safe/jmp/linux-2.6
14 years agotcp: fix tcp_defer_accept to consider the timeout
Willy Tarreau [Tue, 13 Oct 2009 07:27:40 +0000 (00:27 -0700)]
tcp: fix tcp_defer_accept to consider the timeout

I was trying to use TCP_DEFER_ACCEPT and noticed that if the
client does not talk, the connection is never accepted and
remains in SYN_RECV state until the retransmits expire, where
it finally is deleted. This is bad when some firewall such as
netfilter sits between the client and the server because the
firewall sees the connection in ESTABLISHED state while the
server will finally silently drop it without sending an RST.

This behaviour contradicts the man page which says it should
wait only for some time :

       TCP_DEFER_ACCEPT (since Linux 2.4)
          Allows a listener to be awakened only when data arrives
          on the socket.  Takes an integer value  (seconds), this
          can  bound  the  maximum  number  of attempts TCP will
          make to complete the connection. This option should not
          be used in code intended to be portable.

Also, looking at ipv4/tcp.c, a retransmit counter is correctly
computed :

        case TCP_DEFER_ACCEPT:
                icsk->icsk_accept_queue.rskq_defer_accept = 0;
                if (val > 0) {
                        /* Translate value in seconds to number of
                         * retransmits */
                        while (icsk->icsk_accept_queue.rskq_defer_accept < 32 &&
                               val > ((TCP_TIMEOUT_INIT / HZ) <<
                                       icsk->icsk_accept_queue.rskq_defer_accept))
                                icsk->icsk_accept_queue.rskq_defer_accept++;
                        icsk->icsk_accept_queue.rskq_defer_accept++;
                }
                break;

==> rskq_defer_accept is used as a counter of retransmits.

But in tcp_minisocks.c, this counter is only checked. And in
fact, I have found no location which updates it. So I think
that what was intended was to decrease it in tcp_minisocks
whenever it is checked, which the trivial patch below does.

Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years ago3c574_cs: spin_lock the set_multicast_list function
Ken Kawasaki [Tue, 13 Oct 2009 07:32:55 +0000 (00:32 -0700)]
3c574_cs: spin_lock the set_multicast_list function

3c574_cs:
 spin_lock the set_multicast_list function.

Signed-off-by: Ken Kawasaki <ken_kawasaki@spring.nifty.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Teach pegasus driver to ignore bluetoother adapters with clashing Vendor:Product IDs
Chris Rankin [Tue, 13 Oct 2009 07:32:02 +0000 (00:32 -0700)]
net: Teach pegasus driver to ignore bluetoother adapters with clashing Vendor:Product IDs

The Belkin F8T012xx1 bluetooth adaptor has the same vendor and product
IDs as the Belkin F5D5050, so we need to teach the pegasus driver to
ignore adaptors belonging to the "Wireless" class 0xE0. For this one
case anyway, seeing as pegasus is a driver for "Wired" adaptors.

Signed-off-by: Chris Rankin <rankincj@yahoo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonetxen: fix pci bar mapping
Dhananjay Phadke [Tue, 13 Oct 2009 07:26:04 +0000 (00:26 -0700)]
netxen: fix pci bar mapping

Use resource_size_t for PCI resource remapping instead
of unsigned long. Physical addresses can exceed range of
long data type (e.g with PAE).

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoethoc: fix warning from 32bit build
Alan Cox [Mon, 12 Oct 2009 05:27:55 +0000 (05:27 +0000)]
ethoc: fix warning from 32bit build

drivers/net/ethoc.c: In function ‘ethoc_open’:
drivers/net/ethoc.c:667: warning: comparison of distinct pointer types
lacks a cast

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agolibertas: fix build
Alan Cox [Mon, 12 Oct 2009 05:27:48 +0000 (05:27 +0000)]
libertas: fix build

drivers/net/wireless/libertas/cmdresp.c: In function ‘lbs_process_event’:
drivers/net/wireless/libertas/cmdresp.c:519: error: ‘TASK_INTERRUPTIBLE’
undeclared (first use in this function)
drivers/net/wireless/libertas/cmdresp.c:519: error: (Each undeclared
identifier is reported only once
drivers/net/wireless/libertas/cmdresp.c:519: error: for each function it
appears in.)

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: VMware virtual Ethernet NIC driver: vmxnet3
Shreyas Bhatewara [Tue, 13 Oct 2009 07:15:51 +0000 (00:15 -0700)]
net: VMware virtual Ethernet NIC driver: vmxnet3

Ethernet NIC driver for VMware's vmxnet3

From: Shreyas Bhatewara <sbhatewara@vmware.com>

This patch adds driver support for VMware's virtual Ethernet NIC: vmxnet3
Guests running on VMware hypervisors supporting vmxnet3 device will thus have
access to improved network functionalities and performance.

Signed-off-by: Shreyas Bhatewara <sbhatewara@vmware.com>
Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: Ronghua Zhang <ronghua@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Fix IXP 2000 network driver building.
Vincent Sanders [Tue, 13 Oct 2009 07:03:11 +0000 (00:03 -0700)]
net: Fix IXP 2000 network driver building.

The IXP 2000 network driver was failing to build as it has its own
statistics gathering which was not compatible with the recent network
device operations changes. This patch fixes the driver in the obvious
way and has been compile tested. I have been unable to get the ixp2000
maintainer to comment or test this fix.

Signed-off-by: Vincent Sanders <vince@simtec.co.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoacenic: Pass up error code from ace_load_firmware()
Ben Hutchings [Mon, 12 Oct 2009 11:18:48 +0000 (04:18 -0700)]
acenic: Pass up error code from ace_load_firmware()

If ace_load_firmware() fails, ace_init() cleans up but still returns
0, leading to an oops as seen in <http://bugs.debian.org/521383>.
It should pass the error code up.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Fix struct sock bitfield annotation
Eric Dumazet [Thu, 8 Oct 2009 22:50:25 +0000 (22:50 +0000)]
net: Fix struct sock bitfield annotation

Since commit a98b65a3 (net: annotate struct sock bitfield), we lost
8 bytes in struct sock on 64bit arches because of
kmemcheck_bitfield_end(flags) misplacement.

Fix this by putting together sk_shutdown, sk_no_check, sk_userlocks,
sk_protocol and sk_type in the 'flags' 32bits bitfield

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agopkt_sched: pedit use proper struct
jamal [Sun, 11 Oct 2009 04:21:38 +0000 (04:21 +0000)]
pkt_sched: pedit use proper struct

This probably deserves to go into -stable.

Pedit will reject a policy that is large because it
uses the wrong structure in the policy validation.
This fixes it.

Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Add patchwork URL to MAINTAINERS
Eric Dumazet [Sat, 10 Oct 2009 21:40:54 +0000 (21:40 +0000)]
net: Add patchwork URL to MAINTAINERS

Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Link in PHY drivers before others.
David S. Miller [Fri, 9 Oct 2009 21:24:36 +0000 (14:24 -0700)]
net: Link in PHY drivers before others.

We need PHY drivers to initialize in a static kernel before
the MAC drivers that use them.  So link them in first.

Based upon a report by Felix Radensky.

Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoibm_newemac: Added 16K Tx FIFO size support for EMAC4
Dave Mitchell [Thu, 8 Oct 2009 06:32:21 +0000 (06:32 +0000)]
ibm_newemac: Added 16K Tx FIFO size support for EMAC4

Some of the EMAC V4 implementations support 16K Tx FIFOs. This
patch adds support for this functionality and fixes typos in the
Tx FIFO size error messages.

Signed-off-by: Dave Mitchell <dmitchell@appliedmicro.com>
Acked-by: Prodyut Hazarika <phazarika@appliedmicro.com>
Acked-by: Victor Gallardo <vgallardo@appliedmicro.com>
Acked-by: Loc Ho <lho@appliedmicro.com>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Add disable/enable firmare irqs to handler.
Ron Mercer [Thu, 8 Oct 2009 09:54:43 +0000 (09:54 +0000)]
qlge: Add disable/enable firmare irqs to handler.

This was accidentally omitted from one of the previous patches for firmware event
handling.  The handler needs to the enable firmware irq mask when it's done
processing or it may not get any more events interrupts.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Restore rx mode after internal reset.
Ron Mercer [Thu, 8 Oct 2009 09:54:42 +0000 (09:54 +0000)]
qlge: Restore rx mode after internal reset.

Call set_multi API after reset recovery.  This was exposed by tripping
tx_timeout.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Fix chip reset process.
Ron Mercer [Thu, 8 Oct 2009 09:54:41 +0000 (09:54 +0000)]
qlge: Fix chip reset process.

Add wait for NIC fifo and MGMNT fifo to empty before applying reset.
Otherwise broken frames can be processed by management processor and
cause it to hang.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Fix RX multicast filter settings.
Ron Mercer [Thu, 8 Oct 2009 09:54:40 +0000 (09:54 +0000)]
qlge: Fix RX multicast filter settings.

The addresses were being added to the filter properly, but were not
being enabled.  This adds enable bit to filter write.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Fix frame routing for multicast frames.
Ron Mercer [Thu, 8 Oct 2009 09:54:39 +0000 (09:54 +0000)]
qlge: Fix frame routing for multicast frames.

Broadcast/multicast should always be routed to the default (zeroeth) rx
ring.  Broadcast frames are already routed correctly.  This fixes
routing for multicast frames.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Fix frame routing issue related to bonding.
Ron Mercer [Thu, 8 Oct 2009 09:54:38 +0000 (09:54 +0000)]
qlge: Fix frame routing issue related to bonding.

Currently frames are routed based on their type and MAC address. This
patch adds the port number on which the frame arrived to the routing.
This prevents problems in the case where both interfaces have the same MAC address in
a routing configuration.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Fix RSS hashing values.
Ron Mercer [Thu, 8 Oct 2009 09:54:37 +0000 (09:54 +0000)]
qlge: Fix RSS hashing values.

Fix RX queue table size and change from random to default hash values.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Thu, 8 Oct 2009 22:55:21 +0000 (15:55 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6

14 years agocfg80211: fix netns error unwinding bug
Johannes Berg [Thu, 8 Oct 2009 19:02:02 +0000 (21:02 +0200)]
cfg80211: fix netns error unwinding bug

The error unwinding code in set_netns has a bug
that will make it run into a BUG_ON if passed a
bad wiphy index, fix by not trying to unlock a
wiphy that doesn't exist.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agonetlink: fix typo in initialization
Jiri Pirko [Thu, 8 Oct 2009 08:21:46 +0000 (01:21 -0700)]
netlink: fix typo in initialization

Commit 9ef1d4c7c7aca1cd436612b6ca785b726ffb8ed8 ("[NETLINK]: Missing
initializations in dumped data") introduced a typo in
initialization. This patch fixes this.

Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoznet: Don't claim DMA lock around free_dma() calls.
David S. Miller [Thu, 8 Oct 2009 05:15:23 +0000 (22:15 -0700)]
znet: Don't claim DMA lock around free_dma() calls.

It's not necessary and it's illegal too.

Reported-by: Alexander Strakh <strakh@ispras.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoinclude/linux/netdevice.h: fix nanodoc mismatch
Wolfram Sang [Wed, 7 Oct 2009 20:53:11 +0000 (13:53 -0700)]
include/linux/netdevice.h: fix nanodoc mismatch

nanodoc was missing an ndo_-prefix.

Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agob43: do not stack-allocate pio rx/tx header and tail buffers
Albert Herranz [Tue, 6 Oct 2009 22:07:44 +0000 (00:07 +0200)]
b43: do not stack-allocate pio rx/tx header and tail buffers

The DMA-API debugging facility complains about b43 mapping memory from
stack for SDIO-based cards.

Indeed, b43 currently allocates the PIO RX/TX header and tail buffers
from stack. The solution here is to use heap-allocated buffers instead.

Signed-off-by: Albert Herranz <albert_herranz@yahoo.es>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agob43: Fix PPC crash in rfkill polling on unload
Larry Finger [Fri, 2 Oct 2009 17:54:34 +0000 (12:54 -0500)]
b43: Fix PPC crash in rfkill polling on unload

In Bugzilla No. 14181, a PowerMac G4 crashes on ifdown or
module unload because the rfkill polling has not been stopped.
For the x86 architectures, the attempt to reach a now unmapped
register is not fatal as it is on PPC.

(Includes "b43: Fix locking problem when stopping rfkill polling". -- JWL)

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agomac80211: use kfree_skb() to free struct sk_buff pointers
Roel Kluin [Tue, 6 Oct 2009 13:52:35 +0000 (15:52 +0200)]
mac80211: use kfree_skb() to free struct sk_buff pointers

kfree_skb() should be used to free struct sk_buff pointers.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agomac80211: fix ADM8211_SYNCTL_RFtype define
Roel Kluin [Tue, 6 Oct 2009 12:33:14 +0000 (14:33 +0200)]
mac80211: fix ADM8211_SYNCTL_RFtype define

A logical of shifts to the left doesn't make sense.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoiwlwifi: use %pM for formatted MAC addresses
Bjorn Helgaas [Mon, 5 Oct 2009 17:37:01 +0000 (11:37 -0600)]
iwlwifi: use %pM for formatted MAC addresses

User-visible messages should use formatted MAC addresses ("00:01:...")
rather than raw ("0001...") so they match other parts of the system.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
CC: ilw@linux.intel.com
CC: linux-wireless@vger.kernel.org
Acked-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agob43: Protect sanity check against physical device removal
Michael Buesch [Fri, 2 Oct 2009 22:57:58 +0000 (00:57 +0200)]
b43: Protect sanity check against physical device removal

Fix IRQ mask sanity check for physically pulled device.

Tested-by: Andrew Price <andy@andrewprice.me.uk>
Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoiwlagn: fix compile warning in iwl5000_gain_computation
Reinette Chatre [Fri, 2 Oct 2009 20:44:07 +0000 (13:44 -0700)]
iwlagn: fix compile warning in iwl5000_gain_computation

The return type of abs() was recently changed from int to long. With
min()'s type checking we thus need to make sure that values of the same
type are compared.

This fixes:

    CC [M]  drivers/net/wireless/iwlwifi/iwl-5000.o
drivers/net/wireless/iwlwifi/iwl-5000.c: In function ‘iwl5000_gain_computation’:
drivers/net/wireless/iwlwifi/iwl-5000.c:320: warning: comparison of distinct pointer types lacks a cast

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Reported-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoiwlwifi: fix compile warning
Reinette Chatre [Fri, 2 Oct 2009 20:43:57 +0000 (13:43 -0700)]
iwlwifi: fix compile warning

Fixes following on big endian systems:
  CC [M]  drivers/net/wireless/iwlwifi/iwl-rx.o
  drivers/net/wireless/iwlwifi/iwl-rx.c: In function ‘iwl_rx_reply_rx’:
  drivers/net/wireless/iwlwifi/iwl-rx.c:1029: warning: integer overflow in
  expression

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Reported-by: Marcel Holtmann <marcel@holtmann.org>
Tested-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoiwlwifi: fix EEPROM enhance tx power offset
Wey-Yi Guy [Fri, 2 Oct 2009 20:43:56 +0000 (13:43 -0700)]
iwlwifi: fix EEPROM enhance tx power offset

Set the correct EEPROM offset for enhance tx power for 6000 series

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoiwlwifi: incorrect method used for finding valid OTP blocks
Jay Sternberg [Fri, 2 Oct 2009 20:43:55 +0000 (13:43 -0700)]
iwlwifi: incorrect method used for finding valid OTP blocks

The address stored in the next link address is a word address but when
reading the OTP blocks, a byte address is used. Also if the blocks are
full and the last link pointer is not zero, then none of the blocks are
valid so return an error.

The algorithm is simply valid blocks have a next address and that
address's contents is zero.

Using the wrong address for the next link address gets arbitrary data,
obviously. In cases seen, the first block is considered valid when it is not.

If the block has in fact been invalidated there may be old data or
there may be no data, bad data, or partial data, there is no way of
telling. Without this patch it is possible that a device with valid OTP data
is unable to work.

Signed-off-by: Jay Sternberg <jay.e.sternberg@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
CC: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agomac80211: fix vlan and optimise RX
Johannes Berg [Thu, 1 Oct 2009 20:06:29 +0000 (22:06 +0200)]
mac80211: fix vlan and optimise RX

When receiving data frames, we can send them only to
the interface they belong to based on transmitting
station (this doesn't work for probe requests). Also,
don't try to handle other frames for AP_VLAN at all
since those interface should only receive data.

Additionally, the transmit side must check that the
station we're sending a frame to is actually on the
interface we're transmitting on, and not transmit
packets to functions that live on other interfaces,
so validate that as well.

Another bug fix is needed in sta_info.c where in the
VLAN case when adding/removing stations we overwrite
the sdata variable we still need.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agob43: Don't use struct wldev after detach.
Michael Buesch [Thu, 1 Oct 2009 13:54:32 +0000 (15:54 +0200)]
b43: Don't use struct wldev after detach.

Don't use struct wldev after detach. This fixes an oops on access.

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoethoc: limit the number of buffers to 128
Thomas Chou [Tue, 6 Oct 2009 03:25:25 +0000 (03:25 +0000)]
ethoc: limit the number of buffers to 128

Only 128 buffer descriptors are supported in the core. Limit the
number in case we have more memory.

Signed-off-by: Thomas Chou <thomas@wytron.com.tw>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoethoc: use system memory as buffer
Thomas Chou [Sun, 4 Oct 2009 23:33:20 +0000 (23:33 +0000)]
ethoc: use system memory as buffer

This patch enabled the ethoc to allocate system memory as buffer
when there is no dedicated buffer memory.

Some hardware designs may not have dedicated buffer memory such as
on chip or off chip SRAM. In this case, only one memory resource is
supplied in the platform data instead of two. Then a DMA buffer can
be allocated from system memory and used for the transfer.

Signed-off-by: Thomas Chou <thomas@wytron.com.tw>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoethoc: align received packet to make IP header at word boundary
Thomas Chou [Sun, 4 Oct 2009 23:33:19 +0000 (23:33 +0000)]
ethoc: align received packet to make IP header at word boundary

The packet buffer is allocated at 4 bytes boundary, but the IP header
length and version bits is located at byte 14. These bit fields access
as 32 bits word and caused exception on processors that do not support
unaligned access.

The patch adds 2 bytes offset to make the bit fields word aligned.

Signed-off-by: Thomas Chou <thomas@wytron.com.tw>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoethoc: fix buffer address mapping
Thomas Chou [Sun, 4 Oct 2009 23:33:18 +0000 (23:33 +0000)]
ethoc: fix buffer address mapping

The pointer address in buffer descriptors is physical address. The
pointer that processor used to access packet is virtual address.

Though the higher bits of pointer address used by the MAC may be
truncated to zero in special case, it is not always true in larger
designs.

Signed-off-by: Thomas Chou <thomas@wytron.com.tw>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoethoc: fix typo to compute number of tx descriptors
Thomas Chou [Sun, 4 Oct 2009 23:33:17 +0000 (23:33 +0000)]
ethoc: fix typo to compute number of tx descriptors

It should be max() instead of min(). Use 1/4 of available
descriptors for tx, and there should be at least 2 tx
descriptors.

Signed-off-by: Thomas Chou <thomas@wytron.com.tw>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoau1000_eth: Duplicate test of RX_OVERLEN bit in update_rx_stats()
roel kluin [Tue, 6 Oct 2009 09:54:18 +0000 (09:54 +0000)]
au1000_eth: Duplicate test of RX_OVERLEN bit in update_rx_stats()

in update_rx_stats() the RX_OVERLEN bit is set twice, replace it by RX_RUNT.
in au1000_rx() the RX_MISSED_FRAME bit was tested a few lines earlier already

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Manuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonetxen: Fix Unlikely(x) > y
Roel Kluin [Tue, 6 Oct 2009 19:34:39 +0000 (19:34 +0000)]
netxen: Fix Unlikely(x) > y

The closing parenthesis was not on the right location.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agopasemi_mac: ethtool get settings fix
Valentine Barshak [Mon, 5 Oct 2009 03:27:56 +0000 (03:27 +0000)]
pasemi_mac: ethtool get settings fix

Not all pasemi mac interfaces can have a phy attached.
For example, XAUI has no phy and phydev is NULL for it.
In this case ethtool get settings causes kernel crash.
Fix it by returning -EOPNOTSUPP if there's no PHY attached.

Signed-off-by: Valentine Barshak <vbarshak@ru.mvista.com>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoadd maintainer for network drop monitor kernel service
Neil Horman [Mon, 5 Oct 2009 03:56:55 +0000 (03:56 +0000)]
add maintainer for network drop monitor kernel service

I was getting ribbed about this earlier, so I figured I'd make it
official.  Add myself as the maintainer of the drop monitor bits, so
people don't just gripe at Dave when it breaks (I'm sure it will never
break, but just in case :) ).

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotg3: Fix phylib locking strategy
Matt Carlson [Mon, 5 Oct 2009 17:55:29 +0000 (17:55 +0000)]
tg3: Fix phylib locking strategy

Felix Radensky noted that chip resets were generating stack trace dumps.
This is because the driver is attempting to acquire the mdio bus mutex
while holding the tp->lock spinlock.  The fix is to change the code such
that every phy access takes the tp->lock spinlock instead.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agorndis_host: support ETHTOOL_GPERMADDR
John W. Linville [Tue, 6 Oct 2009 11:05:42 +0000 (11:05 +0000)]
rndis_host: support ETHTOOL_GPERMADDR

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoipv4: arp_notify address list bug
Stephen Hemminger [Wed, 7 Oct 2009 10:18:17 +0000 (03:18 -0700)]
ipv4: arp_notify address list bug

This fixes a bug with arp_notify.

If arp_notify is enabled, kernel will crash if address is changed
and no IP address is assigned.
  http://bugzilla.kernel.org/show_bug.cgi?id=14330

Reported-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogigaset: add kerneldoc comments
Tilman Schmidt [Tue, 6 Oct 2009 12:19:01 +0000 (12:19 +0000)]
gigaset: add kerneldoc comments

Add kerneldoc comments to some functions in the Gigaset driver.

Impact: documentation
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogigaset: correct debugging output selection
Tilman Schmidt [Tue, 6 Oct 2009 12:18:56 +0000 (12:18 +0000)]
gigaset: correct debugging output selection

Dump payload data consistently only when DEBUG_STREAM_DUMP debug bit
is set.

Impact: debugging aid
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogigaset: improve error recovery
Tilman Schmidt [Tue, 6 Oct 2009 12:18:51 +0000 (12:18 +0000)]
gigaset: improve error recovery

When the Gigaset base stops responding, try resetting the USB
connection to recover.

Impact: error handling improvement
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogigaset: fix device ERROR response handling
Tilman Schmidt [Tue, 6 Oct 2009 12:18:46 +0000 (12:18 +0000)]
gigaset: fix device ERROR response handling

Clear out pending command that got rejected with 'ERROR' response.
This fixes the bug where unloading the driver module would hang
with the message: "gigaset: not searching scheduled commands: busy"
after a device communication error.

Impact: error handling bugfix
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogigaset: announce if built with debugging
Tilman Schmidt [Tue, 6 Oct 2009 12:18:41 +0000 (12:18 +0000)]
gigaset: announce if built with debugging

Mention in the driver load announcement whether the driver was built
with debugging messages enabled, to facilitate support.

Impact: informational message
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogigaset: handle isoc frame errors more gracefully
Tilman Schmidt [Tue, 6 Oct 2009 12:18:36 +0000 (12:18 +0000)]
gigaset: handle isoc frame errors more gracefully

Don't drop the remainder of an URB if an isochronous frame has an error.

Impact: error handling improvement
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogigaset: linearize skb
Tilman Schmidt [Tue, 6 Oct 2009 12:18:31 +0000 (12:18 +0000)]
gigaset: linearize skb

The code of the Gigaset driver assumes that sk_buff-s coming
from the ISDN4Linux subsystem are always linear. Explicitly
calling skb_linearize() is cheap if they are, but much more
robust in case they ever aren't.

Impact: robustness improvement
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogigaset: fix reject/hangup handling
Tilman Schmidt [Tue, 6 Oct 2009 12:18:26 +0000 (12:18 +0000)]
gigaset: fix reject/hangup handling

Signal D channel disconnect in a few cases where it was missed,
including when an incoming call is disconnected before it was
accepted.

Impact: error handling improvement
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoisdn: make capidrv module parameter "debugmode" writeable
Tilman Schmidt [Tue, 6 Oct 2009 12:18:15 +0000 (12:18 +0000)]
isdn: make capidrv module parameter "debugmode" writeable

Being able to change the debugmode module parameter of capidrv on the
fly is quite useful for debugging and doesn't do any harm.

Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoisdn: avoid races in capidrv
Tilman Schmidt [Tue, 6 Oct 2009 12:18:10 +0000 (12:18 +0000)]
isdn: avoid races in capidrv

In several places, capidrv sends a CAPI message to the ISDN
device and then updates its internal state accordingly.
If the response message from the device arrives before the
state is updated, it may be rejected or processed incorrectly.
Avoid these races by updating the state before emitting the
message.

Impact: bugfix
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Acked-by: Karsten Keil <keil@b1-systems.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoisdn: accept CAPI Informational Info values as success
Tilman Schmidt [Tue, 6 Oct 2009 12:18:05 +0000 (12:18 +0000)]
isdn: accept CAPI Informational Info values as success

Info values in the 0x00xx range are defined in the CAPI standard
as "Informational, message processed successfully". Therefore a
CONNECT_B3_CONF message with an Info value in that range should
open an NCCI just as with Info==0.

Impact: minor bugfix
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Acked-by: Karsten Keil <keil@b1-systems.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoDocumentation: expand isdn/INTERFACE.CAPI document
Tilman Schmidt [Tue, 6 Oct 2009 12:18:00 +0000 (12:18 +0000)]
Documentation: expand isdn/INTERFACE.CAPI document

- Note that send_message() may be called in interrupt context.
- Describe the storage of CAPI messages and payload data in SKBs.
- Add more details to the description of the _cmsg structure.
- Describe kernelcapi debugging output.

Impact: documentation
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: Bug fix to properly update ethtool tx-checksumming after ethtool -K <ifname...
Ajit Khaparde [Mon, 5 Oct 2009 02:22:19 +0000 (02:22 +0000)]
be2net: Bug fix to properly update ethtool tx-checksumming after ethtool -K <ifname> tx off

This is a fix for a bug which was a result of wrong use of checksum offload flag.
The status of tx-checksumming was not changed from on to off
after a 'ethtool -K <ifname> tx off' operation.
Use the proper checksum offload flag NETIF_F_HW_CSUM instead of
NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM.
Patch is against net-2.6 tree.

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: Fix a typo in be_cmds.h
Ajit Khaparde [Mon, 5 Oct 2009 02:22:05 +0000 (02:22 +0000)]
be2net: Fix a typo in be_cmds.h

MCC_STATUS_NOT_SUPPORTED should be decimal 66 not hex 66.
This patch fixes this typo. Patch against net-2.6 tree.

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: Bug Fix while accounting of multicast frames during netdev stats update
Ajit Khaparde [Mon, 5 Oct 2009 02:21:51 +0000 (02:21 +0000)]
be2net: Bug Fix while accounting of multicast frames during netdev stats update

While updating the statistics to be passed via the get_stats,
tx multicast frames were being accounted instead of rx multicast frames.
This patch fixes the bug. This patch is against the net-2.6 tree.

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Fix lock/mutex warnings.
Ron Mercer [Mon, 5 Oct 2009 11:46:49 +0000 (11:46 +0000)]
qlge: Fix lock/mutex warnings.

Get rid of spinlock and private mutex usage for exclusive access to the
HW semaphore register.  rtnl_lock already creates exclusive access to
this register in all driver API.
Add rtnl to firmware worker threads that also use the HW semaphore register.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Fix queueing of firmware handler in ISR.
Ron Mercer [Mon, 5 Oct 2009 11:46:48 +0000 (11:46 +0000)]
qlge: Fix queueing of firmware handler in ISR.

Check that we are not already polling firmware events before we queue the
firmware event worker, then disable firmware interrupts.
Otherwise we can queue the same event multiple times.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlge: Fix some bit definitions for reset register.
Ron Mercer [Mon, 5 Oct 2009 11:46:47 +0000 (11:46 +0000)]
qlge: Fix some bit definitions for reset register.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoconnector: Fix incompatible pointer type warning
Stephen Boyd [Tue, 6 Oct 2009 08:39:51 +0000 (01:39 -0700)]
connector: Fix incompatible pointer type warning

Commit 7069331 (connector: Provide the sender's credentials to the
callback, 2009-10-02) changed callbacks to take two arguments but missed
this one.

drivers/connector/cn_proc.c: In function ‘cn_proc_init’:
drivers/connector/cn_proc.c:263: warning: passing argument 3 of
‘cn_add_callback’ from incompatible pointer type

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agowext: let get_wireless_stats() sleep
Johannes Berg [Mon, 5 Oct 2009 09:22:23 +0000 (02:22 -0700)]
wext: let get_wireless_stats() sleep

A number of drivers (recently including cfg80211-based ones)
assume that all wireless handlers, including statistics, can
sleep and they often also implicitly assume that the rtnl is
held around their invocation. This is almost always true now
except when reading from sysfs:

  BUG: sleeping function called from invalid context at kernel/mutex.c:280
  in_atomic(): 1, irqs_disabled(): 0, pid: 10450, name: head
  2 locks held by head/10450:
   #0:  (&buffer->mutex){+.+.+.}, at: [<c10ceb99>] sysfs_read_file+0x24/0xf4
   #1:  (dev_base_lock){++.?..}, at: [<c12844ee>] wireless_show+0x1a/0x4c
  Pid: 10450, comm: head Not tainted 2.6.32-rc3 #1
  Call Trace:
   [<c102301c>] __might_sleep+0xf0/0xf7
   [<c1324355>] mutex_lock_nested+0x1a/0x33
   [<f8cea53b>] wdev_lock+0xd/0xf [cfg80211]
   [<f8cea58f>] cfg80211_wireless_stats+0x45/0x12d [cfg80211]
   [<c13118d6>] get_wireless_stats+0x16/0x1c
   [<c12844fe>] wireless_show+0x2a/0x4c

Fix this by using the rtnl instead of dev_base_lock.

Reported-by: Miles Lane <miles.lane@gmail.com>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: Fix a bug in preparation of mcc wrb which was causing flash operation to...
Ajit Khaparde [Mon, 5 Oct 2009 09:09:40 +0000 (02:09 -0700)]
be2net: Fix a bug in preparation of mcc wrb which was causing flash operation to fail

This patch fixes a bug that got introduced in commit 76998bc7.
During preparation of mcc wrb, req was being wrongly overwritten
and the flash operation was failing.

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agopcnet_cs: add cis of National Semicondoctor's multifunction pcmcia card
Ken Kawasaki [Mon, 5 Oct 2009 07:40:03 +0000 (00:40 -0700)]
pcnet_cs: add cis of National Semicondoctor's multifunction pcmcia card

pcnet_cs,serial_cs:

add cis of National Semicondoctor's lan&modem mulitifunction pcmcia card,
NE2K, tamarack ethernet card,
and some serial card(COMpad2, COMpad4).

Signed-off-by: Ken Kawasaki <ken_kawasaki@spring.nifty.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Support inclusion of <linux/socket.h> before <sys/socket.h>
Ben Hutchings [Mon, 5 Oct 2009 07:24:36 +0000 (00:24 -0700)]
net: Support inclusion of <linux/socket.h> before <sys/socket.h>

The following user-space program fails to compile:

    #include <linux/socket.h>
    #include <sys/socket.h>
    int main() { return 0; }

The reason is that <linux/socket.h> tests __GLIBC__ to decide whether it
should define various structures and macros that are now defined for
user-space by <sys/socket.h>, but __GLIBC__ is not defined if no libc
headers have yet been included.

It seems safe to drop support for libc 5 now.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Bastian Blank <waldi@debian.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agopktgen: restore nanosec delays
Eric Dumazet [Sat, 3 Oct 2009 01:39:18 +0000 (01:39 +0000)]
pktgen: restore nanosec delays

Commit fd29cf72 (pktgen: convert to use ktime_t)
inadvertantly converted "delay" parameter from nanosec to microsec.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agopktgen: Fix multiqueue handling
Eric Dumazet [Fri, 2 Oct 2009 20:24:59 +0000 (20:24 +0000)]
pktgen: Fix multiqueue handling

It is not currently possible to instruct pktgen to use one selected tx queue.

When Robert added multiqueue support in commit 45b270f8, he added
an interval (queue_map_min, queue_map_max), and his code doesnt take
into account the case of min = max, to select one tx queue exactly.

I suspect a high performance setup on a eight txqueue device wants
to use exactly eight cpus, and assign one tx queue to each sender.

This patchs makes pktgen select the right tx queue, not the first one.

Also updates Documentation to reflect Robert changes.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Robert Olsson <robert.olsson@its.uu.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoe1000e: swap max hw supported frame size between 82574 and 82583
Alexander Duyck [Fri, 2 Oct 2009 12:30:42 +0000 (12:30 +0000)]
e1000e: swap max hw supported frame size between 82574 and 82583

There appears to have been a mixup in the max supported jumbo frame size
between 82574 and 82583 which ended up disabling jumbo frames on the 82574
as a result.  This patch swaps the two so that this issue is resolved.

This patch fixes http://bugzilla.kernel.org/show_bug.cgi?id=14261

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>
14 years agoixgbe: add support for 82599 based X520 10G Dual KX4 Mezz card
Don Skidmore [Fri, 2 Oct 2009 08:58:25 +0000 (08:58 +0000)]
ixgbe: add support for 82599 based X520 10G Dual KX4 Mezz card

This patch adds device support for the 82599 based X520 10GbE
Dual Port KX4 Mezzanine card.

Signed-off-by: Don Skidmore<donald.c.skidmore@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoixgbe: add support for 82599 Combined Backplane
Don Skidmore [Fri, 2 Oct 2009 08:58:04 +0000 (08:58 +0000)]
ixgbe: add support for 82599 Combined Backplane

This patch will add support for the 82599 Dual port Backplane
device (0x10f8).  This device has the ability to link in serial (KR) and
parallel (KX4/KX) modes, depending on what the switch capabilities are in
the blade chassis.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocnic: Fix NETDEV_UP event processing.
Michael Chan [Fri, 2 Oct 2009 18:03:28 +0000 (11:03 -0700)]
cnic: Fix NETDEV_UP event processing.

This fixes the problem of not handling the NETDEV_UP event properly
during hot-plug or modprobe of bnx2 after cnic.  The handling was
skipped by mistakenly using "else if" to check for the event.

Also update version to 2.0.1.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agouvesafb/connector: Disallow unpliviged users to send netlink packets
Philipp Reisner [Fri, 2 Oct 2009 02:40:11 +0000 (02:40 +0000)]
uvesafb/connector: Disallow unpliviged users to send netlink packets

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agopohmelfs/connector: Disallow unpliviged users to configure pohmelfs
Philipp Reisner [Fri, 2 Oct 2009 02:40:10 +0000 (02:40 +0000)]
pohmelfs/connector: Disallow unpliviged users to configure pohmelfs

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agodst/connector: Disallow unpliviged users to configure dst
Philipp Reisner [Fri, 2 Oct 2009 02:40:09 +0000 (02:40 +0000)]
dst/connector: Disallow unpliviged users to configure dst

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agodm/connector: Only process connector packages from privileged processes
Philipp Reisner [Fri, 2 Oct 2009 02:40:08 +0000 (02:40 +0000)]
dm/connector: Only process connector packages from privileged processes

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoconnector: Removed the destruct_data callback since it is always kfree_skb()
Philipp Reisner [Fri, 2 Oct 2009 02:40:07 +0000 (02:40 +0000)]
connector: Removed the destruct_data callback since it is always kfree_skb()

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Acked-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Acked-by: Evgeniy Polyakov <zbr@ioremap.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoconnector/dm: Fixed a compilation warning
Philipp Reisner [Fri, 2 Oct 2009 02:40:06 +0000 (02:40 +0000)]
connector/dm: Fixed a compilation warning

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Acked-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Acked-by: Evgeniy Polyakov <zbr@ioremap.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoconnector: Provide the sender's credentials to the callback
Philipp Reisner [Fri, 2 Oct 2009 02:40:05 +0000 (02:40 +0000)]
connector: Provide the sender's credentials to the callback

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Acked-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Acked-by: Evgeniy Polyakov <zbr@ioremap.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoconnector: Keep the skb in cn_callback_data
Philipp Reisner [Fri, 2 Oct 2009 02:40:04 +0000 (02:40 +0000)]
connector: Keep the skb in cn_callback_data

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Acked-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Acked-by: Evgeniy Polyakov <zbr@ioremap.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoe1000e/igb/ixgbe: Don't report an error if devices don't support AER
Frans Pop [Fri, 2 Oct 2009 17:04:12 +0000 (10:04 -0700)]
e1000e/igb/ixgbe: Don't report an error if devices don't support AER

The only error returned by pci_{en,dis}able_pcie_error_reporting() is
-EIO which simply means that Advanced Error Reporting is not supported.
There is no need to report that, so remove the error check from e1000e,
igb and ixgbe.

Signed-off-by: Frans Pop <elendil@planet.nl>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Fix wrong sizeof
Jean Delvare [Fri, 2 Oct 2009 16:55:19 +0000 (09:55 -0700)]
net: Fix wrong sizeof

Which is why I have always preferred sizeof(struct foo) over
sizeof(var).

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Acked-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: splice() from tcp to pipe should take into account O_NONBLOCK
Eric Dumazet [Thu, 1 Oct 2009 22:26:00 +0000 (15:26 -0700)]
net: splice() from tcp to pipe should take into account O_NONBLOCK

tcp_splice_read() doesnt take into account socket's O_NONBLOCK flag

Before this patch :

splice(socket,0,pipe,0,128*1024,SPLICE_F_MOVE);
causes a random endless block (if pipe is full) and
splice(socket,0,pipe,0,128*1024,SPLICE_F_MOVE | SPLICE_F_NONBLOCK);
will return 0 immediately if the TCP buffer is empty.

User application has no way to instruct splice() that socket should be in blocking mode
but pipe in nonblock more.

Many projects cannot use splice(tcp -> pipe) because of this flaw.

http://git.samba.org/?p=samba.git;a=history;f=source3/lib/recvfile.c;h=ea0159642137390a0f7e57a123684e6e63e47581;hb=HEAD
http://lkml.indiana.edu/hypermail/linux/kernel/0807.2/0687.html

Linus introduced  SPLICE_F_NONBLOCK in commit 29e350944fdc2dfca102500790d8ad6d6ff4f69d
(splice: add SPLICE_F_NONBLOCK flag )

  It doesn't make the splice itself necessarily nonblocking (because the
  actual file descriptors that are spliced from/to may block unless they
  have the O_NONBLOCK flag set), but it makes the splice pipe operations
  nonblocking.

Linus intention was clear : let SPLICE_F_NONBLOCK control the splice pipe mode only

This patch instruct tcp_splice_read() to use the underlying file O_NONBLOCK
flag, as other socket operations do.

Users will then call :

splice(socket,0,pipe,0,128*1024,SPLICE_F_MOVE | SPLICE_F_NONBLOCK );

to block on data coming from socket (if file is in blocking mode),
and not block on pipe output (to avoid deadlock)

First version of this patch was submitted by Octavian Purdila

Reported-by: Volker Lendecke <vl@samba.org>
Reported-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Use sk_mark for routing lookup in more places
Atis Elsts [Thu, 1 Oct 2009 22:16:49 +0000 (15:16 -0700)]
net: Use sk_mark for routing lookup in more places

This patch against v2.6.31 adds support for route lookup using sk_mark in some
more places. The benefits from this patch are the following.
First, SO_MARK option now has effect on UDP sockets too.
Second, ip_queue_xmit() and inet_sk_rebuild_header() could fail to do routing
lookup correctly if TCP sockets with SO_MARK were used.

Signed-off-by: Atis Elsts <atis@mikrotik.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
14 years agosky2: irqname based on pci address
Stephen Hemminger [Thu, 1 Oct 2009 07:11:46 +0000 (07:11 +0000)]
sky2: irqname based on pci address

This is based on Michal Schmidt fix for skge.

Most network drivers request their IRQ when the interface is activated.
sky2 does it in ->probe() instead, because it can work with two-port
cards where the two net_devices use the same IRQ. This works fine most
of the time, except in some situations when the interface gets renamed.
Consider this example:

1. modprobe sky2
   The card is detected as eth0 and requests IRQ 17. Directory
   /proc/irq/17/eth0 is created.
2. There is an udev rule which says this interface should be called
   eth1, so udev renames eth0 -> eth1.
3. modprobe 8139too
   The Realtek card is detected as eth0. It will be using IRQ 17 too.
4. ip link set eth0 up
   Now 8139too requests IRQ 17.

The result is:
WARNING: at fs/proc/generic.c:590 proc_register ...
proc_dir_entry '17/eth0' already registered

The fix is for sky2 to name the irq based on the pci device, as is done
by some other devices DRM, infiniband, ...  ie. sky2@pci:0000:00:00

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Reviewed-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoskge: use unique IRQ name
Michal Schmidt [Thu, 1 Oct 2009 08:13:23 +0000 (08:13 +0000)]
skge: use unique IRQ name

Most network drivers request their IRQ when the interface is activated.
skge does it in ->probe() instead, because it can work with two-port
cards where the two net_devices use the same IRQ. This works fine most
of the time, except in some situations when the interface gets renamed.
Consider this example:

1. modprobe skge
   The card is detected as eth0 and requests IRQ 17. Directory
   /proc/irq/17/eth0 is created.
2. There is an udev rule which says this interface should be called
   eth1, so udev renames eth0 -> eth1.
3. modprobe 8139too
   The Realtek card is detected as eth0. It will be using IRQ 17 too.
4. ip link set eth0 up
   Now 8139too requests IRQ 17.

The result is:
WARNING: at fs/proc/generic.c:590 proc_register ...
proc_dir_entry '17/eth0' already registered
...
And "ls /proc/irq/17" shows two subdirectories, both called eth0.

Fix it by using a unique name for skge's IRQ, based on the PCI address.
The naming from the example then looks like this:
$ grep skge /proc/interrupts
 17:        169   IO-APIC-fasteoi   skge@pci:0000:00:0a.0, eth0

irqbalance daemon will have to be taught to recognize "skge@" as an
Ethernet interrupt. This will be a one-liner addition in classify.c. I
will send a patch to irqbalance if this change is accepted.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoIPv4 TCP fails to send window scale option when window scale is zero
Ori Finkelman [Thu, 1 Oct 2009 06:41:59 +0000 (06:41 +0000)]
IPv4 TCP fails to send window scale option when window scale is zero

Acknowledge TCP window scale support by inserting the proper option in SYN/ACK
and SYN headers even if our window scale is zero.

This fixes the following observed behavior:

1. Client sends a SYN with TCP window scaling option and non zero window scale
value to a Linux box.
2. Linux box notes large receive window from client.
3. Linux decides on a zero value of window scale for its part.
4. Due to compare against requested window scale size option, Linux does not to
 send windows scale TCP option header on SYN/ACK at all.

With the following result:

Client box thinks TCP window scaling is not supported, since SYN/ACK had no
TCP window scale option, while Linux thinks that TCP window scaling is
supported (and scale might be non zero), since SYN had  TCP window scale
option and we have a mismatched idea between the client and server
regarding window sizes.

Probably it also fixes up the following bug (not observed in practice):

1. Linux box opens TCP connection to some server.
2. Linux decides on zero value of window scale.
3. Due to compare against computed window scale size option, Linux does
not to set windows scale TCP  option header on SYN.

With the expected result that the server OS does not use window scale option
due to not receiving such an option in the SYN headers, leading to suboptimal
performance.

Signed-off-by: Gilad Ben-Yossef <gilad@codefidence.com>
Signed-off-by: Ori Finkelman <ori@comsleep.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet/ipv4/tcp.c: fix min() type mismatch warning
Andrew Morton [Thu, 1 Oct 2009 22:02:20 +0000 (15:02 -0700)]
net/ipv4/tcp.c: fix min() type mismatch warning

net/ipv4/tcp.c: In function 'do_tcp_setsockopt':
net/ipv4/tcp.c:2050: warning: comparison of distinct pointer types lacks a cast

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoKconfig: STRIP: Remove stale bits of STRIP help text
Ralf Baechle [Thu, 1 Oct 2009 21:49:14 +0000 (14:49 -0700)]
Kconfig: STRIP: Remove stale bits of STRIP help text

Remove references to dead web site mosquitonet.Stanford.EDU.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoNET: mkiss: Fix typo
Ralf Baechle [Thu, 1 Oct 2009 21:48:25 +0000 (14:48 -0700)]
NET: mkiss: Fix typo

This typo was introduced by 5793f4be23f0171b4999ca68a39a9157b44139f3 on
October 14, 2005 ...

Reported-by: Matti Aarnio <matti.aarnio@zmailer.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotg3: Remove prev_vlan_tag from struct tx_ring_info
Eric Dumazet [Thu, 1 Oct 2009 21:37:34 +0000 (14:37 -0700)]
tg3: Remove prev_vlan_tag from struct tx_ring_info

prev_vlan_tag field is not used.

Patch saves 512*8 bytes per tx queue ring on 64bit arches.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Matthew Carlson <mcarlson@broadcom.com>
14 years agomove virtnet_remove to .devexit.text
Uwe Kleine-König [Wed, 30 Sep 2009 22:28:34 +0000 (22:28 +0000)]
move virtnet_remove to .devexit.text

The function virtnet_remove is used only wrapped by __devexit_p so define
it using __devexit.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Alex Williamson <alex.williamson@hp.com>
Cc: Mark McLoughlin <markmc@redhat.com>
Cc: netdev@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agodon't use __devexit_p to wrap sgiseeq_remove
Uwe Kleine-König [Wed, 30 Sep 2009 22:28:26 +0000 (22:28 +0000)]
don't use __devexit_p to wrap sgiseeq_remove

The function sgiseeq_remove is defined using __exit, so don't use
__devexit_p but __exit_p to wrap it.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Cc: David S. Miller <davem@davemloft.net>
Cc: Wang Chen <wangchen@cn.fujitsu.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Patrick McHardy <kaber@trash.net>
Cc: netdev@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>