pandora-kernel.git
17 years agoPull acpi_os_allocate into test branch
Len Brown [Mon, 10 Jul 2006 06:39:47 +0000 (02:39 -0400)]
Pull acpi_os_allocate into test branch

17 years agoPull acpica-20060707 into test branch
Len Brown [Mon, 10 Jul 2006 06:39:41 +0000 (02:39 -0400)]
Pull acpica-20060707 into test branch

17 years agoPull bugzilla-6687 into test branch
Len Brown [Mon, 10 Jul 2006 06:39:36 +0000 (02:39 -0400)]
Pull bugzilla-6687 into test branch

17 years agoPull dock into test branch
Len Brown [Mon, 10 Jul 2006 06:39:33 +0000 (02:39 -0400)]
Pull dock into test branch

17 years agoPull trivial into test branch
Len Brown [Mon, 10 Jul 2006 06:39:26 +0000 (02:39 -0400)]
Pull trivial into test branch

17 years agoPull battery into test branch
Len Brown [Mon, 10 Jul 2006 06:39:23 +0000 (02:39 -0400)]
Pull battery into test branch

17 years agoACPI: acpi_os_allocate() fixes
Len Brown [Mon, 10 Jul 2006 05:35:51 +0000 (01:35 -0400)]
ACPI: acpi_os_allocate() fixes

Replace acpi_in_resume with a more general hack
to check irqs_disabled() on any kmalloc() from ACPI.
While setting (system_state != SYSTEM_RUNNING) on resume
seemed more general, Andrew Morton preferred this approach.

http://bugzilla.kernel.org/show_bug.cgi?id=3469

Make acpi_os_allocate() into an inline function to
allow /proc/slab_allocators to work.

Delete some memset() that could fault on allocation failure.

Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoACPI: SBS: fix initialization, sem2mutex
Andrew Morton [Mon, 10 Jul 2006 06:34:45 +0000 (02:34 -0400)]
ACPI: SBS: fix initialization, sem2mutex

cm_sbs_sem is being downed (via acpi_ac_init->acpi_lock_ac_dir) before it is
initialised, with grave results.

- Make it a mutex

- Initialise it

- Make it static

- Clean other stuff up.

Thanks to Paul Drynoff <pauldrynoff@gmail.com> for reporting and testing.

Cc: Rich Townsend <rhdt@bartol.udel.edu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoACPI: add 'const' to several ACPI file_operations
Arjan van de Ven [Tue, 4 Jul 2006 17:06:00 +0000 (13:06 -0400)]
ACPI: add 'const' to several ACPI file_operations

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoACPI: delete some defaults from ACPI Kconfig
Andi Kleen [Thu, 6 Jul 2006 16:14:00 +0000 (12:14 -0400)]
ACPI: delete some defaults from ACPI Kconfig

No need for video to be always in
No need for ACPI dock driver to be always in
No need for smart battery driver to be always in

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoACPI: "Device `[%s]' is not power manageable" make message debug only
Len Brown [Sun, 9 Jul 2006 20:33:26 +0000 (16:33 -0400)]
ACPI: "Device `[%s]' is not power manageable" make message debug only

Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoACPI: ACPI_DOCK Kconfig
Len Brown [Mon, 10 Jul 2006 02:09:57 +0000 (22:09 -0400)]
ACPI: ACPI_DOCK Kconfig

HOTPLUG_PCI_ACPI depends on ACPI_DOCK
ACPI_IBM_DOCK depends on ACPI_DOCK=n
ACPI_DOCK is EXPERIMENTAL, though that doesn't seem to mean much

Signed-off-by: Len Brown <len.brown@intel.com>
17 years ago[PATCH] i386: use thread_info flags for debug regs and IO bitmaps
Stephane Eranian [Mon, 10 Jul 2006 01:12:39 +0000 (21:12 -0400)]
[PATCH] i386: use thread_info flags for debug regs and IO bitmaps

Use thread info flags to track use of debug registers and IO bitmaps.

 - add TIF_DEBUG to track when debug registers are active
 - add TIF_IO_BITMAP to track when I/O bitmap is used
 - modify __switch_to() to use the new TIF flags

Performance tested on Pentium II, ten runs of LMbench context switch
benchmark (smaller is better:)

before after
avg 3.65 3.39
min 3.55 3.33

Signed-off-by: Stephane Eranian <eranian@hpl.hp.com>
Signed-off-by: Chuck Ebbert <76306.1226@compuserve.com>
Acked-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoMerge commit master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6 of HEAD
Linus Torvalds [Sun, 9 Jul 2006 22:50:41 +0000 (15:50 -0700)]
Merge commit /linux/kernel/git/davem/net-2.6 of HEAD

* HEAD:
  [AX.25]: Use kzalloc
  [ATM] net/atm/clip.c: fix PROC_FS=n compile
  [PKT_SCHED]: act_api: Fix module leak while flushing actions
  [NET]: Fix IPv4/DECnet routing rule dumping
  [NET] gso: Fix up GSO packets with broken checksums
  [NET] gso: Add skb_is_gso
  [IRDA]: fix drivers/net/irda/ali-ircc.c:ali_ircc_init()
  [ATM]: fix possible recursive locking in skb_migrate()
  [ATM]: Typo in drivers/atm/Kconfig...
  [TG3]: add amd8131 to "write reorder" chipsets
  [NET]: Fix network device interface printk message priority

17 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Sun, 9 Jul 2006 22:50:06 +0000 (15:50 -0700)]
Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/netdev-2.6

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6: (23 commits)
  [PATCH] 8139too deadlock fix
  [netdrvr] 3c59x: snip changelog from source code
  e1000: increase version to 7.1.9-k2
  e1000: add ich8lan device ID's
  e1000: allow user to disable ich8 lock loss workaround
  e1000: integrate ich8 support into driver
  e1000: add ich8lan core functions
  e1000: disable ERT
  e1000: check return value of _get_speed_and_duplex
  e1000: M88 PHY workaround
  e1000: fix adapter led blinking inconsistency
  e1000: disable CRC stripping workaround
  e1000: force register write flushes to circumvent broken platforms
  e1000: rework module param code with uninitialized values
  e1000: recycle skb
  e1000: change printk into DPRINTK
  e1000: add smart power down code
  e1000: small performance tweak by removing double code
  e1000: fix CONFIG_PM blocks
  e1000: Make PHY powerup/down a function
  ...

17 years agoMerge branch 'blktrace' of git://brick.kernel.dk/data/git/linux-2.6-block
Linus Torvalds [Sun, 9 Jul 2006 22:49:36 +0000 (15:49 -0700)]
Merge branch 'blktrace' of git://brick.kernel.dk/data/git/linux-2.6-block

* 'blktrace' of git://brick.kernel.dk/data/git/linux-2.6-block:
  [PATCH] Only the first two bits in bio->bi_rw and rq->flags match
  [PATCH] blktrace: readahead support
  [PATCH] blktrace: fix barrier vs sync typo

17 years agoRevert "Revert "ACPI: dock driver""
Len Brown [Sun, 9 Jul 2006 21:22:28 +0000 (17:22 -0400)]
Revert "Revert "ACPI: dock driver""

This reverts 953969ddf5b049361ed1e8471cc43dc4134d2a6f commit.

17 years agoACPI: acpi_os_get_thread_id() returns current
Len Brown [Sat, 8 Jul 2006 00:11:07 +0000 (20:11 -0400)]
ACPI: acpi_os_get_thread_id() returns current

Linux mutexes and the debug code that that reference
acpi_os_get_thread_id() are happy with 0.
But the AML mutexes in exmutex.c expect a unique non-zero
number for each thread - as they track this thread_id
to permit the mutex re-entrancy defined by the ACPI spec.

http://bugzilla.kernel.org/show_bug.cgi?id=6687

Signed-off-by: Len Brown <len.brown@intel.com>
17 years agoACPI: ACPICA 20060707
Bob Moore [Sat, 8 Jul 2006 00:44:38 +0000 (20:44 -0400)]
ACPI: ACPICA 20060707

Added the ACPI_PACKED_POINTERS_NOT_SUPPORTED macro to
support C compilers that do not allow the initialization
of address pointers within packed structures - even though
the hardware itself may support misaligned transfers. Some
of the debug data structures are packed by default to
minimize size.

Added an error message for the case where
acpi_os_get_thread_id() returns zero. A non-zero value is
required by the core ACPICA code to ensure the proper
operation of AML mutexes and recursive control methods.

The DSDT is now the only ACPI table that determines whether
the AML interpreter is in 32-bit or 64-bit mode. Not really
a functional change, but the hooks for per-table 32/64
switching have been removed from the code. A clarification
to the ACPI specification is forthcoming in ACPI 3.0B.

Fixed a possible leak of an Owner ID in the error
path of tbinstal.c acpi_tb_init_table_descriptor() and
migrated all table OwnerID deletion to a single place in
acpi_tb_uninstall_table() to correct possible leaks when using
the acpi_tb_delete_tables_by_type() interface (with assistance
from Lance Ortiz.)

Fixed a problem with Serialized control methods where the
semaphore associated with the method could be over-signaled
after multiple method invocations.

Fixed two issues with the locking of the internal
namespace data structure. Both the Unload() operator and
acpi_unload_table() interface now lock the namespace during
the namespace deletion associated with the table unload
(with assistance from Linn Crosetto.)

Fixed problem reports (Valery Podrezov) integrated: -
Eliminate unnecessary memory allocation for CreateXxxxField
http://bugzilla.kernel.org/show_bug.cgi?id=5426

Fixed problem reports (Fiodor Suietov) integrated: -
Incomplete cleanup branches in AcpiTbGetTableRsdt (BZ 369)
- On Address Space handler deletion, needless deactivation
call (BZ 374) - AcpiRemoveAddressSpaceHandler: validate
Device handle parameter (BZ 375) - Possible memory leak,
Notify sub-objects of Processor, Power, ThermalZone (BZ
376) - AcpiRemoveAddressSpaceHandler: validate Handler
parameter (BZ 378) - Minimum Length of RSDT should be
validated (BZ 379) - AcpiRemoveNotifyHandler: return
AE_NOT_EXIST if Processor Obj has no Handler (BZ (380)
- AcpiUnloadTable: return AE_NOT_EXIST if no table of
specified type loaded (BZ 381)

Signed-off-by: Bob Moore <robert.moore@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
17 years ago[AX.25]: Use kzalloc
Ralf Baechle [Sun, 9 Jul 2006 19:14:22 +0000 (12:14 -0700)]
[AX.25]: Use kzalloc

Replace kzalloc instead of kmalloc + memset.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[ATM] net/atm/clip.c: fix PROC_FS=n compile
Adrian Bunk [Sun, 9 Jul 2006 19:13:18 +0000 (12:13 -0700)]
[ATM] net/atm/clip.c: fix PROC_FS=n compile

This patch fixes the following compile error with CONFIG_PROC_FS=n by
reverting commit dcdb02752ff13a64433c36f2937a58d93ae7a19e:

<--  snip  -->

...
  CC      net/atm/clip.o
net/atm/clip.c: In function ‘atm_clip_init’:
net/atm/clip.c:975: error: ‘atm_proc_root’ undeclared (first use in this function)
net/atm/clip.c:975: error: (Each undeclared identifier is reported only once
net/atm/clip.c:975: error: for each function it appears in.)
net/atm/clip.c:977: error: ‘arp_seq_fops’ undeclared (first use in this function)
make[2]: *** [net/atm/clip.o] Error 1

<--  snip  -->

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[PKT_SCHED]: act_api: Fix module leak while flushing actions
Thomas Graf [Sun, 9 Jul 2006 18:36:23 +0000 (11:36 -0700)]
[PKT_SCHED]: act_api: Fix module leak while flushing actions

Module reference needs to be given back if message header
construction fails.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years agoRevert "ACPI: dock driver"
Linus Torvalds [Sun, 9 Jul 2006 15:47:46 +0000 (08:47 -0700)]
Revert "ACPI: dock driver"

This reverts commit a5e1b94008f2a96abf4a0c0371a55a56b320c13e.

Adrian Bunk points out that it has build errors, and apparently no
maintenance. Throw it out.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoi386: improve and correct inline asm memory constraints
Linus Torvalds [Sat, 8 Jul 2006 22:24:18 +0000 (15:24 -0700)]
i386: improve and correct inline asm memory constraints

Use "+m" rather than a combination of "=m" and "m" for improved clarity
and consistency.

This also fixes some inlines that incorrectly didn't tell the compiler
that they read the old value at all, potentially causing the compiler to
generate bogus code.  It appear that all of those potential bugs were
hidden by the use of extra "volatile" specifiers on the data structures
in question, though.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agopower: improve inline asm memory constraints
Linus Torvalds [Sat, 8 Jul 2006 22:00:28 +0000 (15:00 -0700)]
power: improve inline asm memory constraints

Use "+m" rather than a combination of "=m" and "m" for improved
clarity and consistency.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years ago[NET]: Fix IPv4/DECnet routing rule dumping
Patrick McHardy [Sat, 8 Jul 2006 20:38:55 +0000 (13:38 -0700)]
[NET]: Fix IPv4/DECnet routing rule dumping

When more rules are present than fit in a single skb, the remaining
rules are incorrectly skipped.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NET] gso: Fix up GSO packets with broken checksums
Herbert Xu [Sat, 8 Jul 2006 20:34:56 +0000 (13:34 -0700)]
[NET] gso: Fix up GSO packets with broken checksums

Certain subsystems in the stack (e.g., netfilter) can break the partial
checksum on GSO packets.  Until they're fixed, this patch allows this to
work by recomputing the partial checksums through the GSO mechanism.

Once they've all been converted to update the partial checksum instead of
clearing it, this workaround can be removed.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NET] gso: Add skb_is_gso
Herbert Xu [Sat, 8 Jul 2006 20:34:32 +0000 (13:34 -0700)]
[NET] gso: Add skb_is_gso

This patch adds the wrapper function skb_is_gso which can be used instead
of directly testing skb_shinfo(skb)->gso_size.  This makes things a little
nicer and allows us to change the primary key for indicating whether an skb
is GSO (if we ever want to do that).

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IRDA]: fix drivers/net/irda/ali-ircc.c:ali_ircc_init()
Adrian Bunk [Sat, 8 Jul 2006 20:33:28 +0000 (13:33 -0700)]
[IRDA]: fix drivers/net/irda/ali-ircc.c:ali_ircc_init()

The Coverity checker spotted, that from the changes from commit
898b1d16f8230fb912a0c2248df685735c6ceda3 the
       if (ret)
               platform_driver_unregister(&ali_ircc_driver);
was dead code.

This patch changes this function to what seems to have been the
intention.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[ATM]: fix possible recursive locking in skb_migrate()
Arjan van de Ven [Sat, 8 Jul 2006 20:30:52 +0000 (13:30 -0700)]
[ATM]: fix possible recursive locking in skb_migrate()

ok this is a real potential deadlock in a way, it takes two locks of 2
skbuffs without doing any kind of lock ordering; I think the following
patch should fix it. Just sort the lock taking order by address of the
skb.. it's not pretty but it's the best this can do in a minimally
invasive way.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Chas Williams <chas@cmf.nrl.navy.mil>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[ATM]: Typo in drivers/atm/Kconfig...
Matt LaPlante [Sat, 8 Jul 2006 20:30:09 +0000 (13:30 -0700)]
[ATM]: Typo in drivers/atm/Kconfig...

From: Matt LaPlante <kernel1@cyberdogtech.com>
Signed-off-by: Charles Williams <chas@cmf.nrl.navy.mil>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[TG3]: add amd8131 to "write reorder" chipsets
John W. Linville [Sat, 8 Jul 2006 20:28:53 +0000 (13:28 -0700)]
[TG3]: add amd8131 to "write reorder" chipsets

Add the AMD 8131 bridge to the list of chipsets that reorder writes.

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Acked-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NET]: Fix network device interface printk message priority
Stephen Hemminger [Fri, 7 Jul 2006 23:54:05 +0000 (16:54 -0700)]
[NET]: Fix network device interface printk message priority

The printk's in the network device interface code should all be tagged
with severity.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[PATCH] Fix cpufreq vs hotplug lockdep recursion.
Dave Jones [Fri, 7 Jul 2006 16:31:27 +0000 (12:31 -0400)]
[PATCH] Fix cpufreq vs hotplug lockdep recursion.

[ There's some not quite baked bits in cpufreq-git right now
  so sending this on as a patch instead ]

On Thu, 2006-07-06 at 07:58 -0700, Tom London wrote:

> After installing .2356 I get this each time I boot:
> =======================================================
> [ INFO: possible circular locking dependency detected ]
> -------------------------------------------------------
> S06cpuspeed/1620 is trying to acquire lock:
>  (dbs_mutex){--..}, at: [<c060d6bb>] mutex_lock+0x21/0x24
>
> but task is already holding lock:
>  (cpucontrol){--..}, at: [<c060d6bb>] mutex_lock+0x21/0x24
>
> which lock already depends on the new lock.
>

make sure the cpu hotplug recursive mutex (yuck) is taken early in the
cpufreq codepaths to avoid a AB-BA deadlock.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
17 years agoMerge branch 'master' into upstream
Jeff Garzik [Thu, 6 Jul 2006 17:07:32 +0000 (13:07 -0400)]
Merge branch 'master' into upstream

17 years ago[PATCH] Only the first two bits in bio->bi_rw and rq->flags match
Jens Axboe [Thu, 6 Jul 2006 08:18:05 +0000 (10:18 +0200)]
[PATCH] Only the first two bits in bio->bi_rw and rq->flags match

Not three, as assumed. This causes the barrier bit to be needlessly set
for some IO.

Signed-off-by: Jens Axboe <axboe@suse.de>
17 years ago[PATCH] blktrace: readahead support
Nathan Scott [Thu, 6 Jul 2006 08:03:28 +0000 (10:03 +0200)]
[PATCH] blktrace: readahead support

Provide the needed kernel support for distinguishing readahead
from regular read requests when tracing block devices.

Signed-off-by: Nathan Scott <nathans@sgi.com>
Signed-off-by: Jens Axboe <axboe@suse.de>
17 years ago[PATCH] blktrace: fix barrier vs sync typo
Nathan Scott [Thu, 6 Jul 2006 07:56:30 +0000 (09:56 +0200)]
[PATCH] blktrace: fix barrier vs sync typo

Signed-off-by: Nathan Scott <nathans@sgi.com>
Signed-off-by: Jens Axboe <axboe@suse.de>
17 years agoLinux 2.6.18-rc1 v2.6.18-rc1
Linus Torvalds [Thu, 6 Jul 2006 04:09:49 +0000 (21:09 -0700)]
Linux 2.6.18-rc1

It's all good.

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Thu, 6 Jul 2006 04:08:43 +0000 (21:08 -0700)]
Merge /pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [PKT_SCHED]: Fix error handling while dumping actions
  [PKT_SCHED]: Return ENOENT if action module is unavailable
  [PKT_SCHED]: Fix illegal memory dereferences when dumping actions

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6
Linus Torvalds [Thu, 6 Jul 2006 04:08:35 +0000 (21:08 -0700)]
Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6:
  [SPARC64]: Fix stack overflow checking in modular non-SMP kernels.
  [SPARC64]: Fix sparc64 build errors when CONFIG_PCI=n.

17 years ago[PKT_SCHED]: Fix error handling while dumping actions
Thomas Graf [Thu, 6 Jul 2006 03:47:28 +0000 (20:47 -0700)]
[PKT_SCHED]: Fix error handling while dumping actions

"return -err" and blindly inheriting the error code in the netlink
failure exception handler causes errors codes to be returned as
positive value therefore making them being ignored by the caller.

May lead to sending out incomplete netlink messages.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[PKT_SCHED]: Return ENOENT if action module is unavailable
Thomas Graf [Thu, 6 Jul 2006 03:45:57 +0000 (20:45 -0700)]
[PKT_SCHED]: Return ENOENT if action module is unavailable

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[PKT_SCHED]: Fix illegal memory dereferences when dumping actions
Thomas Graf [Thu, 6 Jul 2006 03:45:06 +0000 (20:45 -0700)]
[PKT_SCHED]: Fix illegal memory dereferences when dumping actions

The TCA_ACT_KIND attribute is used without checking its
availability when dumping actions therefore leading to a
value of 0x4 being dereferenced.

The use of strcmp() in tc_lookup_action_n() isn't safe
when fed with string from an attribute without enforcing
proper NUL termination.

Both bugs can be triggered with malformed netlink message
and don't require any privileges.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[SPARC64]: Fix stack overflow checking in modular non-SMP kernels.
Mikael Pettersson [Thu, 6 Jul 2006 03:42:58 +0000 (20:42 -0700)]
[SPARC64]: Fix stack overflow checking in modular non-SMP kernels.

The sparc64 kernel's EXPORT_SYMBOL(_mcount) is inside an
#ifdef CONFIG_SMP. This breaks modules in non-SMP kernels
built with stack overflow checking (CONFIG_STACK_DEBUG=y),
as modules_install reports:

WARNING: /lib/modules/2.6.17/kernel/drivers/ide/ide-cd.ko needs unknown symbol _mcount

Trivially fixed by moving EXPORT_SYMBOL(_mcount) outside of
the #ifdef CONFIG_SMP.

Signed-off-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Thu, 6 Jul 2006 03:19:30 +0000 (20:19 -0700)]
Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  [PCI] Add JMicron PCI ID constants
  [PATCH] ahci: Ensure that we don't grab both functions
  [PATCH] libata-core.c: restore configuration boot messages in ata_dev_configure(), v2
  [PATCH] sata_sil24: add suspend/sleep support
  [PATCH] sata_sil24: separate out sil24_init_controller()
  [PATCH] sata_sil: add suspend/sleep support
  [PATCH] sata_sil: separate out sil_init_controller()
  [PATCH] libata: reimplement controller-wide PM
  [PATCH] libata: reimplement per-dev PM
  [PATCH] libata: implement PM EH actions
  [PATCH] libata: separate out __ata_ehi_hotplugged()
  [PATCH] libata: implement ATA_EHI_NO_AUTOPSY and QUIET
  [PATCH] libata: clean up debounce parameters and improve parameter selection
  [PATCH] libata: implement ATA_EHI_RESUME_LINK
  [PATCH] libata: replace ap_lock w/ ap->lock in ata_scsi_error()
  [PATCH] libata: fix ehc->i.action setting in ata_eh_autopsy()
  [PATCH] libata: add ap->pflags and move core dynamic flags to it
  [PATCH] libata: Conditionally set host->max_cmd_len
  [PATCH] sata_vsc: data_xfer should use mmio

17 years ago[SPARC64]: Fix sparc64 build errors when CONFIG_PCI=n.
Randy Dunlap [Thu, 6 Jul 2006 03:18:39 +0000 (20:18 -0700)]
[SPARC64]: Fix sparc64 build errors when CONFIG_PCI=n.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[PCI] Add JMicron PCI ID constants
Jeff Garzik [Thu, 6 Jul 2006 03:02:48 +0000 (23:02 -0400)]
[PCI] Add JMicron PCI ID constants

They will be used in several IDE/libata files.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] ahci: Ensure that we don't grab both functions
root [Thu, 6 Jul 2006 02:58:20 +0000 (22:58 -0400)]
[PATCH] ahci: Ensure that we don't grab both functions

When we force the chip into dual fn mode so we get PATA and AHCI we must
be sure we don't then do anything dumb like try and grab both with the AHCI
driver.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata-core.c: restore configuration boot messages in ata_dev_configure(...
Borislav Petkov [Tue, 27 Jun 2006 12:51:25 +0000 (14:51 +0200)]
[PATCH] libata-core.c: restore configuration boot messages in ata_dev_configure(), v2

This one looks better, IMHO.

This restores the default libata configuration messages printed during booting.

Signed-off-by: <petkov@math.uni-muenster.de>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] sata_sil24: add suspend/sleep support
Tejun Heo [Mon, 3 Jul 2006 07:07:27 +0000 (16:07 +0900)]
[PATCH] sata_sil24: add suspend/sleep support

Add suspend/sleep support.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] sata_sil24: separate out sil24_init_controller()
Tejun Heo [Mon, 3 Jul 2006 07:07:27 +0000 (16:07 +0900)]
[PATCH] sata_sil24: separate out sil24_init_controller()

Separate out controller initialization from sil24_init_one() into
sil24_init_controller().  This will be used by resume.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] sata_sil: add suspend/sleep support
Tejun Heo [Mon, 3 Jul 2006 07:07:27 +0000 (16:07 +0900)]
[PATCH] sata_sil: add suspend/sleep support

Add suspend/sleep support.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] sata_sil: separate out sil_init_controller()
Tejun Heo [Mon, 3 Jul 2006 07:07:27 +0000 (16:07 +0900)]
[PATCH] sata_sil: separate out sil_init_controller()

Separate out controller initialization from sil_init_one() into
sil_init_controller().  This will be used by resume.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: reimplement controller-wide PM
Tejun Heo [Mon, 3 Jul 2006 07:07:27 +0000 (16:07 +0900)]
[PATCH] libata: reimplement controller-wide PM

Reimplement controller-wide PM.  ata_host_set_suspend/resume() are
defined to suspend and resume a host_set.  While suspended, EHs for
all ports in the host_set are pegged using ATA_FLAG_SUSPENDED and
frozen.

Because SCSI device hotplug is done asynchronously against the rest of
libata EH and the same mutex is used when adding new device, suspend
cannot wait for hotplug to complete.  So, if SCSI device hotplug is in
progress, suspend fails with -EBUSY.

In most cases, host_set resume is followed by device resume.  As each
resume operation requires a reset, a single host_set-wide resume
operation may result in multiple resets.  To avoid this, resume waits
upto 1 second giving PM to request resume for devices.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: reimplement per-dev PM
Tejun Heo [Mon, 3 Jul 2006 07:07:26 +0000 (16:07 +0900)]
[PATCH] libata: reimplement per-dev PM

Reimplement per-dev PM.  The original implementation directly put the
device into suspended mode and didn't synchronize w/ EH operations
including hotplug.  This patch reimplements ata_scsi_device_suspend()
and ata_scsi_device_resume() such that they request EH to perform the
respective operations.  Both functions synchronize with hotplug such
that it doesn't operate on detached devices.

Suspend waits for completion but resume just issues request and
returns.  This allows parallel wake up of devices and thus speeds up
system resume.

Due to sdev detach synchronization, it's not feasible to separate out
EH requesting from sdev handling; thus, ata_device_suspend/resume()
are removed and everything is implemented in the respective
libata-scsi functions.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: implement PM EH actions
Tejun Heo [Mon, 3 Jul 2006 07:07:26 +0000 (16:07 +0900)]
[PATCH] libata: implement PM EH actions

Implement two PM per-dev EH actions - ATA_EH_SUSPEND and
ATA_EH_RESUME.  Each action puts the target device into suspended mode
and resumes from it respectively.

Once a device is put to suspended mode, no EH operations other than
RESUME is allowed on the device.  The device will stay suspended till
it gets resumed and thus reset and revalidated.  To implement this, a
new device state helper - ata_dev_ready() - is implemented and used in
EH action implementations to make them operate only on attached &
running devices.

If all possible devices on a port are suspended, reset is skipped too.
This prevents spurious events including hotplug events from disrupting
suspended devices.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: separate out __ata_ehi_hotplugged()
Tejun Heo [Mon, 3 Jul 2006 07:07:26 +0000 (16:07 +0900)]
[PATCH] libata: separate out __ata_ehi_hotplugged()

Separate out __ata_ehi_hotplugged() from ata_ehi_hotplugged().  The
underscored version doesn't set AC_ERR_ATA_BUS.  This will be used for
resume which is a hotplug event but not an ATA bus error.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: implement ATA_EHI_NO_AUTOPSY and QUIET
Tejun Heo [Mon, 3 Jul 2006 07:07:26 +0000 (16:07 +0900)]
[PATCH] libata: implement ATA_EHI_NO_AUTOPSY and QUIET

Implement ATA_EHI_NO_AUTOPSY and QUIET.  These used to be implied by
ATA_PFLAG_LOADING, but new power management and PMP support need to
use these separately.  e.g. Suspend/resume operations shouldn't print
full EH messages and resume shouldn't be recorded as an error.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: clean up debounce parameters and improve parameter selection
Tejun Heo [Mon, 3 Jul 2006 07:07:26 +0000 (16:07 +0900)]
[PATCH] libata: clean up debounce parameters and improve parameter selection

The names of predefined debounce timing parameters didn't exactly
match their usages.  Rename to more generic names and implement param
selection helper sata_ehc_deb_timing() which uses EHI_HOTPLUGGED to
select params.

Combined with the previous EHI_RESUME_LINK differentiation, this makes
parameter selection accurate.  e.g. user scan resumes link but normal
deb param is used instead of hotplug param.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: implement ATA_EHI_RESUME_LINK
Tejun Heo [Mon, 3 Jul 2006 07:07:26 +0000 (16:07 +0900)]
[PATCH] libata: implement ATA_EHI_RESUME_LINK

Implement ATA_EHI_RESUME_LINK, which indicates that the link needs to
be resumed.  This used to be implied by ATA_EHI_HOTPLUGGED.  However,
hotplug isn't the only event which requires link resume and separating
this out allows other places to request link resume.  This
differentiation also allows better debounce timing selection.

This patch converts user scan to use ATA_EHI_RESUME_LINK.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: replace ap_lock w/ ap->lock in ata_scsi_error()
Tejun Heo [Sun, 2 Jul 2006 18:02:15 +0000 (03:02 +0900)]
[PATCH] libata: replace ap_lock w/ ap->lock in ata_scsi_error()

ap_lock was used because &ap->host_set->lock was too long and used a
lot.  Now that &ap->host_set->lock is replaced with ap->lock, there's
no reason to keep ap_lock.

[ed. note: that's not entirely true.  ap_lock is a local variable,
caching the results of a de-ref.  In theory, if the compiler is smart
enough, this patch is cosmetic.  However, since this is not a fast
path (it is the error path), this patch is nonetheless acceptable,
even though it _may_ introduce a performance regression.]

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: fix ehc->i.action setting in ata_eh_autopsy()
Tejun Heo [Sun, 2 Jul 2006 17:54:58 +0000 (02:54 +0900)]
[PATCH] libata: fix ehc->i.action setting in ata_eh_autopsy()

ata_eh_autopsy() used to directly assign determined action mask to
ehc->i.action thus overriding actions set by some of nested analyze
functions.  This patch makes ata_eh_autopsy() add action masks just as
it's done in other places.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: add ap->pflags and move core dynamic flags to it
Tejun Heo [Wed, 28 Jun 2006 16:29:30 +0000 (01:29 +0900)]
[PATCH] libata: add ap->pflags and move core dynamic flags to it

ap->flags is way too clamped.  Separate out core dynamic flags to
ap->pflags.  ATA_FLAG_DISABLED is a dynamic flag but left alone as
it's referenced by a lot of LLDs and it's gonna be removed once all
LLDs are converted to new EH.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] libata: Conditionally set host->max_cmd_len
Brian King [Wed, 28 Jun 2006 13:30:31 +0000 (08:30 -0500)]
[PATCH] libata: Conditionally set host->max_cmd_len

In preparation for SAS attached SATA devices, which will
not have a libata scsi_host, only setup host->max_cmd_len
if ap->host exists.

Signed-off-by: Brian King <brking@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] sata_vsc: data_xfer should use mmio
Martin Hicks [Wed, 28 Jun 2006 16:17:47 +0000 (12:17 -0400)]
[PATCH] sata_vsc: data_xfer should use mmio

Hi,

sata_vsc is an MMIO device, and should use the correct data_xfer
function.  This problem was introduced by:

   commit a6b2c5d4754dc539a560fdf0d3fb78a14174394a
   Author: Alan Cox <alan@lxorguk.ukuu.org.uk>
   Date:   Mon May 22 16:59:59 2006 +0100

        [PATCH] PATCH: libata. Add ->data_xfer method

Signed-off-by: Martin Hicks <mort@bork.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoMerge git://git.linux-nfs.org/pub/linux/nfs-2.6
Linus Torvalds [Wed, 5 Jul 2006 20:05:15 +0000 (13:05 -0700)]
Merge git://git.linux-nfs.org/pub/linux/nfs-2.6

* git://git.linux-nfs.org/pub/linux/nfs-2.6:
  NLM,NFSv4: Wait on local locks before we put RPC calls on the wire
  VFS: Add support for the FL_ACCESS flag to flock_lock_file()
  NFSv4: Ensure nfs4_lock_expired() caches delegated locks
  NLM,NFSv4: Don't put UNLOCK requests on the wire unless we hold a lock
  VFS: Allow caller to determine if BSD or posix locks were actually freed
  NFS: Optimise away an excessive GETATTR call when a file is symlinked
  This fixes a panic doing the first READDIR or READDIRPLUS call when:
  NFS: Fix NFS page_state usage
  Revert "Merge branch 'odirect'"

17 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Wed, 5 Jul 2006 20:03:18 +0000 (13:03 -0700)]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] 3711/1: AT91 timer update
  [ARM] 3709/1: pnx4008: convert to generic irq subsystem
  [ARM] 3710/1: AT91 Serial: Use GPIO API

17 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Wed, 5 Jul 2006 20:02:56 +0000 (13:02 -0700)]
Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/netdev-2.6

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6: (39 commits)
  [PATCH] myri10ge - Export more parameters to ethtool
  [PATCH] myri10ge - Use dev_info() when printing parameters after probe
  [PATCH] myri10ge - Drop ununsed nvidia chipset id
  [PATCH] myri10ge - Drop unused pm_state
  [PATCH] Fix freeing of net device
  [PATCH] remove dead entry in net wan Kconfig
  [PATCH] NI5010 netcard cleanup
  [PATCH] lock validator: fix ns83820.c irq-flags bug
  [PATCH] pcnet32: Cleanup rx buffers after loopback test.
  [PATCH] pcnet32: Suspend the chip rather than restart when changing multicast/promisc
  [PATCH] pcnet32: Handle memory allocation failures cleanly when resizing tx/rx rings
  [PATCH] pcnet32: Use kcalloc instead of kmalloc and memset
  [PATCH] pcnet32: Fix off-by-one in get_ringparam
  [PATCH] pcnet32: Use PCI_DEVICE macro
  [PATCH] pcnet32: Fix Section mismatch error
  [PATCH] Add support for the Cicada 8201 PHY
  [PATCH] zd1211rw: disable TX queue during stop
  [PATCH] ZyDAS ZD1211 USB-WLAN driver
  [PATCH] softmac: fix build-break from 881ee6999d66c8fc903b429b73bbe6045b38c549
  [PATCH] CONFIG_WIRELESS_EXT is neccessary after all
  ...

17 years agoMerge branch 'upstream' of git://lost.foo-projects.org/~ahkok/git/netdev-2.6 into...
Jeff Garzik [Wed, 5 Jul 2006 18:32:39 +0000 (14:32 -0400)]
Merge branch 'upstream' of git://lost.foo-projects.org/~ahkok/git/netdev-2.6 into upstream

Conflicts:

drivers/net/e1000/e1000_main.c

17 years ago[PATCH] 8139too deadlock fix
Arjan van de Ven [Wed, 5 Jul 2006 13:00:40 +0000 (15:00 +0200)]
[PATCH] 8139too deadlock fix

> stack backtrace:

>  [<f9099d31>] rtl8139_start_xmit+0xd9/0xff [8139too]
>  [<c11ad5ea>] netpoll_send_skb+0x98/0xea

This seems to be a real deadlock...

So netpoll_send_skb takes the _xmit_lock, which is all nitty gritty
but then rtl8139_start_xmit comes around while that lock is taken, and
does

      spin_unlock_irq(&tp->lock);

which.. enables interrupts and softirqs; this is quite bad because the
xmit lock is taken in softirq context for the watchdog like this:
  [<c1200376>] _spin_lock+0x23/0x32
  [<c11af282>] dev_watchdog+0x14/0xb1
  [<c101dab2>] run_timer_softirq+0xf2/0x14a
  [<c101a691>] __do_softirq+0x55/0xb0
  [<c1004a8d>] do_softirq+0x58/0xbd

Which would deadlock now that the spin_unlock_irq() has enabled
irqs/softirqs while the _xmit_lock is still held.

The patch below turns this into a irqsave/irqrestore pair so that
interrupts don't get enabled unconditionally.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[netdrvr] 3c59x: snip changelog from source code
Jeff Garzik [Wed, 5 Jul 2006 18:28:34 +0000 (14:28 -0400)]
[netdrvr] 3c59x: snip changelog from source code

Driver source code is not the preferred place to store change history.

Acked-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] myri10ge - Export more parameters to ethtool
Brice Goglin [Mon, 3 Jul 2006 22:16:46 +0000 (18:16 -0400)]
[PATCH] myri10ge - Export more parameters to ethtool

Add the IRQ line, the tx_boundary, and whether Write-combining and MSI
are enabled to the list of parameters that are exported to ethtool.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] myri10ge - Use dev_info() when printing parameters after probe
Brice Goglin [Mon, 3 Jul 2006 22:41:30 +0000 (18:41 -0400)]
[PATCH] myri10ge - Use dev_info() when printing parameters after probe

Displaying the interface name when listing the device parameters
at the end of myri10ge_probe is not a good idea since udev might
rename the interface soon afterwards.
Print the bus id instead, using dev_info().

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] myri10ge - Drop ununsed nvidia chipset id
brice@myri.com [Mon, 3 Jul 2006 22:02:32 +0000 (18:02 -0400)]
[PATCH] myri10ge - Drop ununsed nvidia chipset id

The workaround for the AER capability of the nVidia chipset has been
removed, we don't need this PCI id anymore. Drop it.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] myri10ge - Drop unused pm_state
brice@myri.com [Mon, 3 Jul 2006 22:02:31 +0000 (18:02 -0400)]
[PATCH] myri10ge - Drop unused pm_state

The pm_state field in the myri10ge_priv structure is unused. Drop it.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] Fix freeing of net device
Ralf Baechle [Fri, 30 Jun 2006 12:56:13 +0000 (13:56 +0100)]
[PATCH] Fix freeing of net device

Plus optical sugar.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] remove dead entry in net wan Kconfig
Paul Fulghum [Fri, 30 Jun 2006 09:27:19 +0000 (02:27 -0700)]
[PATCH] remove dead entry in net wan Kconfig

Remove dead entry from net wan Kconfig and net wan Makefile..  This entry is
left over from 2.4 where synclink used syncppp driver directly.  synclink
drivers now use generic HDLC

Signed-off-by: Paul Fulghum <paulkf@microgate.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] NI5010 netcard cleanup
Andreas Mohr [Fri, 30 Jun 2006 09:25:07 +0000 (02:25 -0700)]
[PATCH] NI5010 netcard cleanup

- updated MAINTAINERS entry to new format
- updated Jan-Pascal's (ACKed) and my email address
- driver cleanup/modernization (runtime-, not hardware-tested)

[bunk@stusta.de: build fix]
Signed-off-by: Andreas Mohr <andi@lisas.de>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Jan-Pascal van Best <jvbest@qv3pluto.leidenuniv.nl>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] lock validator: fix ns83820.c irq-flags bug
Ingo Molnar [Fri, 30 Jun 2006 09:25:06 +0000 (02:25 -0700)]
[PATCH] lock validator: fix ns83820.c irq-flags bug

Barry K. Nathan reported the following lockdep warning:

[  197.343948] BUG: warning at kernel/lockdep.c:1856/trace_hardirqs_on()
[  197.345928]  [<c010329b>] show_trace_log_lvl+0x5b/0x105
[  197.346359]  [<c0103896>] show_trace+0x1b/0x20
[  197.346759]  [<c01038ed>] dump_stack+0x1f/0x24
[  197.347159]  [<c012efa2>] trace_hardirqs_on+0xfb/0x185
[  197.348873]  [<c029b009>] _spin_unlock_irq+0x24/0x2d
[  197.350620]  [<e09034e8>] do_tx_done+0x171/0x179 [ns83820]
[  197.350895]  [<e090445c>] ns83820_irq+0x149/0x20b [ns83820]
[  197.351166]  [<c013b4b8>] handle_IRQ_event+0x1d/0x52
[  197.353216]  [<c013c6c2>] handle_level_irq+0x97/0xe1
[  197.355157]  [<c01048c3>] do_IRQ+0x8b/0xac
[  197.355612]  [<c0102d9d>] common_interrupt+0x25/0x2c

this is caused because the ns83820 driver re-enables irq flags
in hardirq context.

While legal in theory, in practice it should only be done if the
hardware is really old and has some very high overhead in its ISR.
(such as PIO IDE)

For modern hardware, running ISRs with irqs enabled is discouraged,
because 1) new hardware is fast enough to not cause latency problems
2) allowing the nesting of hardware interrupts only 'spreads out'
the handling of the current ISR, causing extra cachemisses that would
otherwise not happen. Furthermore, on architectures where ISRs share
the kernel stacks, enabling interrupts in ISRs introduces a much
higher kernel-stack-nesting and thus kernel-stack-overflow risk.
3) not managing irq-flags via the _irqsave / _irqrestore variants
is dangerous: it's easy to forget whether one function nests inside
another, and irq flags might be mismanaged.

In the few cases where re-enabling interrupts in an ISR is considered
useful (and unavoidable), it has to be taught to the lock validator
explicitly (because the lock validator needs the "no ISR ever enables
hardirqs" artificial simplification to keep the IRQ/softirq locking
dependencies manageable).

This teaching is done via the explicit use local_irq_enable_in_hardirq().
On a stock kernel this maps to local_irq_enable(). If the lock validator
is enabled then this does not enable interrupts.

Now, the analysis of drivers/net/ns83820.c's irq flags use: the
irq-enabling in irq context seems intentional, but i dont think it's
justified. Furthermore, the driver suffers from problem #3 above too,
in ns83820_tx_timeout() it disables irqs via local_irq_save(), but
then it calls do_tx_done() which does a spin_unlock_irq(),
re-enabling for a function that does not expect it! While currently
this bug seems harmless (only some debug printout seems to be
affected by it), it's nevertheless something to be fixed.

So this patch makes the ns83820 ISR irq-flags-safe, and cleans up
do_tx_done() use and locking to avoid the ns83820_tx_timeout() bug.

From: Arjan van de Ven <arjan@linux.intel.com>

  ns83820_mib_isr takes the misc_lock in IRQ context.  All other places that
  do this in the ISR already use _irqsave versions, make this consistent at
  least.  At some point in the future someone should audit the driver to see
  if all _irqsave's in the ISR can go away, this is generally an iffy/fragile
  proposition though; for now get it safe, simple and consistent.

From: Arjan van de Ven <arjan@linux.intel.com>

ok this is a real driver deadlock:

The ns83820 driver enabled interrupts (by unlocking the misc_lock with
_irq) while still holding the rx_info.lock, which is required to be irq
safe since it's used in the ISR like this:
                writel(1, dev->base + IER);
                spin_unlock_irq(&dev->misc_lock);
                kick_rx(ndev);
                spin_unlock_irq(&dev->rx_info.lock);

This is can cause a deadlock if an irq was pending at the first
spin_unlock_irq already, or if one would hit during kick_rx().
Simply remove the first _irq solves this

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] pcnet32: Cleanup rx buffers after loopback test.
Don Fry [Thu, 29 Jun 2006 20:55:27 +0000 (13:55 -0700)]
[PATCH] pcnet32: Cleanup rx buffers after loopback test.

More cleanup to pcnet32_loopback_test to release receive buffers if
device is not up.  Created common routine to free rx buffers.

Tested ia32 and ppc64

Signed-off-by: Don Fry <brazilnut@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] pcnet32: Suspend the chip rather than restart when changing multicast/promisc
Don Fry [Thu, 29 Jun 2006 20:55:02 +0000 (13:55 -0700)]
[PATCH] pcnet32: Suspend the chip rather than restart when changing multicast/promisc

Suspend the chip if possible rather than stop and discard all tx and rx
frames, when changing the mcast list or entering/leaving promiscuous
mode.  Created common pcnet32_suspend routine.

Tested ia32 and ppc64

Signed-off-by: Don Fry <brazilnut@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] pcnet32: Handle memory allocation failures cleanly when resizing tx/rx rings
Don Fry [Thu, 29 Jun 2006 20:54:38 +0000 (13:54 -0700)]
[PATCH] pcnet32: Handle memory allocation failures cleanly when resizing tx/rx rings

Fix pcnet32_set_ringparam to handle memory allocation errors without
leaving the adapter in an inoperative state and null pointers waiting to
be dereferenced.

Tested ia32 and ppc64.

Signed-off-by: Don Fry <brazilnut@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] pcnet32: Use kcalloc instead of kmalloc and memset
Don Fry [Thu, 29 Jun 2006 20:53:48 +0000 (13:53 -0700)]
[PATCH] pcnet32: Use kcalloc instead of kmalloc and memset

On 2006-03-08 Eric Sesterhenn wrote:
converts drivers/net to kzalloc usage.

Don Fry modified it to use netif_msg_drv.  Tested ia32 and ppc64.

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: Don Fry <brazilnut@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] pcnet32: Fix off-by-one in get_ringparam
Don Fry [Thu, 29 Jun 2006 20:53:23 +0000 (13:53 -0700)]
[PATCH] pcnet32: Fix off-by-one in get_ringparam

Fix off-by-one in pcnet32_get_ringparam

Signed-off-by: Don Fry <brazilnut@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] pcnet32: Use PCI_DEVICE macro
Don Fry [Thu, 29 Jun 2006 20:52:58 +0000 (13:52 -0700)]
[PATCH] pcnet32: Use PCI_DEVICE macro

Jon Mason wrote on Thu, 12 Jan 2006 17:07:49 -0600:
This patch adds the PCI_DEVICE macro to the pcnet32 driver.

This has been tested on my opteron with my "trident" adapter.

Don Fry modified it slightly and tested on ia32 and ppc64.

Signed-off-by: Jon Mason <jdmason@us.ibm.com>
Signed-off-by: Don Fry <brazilnut@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] pcnet32: Fix Section mismatch error
Don Fry [Thu, 29 Jun 2006 20:52:29 +0000 (13:52 -0700)]
[PATCH] pcnet32: Fix Section mismatch error

Fix Section mismatch error.  Tested ia32 and ppc64.

Signed-off-by: Don Fry <brazilnut@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] Add support for the Cicada 8201 PHY
Kim Phillips [Thu, 29 Jun 2006 02:13:23 +0000 (21:13 -0500)]
[PATCH] Add support for the Cicada 8201 PHY

Add support for the Cicada 8201 PHY, a.k.a Vitesse VSC8201.  This PHY is present on the MPC8349mITX.

Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[PATCH] zd1211rw: disable TX queue during stop
Daniel Drake [Sun, 11 Jun 2006 22:18:54 +0000 (23:18 +0100)]
[PATCH] zd1211rw: disable TX queue during stop

This avoids some potential races.

Signed-off-by: Daniel Drake <dsd@gentoo.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] ZyDAS ZD1211 USB-WLAN driver
Daniel Drake [Fri, 2 Jun 2006 16:11:32 +0000 (17:11 +0100)]
[PATCH] ZyDAS ZD1211 USB-WLAN driver

There are 60+ USB wifi adapters available on the market based on the ZyDAS
ZD1211 chip.

Unlike the predecessor (ZD1201), ZD1211 does not have a hardware MAC, so most
data operations are coordinated by the device driver. The ZD1211 chip sits
alongside an RF transceiver which is also controlled by the driver. Our driver
currently supports 2 RF types, we know of one other available in a few marketed
products which we will be supporting soon.

Our driver also supports the newer revision of ZD1211, called ZD1211B. The
initialization and RF operations are slightly different for the new revision,
but the main difference is 802.11e support. Our driver does not support the
QoS features yet, but we think we know how to use them.

This driver is based on ZyDAS's own GPL driver available from www.zydas.com.tw.
ZyDAS engineers have been responsive and supportive of our efforts, so thumbs
up to them. Additionally, the firmware is redistributable and they have
provided device specs.

This driver has been written primarily by Ulrich Kunitz and myself. Graham
Gower, Greg KH, Remco and Bryan Rittmeyer have also contributed. The
developers of ieee80211 and softmac have made our lives so much easier- thanks!

We maintain a small info-page: http://zd1211.ath.cx/wiki/DriverRewrite

If there is enough time for review, we would like to aim for inclusion in
2.6.18. The driver works nicely as a STA, and can connect to both open and
encrypted networks (we are using software-based encryption for now). We will
work towards supporting more advanced features in the future (ad-hoc, master
mode, 802.11a, ...).

Signed-off-by: Daniel Drake <dsd@gentoo.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] softmac: fix build-break from 881ee6999d66c8fc903b429b73bbe6045b38c549
John W. Linville [Mon, 26 Jun 2006 20:34:29 +0000 (16:34 -0400)]
[PATCH] softmac: fix build-break from 881ee6999d66c8fc903b429b73bbe6045b38c549

Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] CONFIG_WIRELESS_EXT is neccessary after all
Horms [Mon, 26 Jun 2006 08:44:38 +0000 (17:44 +0900)]
[PATCH] CONFIG_WIRELESS_EXT is neccessary after all

WARNING: /lib/modules/2.6.17-mm2/kernel/net/ieee80211/ieee80211.ko
needs unknown symbol wireless_spy_update

Someone removed the `#ifdef CONFIG_WIRELESS_EXT' from around the callsite
in net/ieee80211/ieee80211_rx.c and didn't update Kconfig appropriately.

The offending patchset seems to be 35c14b855f52c49e4f3d078b9532b056005ed321
which is tittled

  [PATCH] ieee80211: remove unnecessary CONFIG_WIRELESS_EXT checking

After a quick look it seems that wireless_spy_update() lives in
net/core/wirless.c, and that file is only compiled if
CONFIG_WIRELESS_EXT is set. Perhaps this is Kconig work, but
in the mean time here is a reversal of the recent change.

Signed-Off-By: Horms <horms@verge.net.au>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] SoftMAC: Add network to ieee80211softmac_call_events when associate times out
Joseph Jezak [Sun, 11 Jun 2006 16:01:03 +0000 (12:01 -0400)]
[PATCH] SoftMAC: Add network to ieee80211softmac_call_events when associate times out

The ieee80211softmac_call_events function, when called with event type
IEEE80211SOFTMAC_EVENT_ASSOCIATE_TIMEOUT should pass the network as the
third parameter.  This patch does that.

Signed-off-by: Joseph Jezak <josejx@gentoo.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] SoftMAC: Prevent multiple authentication attempts on the same network
Joseph Jezak [Sun, 11 Jun 2006 16:00:37 +0000 (12:00 -0400)]
[PATCH] SoftMAC: Prevent multiple authentication attempts on the same network

This patch addresses the "No queue exists" messages commonly seen during
authentication and associating.  These appear due to scheduling multiple
authentication attempts on the same network.  To prevent this, I added a
flag to stop multiple authentication attempts by the association layer.
I also added a check to the wx handler to see if we're connecting to a
different network than the one already in progress.  This scenario was
causing multiple requests on the same network because the network BSSID
was not being updated despite the fact that the ESSID changed.

Signed-off-by: Joseph Jezak <josejx@gentoo.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] bcm43xx: enable shared key authentication
Daniel Drake [Fri, 16 Jun 2006 19:50:22 +0000 (20:50 +0100)]
[PATCH] bcm43xx: enable shared key authentication

I recently patched softmac to enable shared key authentication. This small patch
will enable crazy or unfortunate bcm43xx users to use this new capability.

Signed-off-by: Daniel Drake <dsd@gentoo.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] 2.6.17 missing a call to ieee80211softmac_capabilities from ieee80211softmac_...
Larry Finger [Tue, 20 Jun 2006 14:19:18 +0000 (09:19 -0500)]
[PATCH] 2.6.17 missing a call to ieee80211softmac_capabilities from ieee80211softmac_assoc_req

In commit ba9b28d19a3251bb1dfe6a6f8cc89b96fb85f683, routine
ieee80211softmac_capabilities was added to ieee80211softmac_io.c. As
denoted by its name, it completes the capabilities IE that is
needed in the associate and reassociate requests sent to the
AP. For at least one AP, the Linksys WRT54G V5, the capabilities
field must set the 'short preamble' bit or the AP refuses to
associate. In the commit noted above, there is a call to the
new routine from ieee80211softmac_reassoc_req, but not from
ieee80211softmac_assoc_req. This patch fixes that oversight.

As noted in the subject, v2.6.17 is affected. My bcm43xx card had been
unable to associate since I was forced to buy a new AP. I finally was
able to get a packet dump and traced the problem to the capabilities
info. Although I had heard that a patch was "floating around", I had
not seen it before 2.6.17 was released. As this bug does not affect
security and I seem to have the only AP affected by it, there should
be no problem in leaving it for 2.6.18.

Signed-Off-By: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] Convert bcm43xx-softmac to use the ieee80211_is_valid_channel routine
Larry Finger [Mon, 26 Jun 2006 02:32:21 +0000 (21:32 -0500)]
[PATCH] Convert bcm43xx-softmac to use the ieee80211_is_valid_channel routine

The current version of bcm43xx-softmac uses local routines to check
if a channel is valid. As noted in the comments, these routines do
not take any regulatory information into account. This patch converts
the code to use the equivalent routine in ieee80211, which is being
converted to know about regulatory information.

Signed-Off-By: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] skb used after passing to netif_rx in net/ieee80211/ieee80211_rx.c
Eric Sesterhenn [Wed, 21 Jun 2006 19:05:58 +0000 (21:05 +0200)]
[PATCH] skb used after passing to netif_rx in net/ieee80211/ieee80211_rx.c

this patch fixes coverity id #913. ieee80211_monitor_rx() passes the skb
to netif_rx() and we should not reference it any longer.

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] ieee80211: fix not allocating IV+ICV space when usingencryption in ieee80211_...
Hong Liu [Wed, 21 Jun 2006 03:35:08 +0000 (11:35 +0800)]
[PATCH] ieee80211: fix not allocating IV+ICV space when usingencryption in ieee80211_tx_frame

We should preallocate IV+ICV space when encrypting the frame.
Currently no problem shows up just because dev_alloc_skb aligns the
data len to SMP_CACHE_BYTES which can be used for ICV.

Signed-off-by: Hong Liu <hong.liu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
17 years ago[PATCH] bcm43xx: workaround init_board vs. IRQ race
Michael Buesch [Sat, 17 Jun 2006 13:19:05 +0000 (15:19 +0200)]
[PATCH] bcm43xx: workaround init_board vs. IRQ race

Place the Init-vs-IRQ workaround before any card register
access, because we might not have the wireless core mapped
at all times in init. So this will result in a Machine Check
caused by a bus error.

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>