12 years agoMerge branch 'master' of git://
David S. Miller [Tue, 2 Feb 2010 17:04:58 +0000 (09:04 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/kaber/nf-2.6

12 years agobe2net: use eq-id to calculate cev-isr reg offset
Sathya Perla [Tue, 2 Feb 2010 15:48:40 +0000 (07:48 -0800)]
be2net: use eq-id to calculate cev-isr reg offset

cev-isr reg offset for each function is better calculated using (any) eq-id
alloted to that function instead of using pci-func number(which
does not work in some configurations...)

Signed-off-by: Sathya Perla <>
Signed-off-by: David S. Miller <>
12 years agoMerge branch 'master' of git://
David S. Miller [Mon, 1 Feb 2010 07:34:58 +0000 (23:34 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/holtmann/bluetooth-2.6

12 years agoBluetooth: Use the control channel for raw HID reports
Bastien Nocera [Wed, 20 Jan 2010 12:00:42 +0000 (12:00 +0000)]
Bluetooth: Use the control channel for raw HID reports

In commit 2da31939a42f7a676a0bc5155d6a0a39ed8451f2, support
for Bluetooth hid_output_raw_report was added, but it pushes
the data to the interrupt channel instead of the contol one.

This patch makes hid_output_raw_report use the control channel
instead. Using the interrupt channel was a mistake.

Signed-off-by: Bastien Nocera <>
Signed-off-by: Marcel Holtmann <>
12 years agoBluetooth: Add DFU driver for Atheros Bluetooth chipset AR3011
Vikram Kandukuri [Wed, 6 Jan 2010 13:34:15 +0000 (19:04 +0530)]
Bluetooth: Add DFU driver for Atheros Bluetooth chipset AR3011

Signed-off-by: Vikram Kandukuri <>
Signed-off-by: Alicke Xu <>
Reviewed-by: Luis R. Rodriguez <>
Signed-off-by: Marcel Holtmann <>
12 years agoBluetooth: Redo checks in IRQ handler for shared IRQ support
Mike Frysinger [Mon, 14 Sep 2009 17:43:49 +0000 (13:43 -0400)]
Bluetooth: Redo checks in IRQ handler for shared IRQ support

Commit ac019360fe3 changed the irq handler logic to BUG_ON rather than
returning IRQ_NONE when the incoming argument is invalid.  While this
works in most cases, it doesn't work when the IRQ is shared with other
devices (or when DEBUG_SHIRQ is enabled).

So revert the previous change and replace the warning message with a
comment explaining that we want this behavior.

Signed-off-by: Michael Hennerich <>
Signed-off-by: Mike Frysinger <>
Signed-off-by: Marcel Holtmann <>
12 years agoBluetooth: Fix memory leak in L2CAP
Andrei Emeltchenko [Wed, 23 Dec 2009 11:07:14 +0000 (13:07 +0200)]
Bluetooth: Fix memory leak in L2CAP

Move skb_clone after error confition check so it is not going
potentially out of the scope.

Signed-off-by: Andrei Emeltchenko <>
Acked-by: Gustavo F. Padovan <>
Signed-off-by: Marcel Holtmann <>
12 years agoBluetooth: Remove double free of SKB pointer in L2CAP
Andrei Emeltchenko [Tue, 22 Dec 2009 13:58:08 +0000 (15:58 +0200)]
Bluetooth: Remove double free of SKB pointer in L2CAP

Trivial fix for double free of SKB pointer with kfree_skb to
make code simplier and cleaner. Remove unused variable err.

Signed-off-by: Andrei Emeltchenko <>
Acked-by: Gustavo F. Padovan <>
Signed-off-by: Marcel Holtmann <>
12 years agocdc_ether: Partially revert "usbnet: Set link down initially ..."
Ben Hutchings [Fri, 29 Jan 2010 05:37:18 +0000 (21:37 -0800)]
cdc_ether: Partially revert "usbnet: Set link down initially ..."

Commit 37e8273cd30592d3a82bcb70cbb1bdc4eaeb6b71 ("usbnet: Set link down
initially for drivers that update link state") changed the initial link
state in cdc_ether and other drivers based on the understanding that the
devices they support generate link change interrupts.  However, this is
optional in the CDC Ethernet protocol, and two users have reported in
<> that the link state
for their devices remains down.  Therefore, revert the change in

Signed-off-by: Ben Hutchings <>
Tested-by: Avi Rozen <>
Signed-off-by: David S. Miller <>
12 years agobe2net: Fix memset() arg ordering.
David S. Miller [Fri, 29 Jan 2010 05:36:21 +0000 (21:36 -0800)]
be2net: Fix memset() arg ordering.

Noticed by Ben Hutchings.

Signed-off-by: David S. Miller <>
12 years agobonding: bond_open error return value
stephen hemminger [Mon, 25 Jan 2010 23:34:15 +0000 (23:34 +0000)]
bonding: bond_open error return value

The convention for API functions in kernel is to return errno value;
bond_open would return -1 if alb setup failed. The only reason that
could happen is if kmalloc() failed.

Signed-off-by: Stephen Hemminger <>
Signed-off-by: David S. Miller <>
12 years agoixgbe: if ixgbe_copy_dcb_cfg is going to fail learn about it early
John Fastabend [Wed, 27 Jan 2010 16:38:06 +0000 (16:38 +0000)]
ixgbe: if ixgbe_copy_dcb_cfg is going to fail learn about it early

Call ixgbe_copy_dcb_cfg() earlier in the ixgbe_dcbnl_set_all() so that
we can learn if this is going to fail as early as possible.  Previously,
ixgbe_down or ixgbe_close were being called before this check and the
IXGBE_RESETTING bit was being set and cleared.  Worse if this failed
the corresponding ixgbe_up/ndo_open would not called.

Signed-off-by: John Fastabend <>
Acked-by: PJ Waskiewicz <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agoixgbe: set the correct DCB bit for pg tx settings
John Fastabend [Wed, 27 Jan 2010 16:37:44 +0000 (16:37 +0000)]
ixgbe: set the correct DCB bit for pg tx settings

Set the correct bit BIT_PG_TX when tx PG settings are set.

Signed-off-by: John Fastabend <>
Acked-by: PJ Waskiewicz <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agoigbvf: fix issue w/ mapped_as_page being left set after unmap
Alexander Duyck [Wed, 27 Jan 2010 15:30:39 +0000 (15:30 +0000)]
igbvf: fix issue w/ mapped_as_page being left set after unmap

This change fixes an issue in igbvf with mapped_as_page being left set
after a page is unmapped which results in buffers which are mapped via map
single being unmapped as page.

Signed-off-by: Alexander Duyck <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agodrivers/net: ks8851_mll ethernet network driver
Choi, David [Wed, 27 Jan 2010 06:03:16 +0000 (06:03 +0000)]
drivers/net: ks8851_mll ethernet network driver

Hello David Miller,

I fix a bug in ks8851_mll driver, which has existed since 2.6.32-rc6.

>From : David J. Choi <>

Fix a bug that the data pointers in the interrupt handler are set wrong, which is related with the 5th parameter of request_irq().

Signed-off-by : David J. Choi <>

Signed-off-by: David S. Miller <>
12 years agobe2net: Bug fix to support newer generation of BE ASIC
Ajit Khaparde [Wed, 27 Jan 2010 21:56:44 +0000 (21:56 +0000)]
be2net: Bug fix to support newer generation of BE ASIC

Bug fix in be2net for newer generation of BladeEngine ASIC.

Signed-off-by: Ajit Khaparde <>
Signed-off-by: David S. Miller <>
12 years agoMerge branch 'master' of git://
David S. Miller [Thu, 28 Jan 2010 13:42:33 +0000 (05:42 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6

12 years agostarfire: clean up properly if firmware loading fails
Ben Hutchings [Tue, 26 Jan 2010 18:27:09 +0000 (18:27 +0000)]
starfire: clean up properly if firmware loading fails

netdev_open() will return without cleaning up net device or hardware state
if firmware loading fails.  This results in a BUG() on a second attempt to
bring the interface up, reported in
<>, and probably has even
worse effects if the driver is removed afterwards.

Call netdev_close() to clean up on failure.


Signed-off-by: Ben Hutchings <>
Reported-by: Michael Moffatt <>
Tested-by: Michael Moffatt <>
Cc: "David S. Miller" <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: David S. Miller <>
12 years agomac80211: fix NULL pointer dereference when ftrace is enabled
Zhu Yi [Tue, 26 Jan 2010 07:58:57 +0000 (15:58 +0800)]
mac80211: fix NULL pointer dereference when ftrace is enabled

I got below kernel oops when I try to bring down the network interface if
ftrace is enabled. The root cause is drv_ampdu_action() is passed with a
NULL ssn pointer in the BA session tear down case. We need to check and
avoid dereferencing it in trace entry assignment.

BUG: unable to handle kernel NULL pointer dereference
Modules linked in: at (null)
IP: [<f98fe02a>] ftrace_raw_event_drv_ampdu_action+0x10a/0x160 [mac80211]
*pde = 00000000
Call Trace:
 [<f98fdf20>] ? ftrace_raw_event_drv_ampdu_action+0x0/0x160 [mac80211]
 [<f98dac4c>] ? __ieee80211_stop_rx_ba_session+0xfc/0x220 [mac80211]
 [<f98d97fb>] ? ieee80211_sta_tear_down_BA_sessions+0x3b/0x50 [mac80211]
 [<f98dc6f6>] ? ieee80211_set_disassoc+0xe6/0x230 [mac80211]
 [<f98dc6ac>] ? ieee80211_set_disassoc+0x9c/0x230 [mac80211]
 [<f98dcbb8>] ? ieee80211_mgd_deauth+0x158/0x170 [mac80211]
 [<f98e4bdb>] ? ieee80211_deauth+0x1b/0x20 [mac80211]
 [<f8987f49>] ? __cfg80211_mlme_deauth+0xe9/0x120 [cfg80211]
 [<f898b870>] ? __cfg80211_disconnect+0x170/0x1d0 [cfg80211]

Cc: Johannes Berg <>
Signed-off-by: Zhu Yi <>
Signed-off-by: John W. Linville <>
12 years agonetfilter: ctnetlink: fix expectation mask dump
Patrick McHardy [Tue, 26 Jan 2010 16:04:02 +0000 (17:04 +0100)]
netfilter: ctnetlink: fix expectation mask dump

The protocol number is not initialized, so userspace can't interpret
the layer 4 data properly.

Signed-off-by: Patrick McHardy <>
12 years agoipv6: conntrack: Add member of user to nf_ct_frag6_queue structure
Shan Wei [Tue, 26 Jan 2010 02:40:38 +0000 (02:40 +0000)]
ipv6: conntrack: Add member of user to nf_ct_frag6_queue structure

The commit 0b5ccb2(title:ipv6: reassembly: use seperate reassembly queues for
conntrack and local delivery) has broken the saddr&&daddr member of
nf_ct_frag6_queue when creating new queue.  And then hash value
generated by nf_hashfn() was not equal with that generated by fq_find().
So, a new received fragment can't be inserted to right queue.

The patch fixes the bug with adding member of user to nf_ct_frag6_queue structure.

Signed-off-by: Shan Wei <>
Acked-by: Patrick McHardy <>
Signed-off-by: David S. Miller <>
12 years agovirtio_net: Make delayed refill more reliable
Herbert Xu [Mon, 25 Jan 2010 23:51:01 +0000 (15:51 -0800)]
virtio_net: Make delayed refill more reliable

I have seen RX stalls on a machine that experienced a suspected
OOM.  After the stall, the RX buffer is empty on the guest side
and there are exactly 16 entries available on the host side.  As
the number of entries is less than that required by a maximal
skb, the host cannot proceed.

The guest did not have a refill job scheduled.

My diagnosis is that an OOM had occured, with the delayed refill
job scheduled.  The job was able to allocate at least one skb, but
not enough to overcome the minimum required by the host to proceed.

As the refill job would only reschedule itself if it failed completely
to allocate any skbs, this would lead to an RX stall.

The following patch removes this stall possibility by always
rescheduling the refill job until the ring is totally refilled.

Testing has shown that the RX stall no longer occurs whereas
previously it would occur within a day.

Signed-off-by: Herbert Xu <>
Acked-by: Rusty Russell <>
Signed-off-by: David S. Miller <>
12 years agosfc: Use fixed-size buffers for MCDI NVRAM requests
Ben Hutchings [Mon, 25 Jan 2010 23:49:59 +0000 (15:49 -0800)]
sfc: Use fixed-size buffers for MCDI NVRAM requests

The low-level MCDI code always uses 32-bit MMIO operations, and
callers must pad input and output buffers to multiples of 4 bytes.
The MCDI NVRAM functions are not doing this.  Also, their buffers are
declared as variable-length arrays with no explicit maximum length.

Switch to a fixed buffer size based on the chunk size used by the
MTD driver (which is a multiple of 4).

Signed-off-by: Ben Hutchings <>
Signed-off-by: David S. Miller <>
12 years agosfc: Add workspace for GMAC bug workaround to MCDI MAC_STATS buffer
Guido Barzini [Mon, 25 Jan 2010 23:49:19 +0000 (15:49 -0800)]
sfc: Add workspace for GMAC bug workaround to MCDI MAC_STATS buffer

Due to a hardware bug in the SFC9000 family, the firmware must
transfer raw GMAC statistics to host memory before aggregating them
into the cooked (speed-independent) MAC statistics.  Extend the stats
buffer to support this.

The length of the buffer is explicit in the MAC_STATS command, so this
change is backward-compatible on both sides.

Signed-off-by: Ben Hutchings <>
Signed-off-by: David S. Miller <>
12 years agotcp_probe: avoid modulus operation and wrap fix
Stephen Hemminger [Mon, 25 Jan 2010 23:47:50 +0000 (15:47 -0800)]
tcp_probe: avoid modulus operation and wrap fix

By rounding up the buffer size to power of 2, several expensive
modulus operations can be avoided.  This patch also solves a bug where
the gap need when ring gets full was not being accounted for.

Signed-off-by: Stephen Hemminger <>
Signed-off-by: David S. Miller <>
12 years agoqlge: Only free resources if they were allocated
Breno Leitao [Mon, 25 Jan 2010 23:46:58 +0000 (15:46 -0800)]
qlge: Only free resources if they were allocated

Currently qlge tries to release regions even if they were not allocated.
This causes messages like the following in the kernel log

Trying to free nonexistent resource <00000000006af400-00000000006af4ff>
Trying to free nonexistent resource <00003c04ff9f4000-00003c04ff9f7fff>
Trying to free nonexistent resource <00003c04ffc00000-00003c04ffcfffff>

This patch fixes the goto logic in order to not release the resources
if they were not allocated.

Signed-off-by: Breno Leitao <>
Signed-off-by: David S. Miller <>
12 years agoath9k: fix eeprom INI values override for 2GHz-only cards
Felix Fietkau [Sat, 23 Jan 2010 19:04:18 +0000 (20:04 +0100)]
ath9k: fix eeprom INI values override for 2GHz-only cards

Among other changes, this commit:

   commit 06d0f0663e11cab4ec5f2c143a118d71a12fbbe9
   Author: Sujith <>
   Date:   Thu Feb 12 10:06:45 2009 +0530

   ath9k: Enable Fractional N mode

changed the hw attach code to fix up initialization values only for
dual band devices, however the commit message did not give a reason as
to why this would be useful or necessary.

According to tests by Jorge Boncompte, this breaks at least some
2GHz-only cards, so the code should be changed back to the
unconditional INI fixup.

Signed-off-by: Felix Fietkau <>
Reported-by: Jorge Boncompte <>
Tested-by: Pavel Roskin <>
Signed-off-by: John W. Linville <>
12 years agoiwlwifi: fix pointer signedness warning
Johannes Berg [Fri, 22 Jan 2010 22:22:34 +0000 (14:22 -0800)]
iwlwifi: fix pointer signedness warning

There are a few station addresses that are
char *, instead of the normal u8 *; gcc
gives pointer signedness warnings for some
of those, so use u8 * consistently.

Signed-off-by: Johannes Berg <>
Signed-off-by: Reinette Chatre <>
Signed-off-by: John W. Linville <>
12 years agonetns xfrm: deal with dst entries in netns
Alexey Dobriyan [Mon, 25 Jan 2010 06:47:53 +0000 (22:47 -0800)]
netns xfrm: deal with dst entries in netns

GC is non-existent in netns, so after you hit GC threshold, no new
dst entries will be created until someone triggers cleanup in init_net.

Make xfrm4_dst_ops and xfrm6_dst_ops per-netns.
This is not done in a generic way, because it woule waste
(AF_MAX - 2) * sizeof(struct dst_ops) bytes per-netns.

Reorder GC threshold initialization so it'd be done before registering
XFRM policies.

Signed-off-by: Alexey Dobriyan <>
Signed-off-by: David S. Miller <>
12 years agosky2: revert config space change
stephen hemminger [Sun, 24 Jan 2010 18:46:06 +0000 (18:46 +0000)]
sky2: revert config space change

Obviously, this register had some other impact that is causing
the regression.  Either it is masking some other access or needs
to be reset in some path.

Either, way it is best to just revert the change for 2.6.33

This reverts commit 166a0fd4c788ec7f10ca8194ec6d526afa12db75.

Signed-off-by: David S. Miller <>
12 years agovlan: fix vlan_skb_recv()
Eric Dumazet [Mon, 25 Jan 2010 03:52:24 +0000 (19:52 -0800)]
vlan: fix vlan_skb_recv()

Bruno Prémont found commit 9793241fe92f7d930
(vlan: Precise RX stats accounting) added a regression for non
hw accelerated vlans.

[   26.390576] BUG: unable to handle kernel NULL pointer dereference at (null)
[   26.396369] IP: [<df856b89>] vlan_skb_recv+0x89/0x280 [8021q]

vlan_dev_info() was used with original device, instead of
skb->dev. Also spotted by Américo Wang.

Reported-By: Bruno Prémont <>
Tested-By: Bruno Prémont <>
Signed-off-by: Eric Dumazet <>
Signed-off-by: David S. Miller <>
12 years agonetns xfrm: fix "ip xfrm state|policy count" misreport
Alexey Dobriyan [Sat, 23 Jan 2010 13:37:10 +0000 (13:37 +0000)]
netns xfrm: fix "ip xfrm state|policy count" misreport

"ip xfrm state|policy count" report SA/SP count from init_net,
not from netns of caller process.

Signed-off-by: Alexey Dobriyan <>
Signed-off-by: Herbert Xu <>
Signed-off-by: David S. Miller <>
12 years agosky2: Enable/disable WOL per hardware device
Mike McCormack [Sat, 23 Jan 2010 10:09:26 +0000 (02:09 -0800)]
sky2: Enable/disable WOL per hardware device

Y2_HW_WOL_ON/Y2_HW_WOL_OFF should be set and cleared per chip,
not per port.  On dual port cards, Y2_HW_WOL_ON should be
enabled if either sky2 port has WOL enabled.

Found while reviewing code for a WOL regression, though this is
probably not the cause of the regression.

Signed-off-by: Mike McCormack <>
Signed-off-by: David S. Miller <>
12 years agonet: Fix IPv6 GSO type checks in Intel ethernet drivers
Sridhar Samudrala [Sat, 23 Jan 2010 10:02:21 +0000 (02:02 -0800)]
net: Fix IPv6 GSO type checks in Intel ethernet drivers

Found this problem when testing IPv6 from a KVM guest to a remote
host via e1000e device on the host.
The following patch fixes the check for IPv6 GSO packet in Intel
ethernet drivers to use skb_is_gso_v6(). SKB_GSO_DODGY is also set
when packets are forwarded from a guest.

Signed-off-by: Sridhar Samudrala <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agoigb/igbvf: cleanup exception handling in tx_map_adv
Alexander Duyck [Sat, 23 Jan 2010 09:35:00 +0000 (01:35 -0800)]
igb/igbvf: cleanup exception handling in tx_map_adv

After removing the skb_dma_map/unmap calls the exception handling in
igb_tx_map_adv is not correct.  The issue is that the count value was not
being correctly handled so as a result we were not rewinding the ring as
back as we should have been.

Signed-off-by: Alexander Duyck <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agoMAINTAINERS: Add Intel igbvf maintainer
Jeff Kirsher [Sat, 23 Jan 2010 09:20:22 +0000 (01:20 -0800)]
MAINTAINERS: Add Intel igbvf maintainer

Add igbvf to the list of supported Intel drivers and Alex to the list of

Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agoe1000/e1000e: don't use small hardware rx buffers
Jesse Brandeburg [Fri, 22 Jan 2010 22:56:16 +0000 (22:56 +0000)]
e1000/e1000e: don't use small hardware rx buffers

When testing the "e1000: enhance frame fragment detection" (and e1000e)
patches we found some bugs with reducing the MTU size.  The 1024 byte
descriptor used with the 1000 mtu test also (re) introduced the
(originally) reported bug, and causes us to need the e1000_clean_tx_irq
"enhance frame fragment detection" fix.

So what has occured here is that 2.6.32 is only vulnerable for mtu <
1500 due to the jumbo specific routines in both e1000 and e1000e.
So, 2.6.32 needs the 2kB buffer len fix for those smaller MTUs, but
is not vulnerable to the original issue reported.  It has been pointed
out that this vulnerability needs to be patched in older kernels that
don't have the e1000 jumbo routine.  Without the jumbo routines, we
need the "enhance frame fragment detection" fix the e1000, old
e1000e is only vulnerable for < 1500 mtu, and needs a similar
fix.  We split the patches up to provide easy backport paths.

There is only a slight bit of extra code when this fix and the
original "enhance frame fragment detection" fixes are applied, so
please apply both, even though it is a bit of overkill.

Signed-off-by: Jesse Brandeburg <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agofmvj18x_cs: add new id (Panasonic lan & modem card)
Ken Kawasaki [Fri, 22 Jan 2010 11:56:25 +0000 (11:56 +0000)]
fmvj18x_cs: add new id (Panasonic lan & modem card)

fmvj18x_cs, serial_cs:
 add new id
 Panasonic lan & modem card (model name:AL-VML101)

Signed-off-by: Ken Kawasaki <>
Signed-off-by: David S. Miller <>
12 years agobe2net: swap only first 2 fields of mcc_wrb
Sathya Perla [Thu, 21 Jan 2010 22:51:36 +0000 (22:51 +0000)]
be2net: swap only first 2 fields of mcc_wrb

Only the first two fields of mcc wrb - embedded, payload_len
need to be cpu_to_le32() swapped while issuing a cmd to the hw.
The fields tag0, tag1 are opaque and returned back to cpu as is...

Signed-off-by: Sathya Perla <>
Signed-off-by: David S. Miller <>
12 years agoPlease add support for Microsoft MN-120 PCMCIA network card
Ron Murray [Tue, 19 Jan 2010 08:02:48 +0000 (08:02 +0000)]
Please add support for Microsoft MN-120 PCMCIA network card

Please add support for Microsoft MN-120 PCMCIA network card. It's an
old card, I know, but adding support is very easy. You just need to
get tulip_core.c to recognise its vendor/device ID.

Patch for kernel (and many previous) attached.

 .....Ron Murray

Signed-off-by: Ron Murray <>
Signed-off-by: David S. Miller <>
12 years agoMerge branch 'linux-2.6.33.y' of git://
David S. Miller [Sat, 23 Jan 2010 09:08:39 +0000 (01:08 -0800)]
Merge branch 'linux-2.6.33.y' of git://git./linux/kernel/git/inaky/wimax

12 years agoMerge branch 'master' of /home/davem/src/GIT/linux-2.6/
David S. Miller [Sat, 23 Jan 2010 06:45:46 +0000 (22:45 -0800)]
Merge branch 'master' of /home/davem/src/GIT/linux-2.6/

12 years agobe2net: fix bug in rx page posting
Sathya Perla [Fri, 22 Jan 2010 06:52:08 +0000 (22:52 -0800)]
be2net: fix bug in rx page posting

Pages are posted to the rxq in such a way that more than one frag
can share the page. The last frag that uses the page unmaps the
page.  In the case when a page is not fully used (due to lack of space in rxq)
the last frag that uses the page is not being set as a "last_page_user";
instead, the next frag in the rxq is incorrectly being set.

The fix has also been tested on ppc64 with 64k pages...

Signed-off-by: Sathya Perla <>
Signed-off-by: David S. Miller <>
12 years agowimax/i2400m: Add support for more i6x50 SKUs
Inaky Perez-Gonzalez [Wed, 20 Jan 2010 20:41:13 +0000 (12:41 -0800)]
wimax/i2400m: Add support for more i6x50 SKUs

The Intel WiMax Wireless Link 6050 can show under more than one USB
ID. Add support for all, introducing a generic flag (i2400mu->i6050)
that denotes a 6x50 based device.

Signed-off-by: Inaky Perez-Gonzalez <>
12 years agoLinux 2.6.33-rc5 v2.6.33-rc5
Linus Torvalds [Thu, 21 Jan 2010 23:31:35 +0000 (15:31 -0800)]
Linux 2.6.33-rc5

12 years agoMerge branch 'perf-fixes-for-linus' of git://
Linus Torvalds [Thu, 21 Jan 2010 16:50:04 +0000 (08:50 -0800)]
Merge branch 'perf-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://
  perf: x86: Add support for the ANY bit
  perf: Change the is_software_event() definition
  perf: Honour event state for aux stream data
  perf: Fix perf_event_do_pending() fallback callsite
  perf kmem: Print usage help for unknown commands
  perf kmem: Increase "Hit" column length
  hw-breakpoints, perf: Fix broken mmiotrace due to dr6 by reference change
  perf timechart: Use tid not pid for COMM change

12 years agoMerge branch 'sched-fixes-for-linus' of git://
Linus Torvalds [Thu, 21 Jan 2010 16:49:52 +0000 (08:49 -0800)]
Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://
  sched: Reassign prev and switch_count when reacquire_kernel_lock() fail
  sched: Fix vmark regression on big machines

12 years agoMerge branch 'upstream-linus' of git://
Linus Torvalds [Thu, 21 Jan 2010 15:42:59 +0000 (07:42 -0800)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://
  libata: retry FS IOs even if it has failed with AC_ERR_INVALID

12 years agoMerge git://
Linus Torvalds [Thu, 21 Jan 2010 15:37:20 +0000 (07:37 -0800)]
Merge git://git./linux/kernel/git/gregkh/tty-2.6

* git://
  tty: fix race in tty_fasync
  serial: serial_cs: oxsemi quirk breaks resume
  serial: imx: bit &/| confusion
  serial: Fix crash if the minimum rate of the device is > 9600 baud
  serial-core: resume serial hardware with no_console_suspend
  serial: 8250_pnp: use wildcard for serial Wacom tablets
  nozomi: quick fix for the close/close bug
  compat_ioctl: Supress "unknown cmd" message on serial /dev/console

12 years agoMerge git://
Linus Torvalds [Thu, 21 Jan 2010 15:36:55 +0000 (07:36 -0800)]
Merge git://git./linux/kernel/git/gregkh/staging-2.6

* git://
  Staging: hv: fix smp problems in the hyperv core code
  Staging: et131x: Fix 2.6.33rc1 regression in et131x
  Staging: asus_oled: fix oops in

12 years agoMerge git://
Linus Torvalds [Thu, 21 Jan 2010 15:36:31 +0000 (07:36 -0800)]
Merge git://git./linux/kernel/git/gregkh/driver-core-2.6

* git://
  Revert "sysdev: fix prototype for memory_sysdev_class show/store functions"
  driver-core: fix devtmpfs crash on s390

12 years agoMerge git://
Linus Torvalds [Thu, 21 Jan 2010 15:35:43 +0000 (07:35 -0800)]
Merge git://git./linux/kernel/git/gregkh/usb-2.6

* git://
  USB: isp1362: fix build failure on ARM systems via irq_flags cleanup
  USB: isp1362: better 64bit printf warning fixes
  USB: fix usbstorage for 2770:915d delivers no FAT
  USB: Fix level of isp1760 Reloading ptd error message
  USB: FHCI: avoid NULL pointer dereference
  USB: Fix duplicate sysfs problem after device reset.
  USB: add speed values for USB 3.0 and wireless controllers
  USB: add missing delay during remote wakeup
  USB: EHCI & UHCI: fix race between root-hub suspend and port resume
  USB: EHCI: fix handling of unusual interrupt intervals
  USB: Don't use GFP_KERNEL while we cannot reset a storage device
  USB: fix bitmask merge error
  usb: serial: fix memory leak in generic driver
  USB: serial: fix USB serial fix kfifo_len locking

12 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Thu, 21 Jan 2010 15:32:11 +0000 (07:32 -0800)]
Merge branch 'for-linus' of git://

* 'for-linus' of git://
  fs/bio.c: fix shadows sparse warning
  drbd: The kernel code is now equivalent to out of tree release 8.3.7
  drbd: Allow online resizing of DRBD devices while peer not reachable (needs to be explicitly forced)
  drbd: Don't go into StandAlone mode when authentification failes because of network error
  drivers/block/drbd/drbd_receiver.c: correct NULL test
  cfq-iosched: Respect ioprio_class when preempting
  genhd: overlapping variable definition
  block: removed unused as_io_context
  DM: Fix device mapper topology stacking
  block: bdev_stack_limits wrapper
  block: Fix discard alignment calculation and printing
  block: Correct handling of bottom device misaligment
  drbd: check on CONFIG_LBDAF, not LBD
  drivers/block/drbd: Correct NULL test
  drbd: Silenced an assert that could triggered after changing write ordering method
  drbd: Kconfig fix
  drbd: Fix for a race between IO and a detach operation [Bugz 262]
  drbd: Use drbd_crypto_is_hash() instead of an open coded check

12 years agoMerge branch 'release' of git://
Linus Torvalds [Thu, 21 Jan 2010 15:29:36 +0000 (07:29 -0800)]
Merge branch 'release' of git://git./linux/kernel/git/lenb/linux-acpi-2.6

* 'release' of git:// (23 commits)
  ACPI: delete acpi_processor_power_verify_c2()
  ACPI: allow C3 > 1000usec
  ACPI: enable C2 and Turbo-mode on Nehalem notebooks on A/C
  ACPI: power_meter: remove double kfree()
  ACPI: processor: restrict early _PDC to opt-in platforms
  ACPI: Fix unused variable warning in sbs.c
  acpi: make ACPI device id constant
  sony-laptop - fix using of uninitialized variable
  ACPI: Fix section mismatch error for acpi_early_processor_set_pdc()
  eeepc-laptop: disable wireless hotplug for 1201N
  eeepc-laptop: add hotplug_disable parameter
  eeepc-laptop: switch to using sparse keymap library
  eeepc-laptop: dmi blacklist to disable pci hotplug code
  eeepc-laptop: disable cpu speed control on EeePC 701
  ACPI: don't cond_resched if irq is disabled
  ACPI: Remove unnecessary cast.
  ACPI: Advertise to BIOS in _OSC: _OST on _PPC changes
  ACPI: EC: Add wait for irq storm
  ACPI: SBS: Move SBS HC callback to faster Notify queue
  x86, ACPI: delete acpi_boot_table_init() return value

12 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Thu, 21 Jan 2010 15:28:54 +0000 (07:28 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/ecryptfs/ecryptfs-2.6

* 'for-linus' of git://
  ecryptfs: use after free
  ecryptfs: Eliminate useless code
  ecryptfs: fix interpose/interpolate typos in comments
  ecryptfs: pass matching flags to interpose as defined and used there
  ecryptfs: remove unnecessary d_drop calls in ecryptfs_link
  ecryptfs: don't ignore return value from lock_rename
  ecryptfs: initialize private persistent file before dereferencing pointer
  eCryptfs: Remove mmap from directory operations
  eCryptfs: Add getattr function
  eCryptfs: Use notify_change for truncating lower inodes

12 years agoMerge git://
Linus Torvalds [Thu, 21 Jan 2010 15:28:05 +0000 (07:28 -0800)]
Merge git://git./linux/kernel/git/mason/btrfs-unstable

* git://
  Btrfs: fix possible panic on unmount
  Btrfs: deal with NULL acl sent to btrfs_set_acl
  Btrfs: fix regression in orphan cleanup
  Btrfs: Fix race in btrfs_mark_extent_written
  Btrfs, fix memory leaks in error paths
  Btrfs: align offsets for btrfs_ordered_update_i_size
  btrfs: fix missing last-entry in readdir(3)

12 years agovmalloc: remove BUG_ON due to racy counting of VM_LAZY_FREE
Yongseok Koh [Tue, 19 Jan 2010 08:33:49 +0000 (17:33 +0900)]
vmalloc: remove BUG_ON due to racy counting of VM_LAZY_FREE

In free_unmap_area_noflush(), va->flags is marked as VM_LAZY_FREE first, and
then vmap_lazy_nr is increased atomically.

But, in __purge_vmap_area_lazy(), while traversing of vmap_are_list, nr
is counted by checking VM_LAZY_FREE is set to va->flags.  After counting
the variable nr, kernel reads vmap_lazy_nr atomically and checks a
BUG_ON condition whether nr is greater than vmap_lazy_nr to prevent
vmap_lazy_nr from being negative.

The problem is that, if interrupted right after marking VM_LAZY_FREE,
increment of vmap_lazy_nr can be delayed.  Consequently, BUG_ON
condition can be met because nr is counted more than vmap_lazy_nr.

It is highly probable when vmalloc/vfree are called frequently.  This
scenario have been verified by adding delay between marking VM_LAZY_FREE
and increasing vmap_lazy_nr in free_unmap_area_noflush().

Even the vmap_lazy_nr is for checking high watermark, it never be the
strict watermark.  Although the BUG_ON condition is to prevent
vmap_lazy_nr from being negative, vmap_lazy_nr is signed variable.  So,
it could go down to negative value temporarily.

Consequently, removing the BUG_ON condition is proper.

A possible BUG_ON message is like the below.

   kernel BUG at mm/vmalloc.c:517!
   invalid opcode: 0000 [#1] SMP
   EIP: 0060:[<c04824a4>] EFLAGS: 00010297 CPU: 3
   EIP is at __purge_vmap_area_lazy+0x144/0x150
   EAX: ee8a8818 EBX: c08e77d4 ECX: e7c7ae40 EDX: c08e77ec
   ESI: 000081fe EDI: e7c7ae60 EBP: e7c7ae64 ESP: e7c7ae3c
   DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
   Call Trace:
   [<c0482ad9>] free_unmap_vmap_area_noflush+0x69/0x70
   [<c0482b02>] remove_vm_area+0x22/0x70
   [<c0482c15>] __vunmap+0x45/0xe0
   [<c04831ec>] vmalloc+0x2c/0x30
   Code: 8d 59 e0 eb 04 66 90 89 cb 89 d0 e8 87 fe ff ff 8b 43 20 89 da 8d 48 e0 8d 43 20 3b 04 24 75 e7 fe 05 a8 a5 a3 c0 e9 78 ff ff ff <0f> 0b eb fe 90 8d b4 26 00 00 00 00 56 89 c6 b8 ac a5 a3 c0 31
   EIP: [<c04824a4>] __purge_vmap_area_lazy+0x144/0x150 SS:ESP 0068:e7c7ae3c

[ See also ]

Signed-off-by: Yongseok Koh <>
Reviewed-by: Minchan Kim <>
Cc: Nick Piggin <>
Cc: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
12 years agoMerge branch 'sh/for-2.6.33' of git://
Linus Torvalds [Thu, 21 Jan 2010 15:16:27 +0000 (07:16 -0800)]
Merge branch 'sh/for-2.6.33' of git://git./linux/kernel/git/lethal/sh-2.6

* 'sh/for-2.6.33' of git://
  sh64: wire up sys_accept4.
  sh: unwire sys_recvmmsg.
  sh: ms7724: Correct sh-eth EEPROM polling timeout.

12 years agoMerge
Linus Torvalds [Thu, 21 Jan 2010 15:15:10 +0000 (07:15 -0800)]

  ARM: 5888/1: arm: Update comments in cacheflush.h and remove unnecessary V6 and V7 comments
  ARM: 5886/1: arm: Fix cpu_proc_fin() for proc-v7.S and make kexec work
  ARM: 5885/1: arm: Flush TLB entries in setup_mm_for_reboot()
  ARM: 5884/1: arm: Fix DCC console for v7
  ARM: 5883/1: Revert "disable NX support for OABI-supporting kernels"
  ARM: 5882/1: ARM: Fix uncompress code compile for different defines of flush(void)
  ARM: fix badly placed mach/plat entries in Kconfig & Makefile

12 years agoperf: x86: Add support for the ANY bit
Stephane Eranian [Mon, 18 Jan 2010 08:58:01 +0000 (10:58 +0200)]
perf: x86: Add support for the ANY bit

Propagate the ANY bit into the fixed counter config for v3 and higher.

Signed-off-by: Stephane Eranian <>
[ split from larger patch]
Signed-off-by: Peter Zijlstra <>
LKML-Reference: <>
Signed-off-by: Ingo Molnar <>
12 years agoperf: Change the is_software_event() definition
Peter Zijlstra [Mon, 18 Jan 2010 13:02:16 +0000 (14:02 +0100)]
perf: Change the is_software_event() definition

The is_software_event() definition always confuses me because its an
exclusive expression, make it an inclusive one.

Signed-off-by: Peter Zijlstra <>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <>
12 years agoperf: Honour event state for aux stream data
Peter Zijlstra [Mon, 18 Jan 2010 08:12:32 +0000 (09:12 +0100)]
perf: Honour event state for aux stream data

Anton reported that perf record kept receiving events even after calling
ioctl(PERF_EVENT_IOC_DISABLE). It turns out that FORK,COMM and MMAP
events didn't respect the disabled state and kept flowing in.

Reported-by: Anton Blanchard <>
Signed-off-by: Peter Zijlstra <>
Tested-by: Anton Blanchard <>
LKML-Reference: <1263459187.4244.265.camel@laptop>
Signed-off-by: Ingo Molnar <>
12 years agoperf: Fix perf_event_do_pending() fallback callsite
Peter Zijlstra [Mon, 18 Jan 2010 08:08:26 +0000 (09:08 +0100)]
perf: Fix perf_event_do_pending() fallback callsite

Paul questioned the context in which we should call
perf_event_do_pending(). After looking at that I found that it should be
called from IRQ context these days, however the fallback call-site is
placed in softirq context. Ammend this by placing the callback in the IRQ
timer path.

Reported-by: Paul Mackerras <>
Signed-off-by: Peter Zijlstra <>
LKML-Reference: <1263374859.4244.192.camel@laptop>
Signed-off-by: Ingo Molnar <>
12 years agosched: Reassign prev and switch_count when reacquire_kernel_lock() fail
Yong Zhang [Mon, 11 Jan 2010 06:21:25 +0000 (14:21 +0800)]
sched: Reassign prev and switch_count when reacquire_kernel_lock() fail

Assume A->B schedule is processing, if B have acquired BKL before and it
need reschedule this time. Then on B's context, it will go to
need_resched_nonpreemptible for reschedule. But at this time, prev and
switch_count are related to A. It's wrong and will lead to incorrect
scheduler statistics.

Signed-off-by: Yong Zhang <>
Signed-off-by: Peter Zijlstra <>
LKML-Reference: <>
Signed-off-by: Ingo Molnar <>
12 years agosched: Fix vmark regression on big machines
Mike Galbraith [Mon, 4 Jan 2010 13:44:56 +0000 (14:44 +0100)]
sched: Fix vmark regression on big machines

SD_PREFER_SIBLING is set at the CPU domain level if power saving isn't
enabled, leading to many cache misses on large machines as we traverse
looking for an idle shared cache to wake to.  Change the enabler of
select_idle_sibling() to SD_SHARE_PKG_RESOURCES, and enable same at the
sibling domain level.

Reported-by: Lin Ming <>
Signed-off-by: Mike Galbraith <>
Signed-off-by: Peter Zijlstra <>
LKML-Reference: <>
Signed-off-by: Ingo Molnar <>
12 years agoe1000e: enhance frame fragment detection
Jesse Brandeburg [Tue, 19 Jan 2010 14:15:59 +0000 (14:15 +0000)]
e1000e: enhance frame fragment detection

Originally patched by Neil Horman <>

e1000e could with a jumbo frame enabled interface, and packet split disabled,
receive a packet that would overflow a single rx buffer.  While in practice
very hard to craft a packet that could abuse this, it is possible.

this is related to CVE-2009-4538

Signed-off-by: Jesse Brandeburg <>
CC: Neil Horman <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agoe1000: enhance frame fragment detection
Jesse Brandeburg [Tue, 19 Jan 2010 14:15:38 +0000 (14:15 +0000)]
e1000: enhance frame fragment detection

Originally From: Neil Horman <>
Modified by: Jesse Brandeburg <>

Hey all-
A security discussion was recently given:
And a patch that I submitted awhile back was brought up.  Apparently some of
their testing revealed that they were able to force a buffer fragment in e1000
in which the trailing fragment was greater than 4 bytes.  As a result the
fragment check I introduced failed to detect the fragement and a partial
invalid frame was passed up into the network stack.  I've written this patch
to correct it.  I'm in the process of testing it now, but it makes good
logical sense to me.  Effectively it maintains a per-adapter state variable
which detects a non-EOP frame, and discards it and subsequent non-EOP frames
leading up to _and_ _including_ the next positive-EOP frame (as it is by
definition the last fragment).  This should prevent any and all partial frames
from entering the network stack from e1000.

Signed-off-by: Jesse Brandeburg <>
Acked-by: Neil Horman <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agoe1000/e1000e/igb/igbvf/ixgb/ixgbe: Fix tests of unsigned in *_tx_map()
Roel Kluin [Tue, 19 Jan 2010 14:21:45 +0000 (14:21 +0000)]
e1000/e1000e/igb/igbvf/ixgb/ixgbe: Fix tests of unsigned in *_tx_map()

The variable count and i are unsigned so the (<|>=)0 tests do not work.

Signed-off-by: Roel Kluin <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: David S. Miller <>
12 years agoUSB: isp1362: fix build failure on ARM systems via irq_flags cleanup
Lothar Wassmann [Fri, 15 Jan 2010 19:42:02 +0000 (14:42 -0500)]
USB: isp1362: fix build failure on ARM systems via irq_flags cleanup

There was some left over #ifdef ARM logic that is outdated but no one
really noticed.  So instead of relying on this tricky logic, properly
load and utilize the platform irq_flags resources.

Reported-by: Ben Hutchings <>
Signed-off-by: Lothar Wassmann <>
Signed-off-by: Mike Frysinger <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: isp1362: better 64bit printf warning fixes
Lothar Wassmann [Fri, 15 Jan 2010 13:04:55 +0000 (08:04 -0500)]
USB: isp1362: better 64bit printf warning fixes

Some hosts that treat the return value of sizeof differently from unsigned
long might still hit warnings.  So use %zu for sizeof() values.  This is a
better version of the previous commit b0a9cf297e58721933.

Signed-off-by: Lothar Wassmann <>
Signed-off-by: Mike Frysinger <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: fix usbstorage for 2770:915d delivers no FAT
Ryan May [Wed, 6 Jan 2010 16:09:25 +0000 (10:09 -0600)]
USB: fix usbstorage for 2770:915d delivers no FAT

Resolves bug 14914.

Remove entry for 2770:915d (usb digital camera with mass storage
support) from unusual_devs.h. The fix triggered by the entry causes
the file system on the camera to be completely inaccessible (no
partition table, the device is not mountable).

The patch works, but let me clarify a few things about it.  All the
patch does is remove the entry for this device from the
drivers/usb/storage/unusual_devs.h, which is supposed to help with a
problem with the device's reported size (I think).  I'm pretty sure it
was originally added for a reason, so I'm not sure removing it won't
cause other problems to reappear.  Also, I should note that this
unusual_devs.h entry was present (and activating workarounds) in
2.6.29, but in that version everything works fine.  Starting with
2.6.30, things no longer work.

Signed-off-by: Ryan May <>
Cc: Rohan Hart <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: Fix level of isp1760 Reloading ptd error message
Colin Tuckley [Thu, 7 Jan 2010 11:22:47 +0000 (11:22 +0000)]
USB: Fix level of isp1760 Reloading ptd error message

This error message is not actually an error, it's an information
message. It is triggered when a transfer which ended in a NAQ is
retried successfully by the hardware.

Signed-off-by: Colin Tuckley <>
Cc: Sebastian Andrzej Siewior <>
Cc: Catalin Marinas <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: FHCI: avoid NULL pointer dereference
Alexander Beregalov [Thu, 7 Jan 2010 01:17:32 +0000 (04:17 +0300)]
USB: FHCI: avoid NULL pointer dereference

Assign fhci only if usb is not NULL.

Signed-off-by: Alexander Beregalov <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: Fix duplicate sysfs problem after device reset.
Sarah Sharp [Wed, 6 Jan 2010 18:16:51 +0000 (10:16 -0800)]
USB: Fix duplicate sysfs problem after device reset.

Borislav Petkov reports issues with duplicate sysfs endpoint files after a
resume from a hibernate.  It turns out that the code to support alternate
settings under xHCI has issues when a device with a non-default alternate
setting is reset during the hibernate:

[  427.681810] Restarting tasks ...
[  427.681995] hub 1-0:1.0: state 7 ports 6 chg 0004 evt 0000
[  427.682019] usb usb3: usb resume
[  427.682030] ohci_hcd 0000:00:12.0: wakeup root hub
[  427.682191] hub 1-0:1.0: port 2, status 0501, change 0000, 480 Mb/s
[  427.682205] usb 1-2: usb wakeup-resume
[  427.682226] usb 1-2: finish reset-resume
[  427.682886] done.
[  427.734658] ehci_hcd 0000:00:12.2: port 2 high speed
[  427.734663] ehci_hcd 0000:00:12.2: GetStatus port 2 status 001005 POWER sig=se0 PE CONNECT
[  427.746682] hub 3-0:1.0: hub_reset_resume
[  427.746693] hub 3-0:1.0: trying to enable port power on non-switchable hub
[  427.786715] usb 1-2: reset high speed USB device using ehci_hcd and address 2
[  427.839653] ehci_hcd 0000:00:12.2: port 2 high speed
[  427.839666] ehci_hcd 0000:00:12.2: GetStatus port 2 status 001005 POWER sig=se0 PE CONNECT
[  427.847717] ohci_hcd 0000:00:12.0: GetStatus roothub.portstatus [1] = 0x00010100 CSC PPS
[  427.915497] hub 1-2:1.0: remove_intf_ep_devs: if: ffff88022f9e8800 ->ep_devs_created: 1
[  427.915774] hub 1-2:1.0: remove_intf_ep_devs: bNumEndpoints: 1
[  427.915934] hub 1-2:1.0: if: ffff88022f9e8800: endpoint devs removed.
[  427.916158] hub 1-2:1.0: create_intf_ep_devs: if: ffff88022f9e8800 ->ep_devs_created: 0, ->unregistering: 0
[  427.916434] hub 1-2:1.0: create_intf_ep_devs: bNumEndpoints: 1
[  427.916609]  ep_81: create, parent hub
[  427.916632] ------------[ cut here ]------------
[  427.916644] WARNING: at fs/sysfs/dir.c:477 sysfs_add_one+0x82/0x96()
[  427.916649] Hardware name: System Product Name
[  427.916653] sysfs: cannot create duplicate filename '/devices/pci0000:00/0000:00:12.2/usb1/1-2/1-2:1.0/ep_81'
[  427.916658] Modules linked in: binfmt_misc kvm_amd kvm powernow_k8 cpufreq_ondemand cpufreq_powersave cpufreq_userspace freq_table cpufreq_conservative ipv6 vfat fat
+8250_pnp 8250 pcspkr ohci_hcd serial_core k10temp edac_core
[  427.916694] Pid: 278, comm: khubd Not tainted 2.6.33-rc2-00187-g08d869a-dirty #13
[  427.916699] Call Trace:

The problem is caused by a mismatch between the USB core's view of the
device state and the USB device and xHCI host's view of the device state.

After the device reset and re-configuration, the device and the xHCI host
think they are using alternate setting 0 of all interfaces.  However, the
USB core keeps track of the old state, which may include non-zero
alternate settings.  It uses intf->cur_altsetting to keep the endpoint
sysfs files for the old state across the reset.

The bandwidth allocation functions need to know what the xHCI host thinks
the current alternate settings are, so original patch set
intf->cur_altsetting to the alternate setting 0.  This caused duplicate
endpoint files to be created.

The solution is to not set intf->cur_altsetting before calling
usb_set_interface() in usb_reset_and_verify_device().  Instead, we add a
new flag to struct usb_interface to tell usb_hcd_alloc_bandwidth() to use
alternate setting 0 as the currently installed alternate setting.

Signed-off-by: Sarah Sharp <>
Tested-by: Borislav Petkov <>
Cc: Alan Stern <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: add speed values for USB 3.0 and wireless controllers
Greg Kroah-Hartman [Thu, 14 Jan 2010 18:33:19 +0000 (10:33 -0800)]
USB: add speed values for USB 3.0 and wireless controllers

These controllers say "unknown" for their speed in sysfs, which
obviously isn't correct.

Reported-by: Kurt Garloff <>
Cc: Sarah Sharp <>
Cc: David Vrabel <>
Cc: stable <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: add missing delay during remote wakeup
Alan Stern [Fri, 8 Jan 2010 16:18:38 +0000 (11:18 -0500)]
USB: add missing delay during remote wakeup

This patch (as1330) fixes a bug in khbud's handling of remote
wakeups.  When a device sends a remote-wakeup request, the parent hub
(or the host controller driver, for directly attached devices) begins
the resume sequence and notifies khubd when the sequence finishes.  At
this point the port's SUSPEND feature is automatically turned off.

However the device needs an additional 10-ms resume-recovery time
(TRSMRCY in the USB spec).  Khubd does not wait for this delay if the
SUSPEND feature is off, and as a result some devices fail to behave
properly following a remote wakeup.  This patch adds the missing
delay to the remote-wakeup path.

It also extends the resume-signalling delay used by ehci-hcd and
uhci-hcd from 20 ms (the value in the spec) to 25 ms (the value we use
for non-remote-wakeup resumes).  The extra time appears to help some

Signed-off-by: Alan Stern <>
Cc: stable <>
Cc: Rickard Bellini <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: EHCI & UHCI: fix race between root-hub suspend and port resume
Alan Stern [Fri, 8 Jan 2010 16:18:20 +0000 (11:18 -0500)]
USB: EHCI & UHCI: fix race between root-hub suspend and port resume

This patch (as1321) fixes a problem with EHCI and UHCI root-hub
suspends: If the suspend occurs while a port is trying to resume, the
resume doesn't finish and simply gets lost.  When remote wakeup is
enabled, this is undesirable behavior.

The patch checks first to see if any port resumes are in progress, and
if they are then it fails the root-hub suspend with -EBUSY.

Signed-off-by: Alan Stern <>
Cc: stable <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: EHCI: fix handling of unusual interrupt intervals
Alan Stern [Fri, 8 Jan 2010 16:17:55 +0000 (11:17 -0500)]
USB: EHCI: fix handling of unusual interrupt intervals

This patch (as1320) fixes two problems related to interrupt-URB
scheduling in ehci-hcd.

URBs with an interval of 2 or 4 microframes aren't handled.
For the time being, the patch reduces to interval to 1 uframe.

URBs are constrained to have an interval no larger than 1024
frames by usb_submit_urb().  But some EHCI controllers allow
use of a schedule as short as 256 frames; for these
controllers we may have to decrease the interval to the
actual schedule length.

The second problem isn't very significant since few devices expose
interrupt endpoints with an interval larger than 256 frames.  But the
first problem is critical; it will prevent the kernel from working
with devices having interrupt intervals of 2 or 4 uframes.

Signed-off-by: Alan Stern <>
Cc: stable <>
Tested-by: Glynn Farrow <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: Don't use GFP_KERNEL while we cannot reset a storage device
Oliver Neukum [Tue, 12 Jan 2010 11:32:50 +0000 (12:32 +0100)]
USB: Don't use GFP_KERNEL while we cannot reset a storage device

Memory allocations with GFP_KERNEL can cause IO to a storage
device which can fail resulting in a need to reset the device.
Therefore GFP_KERNEL cannot be safely used between usb_lock_device()
and usb_unlock_device(). Replace by GFP_NOIO.

Signed-off-by: Oliver Neukum <>
Cc: stable <>
Cc: Alan Stern <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: fix bitmask merge error
Alan Stern [Wed, 23 Dec 2009 04:16:32 +0000 (23:16 -0500)]
USB: fix bitmask merge error

This patch adds a mask bit which was mistakenly omitted from the
as1311 patch (usb-storage: add BAD_SENSE flag).

Signed-off-by: Alan Stern <>
Cc: stable <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agousb: serial: fix memory leak in generic driver
Johan Hovold [Wed, 6 Jan 2010 23:48:42 +0000 (15:48 -0800)]
usb: serial: fix memory leak in generic driver

Fix a regression introduced by commit
715b1dc01fe44537e8fce9566e4bb48d6821d84b ("USB: usb_debug,
usb_generic_serial: implement multi urb write").

URB transfer buffer was never freed when using multi-urb writes.
Currently the only driver enabling multi-urb writes is usb_debug.

Signed-off-by: Johan Hovold <>
Cc: Greg KH <>
Acked-by: Jason Wessel <>
Cc: stable <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoUSB: serial: fix USB serial fix kfifo_len locking
Stefani Seibold [Tue, 5 Jan 2010 13:30:31 +0000 (14:30 +0100)]
USB: serial: fix USB serial fix kfifo_len locking

This patch fix a possible race bug in drivers/usb/serial/generic with
the new kfifo API.

Please apply it to the 2.6.33-rc* tree.

Signed-off-by: Stefani Seibold <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoStaging: hv: fix smp problems in the hyperv core code
Greg Kroah-Hartman [Fri, 8 Jan 2010 17:06:40 +0000 (09:06 -0800)]
Staging: hv: fix smp problems in the hyperv core code

This fixes a number of SMP problems that were in the hyperv core code.

Patch originally written by K. Y. Srinivasan <>
but forward ported to the latest in-kernel code and tweaked slightly by

Novell, Inc. hereby disclaims all copyright in any derivative work
copyright associated with this patch.

Signed-off-by: K. Y. Srinivasan <>
Cc: Hank Janssen <>
Cc: Haiyang Zhang <>.
Cc: stable <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoStaging: et131x: Fix 2.6.33rc1 regression in et131x
Alan Cox [Wed, 6 Jan 2010 14:01:26 +0000 (14:01 +0000)]
Staging: et131x: Fix 2.6.33rc1 regression in et131x

et131x: Fix 12bit wrapping

From: Alan Cox <>

The 12bit wrap logic conversion is wrong and this shows up for some
memory sizes and layouts of card. Patch it up for now, once the kernel
view of status is cleaned up it'll become two variables and a lot saner.

Signed-off-by: Alan Cox <>
Cc: stable <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoStaging: asus_oled: fix oops in
Eugeni Dodonov [Wed, 23 Dec 2009 12:27:22 +0000 (10:27 -0200)]
Staging: asus_oled: fix oops in

After updating to 2.6.32 kernel, I started experiencing Oopses caused by
the asus_oled module. After quick investigation, I wrapped this simple
patch which fixes an Oops in by asus_oled module on kernel,
caused by incorrect usage of strict_strtoul function call within
set_enabled and set_disabled functions. This can be triggered by simple
running the userspace client for asus_old (e.g., 'asusoled -e' or
'asusoled -d').

Signed-off-by: Eugeni Dodonov <>
Cc: stable <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agotty: fix race in tty_fasync
Greg Kroah-Hartman [Thu, 17 Dec 2009 15:07:19 +0000 (07:07 -0800)]
tty: fix race in tty_fasync

We need to keep the lock held over the call to __f_setown() to
prevent a PID race.

Thanks to Al Viro for pointing out the problem, and to Travis for
making us look here in the first place.

Cc: Eric W. Biederman <>
Cc: Al Viro <>
Cc: Alan Cox <>
Cc: Linus Torvalds <>
Cc: Tavis Ormandy <>
Cc: Jeff Dike <>
Cc: Julien Tinnes <>
Cc: Matt Mackall <>
Cc: stable <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoserial: serial_cs: oxsemi quirk breaks resume
Pavel Machek [Wed, 9 Dec 2009 20:31:31 +0000 (12:31 -0800)]
serial: serial_cs: oxsemi quirk breaks resume

Quirk is applied on all cards with given manfid (is it that correct?).
Unfortunately, that quirk breaks resume on zaurus with billionton
bluetooth card inserted: c950ctrl is 0 and outb() faults.

I believe it is simply not a multiport card.  (info->multi == 1).  ...
...  confirmed by printks.

Signed-off-by: Pavel Machek <>
Acked-by: Alan Cox <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoserial: imx: bit &/| confusion
Roel Kluin [Wed, 9 Dec 2009 20:31:36 +0000 (12:31 -0800)]
serial: imx: bit &/| confusion

Since UCR1_UARTEN is defined 1, the port was always treated as enabled.

Signed-off-by: Roel Kluin <>
Cc: Alan Cox <>
Acked-by: Oskar Schirmer <>
Cc: Sascha Hauer <>
Cc: Fabian Godehardt <>
Cc: Daniel Glöckner <>
Cc: Russell King <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoserial: Fix crash if the minimum rate of the device is > 9600 baud
Alan Cox [Mon, 4 Jan 2010 16:26:21 +0000 (16:26 +0000)]
serial: Fix crash if the minimum rate of the device is > 9600 baud

In that situation if the old rate is invalid and the new rate is invalid
and the chip cannot do 9600 baud we report zero, which makes all the
drivers explode.

Instead force the rate based on min/max

Signed-off-by: Alan Cox <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoserial-core: resume serial hardware with no_console_suspend
Stanislav Brabec [Wed, 2 Dec 2009 15:20:56 +0000 (16:20 +0100)]
serial-core: resume serial hardware with no_console_suspend

Perform a tricky suspend/resume even with no_console_suspend.

With no_console_suspend, kernel skips serial port suspend/resume and the
serial hardware may remain in undefined state after resume. It actually
happens on devices that don't have BIOS that handle serial
initialization. It makes impossible to use serial console after resume.

Devices affected by this problem include:
Sharp Zaurus devices
Several PXA based ARM embedded boards

The patch does:
- Save the hardware state
- Perform buffer flush in time of its suspend call
- Tell the driver that port is suspended
- But still accept new data
- And keep console hardware in state that allows to send them

It allows to capture late console messages without breaking console
after resume.

This is just a resend of a patch discussed in these threads, as the
patch was not yet applied.

"Possible suspend/resume regression in .32-rc?" (Nov 1-5, 2009, ARM
list, later LKML)

"serial-core: resume serial hardware with no_console_suspend" (Sep
15-Oct 18, 2009, LKML & ARM lists)

Signed-off-by: Stanislav Brabec <>
Tested-by: Haojian Zhuang <>
Tested-by: Daniel Mack <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoserial: 8250_pnp: use wildcard for serial Wacom tablets
Matthew Garrett [Wed, 9 Dec 2009 20:31:37 +0000 (12:31 -0800)]
serial: 8250_pnp: use wildcard for serial Wacom tablets

Wacom claims that the WACF namespace will always be devoted to serial
Wacom tablets.  Remove the existing entries and add a wildcard to avoid
having to update the kernel every time they add a new device.

Signed-off-by: Ping Cheng <>
Signed-off-by: Matthew Garrett <>
Tested-by: Ping Cheng <>
Cc: Alan Cox <>
Signed-off-by: Andrew Morton <>
Cc: stable <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agonozomi: quick fix for the close/close bug
Alan Cox [Mon, 4 Jan 2010 16:26:50 +0000 (16:26 +0000)]
nozomi: quick fix for the close/close bug

Nozomi goes wrong if you get the sequence



which turns out to occur on some ppp type setups.

This is a quick patch up for the problem. It's not really fixing Nozomi
which completely fails to implement tty open/close semantics and all the
other needed stuff. Doing it right is a rather more invasive patch set and
not one that will backport.

Signed-off-by: Alan Cox <>
Cc: stable <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agocompat_ioctl: Supress "unknown cmd" message on serial /dev/console
Atsushi Nemoto [Sun, 10 Jan 2010 14:29:16 +0000 (23:29 +0900)]
compat_ioctl: Supress "unknown cmd" message on serial /dev/console

After the commit fb07a5f8 ("compat_ioctl: remove all VT ioctl
handling"), I got this error message on 64-bit mips kernel with 32-bit
busybox userland:

ioctl32(init:1): Unknown cmd fd(0) cmd(00005600){t:'V';sz:0} arg(7fd76480) on /dev/console

The cmd 5600 is VT_OPENQRY.  The busybox's init issues this ioctl to
know vt-console or serial-console.  If the console was serial console,
VT ioctls are not handled by the serial driver.

And by quick search, I found some programs using VT_GETMODE to check
vt-console is available or not.

Signed-off-by: Atsushi Nemoto <>
Cc: Arnd Bergmann <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agoRevert "sysdev: fix prototype for memory_sysdev_class show/store functions"
Greg Kroah-Hartman [Tue, 19 Jan 2010 21:08:16 +0000 (13:08 -0800)]
Revert "sysdev: fix prototype for memory_sysdev_class show/store functions"

This reverts commit 8ff410daa009c4b44be445ded5b0cec00abc0426

It should not have been sent to Linus's tree yet, as it depends
on changes that are queued up in my driver-core for the .34 kernel

Cc: Wu Fengguang <>
Cc: Andi Kleen <>
Cc: "Zheng, Shaohui" <>
Cc: Andrew Morton <>
Cc: Linus Torvalds
Signed-off-by: Greg Kroah-Hartman <>
12 years agodriver-core: fix devtmpfs crash on s390
Heiko Carstens [Mon, 18 Jan 2010 13:36:12 +0000 (14:36 +0100)]
driver-core: fix devtmpfs crash on s390

On Mon, Jan 18, 2010 at 05:26:20PM +0530, Sachin Sant wrote:
> Hello Heiko,
> Today while trying to boot next-20100118 i came across
> the following Oops :
> Brought up 4 CPUs
> Unable to handle kernel pointer dereference at virtual kernel address 0000000000
> 543000
> Oops: 0004 #1 SMP
> Modules linked in:
> CPU: 0 Not tainted 2.6.33-rc4-autotest-next-20100118-5-default #1
> Process swapper (pid: 1, task: 00000000fd792038, ksp: 00000000fd797a30)
> Krnl PSW : 0704200180000000 00000000001eb0b8 (shmem_parse_options+0xc0/0x328)
>           R:0 T:1 IO:1 EX:1 Key:0 M:1 W:0 P:0 AS:0 CC:2 PM:0 EA:3
> Krnl GPRS: 000000000054388a 000000000000003d 0000000000543836 000000000000003d
>           0000000000000000 0000000000483f28 0000000000536112 00000000fd797d00
>           00000000fd4ba100 0000000000000100 0000000000483978 0000000000543832
>           0000000000000000 0000000000465958 00000000001eb0b0 00000000fd797c58
> Krnl Code: 00000000001eb0aac0e5000994f1       brasl   %r14,31da8c
>           00000000001eb0b0b9020022           ltgr    %r2,%r2
>           00000000001eb0b4a784010b           brc     8,1eb2ca
>          >00000000001eb0b892002000           mvi     0(%r2),0
>           00000000001eb0bca7080000           lhi     %r0,0
>           00000000001eb0c041902001           la      %r9,1(%r2)
>           00000000001eb0c4b9040016           lgr     %r1,%r6
>           00000000001eb0c8b904002b           lgr     %r2,%r11
> Call Trace:
> (<00000000fd797c50> 0xfd797c50)
> <00000000001eb5da> shmem_fill_super+0x13a/0x25c
> <0000000000228cfa> get_sb_single+0xbe/0xdc
> <000000000034ffc0> dev_get_sb+0x2c/0x38
> <000000000066c602> devtmpfs_init+0x46/0xc0
> <000000000066c53e> driver_init+0x22/0x60
> <000000000064d40a> kernel_init+0x24e/0x3d0
> <000000000010a7ea> kernel_thread_starter+0x6/0xc
> <000000000010a7e4> kernel_thread_starter+0x0/0xc
> I never tried to boot a kernel with DEVTMPFS enabled on a s390 box.
> So am wondering if this is supported or not ? If you think this
> is supported i will send a mail to community on this.

There is nothing arch specific to devtmpfs. This part crashes because the
kernel tries to modify the data read-only section which is write protected
on s390.

Signed-off-by: Heiko Carstens <>
Acked-by: Kay Sievers <>
Signed-off-by: Greg Kroah-Hartman <>
12 years agolibata: retry FS IOs even if it has failed with AC_ERR_INVALID
Tejun Heo [Thu, 14 Jan 2010 07:18:09 +0000 (16:18 +0900)]
libata: retry FS IOs even if it has failed with AC_ERR_INVALID

libata currently doesn't retry if a command fails with AC_ERR_INVALID
assuming that retrying won't get it any further even if retried.
However, a failure may be classified as invalid through hardware
glitch (incorrect reading of the error register or firmware bug) and
there isn't whole lot to gain by not retrying as actually invalid
commands will be failed immediately.  Also, commands serving FS IOs
are extremely unlikely to be invalid.  Retry FS IOs even if it's
marked invalid.

Transient and incorrect invalid failure was seen while debugging
firmware related issue on Samsung n130 on bko#14314.

Signed-off-by: Tejun Heo <>
Reported-by: Johannes Stezenbach <>
Signed-off-by: Jeff Garzik <>
12 years agoMerge branch 'bugzilla-14954' into release
Len Brown [Wed, 20 Jan 2010 06:26:22 +0000 (01:26 -0500)]
Merge branch 'bugzilla-14954' into release

12 years agoMerge branch 'misc' into release
Len Brown [Wed, 20 Jan 2010 06:23:27 +0000 (01:23 -0500)]
Merge branch 'misc' into release

12 years agoMerge branch 'osc-bugfix' into release
Len Brown [Wed, 20 Jan 2010 06:23:18 +0000 (01:23 -0500)]
Merge branch 'osc-bugfix' into release

12 years agoMerge branch 'eeepc-laptop' into release
Len Brown [Wed, 20 Jan 2010 06:23:01 +0000 (01:23 -0500)]
Merge branch 'eeepc-laptop' into release