GitHub/LineageOS/android_kernel_samsung_universal7580.git
16 years agosh: Add defconfig for RSK7203.
Paul Mundt [Tue, 3 Jun 2008 11:30:10 +0000 (20:30 +0900)]
sh: Add defconfig for RSK7203.

RSK7203 is supportable through the generic machvec, so we add a defconfig
for those bits. This gets updated with more complete board support later.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
16 years agosh: Update SE7206 defconfig.
Paul Mundt [Tue, 3 Jun 2008 11:29:55 +0000 (20:29 +0900)]
sh: Update SE7206 defconfig.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
16 years agosh: Disable 4KSTACKS on nommu.
Paul Mundt [Mon, 2 Jun 2008 04:48:59 +0000 (13:48 +0900)]
sh: Disable 4KSTACKS on nommu.

4k stacks on nommu ends up blowing up with all sorts of interesting
slab corruption. Disable this by default unless BROKEN is also
enabled.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
16 years agosh: fix miscompilation of ip_fast_csum with gcc >= 4.3
Paul Mundt [Mon, 2 Jun 2008 03:40:14 +0000 (12:40 +0900)]
sh: fix miscompilation of ip_fast_csum with gcc >= 4.3

As noted by Matthew Wilcox:

    Kyle McMartin just tracked down a bug on parisc to a missing
    "memory" clobber in the inline assembly implementation of
    ip_fast_csum.  The FRV, SH and Xtensa ports are also missing a
    memory clobber, so I thought it would be polite to let you know.

    The bug manifests as dropped network packets (obviously they have
    the wrong checksum).  It started appearing for parisc with GCC 4.3.

    The GCC manual says:

     If your assembler instructions access memory in an unpredictable
     fashion, add `memory' to the list of clobbered registers.  This
     will cause GCC to not keep memory values cached in registers
     across the assembler instruction and not optimize stores or loads
     to that memory.

    I see that FRV has a 400 byte memory output which may prevent this
    problem from appearing, but SH and Xtensa have nothing to prevent
    this bug.  Hope this saves you a few days of debugging.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
16 years agosh: module.c use kernel unaligned helpers
Harvey Harrison [Wed, 28 May 2008 23:38:17 +0000 (16:38 -0700)]
sh: module.c use kernel unaligned helpers

Replace the COPY_UNALIGNED_WORD helper.

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
16 years agosh/kernel/cpu/irq/intc-sh5.c build fix
Adrian Bunk [Thu, 29 May 2008 22:04:49 +0000 (01:04 +0300)]
sh/kernel/cpu/irq/intc-sh5.c build fix

This patch fixes the following build error caused by
commit a1dc4b59fa4af97ae68ee214d4d72bbd7c7ec1dc
(sh: intc_sh5 depends on cayman board for IRQ priority table.):

<--  snip  -->

...
  CC      arch/sh/kernel/cpu/irq/intc-sh5.o
/home/bunk/linux/kernel-2.6/git/linux-2.6/arch/sh/kernel/cpu/irq/intc-sh5.c: In function 'plat_irq_setup':
/home/bunk/linux/kernel-2.6/git/linux-2.6/arch/sh/kernel/cpu/irq/intc-sh5.c:257: error: expected declaration or statement at end of input
make[4]: *** [arch/sh/kernel/cpu/irq/intc-sh5.o] Error 1

<--  snip  -->

Reported-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Fri, 30 May 2008 14:45:20 +0000 (07:45 -0700)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (26 commits)
  llc: Fix double accounting of received packets
  netfilter: nf_conntrack_expect: fix error path unwind in nf_conntrack_expect_init()
  bluetooth: fix locking bug in the rfcomm socket cleanup handling
  mac80211: fix alignment issue with compare_ether_addr()
  mac80211: Fix for NULL pointer dereference in sta_info_get()
  mac80211: fix a typo in ieee80211_handle_filtered_frame comment
  rndis_wlan: add missing range check for power_output modparam
  iwlwifi: fix rate scale TLC column selection bug
  iwlwifi: fix exit from stay_in_table state
  rndis_wlan: Make connections to TKIP PSK networks work
  mac80211 : Fixes the status message for iwconfig
  rt2x00: Use atomic interface iteration in irq context
  rt2x00: Reset antenna RSSI after switch
  rt2x00: Don't count retries as failure
  rt2x00: Fix memleak in tx() path
  mac80211: reorder channel and freq reporting in wext scan report
  b43: Fix controller restart crash
  mac80211: fix ieee80211_rx_bss_put/get imbalance
  net/mac80211: always true conditionals
  b43: Upload both beacon templates on initial load
  ...

16 years agoMerge branch 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6
Linus Torvalds [Fri, 30 May 2008 14:44:19 +0000 (07:44 -0700)]
Merge branch 'for-linus' of git://git390.osdl.marist.edu/linux-2.6

* 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6:
  [S390] Update default configuration.
  [S390] disassembler: fix idte instruction format.
  [S390] tape: fix race with stack local wait_queue_head_t.
  [S390] 3270: fix race with stack local wait_queue_head_t.
  [S390] dasd: use a generic wait_queue for sleep_on
  [S390] sclp_vt220: fix scheduling while atomic bug.
  [S390] showmem: Only walk spanned pages.
  [S390] appldata: prevent cpu hotplug when walking cpu_online_map.
  [S390] Fix section mismatch warnings.
  [S390] s390 types: make dma_addr_t 64 bit capable
  [S390] tape: Fix race condition in tape block device driver
  [S390] fix sparsemem related compile error with allnoconfig on s390

16 years agoacpi: fix sparse const errors
Harvey Harrison [Fri, 30 May 2008 00:51:57 +0000 (17:51 -0700)]
acpi: fix sparse const errors

In this case we want a constant pointer to constant chars:

drivers/misc/thinkpad_acpi.c:3824:19: error: Just how const do you want this type to be?

Like the error says.

drivers/misc/thinkpad_acpi.c:3863:19: error: Just how const do you want this type to be?
drivers/misc/thinkpad_acpi.c:3864:19: error: Just how const do you want this type to be?
drivers/misc/thinkpad_acpi.c:3865:19: error: Just how const do you want this type to be?
drivers/misc/thinkpad_acpi.c:3866:19: error: Just how const do you want this type to be?

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Acked-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agollc: Fix double accounting of received packets
Arnaldo Carvalho de Melo [Fri, 30 May 2008 09:57:29 +0000 (02:57 -0700)]
llc: Fix double accounting of received packets

llc_sap_rcv was being preceded by skb_set_owner_r, then calling
llc_state_process that calls sock_queue_rcv_skb, that in turn calls
skb_set_owner_r again making the space allowed to be used by the socket to be
leaked, making the socket to get stuck.

Fix it by setting skb->sk at llc_sap_rcv and leave the accounting to be done
only at sock_queue_rcv_skb.

Reported-by: Dmitry Petukhov <dmgenp@gmail.com>
Tested-by: Dmitry Petukhov <dmgenp@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[S390] Update default configuration.
Martin Schwidefsky [Fri, 30 May 2008 08:03:35 +0000 (10:03 +0200)]
[S390] Update default configuration.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] disassembler: fix idte instruction format.
Martin Schwidefsky [Fri, 30 May 2008 08:03:34 +0000 (10:03 +0200)]
[S390] disassembler: fix idte instruction format.

The correct instruction format of idte is "idte r1,r3,r2" with
r1 at bit 24, r3 at bit 16 and r2 at bit 28.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] tape: fix race with stack local wait_queue_head_t.
Martin Schwidefsky [Fri, 30 May 2008 08:03:33 +0000 (10:03 +0200)]
[S390] tape: fix race with stack local wait_queue_head_t.

A wait_event call with a stack local wait_queue_head_t structure that is
used to do the wake up for the wait_event is inherently racy. After the
wait_event finished the wake_up call might not have completed yet.
Replace the stack local wait_queue_head_t in tape_do_io and
tape_do_io_interruptible with a per device wait queue.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] 3270: fix race with stack local wait_queue_head_t.
Martin Schwidefsky [Fri, 30 May 2008 08:03:32 +0000 (10:03 +0200)]
[S390] 3270: fix race with stack local wait_queue_head_t.

A wait_event call with a stack local wait_queue_head_t structure that is
used to do the wake up for the wait_event is inherently racy. After the
wait_event finished the wake_up call might not have completed yet.
Remove the stack local wait_queue_head_t from raw3270_start_init and
use the global raw3270_wait_queue instead.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] dasd: use a generic wait_queue for sleep_on
Stefan Haberland [Fri, 30 May 2008 08:03:31 +0000 (10:03 +0200)]
[S390] dasd: use a generic wait_queue for sleep_on

Use a generic wait_queue to prevent the wait_queue in dasd_sleep_on_
functions from being referenced by callback_data while it does not
exist any more.

Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] sclp_vt220: fix scheduling while atomic bug.
Heiko Carstens [Fri, 30 May 2008 08:03:30 +0000 (10:03 +0200)]
[S390] sclp_vt220: fix scheduling while atomic bug.

The driver incorrectly assumed that putchar will only be called from
schedulable process context and therefore blocked and waited if no
free output buffers where available.
Since putchar may also be called from BH context this may lead to
deadlocks.
To fix this just return the number of characters accepted and let the
upper layer handle the rest.

The console write function will busy wait (sclp_sync_wait) until a
buffer is available again.

Cc: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] showmem: Only walk spanned pages.
Heiko Carstens [Fri, 30 May 2008 08:03:29 +0000 (10:03 +0200)]
[S390] showmem: Only walk spanned pages.

Convert show_mem() so its nearly the same as on x86/powerpc.
Gives us proper locking and we get also rid of the only use of max_mapnr.
Also the number of pages was contained in an int which might not be
sufficient not too far in the future.

Cc: Johannes Weiner <hannes@saeurebad.de>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] appldata: prevent cpu hotplug when walking cpu_online_map.
Gerald Schaefer [Fri, 30 May 2008 08:03:28 +0000 (10:03 +0200)]
[S390] appldata: prevent cpu hotplug when walking cpu_online_map.

Use get_online_cpus() to prevent cpu hotplug in situations where
for_each_online_cpu() is called.

Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Fix section mismatch warnings.
Heiko Carstens [Fri, 30 May 2008 08:03:27 +0000 (10:03 +0200)]
[S390] Fix section mismatch warnings.

This fixes the last remaining section mismatch warnings in s390
architecture code. It reveals also a real bug introduced by... me
with git commit 2069e978d5a6e7b45d58027e3de7f879b8c5e488
("[S390] sparsemem vmemmap: initialize memmap.")

Calling the generic vmemmap_alloc_block() function to get initialized
memory is a nice idea, however that function is __meminit annotated
and therefore the function might be gone if we try to call it later.
This can happen if a DCSS segment gets added.

So basically revert the patch and clear the memmap explicitly to fix
the original bug.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] s390 types: make dma_addr_t 64 bit capable
Christian Borntraeger [Fri, 30 May 2008 08:03:26 +0000 (10:03 +0200)]
[S390] s390 types: make dma_addr_t 64 bit capable

virtio tests with guests larger than 4 GB revealed that the dma_addr_t
definition for s390 did not make it into the 64bit world.
This patch changes the definition on s390 to have an u64 on 64bit and
u32 on 32bit systems.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] tape: Fix race condition in tape block device driver
Michael Holzheu [Fri, 30 May 2008 08:03:25 +0000 (10:03 +0200)]
[S390] tape: Fix race condition in tape block device driver

Due to incorrect function call sequence it can happen that a tape block
request is finished before the request is taken from the block request queue.

The following sequence leads to that condition:
 * tapeblock_start_request() -> start CCW program
 * Request finishes -> IO interrupt
 * tapeblock_end_request()
 * end_that_request_last()

If blkdev_dequeue_request() has not been called before end_that_request_last(),
a kernel bug is triggered in end_that_request_last() because the request is
still queued. To solve that problem blkdev_dequeue_request() has to be called
before starting the CCW program.

Signed-off-by: Michael Holzheu <holzheu@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] fix sparsemem related compile error with allnoconfig on s390
Hans-Joachim Picht [Fri, 30 May 2008 08:03:24 +0000 (10:03 +0200)]
[S390] fix sparsemem related compile error with allnoconfig on s390

On s390 make allnoconfig fails with the following build error:

arch/s390/mm/init.c: In function 'show_mem':
arch/s390/mm/init.c:55: error: implicit declaration of function 'pfn_valid'
make[1]: *** [arch/s390/mm/init.o] Error 1
make: *** [arch/s390/mm] Error 2

This problem can by fixed ensuring that ARCH_SELECT_MEMORY_MODEL
is always turned on.

Signed-off-by: Hans-Joachim Picht <hans@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-2.6
Linus Torvalds [Fri, 30 May 2008 04:29:39 +0000 (21:29 -0700)]
Merge git://git./linux/kernel/git/gregkh/driver-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-2.6:
  driver-core: prepare for 2.6.27 api change by adding dev_set_name

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6
Linus Torvalds [Fri, 30 May 2008 04:27:53 +0000 (21:27 -0700)]
Merge git://git./linux/kernel/git/gregkh/usb-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6:
  Revert "USB: EHCI: fix performance regression"
  USB: fsl_usb2_udc: fix recursive lock
  USB: usb-serial: option: Don't match Huawei driver CD images
  USB: pl2303: another product ID
  USB: add another scanner quirk
  USB: Add support for ROKR W5 in unusual_devs.h
  USB: Fix M600i unusual_devs entry
  USB: usb-storage: unusual_devs update for Cypress ATACB
  USB: EHCI: fix performance regression
  USB: EHCI: fix bug in Iso scheduling
  USB: EHCI: fix remote-wakeup regression
  USB: EHCI: suppress unwanted error messages
  USB: EHCI: fix up root-hub TT mess
  USB: add all configs to the "descriptors" attribute
  USB: fix possible deadlock involving sysfs attributes
  USB: Firmware loader driver for USB Apple iSight camera
  USB: FTDI_SIO : Add support for Matrix Orbital PID Range

16 years agodriver-core: prepare for 2.6.27 api change by adding dev_set_name
Stephen Rothwell [Fri, 30 May 2008 00:16:40 +0000 (10:16 +1000)]
driver-core: prepare for 2.6.27 api change by adding dev_set_name

Create the dev_set_name function now so that various subsystems can
start changing over to it before other changes in 2.6.27 will make it
compulsory.

Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoRevert "USB: EHCI: fix performance regression"
Greg Kroah-Hartman [Fri, 30 May 2008 02:43:27 +0000 (19:43 -0700)]
Revert "USB: EHCI: fix performance regression"

This reverts commit fa38dfcc56b5f6cce787f9aaa5d1830509213802.

It wasn't really a regression and David and Alan are still working
through the issues reported.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: fsl_usb2_udc: fix recursive lock
Li Yang [Thu, 29 May 2008 13:04:45 +0000 (21:04 +0800)]
USB: fsl_usb2_udc: fix recursive lock

UDC needs to release lock before calling out to gadget driver, since
it may need to reenter.  The change fixes kernel BUG observed on rt
kernel.

> kernel BUG at kernel/rtmutex.c:683!
> stopped custom tracer.
> Oops: Exception in kernel mode, sig: 5 [#1]
> PREEMPT MPC834x ITX
> NIP: c021629c LR: c0216270 CTR: 00000000
> REGS: df761d70 TRAP: 0700   Not tainted  (2.6.23.9-rt13)
> MSR: 00021032 <ME,IR,DR>  CR: 28000022  XER: 00000000
> TASK = df632080[241] 'IRQ-38' THREAD: df760000
> GPR00: 00000001 df761e20 df632080 00000000 11111111 00000000 df761e6c
00000000
> GPR08: df761e48 00000000 df761e50 00000000 80000000 ede5cdde 1fffd000
00800000
> GPR16: ffffffff 00000000 007fff00 00000040 00000000 007ffeb0 00000000
1fff8b08
> GPR24: 00000000 00000026 00000000 df79a320 c026b2e8 c02240bc 00009032
df79a320
> NIP [c021629c] rt_spin_lock_slowlock+0x9c/0x200
> LR [c0216270] rt_spin_lock_slowlock+0x70/0x200
> Call Trace:
> [df761e20] [c0216270] rt_spin_lock_slowlock+0x70/0x200 (unreliable)
> [df761e90] [c0182828] fsl_ep_disable+0xcc/0x154
> [df761eb0] [c0184d30] eth_reset_config+0x88/0x1d0
> [df761ed0] [c0184ec0] eth_disconnect+0x48/0x64
> [df761ef0] [c01831a4] reset_queues+0x60/0x78
> [df761f00] [c0183b74] fsl_udc_irq+0x9b8/0xa58
> [df761f50] [c003ef30] handle_IRQ_event+0x64/0x100
> [df761f80] [c003f758] thread_simple_irq+0x6c/0xc8
> [df761fa0] [c003f888] do_irqd+0xd4/0x2e4
> [df761fd0] [c0032284] kthread+0x50/0x8c
> [df761ff0] [c000f9b4] kernel_thread+0x44/0x60

Signed-off-by: Li Yang <leoli@freescale.com>
Cc: Eugene T. Bordenkircher <Eugene_Bordenkircher@selinc.com>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: usb-serial: option: Don't match Huawei driver CD images
Michael Karcher [Wed, 28 May 2008 21:58:18 +0000 (23:58 +0200)]
USB: usb-serial: option: Don't match Huawei driver CD images

Add the interface info matching to all Huawei cards, as they all also
contain a Mass Storage Device interface (usually containing Windows
drivers) which should not get bound by this driver.

See also drivers/usb/storage/unusual_devs.h

Signed-off-by: Michael Karcher <kernel@mkarcher.dialup.fu-berlin.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: pl2303: another product ID
Steve Murphy [Fri, 23 May 2008 18:09:05 +0000 (23:39 +0530)]
USB: pl2303: another product ID

I've just got a USB GPRS/EDGE modem branded Manufacturer Micromax Model
MMX610U (see http://www.airtel.in/level2_t3data.aspx?path=1/106/179)
working by adding another product ID to pl2303.  Modem info reports same
module as  Max Arnold's i.e.SIMCOM SIM600  but  with product ID 0x0612
(cf Ox0611).

From: Steve Murphy <steve@gnusis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: add another scanner quirk
René Rebe [Tue, 27 May 2008 07:05:46 +0000 (09:05 +0200)]
USB: add another scanner quirk

Like the HP53{00,70} scanner other devices of the OEM Avision require
the USB_QUIRK_STRING_FETCH_255 to correct set a configuration with
"recent" Linux kernels.

Signed-off-by: René Rebe <rene@exactcode.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: Add support for ROKR W5 in unusual_devs.h
Javier Smaldone [Mon, 26 May 2008 19:44:00 +0000 (21:44 +0200)]
USB: Add support for ROKR W5 in unusual_devs.h

This patch adds support for rev 2 of an existing unusual_devs entry
enabling ROKR W5s to work. Greg, please apply.

From: Javier Smaldone <javier@smaldone.com.ar>
Signed-off-by: Phil Dibowitz <phil@ipom.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: Fix M600i unusual_devs entry
Phil Dibowitz [Mon, 26 May 2008 19:33:58 +0000 (21:33 +0200)]
USB: Fix M600i unusual_devs entry

It turns out that the unusual_devs entry for the Motorola M600i needs
another flag. This patch adds it. Thanks to Atte André Jensen
<atte@ballbreaker.dk>.

Signed-off-by: Phil Dibowitz <phil@ipom.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: usb-storage: unusual_devs update for Cypress ATACB
Alan Stern [Wed, 21 May 2008 17:53:01 +0000 (13:53 -0400)]
USB: usb-storage: unusual_devs update for Cypress ATACB

This patch (as1101) updates the unusual_devs entry for the Cypress
ATACB pass-through.  The protocol field is changed from US_PR_BULK to
US_PR_DEVICE, since the Cypress devices already set bInterfaceProtocol
to Bulk-only.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: EHCI: fix performance regression
Alan Stern [Tue, 20 May 2008 20:59:33 +0000 (16:59 -0400)]
USB: EHCI: fix performance regression

This patch (as1099) fixes a performance regression in ehci-hcd.  The
fundamental problem is that queue headers get removed from the
schedule too quickly, since the code checks for a counter advancing
rather than making an actual time-based check.  The latency involved
in removing the queue header and then relinking it can severely
degrade certain kinds of workloads.

The patch replaces a simple counter with a timestamp derived from the
controller's uframe value.  In addition, the delay for unlinking an
idle queue header is increased from 5 ms to 10 ms; since some
controllers (nVidia) have a latency of up to 1 ms for unlinking, this
reduces the relative impact from 20% to 10%.

Finally, a logical error left over from the IAA watchdog-timer
conversion is corrected.  Now the driver will always either unlink an
idle queue header or set up a timer to unlink it later.  The old code
would sometimes fail to do either.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Cc: David Brownell <david-b@pacbell.net>
Cc: Leonid <leonidv11@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: EHCI: fix bug in Iso scheduling
Alan Stern [Tue, 20 May 2008 20:59:10 +0000 (16:59 -0400)]
USB: EHCI: fix bug in Iso scheduling

This patch (as1098) changes the way ehci-hcd schedules its periodic
Iso transfers.  That the current scheduling code is wrong is clear on
the face of it: Sometimes it returns -EL2NSYNC (meaning that an URB
couldn't be scheduled because it was submitted too late), but it does
this even when the URB_ISO_ASAP flag is set (meaning the URB should be
scheduled as soon as possible).

The new code properly implements as-soon-as-possible scheduling,
assigning the next unexpired slot as the URB's starting point.  It
also is more careful about checking for Iso URB completion: It doesn't
bother to check for activity during frames that are already over,
and it allows for the possibility that some of the URB's packets may
have raced the hardware when they were submitted and so never got used
(the packet status is set to -EXDEV).

This fixes problems several people have experienced with USB video
applications.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: EHCI: fix remote-wakeup regression
Alan Stern [Tue, 20 May 2008 20:58:58 +0000 (16:58 -0400)]
USB: EHCI: fix remote-wakeup regression

This patch (as1097) fixes a bug in the remote-wakeup handling in
ehci-hcd.  The driver currently does not keep track of whether the
change-suspend feature is enabled for each port; the feature is
automatically reset the first time it is read.  But recent changes to
the hub driver require that the feature be read at least twice in
order to work properly.

A bit-vector is added for storing the change-suspend feature values.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: EHCI: suppress unwanted error messages
Alan Stern [Tue, 20 May 2008 20:58:29 +0000 (16:58 -0400)]
USB: EHCI: suppress unwanted error messages

This patch (as1096) fixes an annoying problem: When a full-speed or
low-speed device is plugged into an EHCI controller, it fails to
enumerate at high speed and then is handed over to the companion
controller.  But usbcore logs a misleading and unwanted error message
when the high-speed enumeration fails.

The patch adds a new HCD method, port_handed_over, which asks whether
a port has been handed over to a companion controller.  If it has, the
error message is suppressed.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
CC: David Brownell <david-b@pacbell.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: EHCI: fix up root-hub TT mess
Alan Stern [Tue, 20 May 2008 20:58:11 +0000 (16:58 -0400)]
USB: EHCI: fix up root-hub TT mess

This patch (as1095) cleans up the HCD glue and several of the EHCI
bus-glue files.  The ehci->is_tdi_rh_tt flag is redundant, since it
means the same thing as the hcd->has_tt flag, so it is removed and the
other flag used in its place.

Some of the bus-glue files didn't get the relinquish_port method added
to their hc_driver structures.  Although that routine currently
doesn't do anything for controllers with an integrated TT, in the
future it might.  So the patch adds it where it is missing.

Lastly, some of the bus-glue files have erroneous entries for their
hc_driver's suspend and resume methods.  These method pointers are
specific to PCI and shouldn't be used otherwise.

(The patch also includes an invisible whitespace fix.)

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
16 years agoUSB: add all configs to the "descriptors" attribute
Alan Stern [Tue, 20 May 2008 20:40:42 +0000 (16:40 -0400)]
USB: add all configs to the "descriptors" attribute

This patch (as1094) changes the output of the "descriptors" binary
attribute.  Now it will contain the device descriptor followed by all
the configuration descriptors, not just the descriptor for the current
config.

Userspace libraries want to have access to the kernel's cached
descriptor information, so they can learn about device characteristics
without having to wake up suspended devices.  So far the only user of
this attribute is the new libusb-1.0 library; thus changing its
contents shouldn't cause any problems.

This should be considered for 2.6.26, if for no other reason than to
minimize the range of releases in which the attribute contains only the
current config descriptor.

Also, it doesn't hurt that the patch removes the device locking --
which was formerly needed in order to know for certain which config was
indeed current.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: fix possible deadlock involving sysfs attributes
Alan Stern [Tue, 20 May 2008 20:37:34 +0000 (16:37 -0400)]
USB: fix possible deadlock involving sysfs attributes

There is a potential deadlock when the usb_generic driver is unbound
from a device.  The problem is that generic_disconnect() is called
with the device lock held, and it removes a bunch of device attributes
from sysfs.  If a user task happens to be running an attribute method
at the time, the removal will block until the method returns.  But at
least one of the attribute methods (the store routine for power/level)
needs to acquire the device lock!

This patch (as1093) eliminates the deadlock by moving the calls to
create and remove the sysfs attributes from the usb_generic driver
into usb_new_device() and usb_disconnect(), where they can be invoked
without holding the device lock.

Besides, the other sysfs attributes are created when the device is
registered and removed when the device is unregistered.  So it seems
only fitting for the extra attributes to be created and removed at the
same time.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: Firmware loader driver for USB Apple iSight camera
Matthew Garrett [Tue, 20 May 2008 19:06:28 +0000 (20:06 +0100)]
USB: Firmware loader driver for USB Apple iSight camera

Uninitialised Apple iSight drivers present with a distinctive USB ID.
Once firmware has been uploaded, they disconnect and reconnect with a
new ID. At this point they can be driven by the uvcvideo driver. As this
is unique to the Apple cameras and not functionality shared by any other
UVC devices, it makes sense to provide the firmware loading
functionality in a separate driver. This driver will read an isight.fw
file extracted from the Apple driver using the tools at
http://bersace03.free.fr/ift/ and upload it to the camera. It will also
handle the case where the device loses its firmware during hibernation
and must have it reloaded.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: FTDI_SIO : Add support for Matrix Orbital PID Range
Ray Molenkamp [Wed, 21 May 2008 23:06:26 +0000 (17:06 -0600)]
USB: FTDI_SIO : Add support for Matrix Orbital PID Range

This patch adds support for the range of PIDs
that have been allocated for FTDI based devices
at Matrix Orbital.

A small number of units have been shipped early 2008
with a faulty USB Descriptor.  Products that may have
this issue have been marked with the existing quirk to
work around the problem.

Signed-off-by: R. Molenkamp <rmolenkamp@matrixorbital.ca>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq
Linus Torvalds [Thu, 29 May 2008 17:04:54 +0000 (10:04 -0700)]
Merge branch 'fixes' of git://git./linux/kernel/git/davej/cpufreq

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq:
  [CPUFREQ] fix double unlock of cpu_policy_rwsem in drivers/cpufreq/cpufreq.c

16 years agoMerge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 29 May 2008 16:26:17 +0000 (09:26 -0700)]
Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched: re-tune NUMA topologies
  sched: stop wake_affine from causing serious imbalance
  sched: fix sched_clock_cpu()
  revert ("sched: fair-group: SMP-nice for group scheduling")
  sched: cleanup
  show_schedstat(): fix memleak
  sched: unite unlikely pairs in rt_policy() and schedule_debug()
  revert ("sched: fair: weight calculations")

16 years ago[CPUFREQ] fix double unlock of cpu_policy_rwsem in drivers/cpufreq/cpufreq.c
Lothar Waßmann [Thu, 29 May 2008 15:54:52 +0000 (17:54 +0200)]
[CPUFREQ] fix double unlock of cpu_policy_rwsem in drivers/cpufreq/cpufreq.c

In drivers/cpufreq/cpufreq.c the function cpufreq_add_dev() takes the
error exit 'err_out_unregister' from different places once with the
'cpu_policy_rwsem' lock held, once with the lock released:
| if (ret)
| goto err_out_unregister;
| }
|
| policy->governor = NULL; /* to assure that the starting sequence is
|   * run in cpufreq_set_policy */
|
| /* set default policy */
| ret = __cpufreq_set_policy(policy, &new_policy);
| policy->user_policy.policy = policy->policy;
| policy->user_policy.governor = policy->governor;
|
| unlock_policy_rwsem_write(cpu);
|
| if (ret) {
| dprintk("setting policy failed\n");
| goto err_out_unregister;
| }

This leads to the following error message in case of a failing
__cpufreq_set_policy() call:
=====================================
[ BUG: bad unlock balance detected! ]
-------------------------------------
swapper/1 is trying to release lock (&per_cpu(cpu_policy_rwsem, cpu)) at:
[<c01b4564>] unlock_policy_rwsem_write+0x30/0x40
but there are no more locks to release!

other info that might help us debug this:
1 lock held by swapper/1:
 #0:  (sysdev_drivers_lock){--..}, at: [<c018fd18>] sysdev_driver_register+0x74/0x130

stack backtrace:
[<c002f588>] (dump_stack+0x0/0x14) from [<c00692fc>] (print_unlock_inbalance_bug+0xc8/0x104)
[<c0069234>] (print_unlock_inbalance_bug+0x0/0x104) from [<c006b7ac>] (lock_release_non_nested+0xc4/0x19c)
 r6:00000028 r5:c3c1ab80 r4:c01b4564
[<c006b6e8>] (lock_release_non_nested+0x0/0x19c) from [<c006b9e0>] (lock_release+0x15c/0x18c)
 r8:60000013 r7:00000001 r6:c01b4564 r5:c0541bb4 r4:c3c1ab80
[<c006b884>] (lock_release+0x0/0x18c) from [<c0061ba0>] (up_write+0x24/0x30)
 r8:c0541b80 r7:00000000 r6:ffffffea r5:c3c34828 r4:c0541b8c
[<c0061b7c>] (up_write+0x0/0x30) from [<c01b4564>] (unlock_policy_rwsem_write+0x30/0x40)
 r4:c3c34884
[<c01b4534>] (unlock_policy_rwsem_write+0x0/0x40) from [<c01b4c40>] (cpufreq_add_dev+0x324/0x398)
[<c01b491c>] (cpufreq_add_dev+0x0/0x398) from [<c018fd64>] (sysdev_driver_register+0xc0/0x130)
[<c018fca4>] (sysdev_driver_register+0x0/0x130) from [<c01b3574>] (cpufreq_register_driver+0xbc/0x174)

Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: Dave Jones <davej@redhat.com>
16 years agoMerge commit 'linus/master' into sched-fixes-for-linus
Ingo Molnar [Thu, 29 May 2008 14:05:05 +0000 (16:05 +0200)]
Merge commit 'linus/master' into sched-fixes-for-linus

16 years agosched: re-tune NUMA topologies
Ingo Molnar [Thu, 29 May 2008 12:32:23 +0000 (14:32 +0200)]
sched: re-tune NUMA topologies

improve the sysbench ramp-up phase and its peak throughput on
a 16way NUMA box, by turning on WAKE_AFFINE:

             tip/sched   tip/sched+wake-affine
-------------------------------------------------
    1:             700              830    +15.65%
    2:            1465             1391    -5.28%
    4:            3017             3105    +2.81%
    8:            5100             6021    +15.30%
   16:           10725            10745    +0.19%
   32:           10135            10150    +0.16%
   64:            9338             9240    -1.06%
  128:            8599             8252    -4.21%
  256:            8475             8144    -4.07%
-------------------------------------------------
  SUM:           57558            57882    +0.56%

this change also improves lat_ctx from 6.69 usecs to 1.11 usec:

  $ ./lat_ctx -s 0 2
  "size=0k ovr=1.19
  2 1.11

  $ ./lat_ctx -s 0 2
  "size=0k ovr=1.22
  2 6.69

in sysbench it's an overall win with some weakness at the lots-of-clients
side. That happens because we now under-balance this workload
a bit. To counter that effect, turn on NEWIDLE:

              wake-idle          wake-idle+newidle
 -------------------------------------------------
     1:             830              834    +0.43%
     2:            1391             1401    +0.65%
     4:            3105             3091    -0.43%
     8:            6021             6046    +0.42%
    16:           10745            10736    -0.08%
    32:           10150            10206    +0.55%
    64:            9240             9533    +3.08%
   128:            8252             8355    +1.24%
   256:            8144             8384    +2.87%
 -------------------------------------------------
   SUM:           57882            58591    +1.21%

as a bonus this not only improves the many-clients case but
also improves the (more important) rampup phase.

sysbench is a workload that quickly breaks down if the
scheduler over-balances, so since it showed an improvement
under NEWIDLE this change is definitely good.

16 years agonetfilter: nf_conntrack_expect: fix error path unwind in nf_conntrack_expect_init()
Alexey Dobriyan [Thu, 29 May 2008 10:19:37 +0000 (03:19 -0700)]
netfilter: nf_conntrack_expect: fix error path unwind in nf_conntrack_expect_init()

Signed-off-by: Alexey Dobriyan <adobriyan@parallels.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agosched: stop wake_affine from causing serious imbalance
Mike Galbraith [Thu, 29 May 2008 09:11:41 +0000 (11:11 +0200)]
sched: stop wake_affine from causing serious imbalance

Prevent short-running wakers of short-running threads from overloading a single
cpu via wakeup affinity, and wire up disconnected debug option.

Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agosched: fix sched_clock_cpu()
Peter Zijlstra [Thu, 29 May 2008 08:07:15 +0000 (10:07 +0200)]
sched: fix sched_clock_cpu()

Make sched_clock_cpu() return 0 before it has been initialized and avoid
corrupting its state due to doing so.

This fixes the weird printk timestamp jump reported.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
16 years agorevert ("sched: fair-group: SMP-nice for group scheduling")
Ingo Molnar [Thu, 29 May 2008 09:28:57 +0000 (11:28 +0200)]
revert ("sched: fair-group: SMP-nice for group scheduling")

Yanmin Zhang reported:

Comparing with 2.6.25, volanoMark has big regression with kernel 2.6.26-rc1.
It's about 50% on my 8-core stoakley, 16-core tigerton, and Itanium Montecito.

With bisect, I located the following patch:

18d95a2832c1392a2d63227a7a6d433cb9f2037e is first bad commit
| commit 18d95a2832c1392a2d63227a7a6d433cb9f2037e
| Author: Peter Zijlstra <a.p.zijlstra@chello.nl>
| Date:   Sat Apr 19 19:45:00 2008 +0200
|
|     sched: fair-group: SMP-nice for group scheduling

Revert it so that we get v2.6.25 behavior.

Bisected-by: Yanmin Zhang <yanmin_zhang@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agosched: cleanup
Ingo Molnar [Fri, 16 May 2008 15:47:14 +0000 (17:47 +0200)]
sched: cleanup

Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoshow_schedstat(): fix memleak
Adrian Bunk [Wed, 14 May 2008 23:22:59 +0000 (16:22 -0700)]
show_schedstat(): fix memleak

The Coverity checker spotted a memleak introduced by commit
39106dcf85285e78f3b290022122c76f851379b8 (cpumask: use new cpus_scnprintf
function).

It seems the kfree() got lost between v2 and v3 of this patch...

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Cc: Mike Travis <travis@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agosched: unite unlikely pairs in rt_policy() and schedule_debug()
Roel Kluin [Tue, 13 May 2008 21:44:11 +0000 (23:44 +0200)]
sched: unite unlikely pairs in rt_policy() and schedule_debug()

Removes obfuscation and may improve assembly.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agorevert ("sched: fair: weight calculations")
Ingo Molnar [Thu, 29 May 2008 09:23:17 +0000 (11:23 +0200)]
revert ("sched: fair: weight calculations")

Yanmin Zhang reported:

Comparing with kernel 2.6.25, sysbench+mysql(oltp, readonly) has many
regressions with 2.6.26-rc1:

 1) 8-core stoakley: 28%;
 2) 16-core tigerton: 20%;
 3) Itanium Montvale: 50%.

Bisect located this patch:

8f1bc385cfbab474db6c27b5af1e439614f3025c is first bad commit
| commit 8f1bc385cfbab474db6c27b5af1e439614f3025c
| Author: Peter Zijlstra <a.p.zijlstra@chello.nl>
| Date:   Sat Apr 19 19:45:00 2008 +0200
|
|     sched: fair: weight calculations

Revert it to the 2.6.25 state.

Bisected-by: Yanmin Zhang <yanmin_zhang@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoMerge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/linville/wireles...
David S. Miller [Thu, 29 May 2008 08:49:04 +0000 (01:49 -0700)]
Merge branch 'master' of /linux/kernel/git/linville/wireless-2.6

16 years agobluetooth: fix locking bug in the rfcomm socket cleanup handling
Arjan van de Ven [Thu, 29 May 2008 08:32:47 +0000 (01:32 -0700)]
bluetooth: fix locking bug in the rfcomm socket cleanup handling

in net/bluetooth/rfcomm/sock.c, rfcomm_sk_state_change() does the
following operation:

        if (parent && sock_flag(sk, SOCK_ZAPPED)) {
                /* We have to drop DLC lock here, otherwise
                 * rfcomm_sock_destruct() will dead lock. */
                rfcomm_dlc_unlock(d);
                rfcomm_sock_kill(sk);
                rfcomm_dlc_lock(d);
        }
}

which is fine, since rfcomm_sock_kill() will call sk_free() which will call
rfcomm_sock_destruct() which takes the rfcomm_dlc_lock()... so far so good.

HOWEVER, this assumes that the rfcomm_sk_state_change() function always gets
called with the rfcomm_dlc_lock() taken. This is the case for all but one
case, and in that case where we don't have the lock, we do a double unlock
followed by an attempt to take the lock, which due to underflow isn't
going anywhere fast.

This patch fixes this by moving the stragling case inside the lock, like
the other usages of the same call are doing in this code.

This was found with the help of the www.kerneloops.org project, where this
deadlock was observed 51 times at this point in time:
http://www.kerneloops.org/search.php?search=rfcomm_sock_destruct

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agomac80211: fix alignment issue with compare_ether_addr()
Senthil Balasubramanian [Wed, 28 May 2008 17:45:32 +0000 (23:15 +0530)]
mac80211: fix alignment issue with compare_ether_addr()

This addresses an alignment issue with compare_ether_addr().
The addresses passed to compare_ether_addr should be two bytes aligned.
It may function properly in x86 platform. However may not work properly
on IA-64 or ARM processor.

This also fixes a typo in mlme.c where the sk_buff struct name is incorect.
Though sizeof() works for any incorrect structure pointer name as its just
a pointer length that we want, lets just fix it.

Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: Fix for NULL pointer dereference in sta_info_get()
Senthil Balasubramanian [Wed, 28 May 2008 14:38:12 +0000 (20:08 +0530)]
mac80211: Fix for NULL pointer dereference in sta_info_get()

This addresses a NULL pointer dereference in sta_info_get().
TID and sta_info are extracted in ADDBA Timer expiry function
through the timer handler's argument.

The problem is extracging the TID (which was stored in
timer_to_tid[] array of type "u8") through "int *" typecast which
may also yield unwanted bytes for the MSB of TID that results
in incorrect sta_info and ieee80211_local pointers.

ieee80211_local pointer is NULL as illustrated below, it crashes in
sta_info_get(). The problem started when extracting ieee80211_local
pointer out of sta_info iteself and eventually crashed in
stat_info_get().

The proper way to fix is to change the data type of TID to u8
instead of u16. However changing all the occurences requires
some prototype changes as well. We should fix this in upcoming
patches.

Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: Luis Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: fix a typo in ieee80211_handle_filtered_frame comment
Yi Zhu [Tue, 27 May 2008 14:50:50 +0000 (17:50 +0300)]
mac80211: fix a typo in ieee80211_handle_filtered_frame comment

fix a typo in ieee80211_handle_filtered_frame comment

Signed-off-by: Yi Zhu <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agorndis_wlan: add missing range check for power_output modparam
Jussi Kivilinna [Tue, 27 May 2008 08:15:08 +0000 (11:15 +0300)]
rndis_wlan: add missing range check for power_output modparam

Range check for power_output were missing.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoiwlwifi: fix rate scale TLC column selection bug
Guy Cohen [Tue, 27 May 2008 03:29:35 +0000 (11:29 +0800)]
iwlwifi: fix rate scale TLC column selection bug

This patch fixes a case that a wrong maximal rate is selected when
searching for better configurations.

Signed-off-by: Guy Cohen <guy.cohen@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoiwlwifi: fix exit from stay_in_table state
Guy Cohen [Tue, 27 May 2008 03:29:34 +0000 (11:29 +0800)]
iwlwifi: fix exit from stay_in_table state

When exiting from stay in table state (e.g. timer expiration),
all the statistics are reset and the RS flow should not continue
but only after enough statistics are collected again.

Signed-off-by: Guy Cohen <guy.cohen@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agorndis_wlan: Make connections to TKIP PSK networks work
Scott Ashcroft [Mon, 26 May 2008 21:06:15 +0000 (00:06 +0300)]
rndis_wlan: Make connections to TKIP PSK networks work

This patch allows the rndis_wlan driver to connect to TKIP PSK
networks.  It uses the ASSOCIATION_INFORMATION RNDIS call to pull back
the IEs and sends them back to userspace using wireless events. Tested
on a few wireless networks I have access to. Based on the similar
code in ndiswrapper.

Signed-off-by: Scott Ashcroft <scott.ashcroft@talk21.com>
[edit: cleanups]
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211 : Fixes the status message for iwconfig
Abhijeet Kolekar [Fri, 23 May 2008 17:15:26 +0000 (10:15 -0700)]
mac80211 : Fixes the status message for iwconfig

iwconfig was showing incorrect status messages when disassociated.
Patch fixes this by always checking for association status in
ioctl calls for getting ap address.

Signed-off-by: Abhijeet Kolekar <abhijeet.kolekar@intel.com>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agort2x00: Use atomic interface iteration in irq context
Ivo van Doorn [Fri, 23 May 2008 16:14:02 +0000 (18:14 +0200)]
rt2x00: Use atomic interface iteration in irq context

rt2x00lib_beacondone() is called from interrupt context,
this means we cannot use the mac80211 interface iterator
that uses the rtnl lock (since that uses a mutex which can sleep).
Instead we should use the atomic mac80211 interface iterator.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agort2x00: Reset antenna RSSI after switch
Ivo van Doorn [Fri, 23 May 2008 16:13:56 +0000 (18:13 +0200)]
rt2x00: Reset antenna RSSI after switch

When the antenna configuration has changed we should reset
the antenna RSSI value. Otherwise the value will be influenced
by the previous configuration quality which in turn will affect
the antenna diversity.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agort2x00: Don't count retries as failure
Ivo van Doorn [Fri, 23 May 2008 16:13:49 +0000 (18:13 +0200)]
rt2x00: Don't count retries as failure

Link quality estimation became quite low for all rt2x00 drivers
because the number of retries it took to send the frame were
counted as failure.
This does not correspond to the legacy driver link quality calculation,
by not counting it we will send somewhat more optimistic values to
mac80211.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agort2x00: Fix memleak in tx() path
Ivo van Doorn [Fri, 23 May 2008 16:13:41 +0000 (18:13 +0200)]
rt2x00: Fix memleak in tx() path

When the tx() handler runs while the device has disapeared,
we did return NETDEV_TX_OK but didn't free the skb.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: reorder channel and freq reporting in wext scan report
Tomas Winkler [Thu, 22 May 2008 22:36:36 +0000 (01:36 +0300)]
mac80211: reorder channel and freq reporting in wext scan report

This patch switch order of channel and freq (SIOCGIWFREQ) reports
in scan results in order to overcome wpa_supplicant inability
to handle channel numbers in 5.2Ghz band.
Wext reporting channel number is ambiguous as channels 7-12 (802.11j)
exist on both bands.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agob43: Fix controller restart crash
Michael Buesch [Thu, 22 May 2008 14:32:16 +0000 (16:32 +0200)]
b43: Fix controller restart crash

This fixes a kernel crash on rmmod, in the case where the controller
was restarted before doing the rmmod.

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: fix ieee80211_rx_bss_put/get imbalance
Tomas Winkler [Wed, 21 May 2008 15:17:05 +0000 (18:17 +0300)]
mac80211: fix ieee80211_rx_bss_put/get imbalance

This patch fixes iee80211_rx_bss_put/get imbalance
introduced by 'mac80211: enable IBSS merging' patch.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agonet/mac80211: always true conditionals
Nicolas Kaiser [Tue, 20 May 2008 16:42:54 +0000 (18:42 +0200)]
net/mac80211: always true conditionals

Correct always true conditionals.

Signed-off-by: Nicolas Kaiser <nikai@nikai.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agob43: Upload both beacon templates on initial load
Michael Buesch [Tue, 20 May 2008 10:16:28 +0000 (12:16 +0200)]
b43: Upload both beacon templates on initial load

This updates the beacon template code to upload both templates,
if we never uploaded one before.

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agortl8180: fix wrong parameter in grf5101_rf_set_channel
Andrea Merello [Sat, 10 May 2008 11:34:16 +0000 (13:34 +0200)]
rtl8180: fix wrong parameter in grf5101_rf_set_channel

The grf5101 RF code needs to invoke grf5101_write_phy_antenna every time the
channel is being switch.

This should be done passing the channel number to that function.
Incorrectly we were passing the same value that is written on the
channel RF register.
This may cause problems when operating on ch 14.

This patch fixes it.

Thanks to Alessandro Di Marco who found this issue!

Signed-off-by: Andrea Merello <andreamrl@tiscali.it>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agortl8180: fix wrong parameter in max2820_rf_set_channel
Andrea Merello [Sat, 10 May 2008 11:32:34 +0000 (13:32 +0200)]
rtl8180: fix wrong parameter in max2820_rf_set_channel

The max2820 RF code needs to invoke max2820_write_phy_antenna every time the
channel is being switch.

This should be done passing the channel number to that function.
Incorrectly we were passing the same value that is written on the
channel RF register.
This may cause problems when operating on ch 14.

This patch fixes it.

Thanks to Alessandro Di Marco who found this issue!

Signed-off-by: Andrea Merello <andreamrl@tiscali.it>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agortl8180: fix wrong parameter in sa2400_rf_set_channel
Andrea Merello [Sat, 10 May 2008 11:30:12 +0000 (13:30 +0200)]
rtl8180: fix wrong parameter in sa2400_rf_set_channel

The sa2400 RF code needs to invoke sa2400_write_phy_antenna every time the
channel is being switch.

This should be done passing the channel number to that function.
Incorrectly we were passing the same value that is written on the
channel RF register.
This may cause problems when operating on ch 14.

This patch fixes it.

Thanks to Alessandro Di Marco who found this issue!

Signed-off-by: Andrea Merello <andreamrl@tiscali.it>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agortl8180: avoid NULL dereference in max2820_rf_set_channel
John W. Linville [Tue, 27 May 2008 21:01:55 +0000 (17:01 -0400)]
rtl8180: avoid NULL dereference in max2820_rf_set_channel

The static function max2820_rf_set_channel is called with conf == NULL
within its compilation unit.  Originally this defaulted to b/g channel
1, but "cfg80211 API for channels/bitrates, mac80211 and driver
conversion" (commit 8318d78a44d49ac1edf2bdec7299de3617c4232e) mistakenly
dropped this check.  This patch minimally restores the expected
behavior.

Reported-by: Colin Lai <colin_sh@163.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6
Linus Torvalds [Wed, 28 May 2008 19:58:12 +0000 (12:58 -0700)]
Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6:
  [IA64] Workaround for RSE issue

16 years agoFix FRV minimum slab/kmalloc alignment
David Howells [Wed, 28 May 2008 15:49:01 +0000 (16:49 +0100)]
Fix FRV minimum slab/kmalloc alignment

> +#define ARCH_KMALLOC_MINALIGN (sizeof(long) * 2)
> +#define ARCH_SLAB_MINALIGN (sizeof(long) * 2)

This doesn't work if SLAB is selected and slab debugging is enabled as
these are passed to the preprocessor, and the preprocessor doesn't
understand sizeof.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Wed, 28 May 2008 15:00:51 +0000 (08:00 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  cfq-iosched: fix RCU problem in cfq_cic_lookup()
  block: make blktrace use per-cpu buffers for message notes
  Added in elevator switch message to blktrace stream
  Added in MESSAGE notes for blktraces
  block: reorder cfq_queue to save space on 64bit builds
  block: Move the second call to get_request to the end of the loop
  splice: handle try_to_release_page() failure
  splice: fix sendfile() issue with relay

16 years agoFRV: Specify the minimum slab/kmalloc alignment
David Howells [Wed, 28 May 2008 14:36:34 +0000 (15:36 +0100)]
FRV: Specify the minimum slab/kmalloc alignment

Specify the minimum slab/kmalloc alignment to be 8 bytes.  This fixes a
crash when SLOB is selected as the memory allocator.  The FRV arch needs
this so that it can use the load- and store-double instructions without
faulting.  By default SLOB sets the minimum to be 4 bytes.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMN10300: Fix typo in header guard
Vegard Nossum [Wed, 28 May 2008 12:55:24 +0000 (13:55 +0100)]
MN10300: Fix typo in header guard

Fix a typo in the header guard of asm/ipc.h.

Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agocfq-iosched: fix RCU problem in cfq_cic_lookup()
Jens Axboe [Wed, 28 May 2008 12:46:59 +0000 (14:46 +0200)]
cfq-iosched: fix RCU problem in cfq_cic_lookup()

cfq_cic_lookup() needs to properly protect ioc->ioc_data before
dereferencing it and also exclude updaters of ioc->ioc_data as well.

Also add a number of comments documenting why the existing RCU usage
is OK.

Thanks a lot to "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> for
review and comments!

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years agoblock: make blktrace use per-cpu buffers for message notes
Jens Axboe [Wed, 28 May 2008 12:45:33 +0000 (14:45 +0200)]
block: make blktrace use per-cpu buffers for message notes

Currently it uses a single static char array, but that risks
being corrupted when multiple users issue message notes at the
same time. Make the buffers dynamically allocated when the trace
is setup and make them per-cpu instead.

The default max message size of 1k is also very large, the
interface is mainly for small text notes. So shrink it to 128 bytes.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years agoAdded in elevator switch message to blktrace stream
Alan D. Brunelle [Tue, 27 May 2008 12:55:00 +0000 (14:55 +0200)]
Added in elevator switch message to blktrace stream

Signed-off-by: Alan D. Brunelle <alan.brunelle@hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years agoAdded in MESSAGE notes for blktraces
Alan D. Brunelle [Tue, 27 May 2008 12:54:41 +0000 (14:54 +0200)]
Added in MESSAGE notes for blktraces

Allows messages to be inserted into blktrace streams.

Signed-off-by: Alan D. Brunelle <alan.brunelle@hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years agoblock: reorder cfq_queue to save space on 64bit builds
Richard Kennedy [Fri, 23 May 2008 04:52:00 +0000 (06:52 +0200)]
block: reorder cfq_queue to save space on 64bit builds

saves 8 bytes of padding & increases objects/slab from 30 to 32 on my
AMD64 config

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years agoblock: Move the second call to get_request to the end of the loop
Zhang, Yanmin [Thu, 22 May 2008 13:13:29 +0000 (15:13 +0200)]
block: Move the second call to get_request to the end of the loop

In function get_request_wait, the second call to get_request could be
moved to the end of the while loop, because if the first call to
get_request fails, the second call will fail without sleep.

Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years agosplice: handle try_to_release_page() failure
Jens Axboe [Tue, 20 May 2008 19:27:41 +0000 (21:27 +0200)]
splice: handle try_to_release_page() failure

splice currently assumes that try_to_release_page() always suceeds,
but it can return failure. If it does, we cannot steal the page.

Acked-by: Mingming Cao <cmm@us.ibm.com
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years agosplice: fix sendfile() issue with relay
Tom Zanussi [Fri, 9 May 2008 11:28:36 +0000 (13:28 +0200)]
splice: fix sendfile() issue with relay

Splice isn't always incrementing the ppos correctly, which broke
relay splice.

Signed-off-by: Tom Zanussi <zanussi@comcast.net>
Tested-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes...
Linus Torvalds [Wed, 28 May 2008 01:47:59 +0000 (18:47 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jbarnes/pci-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
  pciehp: add message about pciehp_slot_with_bus option
  pci hotplug core: add check of duplicate slot name
  pciehp: move msleep after power off
  pciehp: poll cmd completion if hotplug interrupt is disabled
  pciehp: fix slow probing
  pciehp: fix NULL dereference in interrupt handler
  shpchp: add message about shpchp_slot_with_bus option
  PCI: don't enable ASPM on devices with mixed PCIe/PCI functions

16 years agopciehp: add message about pciehp_slot_with_bus option
Kenji Kaneshige [Tue, 27 May 2008 10:07:33 +0000 (19:07 +0900)]
pciehp: add message about pciehp_slot_with_bus option

Some (broken?) platform assign the same slot name to multiple hotplug
slots. On such system, slot initialization would fail because of name
collision. The pciehp driver already have a "slot_with_bus" module
option which adds the bus number into the slot name. This patch adds
the message about this module option that will be displayed when slot
name collision is detected.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
16 years agopci hotplug core: add check of duplicate slot name
Kenji Kaneshige [Tue, 27 May 2008 10:07:01 +0000 (19:07 +0900)]
pci hotplug core: add check of duplicate slot name

Fix the following errors reported by Jan C. Nordholz in
http://bugzilla.kernel.org/show_bug.cgi?id=10751.

kobject_add_internal failed for 2 with -EEXIST, don't try to register things with the same name in the same directory.
Pid: 1, comm: swapper Tainted: G        W 2.6.26-rc3 #1
 [<c0266980>] kobject_add_internal+0x140/0x190
 [<c0266afd>] kobject_init_and_add+0x2d/0x40
 [<c027bc91>] pci_hp_register+0x81/0x2f0
 [<c027fd07>] pciehp_probe+0x1a7/0x470
 [<c01b3b84>] sysfs_add_one+0x44/0xa0
 [<c01b3c1f>] sysfs_addrm_start+0x3f/0xb0
 [<c01b497a>] sysfs_create_link+0x8a/0xf0
 [<c0279570>] pcie_port_probe_service+0x50/0x80
 [<c02e0545>] driver_sysfs_add+0x55/0x70
 [<c02e0662>] driver_probe_device+0x82/0x180
 [<c02e07cc>] __driver_attach+0x6c/0x70
 [<c02dfe0a>] bus_for_each_dev+0x3a/0x60
 [<c05db2d0>] pcied_init+0x0/0x80
 [<c02e04e6>] driver_attach+0x16/0x20
 [<c02e0760>] __driver_attach+0x0/0x70
 [<c02e0341>] bus_add_driver+0x1a1/0x220
 [<c05db2d0>] pcied_init+0x0/0x80
 [<c02e09cd>] driver_register+0x4d/0x120
 [<c05db050>] ibm_acpiphp_init+0x0/0x190
 [<c0125aab>] printk+0x1b/0x20
 [<c05db2d0>] pcied_init+0x0/0x80
 [<c05db2de>] pcied_init+0xe/0x80
 [<c05c751a>] kernel_init+0x10a/0x300
 [<c0120138>] schedule_tail+0x18/0x50
 [<c0103b9a>] ret_from_fork+0x6/0x1c
 [<c05c7410>] kernel_init+0x0/0x300
 [<c05c7410>] kernel_init+0x0/0x300
 [<c010485b>] kernel_thread_helper+0x7/0x1c
 =======================
pci_hotplug: Unable to register kobject '2'<3>pciehp: pci_hp_register failed with error -22

Slot with the same name can be registered multiple times if shpchp or
pciehp driver is loaded after acpiphp is loaded because ACPI based
hotplug driver and Native OS hotplug driver trying to handle the same
physical slot. In this case, current pci_hotplug core will call
kobject_init_and_add() muliple time with the same name. This is the
cause of this problem. To fix this problem, this patch adds the check
into pci_hp_register() to see if the slot with the same name.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
16 years agopciehp: move msleep after power off
Kenji Kaneshige [Tue, 27 May 2008 10:06:22 +0000 (19:06 +0900)]
pciehp: move msleep after power off

According to the PCI Express specification, we must wait for at least
1 second after turning power off before taking any action that relies
on power having been removed from the slot/adapter. For this, current
pciehp wait for 1 second after issuing the power off command in
hpc_power_off_slot() function. But waiting for 1 second in
hpc_power_off_slot() can make pciehp probing slow-down because pciehp
probe code calls hpc_power_off_slot() if the slot is not occupied just
in case. We don't need to wait for 1 second at the pciehp probe time
because there is no action on that empty slot. So move 1 second wait
from hpc_power_off_slot() to the caller of hpc_power_off_slot().

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
16 years agopciehp: poll cmd completion if hotplug interrupt is disabled
Kenji Kaneshige [Tue, 27 May 2008 10:05:26 +0000 (19:05 +0900)]
pciehp: poll cmd completion if hotplug interrupt is disabled

Fix improper long wait for command completion in pciehp probing.

As described in PCI Express specification, software notification is
not generated if the command that occurs as a result of a write to the
Slot Control register that disables software notification of command
completed events. Since pciehp driver doesn't take it into account,
such command is issued in pciehp probing, and it causes improper long
wait for command completion.

This patch changes the pciehp driver to take such command into
account.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
16 years agopciehp: fix slow probing
Kenji Kaneshige [Tue, 27 May 2008 10:04:30 +0000 (19:04 +0900)]
pciehp: fix slow probing

Fix the "pciehp probing slow" problem reported from Jan C. Nordholz in
http://bugzilla.kernel.org/show_bug.cgi?id=10751.

The command completed bit in Slot Status register applies only to
commands issued to control the attention indicator, power indicator,
power controller, or electromechanical interlock. However, writes to
other parts of the Slot Control register would end up writing to the
control fields. Hence, any write to Slot Control register is
considered as a command. However, if the controller doesn't support
any of attention indicator, power indicator, power controller and
electromechanical interlock, command completed bit would not set in
writing to Slot Control register. In this case, we should not wait for
command completed bit set, otherwise all commands would be considered
not completed in timeout seconds (1 sec.).

The cause of the problem is pciehp driver didn't take this situation
into account. This patch changes pciehp to take it into account. This
patch also add the check for "No Command Completed Support" bit in
Slot Capability register. If it is set, we should not wait for command
completed bit set as well.

This problem seems to be revealed by the commit
c27fb883dffe11aa4cb35ecea1fa1832ba45d4da that fixed the bug that
pciehp did not wait for command completed properly (pciehp just
ignored the command completion event).

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
16 years agopciehp: fix NULL dereference in interrupt handler
Kenji Kaneshige [Tue, 27 May 2008 10:03:16 +0000 (19:03 +0900)]
pciehp: fix NULL dereference in interrupt handler

Fix the following NULL dereference problem reported from Pierre Ossman
and Ingo Molnar.

pciehp: HPC vendor_id 8086 device_id 27d0 ss_vid 0 ss_did 0
pciehp: pciehp_find_slot: slot (device=0x0) not found
BUG: unable to handle kernel NULL pointer dereference at 0000000000000070
IP: [<ffffffff80494a8b>] pciehp_handle_presence_change+0x7e/0x113
PGD 0
Oops: 0000 [1]
CPU 0
Modules linked in:
Pid: 1, comm: swapper Tainted: G        W 2.6.26-rc3-sched-devel.git-00001-g2b99b26-dirty #170
RIP: 0010:[<ffffffff80494a8b>]  [<ffffffff80494a8b>] pciehp_handle_presence_change+0x7e/0x113
RSP: 0000:ffff81003f83fbb0  EFLAGS: 00010046
RAX: 0000000000000039 RBX: 0000000000000000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000001 RDI: 0000000000000046
RBP: ffff81003f83fbd0 R08: 0000000000000001 R09: ffffffff80245103
R10: 0000000000000020 R11: 0000000000000000 R12: ffff81003ea53a30
R13: 0000000000000000 R14: 0000000000000011 R15: ffffffff80495926
FS:  0000000000000000(0000) GS:ffffffff80be7400(0000) knlGS:0000000000000000
CS:  0010 DS: 0018 ES: 0018 CR0: 000000008005003b
CR2: 0000000000000070 CR3: 0000000000201000 CR4: 00000000000006a0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process swapper (pid: 1, threadinfo ffff81003f83e000, task ffff81003f840000)
Stack:  0000000000000008 ffff81003f83fbf6 ffff81003ea53a30 0000000000000008
 ffff81003f83fc10 ffffffff80495ab4 0000000000000011 0000000000000002
 0000000000000202 0000000000000202 00000000fffffff4 ffff81003ea53a30
Call Trace:
 [<ffffffff80495ab4>] pcie_isr+0x18e/0x1bc
 [<ffffffff80260831>] request_irq+0x106/0x12f
 [<ffffffff80495fb6>] pcie_init+0x15e/0x6cc
 [<ffffffff804933a3>] pciehp_probe+0x64/0x541
 [<ffffffff8048f4e7>] pcie_port_probe_service+0x4c/0x76
 [<ffffffff8054af70>] driver_probe_device+0xd4/0x1f0
 [<ffffffff8054b108>] __driver_attach+0x7c/0x7e
 [<ffffffff8054b08c>] ? __driver_attach+0x0/0x7e
 [<ffffffff8054a4b6>] bus_for_each_dev+0x53/0x7d
 [<ffffffff8054ad3c>] driver_attach+0x1c/0x1e
 [<ffffffff8054a9c2>] bus_add_driver+0xdd/0x25b
 [<ffffffff80c09d3d>] ? pcied_init+0x0/0x8b
 [<ffffffff8054b288>] driver_register+0x5f/0x13e
 [<ffffffff80c09d3d>] ? pcied_init+0x0/0x8b
 [<ffffffff8048f441>] pcie_port_service_register+0x47/0x49
 [<ffffffff80c09d52>] pcied_init+0x15/0x8b
 [<ffffffff80bf3938>] kernel_init+0x75/0x243
 [<ffffffff808639d2>] ? _spin_unlock_irq+0x2b/0x3a
 [<ffffffff80228d1f>] ? finish_task_switch+0x57/0x9a
 [<ffffffff8020c258>] child_rip+0xa/0x12
 [<ffffffff8020bcec>] ? restore_args+0x0/0x30
 [<ffffffff80bf38c3>] ? kernel_init+0x0/0x243
 [<ffffffff8020c24e>] ? child_rip+0x0/0x12

Code: 83 80 00 00 00 48 39 f0 75 e1 0f b6 c9 48 c7 c2 00 0e 8d 80 48 c7 c6 8a 60 a6 80 48 c7 c7 10 db a8 80 31 c0 e8 3f 8d d9 ff 31 db <48> 8b 43 70 48 8d 75 ef 48 89 df ff 50 30 80 7d ef 00 74 37 48
RIP  [<ffffffff80494a8b>] pciehp_handle_presence_change+0x7e/0x113
 RSP <ffff81003f83fbb0>
CR2: 0000000000000070
Kernel panic - not syncing: Fatal exception

The situation under which it occurs is hw and timing related: it appears
to happen on a system that has PCI hotplug hardware but with no active
hotplug cards, and another interrupt in the same (shared) IRQ line
arrives too early, before the hotplug-slot entry has been set up - as
triggered by CONFIG_DEBUG_SHIRQ=y:

This patch contains the following two fixes.

(1) Clear all events bits in Slot Status register to prevent the pciehp
    driver from detecting the spurious events that would have been occur
    before pciehp loading.

(2) Add check whether slot initialization had been already done.

This is short term fix. We need more structural fixes to install
interrupt handler after slot initialization is done.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
16 years agoshpchp: add message about shpchp_slot_with_bus option
Kenji Kaneshige [Tue, 27 May 2008 10:08:23 +0000 (19:08 +0900)]
shpchp: add message about shpchp_slot_with_bus option

Some (broken?) platform assign the same slot name to multiple hotplug
slots. On such system, slot initialization would fail because of name
collision. The shpchp driver already have a "slot_with_bus" module
option which adds the bus number into the slot name. This patch adds
the message about this module option that will be displayed when slot
name collision is detected.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
16 years ago[IA64] Workaround for RSE issue
Tony Luck [Tue, 27 May 2008 20:23:16 +0000 (13:23 -0700)]
[IA64] Workaround for RSE issue

Problem: An application violating the architectural rules regarding
operation dependencies and having specific Register Stack Engine (RSE)
state at the time of the violation, may result in an illegal operation
fault and invalid RSE state.  Such faults may initiate a cascade of
repeated illegal operation faults within OS interruption handlers.
The specific behavior is OS dependent.

Implication: An application causing an illegal operation fault with
specific RSE state may result in a series of illegal operation faults
and an eventual OS stack overflow condition.

Workaround: OS interruption handlers that switch to kernel backing
store implement a check for invalid RSE state to avoid the series
of illegal operation faults.

The core of the workaround is the RSE_WORKAROUND code sequence
inserted into each invocation of the SAVE_MIN_WITH_COVER and
SAVE_MIN_WITH_COVER_R19 macros.  This sequence includes hard-coded
constants that depend on the number of stacked physical registers
being 96.  The rest of this patch consists of code to disable this
workaround should this not be the case (with the presumption that
if a future Itanium processor increases the number of registers, it
would also remove the need for this patch).

Move the start of the RBS up to a mod32 boundary to avoid some
corner cases.

The dispatch_illegal_op_fault code outgrew the spot it was
squatting in when built with this patch and CONFIG_VIRT_CPU_ACCOUNTING=y
Move it out to the end of the ivt.

Signed-off-by: Tony Luck <tony.luck@intel.com>