GitHub/moto-9609/android_kernel_motorola_exynos9610.git
13 years agoixgbe: allow eeprom writes via ethtool
Emil Tantilov [Thu, 6 Oct 2011 08:57:04 +0000 (08:57 +0000)]
ixgbe: allow eeprom writes via ethtool

Implement support for ethtool -E

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Tested-by: Stephen Ko <stephen.s.ko@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: fix endianess when writing driver version to firmware
Emil Tantilov [Tue, 11 Oct 2011 08:24:57 +0000 (08:24 +0000)]
ixgbe: fix endianess when writing driver version to firmware

This patch makes sure that register writes are in little endian and
also converts the reads back to big-endian.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Stephen Ko <stephen.s.ko@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Check if subordinate VFs are assigned to virtual machines
Greg Rose [Fri, 14 Oct 2011 02:57:14 +0000 (02:57 +0000)]
igb: Check if subordinate VFs are assigned to virtual machines

Kvm and the Xen pci-back driver will set a flag in the virtual function
pci device dev_flags when the VF is assigned to a guest VM.  Before
destroying subordinate VFs check to see if the flag is set and if so
skip the call to pci_disable_sriov() to avoid system crashes.

Copy the maintainer for the Xen pci-back driver.  Also CC'ing
maintainers of all drivers found to call pci_disable_sriov().

V2 - Fix  uninitialized variable warning

Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Christian Benvenuti <benve@cisco.com>
Cc: Sathya Perla <sathya.perla@emulex.com>
Cc: Dimitris Michailidis <dm@chelsio.com>
Cc: Jon Mason <jdmason@kudzu.us>
Cc: James Smart <james.smart@emulex.com>
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigbvf: Fix trunk vlan
Greg Rose [Sat, 27 Aug 2011 06:24:59 +0000 (06:24 +0000)]
igbvf: Fix trunk vlan

Changes to clean up the VLAN Rx path by Jiri Pirko broke trunk VLAN.
Trunk VLANs in a VF driver are those set using

"ip link set <pfdev> vf <n> <vlanid>"

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
CC: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigbvf: convert to ndo_fix_features
Michał Mirosław [Tue, 30 Aug 2011 17:07:11 +0000 (17:07 +0000)]
igbvf: convert to ndo_fix_features

Private rx_csum flags are now duplicate of netdev->features & NETIF_F_RXCSUM.
Removing this needs deeper surgery.

Things noticed:
 - HW VLAN acceleration probably can be toggled, but it's left as is
 - the resets on RX csum offload change can probably be avoided
 - there is A LOT of copy-and-pasted code here

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: enable l4 timestamping for v2 event packets
Jacob Keller [Wed, 12 Oct 2011 00:51:54 +0000 (00:51 +0000)]
igb: enable l4 timestamping for v2 event packets

When enabling hardware timestamping for ptp v2 event packets, the
software does not setup the queue for l4 packets, although layer 4
packets are valid for v2. This patch adds the flag which enables
setting up a queue and enabling udp packet timestamping.

Signed-off-by: Jacob E Keller <jacob.e.keller@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: Add new netdev op to turn spoof checking on or off per VF
Greg Rose [Thu, 29 Sep 2011 05:57:33 +0000 (05:57 +0000)]
ixgbe: Add new netdev op to turn spoof checking on or off per VF

Implements the new netdev op to allow user configuration of spoof
checking on a per VF basis.

V2 - Change netdev spoof check op setting to bool

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoif_link: Add additional parameter to IFLA_VF_INFO for spoof checking
Greg Rose [Sat, 8 Oct 2011 03:05:24 +0000 (03:05 +0000)]
if_link: Add additional parameter to IFLA_VF_INFO for spoof checking

Add configuration setting for drivers to turn spoof checking on or off
for discrete VFs.

v2 - Fix indentation problem, wrap the ifla_vf_info structure in
     #ifdef __KERNEL__ to prevent user space from accessing and
     change function paramater for the spoof check setting netdev
     op from u8 to bool.
v3 - Preset spoof check setting to -1 so that user space tools such
     as ip can detect that the driver didn't report a spoofcheck
     setting.  Prevents incorrect display of spoof check settings
     for drivers that don't report it.

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoe1000e: locking bug introduced by commit 67fd4fcb
Bruce Allan [Fri, 7 Oct 2011 03:50:38 +0000 (03:50 +0000)]
e1000e: locking bug introduced by commit 67fd4fcb

Commit 67fd4fcb (e1000e: convert to stats64) added the ability to update
statistics more accurately and on-demand through the net_device_ops
.ndo_get_stats64 hook, but introduced a locking bug on 82577/8/9 when
linked at half-duplex (seen on kernels with CONFIG_DEBUG_ATOMIC_SLEEP=y and
CONFIG_PROVE_LOCKING=y).  The commit introduced code paths that caused a
mutex to be locked in atomic contexts, e.g. an rcu_read_lock is held when
irqbalance reads the stats from /sys/class/net/ethX/statistics causing the
mutex to be locked to read the Phy half-duplex statistics registers.

The mutex was originally introduced to prevent concurrent accesses of
resources (the NVM and Phy) shared by the driver, firmware and hardware
a few years back when there was an issue with the NVM getting corrupted.
It was later split into two mutexes - one for the NVM and one for the Phy
when it was determined the NVM, unlike the Phy, should not be protected by
the software/firmware/hardware semaphore (arbitration of which is done in
part with the SWFLAG bit in the EXTCNF_CTRL register).  This latter
semaphore should be sufficient to prevent resource contention of the Phy in
the driver (i.e. the mutex for Phy accesses is not needed), but to be sure
the mutex is replaced with an atomic bit flag which will warn if any
contention is possible.

Also add additional debug output to help determine when the sw/fw/hw
semaphore is owned by the firmware or hardware.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Reported-by: Francois Romieu <romieu@fr.zoreil.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
13 years agoftmac100: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 11:20:08 +0000 (11:20 +0000)]
ftmac100: fix skb truesize underestimation

ftmac100 allocates a page per skb fragment. We must account
PAGE_SIZE increments on skb->truesize, not the actual frag length.

If frame is under 64 bytes, page is freed, so increase truesize only for
bigger frames.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Po-Yu Chuang <ratbert@faraday-tech.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoniu: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 12:39:27 +0000 (12:39 +0000)]
niu: fix skb truesize underestimation

Add a 'truesize' argument to niu_rx_skb_append(), filled with rcr_size
by the caller to properly account frag sizes in skb->truesize

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agovmxnet3: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 11:38:17 +0000 (11:38 +0000)]
vmxnet3: fix skb truesize underestimation

vmxnet3 allocates a page per skb fragment. We must account
PAGE_SIZE increments on skb->truesize, not the actual frag length.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Shreyas Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoftgmac100: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 11:30:52 +0000 (11:30 +0000)]
ftgmac100: fix skb truesize underestimation

ftgmac100 allocates a page per skb fragment. We must account
PAGE_SIZE increments on skb->truesize, not the actual frag length.

If frame is under 64 bytes, page is freed, and truesize adjusted.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Po-Yu Chuang <ratbert@faraday-tech.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agosky2: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 21:12:46 +0000 (17:12 -0400)]
sky2: fix skb truesize underestimation

sky2 allocates a page per skb fragment. We must account
PAGE_SIZE increments on skb->truesize, not the actual frag length.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoe1000e: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 08:03:36 +0000 (08:03 +0000)]
e1000e: fix skb truesize underestimation

e1000e allocates a page per skb fragment. We must account
PAGE_SIZE increments on skb->truesize, not the actual frag length.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoixgbe: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 07:59:41 +0000 (07:59 +0000)]
ixgbe: fix skb truesize underestimation

ixgbe allocates half a page per skb fragment. We must account
PAGE_SIZE/2 increments on skb->truesize, not the actual frag length.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoigb: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 07:56:41 +0000 (07:56 +0000)]
igb: fix skb truesize underestimation

e1000 allocates half a page per skb fragment. We must account
PAGE_SIZE/2 increments on skb->truesize, not the actual frag length.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoe1000: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 07:53:42 +0000 (07:53 +0000)]
e1000: fix skb truesize underestimation

e1000 allocates a full page per skb fragment. We must account PAGE_SIZE
increments on skb->truesize, not the actual frag length.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobnx2: fix skb truesize underestimation
Eric Dumazet [Thu, 13 Oct 2011 07:50:19 +0000 (07:50 +0000)]
bnx2: fix skb truesize underestimation

bnx2 allocates a full page per fragment. We must account PAGE_SIZE
increments on skb->truesize, not the actual frag length.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobe2net: fix truesize errors
Eric Dumazet [Thu, 13 Oct 2011 06:31:02 +0000 (06:31 +0000)]
be2net: fix truesize errors

Fix skb truesize underestimations of this driver.

Each frag truesize is exactly rx_frag_size bytes. (2048 bytes per
default)

A driver should not use "sizeof(struct sk_buff)" at all.

Signed-off-by: Eric Dumazet <eric.dumazet>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: more accurate skb truesize
Eric Dumazet [Thu, 13 Oct 2011 07:28:54 +0000 (07:28 +0000)]
net: more accurate skb truesize

skb truesize currently accounts for sk_buff struct and part of skb head.
kmalloc() roundings are also ignored.

Considering that skb_shared_info is larger than sk_buff, its time to
take it into account for better memory accounting.

This patch introduces SKB_TRUESIZE(X) macro to centralize various
assumptions into a single place.

At skb alloc phase, we put skb_shared_info struct at the exact end of
skb head, to allow a better use of memory (lowering number of
reallocations), since kmalloc() gives us power-of-two memory blocks.

Unless SLUB/SLUB debug is active, both skb->head and skb_shared_info are
aligned to cache lines, as before.

Note: This patch might trigger performance regressions because of
misconfigured protocol stacks, hitting per socket or global memory
limits that were previously not reached. But its a necessary step for a
more accurate memory accounting.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Andi Kleen <ak@linux.intel.com>
CC: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net-next
David S. Miller [Thu, 13 Oct 2011 18:00:06 +0000 (14:00 -0400)]
Merge git://git./linux/kernel/git/jkirsher/net-next

13 years agoigb: Version bump.
Carolyn Wyborny [Fri, 7 Oct 2011 07:00:27 +0000 (07:00 +0000)]
igb: Version bump.

This change updates the driver version to 3.2.10.

Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Loopback functionality supports for i350 devices
Akeem G. Abodunrin [Thu, 8 Sep 2011 20:39:48 +0000 (20:39 +0000)]
igb: Loopback functionality supports for i350 devices

This patch adds VMDq loopback pf support for i350 devices. The patch
is necessary since the register that enabled loopback was moved and
renamed from DTXSWC to TXSWC.

Signed-off-by: "Akeem G. Abodunrin" <akeem.g.abodunrin@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: fix static function warnings reported by sparse
Emil Tantilov [Tue, 30 Aug 2011 06:35:04 +0000 (06:35 +0000)]
igb: fix static function warnings reported by sparse

igb_update/validate_nvm_checksum_with_offset() should be static.
Also removes unneeded prototypes for the above functions.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Add workaround for byte swapped VLAN on i350 local traffic
Alexander Duyck [Fri, 26 Aug 2011 07:47:11 +0000 (07:47 +0000)]
igb: Add workaround for byte swapped VLAN on i350 local traffic

On i350 when traffic is looped back from a VF to the PF the value is byte
swapped from the normal format.  In order to address this we need to add a
flag indicating that the ring will need to byte swap the loopback packets
prior to processing them.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Drop unnecessary write of E1000_IMS from igb_msix_other
Alexander Duyck [Fri, 26 Aug 2011 07:47:01 +0000 (07:47 +0000)]
igb: Drop unnecessary write of E1000_IMS from igb_msix_other

Since we mask interrupts in EIMS not in IMS there is no need to re-enable
mask bits in that register.  As such we can remove the write to IMS from
the end of igb_msix_other.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Fix features that are currently 82580 only and should also be i350
Alexander Duyck [Fri, 26 Aug 2011 07:46:55 +0000 (07:46 +0000)]
igb: Fix features that are currently 82580 only and should also be i350

This change allows support for per packet timesync and global device reset
on the i350 adapter.  These features were supported on both 82580 and i350
however it looks like several checks where not updated and as such the i350
support was not enabled.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Make certain one vector is always assigned in igb_request_irq
Alexander Duyck [Fri, 26 Aug 2011 07:46:45 +0000 (07:46 +0000)]
igb: Make certain one vector is always assigned in igb_request_irq

This change makes certain that one interrupt is always initialized in
igb_request_irq.  In addition we drop the use of adapter->pdev and
instead just call pdev since we made a local copy of the pointer earlier in
the function.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: avoid unnecessarily creating a local copy of the q_vector
Alexander Duyck [Fri, 26 Aug 2011 07:46:34 +0000 (07:46 +0000)]
igb: avoid unnecessarily creating a local copy of the q_vector

This is mostly a drop of unnecessary pointer defines for q_vector when we
don't have issues with line width and don't have multiple references to
the pointer.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: Correct check for change in FCoE priority
Mark Rustad [Tue, 20 Sep 2011 03:00:22 +0000 (03:00 +0000)]
ixgbe: Correct check for change in FCoE priority

Correct a check for change in FCoE priority when IEEE mode DCB is in use.
In IEEE mode a different function has to be used to get the FCoE priority
mask. Also, the check for the mask assumed that only one priority was set.
In case there should be more than one, check just the bit.

These changes help avoid link flapping issues that can come up when IEEE
DCB is in use.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: Add FCoE DDP allocation failure counters to ethtool stats.
Amir Hanania [Wed, 31 Aug 2011 02:07:55 +0000 (02:07 +0000)]
ixgbe: Add FCoE DDP allocation failure counters to ethtool stats.

Add 2 new counters to ethtool:
1. Count DDP allocation failure since we max the number of buffers
allowed in one DDP context.
2. Count DDP allocation failure since we max the number of buffers
allowed in one DDP context when we alloc an extra buffer.

Signed-off-by: Amir Hanania <amir.hanania@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: Add protection from VF invalid target DMA
Greg Rose [Wed, 7 Sep 2011 05:59:35 +0000 (05:59 +0000)]
ixgbe: Add protection from VF invalid target DMA

It is possible for a VF to set an invalid target DMA address in its
Tx/Rx descriptor buffer pointers.  The workarounds in this patch
will guard against such an event and issue a VFLR to the VF in response.
The VFLR will shut down the VF until an administrator can take action
to investigate the event and correct the problem.

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agonet-netlink: Add a new attribute to expose TOS values via netlink
Murali Raja [Wed, 12 Oct 2011 09:00:35 +0000 (09:00 +0000)]
net-netlink: Add a new attribute to expose TOS values via netlink

This patch exposes the tos value for the TCP sockets when the TOS flag
is requested in the ext_flags for the inet_diag request. This would mainly be
used to expose TOS values for both for TCP and UDP sockets. Currently it is
supported for TCP. When netlink support for UDP would be added the support
to expose the TOS values would alse be done. For IPV4 tos value is exposed
and for IPV6 tclass value is exposed.

Signed-off-by: Murali Raja <muralira@google.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agocipso: remove an unneeded NULL check in cipso_v4_doi_add()
Dan Carpenter [Tue, 11 Oct 2011 22:43:53 +0000 (18:43 -0400)]
cipso: remove an unneeded NULL check in cipso_v4_doi_add()

We dereference doi_def on the line before the NULL check.  It has
been this way since 2008.  I checked all the callers and doi_def is
always non-NULL here.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'for-davem' of git://git.infradead.org/users/linville/wireless-next
David S. Miller [Tue, 11 Oct 2011 19:39:21 +0000 (15:39 -0400)]
Merge branch 'for-davem' of git://git.infradead.org/users/linville/wireless-next

13 years agoMerge branch 'master' of git://git.infradead.org/users/linville/wireless-next into...
John W. Linville [Tue, 11 Oct 2011 19:35:42 +0000 (15:35 -0400)]
Merge branch 'master' of git://git.infradead.org/users/linville/wireless-next into for-davem

Conflicts:
Documentation/feature-removal-schedule.txt

13 years agoath6kl: fixup merge damage in ath6kl_mgmt_tx
John W. Linville [Tue, 11 Oct 2011 19:33:10 +0000 (15:33 -0400)]
ath6kl: fixup merge damage in ath6kl_mgmt_tx

  CC [M]  drivers/net/wireless/ath/ath6kl/cfg80211.o
drivers/net/wireless/ath/ath6kl/cfg80211.c:1838:2: warning: initialization from incompatible pointer type

Caused by commit e9f935e3e8dc0bddd0df6d148165d95925422502...

Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net-next
David S. Miller [Tue, 11 Oct 2011 19:24:56 +0000 (15:24 -0400)]
Merge git://git./linux/kernel/git/jkirsher/net-next

13 years agoll_temac: convert to SKB paged frag API.
Ian Campbell [Mon, 10 Oct 2011 01:11:40 +0000 (01:11 +0000)]
ll_temac: convert to SKB paged frag API.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoemac: convert to SKB paged frag API.
Ian Campbell [Mon, 10 Oct 2011 01:11:39 +0000 (01:11 +0000)]
emac: convert to SKB paged frag API.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoehea: convert to SKB paged frag API
Ian Campbell [Mon, 10 Oct 2011 01:11:38 +0000 (01:11 +0000)]
ehea: convert to SKB paged frag API

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Breno Leitao <leitao@linux.vnet.ibm.com>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoaf_packet: remove unnecessary BUG_ON() in tpacket_destruct_skb
danborkmann@iogearbox.net [Mon, 10 Oct 2011 06:52:46 +0000 (06:52 +0000)]
af_packet: remove unnecessary BUG_ON() in tpacket_destruct_skb

If skb is NULL, then stack trace is thrown anyway on dereference.
Therefore, the stack trace triggered by BUG_ON is duplicate.

Signed-off-by: Daniel Borkmann <danborkmann@googlemail.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agor6040: bump version to 0.28 and date to 07Oct2011.
Florian Fainelli [Thu, 6 Oct 2011 23:36:28 +0000 (23:36 +0000)]
r6040: bump version to 0.28 and date to 07Oct2011.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agor6040: invoke phy_{start,stop} when appropriate
Florian Fainelli [Thu, 6 Oct 2011 23:36:22 +0000 (23:36 +0000)]
r6040: invoke phy_{start,stop} when appropriate

Joe reported to me that right after a bring up of a r6040 interface
the ethtool output had no consistent output with respect to link duplex
and speed. Fix this by adding a missing phy_start call in r6040_up and
conversely a phy_stop call in r6040_down to properly initialize phy states.

Reported-by: Joe Chou <Joe.Chou@rdc.com.tw>
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agomlx4_en: Adding 40gb speed report for ethtool
Alexander Guller [Sun, 9 Oct 2011 05:29:42 +0000 (05:29 +0000)]
mlx4_en: Adding 40gb speed report for ethtool

Query port will now identify a 40G Ethernet speed.

Signed-off-by: Alexander Guller <alexg@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agomlx4_en: Fix crash upon device initialization error
Alexander Guller [Sun, 9 Oct 2011 05:29:35 +0000 (05:29 +0000)]
mlx4_en: Fix crash upon device initialization error

Netdevice was being freed without being unregistered first if
mlx4_SET_PORT_general or mlx4_INIT_PORT failed.

Signed-off-by: Alexander Guller <alexg@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agomlx4_en: Fix QP number calculation according to module param
Alexander Guller [Sun, 9 Oct 2011 05:29:26 +0000 (05:29 +0000)]
mlx4_en: Fix QP number calculation according to module param

Number of bits taken from mac table index in QP
calculation should be based on log_num_mac parameter.

Signed-off-by: Alexander Guller <alexg@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agomlx4_en: Added missing iounmap upon releasing a device
Alexander Guller [Sun, 9 Oct 2011 05:27:11 +0000 (05:27 +0000)]
mlx4_en: Added missing iounmap upon releasing a device

Fixed a memory leak caused by missing iounmap when device
is being released.

Signed-off-by: Alexander Guller <alexg@mellanox.co.il>
Signed-off-by: Sharon Cohen <sharonc@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agomlx4_en: Adjusting moderation per each ring
Alexander Guller [Sun, 9 Oct 2011 05:38:23 +0000 (05:38 +0000)]
mlx4_en: Adjusting moderation per each ring

Moderation is now done per ring and coalescing is enabled
by set_ring_param in ethtool.

Signed-off-by: Alexander Guller <alexg@mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agomlx4_en: Removing reserve vectors
Alexander Guller [Sun, 9 Oct 2011 05:26:46 +0000 (05:26 +0000)]
mlx4_en: Removing reserve vectors

Fixed a bug where ring size change caused insufficient memory
upon driver restart due to unreleased EQs.

Signed-off-by: Alexander Guller <alexg@mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agomlx4_en: Assigning TX irq per ring
Alexander Guller [Sun, 9 Oct 2011 05:26:31 +0000 (05:26 +0000)]
mlx4_en: Assigning TX irq per ring

Until now only RX rings used irq per ring
and TX used only one per port.
>From now on, both of them will use the
irq per ring while RX & TX ring[i] will
use the same irq.

Signed-off-by: Alexander Guller <alexg@mellanox.co.il>
Signed-off-by: Sharon Cohen <sharonc@mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoigb: add support for NETIF_F_RXHASH
Alexander Duyck [Fri, 26 Aug 2011 07:46:29 +0000 (07:46 +0000)]
igb: add support for NETIF_F_RXHASH

This patch adds support for Rx hashing.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: move TX hang check flag into ring->flags
Alexander Duyck [Fri, 26 Aug 2011 07:46:19 +0000 (07:46 +0000)]
igb: move TX hang check flag into ring->flags

This change moves the Tx hang check into the ring flags.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: fix recent VLAN changes that would leave VLANs disabled after reset
Alexander Duyck [Fri, 26 Aug 2011 07:46:08 +0000 (07:46 +0000)]
igb: fix recent VLAN changes that would leave VLANs disabled after reset

This patch cleans up several issues with VLANs on igb after the recent
changes that were meant to leave the VLANs enabled/disable via the
netdev->features flags.

Specifically the Rx VLAN settings were being dropped after reset due to the
fact that they were not being restored correctly.  In addition I removed
the IRQ disable/enable since those were in place to protect the setting of
vlgrp.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: leave staterr in place and instead us a helper function to check bits
Alexander Duyck [Fri, 26 Aug 2011 07:46:03 +0000 (07:46 +0000)]
igb: leave staterr in place and instead us a helper function to check bits

Instead of doing a byte swap on the staterr bits in the Rx descriptor we can
save ourselves a bit of space and some CPU time by instead just testing for
the various bits out of the Rx descriptor directly.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: retire the RX_CSUM flag and use the netdev flag instead
Alexander Duyck [Fri, 26 Aug 2011 07:45:57 +0000 (07:45 +0000)]
igb: retire the RX_CSUM flag and use the netdev flag instead

Since the netdev now has its' own checksum flag to indicate if Rx checksum
is enabled we might as well use that instead of using the ring flag.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: cleanup IVAR configuration
Alexander Duyck [Fri, 26 Aug 2011 07:45:52 +0000 (07:45 +0000)]
igb: cleanup IVAR configuration

This change is meant to cleanup some of the IVAR register configuration.
igb_assign_vector had become pretty large with multiple copies of the same
general code for setting the IVAR. This change consolidates most of that
code by adding the igb_write_ivar function which allows us just to compute
the index and offset and then use that information to setup the IVAR.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Move ITR related data into work container within the q_vector
Alexander Duyck [Fri, 26 Aug 2011 07:45:47 +0000 (07:45 +0000)]
igb: Move ITR related data into work container within the q_vector

This change moves information related to interrupt throttle rate
configuration into a separate q_vector sub-structure called a work
container. A similar change has already been made for ixgbe and this work
is based off of that.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Consolidate all of the ring feature flags into a single value
Alexander Duyck [Fri, 26 Aug 2011 07:45:36 +0000 (07:45 +0000)]
igb: Consolidate all of the ring feature flags into a single value

This change moves all of the ring flags into a single value.  The advantage
to this is that there is one central area for all of these flags and they
can all make use of the set/test bit operations.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: avoid unnecessary conversions from u16 to int
Alexander Duyck [Fri, 26 Aug 2011 07:45:26 +0000 (07:45 +0000)]
igb: avoid unnecessary conversions from u16 to int

There are a number of places where we have values that are stored as u16
but are being converted to int unnecessarily.  In order to avoid that we
should convert all variables that deal with the next_to_clean, next_to_use,
and count to u16 values.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Use node specific allocations for the q_vectors and rings
Alexander Duyck [Fri, 26 Aug 2011 07:45:20 +0000 (07:45 +0000)]
igb: Use node specific allocations for the q_vectors and rings

This change is meant to update the ring and vector allocations so that they
are per node instead of allocating everything on the node that
ifconfig/modprobe is called on.  By doing this we can cut down
significantly on cross node traffic.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: push data into first igb_tx_buffer sooner to reduce stack usage
Alexander Duyck [Fri, 26 Aug 2011 07:45:15 +0000 (07:45 +0000)]
igb: push data into first igb_tx_buffer sooner to reduce stack usage

Instead of storing most of the data for the TX hot path in the stack until
we are ready to write the descriptor we can save ourselves some time and
effort by pushing the SKB, tx_flags, gso_size, bytecount, and protocol into
the first igb_tx_buffer since that is where we will end up putting it
anyway.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agonet: Remove unnecessary driver assignments of ethtool_ringparam fields to zero
Rick Jones [Fri, 7 Oct 2011 23:13:28 +0000 (19:13 -0400)]
net: Remove unnecessary driver assignments of ethtool_ringparam fields to zero

Per comments from Ben Hutchings on a previous patch, sweep the floors
a little removing unnecessary assignments of zero to fields of struct
ethtool_ringparam in driver code supporting ethtool -g.

Signed-off-by: Rick Jones <rick.jones2@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoAdd ethtool -g support to 8139cp
Rick Jones [Fri, 7 Oct 2011 06:42:21 +0000 (06:42 +0000)]
Add ethtool -g support to 8139cp

Add support for reporting ring sizes via ethtool -g to the 8139cp driver.

Signed-off-by: Rick Jones <rick.jones2@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'master' of github.com:davem330/net
David S. Miller [Fri, 7 Oct 2011 17:38:43 +0000 (13:38 -0400)]
Merge branch 'master' of github.com:davem330/net

Conflicts:
net/batman-adv/soft-interface.c

13 years agonet: use sock_valbool_flag to set/clear SOCK_RXQ_OVFL
Johannes Berg [Fri, 7 Oct 2011 03:30:20 +0000 (03:30 +0000)]
net: use sock_valbool_flag to set/clear SOCK_RXQ_OVFL

There's no point in open-coding sock_valbool_flag().

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge git://github.com/Jkirsher/net-next
David S. Miller [Fri, 7 Oct 2011 16:34:52 +0000 (12:34 -0400)]
Merge git://github.com/Jkirsher/net-next

13 years agoigb: consolidate creation of Tx buffer info and data descriptor
Alexander Duyck [Fri, 26 Aug 2011 07:45:09 +0000 (07:45 +0000)]
igb: consolidate creation of Tx buffer info and data descriptor

This change will combine the writes of tx_buffer_info and the Tx data
descriptors into a single function. The advantage of this is that we can
avoid needless memory reads from the buffer info struct and speed things up
by keeping the accesses to the local registers.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Combine all flag info fields into a single tx_flags structure
Alexander Duyck [Fri, 26 Aug 2011 07:44:59 +0000 (07:44 +0000)]
igb: Combine all flag info fields into a single tx_flags structure

This change is meant to combine all of the TX flags fields into one u32
flags field so that it can be stored into the tx_buffer_info structure.
This includes the time stamp flag as well as mapped_as_page flag info.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Cleanup protocol handling in transmit path
Alexander Duyck [Fri, 26 Aug 2011 07:44:53 +0000 (07:44 +0000)]
igb: Cleanup protocol handling in transmit path

This change is meant to cleanup the protocol handling in the transmit path
so that it correctly offloads software VLAN tagged frames.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Create separate functions for generating cmd_type and olinfo
Alexander Duyck [Fri, 26 Aug 2011 07:44:48 +0000 (07:44 +0000)]
igb: Create separate functions for generating cmd_type and olinfo

This change is meant to improve the readability of the driver by separating
out the cmd_type configuration and the olinfo configuration into their own
functions.  By doing this it is much easier to determine which ingredients
go into setting up these to portions of the descriptor.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Make first and tx_buffer_info->next_to_watch into pointers
Alexander Duyck [Fri, 26 Aug 2011 07:44:43 +0000 (07:44 +0000)]
igb: Make first and tx_buffer_info->next_to_watch into pointers

This change converts two tx_buffer_info index values into pointers.  The
advantage to this is that we reduce unnecessary computations and in the case
of next_to_watch we get an added bonus of the value being able to provide
additional information as a NULL value indicates it is unset versus a 0 not
having any meaning for the index value.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Consolidate creation of Tx context descriptors into a single function
Alexander Duyck [Fri, 26 Aug 2011 07:44:32 +0000 (07:44 +0000)]
igb: Consolidate creation of Tx context descriptors into a single function

This patch is meant to simplify the transmit path by reducing the overhead
for creating a transmit context descriptor.  The current implementation is
split with igb_tso and igb_tx_csum doing two separate implementations on
how to setup the tx_buffer_info structure and the tx_desc.  By combining
them it is possible to reduce code and simplify things since now only one
function will create context descriptors.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: split buffer_info into tx_buffer_info and rx_buffer_info
Alexander Duyck [Fri, 26 Aug 2011 07:44:22 +0000 (07:44 +0000)]
igb: split buffer_info into tx_buffer_info and rx_buffer_info

In order to be able to improve the performance of the TX path it has been
necessary to add addition info to the tx_buffer_info structure.  However a
side effect is that the structure has gotten larger and this in turn has
also increased the size of the RX buffer info structure.  In order to avoid
this in the future I am splitting the single buffer_info structure into two
separate ones and instead I will join them by making the buffer_info
pointer in the ring a union of the two.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Make Tx budget for NAPI user adjustable
Alexander Duyck [Wed, 5 Oct 2011 13:35:24 +0000 (13:35 +0000)]
igb: Make Tx budget for NAPI user adjustable

This change is to make the NAPI budget limits for transmit
adjustable.  Currently they are only set to 128, and when
the changes/improvements to NAPI occur to allow for adjustability,
it would be possible to tune the value for optimal
performance with applications such as routing.

v2: remove tie between NAPI and interrupt moderation
    fix work limit define name (s/IXGBE/IGB/)
    Update patch description to better reflect patch

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
13 years agoe1000e: bad short packets received when jumbos enabled on 82579
Bruce Allan [Fri, 30 Sep 2011 08:07:00 +0000 (08:07 +0000)]
e1000e: bad short packets received when jumbos enabled on 82579

When short packets are received with jumbos enabled on 82579, they can be
interpreted to have a receive address that does not match any configured
address.  This is due to a hardware bug that can be worked around by
reducing the number of IPG octets added when the packet is transferred from
the PHY to the MAC.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoe1000: convert to private mutex from rtnl
Jesse Brandeburg [Wed, 5 Oct 2011 07:24:51 +0000 (07:24 +0000)]
e1000: convert to private mutex from rtnl

The e1000 driver when running with lockdep could run into
some possible deadlocks between the work items acquiring
rtnl and the rtnl lock being acquired before work items
were cancelled.

Use a private mutex to make sure lock ordering isn't violated.
The private mutex is only used to protect areas not generally
covered by the rtnl lock already.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Tushar Dave <tushar.n.dave@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoe1000: convert mdelay to msleep
Jesse Brandeburg [Wed, 5 Oct 2011 07:24:46 +0000 (07:24 +0000)]
e1000: convert mdelay to msleep

With the previous commit, there are several functions
that are only ever called from thread context, and are
able to sleep with msleep instead of mdelay.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Tushar Dave <tushar.n.dave@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoe1000: convert hardware management from timers to threads
Jesse Brandeburg [Wed, 5 Oct 2011 07:24:41 +0000 (07:24 +0000)]
e1000: convert hardware management from timers to threads

Thomas Gleixner (tglx) reported that e1000 was delaying for many milliseconds
(using mdelay) from inside timer/interrupt context.  None of these paths are
performance critical and can be moved into threads/work items.  This patch
implements the work items and the next patch changes the mdelays to msleeps.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Tushar Dave <tushar.n.dave@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: bump version number
Don Skidmore [Fri, 7 Oct 2011 03:53:51 +0000 (03:53 +0000)]
ixgbe: bump version number

Bump the version string to better match pair up with the out of tree
driver that contains the same functionality.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoMerge git://github.com/davem330/net
Linus Torvalds [Thu, 6 Oct 2011 23:15:10 +0000 (16:15 -0700)]
Merge git://github.com/davem330/net

* git://github.com/davem330/net:
  net: fix typos in Documentation/networking/scaling.txt
  bridge: leave carrier on for empty bridge
  netfilter: Use proper rwlock init function
  tcp: properly update lost_cnt_hint during shifting
  tcp: properly handle md5sig_pool references
  macvlan/macvtap: Fix unicast between macvtap interfaces in bridge mode

13 years agox86/PCI: use host bridge _CRS info on ASUS M2V-MX SE
Paul Menzel [Wed, 31 Aug 2011 15:07:10 +0000 (17:07 +0200)]
x86/PCI: use host bridge _CRS info on ASUS M2V-MX SE

In summary, this DMI quirk uses the _CRS info by default for the ASUS
M2V-MX SE by turning on `pci=use_crs` and is similar to the quirk
added by commit 2491762cfb47 ("x86/PCI: use host bridge _CRS info on
ASRock ALiveSATA2-GLAN") whose commit message should be read for further
information.

Since commit 3e3da00c01d0 ("x86/pci: AMD one chain system to use pci
read out res") Linux gives the following oops:

    parport0: PC-style at 0x378, irq 7 [PCSPP,TRISTATE]
    HDA Intel 0000:20:01.0: PCI INT A -> GSI 17 (level, low) -> IRQ 17
    HDA Intel 0000:20:01.0: setting latency timer to 64
    BUG: unable to handle kernel paging request at ffffc90011c08000
    IP: [<ffffffffa0578402>] azx_probe+0x3ad/0x86b [snd_hda_intel]
    PGD 13781a067 PUD 13781b067 PMD 1300ba067 PTE 800000fd00000173
    Oops: 0009 [#1] SMP
    last sysfs file: /sys/module/snd_pcm/initstate
    CPU 0
    Modules linked in: snd_hda_intel(+) snd_hda_codec snd_hwdep snd_pcm_oss snd_mixer_oss snd_pcm snd_seq_midi snd_rawmidi snd_seq_midi_event tpm_tis tpm snd_seq tpm_bios psmouse parport_pc snd_timer snd_seq_device parport processor evdev snd i2c_viapro thermal_sys amd64_edac_mod k8temp i2c_core soundcore shpchp pcspkr serio_raw asus_atk0110 pci_hotplug edac_core button snd_page_alloc edac_mce_amd ext3 jbd mbcache sha256_generic cryptd aes_x86_64 aes_generic cbc dm_crypt dm_mod raid1 md_mod usbhid hid sg sd_mod crc_t10dif sr_mod cdrom ata_generic uhci_hcd sata_via pata_via libata ehci_hcd usbcore scsi_mod via_rhine mii nls_base [last unloaded: scsi_wait_scan]
    Pid: 1153, comm: work_for_cpu Not tainted 2.6.37-1-amd64 #1 M2V-MX SE/System Product Name
    RIP: 0010:[<ffffffffa0578402>]  [<ffffffffa0578402>] azx_probe+0x3ad/0x86b [snd_hda_intel]
    RSP: 0018:ffff88013153fe50  EFLAGS: 00010286
    RAX: ffffc90011c08000 RBX: ffff88013029ec00 RCX: 0000000000000006
    RDX: 0000000000000000 RSI: 0000000000000246 RDI: 0000000000000246
    RBP: ffff88013341d000 R08: 0000000000000000 R09: 0000000000000040
    R10: 0000000000000286 R11: 0000000000003731 R12: ffff88013029c400
    R13: 0000000000000000 R14: 0000000000000000 R15: ffff88013341d090
    FS:  0000000000000000(0000) GS:ffff8800bfc00000(0000) knlGS:00000000f7610ab0
    CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
    CR2: ffffc90011c08000 CR3: 0000000132f57000 CR4: 00000000000006f0
    DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
    DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
    Process work_for_cpu (pid: 1153, threadinfo ffff88013153e000, task ffff8801303c86c0)
    Stack:
     0000000000000005 ffffffff8123ad65 00000000000136c0 ffff88013029c400
     ffff8801303c8998 ffff88013341d000 ffff88013341d090 ffff8801322d9dc8
     ffff88013341d208 0000000000000000 0000000000000000 ffffffff811ad232
    Call Trace:
     [<ffffffff8123ad65>] ? __pm_runtime_set_status+0x162/0x186
     [<ffffffff811ad232>] ? local_pci_probe+0x49/0x92
     [<ffffffff8105afc5>] ? do_work_for_cpu+0x0/0x1b
     [<ffffffff8105afc5>] ? do_work_for_cpu+0x0/0x1b
     [<ffffffff8105afd0>] ? do_work_for_cpu+0xb/0x1b
     [<ffffffff8105fd3f>] ? kthread+0x7a/0x82
     [<ffffffff8100a824>] ? kernel_thread_helper+0x4/0x10
     [<ffffffff8105fcc5>] ? kthread+0x0/0x82
     [<ffffffff8100a820>] ? kernel_thread_helper+0x0/0x10
    Code: f4 01 00 00 ef 31 f6 48 89 df e8 29 dd ff ff 85 c0 0f 88 2b 03 00 00 48 89 ef e8 b4 39 c3 e0 8b 7b 40 e8 fc 9d b1 e0 48 8b 43 38 <66> 8b 10 66 89 14 24 8b 43 14 83 e8 03 83 f8 01 77 32 31 d2 be
    RIP  [<ffffffffa0578402>] azx_probe+0x3ad/0x86b [snd_hda_intel]
     RSP <ffff88013153fe50>
    CR2: ffffc90011c08000
    ---[ end trace 8d1f3ebc136437fd ]---

Trusting the ACPI _CRS information (`pci=use_crs`) fixes this problem.

    $ dmesg | grep -i crs # with the quirk
    PCI: Using host bridge windows from ACPI; if necessary, use "pci=nocrs" and report a bug

The match has to be against the DMI board entries though since the vendor entries are not populated.

    DMI: System manufacturer System Product Name/M2V-MX SE, BIOS 0304    10/30/2007

This quirk should be removed when `pci=use_crs` is enabled for machines
from 2006 or earlier or some other solution is implemented.

Using coreboot [1] with this board the problem does not exist but this
quirk also does not affect it either. To be safe though the check is
tightened to only take effect when the BIOS from American Megatrends is
used.

        15:13 < ruik> but coreboot does not need that
        15:13 < ruik> because i have there only one root bus
        15:13 < ruik> the audio is behind a bridge

        $ sudo dmidecode
        BIOS Information
                Vendor: American Megatrends Inc.
                Version: 0304
                Release Date: 10/30/2007

[1] http://www.coreboot.org/

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=30552

Cc: stable@kernel.org (2.6.34)
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: x86@kernel.org
Signed-off-by: Paul Menzel <paulepanter@users.sourceforge.net>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agonet: use DMA_x_DEVICE and dma_mapping_error with skb_frag_dma_map
Ian Campbell [Thu, 6 Oct 2011 10:10:48 +0000 (11:10 +0100)]
net: use DMA_x_DEVICE and dma_mapping_error with skb_frag_dma_map

When I converted some drivers from pci_map_page to skb_frag_dma_map I
neglected to convert PCI_DMA_xDEVICE into DMA_x_DEVICE and
pci_dma_mapping_error into dma_mapping_error.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: fix typos in Documentation/networking/scaling.txt
Benjamin Poirier [Tue, 4 Oct 2011 04:00:30 +0000 (04:00 +0000)]
net: fix typos in Documentation/networking/scaling.txt

The second hunk fixes rps_sock_flow_table but has to re-wrap the paragraph.

Signed-off-by: Benjamin Poirier <benjamin.poirier@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agodcb: Add stub routines for !CONFIG_DCB
John Fastabend [Thu, 6 Oct 2011 08:52:44 +0000 (08:52 +0000)]
dcb: Add stub routines for !CONFIG_DCB

To avoid ifdefs in the other code that supports DCB notifiers
add stub routines. This method seems popular in other net code
for example 8021Q.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agodcb: add DCBX mode to event notifier attributes
John Fastabend [Thu, 6 Oct 2011 08:52:38 +0000 (08:52 +0000)]
dcb: add DCBX mode to event notifier attributes

Add DCBX mode to event notifiers so listeners can learn
currently enabled mode.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agodcb: Use ifindex instead of ifname
Mark Rustad [Thu, 6 Oct 2011 08:52:33 +0000 (08:52 +0000)]
dcb: Use ifindex instead of ifname

Use ifindex instead of ifname in the DCB app ring. This makes for a smaller
data structure and faster comparisons. It also avoids possible issues when
a net device is renamed.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agovirtio-net: Verify page list size before fitting into skb
Sasha Levin [Wed, 28 Sep 2011 04:40:54 +0000 (04:40 +0000)]
virtio-net: Verify page list size before fitting into skb

This patch verifies that the length of a buffer stored in a linked list
of pages is small enough to fit into a skb.

If the size is larger than a max size of a skb, it means that we shouldn't
go ahead building skbs anyway since we won't be able to send the buffer as
the user requested.

Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: virtualization@lists.linux-foundation.org
Cc: netdev@vger.kernel.org
Cc: kvm@vger.kernel.org
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobridge: allow forwarding some link local frames
stephen hemminger [Mon, 3 Oct 2011 18:14:46 +0000 (18:14 +0000)]
bridge: allow forwarding some link local frames

This is based on an earlier patch by Nick Carter with comments
by David Lamparter but with some refinements. Thanks for their patience
this is a confusing area with overlap of standards, user requirements,
and compatibility with earlier releases.

It adds a new sysfs attribute
   /sys/class/net/brX/bridge/group_fwd_mask
that controls forwarding of frames with address of: 01-80-C2-00-00-0X
The default setting has no forwarding to retain compatibility.

One change from earlier releases is that forwarding of group
addresses is not dependent on STP being enabled or disabled. This
choice was made based on interpretation of tie 802.1 standards.
I expect complaints will arise because of this, but better to follow
the standard than continue acting incorrectly by default.

The filtering mask is writeable, but only values that don't forward
known control frames are allowed. It intentionally blocks attempts
to filter control protocols. For example: writing a 8 allows
forwarding 802.1X PAE addresses which is the most common request.

Reported-by: David Lamparter <equinox@diac24.net>
Original-patch-by: Nick Carter <ncarter100@gmail.com>
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Tested-by: Benjamin Poirier <benjamin.poirier@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobridge: leave carrier on for empty bridge
stephen hemminger [Mon, 3 Oct 2011 18:14:45 +0000 (18:14 +0000)]
bridge: leave carrier on for empty bridge

This resolves a regression seen by some users of bridging.
Some users use the bridge like a dummy device.
They expect to be able to put an IPv6 address on the device
with no ports attached. Although there are better ways of doing
this, there is no reason to not allow it.

Note: the bridge still will reflect the state of ports in the
bridge if there are any added.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'for-linus' of http://people.redhat.com/agk/git/linux-dm
Linus Torvalds [Thu, 6 Oct 2011 15:31:47 +0000 (08:31 -0700)]
Merge branch 'for-linus' of people.redhat.com/agk/git/linux-dm

* 'for-linus' of http://people.redhat.com/agk/git/linux-dm:
  dm crypt: always disable discard_zeroes_data
  dm: raid fix write_mostly arg validation
  dm table: avoid crash if integrity profile changes
  dm: flakey fix corrupt_bio_byte error path

13 years agoMerge branch 'for-linus' of git://neil.brown.name/md
Linus Torvalds [Thu, 6 Oct 2011 15:30:03 +0000 (08:30 -0700)]
Merge branch 'for-linus' of git://neil.brown.name/md

* 'for-linus' of git://neil.brown.name/md:
  md: Avoid waking up a thread after it has been freed.

13 years agoigb: Alternate MAC Address Updates for Func2&3
Akeem G. Abodunrin [Fri, 2 Sep 2011 23:09:30 +0000 (23:09 +0000)]
igb: Alternate MAC Address Updates for Func2&3

Only function 1 has support for Alternate MAC Address in the EEPROM before,
this update now allow function 2 and 3 to have support for Alternate MAC
Address in the EEPROM.

Signed-off-by: "Akeem G. Abodunrin" <akeem.g.abodunrin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Alternate MAC Address EEPROM Updates
Akeem G. Abodunrin [Fri, 2 Sep 2011 23:08:55 +0000 (23:08 +0000)]
igb: Alternate MAC Address EEPROM Updates

This code check word 0x37 in the EEPROM, if it is 0xFFFF _or_ 0x0000, then
there is no Alternate MAC Address in the EEPROM.

Signed-off-by: "Akeem G. Abodunrin" <akeem.g.abodunrin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoigb: Code to prevent overwriting SFP I2C
Akeem G. Abodunrin [Fri, 2 Sep 2011 23:11:19 +0000 (23:11 +0000)]
igb: Code to prevent overwriting SFP I2C

This patch fixes "overwrite" problem. without this fix, SFP I2C EEPROM
data, which is located at A0 can be overwritten by the phy write function.

Signed-off-by: "Akeem G. Abodunrin" <akeem.g.abodunrin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: X540 devices RX PFC frames pause traffic even if disabled
John Fastabend [Wed, 21 Sep 2011 14:44:10 +0000 (14:44 +0000)]
ixgbe: X540 devices RX PFC frames pause traffic even if disabled

Receiving PFC (priority flow control) frames while the feature
is off should not pause the traffic class. On the X540 devices
the traffic class react to frames if it was previously enabled
because the field is incorrectly cleared.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: DCB X540 devices support max traffic class of 4
John Fastabend [Tue, 27 Sep 2011 03:52:01 +0000 (03:52 +0000)]
ixgbe: DCB X540 devices support max traffic class of 4

X540 devices can only support up to 4 traffic classes and
guarantee a "lossless" traffic class on some platforms.
This patch sets the X540 devices to initialize a max
traffic class value of 4 at probe time.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: fixup hard dependencies on supporting 8 traffic classes
John Fastabend [Tue, 27 Sep 2011 03:51:56 +0000 (03:51 +0000)]
ixgbe: fixup hard dependencies on supporting 8 traffic classes

This patch correctly configures DCB when less than 8 traffic classes
are available in hardware.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: Fix PFC mask generation
Mark Rustad [Tue, 20 Sep 2011 03:00:27 +0000 (03:00 +0000)]
ixgbe: Fix PFC mask generation

Fix PFC mask generation to OR in only a single bit for each priority in
the PFC mask returned via netlink.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>