoured with GCC3, and make sure it cannot
  be set for Alpha? ]

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Fix 'make headers_check' on x86_64
David Woodhouse [Sat, 16 Sep 2006 19:15:48 +0000 (12:15 -0700)]
[PATCH] Fix 'make headers_check' on x86_64

On Tue, 2006-09-12 at 17:44 +0100, David Woodhouse wrote:
> asm-x86_64/elf.h requires asm/processor.h, which does not exist
> asm-x86_64/signal.h requires linux/linkage.h, which does not exist
> asm-x86_64/unistd.h requires linux/linkage.h, which does not exist
> asm-x86_64/vsyscall.h requires linux/seqlock.h, which does not exist

Again, move stuff which shouldn't be visible inside (mostly already existing)
#ifdef __KERNEL__.

This fixes a bunch of mislabelled and unlabelled #endifs in unistd.h and also
cleans that up to conform with what's visible on other architectures, since
the minimal fix for the error reported about would have involved a more
intrusive patch, renesting other ifdefs.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Fix 'make headers_check' on i386
David Woodhouse [Sat, 16 Sep 2006 19:15:48 +0000 (12:15 -0700)]
[PATCH] Fix 'make headers_check' on i386

This brings i386 asm/unistd.h into consistency with other architectures by not
exporting functionality which is not necessary.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Fix 'make headers_check' on ia64
David Woodhouse [Sat, 16 Sep 2006 19:15:47 +0000 (12:15 -0700)]
[PATCH] Fix 'make headers_check' on ia64

On Tue, 2006-09-12 at 17:44 +0100, David Woodhouse wrote:
> asm-ia64/ptrace.h requires asm/asm-offsets.h, which does not exist
> asm-ia64/resource.h requires asm/ustack.h, which does not exist

Hide parts which shouldn't be visible to userspace.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Fix 'make headers_check' on s390
David Woodhouse [Sat, 16 Sep 2006 19:15:46 +0000 (12:15 -0700)]
[PATCH] Fix 'make headers_check' on s390

On Tue, 2006-09-12 at 17:44 +0100, David Woodhouse wrote:
> asm-s390/debug.h requires linux/string.h, which does not exist
> asm-s390/elf.h requires asm/system.h, which does not exist

Move things around slightly so the right things end up within
#ifdef __KERNEL__ and thus don't pollute the exported headers.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Fix 'make headers_check' on biarch architectures
David Woodhouse [Sat, 16 Sep 2006 19:15:45 +0000 (12:15 -0700)]
[PATCH] Fix 'make headers_check' on biarch architectures

We generate an <asm/foo.h> which includes either <asm-$ARCH/foo.h> or
<asm-$ALTARCH/foo.h> as appropriate.  But we were doing this dependent on
whether the file in question existed in the _unexported_ tree, not the
exported tree.  So if a file was exported to userspace in one asm- directory
but not the other, the generated file in asm/ was incorrect.

This only changed the failure mode if it _was_ included from a nice #error to
a less explicable #include failure -- but it also gave false errors in 'make
headers_check' output.  Fix it by looking in the right place instead.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Add symbol type files (*.symtypes) to .gitignore
Josh Triplett [Sat, 16 Sep 2006 19:15:44 +0000 (12:15 -0700)]
[PATCH] Add symbol type files (*.symtypes) to .gitignore

The kernel build system supports making symbol type files (*.symtypes) from C
source files.  Add these files to .gitignore.

Signed-off-by: Josh Triplett <josh@freedesktop.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Add mixed source and assembly listings (*.lst) to .gitignore
Josh Triplett [Sat, 16 Sep 2006 19:15:43 +0000 (12:15 -0700)]
[PATCH] Add mixed source and assembly listings (*.lst) to .gitignore

The kernel build system supports making mixed source and assembly listings
(*.lst) from C source files.  Add these files to .gitignore.

Signed-off-by: Josh Triplett <josh@freedesktop.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Add preprocessed files (*.i) to .gitignore
Josh Triplett [Sat, 16 Sep 2006 19:15:42 +0000 (12:15 -0700)]
[PATCH] Add preprocessed files (*.i) to .gitignore

The kernel build system supports making preprocessed files (*.i) from C source
files.  Add these files to .gitignore.

Signed-off-by: Josh Triplett <josh@freedesktop.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] IPMI: fix handling of OEM flags
Corey Minyard [Sat, 16 Sep 2006 19:15:41 +0000 (12:15 -0700)]
[PATCH] IPMI: fix handling of OEM flags

If one of the OEM flags becomes set in the flags from the hardware, the
driver could hang if no OEM handler was set.  Fix the code to handle this.
This was tested by setting the flags by hand after they were fetched.

Signed-off-by: Corey Minyard <minyard@acm.org>
Ackde-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Add a missing space that prevents building modules that require host programs
Ross Biro [Sat, 16 Sep 2006 19:15:39 +0000 (12:15 -0700)]
[PATCH] Add a missing space that prevents building modules that require host programs

Signed-off-by: Ross Biro <rossb@google.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] hvc_console suspend fix
Andrew Morton [Sat, 16 Sep 2006 19:15:38 +0000 (12:15 -0700)]
[PATCH] hvc_console suspend fix

Fix http://bugzilla.kernel.org/show_bug.cgi?id=7152

Cc: Michael Tautschnig <tautschn@model.in.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] knfsd: Make ext3 reject filehandles referring to invalid inode number
NeilBrown [Sat, 16 Sep 2006 19:15:38 +0000 (12:15 -0700)]
[PATCH] knfsd: Make ext3 reject filehandles referring to invalid inode number

Inodes earlier than the 'first' inode (e.g.  journal, resize) should be
rejected early - except the root inode.  Also inode numbers that are too
big should be rejected early.

[akpm@osdl.org: cleanup]
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] knfsd: Have ext2 reject file handles with bad inode numbers early
NeilBrown [Sat, 16 Sep 2006 19:15:37 +0000 (12:15 -0700)]
[PATCH] knfsd: Have ext2 reject file handles with bad inode numbers early

This prevents bad inode numbers from triggering errors in ext2_get_inode.

[akpm@osdl.org: speedup, cleanup]
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] IPMI: Fix oops on ipmi_msghandler removal for non ipmi systems
Arnaud Patard [Sat, 16 Sep 2006 19:15:36 +0000 (12:15 -0700)]
[PATCH] IPMI: Fix oops on ipmi_msghandler removal for non ipmi systems

When the ipmi_si module is loaded on a system without any ipmi device, it
fails with nodev.  It would be fine if all resources were freed.  A call to
device_unregister() is missing, resulting to a oops when you remove the
ipmi_msghandler.

Signed-off-by: Arnaud Patard <apatard@mandriva.com>
Acked-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] genirq: fix typo in IRQ resend
Imre Deak [Sat, 16 Sep 2006 19:15:35 +0000 (12:15 -0700)]
[PATCH] genirq: fix typo in IRQ resend

Fix a bug where the IRQ_PENDING flag is never cleared and the ISR is called
endlessly without an actual interrupt.

Signed-off-by: Imre Deak <imre.deak@solidboot.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years agoRDMA/cma: Increase the IB CM retry count in CMA
Michael S. Tsirkin [Wed, 13 Sep 2006 12:01:54 +0000 (15:01 +0300)]
RDMA/cma: Increase the IB CM retry count in CMA

3 seems like a low number of IB Communication Manager retries to set;
we see connections failing under stress, and in any case 3 just looks
like an arbitrary number.  15 is the max value allowed by the
InfiniBand spec.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Acked-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
18 years agoIPoIB: Retry failed send-only multicast group joins
Eli Cohen [Thu, 14 Sep 2006 20:51:41 +0000 (13:51 -0700)]
IPoIB: Retry failed send-only multicast group joins

When a send-only multicast group join fails, mcast->query must be set
to NULL.  Otherwise, IPoIB will never retry the join and the multicast
group will never be reachable.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
18 years agoIB/srp: Don't schedule reconnect from srp
Ishai Rabinovitz [Thu, 7 Sep 2006 12:00:01 +0000 (15:00 +0300)]
IB/srp: Don't schedule reconnect from srp

If there is a problem in the connection, the SCSI mid-layer will
eventually call srp_reset_host(), which will call srp_reconnect(), so
we do not need to schedule a call to srp_reconnect_work() from
srp_completion().

Removing this prevents srp_reset_host() from failing if a reconnect
scheduled from srp_completion() is already in progress, which in turn
was causing crashes as both SCSI midlayer and srp_reconnect() were
cancelling commands.

Signed-off-by: Ishai Rabinovitz <ishai@mellanox.co.il>
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
18 years ago[PATCH] mv643xx_eth: Unmap DMA buffers in receive path
Dale Farnsworth [Wed, 13 Sep 2006 16:21:08 +0000 (09:21 -0700)]
[PATCH] mv643xx_eth: Unmap DMA buffers in receive path

Fix a missing call to dma_unmap_single() in the receive path.  Without
this call, errors have been observed on non-cache-coherent systems.

Signed-off-by Dale Farnsworth <dale@farnsworth.org>

Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
Linus Torvalds [Wed, 13 Sep 2006 15:01:41 +0000 (08:01 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/paulus/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc:
  [POWERPC] Fix G5 DART (IOMMU) race causing occasional data corruption
  [POWERPC] Fix MMIO ops to provide expected barrier behaviour
  [POWERPC] Fix interrupt clearing in kdump shutdown sequence
  [POWERPC] update prep_defconfig
  [POWERPC] kdump: Support kernels having 64k page size.
  [POWERPC] Implement PowerPC futex_atomic_cmpxchg_inatomic().
  [POWERPC] Add new, missing argument to of_irq_map_raw() for 86xx.
  [POWERPC] Update defconfigs

18 years ago[PATCH] cciss: version update, new hw
Mike Miller [Wed, 13 Sep 2006 03:36:07 +0000 (20:36 -0700)]
[PATCH] cciss: version update, new hw

Add support for new hardware and bumps the version to 3.6.10.  It seems
there were several changes introduced including soft_irq.  I decided to
bump the major number to reflect these changes.  Since we're still
supporting older vendor kernels I need some way differentiate between
kernel versions <=2.6.10 and newer kernels >=2.6.16.

Signed-off-by: Mike Miller <mike.miller@hp.com>
Cc: Jens Axboe <axboe@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] headers_check: fix userspace build of asm-mips/page.h
David Woodhouse [Wed, 13 Sep 2006 03:36:06 +0000 (20:36 -0700)]
[PATCH] headers_check: fix userspace build of asm-mips/page.h

MIPS asm/page.h unconditionally includes <asm-generic/memory_model.h>, which
doesn't exist in userspace.  Move an #endif /* __KERNEL__ */ down a few lines
to prevent that.

Also, remove the broken definition of PAGE_SIZE which is never going to be
correct -- in the absence of PAGE_SIZE, non-broken userspace will fall back to
using sysconf() or getpagesize() instead.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] headers_check: don't expose PFN stuff to userspace in <asm-i386/setup.h>
David Woodhouse [Wed, 13 Sep 2006 03:36:05 +0000 (20:36 -0700)]
[PATCH] headers_check: don't expose PFN stuff to userspace in <asm-i386/setup.h>

The header file <linux/pfn.h> doesn't exist in userspace and probably
shouldn't -- but it's used unconditionally in <asm-i386/setup.h>.  Protect it
with #ifdef __KERNEL__ and move setup.h from $(header-y) to $(unifdef-y) in
Kbuild accordingly.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] headers_check: move kernel-only #includes within <asm-i386/elf.h>
David Woodhouse [Wed, 13 Sep 2006 03:36:04 +0000 (20:36 -0700)]
[PATCH] headers_check: move kernel-only #includes within <asm-i386/elf.h>

Some files which don't exist in userspace were being included unconditionally
in asm-i386/elf.h.  Move the offending #includes down a few lines so that
they're protected by #ifdef __KERNEL__

In fact, we probably want to kill off all userspace use of asm/elf.h -- but we
aren't there yet, so we should at least make it possible to include it for
now.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] headers_check: move inclusion of <linux/linkage.h> in <asm-i386/signal.h>
David Woodhouse [Wed, 13 Sep 2006 03:36:04 +0000 (20:36 -0700)]
[PATCH] headers_check: move inclusion of <linux/linkage.h> in <asm-i386/signal.h>

Because <linux/linkage.h> doesn't exist in userspace, it should be only
included from within #ifdef __KERNEL__.  Move the corresponding #include

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] headers_check: remove <asm/timex.h> from user export
David Woodhouse [Wed, 13 Sep 2006 03:36:03 +0000 (20:36 -0700)]
[PATCH] headers_check: remove <asm/timex.h> from user export

There's useful stuff in <linux/timex.h> but <asm/timex.h> has nothing for
userspace.  Stop exporting it, and include it only from within the existing
#ifdef __KERNEL__ part of <linux/timex.h>

This fixes a 'make headers_check' failure on i386 because asm-i386/timex.h
includes both asm-i386/tsc.h and asm-i386/processor.h, neither of which are
exported to userspace.  It's not entirely clear _why_ it includes either of
these, but it does.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] headers_check: reduce user-visible noise in <linux/nfs_fs.h>
David Woodhouse [Wed, 13 Sep 2006 03:36:02 +0000 (20:36 -0700)]
[PATCH] headers_check: reduce user-visible noise in <linux/nfs_fs.h>

We don't need any of this crap included from the user-visible part of nfs_fs.h
-- remove it all.

In fact, we probably don't need anything but NFS_SUPER_MAGIC to be defined; is
there any need for anything else?  And magic numbers should probably move to
<linux/magic.h> rather than being strewn across various fs-specific include
files which exist in userspace for solely that purpose.

With this patch, 'make header_check' works again at least on PowerPC.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] headers_check: clarify error message
Alexey Dobriyan [Wed, 13 Sep 2006 03:36:01 +0000 (20:36 -0700)]
[PATCH] headers_check: clarify error message

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] headers_check: improve #include regexp
Alexey Dobriyan [Wed, 13 Sep 2006 03:36:00 +0000 (20:36 -0700)]
[PATCH] headers_check: improve #include regexp

The following combinations of pp-tokens are used

#include
 #include
# include

so, script'd better check for all of them.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86: reserve a boot-loader ID number for Xen
Jeremy Fitzhardinge [Wed, 13 Sep 2006 03:35:57 +0000 (20:35 -0700)]
[PATCH] x86: reserve a boot-loader ID number for Xen

Claim an ID number for Xen in the LOADER_TYPE field.

Also, keep the table in zero-page.txt consistent with boot.txt.

[hpa says: 6 was skipped because I couldn't rule out that it hadn't been
 unofficially used.  It seemed easier to skip it for now.]

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] rcu_do_batch: make ->qlen decrement irq safe
Oleg Nesterov [Wed, 13 Sep 2006 03:35:55 +0000 (20:35 -0700)]
[PATCH] rcu_do_batch: make ->qlen decrement irq safe

rcu_do_batch() decrements rdp->qlen with irqs enabled.  This is not good,
it can also be modified by call_rcu() from interrupt.

Decrement ->qlen once with irqs disabled, after a main loop.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Dipankar Sarma <dipankar@in.ibm.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] SCSI: lockdep annotation in scsi_send_eh_cmnd
Stefan Richter [Wed, 13 Sep 2006 03:35:54 +0000 (20:35 -0700)]
[PATCH] SCSI: lockdep annotation in scsi_send_eh_cmnd

Fixup for lockdep enabled kernels: Annotate an on-stack completion.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] alim15x3.c: M5229 (rev c8) support for DMA cd-writer
Michael De Backer [Wed, 13 Sep 2006 03:35:53 +0000 (20:35 -0700)]
[PATCH] alim15x3.c: M5229 (rev c8) support for DMA cd-writer

Configuration bits are not set properly for DMA on some chipset revisions.
It has already been corrected for M5229 (rev c7) but not for M5229 (rev
c8).  This leads to the bug described at
http://bugzilla.kernel.org/show_bug.cgi?id=5786 (lost interrupt + ide bus
hangs).

Signed-off-by: Michael De Backer <micdb@skynet.be>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] We can not allow anonymous contributions to the kernel
Greg KH [Wed, 13 Sep 2006 03:35:52 +0000 (20:35 -0700)]
[PATCH] We can not allow anonymous contributions to the kernel

The DCO does not mean anything if we allow anonymous contributors to the
kernel.  As this is an open source project, we need to do everything in the
open.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] lockdep: double the number of stack-trace entries
Ingo Molnar [Wed, 13 Sep 2006 03:35:50 +0000 (20:35 -0700)]
[PATCH] lockdep: double the number of stack-trace entries

Miles Lane reported the "BUG: MAX_STACK_TRACE_ENTRIES too low!" message,
which means that during normal use his system produced enough lockdep
events so that the 128-thousand entries stack-trace array got exhausted.
Double the size of the array.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Miles Lane <miles.lane@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] libata: ignore CFA signature while sanity-checking an ATAPI device
Tejun Heo [Wed, 13 Sep 2006 03:35:49 +0000 (20:35 -0700)]
[PATCH] libata: ignore CFA signature while sanity-checking an ATAPI device

0x848a in ID word 0 indicates CFA device iff the ID data is obtained from
IDENTIFY DEVICE.  For ATAPI devices, 0x848a in ID work 0 indicates valid
ATAPI device.  Fix sanity check in ata_dev_read_id() such that ATAPI
devices reporting 0x848a in ID word 0 is not handled as error.

The problem is identified by J.A.  Magallon with HL-DT-ST DVDRAM GSA-4120B.

Signed-off-by: Tejun Helo <htejun@gmail.com>
Cc: J.A. Magallon <jamagallon@ono.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[POWERPC] Fix G5 DART (IOMMU) race causing occasional data corruption
Benjamin Herrenschmidt [Wed, 13 Sep 2006 12:12:52 +0000 (22:12 +1000)]
[POWERPC] Fix G5 DART (IOMMU) race causing occasional data corruption

It seems that the occasional data corruption observed with the tg3
driver wasn't due to missing barriers after all, but rather seems to
be due to the DART (= IOMMU) in the U4 northbridge reading stale
IOMMU table entries from memory due to a race.  This fixes it by
making the CPU read the entry back from memory before using it.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Fix MMIO ops to provide expected barrier behaviour
Paul Mackerras [Wed, 13 Sep 2006 12:08:26 +0000 (22:08 +1000)]
[POWERPC] Fix MMIO ops to provide expected barrier behaviour

This changes the writeX family of functions to have a sync instruction
before the MMIO store rather than after, because the generally expected
behaviour is that the device receiving the MMIO store can be guaranteed
to see the effects of any preceding writes to normal memory.

To preserve ordering between writeX and readX, and to preserve ordering
between preceding stores and the readX, the readX family of functions
have had an sync added before the load.

Although writeX followed by spin_unlock is not officially guaranteed
to keep the writeX inside the spin-locked region unless an mmiowb()
is used, there are currently drivers that depend on the previous
behaviour on powerpc, which was that the mmiowb wasn't actually required.
Therefore we have a per-cpu flag that is set by writeX, cleared by
__raw_spin_lock and mmiowb, and tested by __raw_spin_unlock.  If it is
set, __raw_spin_unlock does a sync and clears it.

This changes both 32-bit and 64-bit readX/writeX.  32-bit already has a
sync in __raw_spin_unlock (since lwsync doesn't exist on 32-bit), and thus
doesn't need the per-cpu flag.

Tested on G5 (PPC970) and POWER5.

Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Fix interrupt clearing in kdump shutdown sequence
Mohan Kumar M [Tue, 12 Sep 2006 12:18:21 +0000 (17:48 +0530)]
[POWERPC] Fix interrupt clearing in kdump shutdown sequence

Call chip->eoi(irq) to clear any pending interrupt in case of kdump
shutdown sequence.  chip->end(irq) does not serve this purpose.

Signed-off-by: Mohan Kumar M <mohan@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] update prep_defconfig
Olaf Hering [Thu, 7 Sep 2006 12:44:45 +0000 (14:44 +0200)]
[POWERPC] update prep_defconfig

Update PReP defconfig, disable some drivers for hardware that is not
used on those systems; enable SL82C105 IDE driver for Powerstack.

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] kdump: Support kernels having 64k page size.
Sachin P. Sant [Fri, 8 Sep 2006 02:29:52 +0000 (07:59 +0530)]
[POWERPC] kdump: Support kernels having 64k page size.

This is required to generate proper core files using kdump on ppc64.

Create a backup region of 64K size irrespective of the PAGE SIZE.
At present 32K was used as backup size. In the case of 64K page size,
second PT_LOAD segments starts at 32K and the first one is not page
aligned.  __ioremap() (crash_dump.c) fails if pfn = 0 which is the
case for the second PT_LOAD segment. This is not an issue for 4K page
size because the the first page (32K backup) is copied to second
kernel memory and thus referencing with the second kernel pfn.

Signed-off-by: Sachin Sant <sachinp@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Implement PowerPC futex_atomic_cmpxchg_inatomic().
David Woodhouse [Tue, 5 Sep 2006 04:53:14 +0000 (21:53 -0700)]
[POWERPC] Implement PowerPC futex_atomic_cmpxchg_inatomic().

The sys_[gs]et_robust_list() syscalls were wired up on PowerPC but
didn't work correctly because futex_atomic_cmpxchg_inatomic() wasn't
implemented.  Implement it, based on __cmpxchg_u32().

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Add new, missing argument to of_irq_map_raw() for 86xx.
Jon Loeliger [Fri, 1 Sep 2006 15:17:20 +0000 (10:17 -0500)]
[POWERPC] Add new, missing argument to of_irq_map_raw() for 86xx.

Ben speaks; we follow.

Signed-off-by: Jon Loeliger <jdl@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years agoMerge branch 'master' into upstream-fixes
Jeff Garzik [Wed, 13 Sep 2006 04:34:34 +0000 (00:34 -0400)]
Merge branch 'master' into upstream-fixes

18 years agoLinux v2.6.18-rc7
Linus Torvalds [Wed, 13 Sep 2006 01:41:36 +0000 (18:41 -0700)]
Linux v2.6.18-rc7

One last time..

18 years agoMerge branch 'audit.b29' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit...
Linus Torvalds [Wed, 13 Sep 2006 00:37:24 +0000 (17:37 -0700)]
Merge branch 'audit.b29' of git://git./linux/kernel/git/viro/audit-current

* 'audit.b29' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit-current:
  [PATCH] sparc64 audit syscall classes hookup
  [PATCH] syscall class hookup for all normal targets

18 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Wed, 13 Sep 2006 00:34:46 +0000 (17:34 -0700)]
Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  [CIFS] Fix CIFS readdir access denied when SE Linux enabled

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/mchehab/v4l-dvb
Linus Torvalds [Wed, 13 Sep 2006 00:33:42 +0000 (17:33 -0700)]
Merge /pub/scm/linux/kernel/git/mchehab/v4l-dvb

* master.kernel.org:/pub/scm/linux/kernel/git/mchehab/v4l-dvb:
  V4L/DVB (4608c): Fix I2C dependencies for saa7146 modules
  V4L/DVB (4608b): i2c deps fix on DVB
  V4L/DVB (4605): Fixes an issue with V4L1 and make headers-install
  V4L/DVB (4520): Fix an error when loading bttv driver on PV M4900.
  V4L/DVB (4511): Restore tuner_ymec_tvf66t5_b_dff_pal_ranges[] to fix UHF switch functionality
  V4L/DVB (4494a): Fix compilation when V4L1 support is not present

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/gregkh/usb-2.6
Linus Torvalds [Wed, 13 Sep 2006 00:32:21 +0000 (17:32 -0700)]
Merge /pub/scm/linux/kernel/git/gregkh/usb-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/gregkh/usb-2.6:
  USB: Fix unload oops and memory leak in yealink driver
  usbserial: Reference leak

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/lethal/sh64-2.6
Linus Torvalds [Wed, 13 Sep 2006 00:31:36 +0000 (17:31 -0700)]
Merge /pub/scm/linux/kernel/git/lethal/sh64-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/lethal/sh64-2.6:
  sh64: Add a sane pm_power_off implementation.
  sh64: Use generic BUG_ON()/WARN_ON().
  sh64: Trivial build fixes.
  sh64: Drop deprecated ISA tuning for legacy toolchains.

18 years agoMerge git://oss.sgi.com:8090/xfs/xfs-2.6
Linus Torvalds [Wed, 13 Sep 2006 00:31:16 +0000 (17:31 -0700)]
Merge git://oss.sgi.com:8090/xfs/xfs-2.6

* git://oss.sgi.com:8090/xfs/xfs-2.6:
  [XFS] Fix a bad pointer dereference in the quota statvfs handling.
  [XFS] Fix xfs_splice_write() so appended data gets to disk.
  [XFS] Fix ABBA deadlock between i_mutex and iolock. Avoid calling
  [XFS] Prevent free space oversubscription and xfssyncd looping.

18 years ago[PATCH] e1000: fix TX timout hang regression for 82542rev3
Auke Kok [Mon, 11 Sep 2006 21:00:21 +0000 (14:00 -0700)]
[PATCH] e1000: fix TX timout hang regression for 82542rev3

Commit 581d708eb47cccb5f41bc0817e50c9b004011ba8 (oct. 5 2005) introduced
partial Multiqueue support for e1000 which broke macro smartness in setting
up head/tail registers for 82542 rev3 chipsets, making these adapters
completely non-working since 2.6.15.

This commit sets the proper head and tail registers for read and write
descriptor rings. Ths fix was tested on an 82542 rev3 NIC and newer NICs.

Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years agoMerge branch 'upstream-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/linvil...
Jeff Garzik [Tue, 12 Sep 2006 15:42:33 +0000 (11:42 -0400)]
Merge branch 'upstream-fixes' of git://git./linux/kernel/git/linville/wireless-2.6 into upstream-fixes

18 years agoUSB: Fix unload oops and memory leak in yealink driver
Henk Vergonet [Fri, 8 Sep 2006 00:21:01 +0000 (02:21 +0200)]
USB: Fix unload oops and memory leak in yealink driver

This patch fixes a memory leak and a kernel oops when trying to unload
the driver, due to an unbalanced cleanup.
Thanks Ivar Jensen for spotting my mistake.

Signed-off-by: Henk Vergonet <henk.vergonet@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agousbserial: Reference leak
Matthias Urlichs [Mon, 11 Sep 2006 10:35:20 +0000 (12:35 +0200)]
usbserial: Reference leak

A sufficiently-large number of USB serial devices causes a reference leak
when /proc/tty/drivers/usbserial is read.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years ago[PATCH] sparc64 audit syscall classes hookup
Al Viro [Tue, 12 Sep 2006 06:59:45 +0000 (02:59 -0400)]
[PATCH] sparc64 audit syscall classes hookup

... that should do it for all targets; the only remaining issues are
mips (currently treated as non-biarch) and handling of other OS
emulations (OSF/SunOS/Solaris/???).  The latter would need to be
assigned new AUDIT_ARCH_... ABI numbers anyway...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
18 years ago[PATCH] syscall class hookup for all normal targets
Al Viro [Tue, 12 Sep 2006 07:04:40 +0000 (03:04 -0400)]
[PATCH] syscall class hookup for all normal targets

Take default arch/*/kernel/audit.c to lib/, have those with special
needs (== biarch) define AUDIT_ARCH in their Kconfig.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>