platform/kernel/linux-rpi3.git
9 years agonet: phy: micrel: add has-broadcast-disable flag to type data
Johan Hovold [Wed, 19 Nov 2014 11:59:17 +0000 (12:59 +0100)]
net: phy: micrel: add has-broadcast-disable flag to type data

Add has_broadcast_disable flag to type-data and generic config_init.

This allows us to remove the ksz8081 config_init callback.

Note that ksz8021_config_init is kept for now due to a95a18afe4c8
("phy/micrel: KSZ8031RNL RMII clock reconfiguration bug").

Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: phy: micrel: parse of nodes at probe
Johan Hovold [Wed, 19 Nov 2014 11:59:16 +0000 (12:59 +0100)]
net: phy: micrel: parse of nodes at probe

Parse the "micrel,led-mode" property at probe, rather than at config_init
time in the led-setup helper itself.

Note that the bogus parent->of_node bit is removed.

Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: phy: micrel: add device-type abstraction
Johan Hovold [Wed, 19 Nov 2014 11:59:15 +0000 (12:59 +0100)]
net: phy: micrel: add device-type abstraction

Add structured device-type information and support for generic led-mode
setup to the generic config_init callback.

This is a first step in ultimately getting rid of device-type specific
callbacks.

Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: phy: add static data field to struct phy_driver
Johan Hovold [Wed, 19 Nov 2014 11:59:14 +0000 (12:59 +0100)]
net: phy: add static data field to struct phy_driver

Add static driver-data field to struct phy_driver, which can be used to
store structured device-type information.

Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agosky2: use new netdev_rss_key_fill() helper
Ian Morris [Wed, 19 Nov 2014 09:06:51 +0000 (09:06 +0000)]
sky2: use new netdev_rss_key_fill() helper

Switch to a random RSS key rather than a fixed one.
Using netdev_rss_key_fill helper also ensures that all ports share
a common key.

See also commit 960fb622f85180f36d3aff82af53e2be3db2f888.

Signed-off-by: Ian Morris <ipm@chirality.org.uk>
Cc: Mirko Lindner <mlindner@marvell.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: Eric Dumazet <edumazet@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoenic: support skb->xmit_more
Govindarajulu Varadarajan [Wed, 19 Nov 2014 07:29:32 +0000 (12:59 +0530)]
enic: support skb->xmit_more

Check and update posted_index only when skb->xmit_more is 0 or tx queue is full.

v2:
use txq_map instead of skb_get_queue_mapping(skb)

Signed-off-by: Govindarajulu Varadarajan <_govind@gmx.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agomISDN: Deletion of unnecessary checks before the function call "vfree"
Markus Elfring [Wed, 19 Nov 2014 19:48:26 +0000 (20:48 +0100)]
mISDN: Deletion of unnecessary checks before the function call "vfree"

The vfree() function performs also input parameter validation. Thus the test
around the call is not needed.

This issue was detected by using the Coccinelle software.

Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
David S. Miller [Fri, 21 Nov 2014 17:17:35 +0000 (12:17 -0500)]
Merge branch 'for-davem' of git://git./linux/kernel/git/viro/vfs

9 years agoMerge branch 'bonding_4ad'
David S. Miller [Thu, 20 Nov 2014 00:10:17 +0000 (19:10 -0500)]
Merge branch 'bonding_4ad'

Xie Jianhua says:

====================
bonding: Introduce 4 AD link speed

The speed field of AD Port Key was based on bitmask, it supported 5
kinds of link speed at most, as there were only 5 bits in the speed
field of the AD Port Key.  This patches series change the speed type
(AD_LINK_SPEED_BITMASK) from bitmask to enum type in order to enhance
speed type from 5 to 32, and then introduce 4 AD link speed to fix
agg_bandwidth.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobonding: Introduce 4 AD link speed to fix agg_bandwidth
Jianhua Xie [Wed, 19 Nov 2014 08:48:59 +0000 (16:48 +0800)]
bonding: Introduce 4 AD link speed to fix agg_bandwidth

This patch adds [2.5|20|40|56] Gbps enum definition, and fixes
aggregated bandwidth calculation based on above slave links.

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: David S. Miller <davem@davemloft.net>
Signed-off-by: Jianhua Xie <jianhua.xie@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobonding: change AD_LINK_SPEED_BITMASK to enum to suport more speed
Jianhua Xie [Wed, 19 Nov 2014 08:48:58 +0000 (16:48 +0800)]
bonding: change AD_LINK_SPEED_BITMASK to enum to suport more speed

Port Key was determined as 16 bits according to the link speed,
duplex and user key (which is yet not supported).  In the old
speed field, 5 bits are for speed [1|10|100|1000|10000]Mbps as
below:
--------------------------------------------------------------
Port key :| User key        | Speed         |       Duplex|
--------------------------------------------------------------
    16                  6               1               0
This patch keeps the old layout, but changes AD_LINK_SPEED_BITMASK
from bit type to an enum type.  In this way, the speed field can
expand speed type from 5 to 32.

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: David S. Miller <davem@davemloft.net>
Signed-off-by: Jianhua Xie <jianhua.xie@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobury skb_copy_to_page()
Al Viro [Mon, 10 Nov 2014 22:30:00 +0000 (17:30 -0500)]
bury skb_copy_to_page()

no callers since 3.0

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agofold verify_iovec() into copy_msghdr_from_user()
Al Viro [Tue, 11 Nov 2014 01:23:13 +0000 (20:23 -0500)]
fold verify_iovec() into copy_msghdr_from_user()

... and do the same on the compat side of things.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years ago{compat_,}verify_iovec(): switch to generic copying of iovecs
Al Viro [Mon, 10 Nov 2014 03:33:45 +0000 (22:33 -0500)]
{compat_,}verify_iovec(): switch to generic copying of iovecs

use {compat_,}rw_copy_check_uvector().  As the result, we are
guaranteed that all iovecs seen in ->msg_iov by ->sendmsg()
and ->recvmsg() will pass access_ok().

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoseparate kernel- and userland-side msghdr
Al Viro [Sun, 6 Apr 2014 18:03:05 +0000 (14:03 -0400)]
separate kernel- and userland-side msghdr

Kernel-side struct msghdr is (currently) using the same layout as
userland one, but it's not a one-to-one copy - even without considering
32bit compat issues, we have msg_iov, msg_name and msg_control copied
to kernel[1].  It's fairly localized, so we get away with a few functions
where that knowledge is needed (and we could shrink that set even
more).  Pretty much everything deals with the kernel-side variant and
the few places that want userland one just use a bunch of force-casts
to paper over the differences.

The thing is, kernel-side definition of struct msghdr is *not* exposed
in include/uapi - libc doesn't see it, etc.  So we can add struct user_msghdr,
with proper annotations and let the few places that ever deal with those
beasts use it for userland pointers.  Saner typechecking aside, that will
allow to change the layout of kernel-side msghdr - e.g. replace
msg_iov/msg_iovlen there with struct iov_iter, getting rid of the need
to modify the iovec as we copy data to/from it, etc.

We could introduce kernel_msghdr instead, but that would create much more
noise - the absolute majority of the instances would need to have the
type switched to kernel_msghdr and definition of struct msghdr in
include/linux/socket.h is not going to be seen by userland anyway.

This commit just introduces user_msghdr and switches the few places that
are dealing with userland-side msghdr to it.

[1] actually, it's even trickier than that - we copy msg_control for
sendmsg, but keep the userland address on recvmsg.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agobpf: fix arraymap NULL deref and missing overflow and zero size checks
Alexei Starovoitov [Wed, 19 Nov 2014 01:32:16 +0000 (17:32 -0800)]
bpf: fix arraymap NULL deref and missing overflow and zero size checks

- fix NULL pointer dereference:
kernel/bpf/arraymap.c:41 array_map_alloc() error: potential null dereference 'array'.  (kzalloc returns null)
kernel/bpf/arraymap.c:41 array_map_alloc() error: we previously assumed 'array' could be null (see line 40)

- integer overflow check was missing in arraymap
(hashmap checks for overflow via kmalloc_array())

- arraymap can round_up(value_size, 8) to zero. check was missing.

- hashmap was missing zero size check as well, since roundup_pow_of_two() can
truncate into zero

- found a typo in the arraymap comment and unnecessary empty line

Fix all of these issues and make both overflow checks explicit U32 in size.

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonetlink: Deletion of an unnecessary check before the function call "__module_get"
Markus Elfring [Tue, 18 Nov 2014 20:03:13 +0000 (21:03 +0100)]
netlink: Deletion of an unnecessary check before the function call "__module_get"

The __module_get() function tests whether its argument is NULL and then
returns immediately. Thus the test around the call is not needed.

This issue was detected by using the Coccinelle software.

Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: pktgen: Deletion of an unnecessary check before the function call "proc_remove"
Markus Elfring [Tue, 18 Nov 2014 19:10:34 +0000 (20:10 +0100)]
net: pktgen: Deletion of an unnecessary check before the function call "proc_remove"

The proc_remove() function tests whether its argument is NULL and then
returns immediately. Thus the test around the call is not needed.

This issue was detected by using the Coccinelle software.

Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agousbnet: rtl8150: remove unused variable
Sudip Mukherjee [Tue, 18 Nov 2014 16:25:21 +0000 (21:55 +0530)]
usbnet: rtl8150: remove unused variable

remove unused variable

Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'stmmac-next'
David S. Miller [Wed, 19 Nov 2014 20:04:58 +0000 (15:04 -0500)]
Merge branch 'stmmac-next'

Giuseppe Cavallaro says:

====================
stmmac: update driver documentation

Recently many changes have been done inside the driver
so this patch updates the driver's doc for example reviewing
information for the rx and tx processes that are managed
by napi method, adding new information for missing glue-logic files
etc.
Also this reviews and fixes what is reported when run kernel-doc script.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agostmmac: review driver when run kernel-doc
Giuseppe CAVALLARO [Tue, 18 Nov 2014 08:47:01 +0000 (09:47 +0100)]
stmmac: review driver when run kernel-doc

When run ./scripts/kernel-doc several warnings are reported
so this patch fix them.
Also it reviews many comments and adds new ones.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agostmmac: document common header file
Giuseppe CAVALLARO [Tue, 18 Nov 2014 08:47:00 +0000 (09:47 +0100)]
stmmac: document common header file

This patch adds some useful comments inside the common header
file to provide information about the APIs exposed by the driver.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agostmmac: update driver documentation
Giuseppe CAVALLARO [Tue, 18 Nov 2014 08:46:59 +0000 (09:46 +0100)]
stmmac: update driver documentation

Recently many changes have been done inside the driver
so this patch updates the driver's doc for example reviewing
information for the rx and tx processes that are managed
by napi method, adding new information for missing glue-logic files
etc.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotcp: make connect() mem charging friendly
Eric Dumazet [Tue, 18 Nov 2014 07:06:20 +0000 (23:06 -0800)]
tcp: make connect() mem charging friendly

While working on sk_forward_alloc problems reported by Denys
Fedoryshchenko, we found that tcp connect() (and fastopen) do not call
sk_wmem_schedule() for SYN packet (and/or SYN/DATA packet), so
sk_forward_alloc is negative while connect is in progress.

We can fix this by calling regular sk_stream_alloc_skb() both for the
SYN packet (in tcp_connect()) and the syn_data packet in
tcp_send_syn_data()

Then, tcp_send_syn_data() can avoid copying syn_data as we simply
can manipulate syn_data->cb[] to remove SYN flag (and increment seq)

Instead of open coding memcpy_fromiovecend(), simply use this helper.

This leaves in socket write queue clean fast clone skbs.

This was tested against our fastopen packetdrill tests.

Reported-by: Denys Fedoryshchenko <nuclearcat@nuclearcat.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotun: return NET_XMIT_DROP for dropped packets
Jason Wang [Tue, 18 Nov 2014 05:20:41 +0000 (13:20 +0800)]
tun: return NET_XMIT_DROP for dropped packets

After commit 5d097109257c03a71845729f8db6b5770c4bbedc
("tun: only queue packets on device"), NETDEV_TX_OK was returned for
dropped packets. This will confuse pktgen since dropped packets were
counted as sent ones.

Fixing this by returning NET_XMIT_DROP to let pktgen count it as error
packet.

Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoicmp: Remove some spurious dropped packet profile hits from the ICMP path
Rick Jones [Mon, 17 Nov 2014 22:04:29 +0000 (14:04 -0800)]
icmp: Remove some spurious dropped packet profile hits from the ICMP path

If icmp_rcv() has successfully processed the incoming ICMP datagram, we
should use consume_skb() rather than kfree_skb() because a hit on the likes
of perf -e skb:kfree_skb is not called-for.

Signed-off-by: Rick Jones <rick.jones2@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodev_ioctl: use sizeof(x) instead of sizeof x
Fabian Frederick [Mon, 17 Nov 2014 21:23:17 +0000 (22:23 +0100)]
dev_ioctl: use sizeof(x) instead of sizeof x

Also remove spaces after cast.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/core: include linux/types.h instead of asm/types.h
Fabian Frederick [Mon, 17 Nov 2014 21:08:22 +0000 (22:08 +0100)]
net/core: include linux/types.h instead of asm/types.h

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fix spelling for synchronized
Fabian Frederick [Mon, 17 Nov 2014 21:04:03 +0000 (22:04 +0100)]
net: fix spelling for synchronized

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodccp: spelling s/reseting/resetting
Fabian Frederick [Mon, 17 Nov 2014 21:00:22 +0000 (22:00 +0100)]
dccp: spelling s/reseting/resetting

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodccp: replace min/casting by min_t
Fabian Frederick [Mon, 17 Nov 2014 20:58:37 +0000 (21:58 +0100)]
dccp: replace min/casting by min_t

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodccp: remove blank lines between function/EXPORT_SYMBOL
Fabian Frederick [Mon, 17 Nov 2014 20:54:58 +0000 (21:54 +0100)]
dccp: remove blank lines between function/EXPORT_SYMBOL

See Documentation/CodingStyle chapter 6.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodccp: kerneldoc warning fixes
Fabian Frederick [Mon, 17 Nov 2014 20:51:21 +0000 (21:51 +0100)]
dccp: kerneldoc warning fixes

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge tag 'linux-can-next-for-3.19-20141117' of git://gitorious.org/linux-can/linux...
David S. Miller [Tue, 18 Nov 2014 20:24:53 +0000 (15:24 -0500)]
Merge tag 'linux-can-next-for-3.19-20141117' of git://gitorious.org/linux-can/linux-can-next

Marc Kleine-Budde says:

====================
this is a pull request of 9 patches for net-next/master.

All 9 patches are by Roger Quadros and update the c_can platform
driver. First by improving the initialization sequence of the message
RAM, making use of syscon/regmap. In the later patches support for
various TI SoCs is added.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'fec-next'
David S. Miller [Tue, 18 Nov 2014 20:20:12 +0000 (15:20 -0500)]
Merge branch 'fec-next'

Lothar Waßmann says:

====================
net: fec: assorted cleanup patches

This patch series is a followup to:
<1415350967-2238-1-git-send-email-LW@KARO-electronics.de>
[PATCHv4 1/1] net: fec: fix regression on i.MX28 introduced by rx_copybreak support
to apply the cleanup patches that were originally sent along with the
bugfix patch.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: remove unused return value from swap_buffer()
Lothar Waßmann [Mon, 17 Nov 2014 09:51:24 +0000 (10:51 +0100)]
net: fec: remove unused return value from swap_buffer()

The return value of swap_buffer() is not used by any caller, thus
remove it.

Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: simplify loop counter handling in swap_buffer()
Lothar Waßmann [Mon, 17 Nov 2014 09:51:23 +0000 (10:51 +0100)]
net: fec: simplify loop counter handling in swap_buffer()

Eliminate the DIV_ROUND_UP() and change the loop counter increment to
4 instead. This results in saving 6 instructions in the functions
assembly code.

Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: use swab32s() instead of cpu_to_be32()
Lothar Waßmann [Mon, 17 Nov 2014 09:51:22 +0000 (10:51 +0100)]
net: fec: use swab32s() instead of cpu_to_be32()

when swap_buffer() is being called, we know for sure, that we need to
byte swap the data. Furthermore, this function is called for swapping
data in both directions. Thus cpu_to_be32() is semantically not
correct for all use cases. Use swab32s() to reflect this.

Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: improve access to quirk flags by copying them into fec_enet_private struct
Lothar Waßmann [Mon, 17 Nov 2014 09:51:21 +0000 (10:51 +0100)]
net: fec: improve access to quirk flags by copying them into fec_enet_private struct

Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: change type of 'bufdesc_ex' to bool
Lothar Waßmann [Mon, 17 Nov 2014 09:51:20 +0000 (10:51 +0100)]
net: fec: change type of 'bufdesc_ex' to bool

fep->bufdesc_ex is treated as a boolean value, thus declare it as
such.

Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: properly parenthesize macro args
Lothar Waßmann [Mon, 17 Nov 2014 09:51:19 +0000 (10:51 +0100)]
net: fec: properly parenthesize macro args

Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: consistently use lower case chars as hex digits
Lothar Waßmann [Mon, 17 Nov 2014 09:51:18 +0000 (10:51 +0100)]
net: fec: consistently use lower case chars as hex digits

Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: indentation cleanup
Lothar Waßmann [Mon, 17 Nov 2014 09:51:17 +0000 (10:51 +0100)]
net: fec: indentation cleanup

consistently use TABs for indentation

Signed-off-by: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'ebpf_maps'
David S. Miller [Tue, 18 Nov 2014 18:44:06 +0000 (13:44 -0500)]
Merge branch 'ebpf_maps'

Alexei Starovoitov says:

====================
implementation of eBPF maps

v1->v2:
renamed flags for MAP_UPDATE_ELEM command to be more concise,
clarified commit logs and improved comments in patches 1,3,7
per discussions with Daniel

Old v1 cover:

this set of patches adds implementation of HASH and ARRAY types of eBPF maps
which were described in manpage in commit b4fc1a460f30("Merge branch 'bpf-next'")

The difference vs previous version of these patches from August:
- added 'flags' attribute to BPF_MAP_UPDATE_ELEM
- in HASH type implementation removed per-map kmem_cache.
  I was doing kmem_cache_create() for every map to enable selective slub
  debugging to check for overflows and leaks. Now it's not needed, so just
  use normal kmalloc() for map elements.
- added ARRAY type which was mentioned in manpage, but wasn't public yet
- added map testsuite and removed temporary bits from test_stubs

Note, eBPF programs cannot be attached to events yet.
It will come in the next set.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobpf: remove test map scaffolding and user proper types
Alexei Starovoitov [Fri, 14 Nov 2014 01:36:50 +0000 (17:36 -0800)]
bpf: remove test map scaffolding and user proper types

proper types and function helpers are ready. Use them in verifier testsuite.
Remove temporary stubs

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobpf: allow eBPF programs to use maps
Alexei Starovoitov [Fri, 14 Nov 2014 01:36:49 +0000 (17:36 -0800)]
bpf: allow eBPF programs to use maps

expose bpf_map_lookup_elem(), bpf_map_update_elem(), bpf_map_delete_elem()
map accessors to eBPF programs

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobpf: add a testsuite for eBPF maps
Alexei Starovoitov [Fri, 14 Nov 2014 01:36:48 +0000 (17:36 -0800)]
bpf: add a testsuite for eBPF maps

. check error conditions and sanity of hash and array map APIs
. check large maps (that kernel gracefully switches to vmalloc from kmalloc)
. check multi-process parallel access and stress test

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobpf: fix BPF_MAP_LOOKUP_ELEM command return code
Alexei Starovoitov [Fri, 14 Nov 2014 01:36:47 +0000 (17:36 -0800)]
bpf: fix BPF_MAP_LOOKUP_ELEM command return code

fix errno of BPF_MAP_LOOKUP_ELEM command as bpf manpage
described it in commit b4fc1a460f30("Merge branch 'bpf-next'"):
-----
BPF_MAP_LOOKUP_ELEM
    int bpf_lookup_elem(int fd, void *key, void *value)
    {
        union bpf_attr attr = {
            .map_fd = fd,
            .key = ptr_to_u64(key),
            .value = ptr_to_u64(value),
        };

        return bpf(BPF_MAP_LOOKUP_ELEM, &attr, sizeof(attr));
    }
    bpf() syscall looks up an element with given key in  a  map  fd.
    If  element  is found it returns zero and stores element's value
    into value.  If element is not found  it  returns  -1  and  sets
    errno to ENOENT.

and further down in manpage:

   ENOENT For BPF_MAP_LOOKUP_ELEM or BPF_MAP_DELETE_ELEM,  indicates  that
          element with given key was not found.
-----

In general all BPF commands return ENOENT when map element is not found
(including BPF_MAP_GET_NEXT_KEY and BPF_MAP_UPDATE_ELEM with
 flags == BPF_MAP_UPDATE_ONLY)

Subsequent patch adds a testsuite to check return values for all of
these combinations.

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobpf: add array type of eBPF maps
Alexei Starovoitov [Fri, 14 Nov 2014 01:36:46 +0000 (17:36 -0800)]
bpf: add array type of eBPF maps

add new map type BPF_MAP_TYPE_ARRAY and its implementation

- optimized for fastest possible lookup()
  . in the future verifier/JIT may recognize lookup() with constant key
    and optimize it into constant pointer. Can optimize non-constant
    key into direct pointer arithmetic as well, since pointers and
    value_size are constant for the life of the eBPF program.
    In other words array_map_lookup_elem() may be 'inlined' by verifier/JIT
    while preserving concurrent access to this map from user space

- two main use cases for array type:
  . 'global' eBPF variables: array of 1 element with key=0 and value is a
    collection of 'global' variables which programs can use to keep the state
    between events
  . aggregation of tracing events into fixed set of buckets

- all array elements pre-allocated and zero initialized at init time

- key as an index in array and can only be 4 byte

- map_delete_elem() returns EINVAL, since elements cannot be deleted

- map_update_elem() replaces elements in an non-atomic way
  (for atomic updates hashtable type should be used instead)

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobpf: add hashtable type of eBPF maps
Alexei Starovoitov [Fri, 14 Nov 2014 01:36:45 +0000 (17:36 -0800)]
bpf: add hashtable type of eBPF maps

add new map type BPF_MAP_TYPE_HASH and its implementation

- maps are created/destroyed by userspace. Both userspace and eBPF programs
  can lookup/update/delete elements from the map

- eBPF programs can be called in_irq(), so use spin_lock_irqsave() mechanism
  for concurrent updates

- key/value are opaque range of bytes (aligned to 8 bytes)

- user space provides 3 configuration attributes via BPF syscall:
  key_size, value_size, max_entries

- map takes care of allocating/freeing key/value pairs

- map_update_elem() must fail to insert new element when max_entries
  limit is reached to make sure that eBPF programs cannot exhaust memory

- map_update_elem() replaces elements in an atomic way

- optimized for speed of lookup() which can be called multiple times from
  eBPF program which itself is triggered by high volume of events
  . in the future JIT compiler may recognize lookup() call and optimize it
    further, since key_size is constant for life of eBPF program

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobpf: add 'flags' attribute to BPF_MAP_UPDATE_ELEM command
Alexei Starovoitov [Fri, 14 Nov 2014 01:36:44 +0000 (17:36 -0800)]
bpf: add 'flags' attribute to BPF_MAP_UPDATE_ELEM command

the current meaning of BPF_MAP_UPDATE_ELEM syscall command is:
either update existing map element or create a new one.
Initially the plan was to add a new command to handle the case of
'create new element if it didn't exist', but 'flags' style looks
cleaner and overall diff is much smaller (more code reused), so add 'flags'
attribute to BPF_MAP_UPDATE_ELEM command with the following meaning:
 #define BPF_ANY 0 /* create new element or update existing */
 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
 #define BPF_EXIST 2 /* update existing element */

bpf_update_elem(fd, key, value, BPF_NOEXIST) call can fail with EEXIST
if element already exists.

bpf_update_elem(fd, key, value, BPF_EXIST) can fail with ENOENT
if element doesn't exist.

Userspace will call it as:
int bpf_update_elem(int fd, void *key, void *value, __u64 flags)
{
    union bpf_attr attr = {
        .map_fd = fd,
        .key = ptr_to_u64(key),
        .value = ptr_to_u64(value),
        .flags = flags;
    };

    return bpf(BPF_MAP_UPDATE_ELEM, &attr, sizeof(attr));
}

First two bits of 'flags' are used to encode style of bpf_update_elem() command.
Bits 2-63 are reserved for future use.

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Tue, 18 Nov 2014 18:37:29 +0000 (13:37 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2014-11-18

This series contains updates to i40e only.

Shannon provides a patch to clean up the driver to only warn once that
PTP is not supported when linked at 100Mbps.

Mitch provides a fix for i40e where the VF interrupt processing takes
a long time and it is possible that we could lose a VFLR event if it
happens while processing a VFLR on another VF.  To correct this situation,
we enable the VFLR interrupt cause before we begin processing any pending
resets.

Neerav provides several patches to update DCB support in i40e.  When
there are DCB configuration changes based on DCBx, the firmware suspends
the port's Tx and generates an event to the PF.  The PF is then
responsible to reconfigure the PF VSIs and switching topology as per the
updated DCB configuration and then resume the port's Tx by calling the
"Resume Port Tx" AQ command, so add this call to the flow that handles
DCB re-configuration in the PF.  Allow the driver to query and use DCB
configuration from firmware when firmware DCBx agent is in CEE mode.
Add a check whether LLDP Agent's default AdminStatus is enabled or
disabled on a given port, and sets DCBx status to disabled if the
status is disabled.  Fix an issue when the port TC configuration
changes as a result of DCBx and the driver modifies the enabled TCs for
the VEBs it manages but does not update the enabled_tc value that
was cached on a per VEB basis.  Add a new PF state so that if a port's
Tx is in suspended state the Tx queue disable flow would just put the
request for the queue to be disabled and return without waiting for the
queue to be actually disabled.  Allows the driver to enable/disable
the XPS based on the number of TCs being enabled for the given VSI.

v2: Dropped patch "i40e: Handle a single mss packet with more than 8 frags"
    while we rework the patch after we test a bit more based on feedback from
    Eric Dumazet.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoPPC: bpf_jit_comp: Unify BPF_MOD | BPF_X and BPF_DIV | BPF_X
Denis Kirjanov [Mon, 17 Nov 2014 20:07:41 +0000 (23:07 +0300)]
PPC: bpf_jit_comp: Unify BPF_MOD | BPF_X and BPF_DIV | BPF_X

Reduce duplicated code by unifying
BPF_ALU | BPF_MOD | BPF_X and BPF_ALU | BPF_DIV | BPF_X

CC: Alexei Starovoitov<alexei.starovoitov@gmail.com>
CC: Daniel Borkmann<dborkman@redhat.com>
CC: Philippe Bergheaud<felix@linux.vnet.ibm.com>
Signed-off-by: Denis Kirjanov <kda@linux-powerpc.org>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoi40e: Set XPS bit mask to zero in DCB mode
Neerav Parikh [Wed, 12 Nov 2014 00:19:02 +0000 (00:19 +0000)]
i40e: Set XPS bit mask to zero in DCB mode

Due to DCBX configuration change if the VSI needs to use more than 1 TC;
it needs to disable the XPS maps that were set when operating in 1 TC mode.
Without disabling XPS the netdev layer will select queues based on those
settings and not use the TC queue mapping to make the queue selection.

This patch allows the driver to enable/disable the XPS based on the number
of TCs being enabled for the given VSI.

Change-ID: Idc4dec47a672d2a509f6d7fe11ed1ee65b4f0e08
Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Tested-By: Jack Morgan <jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: Prevent link flow control settings when PFC is enabled
Neerav Parikh [Wed, 12 Nov 2014 00:18:57 +0000 (00:18 +0000)]
i40e: Prevent link flow control settings when PFC is enabled

When PFC is enabled we should not proceed with setting the link flow control
parameters.  Also, always report the link flow Tx/Rx settings as off when
PFC is enabled.

Change-ID: Ib09ec58afdf0b2e587ac9d8851a5c80ad58206c4
Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Tested-By: Jack Morgan <jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: Do not disable/enable FCoE VSI with DCB reconfig
Neerav Parikh [Wed, 12 Nov 2014 00:18:51 +0000 (00:18 +0000)]
i40e: Do not disable/enable FCoE VSI with DCB reconfig

FCoE VSI Tx queue disable times out when reconfiguring as a result of
DCB TC configuration change event.

The hardware allows us to skip disabling and enabling of Tx queues for
VSIs with single TC enabled. As FCoE VSI is configured to have only
single TC we skip it from disable/enable flow.

Change-ID: Ia73ff3df8785ba2aa3db91e6f2c9005e61ebaec2
Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Tested-By: Jack Morgan <jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: Modify Tx disable wait flow in case of DCB reconfiguration
Neerav Parikh [Wed, 12 Nov 2014 00:18:46 +0000 (00:18 +0000)]
i40e: Modify Tx disable wait flow in case of DCB reconfiguration

When DCB TC configuration changes the firmware suspends the port's Tx.
Now, as DCB TCs may have changed the PF driver tries to reconfigure the
TC configuration of the VSIs it manages. As part of this process it disables
the VSI queues but the Tx queue disable will not complete as the port's
Tx has been suspended. So, waiting for Tx queues to go to disable state
in this flow may lead to detection of Tx queue disable timeout errors.

Hence, this patch adds a new PF state so that if a port's Tx is in
suspended state the Tx queue disable flow would just put the request for
the queue to be disabled and return without waiting for the queue to be
actually disabled.
Once the VSI(s) TC reconfiguration has been done and driver has called
firmware AQC "Resume PF Traffic" the driver checks the Tx queues requested
to be disabled are actually disabled before re-enabling them again.

Change-ID: If3e03ce4813a4e342dbd5a1eb1d2861e952b7544
Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Tested-By: Jack Morgan <jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: Update VEB's enabled_tc after reconfiguration
Neerav Parikh [Wed, 12 Nov 2014 00:18:41 +0000 (00:18 +0000)]
i40e: Update VEB's enabled_tc after reconfiguration

When the port TC configuration changes as a result of DCBx the driver
modifies the enabled TCs for the VEBs it manages. But, in the process
it did not update the enabled_tc value that it caches on a per VEB basis.

So, when the next reconfiguration event occurs where the number of TC
value is same as the value cached in enabled_tc for a given VEB; driver
does not modify it's TC configuration by calling appropriate AQ command
believing it is running with the same configuration as requested.
Now, as the VEB is not actually enabled for the TCs that are there any
TC configuration command for VSI attached to that VEB with TCs that are
not enabled for the VEB fails.

This patch fixes this issue.

Change-ID: Ife5694469b05494228e0d850429ea1734738cf29
Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Tested-By: Jack Morgan <jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: Check for LLDP AdminStatus before querying DCBX
Neerav Parikh [Wed, 12 Nov 2014 00:18:30 +0000 (00:18 +0000)]
i40e: Check for LLDP AdminStatus before querying DCBX

This patch adds a check whether LLDP Agent's default AdminStatus is
enabled or disabled on a given port. If it is disabled then it sets
the DCBX status to disabled as well; and would not query firmware for
any DCBX configuration data.

Change-ID: I73c0b9f0adbf4cae177d14914b20a48c9a8f50fd
Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Tested-By: Jack Morgan <jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: Add support to firmware CEE DCBX mode
Neerav Parikh [Wed, 12 Nov 2014 00:18:25 +0000 (00:18 +0000)]
i40e: Add support to firmware CEE DCBX mode

This patch allows i40e driver to query and use DCB configuration from
firmware when firmware DCBX agent is in CEE mode.

Change-ID: I30f92a67eb890f0f024f35339696e6e83d49a274
Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Tested-By: Jack Morgan <jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: Resume Port Tx after DCB event
Neerav Parikh [Wed, 12 Nov 2014 00:18:20 +0000 (00:18 +0000)]
i40e: Resume Port Tx after DCB event

When there are DCB configuration changes based on DCBX the firmware suspends
the port's Tx and generates an event to the PF. The PF is then responsible
to reconfigure the PF VSIs and switching topology as per the updated DCB
configuration and then resume the port's Tx by calling the "Resume Port Tx"
AQ command.

This patch adds this call to the flow that handles DCB re-configuration in
the PF.

Change-ID: I5b860ad48abfbf379b003143c4d3453e2ed5cc1c
Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Tested-By: Jack Morgan <jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: Bump version to 1.1.23
Catherine Sullivan [Tue, 11 Nov 2014 03:15:06 +0000 (03:15 +0000)]
i40e: Bump version to 1.1.23

Bumping minor version as this will be the second SW release and it
should be 1.

Change-ID: If0bd102095d2f059ae0c9b7f4ad625535ffbbdee
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: re-enable VFLR interrupt sooner
Mitch Williams [Tue, 11 Nov 2014 03:15:04 +0000 (03:15 +0000)]
i40e: re-enable VFLR interrupt sooner

VF interrupt processing takes a looooong time, and it's possible that we
could lose a VFLR event if it happens while we're processing a VFLR on
another VF. This would leave the VF in a semi-permanent reset state,
which would not be cleared until yet another VF experiences a VFLR.

To correct this situation, we enable the VFLR interrupt cause before we
begin processing any pending resets. This means that any VFLR that
occurs during reset processing will generate another interrupt and this
routine will get called again.

This change may cause a spurious interrupt when multiple VFLRs occur
very close together in time. If this happens, then this routine will be
called again and it will detect no outstanding VFLR events and do
nothing. No harm, no foul.

Change-ID: Id0451f3e6e73a2cf6db1668296c71e129b59dc19
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: only warn once of PTP nonsupport in 100Mbit speed
Shannon Nelson [Tue, 11 Nov 2014 03:15:03 +0000 (03:15 +0000)]
i40e: only warn once of PTP nonsupport in 100Mbit speed

Only warn once that PTP is not supported when linked at 100Mbit.

Yes, using a static this way means that this once-only message is not
port specific, but once only for the life of the driver, regardless of
the number of ports.  That should be plenty.

Change-ID: Ie6476530056df408452e195ef06afd4f57caa4b2
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agonet: can: c_can: Add support for TI am4372 DCAN
Roger Quadros [Mon, 17 Nov 2014 12:09:03 +0000 (14:09 +0200)]
net: can: c_can: Add support for TI am4372 DCAN

AM4372 SoC has 2 DCAN modules. Add compatible id and
raminit driver data for it. The driver data is same as AM3352
but this gives us flexibility to add AM4372 specific quirks
if required later.

Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agocan: c_can: Add support for TI am3352 DCAN
Roger Quadros [Fri, 7 Nov 2014 14:49:22 +0000 (16:49 +0200)]
can: c_can: Add support for TI am3352 DCAN

AM3352 SoC has 2 DCAN modules. Add compatible id and
raminit driver data for am3352 DCAN.

Signed-off-by: Roger Quadros <rogerq@ti.com>
Acked-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agocan: c_can: Add support for TI DRA7 DCAN
Roger Quadros [Fri, 7 Nov 2014 14:49:21 +0000 (16:49 +0200)]
can: c_can: Add support for TI DRA7 DCAN

DRA7 SoC has 2 CAN IPs. Provide compatible IDs and RAMINIT
register data for both.

Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agocan: c_can: Disable pins when CAN interface is down
Roger Quadros [Fri, 14 Nov 2014 15:40:13 +0000 (17:40 +0200)]
can: c_can: Disable pins when CAN interface is down

DRA7 CAN IP suffers from a problem which causes it to be prevented
from fully turning OFF (i.e. stuck in transition) if the module was
disabled while there was traffic on the CAN_RX line.

To work around this issue we select the SLEEP pin state by default
on probe and use the DEFAULT pin state on CAN up and back to the
SLEEP pin state on CAN down.

Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agocan: c_can: Add support for START pulse in RAMINIT sequence
Roger Quadros [Fri, 7 Nov 2014 14:49:19 +0000 (16:49 +0200)]
can: c_can: Add support for START pulse in RAMINIT sequence

Some SoCs e.g. (TI DRA7xx) need a START pulse to start the
RAMINIT sequence i.e. START bit must be set and cleared before
checking for the DONE bit status.

Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agocan: c_can: Add syscon/regmap RAMINIT mechanism
Roger Quadros [Fri, 14 Nov 2014 15:37:39 +0000 (17:37 +0200)]
can: c_can: Add syscon/regmap RAMINIT mechanism

Some TI SoCs like DRA7 have a RAMINIT register specification
different from the other AMxx SoCs and as expected by the
existing driver.

To add more insanity, this register is shared with other
IPs like DSS, PCIe and PWM.

Provides a more generic mechanism to specify the RAMINIT
register location and START/DONE bit position and use the
syscon/regmap framework to access the register.

Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agocan: c_can: Add RAMINIT register information to driver data
Roger Quadros [Fri, 7 Nov 2014 14:49:17 +0000 (16:49 +0200)]
can: c_can: Add RAMINIT register information to driver data

Some platforms (e.g. TI) need special RAMINIT register handling.
Provide a way to store RAMINIT register description in driver data.

Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agocan: c_can: Introduce c_can_driver_data structure
Roger Quadros [Fri, 7 Nov 2014 14:49:16 +0000 (16:49 +0200)]
can: c_can: Introduce c_can_driver_data structure

We want to have more data than just can_dev_id to be present
in the driver data e.g. TI platforms need RAMINIT register
description. Introduce the c_can_driver_data structure and move
the can_dev_id into it.

Tidy up the way it is used on probe().

Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agocan: c_can: Add timeout to c_can_hw_raminit_ti()
Roger Quadros [Fri, 7 Nov 2014 14:49:15 +0000 (16:49 +0200)]
can: c_can: Add timeout to c_can_hw_raminit_ti()

TI's RAMINIT DONE mechanism is buggy on AM43xx SoC and may not always
be set after the START bit is set. Although it seems to work fine even
in that case. So add a timeout mechanism to c_can_hw_raminit_wait_ti().
Don't bail out in that failure case but just print an error message.

Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agoMerge branch 'rss_key_fill'
David S. Miller [Sun, 16 Nov 2014 20:59:19 +0000 (15:59 -0500)]
Merge branch 'rss_key_fill'

Eric Dumazet says:

====================
net: provide common RSS key infrastructure

RSS (Receive Side Scaling) uses a 40 bytes key to provide hash for incoming
packets to select appropriate incoming queue on NIC.

Hash algo (Toeplitz) is also well known and documented by Microsoft
(search for "Verifying the RSS Hash Calculation")

Problem is that some drivers use a well known key.
It makes very easy for attackers to target one particular RX queue,
knowing that number of RX queues is a power of two, or at least some
small number.

Other drivers use a random value per port, making difficult
tuning on bonding setups.

Lets add a common infrastructure, so that host gets an unique
RSS key, and drivers do not have to worry about this.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovmxnet3: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:18 +0000 (06:23 -0800)]
vmxnet3: use netdev_rss_key_fill() helper

Use of well known RSS key increases attack surface.
Switch to a random one, using generic helper so that all
ports share a common key.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Shreyas Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agosfc: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:17 +0000 (06:23 -0800)]
sfc: use netdev_rss_key_fill() helper

Use netdev_rss_key_fill() helper, as it provides better support for some
bonding setups.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agomlx4: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:16 +0000 (06:23 -0800)]
mlx4: use netdev_rss_key_fill() helper

Use of well known RSS key increases attack surface.
Switch to a random one, using generic helper so that all
ports share a common key.

Also provide ethtool -x support to fetch RSS key

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoixgbe: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:15 +0000 (06:23 -0800)]
ixgbe: use netdev_rss_key_fill() helper

Use of well known RSS key increases attack surface.
Switch to a random one, using generic helper so that all
ports share a common key.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoigb: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:14 +0000 (06:23 -0800)]
igb: use netdev_rss_key_fill() helper

Use of well known RSS key increases attack surface.
Switch to a random one, using generic helper so that all
ports share a common key.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoi40e: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:13 +0000 (06:23 -0800)]
i40e: use netdev_rss_key_fill() helper

Use of well known RSS key increases attack surface.
Switch to a random one, using generic helper so that all
ports share a common key.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agofm10k: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:12 +0000 (06:23 -0800)]
fm10k: use netdev_rss_key_fill() helper

Use of well known RSS key increases attack surface.
Switch to a random one, using generic helper so that all
ports share a common key.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoe100e: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:11 +0000 (06:23 -0800)]
e100e: use netdev_rss_key_fill() helper

Use of well known RSS key increases attack surface.
Switch to a random one, using generic helper so that all
ports share a common key.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobe2net:use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:10 +0000 (06:23 -0800)]
be2net:use netdev_rss_key_fill() helper

Use netdev_rss_key_fill() helper, as it provides better support for some
bonding setups.
Rename rss_hkey local variable to rss_key to have consistent name among
drivers.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Sathya Perla <sathya.perla@emulex.com>
Cc: Subbu Seetharaman <subbu.seetharaman@emulex.com>
Cc: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobna: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:09 +0000 (06:23 -0800)]
bna: use netdev_rss_key_fill() helper

Use netdev_rss_key_fill() helper, as it provides better support for some
bonding setups.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Rasesh Mody <rasesh.mody@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotg3: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:08 +0000 (06:23 -0800)]
tg3: use netdev_rss_key_fill() helper

Use of well known RSS key increases attack surface.
Switch to a random one, using generic helper so that all
ports share a common key.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Prashant Sreedharan <prashant@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobnx2x: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:07 +0000 (06:23 -0800)]
bnx2x: use netdev_rss_key_fill() helper

Use netdev_rss_key_fill() helper, as it provides better support for some
bonding setups.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ariel Elior <ariel.elior@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoamd-xgbe: use netdev_rss_key_fill() helper
Eric Dumazet [Sun, 16 Nov 2014 14:23:06 +0000 (06:23 -0800)]
amd-xgbe: use netdev_rss_key_fill() helper

Use netdev_rss_key_fill() helper, as it provides better support for some
bonding setups.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Lendacky, Thomas <Thomas.Lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: provide a per host RSS key generic infrastructure
Eric Dumazet [Sun, 16 Nov 2014 14:23:05 +0000 (06:23 -0800)]
net: provide a per host RSS key generic infrastructure

RSS (Receive Side Scaling) typically uses Toeplitz hash and a 40 or 52 bytes
RSS key.

Some drivers use a constant (and well known key), some drivers use a random
key per port, making bonding setups hard to tune. Well known keys increase
attack surface, considering that number of queues is usually a power of two.

This patch provides infrastructure to help drivers doing the right thing.

netdev_rss_key_fill() should be used by drivers to initialize their RSS key,
even if they provide ethtool -X support to let user redefine the key later.

A new /proc/sys/net/core/netdev_rss_key file can be used to get the host
RSS key even for drivers not providing ethtool -x support, in case some
applications want to precisely setup flows to match some RX queues.

Tested:

myhost:~# cat /proc/sys/net/core/netdev_rss_key
11:63:99:bb:79:fb:a5:a7:07:45:b2:20:bf:02:42:2d:08:1a:dd:19:2b:6b:23:ac:56:28:9d:70:c3:ac:e8:16:4b:b7:c1:10:53:a4:78:41:36:40:74:b6:15:ca:27:44:aa:b3:4d:72

myhost:~# ethtool -x eth0
RX flow hash indirection table for eth0 with 8 RX ring(s):
    0:      0     1     2     3     4     5     6     7
RSS hash key:
11:63:99:bb:79:fb:a5:a7:07:45:b2:20:bf:02:42:2d:08:1a:dd:19:2b:6b:23:ac:56:28:9d:70:c3:ac:e8:16:4b:b7:c1:10:53:a4:78:41

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'mv88e6171_temps'
David S. Miller [Sun, 16 Nov 2014 20:48:01 +0000 (15:48 -0500)]
Merge branch 'mv88e6171_temps'

Andrew Lunn says:

====================
Add temperature reading and registers dump to mv88e6171

These patches centralize the temperature sensor reading code, and then
make use of it with the mv88e6171 which has a compatible
sensor. Additionally, support is added for reading the mv88e6171 via
ethtool.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: dsa: mv88e6171: Add support for reading switch registers
Andrew Lunn [Sat, 15 Nov 2014 21:24:53 +0000 (22:24 +0100)]
net: dsa: mv88e6171: Add support for reading switch registers

The infrastructure can now report switch registers to ethtool.
Add support for it to the mv88e6171 driver.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: dsa: mv88e6171: Add support for reading the temperature
Andrew Lunn [Sat, 15 Nov 2014 21:24:52 +0000 (22:24 +0100)]
net: dsa: mv88e6171: Add support for reading the temperature

This chip also has a temperature sensor which can be read using the
common code. In order to use it, add the needed mutex protection for
accessing registers via the shared code.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: dsa: Centralise code for reading the temperature sensor
Andrew Lunn [Sat, 15 Nov 2014 21:24:51 +0000 (22:24 +0100)]
net: dsa: Centralise code for reading the temperature sensor

The method to read the temperature used in the mve6123_61_65 driver
can also be used for other chips. Move the code into the shared code
base of mv88e6xxx.c.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: dsa: replace count*size kzalloc by kcalloc
Fabian Frederick [Fri, 14 Nov 2014 18:38:23 +0000 (19:38 +0100)]
net: dsa: replace count*size kzalloc by kcalloc

kcalloc manages count*sizeof overflow.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: dsa: replace count*size kmalloc by kmalloc_array
Fabian Frederick [Fri, 14 Nov 2014 18:36:42 +0000 (19:36 +0100)]
net: dsa: replace count*size kmalloc by kmalloc_array

kmalloc_array manages count*sizeof overflow.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoopenvswitch: use PTR_ERR_OR_ZERO
Fabian Frederick [Fri, 14 Nov 2014 18:32:58 +0000 (19:32 +0100)]
openvswitch: use PTR_ERR_OR_ZERO

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotipc: allow one link per bearer to neighboring nodes
Holger Brunck [Fri, 14 Nov 2014 17:33:19 +0000 (18:33 +0100)]
tipc: allow one link per bearer to neighboring nodes

There is no reason to limit the amount of possible links to a
neighboring node to 2. If we have more then two bearers we can also
establish more links.

Signed-off-by: Holger Brunck <holger.brunck@keymile.com>
Reviewed-By: Jon Maloy <jon.maloy@ericsson.com>
cc: Ying Xue <ying.xue@windriver.com>
cc: Erik Hugne <erik.hugne@ericsson.com>
cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoRevert "fast_hash: avoid indirect function calls"
Jay Vosburgh [Fri, 14 Nov 2014 19:05:06 +0000 (11:05 -0800)]
Revert "fast_hash: avoid indirect function calls"

This reverts commit e5a2c899957659cd1a9f789bc462f9c0b35f5150.

Commit e5a2c899 introduced an alternative_call, arch_fast_hash2,
that selects between __jhash2 and __intel_crc4_2_hash based on the
X86_FEATURE_XMM4_2.

Unfortunately, the alternative_call system does not appear to be
suitable for use with C functions, as register usage is not handled
properly for the called functions.  The __jhash2 function in particular
clobbers registers that are not preserved when called via
alternative_call, resulting in a panic for direct callers of
arch_fast_hash2 on older CPUs lacking sse4_2.  It is possible that
__intel_crc4_2_hash works merely by chance because it uses fewer
registers.

This commit was suggested as the source of the problem by Jesse
Gross <jesse@nicira.com>.

Signed-off-by: Jay Vosburgh <jay.vosburgh@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoopenvswitch: Fix build failure.
Pravin B Shelar [Fri, 14 Nov 2014 06:21:30 +0000 (22:21 -0800)]
openvswitch: Fix build failure.

Add dependency on INET to fix following build error. I have also
fixed MPLS dependency.

ERROR: "ip_route_output_flow" [net/openvswitch/openvswitch.ko]
undefined!
make[1]: *** [__modpost] Error 1

Reported-by: Jim Davis <jim.epost@gmail.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Fri, 14 Nov 2014 06:01:12 +0000 (01:01 -0500)]
Merge git://git./linux/kernel/git/davem/net

Conflicts:
drivers/net/ethernet/chelsio/cxgb4vf/sge.c
drivers/net/ethernet/intel/ixgbe/ixgbe_phy.c

sge.c was overlapping two changes, one to use the new
__dev_alloc_page() in net-next, and one to use s->fl_pg_order in net.

ixgbe_phy.c was a set of overlapping whitespace changes.

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Linus Torvalds [Fri, 14 Nov 2014 02:07:52 +0000 (18:07 -0800)]
Merge tag 'for_linus' of git://git./linux/kernel/git/mst/vhost

Pull virtio bugfix from Michael S Tsirkin:
 "This fixes a crash in virtio console multi-channel mode that got
  introduced in -rc1"

* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
  virtio_console: move early VQ enablement