profile/ivi/kernel-x86-ivi.git
16 years agonet: Remove __skb_insert() calls outside of skbuff internals.
David S. Miller [Mon, 22 Sep 2008 04:28:51 +0000 (21:28 -0700)]
net: Remove __skb_insert() calls outside of skbuff internals.

This minor cleanup simplifies later changes which will convert
struct sk_buff and friends over to using struct list_head.

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: advertise MSS requested by user
Tom Quetchenbach [Sun, 21 Sep 2008 07:21:51 +0000 (00:21 -0700)]
tcp: advertise MSS requested by user

I'm trying to use the TCP_MAXSEG option to setsockopt() to set the MSS
for both sides of a bidirectional connection.

man tcp says: "If this option is set before connection establishment, it
also changes the MSS value announced to the other end in the initial
packet."

However, the kernel only uses the MTU/route cache to set the advertised
MSS. That means if I set the MSS to, say, 500 before calling connect(),
I will send at most 500-byte packets, but I will still receive 1500-byte
packets in reply.

This is a bug, either in the kernel or the documentation.

This patch (applies to latest net-2.6) reduces the advertised value to
that requested by the user as long as setsockopt() is called before
connect() or accept(). This seems like the behavior that one would
expect as well as that which is documented.

I've tried to make sure that things that depend on the advertised MSS
are set correctly.

Signed-off-by: Tom Quetchenbach <virtualphtn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agonet: Use hton[sl]() instead of __constant_hton[sl]() where applicable
Arnaldo Carvalho de Melo [Sun, 21 Sep 2008 05:20:49 +0000 (22:20 -0700)]
net: Use hton[sl]() instead of __constant_hton[sl]() where applicable

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agomultiq: requeue should rewind the current_band
Alexander Duyck [Sun, 21 Sep 2008 05:07:34 +0000 (22:07 -0700)]
multiq: requeue should rewind the current_band

Currently dequeueing a packet and requeueing the same packet will cause a
different packet to be pulled on the next dequeue.  This change forces
requeue to rewind the current_band.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: back retransmit_high when it over-estimated
Ilpo Järvinen [Sun, 21 Sep 2008 04:26:22 +0000 (21:26 -0700)]
tcp: back retransmit_high when it over-estimated

If lost skb is sacked, we might have nothing to retransmit
as high as the retransmit_high is pointing to, so place
it lower to avoid unnecessary walking.

This is mainly for the case where high L'ed skbs gets sacked.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: don't clear lost_skb_hint when not necessary
Ilpo Järvinen [Sun, 21 Sep 2008 04:25:52 +0000 (21:25 -0700)]
tcp: don't clear lost_skb_hint when not necessary

Most importantly avoid doing it with cumulative ACK. However,
since we have lost_cnt_hint in the picture as well needing
adjustments, it's not as trivial as dealing with
retransmit_skb_hint (and cannot be done in the all place we
could trivially leave retransmit_skb_hint untouched).

With the previous patch, this should mostly remove O(n^2)
behavior while cumulative ACKs start flowing once rexmit
after a lossy round-trip made it through.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: don't clear retransmit_skb_hint when not necessary
Ilpo Järvinen [Sun, 21 Sep 2008 04:25:15 +0000 (21:25 -0700)]
tcp: don't clear retransmit_skb_hint when not necessary

Most importantly avoid doing it with cumulative ACK. Not clearing
means that we no longer need n^2 processing in resolution of each
fast recovery.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: remove retransmit_skb_hint clearing from failure
Ilpo Järvinen [Sun, 21 Sep 2008 04:24:49 +0000 (21:24 -0700)]
tcp: remove retransmit_skb_hint clearing from failure

This doesn't much sense here afaict, probably never has. Since
fragmenting and collapsing deal the hints by themselves, there
should be very little reason for the rexmit loop to do that.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: reorganize retransmit code loops
Ilpo Järvinen [Sun, 21 Sep 2008 04:24:21 +0000 (21:24 -0700)]
tcp: reorganize retransmit code loops

Both loops are quite similar, so they can be combined
with little effort. As a result, forward_skb_hint becomes
obsolete as well.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: remove tp->lost_out guard to make joining diff nicer
Ilpo Järvinen [Sun, 21 Sep 2008 04:23:49 +0000 (21:23 -0700)]
tcp: remove tp->lost_out guard to make joining diff nicer

The validity of the retransmit_high must then be ensured
if no L'ed skb exits!

This makes a minor change to behavior, we now have to
iterate the head to find out that the loop terminates.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: Reorganize skb tagbit checks
Ilpo Järvinen [Sun, 21 Sep 2008 04:22:59 +0000 (21:22 -0700)]
tcp: Reorganize skb tagbit checks

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: remove obsolete validity concern
Ilpo Järvinen [Sun, 21 Sep 2008 04:22:17 +0000 (21:22 -0700)]
tcp: remove obsolete validity concern

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: add tcp_can_forward_retransmit
Ilpo Järvinen [Sun, 21 Sep 2008 04:21:54 +0000 (21:21 -0700)]
tcp: add tcp_can_forward_retransmit

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: No need to clear retransmit_skb_hint when SACKing
Ilpo Järvinen [Sun, 21 Sep 2008 04:21:16 +0000 (21:21 -0700)]
tcp: No need to clear retransmit_skb_hint when SACKing

Because lost counter no longer requires tuning, this is
trivial to remove (the tuning wouldn't have been too
hard either) because no "new" retransmittable skb appeared
below retransmit_skb_hint when SACKing for sure.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: Kill precaution that's very likely obsolete
Ilpo Järvinen [Sun, 21 Sep 2008 04:20:50 +0000 (21:20 -0700)]
tcp: Kill precaution that's very likely obsolete

I suspect it might have been related to the changed amount
of lost skbs, which was counted by retransmit_cnt_hint that
got changed.

The place for this clearing was very illogical anyway,
it should have been after the LOST-bit clearing loop to
make any sense.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: convert retransmit_cnt_hint to seqno
Ilpo Järvinen [Sun, 21 Sep 2008 04:20:20 +0000 (21:20 -0700)]
tcp: convert retransmit_cnt_hint to seqno

Main benefit in this is that we can then freely point
the retransmit_skb_hint to anywhere we want to because
there's no longer need to know what would be the count
changes involve, and since this is really used only as a
terminator, unnecessary work is one time walk at most,
and if some retransmissions are necessary after that
point later on, the walk is not full waste of time
anyway.

Since retransmit_high must be kept valid, all lost
markers must ensure that.

Now I also have learned how those "holes" in the
rexmittable skbs can appear, mtu probe does them. So
I removed the misleading comment as well.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: add helper for lost bit toggling
Ilpo Järvinen [Sun, 21 Sep 2008 04:19:22 +0000 (21:19 -0700)]
tcp: add helper for lost bit toggling

This useful because we'd need to verifying soon in many places
which makes things slightly more complex than it used to be.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: move tcp_verify_retransmit_hint
Ilpo Järvinen [Sun, 21 Sep 2008 04:18:55 +0000 (21:18 -0700)]
tcp: move tcp_verify_retransmit_hint

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp: Partial hint clearing has again become meaningless
Ilpo Järvinen [Sun, 21 Sep 2008 04:18:32 +0000 (21:18 -0700)]
tcp: Partial hint clearing has again become meaningless

Ie., the difference between partial and all clearing doesn't
exists anymore since the SACK optimizations got dropped by
an sacktag rewrite.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Fri, 19 Sep 2008 23:17:12 +0000 (16:17 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next-2.6

16 years agoqlge: Protect qlge_resume() with CONFIG_PM
David S. Miller [Fri, 19 Sep 2008 23:14:24 +0000 (16:14 -0700)]
qlge: Protect qlge_resume() with CONFIG_PM

Fixes the following build warning:

drivers/net/qlge/qlge_main.c:3897: warning: ‘qlge_resume’ defined but not used

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoqlge: Fix warnings in debugging code.
David S. Miller [Fri, 19 Sep 2008 23:13:05 +0000 (16:13 -0700)]
qlge: Fix warnings in debugging code.

drivers/net/qlge/qlge_dbg.c: In function ‘ql_dump_qdev’:
drivers/net/qlge/qlge_dbg.c:369: warning: cast to pointer from integer of different size
drivers/net/qlge/qlge_dbg.c:373: warning: cast to pointer from integer of different size
drivers/net/qlge/qlge_dbg.c: In function ‘ql_dump_tx_ring’:
drivers/net/qlge/qlge_dbg.c:457: warning: format ‘%llx’ expects type ‘long long unsigned int’, but argument 2 has type ‘long unsigned int’
drivers/net/qlge/qlge_dbg.c:461: warning: format ‘%llx’ expects type ‘long long unsigned int’, but argument 2 has type ‘long unsigned int’
drivers/net/qlge/qlge_dbg.c: In function ‘ql_dump_rx_ring’:
drivers/net/qlge/qlge_dbg.c:557: warning: format ‘%llx’ expects type ‘long long unsigned int’, but argument 2 has type ‘long unsigned int’
drivers/net/qlge/qlge_dbg.c:565: warning: format ‘%llx’ expects type ‘long long unsigned int’, but argument 2 has type ‘long unsigned int’
drivers/net/qlge/qlge_dbg.c:575: warning: format ‘%llx’ expects type ‘long long unsigned int’, but argument 2 has type ‘long unsigned int’
drivers/net/qlge/qlge_dbg.c:579: warning: format ‘%llx’ expects type ‘long long unsigned int’, but argument 2 has type ‘long unsigned int’
drivers/net/qlge/qlge_dbg.c:598: warning: format ‘%llx’ expects type ‘long long unsigned int’, but argument 2 has type ‘long unsigned int’
drivers/net/qlge/qlge_dbg.c:602: warning: format ‘%llx’ expects type ‘long long unsigned int’, but argument 2 has type ‘long unsigned int’

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoMerge branch 'new-drivers' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
David S. Miller [Fri, 19 Sep 2008 22:51:35 +0000 (15:51 -0700)]
Merge branch 'new-drivers' of /linux/kernel/git/jgarzik/netdev-2.6

16 years agonet: Fix build with ARCH=um
David S. Miller [Fri, 19 Sep 2008 19:44:54 +0000 (12:44 -0700)]
net: Fix build with ARCH=um

If UM is going to claim that it supports DMA by setting
HAS_DMA, it should provide a dma_mapping_error() implementation.

Based upon a report by Julius Volz.

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoISDN sockets: add missing lockdep strings
Rémi Denis-Courmont [Thu, 18 Sep 2008 23:44:31 +0000 (16:44 -0700)]
ISDN sockets: add missing lockdep strings

Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agobnx2: Remove timer_interval field from the bnx2 structure
Benjamin Li [Thu, 18 Sep 2008 23:40:49 +0000 (16:40 -0700)]
bnx2: Remove timer_interval field from the bnx2 structure

The timer_interval field is only assigned once, and never reassigned.
We can safely replace all instances of the timer_interval with a
constant value.

Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agobnx2: Remove name field from bnx2 structure
Benjamin Li [Thu, 18 Sep 2008 23:40:03 +0000 (16:40 -0700)]
bnx2: Remove name field from bnx2 structure

The name of the board is only used during the initialization of
the adapter.  We can save the space of a pointer by not storing
this information.

Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agobnx2: Update MODULE_DESCRIPTION to include the 5716
Benjamin Li [Thu, 18 Sep 2008 23:39:16 +0000 (16:39 -0700)]
bnx2: Update MODULE_DESCRIPTION to include the 5716

Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agobnx2: In bnx2_set_mac_link() return void rather then int
Benjamin Li [Thu, 18 Sep 2008 23:38:24 +0000 (16:38 -0700)]
bnx2: In bnx2_set_mac_link() return void rather then int

bnx2_set_mac_link() doesn't need to return any error codes.  And
all the callers don't check the return code. It is safe to
change the return type to a void.

Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoqlge: New Qlogic 10Gb Ethernet Driver.
Ron Mercer [Thu, 18 Sep 2008 15:56:28 +0000 (11:56 -0400)]
qlge: New Qlogic 10Gb Ethernet Driver.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
16 years agojme: JMicron Gigabit Ethernet Driver
Guo-Fu Tseng [Mon, 15 Sep 2008 17:00:11 +0000 (01:00 +0800)]
jme: JMicron Gigabit Ethernet Driver

Supporting JMC250, and JMC260.

Signed-off-by: Guo-Fu Tseng <cooldavid@cooldavid.org>
Acked-and-tested-by: Ethan Hsiao <ethanhsiao@jmicron.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
16 years agoenic: add Cisco 10G Ethernet NIC driver
Scott Feldman [Mon, 15 Sep 2008 16:17:11 +0000 (09:17 -0700)]
enic: add Cisco 10G Ethernet NIC driver

Signed-off-by: Scott Feldman <scofeldm@cisco.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
16 years agoatl2: add atl2 driver
Chris Snook [Mon, 15 Sep 2008 00:56:10 +0000 (19:56 -0500)]
atl2: add atl2 driver

Driver for Atheros L2 10/100 network device. Includes necessary
changes for Kconfig, Makefile, and pci_ids.h.

Signed-off-by: Chris Snook <csnook@redhat.com>
Signed-off-by: Jay Cliburn <jacliburn@bellsouth.net>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
16 years agoipvs: change some __constant_htons() to htons()
Brian Haley [Tue, 16 Sep 2008 15:11:11 +0000 (11:11 -0400)]
ipvs: change some __constant_htons() to htons()

Change __contant_htons() to htons() in the IPVS code when not in an
initializer.

-Brian

Signed-off-by: Brian Haley <brian.haley@hp.com>
Acked-by: Julius Volz <juliusv@google.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
16 years agoipvs: add __aquire/__release annotations to ip_vs_info_seq_start/ip_vs_info_seq_stop
Simon Horman [Wed, 17 Sep 2008 00:10:42 +0000 (10:10 +1000)]
ipvs: add __aquire/__release annotations to ip_vs_info_seq_start/ip_vs_info_seq_stop

This teaches sparse that the following are not problems:

make C=1
  CHECK   net/ipv4/ipvs/ip_vs_ctl.c
net/ipv4/ipvs/ip_vs_ctl.c:1793:14: warning: context imbalance in 'ip_vs_info_seq_start' - wrong count at exit
net/ipv4/ipvs/ip_vs_ctl.c:1842:13: warning: context imbalance in 'ip_vs_info_seq_stop' - unexpected unlock

Acked-by: Sven Wegener <sven.wegener@stealer.net>
Acked-by: Julius Volz <juliusv@google.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
16 years agoipvs: supply a valid 0 address to ip_vs_conn_new()
Simon Horman [Wed, 17 Sep 2008 00:10:42 +0000 (10:10 +1000)]
ipvs: supply a valid 0 address to ip_vs_conn_new()

ip_vs_conn_new expects a union nf_inet_addr as the type for its address
parameters, not a plain integer.

This problem was detected by sparse.

make C=1
  CHECK   net/ipv4/ipvs/ip_vs_core.c
net/ipv4/ipvs/ip_vs_core.c:469:9: warning: Using plain integer as NULL pointer

Acked-by: Sven Wegener <sven.wegener@stealer.net>
Acked-by: Julius Volz <juliusv@google.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
16 years agoipvs: only unlock in ip_vs_edit_service() if already locked
Simon Horman [Wed, 17 Sep 2008 00:10:41 +0000 (10:10 +1000)]
ipvs: only unlock in ip_vs_edit_service() if already locked

Jumping to out unlocks __ip_vs_svc_lock, but that lock is not taken until
after code that may jump to out.

This problem was detected by sparse.

make C=1
  CHECK   net/ipv4/ipvs/ip_vs_ctl.c
net/ipv4/ipvs/ip_vs_ctl.c:1332:2: warning: context imbalance in 'ip_vs_edit_service' - unexpected unlock

Acked-by: Sven Wegener <sven.wegener@stealer.net>
Acked-by: Julius Volz <juliusv@google.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/horms/lvs-2.6 into lvs-next-2.6
Simon Horman [Wed, 17 Sep 2008 00:08:26 +0000 (10:08 +1000)]
Merge git://git./linux/kernel/git/horms/lvs-2.6 into lvs-next-2.6

16 years agolibertas: correct "limited range of data type" warning
John W. Linville [Mon, 15 Sep 2008 21:25:54 +0000 (17:25 -0400)]
libertas: correct "limited range of data type" warning

  CC [M]  drivers/net/wireless/libertas/wext.o
drivers/net/wireless/libertas/wext.c: In function ‘lbs_get_rts’:
drivers/net/wireless/libertas/wext.c:307: warning: comparison is always
false due to limited range of data type

Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoorinoco: Fix compile warnings
David Kilroy [Fri, 12 Sep 2008 21:28:18 +0000 (22:28 +0100)]
orinoco: Fix compile warnings

Use min_t to avoid warnings when the typesafe version is used.

Explicitly cast u64s to unsigned long long when being passed to printk.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211 : Fix mode change hard_start_xmit function
Abhijeet Kolekar [Fri, 12 Sep 2008 20:44:08 +0000 (13:44 -0700)]
mac80211 : Fix mode change hard_start_xmit function

When monitor mode is changed to BSS or IBSS, data trasnfer can not happen
because proper transmit function is not assigend for BSS ,IBSS mode.
This patch fixes this problem by assigning the ieee80211_subif_start_xmit
to device's hard_start_xmit function.

Signed-off-by: Abhijeet Kolekar <abhijeet.kolekar@intel.com>
Acked-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: remove debug frame dumping
Johannes Berg [Fri, 12 Sep 2008 13:22:53 +0000 (15:22 +0200)]
mac80211: remove debug frame dumping

You can just pull up a monitor interface to get much more
detailed information, or, when debugging a driver, insert
dump code into the driver (which usually you will have to
do anyway to dump the driver-specific information). Hence
this option is useless.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: remove beacon counters
Johannes Berg [Fri, 12 Sep 2008 09:05:39 +0000 (11:05 +0200)]
mac80211: remove beacon counters

The beacon counters mac80211 keeps are only used for debugfs,
unfortunately, they are incorrect for many hardware designs,
namely any design that has a beacon template. Hence, remove the
counters so we don't create the impression they are usable.

This also allows removing the beacon MESH #ifdef again.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: clean up kdoc
Johannes Berg [Fri, 12 Sep 2008 08:18:44 +0000 (10:18 +0200)]
mac80211: clean up kdoc

A few errors sneaked in over time, some functions no longer exist,
for some alternatives exist. This changes the docbook template to
include the right things.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: fix sta_info kernel-doc warning
Johannes Berg [Fri, 12 Sep 2008 07:53:27 +0000 (09:53 +0200)]
mac80211: fix sta_info kernel-doc warning

Sorry, forgot to run kernel-doc and just got the output from the nightly
run by email, this fixes a warning which I introduced when doing the
first RC API cleanups.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211 hwsim: make radio list dynamic
Johannes Berg [Thu, 11 Sep 2008 22:39:22 +0000 (00:39 +0200)]
mac80211 hwsim: make radio list dynamic

This paves the way for dynamic radio additions while the module
is loaded. Also restrict the number of radios to 100 because
creating that many already takes forever.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: fix virtual interfaces vs. injection
Johannes Berg [Fri, 12 Sep 2008 20:52:47 +0000 (22:52 +0200)]
mac80211: fix virtual interfaces vs. injection

Currently, virtual interface pointers passed to drivers might be
from monitor interfaces and as such completely uninitialised
because we do not tell the driver about monitor interfaces when
those are created. Instead of passing them, we should therefore
indicate to the driver that there is no information; do that by
passing a NULL value and adjust drivers to cope with it.

As a result, some mac80211 API functions also need to cope with
a NULL vif pointer so drivers can still call them unconditionally.

Also, when injecting frames we really don't want to pass NULL all
the time, if we know we are the source address of a frame and have
a local interface for that address, we can to use that interface.
This also helps with processing the frame correctly for that
interface which will help the 802.11w implementation. It's not
entirely correct for VLANs or WDS interfaces because there the MAC
address isn't unique, but it's already a lot better than what we
do now.

Finally, when injecting without a matching local interface, don't
assign sequence numbers at all.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agolibertas: Reduce the WPA key installation time.
Javier Cardona [Thu, 11 Sep 2008 22:32:50 +0000 (15:32 -0700)]
libertas: Reduce the WPA key installation time.

WPA requires that the PTK is installed immediately after the 4-way handshake
in order to properly decrypt the subsequent incoming EAPOL-GTK frame.  If the
PTK is not enabled by the time the EAPOL-GTK frame arrives, the frame is
dropped and the supplicant does not receive the group key.

This will happen with fast Access Points that send the EAPOL-GTK frame before
the suplicant has successfully installed and enabled the PTK.  To mitigate
this situation, this patch simplifies and accelerates the SIOCSIWENCODEEXT
execution.

This patch resolves OLPC ticket 7825 (http://dev.laptop.org/ticket/7825)

Signed-off-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agorfkill: remove transmitter blocking on suspend
Henrique de Moraes Holschuh [Tue, 26 Aug 2008 14:58:01 +0000 (11:58 -0300)]
rfkill: remove transmitter blocking on suspend

Currently, rfkill would stand in the way of properly supporting wireless
devices that are capable of waking the system up from sleep or hibernation
when they receive a special wireless message.  It would also get in the way
of mesh devices that need to remain operational even during platform
suspend.

To avoid that, stop trying to block the transmitters on the rfkill class
suspend handler.

Drivers that need rfkill's older behaviour will have to implement it by
themselves in their own suspend handling.

Do note that rfkill *will* attempt to restore the transmitter state on
resume in any situation.  This happens after the driver's resume method is
called by the suspend core (class devices resume after the devices they are
attached to have been resumed).

The following drivers need to check if they need to explicitly block
their transmitters in their own suspend handlers (maintainers Cc'd):
arch/arm/mach-pxa/tosa-bt.c
drivers/net/usb/hso.c
drivers/net/wireless/rt2x00/* (USB might need it?)
drivers/net/wireless/b43/ (SSB over USB might need it?)
drivers/misc/hp-wmi.c
eeepc-laptop w/rfkill support (not in mainline yet)
Compal laptop w/rfkill support (not in mainline yet)
toshiba-acpi w/rfkill support (not in mainline yet)

Signed-off-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Cc: Ivo van Doorn <IvDoorn@gmail.com>
Cc: Matthew Garrett <mjg@redhat.com>
Cc: Andrew Bird <ajb@spheresystems.co.uk>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Cezary Jackiewicz <cezary.jackiewicz@gmail.com>
Cc: Philip Langdale <philipl@overt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agob43: fix QoS parameters initialization
Lorenzo Nava [Thu, 11 Sep 2008 13:06:24 +0000 (15:06 +0200)]
b43: fix QoS parameters initialization

This fixes the initialization of QoS parameters.

Reported-by: Lorenzo Nava, Francesco Gringoli
Signed-off-by: Francesco Gringoli <francesco.gringoli@ing.unibs.it>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoiwl3945: add MODULE_FIRMWARE for 3945 HW
Zhu Yi [Thu, 11 Sep 2008 03:45:22 +0000 (11:45 +0800)]
iwl3945: add MODULE_FIRMWARE for 3945 HW

This patch adds MODULE_FIRMWARE statement for 3945 HW.

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoiwlwifi: add MODULE_FIRMWARE for 5000 HW
Tomas Winkler [Thu, 11 Sep 2008 03:45:21 +0000 (11:45 +0800)]
iwlwifi: add MODULE_FIRMWARE for 5000 HW

This patch adds MODULE_FIRMWARE statement for 5000 HW.

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>
16 years agoiwlwifi: update 5350 Wifi PCI IDs
Tomas Winkler [Thu, 11 Sep 2008 03:45:20 +0000 (11:45 +0800)]
iwlwifi: update 5350 Wifi PCI IDs

This patch updates PCI IDs for 5350 Wifi/WiMax.

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>
16 years agomac80211: pass AP vif pointer for VLANs
Johannes Berg [Thu, 11 Sep 2008 03:27:40 +0000 (05:27 +0200)]
mac80211: pass AP vif pointer for VLANs

We cannot pass a VLAN vif pointer to the driver since those are
entirely virtual and we never tell the driver.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoiwlwifi: don't access mac80211's AMPDU state machine
Johannes Berg [Thu, 11 Sep 2008 01:17:05 +0000 (03:17 +0200)]
iwlwifi: don't access mac80211's AMPDU state machine

There really is no need, at worst ieee80211_start_tx_ba_session
will log a message when debugging is enabled, and poking such
internals of mac80211 definitely doesn't belong into an RC
algorithm.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: share sta_info->ht_info
Johannes Berg [Thu, 11 Sep 2008 01:14:11 +0000 (03:14 +0200)]
mac80211: share sta_info->ht_info

Rate control algorithms may need access to a station's
HT capabilities, so share the ht_info struct in the
public station API.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: move txrate_idx into RC algorithms
Johannes Berg [Thu, 11 Sep 2008 01:04:36 +0000 (03:04 +0200)]
mac80211: move txrate_idx into RC algorithms

The sta_info->txrate_idx member isn't used by all RC algorithms
in the way it was intended to be used, move it into those that
require it (only PID) and keep track in the core code of which
rate was last used for reporting to userspace and the mesh MLME.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: share sta->supp_rates
Johannes Berg [Thu, 11 Sep 2008 00:45:11 +0000 (02:45 +0200)]
mac80211: share sta->supp_rates

As more preparation for a saner rate control algorithm API,
share the supported rates bitmap in the public API.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: move last_txrate_idx into RC algorithms
Johannes Berg [Thu, 11 Sep 2008 00:22:58 +0000 (02:22 +0200)]
mac80211: move last_txrate_idx into RC algorithms

This variable in sta_info is only used in a meaningful way
by the Intel RC algorithms, so move it into those.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: small rate control changes
Johannes Berg [Thu, 11 Sep 2008 00:03:28 +0000 (02:03 +0200)]
mac80211: small rate control changes

This patch fixes mac80211 to not rely on the rate control
algorithm to update sta->tx_retry_failed and sta->tx_retry_count
(even if we don't currently use them), removes a number of
completely unused values we don't even show in debugfs and
changes the code in ieee80211_tx_status() to not look up the
sta_info repeatedly.

The only behaviour change here would be not calling the rate
control function rate_control_tx_status() when no sta_info is
found, but all rate control algorithms ignore such calls anyway.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211 hwsim: verify sta pointers
Johannes Berg [Thu, 11 Sep 2008 00:17:01 +0000 (02:17 +0200)]
mac80211 hwsim: verify sta pointers

In analogy with the previous patch to make mac80211-hwsim
verify that the virtual interface pointers are correct,
this makes it very that it knows about all station structs.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: share STA information with driver
Johannes Berg [Wed, 10 Sep 2008 22:02:02 +0000 (00:02 +0200)]
mac80211: share STA information with driver

This patch changes mac80211 to share some more data about
stations with drivers. Should help iwlwifi and ath9k when
 they get around to updating, and might also help with
implementing rate control algorithms without internals.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Cc: Sujith Manoharan <Sujith.Manoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211 hwsim: verify vif pointers
Johannes Berg [Thu, 11 Sep 2008 00:16:36 +0000 (02:16 +0200)]
mac80211 hwsim: verify vif pointers

mac80211-hwsim is a debugging tool for mac80211, and as such
it can very well verify that mac80211 isn't passing junk to
drivers, especially the vif pointer is prone to this because
for vlan interfaces the AP interface pointer needs to be passed.
This makes mac80211-hwsim add a magic cookie to the private vif
area and verify it whenever an operation is called that gets a
vif pointer.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: warn on some invalid vlan operations
Johannes Berg [Wed, 10 Sep 2008 22:02:00 +0000 (00:02 +0200)]
mac80211: warn on some invalid vlan operations

These should never happen, but better warn about them than
crashing a driver, the fact that they never happen is rather
subtle throughout mac80211.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: move regular interface handling
Johannes Berg [Wed, 10 Sep 2008 22:01:59 +0000 (00:01 +0200)]
mac80211: move regular interface handling

Move the code to handle regular interfaces out of main.c and
into iface.c, keep only the master interface stuff in main.c.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: use nl80211 interface types
Johannes Berg [Wed, 10 Sep 2008 22:01:58 +0000 (00:01 +0200)]
mac80211: use nl80211 interface types

There's really no reason for mac80211 to be using its
own interface type defines. Use the nl80211 types and
simplify the configuration code a bit: there's no need
to translate them any more now.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: inform driver of basic rateset
Johannes Berg [Wed, 10 Sep 2008 22:01:57 +0000 (00:01 +0200)]
mac80211: inform driver of basic rateset

Drivers need to know the basic rateset to be able to configure
the ACK/CTS programming in hardware correctly.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: clean up some comments
Johannes Berg [Wed, 10 Sep 2008 22:01:56 +0000 (00:01 +0200)]
mac80211: clean up some comments

Some comments refer to 80211.o or similar; also remove
a comment about implementing fragments better, we really
have better things to do.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: clean up scan namespace
Johannes Berg [Wed, 10 Sep 2008 22:01:55 +0000 (00:01 +0200)]
mac80211: clean up scan namespace

Most of the scan functions are called ieee80211_sta_scan_*
or similar, make clean it up so they are all just called
ieee80211_scan_*.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: make bridge_packets a virtual interface option
Johannes Berg [Wed, 10 Sep 2008 22:01:54 +0000 (00:01 +0200)]
mac80211: make bridge_packets a virtual interface option

The bridge_packets configuration really should be per virtual
interface (theoretically per AP/VLAN, but this is much easier);
there currently is no way to set it yet though. Also invert
the option to "NO_BRIDGE_PACKETS" so the default is to bridge.

While at it, also document the flags properly.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: move ieee80211_set_freq to utils
Johannes Berg [Wed, 10 Sep 2008 22:01:53 +0000 (00:01 +0200)]
mac80211: move ieee80211_set_freq to utils

It really doesn't belong into the wireless extensions code.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: reorder MLME code more
Johannes Berg [Wed, 10 Sep 2008 22:01:52 +0000 (00:01 +0200)]
mac80211: reorder MLME code more

This way all the utility functions are at the top, then the
state machine and externally callable functions are moved to
the bottom. Also clean up ieee80211_i.h a bit and add a few
comments about which functions are called from where.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: fix scan vs. interface removal race
Johannes Berg [Wed, 10 Sep 2008 22:01:51 +0000 (00:01 +0200)]
mac80211: fix scan vs. interface removal race

When we remove an interface, we can currently end up having
a pointer to it left in local->scan_sdata after it has been
set down, and then with a hardware scan the scan completion
can try to access it which is a bug. Alternatively, a scan
that started as a hardware scan may terminate as though it
was a software scan, if the timing is just right.

On SMP systems, software scan also has a similar problem,
just canceling the delayed work and setting a flag isn't
enough since it may be running concurrently; in this case
we would also never restore state of other interfaces.

This patch hopefully fixes the problems by always invoking
ieee80211_scan_completed or requiring it to be invoked by
the driver, I suspect the drivers that have ->hw_scan() are
buggy. The bug will not manifest itself unless you remove
the interface while hw-scanning which will also turn off
the hw, and then add a new interface which will be unusable
until you scan once.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: fix work race
Johannes Berg [Wed, 10 Sep 2008 22:01:50 +0000 (00:01 +0200)]
mac80211: fix work race

When we stop an interface, the work on it may still be pending
or running. We do cancel the timer, but we do not currently
protect against the work struct. The race is very unlikely to
hit -- it'll happen only when the driver is using mac80211's
workqueue to run long-running tasks and the sta/mesh works are
delayed for quite a bit.

This patch fixes it by cancelling the work explicitly.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: split off mesh handling entirely
Johannes Berg [Wed, 10 Sep 2008 22:01:49 +0000 (00:01 +0200)]
mac80211: split off mesh handling entirely

This patch splits off mesh handling from the STA/IBSS.
Unfortunately it increases mesh code size a bit, but I
think it makes things clearer. The patch also reduces
per-interface run-time memory usage.

Also clean up a few places where ifdef is not required.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: dont set REQ_RUN when scan finishes
Johannes Berg [Wed, 10 Sep 2008 22:01:48 +0000 (00:01 +0200)]
mac80211: dont set REQ_RUN when scan finishes

The timer restart is done wrongly, we shouldn't set the REQ_RUN
bit when the scan has finished if it hadn't been set before the
scan started. If the timer fires during the scan, it will set
REQ_RUN and then we can run the work for it, if it didn't fire
then we shouldn't run its work either.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: move STA timer restart
Johannes Berg [Wed, 10 Sep 2008 22:01:47 +0000 (00:01 +0200)]
mac80211: move STA timer restart

This I shouldn't have moved to the scan implementation, move
it back to the MLME where it belongs, to the notification.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: move ieee80211_sta_expire
Johannes Berg [Wed, 10 Sep 2008 22:01:46 +0000 (00:01 +0200)]
mac80211: move ieee80211_sta_expire

ieee80211_sta_expire uses the internal __sta_info_unlink
function which can become static if this function is moved
to sta_info.c.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agolibertas: convert SNMP_MIB to a direct command
Dan Williams [Wed, 10 Sep 2008 16:49:00 +0000 (12:49 -0400)]
libertas: convert SNMP_MIB to a direct command

And support setting both long and short retries independently.

Signed-off-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoath9k: Add RF kill support
Vasanthakumar Thiagarajan [Wed, 10 Sep 2008 13:20:17 +0000 (18:50 +0530)]
ath9k: Add RF kill support

RF kill support is enabled when CONFIG_RFKILL
is set.

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoath9k: Re-order functions in main.c
Vasanthakumar Thiagarajan [Wed, 10 Sep 2008 13:19:27 +0000 (18:49 +0530)]
ath9k: Re-order functions in main.c

Some of the functions in main.c are re-ordered in such
a way that all local functions are defined before mac80211
and pci callbacks.

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agolibertas: remove unused generic RESET command
Dan Williams [Wed, 10 Sep 2008 13:04:33 +0000 (09:04 -0400)]
libertas: remove unused generic RESET command

The generic reset command is unused.  Each interface type needs to
handle the reset command differently since after reset, the firmware is
dead and interface-specific mechanisms must be used to reinitialize the
card.

Signed-off-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agozd1211rw: make use of new regulatory_hint()
Luis R. Rodriguez [Wed, 10 Sep 2008 06:19:49 +0000 (23:19 -0700)]
zd1211rw: make use of new regulatory_hint()

This cleans up zd1211rw's own regulatory work, and makes use of
the new cfg80211 regulatory_hint().

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agocfg80211: Add new wireless regulatory infrastructure
Luis R. Rodriguez [Wed, 10 Sep 2008 06:19:48 +0000 (23:19 -0700)]
cfg80211: Add new wireless regulatory infrastructure

This adds the new wireless regulatory infrastructure. The
main motiviation behind this was to centralize regulatory
code as each driver was implementing their own regulatory solution,
and to replace the initial centralized code we have where:

* only 3 regulatory domains are supported: US, JP and EU
* regulatory domains can only be changed through module parameter
* all rules were built statically in the kernel

We now have support for regulatory domains for many countries
and regulatory domains are now queried through a userspace agent
through udev allowing distributions to update regulatory rules
without updating the kernel.

Each driver can regulatory_hint() a regulatory domain
based on either their EEPROM mapped regulatory domain value to a
respective ISO/IEC 3166-1 country code or pass an internally built
regulatory domain. We also add support to let the user set the
regulatory domain through userspace in case of faulty EEPROMs to
further help compliance.

Support for world roaming will be added soon for cards capable of
this.

For more information see:

http://wireless.kernel.org/en/developers/Regulatory/CRDA

For now we leave an option to enable the old module parameter,
ieee80211_regdom, and to build the 3 old regdomains statically
(US, JP and EU). This option is CONFIG_WIRELESS_OLD_REGULATORY.
These old static definitions and the module parameter is being
scheduled for removal for 2.6.29. Note that if you use this
you won't make use of a world regulatory domain as its pointless.
If you leave this option enabled and if CRDA is present and you
use US or JP we will try to ask CRDA to update us a regulatory
domain for us.

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agonet: ip_vs_proto_{tcp,udp} build fix
Stephen Rothwell [Sat, 13 Sep 2008 06:23:50 +0000 (23:23 -0700)]
net: ip_vs_proto_{tcp,udp} build fix

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agomultiq: Further multiqueue cleanup
Alexander Duyck [Sat, 13 Sep 2008 00:57:23 +0000 (17:57 -0700)]
multiq: Further multiqueue cleanup

This patch resolves a few issues found with multiq including wording
suggestions and a problem seen in the allocation of queues.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoskbedit: Fix a typo in the documentation
Alexander Duyck [Sat, 13 Sep 2008 00:56:50 +0000 (17:56 -0700)]
skbedit: Fix a typo in the documentation

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agopkt_action: add new action skbedit
Alexander Duyck [Fri, 12 Sep 2008 23:30:20 +0000 (16:30 -0700)]
pkt_action: add new action skbedit

This new action will have the ability to change the priority and/or
queue_mapping fields on an sk_buff.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agopkt_sched: Add multiqueue scheduler support
Alexander Duyck [Fri, 12 Sep 2008 23:29:34 +0000 (16:29 -0700)]
pkt_sched: Add multiqueue scheduler support

This patch is intended to add a qdisc to support the new tx multiqueue
architecture by providing a band for each hardware queue.  By doing
this it is possible to support a different qdisc per physical hardware
queue.

This qdisc uses the skb->queue_mapping to select which band to place
the traffic onto.  It then uses a round robin w/ a check to see if the
subqueue is stopped to determine which band to dequeue the packet from.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agotcp_ipv6: fix use of uninitialized memory
Vegard Nossum [Fri, 12 Sep 2008 23:17:43 +0000 (16:17 -0700)]
tcp_ipv6: fix use of uninitialized memory

inet6_rsk() is called on a struct request_sock * before we
have checked whether the socket is an ipv6 socket or a ipv6-
mapped ipv4 socket. The access that triggers this is the
inet_rsk(rsk)->inet6_rsk_offset dereference in inet6_rsk().

This is arguably not a critical error as the inet6_rsk_offset
is only used to compute a pointer which is never really used
(in the code path in question) anyway. But it might be a
latent error, so let's fix it.

Spotted by kmemcheck.

Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agonet: fix scheduling of dst_gc_task by __dst_free
Benjamin Thery [Fri, 12 Sep 2008 23:16:37 +0000 (16:16 -0700)]
net: fix scheduling of dst_gc_task by __dst_free

The dst garbage collector dst_gc_task() may not be scheduled as we
expect it to be in __dst_free().

Indeed, when the dst_gc_timer was replaced by the delayed_work
dst_gc_work, the mod_timer() call used to schedule the garbage
collector at an earlier date was replaced by a schedule_delayed_work()
(see commit 86bba269d08f0c545ae76c90b56727f65d62d57f).

But, the behaviour of mod_timer() and schedule_delayed_work() is
different in the way they handle the delay.

mod_timer() stops the timer and re-arm it with the new given delay,
whereas schedule_delayed_work() only check if the work is already
queued in the workqueue (and queue it (with delay) if it is not)
BUT it does NOT take into account the new delay (even if the new delay
is earlier in time).
schedule_delayed_work() returns 0 if it didn't queue the work,
but we don't check the return code in __dst_free().

If I understand the code in __dst_free() correctly, we want dst_gc_task
to be queued after DST_GC_INC jiffies if we pass the test (and not in
some undetermined time in the future), so I think we should add a call
to cancel_delayed_work() before schedule_delayed_work(). Patch below.

Or we should at least test the return code of schedule_delayed_work(),
and reset the values of dst_garbage.timer_inc and dst_garbage.timer_expires
back to their former values if schedule_delayed_work() failed.
Otherwise the subsequent calls to __dst_free will test the wrong values
and assume wrong thing about when the garbage collector is supposed to
be scheduled.

dst_gc_task() also calls schedule_delayed_work() without checking
its return code (or calling cancel_scheduled_work() first), but it
should fine there: dst_gc_task is the routine of the delayed_work, so
no dst_gc_work should be pending in the queue when it's running.

Signed-off-by: Benjamin Thery <benjamin.thery@bull.net>
Acked-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agovlan: vlan device not reading gso max size of parent.
Alexander Duyck [Fri, 12 Sep 2008 03:17:05 +0000 (20:17 -0700)]
vlan: vlan device not reading gso max size of parent.

The vlan devices are not reading the gso max size of the parent device.  As
a result devices that do not support 64K max gso size are currently
failing.

This issue is seen on 2.6.26 kernels as well and the same patch should be
able to be applied without any issues.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoMerge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
David S. Miller [Thu, 11 Sep 2008 23:08:24 +0000 (16:08 -0700)]
Merge branch 'master' of /linux/kernel/git/davem/net-2.6

Conflicts:

drivers/net/wireless/ath9k/beacon.c
drivers/net/wireless/ath9k/core.h

16 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Thu, 11 Sep 2008 22:46:02 +0000 (15:46 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next-2.6

16 years agotg3: Fix DMA mapping leak in tigon3_dma_hwbug_workaround().
David S. Miller [Thu, 11 Sep 2008 22:45:19 +0000 (15:45 -0700)]
tg3: Fix DMA mapping leak in tigon3_dma_hwbug_workaround().

Noticed by Michael Chan.

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agolibertas: Improvements on automatic tx power control via SIOCSIWTXPOW.
Anna Neal [Thu, 11 Sep 2008 18:17:25 +0000 (11:17 -0700)]
libertas: Improvements on automatic tx power control via SIOCSIWTXPOW.

iwconfig txpower can now be used to set tx power to fixed or auto. If set to
auto the default firmware settings are used.

The command CMD_802_11_PA_CFG is only sent to older firmware, as Dan Williams
noted the command was no longer supported in firmware V9+.

Signed-off-by: Anna Neal <anna@cozybit.com>
Signed-off-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: Reorder debugfs calls during netdev deinit
Jouni Malinen [Tue, 9 Sep 2008 14:33:15 +0000 (16:33 +0200)]
mac80211: Reorder debugfs calls during netdev deinit

ieee80211_free_keys() must be called before
ieee80211_debugfs_remove_netdev() in order to make sure that the
possible default_key symlink is removed before attempting to
remove the netdev debugfs directory.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: move frame TX function
Johannes Berg [Tue, 9 Sep 2008 13:07:09 +0000 (15:07 +0200)]
mac80211: move frame TX function

The ieee80211_sta_tx function isn't MLME code any more,
it's getting used by a lot of code. Move it to utils and
rename it to ieee80211_tx_skb.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: make ieee80211_rx_h_mgmt more readable
Johannes Berg [Tue, 9 Sep 2008 12:55:09 +0000 (14:55 +0200)]
mac80211: make ieee80211_rx_h_mgmt more readable

That function isn't exactly easy to read especially since it
does something in an if branch that continues after the if
because the else returns. Express it in a more readable way.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agomac80211: move spectrum management code out
Johannes Berg [Tue, 9 Sep 2008 12:49:03 +0000 (14:49 +0200)]
mac80211: move spectrum management code out

Like the HT code, this doesn't depend on the STA-mode implementation
and can be handled entirely independently. There's only stub code
for now, but when it gets filled having it in its own file will be
beneficial.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>