openwrt/staging/blogic.git
15 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Mon, 26 Jan 2009 05:36:34 +0000 (21:36 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6

15 years agoaf_key: initialize xfrm encap_oa
Timo Teras [Mon, 26 Jan 2009 04:49:14 +0000 (20:49 -0800)]
af_key: initialize xfrm encap_oa

Currently encap_oa is left uninitialized, so it contains garbage data which
is visible to userland via Netlink. Initialize it by zeroing it out.

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agovirtio_net: Fix MAX_PACKET_LEN to support 802.1Q VLANs
Alex Williamson [Mon, 26 Jan 2009 02:06:26 +0000 (18:06 -0800)]
virtio_net: Fix MAX_PACKET_LEN to support 802.1Q VLANs

802.1Q expanded the maximum ethernet frame size by 4 bytes for the
VLAN tag.  We're not taking this into account in virtio_net, which
means the buffers we provide to the backend in the virtqueue RX ring
aren't big enough to hold a full MTU VLAN packet.  For QEMU/KVM,
this results in the backend exiting with a packet truncation error.

Signed-off-by: Alex Williamson <alex.williamson@hp.com>
Acked-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agolcs: fix compilation for !CONFIG_IP_MULTICAST
Heiko Carstens [Tue, 20 Jan 2009 06:14:32 +0000 (06:14 +0000)]
lcs: fix compilation for !CONFIG_IP_MULTICAST

drivers/s390/net/lcs.c: In function 'lcs_new_device':
drivers/s390/net/lcs.c:2179: error: implicit declaration of function 'lcs_set_multicast_list'

Reported-by: Kamalesh Babulal <kamalesh@linux.vnet.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agortl8187: Add termination packet to prevent stall
Larry Finger [Fri, 23 Jan 2009 17:46:32 +0000 (11:46 -0600)]
rtl8187: Add termination packet to prevent stall

The RTL8187 and RTL8187B devices can stall unless an explicit termination
packet is sent.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: fix rs_get_rate WARN_ON()
Abbas, Mohamed [Wed, 21 Jan 2009 18:58:02 +0000 (10:58 -0800)]
iwlwifi: fix rs_get_rate WARN_ON()

In ieee80211_sta structure there is u64 supp_rates[IEEE80211_NUM_BANDS]
this is filled with all support rate from assoc_resp.  If we associate
with G-band AP only supp_rates of G-band will be set the other band
supp_rates will be set to 0. If the user type this command
this will cause mac80211 to set to new channel, mac80211
does not disassociate in setting new channel, so the active
band is now A-band. then in handling the new essid mac80211 will
kick in the assoc steps which involve sending disassociation frame.
in this mac80211 will WARN_ON sta->supp_rates[A_BAND] == 0.

This fixes:
http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1822
http://www.kerneloops.org/searchweek.php?search=rs_get_rate

Signed-off-by: mohamed abbas <mohamed.abbas@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54usb: fix packet loss with first generation devices
Christian Lamparter [Tue, 20 Jan 2009 22:11:21 +0000 (23:11 +0100)]
p54usb: fix packet loss with first generation devices

Artur Skawina confirmed that the first generation devices needs the same
URB_ZERO_PACKET flag, in oder to finish the pending transfer properly.
The second generation has been successfully fixed by
"p54usb: fix random traffic stalls (LM87)" (43af18f06d5)

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agosctp: Fix another socket race during accept/peeloff
Vlad Yasevich [Thu, 22 Jan 2009 22:53:23 +0000 (14:53 -0800)]
sctp: Fix another socket race during accept/peeloff

There is a race between sctp_rcv() and sctp_accept() where we
have moved the association from the listening socket to the
accepted socket, but sctp_rcv() processing cached the old
socket and continues to use it.

The easy solution is to check for the socket mismatch once we've
grabed the socket lock.  If we hit a mis-match, that means
that were are currently holding the lock on the listening socket,
but the association is refrencing a newly accepted socket.  We need
to drop the lock on the old socket and grab the lock on the new one.

A more proper solution might be to create accepted sockets when
the new association is established, similar to TCP.  That would
eliminate the race for 1-to-1 style sockets, but it would still
existing for 1-to-many sockets where a user wished to peeloff an
association.  For now, we'll live with this easy solution as
it addresses the problem.

Reported-by: Michal Hocko <mhocko@suse.cz>
Reported-by: Karsten Keil <kkeil@suse.de>
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosctp: Properly timestamp outgoing data chunks for rtx purposes
Vlad Yasevich [Thu, 22 Jan 2009 22:53:01 +0000 (14:53 -0800)]
sctp: Properly timestamp outgoing data chunks for rtx purposes

Recent changes to the retransmit code exposed a long standing
bug where it was possible for a chunk to be time stamped
after the retransmit timer was reset.  This caused a rare
situation where the retrnamist timer has expired, but
nothing was marked for retrnasmission because all of
timesamps on data were less then 1 rto ago.  As result,
the timer was never restarted since nothing was retransmitted,
and this resulted in a hung association that did couldn't
complete the data transfer.  The solution is to timestamp
the chunk when it's added to the packet for transmission
purposes.  After the packet is trsnmitted the rtx timer
is restarted.  This guarantees that when the timer expires,
there will be data to retransmit.

Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosctp: Correctly start rtx timer on new packet transmissions.
Vlad Yasevich [Thu, 22 Jan 2009 22:52:43 +0000 (14:52 -0800)]
sctp: Correctly start rtx timer on new packet transmissions.

Commit 62aeaff5ccd96462b7077046357a6d7886175a57
(sctp: Start T3-RTX timer when fast retransmitting lowest TSN)
introduced a regression where it was possible to forcibly
restart the sctp retransmit timer at the transmission of any
new chunk.  This resulted in much longer timeout times and
sometimes hung sctp connections.

Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosctp: Fix crc32c calculations on big-endian arhes.
Vlad Yasevich [Thu, 22 Jan 2009 22:52:23 +0000 (14:52 -0800)]
sctp: Fix crc32c calculations on big-endian arhes.

crc32c algorithm provides a byteswaped result.  On little-endian
arches, the result ends up in big-endian/network byte order.
On big-endinan arches, the result ends up in little-endian
order and needs to be byte swapped again.  Thus calling cpu_to_le32
gives the right output.

Tested-by: Jukka Taimisto <jukka.taimisto@mail.suomi.net>
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agophy: Add suspend/resume support to SMSC PHYs
Steve Glendinning [Thu, 22 Jan 2009 22:07:43 +0000 (14:07 -0800)]
phy: Add suspend/resume support to SMSC PHYs

All supported SMSC PHYs implement the standard "power down" bit 11 of
BMCR, so this patch adds support using the generic genphy_{suspend,resume}
functions.

Signed-off-by: Steve Glendinning <steve.glendinning@smsc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agophylib: Fix oops in suspend/resume paths
Anton Vorontsov [Thu, 22 Jan 2009 21:51:24 +0000 (13:51 -0800)]
phylib: Fix oops in suspend/resume paths

Suspend/resume routines check for phydrv != NULL, but that is
wrong because "phydrv" comes from container_of(drv). If drv is NULL,
then container_of(drv) will return non-NULL result, and the checks
won't work.

The Freescale TBI PHYs are driver-less, so "drv" is NULL, and that
leads to the following oops:

Unable to handle kernel paging request for data at address 0xffffffe4
Faulting instruction address: 0xc0215554
Oops: Kernel access of bad area, sig: 11 [#1]
[...]
NIP [c0215554] mdio_bus_suspend+0x34/0x70
LR [c01cc508] suspend_device+0x258/0x2bc
Call Trace:
[cfad3da0] [cfad3db8] 0xcfad3db8 (unreliable)
[cfad3db0] [c01cc508] suspend_device+0x258/0x2bc
[cfad3dd0] [c01cc62c] dpm_suspend+0xc0/0x140
[cfad3e20] [c01cc6f4] device_suspend+0x48/0x5c
[cfad3e40] [c0068dd8] suspend_devices_and_enter+0x8c/0x148
[cfad3e60] [c00690f8] enter_state+0x100/0x118
[cfad3e80] [c00691c0] state_store+0xb0/0xe4
[cfad3ea0] [c018c938] kobj_attr_store+0x24/0x3c
[cfad3eb0] [c00ea9a8] flush_write_buffer+0x58/0x7c
[cfad3ed0] [c00eadf0] sysfs_write_file+0x58/0xa0
[cfad3ef0] [c009e810] vfs_write+0xb4/0x16c
[cfad3f10] [c009ed40] sys_write+0x4c/0x90
[cfad3f40] [c0014954] ret_from_syscall+0x0/0x38
[...]

This patch fixes the issue, plus removes unneeded parentheses
and fixes indentation level in mdio_bus_suspend().

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Version
Eilon Greenstein [Thu, 22 Jan 2009 06:01:37 +0000 (06:01 +0000)]
bnx2x: Version

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: loopback test failure
Eilon Greenstein [Thu, 22 Jan 2009 06:01:32 +0000 (06:01 +0000)]
bnx2x: loopback test failure

A link change interrupt might be queued and activated after the loopback was set
and it will cause the loopback to fail. The PHY lock should be kept until the
loopback test is over.

That implies that the bnx2x_test_link should used within the loopback function
and not bnx2x_wait_for_link since that function also takes the PHY link

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Missing rmb when waiting for FW response
Eilon Greenstein [Thu, 22 Jan 2009 06:01:29 +0000 (06:01 +0000)]
bnx2x: Missing rmb when waiting for FW response

Waiting for the FW to response requires a memory barrier

Signed-off-by: Michal Kalderon <michals@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Calling napi_del
Eilon Greenstein [Thu, 22 Jan 2009 06:01:25 +0000 (06:01 +0000)]
bnx2x: Calling napi_del

rmmod might hang without this patch since the reference counter is not going
down

Signed-off-by: Yitchak Gertner <gertner@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Carrier off first call
Eilon Greenstein [Thu, 22 Jan 2009 03:37:48 +0000 (03:37 +0000)]
bnx2x: Carrier off first call

Call carrier off should not be called after register_netdev since after
register netdev open can be called at any time followed by an interrupt that
will set it to carrier_on and the probe will resume control and set it to off

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Handling load failures
Eilon Greenstein [Thu, 22 Jan 2009 03:37:44 +0000 (03:37 +0000)]
bnx2x: Handling load failures

Failures on load were not handled correctly - separate the flow to handle

different failures

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Disable napi
Eilon Greenstein [Thu, 22 Jan 2009 03:37:36 +0000 (03:37 +0000)]
bnx2x: Disable napi

Calling napi disabled unconditionally at netif stop

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Reset HW before use
Eilon Greenstein [Thu, 22 Jan 2009 03:37:31 +0000 (03:37 +0000)]
bnx2x: Reset HW before use

To avoid complications, make sure that the HW is in reset (as it should be)
before trying to take it out of reset. In normal flows, the HW is indeed in rest
so this should have no effect

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoiwlwifi: return NETDEV_TX_OK from _tx ops
Reinette Chatre [Mon, 19 Jan 2009 23:30:32 +0000 (15:30 -0800)]
iwlwifi: return NETDEV_TX_OK from _tx ops

be consistent with mac80211 drivers and return correct return code.
NETDEV_TX_OK is 0, but we need to be consistent wrt formatting amongst
implementations

re: http://marc.info/?l=linux-wireless&m=123119327419865&w=2

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Reviewed-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agozd1211rw: adding Sitecom WL-603 (0df6:0036) to the USB id list
Hin-Tak Leung [Mon, 19 Jan 2009 23:39:09 +0000 (23:39 +0000)]
zd1211rw: adding Sitecom WL-603 (0df6:0036) to the USB id list

Giuseppe Cala <jiveaxe@gmail.com> (The second "a" in "Cala" should be
a grave, U+00E0) reported success on zd1211-devs@lists.sourceforge.net.
The chip info is:
zd1211b chip 0df6:0036 v4810 high 00-0c-f6 AL2230_RF pa0 g--N-

The Sitecom WL-603 is detected as a zd1211b with a AL2230 RF transceiver chip.

Signed-off-by: Giuseppe Cala <jiveaxe@gmail.com>
Signed-off-by: Hin-Tak Leung <htl10@users.sourceforge.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54usb: fix nasty use after free
Christian Lamparter [Mon, 19 Jan 2009 23:27:57 +0000 (00:27 +0100)]
p54usb: fix nasty use after free

In theory, the firmware acks the received a data frame, before signaling the driver to free it again.
However Artur Skawina <art.08.09@gmail.com> has shown that it can happen in reverse order as well.
This is very bad and could lead to memory corruptions, oopses and panics.

Thanks to Artur Skawina <art.08.09@gmail.com> for reporting and debugging this issue.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Tested-by: Artur Skawina <art.08.09@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: more cryptographic accelerator fixes
Christian Lamparter [Mon, 19 Jan 2009 15:08:48 +0000 (16:08 +0100)]
p54: more cryptographic accelerator fixes

If we let the firmware do the data encryption, we have to remove the ICV and
(M)MIC at the end of the frame before we can give it back to mac80211.
Or, these data frames have a few trailing bytes on cooked monitor interfaces.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: add missing break in eeprom parser
Chr [Mon, 19 Jan 2009 13:30:26 +0000 (14:30 +0100)]
p54: add missing break in eeprom parser

This patch fixes a obvious memory leak in the eeprom parser.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoorinoco: use KERN_DEBUG for link status messages
Pavel Roskin [Mon, 19 Jan 2009 04:20:58 +0000 (23:20 -0500)]
orinoco: use KERN_DEBUG for link status messages

KERN_INFO is too "loud" for messages that are generated by the ordinary
events, such as accociation.  Use of KERN_DEBUG is consistent with
mac80211.

Suggested by Michael Gilbert <michael.s.gilbert@gmail.com>

Signed-off-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Fix TX rate short preamble detection
Ivo van Doorn [Sun, 18 Jan 2009 19:15:24 +0000 (20:15 +0100)]
rt2x00: Fix TX rate short preamble detection

Mac80211 provides 2 structures to handle bitrates, namely
ieee80211_rate and ieee80211_tx_rate. To determine the short preamble
mode for an outgoing frame, the flag IEEE80211_TX_RC_USE_SHORT_PREAMBLE
must be checked on ieee80211_tx_rate and not ieee80211_rate (which rt2x00 did).

This fixes a regression which was triggered in 2.6.29-rcX as reported by Chris Clayton.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Tested-By: Chris Clayton <chris2553@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: fix slot time debug message
Christian Lamparter [Sun, 18 Jan 2009 01:24:15 +0000 (02:24 +0100)]
mac80211: fix slot time debug message

wlan0: switched to short barker preamble (BSSID=00:01:aa:bb:cc:dd)
wlan0: switched to short slot (BSSID=) <something is missing here>

should be:

wlan0: switched to short barker preamble (BSSID=00:01:aa:bb:cc:dd)
wlan0: switched to short slot (BSSID=00:01:aa:bb:cc:dd)

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoorinoco: move kmalloc(..., GFP_KERNEL) outside spinlock in orinoco_ioctl_set_genie
Andrey Borzenkov [Tue, 20 Jan 2009 17:26:46 +0000 (20:26 +0300)]
orinoco: move kmalloc(..., GFP_KERNEL) outside spinlock in orinoco_ioctl_set_genie

[   56.923623] BUG: sleeping function called from invalid context at /home/bor/src/linux-git/mm/slub.c:1599
[   56.923644] in_atomic(): 0, irqs_disabled(): 1, pid: 3031, name: wpa_supplicant
[   56.923656] 2 locks held by wpa_supplicant/3031:
[   56.923662]  #0:  (rtnl_mutex){--..}, at: [<c02abd1f>] rtnl_lock+0xf/0x20
[   56.923703]  #1:  (&priv->lock){++..}, at: [<dfc840c2>] orinoco_ioctl_set_genie+0x52/0x130 [orinoco]
[   56.923782] irq event stamp: 910
[   56.923788] hardirqs last  enabled at (909): [<c01957db>] __kmalloc+0x7b/0x140
[   56.923820] hardirqs last disabled at (910): [<c0309419>] _spin_lock_irqsave+0x19/0x80
[   56.923847] softirqs last  enabled at (880): [<c0124f54>] __do_softirq+0xc4/0x110
[   56.923865] softirqs last disabled at (871): [<c01049ae>] do_softirq+0x8e/0xe0
[   56.923895] Pid: 3031, comm: wpa_supplicant Not tainted 2.6.29-rc2-1avb #1
[   56.923905] Call Trace:
[   56.923919]  [<c01049ae>] ? do_softirq+0x8e/0xe0
[   56.923941]  [<c011ad12>] __might_sleep+0xd2/0x100
[   56.923952]  [<c0195837>] __kmalloc+0xd7/0x140
[   56.923963]  [<c030946a>] ? _spin_lock_irqsave+0x6a/0x80
[   56.923981]  [<dfc840e9>] ? orinoco_ioctl_set_genie+0x79/0x130 [orinoco]
[   56.923999]  [<dfc840c2>] ? orinoco_ioctl_set_genie+0x52/0x130 [orinoco]
[   56.924017]  [<dfc840e9>] orinoco_ioctl_set_genie+0x79/0x130 [orinoco]
[   56.924036]  [<c0209325>] ? copy_from_user+0x35/0x130
[   56.924061]  [<c02ffd96>] ioctl_standard_call+0x196/0x380
[   56.924085]  [<c029f945>] ? __dev_get_by_name+0x85/0xb0
[   56.924096]  [<c02ff88f>] wext_handle_ioctl+0x14f/0x230
[   56.924113]  [<dfc84070>] ? orinoco_ioctl_set_genie+0x0/0x130 [orinoco]
[   56.924132]  [<c02a3da5>] dev_ioctl+0x495/0x570
[   56.924155]  [<c0293e05>] ? sys_sendto+0xa5/0xd0
[   56.924171]  [<c0142fe8>] ? mark_held_locks+0x48/0x90
[   56.924183]  [<c0292880>] ? sock_ioctl+0x0/0x280
[   56.924193]  [<c029297d>] sock_ioctl+0xfd/0x280
[   56.924203]  [<c0292880>] ? sock_ioctl+0x0/0x280
[   56.924235]  [<c01a51d0>] vfs_ioctl+0x20/0x80
[   56.924246]  [<c01a53e2>] do_vfs_ioctl+0x72/0x570
[   56.924257]  [<c0293e62>] ? sys_send+0x32/0x40
[   56.924268]  [<c02947c0>] ? sys_socketcall+0x1d0/0x2a0
[   56.924280]  [<c010339f>] ? sysenter_exit+0xf/0x16
[   56.924292]  [<c01a5919>] sys_ioctl+0x39/0x70
[   56.924302]  [<c0103371>] sysenter_do_call+0x12/0x31

Signed-off-by: Andrey Borzenkov <arvidjaar@mail.ru>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: decrement ref count to netdev after launching mesh discovery
Brian Cavagnolo [Sat, 17 Jan 2009 03:04:49 +0000 (19:04 -0800)]
mac80211: decrement ref count to netdev after launching mesh discovery

After launching mesh discovery in tx path, reference count was not being
decremented.  This was preventing module unload.

Signed-off-by: Brian Cavagnolo <brian@cozybit.com>
Signed-off-by: Andrey Yurovsky <andrey@cozybit.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agousb/mcs7830: Don't use buffers from stack for USB transfers
Christian Eggers [Wed, 21 Jan 2009 20:56:24 +0000 (12:56 -0800)]
usb/mcs7830: Don't use buffers from stack for USB transfers

mcs7830_set_reg() and mcs7830_get_reg() are called with buffers
from stack which must not be used directly for USB transfers.
This causes corruption of the stack particulary on non x86
architectures because DMA may be used for these transfers.

Signed-off-by: Christian Eggers <christian.eggers@kathrein.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agousbnet: allow type check of devdbg arguments in non-debug build
Steve Glendinning [Wed, 21 Jan 2009 20:35:43 +0000 (12:35 -0800)]
usbnet: allow type check of devdbg arguments in non-debug build

Improve usbnet's devdbg to always type-check diagnostic arguments,
like dev_dbg (device.h).  This makes no change to the resulting size of
usbnet modules.

This patch also removes an #ifdef DEBUG directive from rndis_wlan so
it's devdbg statements are always type-checked at compile time.

Signed-off-by: Steve Glendinning <steve.glendinning@smsc.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetfilter: ctnetlink: fix scheduling while atomic
Patrick McHardy [Wed, 21 Jan 2009 20:19:49 +0000 (12:19 -0800)]
netfilter: ctnetlink: fix scheduling while atomic

Caused by call to request_module() while holding nf_conntrack_lock.

Reported-and-tested-by: Kövesdi György <kgy@teledigit.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agogro: Fix merging of paged packets
Herbert Xu [Sat, 17 Jan 2009 19:48:13 +0000 (19:48 +0000)]
gro: Fix merging of paged packets

The previous fix to paged packets broke the merging because it
reset the skb->len before we added it to the merged packet.  This
wasn't detected because it simply resulted in the truncation of
the packet while the missing bit is subsequently retransmitted.

The fix is to store skb->len before we clobber it.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agogro: Fix error handling on extremely short frags
Herbert Xu [Sat, 17 Jan 2009 19:47:18 +0000 (19:47 +0000)]
gro: Fix error handling on extremely short frags

When a frag is shorter than an Ethernet header, we'd return a
zeroed packet instead of aborting.  This patch fixes that.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agogro: Fix handling of complete checksums in IPv6
Herbert Xu [Sat, 17 Jan 2009 19:46:16 +0000 (19:46 +0000)]
gro: Fix handling of complete checksums in IPv6

We need to perform skb_postpull_rcsum after pulling the IPv6
header in order to maintain the correctness of the complete
checksum.

This patch also adds a missing iph reload after pulling.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agovia-velocity: fix hot spin
roel kluin [Sat, 17 Jan 2009 11:14:31 +0000 (11:14 +0000)]
via-velocity: fix hot spin

while(--j >= 0) keeps spinning when j is unsigned:

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomyri10ge: don't forget pci_disable_device()
Brice Goglin [Sat, 17 Jan 2009 08:27:19 +0000 (08:27 +0000)]
myri10ge: don't forget pci_disable_device()

Don't forget to call pci_disable_device() in myri10ge_remove()
and when myri10ge_probe() fails.

By the way, update the copyright years.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoNET: net_namespace, fix lock imbalance
Jiri Slaby [Sat, 17 Jan 2009 06:47:12 +0000 (06:47 +0000)]
NET: net_namespace, fix lock imbalance

register_pernet_gen_subsys omits mutex_unlock in one fail path.
Fix it.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Tue, 20 Jan 2009 22:37:10 +0000 (14:37 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6

15 years agoRevert "xfrm: For 32/64 compatability wrt. xfrm_usersa_info"
David S. Miller [Tue, 20 Jan 2009 17:49:51 +0000 (09:49 -0800)]
Revert "xfrm: For 32/64 compatability wrt. xfrm_usersa_info"

This reverts commit fc8c7dc1b29560c016a67a34ccff32a712b5aa86.

As indicated by Jiri Klimes, this won't work.  These numbers are
not only used the size validation, they are also used to locate
attributes sitting after the message.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agowimax/i2400m: error paths that need to free an skb should use kfree_skb()
Inaky Perez-Gonzalez [Mon, 19 Jan 2009 13:19:30 +0000 (13:19 +0000)]
wimax/i2400m: error paths that need to free an skb should use kfree_skb()

Roel Kluin reported a bug in two error paths where skbs were wrongly
being freed using kfree(). He provided a fix where it was replaced to
kfree_skb(), as it should be.

However, in i2400mu_rx(), the error path was missing returning an
indication of the failure. Changed to reset rx_skb to NULL and return
it to the caller, i2400mu_rxd(). It will be treated as a transient
error and just ignore the packet.

Depending on the buffering conditions inside the device, the data
packet might be dropped or the device will signal the host again for
data-ready-to-read and the host will retry.

Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomv643xx_eth: prevent interrupt storm on ifconfig down
Gabriel Paubert [Tue, 20 Jan 2009 01:18:09 +0000 (17:18 -0800)]
mv643xx_eth: prevent interrupt storm on ifconfig down

Contrary to what the docs say, the 'extended interrupt cause' bit in
the interrupt cause register (bit 1) appears to not be maskable on at
least some of the mv643xx_eth platforms, making writing zeroes to the
interrupt mask register but not the extended interrupt mask register
insufficient to stop interrupts from occuring.  Therefore, also write
zeroes to the extended interrupt mask register when shutting down the
port.

This fixes the interrupt storm seen on the Pegasos board when shutting
down the interface.

Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomv643xx_eth: fix multicast filter programming
Lennert Buytenhek [Tue, 20 Jan 2009 01:17:18 +0000 (17:17 -0800)]
mv643xx_eth: fix multicast filter programming

Commit 66e63ffbc04706568d8789cbb00eaa8ddbcae648 ("mv643xx_eth:
implement ->set_rx_mode()") cleaned up mv643xx_eth's multicast filter
programming, but broke it as well.

The non-special multicast filter table (for multicast addresses that
are not of the form 01:00:5e:00:00:xx) consists of 256 hash table
buckets organised as 64 32-bit words, where the 'accept' bits are
in the LSB of each byte, so in bits 24 16 8 0 of each 32-bit word.
The old code got this right, but the referenced commit broke this by
using bits 3 2 1 0 instead.  This commit fixes this up.

Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRevert "mv643xx_eth: use longer DMA bursts".
Lennert Buytenhek [Tue, 20 Jan 2009 01:16:00 +0000 (17:16 -0800)]
Revert "mv643xx_eth: use longer DMA bursts".

This reverts commit cd4ccf76bfd2c36d351e68be7e6a597268f98a1a.

On the Pegasos board, we can't do DMA burst that are longer than
one cache line.  For now, go back to using 32 byte DMA bursts for
all mv643xx_eth platforms -- we can switch the ARM-based platforms
back to doing long 128 byte bursts in the next development cycle.

Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Reported-by: Alan Curry <pacman@kosh.dhis.org>
Reported-by: Gabriel Paubert <paubert@iram.es>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Fix data corruption when splicing from sockets.
Jarek Poplawski [Tue, 20 Jan 2009 01:03:56 +0000 (17:03 -0800)]
net: Fix data corruption when splicing from sockets.

The trick in socket splicing where we try to convert the skb->data
into a page based reference using virt_to_page() does not work so
well.

The idea is to pass the virt_to_page() reference via the pipe
buffer, and refcount the buffer using a SKB reference.

But if we are splicing from a socket to a socket (via sendpage)
this doesn't work.

The from side processing will grab the page (and SKB) references.
The sendpage() calls will grab page references only, return, and
then the from side processing completes and drops the SKB ref.

The page based reference to skb->data is not enough to keep the
kmalloc() buffer backing it from being reused.  Yet, that is
all that the socket send side has at this point.

This leads to data corruption if the skb->data buffer is reused
by SLAB before the send side socket actually gets the TX packet
out to the device.

The fix employed here is to simply allocate a page and copy the
skb->data bytes into that page.

This will hurt performance, but there is no clear way to fix this
properly without a copy at the present time, and it is important
to get rid of the data corruption.

With fixes from Herbert Xu.

Tested-by: Willy Tarreau <w@1wt.eu>
Foreseen-by: Changli Gao <xiaosuo@gmail.com>
Diagnosed-by: Willy Tarreau <w@1wt.eu>
Reported-by: Willy Tarreau <w@1wt.eu>
Fixed-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agotg3: Fix firmware loading
Matt Carlson [Tue, 20 Jan 2009 00:57:45 +0000 (16:57 -0800)]
tg3: Fix firmware loading

This patch modifies how the tg3 driver handles device firmware.

The patch starts by consolidating David Woodhouse's earlier patch under
the same name.  Specifically, the patch moves the request_firmware call
into a separate tg3_request_firmware() function and calls that function
from tg3_open() rather than tg3_init_one().

The patch then goes on to limit the number of devices that will make
request_firmware calls.  The original firmware patch unnecessarily
requested TSO firmware for devices that did not need it.  This patch
reduces the set of devices making TSO firmware patches to approximately
the following device set : 5703, 5704, and 5705.

Finally, the patch reduces the effects of a request_firmware() failure.
For those devices that are requesting TSO firmware, the driver will turn
off the TSO capability.  If TSO firmware becomes available at a later
time, the device can be closed and then opened again to reacquire the
TSO capability.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Fix usage of netif_*_all_queues() with netif_carrier_{off|on}()
Peter P Waskiewicz Jr [Tue, 20 Jan 2009 00:55:03 +0000 (16:55 -0800)]
ixgbe: Fix usage of netif_*_all_queues() with netif_carrier_{off|on}()

netif_carrier_off() is sufficient to stop Tx into the driver.  Stopping the Tx
queues is redundant and unnecessary.  By the same token, netif_carrier_on()
will be sufficient to re-enable Tx, so waking the queues is unnecessary.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: fix tag stripping for VLAN ID 0
Don Skidmore [Tue, 20 Jan 2009 00:54:36 +0000 (16:54 -0800)]
ixgbe: fix tag stripping for VLAN ID 0

Register VLAN ID 0 so that frames with VLAN ID 0 are received and get
their tag stripped when ixgbe is not in DCB mode.  VLAN ID 0 means
that the frame is 'priority tagged' only - it is not a VLAN, but the
priority value is the tag in valid.  The functions
ixgbe_vlan_rx_register() and ixgbe_vlan_rx_kill_vid() were moved up a
couple functions to correct compiling issues with this change.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Eric W Multanen <eric.w.multanen@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: fix dca issue with relaxed ordering turned on
Don Skidmore [Tue, 20 Jan 2009 00:54:13 +0000 (16:54 -0800)]
ixgbe: fix dca issue with relaxed ordering turned on

The is an issue where setting Relaxed Ordering (RO) bit
(in a PCI-E write transaction) on 82598 causing the chipset
to drop DCA hints.  This patch forces RO not to be set for
descriptors as well as payload.  This will only be in effect
while DCA is enabled and no performance difference was
noticed in testing.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Add debug info to track down GSO checksum bug
Herbert Xu [Tue, 20 Jan 2009 00:26:44 +0000 (16:26 -0800)]
net: Add debug info to track down GSO checksum bug

I'm trying to track down why people're hitting the checksum warning
in skb_gso_segment.  As the problem seems to be hitting lots of
people and I can't reproduce it or locate the bug, here is a patch
to print out more details which hopefully should help us to track
this down.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agokorina: drop leftover assignment
Phil Sutter [Thu, 15 Jan 2009 12:29:57 +0000 (12:29 +0000)]
korina: drop leftover assignment

As the assigned value is being overwritten shortly after, it can be
dropped and so the whole variable definition moved to the start of the
function.

Signed-off-by: Phil Sutter <n0-1@freewrt.org>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agokorina: adjust headroom for new skb's also
Phil Sutter [Thu, 15 Jan 2009 12:29:56 +0000 (12:29 +0000)]
korina: adjust headroom for new skb's also

This is copy and paste from the original driver. As skb_reserve() is
also called within korina_alloc_ring() when initially allocating the
receive descriptors, the same should be done when allocating new space
after passing an skb to upper layers.

Signed-off-by: Phil Sutter <n0-1@freewrt.org>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agokorina: fix loop back of receive descriptors
Phil Sutter [Thu, 15 Jan 2009 12:29:55 +0000 (12:29 +0000)]
korina: fix loop back of receive descriptors

After the last loop iteration, i has the value RC32434_NUM_RDS and
therefore leads to an index overflow when used afterwards to address the
last element. This is yet another another bug introduced when rewriting
parts of the driver for upstream preparation, as the original driver
used 'RC32434_NUM_RDS - 1' instead.

Signed-off-by: Phil Sutter <n0-1@freewrt.org>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agocxgb3: Fix LRO misalignment
Divy Le Ray [Mon, 19 Jan 2009 06:01:32 +0000 (22:01 -0800)]
cxgb3: Fix LRO misalignment

The lro manager's frag_align_pad setting was missing,
leading to misaligned access to the skb passed up
to the stack.

Tested-by: Rick Jones <rick.jones2@hp.com>
Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomacb: avoid lockup when TGO during underrun
Richard Röjfors [Mon, 19 Jan 2009 05:57:35 +0000 (21:57 -0800)]
macb: avoid lockup when TGO during underrun

In rare cases when an underrun occur, all macb buffers where consumed
and the netif_queue was stopped infinitely. This happens then the TGO
(transfer ongoing) bit in the TSR is set (and UND). It seems like
clening up after the underrun makes the driver and the macb hardware
end up in an inconsistent state. The result of this is that in the
following calls to macb_tx no TX buffers are released -> the
netif_queue was stopped, and never woken up again.

The solution is to disable the transmitter, if TGO is set, before
clening up after the underrun, and re-enable the transmitter when the
cleaning up is done.

Signed-off-by: Richard Röjfors <richard.rojfors@endian.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet/9p: fid->fid is used uninitialized
Roel Kluin [Mon, 19 Jan 2009 05:32:11 +0000 (21:32 -0800)]
net/9p: fid->fid is used uninitialized

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Version update
Eilon Greenstein [Sun, 18 Jan 2009 07:26:13 +0000 (23:26 -0800)]
bnx2x: Version update

Updating the version and the year of updated files

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Barriers for the compiler
Eilon Greenstein [Wed, 14 Jan 2009 06:42:44 +0000 (06:42 +0000)]
bnx2x: Barriers for the compiler

To make sure no swapping are made by the compiler, changed HAS_WORK to inline
functions and added all the necessary barriers

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Driver description update
Eilon Greenstein [Wed, 14 Jan 2009 06:44:28 +0000 (06:44 +0000)]
bnx2x: Driver description update

The Driver supports the 57711 and 57711E as well but the description was out of
date

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Handling PHY FW load failure
Eilon Greenstein [Wed, 14 Jan 2009 06:44:24 +0000 (06:44 +0000)]
bnx2x: Handling PHY FW load failure

If the default PHY version (0x4321) is read - the PHY FW load failed

Signed-off-by: Yaniv Rosner <yanivr@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Legacy speeds autoneg failures
Eilon Greenstein [Wed, 14 Jan 2009 06:44:21 +0000 (06:44 +0000)]
bnx2x: Legacy speeds autoneg failures

10M/100M autoneg was not establishing link.

Signed-off-by: Yaniv Rosner <yanivr@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Prevent self test loopback failures
Eilon Greenstein [Wed, 14 Jan 2009 06:44:16 +0000 (06:44 +0000)]
bnx2x: Prevent self test loopback failures

Setting loopback requires time to take effect

Signed-off-by: Yaniv Rosner <yanivr@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: 1G-10G toggling race
Eilon Greenstein [Wed, 14 Jan 2009 06:44:13 +0000 (06:44 +0000)]
bnx2x: 1G-10G toggling race

The HW should be configured so fast toggling between 1G and 10G will not be
missed. Make sure that the HW is re-configured in full

Signed-off-by: Yaniv Rosner <yanivr@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Overstepping array bounds
Eilon Greenstein [Wed, 14 Jan 2009 06:44:10 +0000 (06:44 +0000)]
bnx2x: Overstepping array bounds

If the page size is > 8KB this violation happens

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Block nvram access when the device is inactive
Eilon Greenstein [Wed, 14 Jan 2009 06:44:07 +0000 (06:44 +0000)]
bnx2x: Block nvram access when the device is inactive

Don't dump eeprom when bnx2x adapter is down.  Running ethtool -e causes an eeh
without it when the device is down

Signed-off-by: Paul Larson <pl@linux.vnet.ibm.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Read chip ID
Eilon Greenstein [Wed, 14 Jan 2009 06:44:04 +0000 (06:44 +0000)]
bnx2x: Read chip ID

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Wrong HDR offset in CAM
Eilon Greenstein [Wed, 14 Jan 2009 06:43:59 +0000 (06:43 +0000)]
bnx2x: Wrong HDR offset in CAM

Has a negative side effect when sending MAC update with no content (as done in
the self-test)

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Potential race after iSCSI boot
Eilon Greenstein [Wed, 14 Jan 2009 06:43:56 +0000 (06:43 +0000)]
bnx2x: Potential race after iSCSI boot

The lock was release too soon. Make sure the HW is marked as locked until the
boot driver was unloaded from FW perspective

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Handling probe failures
Eilon Greenstein [Wed, 14 Jan 2009 06:43:52 +0000 (06:43 +0000)]
bnx2x: Handling probe failures

Failures in the probe not handled correctly - separate the flow to handle

different failures

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: Free IRQ
Eilon Greenstein [Wed, 14 Jan 2009 06:43:48 +0000 (06:43 +0000)]
bnx2x: Free IRQ

Error check could result with not freeing the IRQ

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Sat, 17 Jan 2009 22:28:53 +0000 (14:28 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  IB/ehca: Use consistent types for ehca_plpar_hcall9()
  IB/ehca: Fix printk format warnings from u64 type change
  IPoIB: Do not print error messages for multicast join retries
  IB/mlx4: Fix memory ordering problem when posting LSO sends
  mlx4_core: Fix min() warning
  IPoIB: Fix deadlock between ipoib_open() and child interface create
  IPoIB: Fix hang in napi_disable() if P_Key is never found

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/lrg/voltage-2.6
Linus Torvalds [Sat, 17 Jan 2009 22:28:23 +0000 (14:28 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/lrg/voltage-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/lrg/voltage-2.6:
  move wm8400-regulator's probe function to .devinit.text

15 years agodell_rbu: use scnprintf() instead of less secure sprintf()
Pavel Roskin [Sat, 17 Jan 2009 18:33:03 +0000 (13:33 -0500)]
dell_rbu: use scnprintf() instead of less secure sprintf()

Reading 0 bytes from /sys/devices/platform/dell_rbu/image_type or
/sys/devices/platform/dell_rbu/packet_size by an ordinary user causes an
oops.

Signed-off-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agosecurity: introduce missing kfree
Vegard Nossum [Sat, 17 Jan 2009 16:45:45 +0000 (17:45 +0100)]
security: introduce missing kfree

Plug this leak.

Acked-by: David Howells <dhowells@redhat.com>
Cc: James Morris <jmorris@namei.org>
Cc: <stable@kernel.org>
Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomove wm8400-regulator's probe function to .devinit.text
Uwe Kleine-König [Mon, 12 Jan 2009 22:25:05 +0000 (23:25 +0100)]
move wm8400-regulator's probe function to .devinit.text

A pointer to wm8400_regulator_probe is passed to the core via
platform_driver_register and so the function must not disappear when the
.init sections are discarded.  Otherwise (if also having HOTPLUG=y)
unbinding and binding a device to the driver via sysfs will result in an
oops as does a device being registered late.

An alternative to this patch is using platform_driver_probe instead of
platform_driver_register plus removing the pointer to the probe function
from the struct platform_driver.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Liam Girdwood <lrg@slimlogic.co.uk>
15 years agoMerge branches 'ehca', 'ipoib' and 'mlx4' into for-linus
Roland Dreier [Fri, 16 Jan 2009 23:05:54 +0000 (15:05 -0800)]
Merge branches 'ehca', 'ipoib' and 'mlx4' into for-linus

15 years agoIB/ehca: Use consistent types for ehca_plpar_hcall9()
Stephen Rothwell [Fri, 16 Jan 2009 22:55:40 +0000 (14:55 -0800)]
IB/ehca: Use consistent types for ehca_plpar_hcall9()

ehca_plpar_hcall9() takes an unsigned long array, so make all callers
pass that in.  This fixes warnings introduced by commit fe333321
("powerpc: Change u64/s64 to a long long integer type"), which changed
u64 from unsigned long to unsigned long long.

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agoIB/ehca: Fix printk format warnings from u64 type change
Stephen Rothwell [Fri, 16 Jan 2009 22:55:28 +0000 (14:55 -0800)]
IB/ehca: Fix printk format warnings from u64 type change

Commit fe333321 ("powerpc: Change u64/s64 to a long long integer
type") changed u64 from unsigned long to unsigned long long, which
means that printk formats for printing u64 values should use "ll"
instead of "l" to avoid warnings.  Fix all the places affected by this
in ehca.

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agop54: fix p54_set_key's return code
Christian Lamparter [Fri, 16 Jan 2009 18:46:28 +0000 (19:46 +0100)]
p54: fix p54_set_key's return code

p54 doesn't support AES-128-CMAC offload.

This patch will fix the noisy mac80211 warnings, when 802.11w is enabled:
mac80211-phy189: failed to set key (4, ff:ff:ff:ff:ff:ff) to hardware (-22)
mac80211-phy189: failed to set key (5, ff:ff:ff:ff:ff:ff) to hardware (-22)

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: set_tim must be atomic.
Artur Skawina [Thu, 15 Jan 2009 20:07:03 +0000 (21:07 +0100)]
p54: set_tim must be atomic.

Fix for:

BUG: scheduling while atomic: named/2004/0x10000200
Pid: 2004, comm: named Not tainted 2.6.29-rc1-00271-ge9fa6b0 #45
Call Trace:
 [<c04d4ef7>] schedule+0x2a7/0x320
 [<c03aed74>] __alloc_skb+0x34/0x110
 [<c011f5b3>] __cond_resched+0x13/0x30
 [<c04d501d>] _cond_resched+0x2d/0x40
 [<c016d8c5>] kmem_cache_alloc+0x95/0xc0
 [<c016b8d4>] check_object+0xc4/0x230
 [<c03aed74>] __alloc_skb+0x34/0x110
 [<c02ede91>] p54_alloc_skb+0x71/0xf0
 [<c02ee36f>] p54_set_tim+0x3f/0xa0
 [<c04ae064>] sta_info_set_tim_bit+0x64/0x80
 [<c04c1017>] invoke_tx_handlers+0xd57/0xd80
 [<c016c397>] free_debug_processing+0x197/0x210
 [<c03ae215>] pskb_expand_head+0xf5/0x170
 [<c04bfd94>] __ieee80211_tx_prepare+0x164/0x2f0
 [<c04c1a8d>] ieee80211_skb_resize+0x6d/0xe0
 [<c04c250f>] ieee80211_master_start_xmit+0x23f/0x550
 [<c016d188>] __slab_alloc+0x2b8/0x4f0
 [<c013a711>] getnstimeofday+0x51/0x120
 [<c03b5e7b>] dev_hard_start_xmit+0x1db/0x240
 [<c03c6a4b>] __qdisc_run+0x1ab/0x200
 [<c0136aa1>] __run_hrtimer+0x31/0xf0
 [<c03b6247>] dev_queue_xmit+0x247/0x500
 [<c04c1e56>] ieee80211_subif_start_xmit+0x356/0x7d0
 [<c0466ff7>] packet_rcv_spkt+0x37/0x150
 [<c0466ff7>] packet_rcv_spkt+0x37/0x150
 [<c03b5e7b>] dev_hard_start_xmit+0x1db/0x240
 [<c03c6a4b>] __qdisc_run+0x1ab/0x200
 [<c03b6247>] dev_queue_xmit+0x247/0x500
 [<c03bc1e2>] neigh_resolve_output+0xe2/0x200
 [<c0410080>] ip_finish_output+0x0/0x290
 [<c0410267>] ip_finish_output+0x1e7/0x290
 [<c040f355>] ip_local_out+0x15/0x20
 [<c040f5d2>] ip_push_pending_frames+0x272/0x380
 [<c042bbc6>] udp_push_pending_frames+0x146/0x3a0
 [<c042d52a>] udp_sendmsg+0x2fa/0x6b0
 [<c0433bc7>] inet_sendmsg+0x37/0x70
 [<c03a7b7e>] sock_sendmsg+0xbe/0x100
 [<c0133cd0>] autoremove_wake_function+0x0/0x50
 [<c011c043>] __wake_up_common+0x43/0x70
 [<c024a892>] copy_from_user+0x32/0x130
 [<c024a892>] copy_from_user+0x32/0x130
 [<c03b001e>] verify_iovec+0x2e/0xb0
 [<c03a7d3f>] sys_sendmsg+0x17f/0x290
 [<c017730a>] pipe_write+0x29a/0x570
 [<c013a172>] update_wall_time+0x492/0x8e0
 [<c013a711>] getnstimeofday+0x51/0x120
 [<c011b05d>] sched_slice+0x3d/0x80
 [<c013a711>] getnstimeofday+0x51/0x120
 [<c0136657>] hrtimer_forward+0x147/0x1a0
 [<c01101b0>] lapic_next_event+0x10/0x20
 [<c013ccb3>] clockevents_program_event+0xa3/0x170
 [<c03a9054>] sys_socketcall+0xa4/0x290
 [<c0110920>] smp_apic_timer_interrupt+0x40/0x70
 [<c0103165>] sysenter_do_call+0x12/0x25

Signed-off-by: Artur Skawina <art.08.09@gmail.com>
Acked-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: fix a wrong parameter for __test_and_clear_bit() in rt2x00rfkill_free().
Rami Rosen [Tue, 13 Jan 2009 22:00:13 +0000 (00:00 +0200)]
rt2x00: fix a wrong parameter for __test_and_clear_bit() in rt2x00rfkill_free().

When running modprobe rt73usb, and then rmmod rt73usb, and then
iwconfig, the wlan0 device does not disappear. When repeating this
process again, we get a kernel Oops errors and "BUG: unable to handle
kernel paging request..." message in the kernel log.

The reason for this is that there is an error in rt2x00rfkill_free(),
which is called in the process of removing the device
(rt2x00lib_remove_dev() in rt2x00dev.c).
rt2x00rfkill_free() clears the RFKILL_STATE_ALLOCATED bit , which is
bit number 1 () in rt2x00dev->flags instead of in
rt2x00dev->rfkill_state. As a result, when checking the
DEVICE_STATE_REGISTERED_HW bit (bit number 1 in rt2x00dev->flags) in
rt2x00lib_remove_hw() it is **unset**, and we wrongly **don't** call
ieee80211_unregister_hw().

This patch corrects this: the parameter for __test_and_clear_bit() in
rt2x00rfkill_free() should be &rt2x00dev->rfkill_state and not
&rt2x00dev->flags.

Signed-off-by: Rami Rosen <ramirose@gmail.com>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agocfg80211: Fix parsed country IE info for 5 GHz
Luis R. Rodriguez [Thu, 8 Jan 2009 01:43:37 +0000 (17:43 -0800)]
cfg80211: Fix parsed country IE info for 5 GHz

The country IE number of channels on 5 GHz specifies the number
of 5 GHz channels, not the number of sequential channel numbers.
For example, if in a country IEs if the first channel given is 36
and the number of channels passed is 4 then the individual channel
numbers defined for the 5 GHz PHY by these parameters

are: 36, 40, 44, 48
not: 36, 37, 38, 39

See: http://tinyurl.com/11d-clarification

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agocfg80211: Fix regression with 11d on bands
Luis R. Rodriguez [Thu, 8 Jan 2009 01:43:36 +0000 (17:43 -0800)]
cfg80211: Fix regression with 11d on bands

This fixes a regression on disallowing bands introduced with the new
802.11d support. The issue is that IEEE-802.11 allows APs to send
a subset of what a country regulatory domain defines. This was clarified
in this document:

http://tinyurl.com/11d-clarification

As such it is possible, and this is what is done in practice, that a
single band 2.4 GHz AP will only send 2.4 GHz band regulatory information
through the 802.11 country information element and then the current
intersection with what CRDA provided yields a regulatory domain with
no 5 GHz information -- even though that country may actually allow
5 GHz operation. We correct this by only applying the intersection rules
on a channel if the the intersection yields a regulatory rule on the
same band the channel is on.

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agocfg80211: make handle_band() and handle_channel() wiphy specific
Luis R. Rodriguez [Thu, 8 Jan 2009 01:43:33 +0000 (17:43 -0800)]
cfg80211: make handle_band() and handle_channel() wiphy specific

This allows us to make more wiphy specific judgements when
handling the channels later on.

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: remove CMD_WANT_SKB flag if send_cmd_sync failure
Zhu Yi [Thu, 8 Jan 2009 18:19:58 +0000 (10:19 -0800)]
iwlwifi: remove CMD_WANT_SKB flag if send_cmd_sync failure

In function iwl_send_cmd_sync(), if the flag CMD_WANT_SKB is set but
we are not provided with a valid SKB (cmd->meta.u.skb == NULL), we need
to remove the CMD_WANT_SKB flag from the TX cmd queue. Otherwise in case
the cmd comes in later, it will possibly set an invalid address. Thus
it causes an invalid memory access.

This fixed the bug http://bugzilla.kernel.org/show_bug.cgi?id=11326.

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Fix an operator typo in REG_DOMAIN_2GHZ_MASK
Jouni Malinen [Tue, 13 Jan 2009 12:35:08 +0000 (14:35 +0200)]
ath9k: Fix an operator typo in REG_DOMAIN_2GHZ_MASK

Incorrect operator causes the REG_DOMAIN_2GHZ_MASK to be zero which
surely was not the goal of this definition. Mask out the 11a flags
correctly.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Fix an operator typo in phy rate validation
Jouni Malinen [Tue, 13 Jan 2009 12:33:49 +0000 (14:33 +0200)]
ath9k: Fix an operator typo in phy rate validation

This was not supposed to be a bitwise AND operation, but a check of
two separate conditions. Anyway, the old code happened to result in
the same behavior, so this is just changing the code to be easier to
understand and also to keep sparse from warning about dubious
operators.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agolibertas: Fix alignment issues in libertas core
Andrey Yurovsky [Mon, 12 Jan 2009 21:14:27 +0000 (13:14 -0800)]
libertas: Fix alignment issues in libertas core

Data structures that come over the wire from the WLAN firmware must be packed.
This fixes alignment problems on the blackfin architecture and, reportedly, on
the AVR32.

This is a replacement for the previous version of this patch which had also
explicitly used get_unaligned_ macros.  As Johannes Berg pointed out, these
macros were unnecessary.

Signed-off-by: Andrey Yurovsky <andrey@cozybit.com>
Signed-off-by: Colin McCabe <colin@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: more kernel-doc fixes
Randy Dunlap [Fri, 9 Jan 2009 23:25:09 +0000 (15:25 -0800)]
mac80211: more kernel-doc fixes

Fix (delete) more mac80211 kernel-doc:

Warning(linux-2.6.28-git13//include/net/mac80211.h:375): Excess struct/union/enum/typedef member 'retry_count' description in 'ieee80211_tx_info'
Warning(linux-2.6.28-git13//net/mac80211/sta_info.h:308): Excess struct/union/enum/typedef member 'last_txrate' description in 'sta_info'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: fix oops caused by bad eeproms
Christian Lamparter [Fri, 9 Jan 2009 20:06:30 +0000 (21:06 +0100)]
p54: fix oops caused by bad eeproms

This patch fixes a bug that could occur, if it the eeprom is incomplete or partly corrupted.

BUG: unable to handle kernel NULL pointer dereference at 00000008
IP: p54_assign_address+0x108/0x15d [p54common]
Oops: 0002 [#1] SMP
Pid: 12988, comm: phy1 Tainted: P        W  2.6.28-rc6-wl #3
RIP: 0010: p54_assign_address+0x108/0x15d [p54common]
 [...]
Call Trace:
 p54_alloc_skb+0xa3/0xc0 [p54common]
 p54_scan+0x37/0x204 [p54common]
 [...]

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoIPoIB: Do not print error messages for multicast join retries
Yossi Etigin [Fri, 16 Jan 2009 21:42:59 +0000 (13:42 -0800)]
IPoIB: Do not print error messages for multicast join retries

When IPoIB tries to join a multicast group, and the SA module's SM
address handle is NULL (because of an SM change, etc), the join
returns with -EAGAIN status.  In that case, don't print an error
message unless multicast debugging is enabled.

Signed-off-by: Yossi Etigin <yosefe@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agoIB/mlx4: Fix memory ordering problem when posting LSO sends
Roland Dreier [Fri, 16 Jan 2009 20:47:47 +0000 (12:47 -0800)]
IB/mlx4: Fix memory ordering problem when posting LSO sends

The current work request posting code writes the LSO segment before
writing any data segments.  This leaves a window where the LSO segment
overwrites the stamping in one cacheline that the HCA prefetches
before the rest of the cacheline is filled with the correct data
segments.  When the HCA processes this work request, a local
protection error may result.

Fix this by saving the LSO header size field off and writing it only
after all data segments are written.  This fix is a cleaned-up version
of a patch from Jack Morgenstein <jackm@dev.mellanox.co.il>.

This fixes <https://bugs.openfabrics.org/show_bug.cgi?id=1383>.

Reported-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agoLinux 2.6.29-rc2
Linus Torvalds [Fri, 16 Jan 2009 20:43:00 +0000 (12:43 -0800)]
Linux 2.6.29-rc2

15 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux...
Linus Torvalds [Fri, 16 Jan 2009 20:40:37 +0000 (12:40 -0800)]
Merge branch 'release' of git://git./linux/kernel/git/lenb/linux-acpi-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6: (23 commits)
  ACPI PCI hotplug: harden against panic regression
  ACPI: rename main.c to sleep.c
  dell-laptop: move to drivers/platform/x86/ from drivers/misc/
  eeepc-laptop: enable Bluetooth ACPI details
  ACPI: fix ACPI_FADT_S4_RTC_WAKE comment
  kprobes: check CONFIG_FREEZER instead of CONFIG_PM
  PM: Fix freezer compilation if PM_SLEEP is unset
  thermal fixup for broken BIOS which has invalid trip points.
  ACPI: EC: Don't trust ECDT tables from ASUS
  ACPI: EC: Limit workaround for ASUS notebooks even more
  ACPI: thinkpad-acpi: bump up version to 0.22
  ACPI: thinkpad-acpi: handle HKEY event 6030
  ACPI: thinkpad-acpi: clean-up fan subdriver quirk
  ACPI: thinkpad-acpi: start the event hunt season
  ACPI: thinkpad-acpi: handle HKEY thermal and battery alarms
  ACPI: thinkpad-acpi: clean up hotkey_notify()
  ACPI: thinkpad-acpi: use killable instead of interruptible mutexes
  ACPI: thinkpad-acpi: add UWB radio support
  ACPI: thinkpad-acpi: preserve radio state across shutdown
  ACPI: thinkpad-acpi: resume with radios disabled
  ...

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Fri, 16 Jan 2009 20:40:11 +0000 (12:40 -0800)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  netxen: include ipv6.h (fixes build failure)
  netxen: avoid invalid iounmap

15 years agoACPI PCI hotplug: harden against panic regression
James Bottomley [Thu, 15 Jan 2009 20:12:27 +0000 (15:12 -0500)]
ACPI PCI hotplug: harden against panic regression

ACPI hotplug panic with current git head
http://lkml.org/lkml/2009/1/10/136

Rather than reverting the entire commit that causes the crash:
e8c331e963c58b83db24b7d0e39e8c07f687dbc6
"PCI hotplug: introduce functions for ACPI slot detection"

simply harden against it while the changes to
the hotplug code on this particularl machine are understood.

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Len Brown <len.brown@intel.com>
15 years agoMerge branch 'misc' into release
Len Brown [Fri, 16 Jan 2009 19:45:34 +0000 (14:45 -0500)]
Merge branch 'misc' into release

15 years agoMerge branch 'thinkpad-acpi' into release
Len Brown [Fri, 16 Jan 2009 19:45:24 +0000 (14:45 -0500)]
Merge branch 'thinkpad-acpi' into release