openwrt/staging/blogic.git
15 years agonet: PHYLIB mdio fixes #2
Krzysztof Halasa [Fri, 26 Dec 2008 00:50:41 +0000 (16:50 -0800)]
net: PHYLIB mdio fixes #2

The PHYLIB mdio code has more problems in error paths:
- mdiobus_release can be called before bus->state is set to
  MDIOBUS_REGISTERED
- mdiobus_scan allocates resources which need to be freed
- the comment is wrong, the resistors used are actually pull-ups.

Signed-off-by: Krzysztof Halasa <khc@pm.waw.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agovlan: fix convertion to net_device_ops
Eric Dumazet [Fri, 26 Dec 2008 00:45:19 +0000 (16:45 -0800)]
vlan: fix convertion to net_device_ops

commit 656299f706e52e0409733d704c2761f1b12d6954
(vlan: convert to net_device_ops) added a net_device_ops
with a NULL ndo_start_xmit field.

This gives a crash in dev_hard_start_xmit()

Fix it using two net_device_ops structures, one for hwaccel vlan,
one for non hwaccel vlan.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosmsc911x: fix platform resource call during module unload
Steve Glendinning [Fri, 26 Dec 2008 00:44:01 +0000 (16:44 -0800)]
smsc911x: fix platform resource call during module unload

This patch fixes a typo, the platform_get_resource calls in
smsc911x_drv_remove are supposed to look the same as those in
smsc911x_drv_probe.

Reported and fixed by dfoley@telus.net.

Signed-off-by: Steve Glendinning <steve.glendinning@smsc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosmsc911x: don't clobber driver_data
Steve Glendinning [Fri, 26 Dec 2008 00:43:34 +0000 (16:43 -0800)]
smsc911x: don't clobber driver_data

smsc911x uses driver_data to store our net_device, don't overwrite this
with the mii_bus.

Reported and fixed by dfoley@telus.net.

Signed-off-by: Steve Glendinning <steve.glendinning@smsc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetns: igmp: make /proc/net/{igmp,mcfilter} per netns
Alexey Dobriyan [Fri, 26 Dec 2008 00:42:51 +0000 (16:42 -0800)]
netns: igmp: make /proc/net/{igmp,mcfilter} per netns

This patch makes the followinf proc entries per-netns:
/proc/net/igmp
/proc/net/mcfilter

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: Daniel Lezcano <dlezcano@fr.ibm.com>
Acked-by: Benjamin Thery <benjamin.thery@bull.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetns: igmp: allow IPPROTO_IGMP sockets in netns
Alexey Dobriyan [Fri, 26 Dec 2008 00:42:23 +0000 (16:42 -0800)]
netns: igmp: allow IPPROTO_IGMP sockets in netns

Looks like everything is already ready.

Required for ebtables(8) for one thing.

Also, required for ipmr per-netns (coming soon). (Benjamin)

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: Benjamin Thery <benjamin.thery@bull.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosmsc911x: make smsc911x_ethtool_ops constant
Steve Glendinning [Fri, 26 Dec 2008 00:41:09 +0000 (16:41 -0800)]
smsc911x: make smsc911x_ethtool_ops constant

Signed-off-by: Steve Glendinning <steve.glendinning@smsc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosmsc911x: convert driver to use net_device_ops
Steve Glendinning [Fri, 26 Dec 2008 00:40:47 +0000 (16:40 -0800)]
smsc911x: convert driver to use net_device_ops

Signed-off-by: Steve Glendinning <steve.glendinning@smsc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosmsc911x: fix BUG if module is removed while interface is up
Steve Glendinning [Fri, 26 Dec 2008 00:40:19 +0000 (16:40 -0800)]
smsc911x: fix BUG if module is removed while interface is up

If the module is removed while its interface is up, smsc911x_stop is not
called until after smsc911x_drv_remove has disposed of phy_dev.  This
patch changes the stop method to handle this situation.

This is a different problem to the one reported by dfoley@telus.net.

Signed-off-by: Steve Glendinning <steve.glendinning@smsc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoMerge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
David S. Miller [Wed, 24 Dec 2008 01:56:23 +0000 (17:56 -0800)]
Merge branch 'master' of /linux/kernel/git/davem/net-2.6

Conflicts:

drivers/net/ppp_generic.c

15 years agonet: Remove unused netdev arg from some NAPI interfaces.
Neil Horman [Tue, 23 Dec 2008 04:43:12 +0000 (20:43 -0800)]
net: Remove unused netdev arg from some NAPI interfaces.

When the napi api was changed to separate its 1:1 binding to the net_device
struct, the netif_rx_[prep|schedule|complete] api failed to remove the now
vestigual net_device structure parameter.  This patch cleans up that api by
properly removing it..

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: startup race in hso driver
Oliver Neukum [Thu, 18 Dec 2008 03:57:35 +0000 (03:57 +0000)]
net: startup race in hso driver

The flag marking a device running must be set before the URBs for
recption are submitted or they may complete too early and fail to resubmit.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agopkt_sched: Remove smp_wmb() in qdisc_watchdog()
Jarek Poplawski [Tue, 23 Dec 2008 03:44:13 +0000 (19:44 -0800)]
pkt_sched: Remove smp_wmb() in qdisc_watchdog()

While implementing a TCQ_F_THROTTLED flag there was used an smp_wmb()
in qdisc_watchdog(), but since this flag is practically used only in
sch_netem(), and since it's not even clear what reordering is avoided
here (TCQ_F_THROTTLED vs. __QDISC_STATE_SCHED?) it seems the barrier
could be safely removed.

Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agor6040: bump release number to 0.19
Florian Fainelli [Tue, 23 Dec 2008 03:40:38 +0000 (19:40 -0800)]
r6040: bump release number to 0.19

This patch bumps the release number of the driver.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agor6040: save and restore MIER correctly in the interrupt routine
Joe Chou [Tue, 23 Dec 2008 03:40:02 +0000 (19:40 -0800)]
r6040: save and restore MIER correctly in the interrupt routine

This patch saves the MIER register contents before treating
interrupts, then restores them correcty at the end of the
interrupt routine.

Signed-off-by: Joe Chou <Joe.Chou@rdc.com.tw>
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agor6040: fix wrong logic in mdio code
Joe Chou [Tue, 23 Dec 2008 03:38:17 +0000 (19:38 -0800)]
r6040: fix wrong logic in mdio code

This patch fixes a reverse logic in the MDIO code.

Signed-off-by: Joe Chou <Joe.Chou@rdc.com.tw>
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Fix oops in dev_ifsioc()
Jarek Poplawski [Tue, 23 Dec 2008 03:35:28 +0000 (19:35 -0800)]
net: Fix oops in dev_ifsioc()

A command like this: "brctl addif br1 eth1" issued as a user gave me
an oops when bridge module wasn't loaded. It's caused by using a dev
pointer before checking for NULL.

Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agotg3: Update version to 3.97
Matt Carlson [Mon, 22 Dec 2008 04:21:52 +0000 (20:21 -0800)]
tg3: Update version to 3.97

This patch updates the version number to 3.97.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agotg3: tg3.h cleanups
Matt Carlson [Mon, 22 Dec 2008 04:21:18 +0000 (20:21 -0800)]
tg3: tg3.h cleanups

This patch cleans up the tg3 header file by removing the preprocessor
definitions for standard PCI configuration space registers.  The driver
should be using the standard definitions when needed.  The patch
continues by removing redundant PHY related definitions and reorganizes
some of the remaining entries.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agotg3: Remove unused cfgspc device members
Matt Carlson [Mon, 22 Dec 2008 04:19:57 +0000 (20:19 -0800)]
tg3: Remove unused cfgspc device members

This patch removes the pci_bist and pci_hdr_type members from the
device structure and removes the code that references them.  They are
not really used.

The patch rounds out the changes by moving the pci_cmd member to plug
a structure hole that would have been created.  On 32-bit systems, this
movement removes a subsequent structure hole later in the structure. On
64-bit systems though, the movement merely consolidates two holes into
one larger hole.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agotg3: Cleanup IPV6 LSO
Matt Carlson [Mon, 22 Dec 2008 04:19:30 +0000 (20:19 -0800)]
tg3: Cleanup IPV6 LSO

This patch attempts to make the relationship between IPV6 checksum
offload and IPV6 LSO more obvious.  The patch also toggles a bit
needed for IPV6 LSO on 5785 and 57780 devices.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agotg3: WOL fixes
Matt Carlson [Mon, 22 Dec 2008 04:18:56 +0000 (20:18 -0800)]
tg3: WOL fixes

The first hunk of this patch inverts a flag that was accidentally
toggled as part of commit 0a459aac9d151c2e36ec65723b9b845b24c5cbc3
("tg3: Allow WOL for phylib controlled Broadcom phys").

The second hunk of the patch removes the call to
device_may_wakeup() in the 5906 config detection path.  At the
point of the call, the driver shouldn't be querying for WOL
capability.  It should be detecting and setting it.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agopkt_sched: Annotate uninitialized var in sfq_enqueue()
Jarek Poplawski [Mon, 22 Dec 2008 04:14:48 +0000 (20:14 -0800)]
pkt_sched: Annotate uninitialized var in sfq_enqueue()

Some gcc versions warn that ret may be used uninitialized in
sfq_enqueue(). It's a false positive, so let's annotate this.

Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: fix the display of DCB control stats in ethtool
Don Skidmore [Mon, 22 Dec 2008 04:11:04 +0000 (20:11 -0800)]
ixgbe: fix the display of DCB control stats in ethtool

Priority flow contol statistics for Data Center Bridging (DCB) weren't
included in ethtool.  This patch adds them.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Eric W Multanen <eric.w.multanen@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 DCNA attribute to the BCN interface for DCB
Don Skidmore [Mon, 22 Dec 2008 04:10:29 +0000 (20:10 -0800)]
net: add DCNA attribute to the BCN interface for DCB

Adds the Backward Congestion Notification Address (BCNA) attribute to the
Backward Congestion Notification (BCN) interface for Data Center Bridging
(DCB), which was missing.  Receive the BCNA attribute in the ixgbe driver.
The BCNA attribute is for a switch to inform the endstation about the physical
port identification in order to support BCN on aggregated links.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Eric W Multanen <eric.w.multanen@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
15 years agonet: fix DCB setstate to return success/failure
Don Skidmore [Mon, 22 Dec 2008 04:09:50 +0000 (20:09 -0800)]
net: fix DCB setstate to return success/failure

Data Center Bridging (DCB) had no way to know if setstate had failed in the
driver.  This patch enables dcb netlink code to handle the status for the DCB
setstate interface.  Likewise it allows the driver to return a failed status
if MSI-X isn't enabled.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Eric W Multanen <eric.w.multanen@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.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 [Mon, 22 Dec 2008 03:57:10 +0000 (19:57 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next-2.6

15 years agob43: Add key memory dumping
Michael Buesch [Fri, 19 Dec 2008 19:24:30 +0000 (20:24 +0100)]
b43: Add key memory dumping

This adds an option to dump all crypto related memory to
the kernel log.
Obviously, it should not be enabled on productive systems. ;)

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agob43: Suspend MAC while killing the radio
Michael Buesch [Fri, 19 Dec 2008 17:40:00 +0000 (18:40 +0100)]
b43: Suspend MAC while killing the radio

We should suspend the MAC, before we kill the radio. This gives
the MAC a chance to leave any TX/RX state and it avoids races on
the PHY/RADIO registers.

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: correct packet length in tx descriptors
Bob Copeland [Fri, 19 Dec 2008 04:23:05 +0000 (23:23 -0500)]
ath5k: correct packet length in tx descriptors

Packet length calculation (which includes frame check sequence)
should take into account whether we add a pad field or not.
Extract the calculation into a helper and use it in both places.

Changes to desc.c
Changes-licensed-under: ISC

Changes to ath5k.h, base.c
Changes-licensed-under: 3-Clause-BSD

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: implement dynamic power save
Kalle Valo [Thu, 18 Dec 2008 21:35:27 +0000 (23:35 +0200)]
mac80211: implement dynamic power save

This patch implements dynamic power save for mac80211. Basically it
means enabling power save mode after an idle period. Implementing it
dynamically gives a good compromise of low power consumption and low
latency. Some hardware have support for this in firmware, but some
require the host to do it.

The dynamic power save is implemented by adding an timeout to
ieee80211_subif_start_xmit(). The timeout can be enabled from userspace
with Wireless Extensions. For example, the command below enables the
dynamic power save and sets the time timeout to 500 ms:

iwconfig wlan0 power timeout 500m

Power save now only works with devices which handle power save in firmware.
It's also disabled by default and the heuristics when and how to enable is
considered as a policy decision and will be left for the userspace to handle.
In case the firmware has support for this, drivers can disable this feature
with IEEE80211_HW_NO_STACK_DYNAMIC_PS.

Big thanks to Johannes Berg for the help with the design and code.

Signed-off-by: Kalle Valo <kalle.valo@nokia.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: track master queue status
Kalle Valo [Thu, 18 Dec 2008 21:35:20 +0000 (23:35 +0200)]
mac80211: track master queue status

This is a preparation for the dynamic power save support. In future there are
two paths to stop the master queues and we need to track this properly to
avoid starting queues incorrectly. Implement this by adding a status
array for each queue.

The original idea and design is from Johannes Berg, I just did
the implementation based on his notes. All the bugs are mine, of course.

Signed-off-by: Kalle Valo <kalle.valo@nokia.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: enable IEEE80211_CONF_PS only when associated
Kalle Valo [Thu, 18 Dec 2008 21:35:13 +0000 (23:35 +0200)]
mac80211: enable IEEE80211_CONF_PS only when associated

Also disable power save when disassociated. It makes no sense to have
power save enabled while disassociated.

iwlwifi seems to have this check in the driver, but it's better to do this
in mac80211 instead.

Signed-off-by: Kalle Valo <kalle.valo@nokia.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agob43: Fix some MAC locking
Michael Buesch [Thu, 18 Dec 2008 21:13:39 +0000 (22:13 +0100)]
b43: Fix some MAC locking

This fixes some locking w.r.t. the lower MAC (firmware).
It also removes a lot of ancient IRQ-locking that's not needed anymore.
We simply suspend the MAC. That's easier and causes less trouble.

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: iwl-tx.c cleanup (remove unused parameter and unused local variable).
Rami Rosen [Thu, 18 Dec 2008 16:04:51 +0000 (18:04 +0200)]
iwlwifi: iwl-tx.c cleanup (remove unused parameter and unused local variable).

This patch removes unused parameter and unused local variable in
methods in iwl-tx.c:

- Remove a parameter (is_unicast) from iwl_tx_cmd_build_basic().

- Remove an unused variable name unicast from iwl_tx_skb().

Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Fixed RX decryption status reporting
Jouni Malinen [Thu, 18 Dec 2008 12:33:00 +0000 (14:33 +0200)]
ath9k: Fixed RX decryption status reporting

The RX code in ath9k uses sc_keymap to figure out whether a default
key was used. However, the default key entries in sc_keymap were
always set and as such, frames could have been claimed to be decrypted
by hardware when they were not. This can cause problems especially
with TKIP since mac80211 is validating the Michael MIC in the frame
and this will result in MIC failure and potentially TKIP
countermeasures if the frame was not decrypted correctly.

Change key cache slot allocation to mark only the keys that really
have been used in sc_keymap to avoid the issue. The key cache slot
selection routines are now internally avoiding the slots that may be
needed for TKIP group keys.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Fix key cache slot selection for splitmic
Jouni Malinen [Thu, 18 Dec 2008 12:31:10 +0000 (14:31 +0200)]
ath9k: Fix key cache slot selection for splitmic

ath_reserve_key_cache_slot() was obviously supposed to return an index
to a free slot, not reserved one. This could have caused problems with
hardware revisions that use splitmic.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Protect config() callback with a mutex
Sujith [Thu, 18 Dec 2008 06:10:16 +0000 (11:40 +0530)]
ath9k: Protect config() callback with a mutex

This should fix the timeout issues seen when using wpa_supplicant.

Signed-off-by: Sujith <Sujith.Manoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: Print unknown packet type in tasklet_handler
Larry Finger [Wed, 17 Dec 2008 20:26:59 +0000 (14:26 -0600)]
mac80211: Print unknown packet type in tasklet_handler

In stress testing p54usb, the WARN_ON() in ieee80211_tasklet_handler() was
triggered; however, there is no logging of the received value for packet
type. Adding that feature will improve the warning.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54usb: bring first generation devices back to life
Christian Lamparter [Wed, 17 Dec 2008 11:04:43 +0000 (12:04 +0100)]
p54usb: bring first generation devices back to life

This patch fixes a serious regression (introduced by:
"p54: fix memory management")
that affected isl3886+net2280 usb devices operation.

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 agoath9k: Key cache allocation for AP mode
Jouni Malinen [Wed, 17 Dec 2008 11:32:17 +0000 (13:32 +0200)]
ath9k: Key cache allocation for AP mode

Enhance allocation of key cache entries to support multiple pairwise
keys to fix AP mode with more than one associated STA.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Fix a NULL pointer dereference in ath_rate_get
Jouni Malinen [Wed, 17 Dec 2008 11:30:31 +0000 (13:30 +0200)]
ath9k: Fix a NULL pointer dereference in ath_rate_get

It looks like mac80211 may try to send unicast frames to a STA that
does not have a STA entry. We need to make sure that that is caught in
the rate control code before dereferencing STA data.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: indicate txpower is off in sysfs
Jay Sternberg [Wed, 17 Dec 2008 08:52:34 +0000 (16:52 +0800)]
iwlwifi: indicate txpower is off in sysfs

The patch checks if the radio is disabled before displaying the tx power
level. Previously when the txpower was set off show_tx_power still
returned the prior power level. Now it will indicate the power has been
turned off.

Signed-off-by: Jay Sternberg <jay.e.sternberg@linux.intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: use GFP_KERNEL to allocate Rx SKB memory
Zhu Yi [Wed, 17 Dec 2008 08:52:33 +0000 (16:52 +0800)]
iwlwifi: use GFP_KERNEL to allocate Rx SKB memory

Previously we allocate Rx SKB with GFP_ATOMIC flag. This is because we need
to hold a spinlock to protect the two rx_used and rx_free lists operation
in the rxq.

spin_lock();
...
element = rxq->rx_used.next;
element->skb = alloc_skb(..., GFP_ATOMIC);
list_del(element);
list_add_tail(&element->list, &rxq->rx_free);
...
spin_unlock();

After spliting the rx_used delete and rx_free insert into two operations,
we don't require the skb allocation in an atomic context any more (the
function itself is scheduled in a workqueue).

spin_lock();
...
element = rxq->rx_used.next;
list_del(element);
...
spin_unlock();
...
element->skb = alloc_skb(..., GFP_KERNEL);
...
spin_lock()
...
list_add_tail(&element->list, &rxq->rx_free);
...
spin_unlock();

This patch should fix the "iwlagn: Can not allocate SKB buffers" warning
we see recently.

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Tomas Winkler <tomas.winkler@intel.com>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: use meaningful vars in _iwl_poll_bit()
Wu, Fengguang [Wed, 17 Dec 2008 08:52:32 +0000 (16:52 +0800)]
iwlwifi: use meaningful vars in _iwl_poll_bit()

Rename vars in _iwl_poll_bit() to better reflect the truth.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: remove useless goto
Wu Fengguang [Wed, 17 Dec 2008 08:52:31 +0000 (16:52 +0800)]
iwlwifi: remove useless goto

The patch removes some useless goto in code cleanup.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: use type uint for module param debug
Wu, Fengguang [Wed, 17 Dec 2008 08:52:30 +0000 (16:52 +0800)]
iwlwifi: use type uint for module param debug

This enables one to change the debug level at bit 31.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: update comments on the debug interface
Wu, Fengguang [Wed, 17 Dec 2008 08:52:29 +0000 (16:52 +0800)]
iwlwifi: update comments on the debug interface

Bring up-to-date some comments on the location of debug files.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Acked-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: add line feed to printk
Wu, Fengguang [Wed, 17 Dec 2008 08:52:28 +0000 (16:52 +0800)]
iwlwifi: add line feed to printk

This adds line feed to printk.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: cleanup iwl-dev.h
Winkler, Tomas [Wed, 17 Dec 2008 08:52:27 +0000 (16:52 +0800)]
iwlwifi: cleanup iwl-dev.h

The patch removes unused definition and moves code to proper places.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: fix a typo in ieee80211_send_assoc() method.
Rami Rosen [Tue, 16 Dec 2008 20:38:29 +0000 (22:38 +0200)]
mac80211: fix a typo in ieee80211_send_assoc() method.

This patch fixes a typo in ieee80211_send_assoc(), net/mac80211/mlme.c.

The error is usage of a wrong member when building
the ie80211 management frame (it should be assoc_req, and not reassoc_req).

Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: remove a parameter (dest) from *_fill_beacon_frame() methods.
Rami Rosen [Tue, 16 Dec 2008 07:37:07 +0000 (09:37 +0200)]
iwlwifi: remove a parameter (dest) from *_fill_beacon_frame() methods.

This patch removes a parameter (dest) from iwl_fill_beacon_frame() (iwl-agn.c)
 and from iwl3945_fill_beacon_frame(). (iwl-3945.c,iwl-3945.h)

Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Move rate control alg register/unregister to appropriate place
Vasanthakumar Thiagarajan [Tue, 16 Dec 2008 07:07:38 +0000 (12:37 +0530)]
ath9k: Move rate control alg register/unregister to appropriate place

This patch makes sure the rate control alg is registered/unregistered
only once for this module.

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Synchronize DMA transfer with CPU at right place
Vasanthakumar Thiagarajan [Mon, 15 Dec 2008 15:10:46 +0000 (20:40 +0530)]
ath9k: Synchronize DMA transfer with CPU at right place

This patch does pci_dma_sync_single_for_cpu() before accessing
the header of the frame and queueing the same buffer into h/w.

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Remove MAC header pad before reporting TX status
Jouni Malinen [Mon, 15 Dec 2008 14:02:04 +0000 (16:02 +0200)]
ath9k: Remove MAC header pad before reporting TX status

Remove the possible MAC header pad before reporting TX status to
mac80211. This pad is hardware specific operation and should not be
exposed outside the driver. This fixes the frame body in monitor
interfaces that could be used to check on TX status for transmitted
frames.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Fix TX status reporting for retries and MCS index
Jouni Malinen [Mon, 15 Dec 2008 13:44:53 +0000 (15:44 +0200)]
ath9k: Fix TX status reporting for retries and MCS index

The count field in struct ieee80211_tx_rate does not include the final
successful attempt, so only report retries here.

Fix the struct ieee80211_tx_rate::idx field when MCS was used. It is
supposed to be the MCS index, not an internal index to the rate
control algorithm table.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: Remove radiotap rate-present flag for HT
Jouni Malinen [Mon, 15 Dec 2008 08:37:50 +0000 (10:37 +0200)]
mac80211: Remove radiotap rate-present flag for HT

Since we do not currently report HT rates (MCS index) in radiotap
header for HT rates, we should not claim the rate is present. The rate
octet itself is used as padding in this case, so only the it_present
flag needs to be removed in case of HT rates.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agortl8187: Fix crash on unload when using SLUB debug
Herton Ronaldo Krzesinski [Sun, 14 Dec 2008 18:18:09 +0000 (12:18 -0600)]
rtl8187: Fix crash on unload when using SLUB debug

Reported-by: Hin-Tak Leung <htl10@users.sourceforge.net>
After the code was modified to use urb anchors ("rtl8187: Use usb anchor
facilities to manage urbs"), rtl8187 began generating an intermittent GPF
on shutdown when using SLUB with debugging enabled. Furthermore, rebooting
the system with a ping running caused a GPF every time. There are two problems:
(1) incorrect locking in the rtl8187_rx_cb() routine, a pre-existing bug that
apparently had not been triggered before, and (2) duplicate freeing of receive
skbs that was probably introduced with the change to anchors.

Signed-off-by: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Tested-by: Hin-Tak Leung <htl10@users.sourceforge.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: remove free_on_tx parameter
Christian Lamparter [Sun, 14 Dec 2008 14:05:42 +0000 (15:05 +0100)]
p54: remove free_on_tx parameter

This patch simplifies the tx code a bit and will be necessary for
the upcoming stlc45xx<->p54 port.

In detail: we no longer have to tell all back-end drivers directly,
if we want to free a frame right after it was send to the firmware,
or if we do it in the library callback later.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: more accurate rssi to dBm conversion
Christian Lamparter [Sun, 14 Dec 2008 13:45:30 +0000 (14:45 +0100)]
p54: more accurate rssi to dBm conversion

This patch replaces the static rssi auto calibration data
with more precise values out of the device's eeprom.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agortl8187: Fix error returns
Larry Finger [Sun, 14 Dec 2008 00:09:56 +0000 (18:09 -0600)]
rtl8187: Fix error returns

There are two places in the rtl8187 code where a routine was returning zero
(OK) when it should have been returning an error.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Acked-by: Hin-Tak Leung <htl10@users.sourceforge.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agolibertas: use roundup instead of opencoding
Ilpo Järvinen [Sat, 13 Dec 2008 19:33:13 +0000 (21:33 +0200)]
libertas: use roundup instead of opencoding

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: update ACK failure statistic counter in real-time
Christian Lamparter [Sat, 13 Dec 2008 15:45:38 +0000 (16:45 +0100)]
p54: update ACK failure statistic counter in real-time

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: move statistic timer update routine into a workqueue
Christian Lamparter [Sat, 13 Dec 2008 13:14:20 +0000 (14:14 +0100)]
p54: move statistic timer update routine into a workqueue

This patch moves a good chunk of code from the former statistic update
timer routine into a workqueue, which is kindly provided by mac80211.

Also as a nice side-effect we can lay the foundation for other
essential housekeeping features we want to do in the future.
e.g:
 - drain the (clogged) tx_queue.
 - initiate bursts.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlagn: fix agn rate scaling
Mohamed Abbas [Fri, 12 Dec 2008 16:22:15 +0000 (08:22 -0800)]
iwlagn: fix agn rate scaling

    Rate scaling in agn can be broken because of these two problems
    1- the LQ command could not sent, we try to call LQ command before
       post association will be rejected, this will cause the driver to
       break in rs_tx_status reporting initial rate does not match.
    2- In rs_rate_init we can get a value of 12, rate 60, for
       last_txrate_idx, this rate only valid in SISO/MIMO mode only

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 agomac80211: Send Layer 2 Update frame on reassociation
Jouni Malinen [Fri, 12 Dec 2008 15:08:31 +0000 (17:08 +0200)]
mac80211: Send Layer 2 Update frame on reassociation

When a STA roams back to the same AP before the previous STA entry has
expired, a new STA entry is not added in mac80211. However, a Layer 2
Update frame still needs to be transmitted to update layer 2 devices
about the new location for the STA. Without this, switches may
continue to forward frames to the previous (now incorrect) port when
STA roams between APs.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: fix 802.11 header padding on RX, unpadding on TX
Benoit PAPILLAULT [Fri, 12 Dec 2008 14:29:58 +0000 (15:29 +0100)]
ath5k: fix 802.11 header padding on RX, unpadding on TX

Padding the 802.11 header to a multiple of 4 bytes needs to be done only for
frames with a body. This fixes a bug where 2 bytes were missing in monitor
mode for ACK frames. Inspired by a patch from Jouni Malinen on ath9k.

Ref: http://bugzilla.kernel.org/show_bug.cgi?id=12101 :
Signed-off-by: Benoit Papillault <benoit.papillault@free.fr>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Report HT rates in RX status
Jouni Malinen [Fri, 12 Dec 2008 12:38:34 +0000 (14:38 +0200)]
ath9k: Report HT rates in RX status

Fix and clean up the RX status reporting by getting rid of code that
used internal rate tables and ratekbps calculation. The correct value
is now reported with MCS index instead of the old mechanism that
defaulted to using the highest legacy rate.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: Add HT rates into RX status reporting
Jouni Malinen [Fri, 12 Dec 2008 12:38:33 +0000 (14:38 +0200)]
mac80211: Add HT rates into RX status reporting

This patch adds option for HT-enabled drivers to report HT rates
(HT20/HT40, short GI, MCS index) to mac80211. These rates are
currently not in the rate table, so the rate_idx is used to indicate
MCS index.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: Do not use association state to update rate table
Sujith [Fri, 12 Dec 2008 06:29:07 +0000 (11:59 +0530)]
ath9k: Do not use association state to update rate table

Now that HT information is made available to the driver through
config() callback, the channel type can be calculated whenever
setting the channel. Update the rate table accordingly with the new
channel type.

Signed-off-by: Sujith <Sujith.Manoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: Fix HT channel selection
Sujith [Fri, 12 Dec 2008 06:27:43 +0000 (11:57 +0530)]
mac80211: Fix HT channel selection

HT management is done differently for AP and STA modes, unify
to just the ->config() callback since HT is fundamentally a
PHY property and cannot be per-BSS.

Rename enum nl80211_sec_chan_offset as nl80211_channel_type to denote
the channel type ( NO_HT, HT20, HT40+, HT40- ).

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Sujith <Sujith.Manoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agonl80211: Add signal strength and bandwith to nl80211station info
Henning Rogge [Thu, 11 Dec 2008 21:04:19 +0000 (22:04 +0100)]
nl80211: Add signal strength and bandwith to nl80211station info

This patch adds signal strength and transmission bitrate
to the station_info of nl80211.

Signed-off-by: Henning Rogge <rogge@fgan.de>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agonet: prepare usb net drivers for addition of status as a parameter
Oliver Neukum [Fri, 19 Dec 2008 07:00:59 +0000 (23:00 -0800)]
net: prepare usb net drivers for addition of status as a parameter

USB is going to switch the signature of the callbacks to
void callback(struct urb *urb, int status)
This patch will ease the transition.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: ppp_generic - use DEFINE_IDR for static initialization
Cyrill Gorcunov [Fri, 19 Dec 2008 06:59:32 +0000 (22:59 -0800)]
net: ppp_generic - use DEFINE_IDR for static initialization

We could use DEFINE_IDR for statically allocated idr
that allow us to save a few lines of code.

And spell fix.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agocxgb3: manage private iSCSI IP address
Karen Xie [Fri, 19 Dec 2008 06:56:20 +0000 (22:56 -0800)]
cxgb3: manage private iSCSI IP address

The accelerated iSCSI traffic could use a private IP address unknown to the OS:
- The IP address is required in both drivers to manage ARP requests and connection set up.
- Added an control call to retrieve the ip address.
- Reply to ARP requests dedicated to the private IP address.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Karen Xie <kxie@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoucc_geth: Remove UGETH_FILTERING dead code
Anton Vorontsov [Thu, 18 Dec 2008 08:23:35 +0000 (08:23 +0000)]
ucc_geth: Remove UGETH_FILTERING dead code

The code appears to be dead: nobody call these functions, plus build
breaks when UGETH_FILTERING is enabled:

ucc_geth.c:1848: warning: 'struct enet_addr' declared inside parameter list
ucc_geth.c:1848: warning: its scope is only this definition or declaration, which is probably not what you want
ucc_geth.c: In function 'ugeth_82xx_filtering_get_match_addr_in_hash':
ucc_geth.c:1856: error: dereferencing pointer to incomplete type
ucc_geth.c:1874: error: dereferencing pointer to incomplete type
ucc_geth.c:1877: warning: return from incompatible pointer type
ucc_geth.c: At top level:
ucc_geth.c:1885: warning: 'struct enet_addr' declared inside parameter list
ucc_geth.c: In function 'ugeth_82xx_filtering_add_addr_in_hash':
ucc_geth.c:1894: error: dereferencing pointer to incomplete type
ucc_geth.c:1909: warning: passing argument 2 of 'ugeth_82xx_filtering_get_match_addr_in_hash' from incompatible pointer type
ucc_geth.c:1909: warning: assignment from incompatible pointer type
ucc_geth.c:1918: error: dereferencing pointer to incomplete type
ucc_geth.c: At top level:
ucc_geth.c:1928: warning: 'struct enet_addr' declared inside parameter list
ucc_geth.c: In function 'ugeth_82xx_filtering_clear_addr_in_hash':
ucc_geth.c:1947: warning: passing argument 2 of 'ugeth_82xx_filtering_get_match_addr_in_hash' from incompatible pointer type
ucc_geth.c:1947: warning: assignment from incompatible pointer type
ucc_geth.c:1954: error: dereferencing pointer to incomplete type
ucc_geth.c: At top level:
ucc_geth.c:2060: warning: 'struct enet_addr' declared inside parameter list
ucc_geth.c: In function 'ugeth_82xx_filtering_add_addr_in_paddr':
ucc_geth.c:2064: error: dereferencing pointer to incomplete type
ucc_geth.c:2073: error: dereferencing pointer to incomplete type
ucc_geth.c:2075: warning: passing argument 2 of 'hw_add_addr_in_paddr' from incompatible pointer type
make[2]: *** [ucc_geth.o] Error 1

The code is there since the driver was merged, and nobody seem to be
interested in fixing or actually using it. If we ever want the
filtering support, we can always revert the patch and fix it, but so
far it just draws reader's attention.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoucc_geth: Fix IO memory (un)mapping code
Anton Vorontsov [Thu, 18 Dec 2008 08:23:33 +0000 (08:23 +0000)]
ucc_geth: Fix IO memory (un)mapping code

The driver doesn't check ioremap() return value, and doesn't free the
remapped memory.

This patch fixes it.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoucc_geth: Cleanup repetitive ucc_geth_memclean() calls
Anton Vorontsov [Thu, 18 Dec 2008 08:23:31 +0000 (08:23 +0000)]
ucc_geth: Cleanup repetitive ucc_geth_memclean() calls

No need to call ucc_geth_memclean() so many times, just check for
errors in ucc_geth_open(), and call ucc_geth_stop() in case of errors.

The ucc_geth_stop() may be called anytime and will do the right thing.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoucc_geth: Fix IRQ freeing code in ucc_geth_open()
Anton Vorontsov [Thu, 18 Dec 2008 08:23:29 +0000 (08:23 +0000)]
ucc_geth: Fix IRQ freeing code in ucc_geth_open()

open() routine calls stop() in case of errors, the function will try
to free the requested IRQ. But we don't know if it was actually
requested, so the code might issue bogus free_irq(0, dev) call.

Fix this by rearranging the code so that now request_irq() is the last
call in the open() routine, and move free_irq() into the close().

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoucc_geth: Fix TX watchdog timeout handling
Anton Vorontsov [Thu, 18 Dec 2008 08:23:26 +0000 (08:23 +0000)]
ucc_geth: Fix TX watchdog timeout handling

The timeout handling code is currently broken in several ways:

- It calls stop() (which frees all the memory and IRQ), and then
  calls startup() (which won't re-request IRQ, neither it will
  re-init the Fast UCC structure).
- It calls these routines from the softirq context, which is wrong,
  since stop() calls free_irq() (which might sleep) and startup()
  allocates things with GFP_KERNEL.
- It won't soft-reset the PHY. We need the PHY reset for at least
  MPC8360E-MDS boards with Marvell 88E1111 PHY, the PHY won't recover
  from timeouts w/o the reset.

So the patch fixes these problems by implementing the workqueue for the
timeout handling, and there we fully re-open the device via close() and
open() calls. The close/open paths do the right things, and I can see
that the driver actually survive the timeouts.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoucc_geth: Fix endless loop in stop_{tx,rx} routines
Anton Vorontsov [Thu, 18 Dec 2008 08:23:22 +0000 (08:23 +0000)]
ucc_geth: Fix endless loop in stop_{tx,rx} routines

Currently the routines wait for the various bits w/o an assumption that
bits may never get set. When timeouts happen I see that these bits never
get set and so the routines hang the kernel.

With this patch we'll wait the graceful stop for 100 ms, and then will
simply exit. There is nothing* we can do about that, but it's OK since
we'll do full reset later.

* Well, actually, there is also not-graceful variant for the TX stop,
  but specs says that we never should use it.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agotcp: Stop scaring users with "treason uncloaked!"
Matt Mackall [Fri, 19 Dec 2008 03:54:22 +0000 (19:54 -0800)]
tcp: Stop scaring users with "treason uncloaked!"

The original message was unhelpful and extremely alarming to our poor
users, despite its charm. Make it less frightening.

Signed-off-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoNIU: Implement discard counters, optimize
Jesper Dangaard Brouer [Fri, 19 Dec 2008 03:51:56 +0000 (19:51 -0800)]
NIU: Implement discard counters, optimize

Optimize the lightly loaded case, by only synchronizing discards stats
when qlen > 10 indicate potential for drops.

Notice Robert Olsson might disagree with this patch.

Signed-off-by: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoNIU: Implement discard counters, info/debug statements.
Jesper Dangaard Brouer [Fri, 19 Dec 2008 03:51:26 +0000 (19:51 -0800)]
NIU: Implement discard counters, info/debug statements.

Discard packet counter debug statements that can be turned on
at runtime by users to assist debugging of the driver code.

Signed-off-by: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoNIU: Implement discard counters
Jesper Dangaard Brouer [Fri, 19 Dec 2008 03:50:49 +0000 (19:50 -0800)]
NIU: Implement discard counters

Implementing discard counters for the NIU driver turned out to be more
complicated than first assumed.

The discard counters for the NIU neptune chip are only 16-bit (even
though this is a 64-bit chip).  These 16-bit counters can overflow
quickly, especially considering this is a 10Gbit/s ethernet card.

The overflow indication bit is, unfortunatly, not usable as the
counter value does not wrap, but remains at max value 0xFFFF.
Resulting in lost counts until the counter is reset.

The read and reset scheme also poses a problem. Both in theory and in
practice counters can be lost in between reading nr64() and clearing
the counter nw64().  For this reason, the number of counter clearings
nw64() is limited/reduced.  On the fast-path the counters are only
syncronized once it exceeds 0x7FFF.  When read by userspace, its
syncronized fully.

Signed-off-by: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoenc28j60: reduce the number of spi transfers in enc28j60_set_bank()
Baruch Siach [Fri, 19 Dec 2008 03:39:14 +0000 (19:39 -0800)]
enc28j60: reduce the number of spi transfers in enc28j60_set_bank()

A major source of overhead in the enc28j60 driver is the SPI transfers. Each
SPI transfer entails two kernel thread context switches. One major source of
SPI transfers is the enc28j60_set_bank() functions which runs before every
register access. This patch reduces the number of SPI transfers that
enc28j60_set_bank() performs in two ways:

  1. removes unnecessary bank switch for the registers that are present in all
 banks

  2. when switching from banks 0 or 3 to banks 1 or 2 (i.e. only one bit
 changes) enc28j60_set_bank() does only one SPI transfer instead of two

According to my tests these changes reduce the number of SPI transfers in
about 25%.

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetdevice zd1201: Use after free
Wang Chen [Fri, 19 Dec 2008 03:36:46 +0000 (19:36 -0800)]
netdevice zd1201: Use after free

| commit 3d29b0c33d431ecc69ec778f8c236d382f59a85f
| Author: John W. Linville <linville@tuxdriver.com>
| Date:   Fri Oct 31 14:13:12 2008 -0400
|
|     netdevice zd1201: Convert directly reference of netdev->priv to netdev_priv()
|
|     We have some reasons to kill netdev->priv:
|     1. netdev->priv is equal to netdev_priv().
|     2. netdev_priv() wraps the calculation of netdev->priv's offset, obviously
|        netdev_priv() is more flexible than netdev->priv.
|     But we cann't kill netdev->priv, because so many drivers reference to it
|     directly.
|
|     OK, becasue Dave S. Miller said, "every direct netdev->priv usage is a bug",
|     and I want to kill netdev->priv later, I decided to convert all the direct
|     reference of netdev->priv first.
|
|     (Original patch posted by Wang Chen <wangchen@cn.fujitsu.com> w/ above
|     changelog but using dev->ml_priv.  That doesn't seem appropriate
|     to me for this driver, so I've revamped it to use netdev_priv()
|     instead. -- JWL)

This commit changed the allocation of netdev, but didn't change
the free method of it.
This causes "zd" be used after the memory, which is pointed by "zd", being
freed by free_netdev().

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRevert "xfrm: Accept ESP packets regardless of UDP encapsulation mode"
David S. Miller [Fri, 19 Dec 2008 03:23:56 +0000 (19:23 -0800)]
Revert "xfrm: Accept ESP packets regardless of UDP encapsulation mode"

This reverts commit e061b165c7f4ec5e2e160d990b49011b5b6e5c6a.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoppp: fix segfaults introduced by netdev_priv changes
James Chapman [Wed, 17 Dec 2008 12:02:16 +0000 (12:02 +0000)]
ppp: fix segfaults introduced by netdev_priv changes

This patch fixes a segfault in ppp_shutdown_interface() and
ppp_destroy_interface() when a PPP connection is closed. I bisected
the problem to the following commit:

  commit c8019bf3aff653cceb64f66489fc299ee5957b57
  Author: Wang Chen <wangchen@cn.fujitsu.com>
  Date:   Thu Nov 20 04:24:17 2008 -0800

    netdevice ppp: Convert directly reference of netdev->priv

    1. Use netdev_priv(dev) to replace dev->priv.
    2. Alloc netdev's private data by alloc_netdev().

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The original ppp_generic code treated the netdev and struct ppp as
independent data structures which were freed separately. In moving the
ppp struct into the netdev, it is now possible for the private data to
be freed before the call to ppp_shutdown_interface(), which is bad.

The kfree(ppp) in ppp_destroy_interface() is also wrong; presumably
ppp hasn't worked since the above commit.

The following patch fixes both problems.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Fix module refcount leak in kernel_accept()
Wei Yongjun [Fri, 19 Dec 2008 03:35:10 +0000 (19:35 -0800)]
net: Fix module refcount leak in kernel_accept()

The kernel_accept() does not hold the module refcount of newsock->ops->owner,
so we need __module_get(newsock->ops->owner) code after call kernel_accept()
by hand.
In sunrpc, the module refcount is missing to hold. So this cause kernel panic.

Used following script to reproduct:

while [ 1 ];
do
    mount -t nfs4 192.168.0.19:/ /mnt
    touch /mnt/file
    umount /mnt
    lsmod | grep ipv6
done

This patch fixed the problem by add __module_get(newsock->ops->owner) to
kernel_accept(). So we do not need to used __module_get(newsock->ops->owner)
in every place when used kernel_accept().

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoRevert "net: release skb->dst in sock_queue_rcv_skb()"
David S. Miller [Thu, 18 Dec 2008 06:11:38 +0000 (22:11 -0800)]
Revert "net: release skb->dst in sock_queue_rcv_skb()"

This reverts commit 70355602879229c6f8bd694ec9c0814222bc4936.

As pointed out by Mark McLoughlin IP_PKTINFO cmsg data is one
post-queueing user, so this optimization is not valid right
now.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobonding: use port_params in __update_lacpdu_from_port
Holger Eitzenberger [Thu, 18 Dec 2008 03:13:53 +0000 (19:13 -0800)]
bonding: use port_params in __update_lacpdu_from_port

Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobonding: use port_params in __update_default_selected()
Holger Eitzenberger [Thu, 18 Dec 2008 03:13:27 +0000 (19:13 -0800)]
bonding: use port_params in __update_default_selected()

I also removed some of the unneeded braces in the if condition to
improve readability and a little bit of reformatting.

Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobonding: use port_params in __update_selected()
Holger Eitzenberger [Thu, 18 Dec 2008 03:13:07 +0000 (19:13 -0800)]
bonding: use port_params in __update_selected()

Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobonding: remove various function declarations
Holger Eitzenberger [Thu, 18 Dec 2008 03:12:40 +0000 (19:12 -0800)]
bonding: remove various function declarations

They are all defined before used, it's therefore ok to remove
them.

Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobonding: init port_params from template
Holger Eitzenberger [Thu, 18 Dec 2008 03:12:07 +0000 (19:12 -0800)]
bonding: init port_params from template

Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobonding: use plain memcpy in __record_default()
Holger Eitzenberger [Thu, 18 Dec 2008 03:08:46 +0000 (19:08 -0800)]
bonding: use plain memcpy in __record_default()

Also remove the pointless comment at the top.

Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobonding: use port_params in __record_pdu()
Holger Eitzenberger [Thu, 18 Dec 2008 03:08:14 +0000 (19:08 -0800)]
bonding: use port_params in __record_pdu()

It helps in maintaining the various partner information values from
the LACPDU.  It also removes the pointless comment at the top.

Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobonding: introduce and use port_params structure
Holger Eitzenberger [Thu, 18 Dec 2008 03:07:38 +0000 (19:07 -0800)]
bonding: introduce and use port_params structure

It generally helps to handle those values in various places, using it
might make the code more readable and gives room for other improvements.

The IEEE standard talks about them as "parameter values".

Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobonding: improve elaborate port_state assignment
Holger Eitzenberger [Thu, 18 Dec 2008 03:07:16 +0000 (19:07 -0800)]
bonding: improve elaborate port_state assignment

The previous code was just a funny way of assigning both values (they
are both of type u8).

Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: David S. Miller <davem@davemloft.net>