openwrt/staging/blogic.git
8 years agoMerge branch 'reset_mac_header'
David S. Miller [Sat, 5 Mar 2016 03:45:14 +0000 (22:45 -0500)]
Merge branch 'reset_mac_header'

Zhang Shengju says:

====================
use reset to set header pointers

This patch series replace set function with reset when offset is zero.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agowireless: use reset to set mac header
Zhang Shengju [Thu, 3 Mar 2016 01:16:57 +0000 (01:16 +0000)]
wireless: use reset to set mac header

Since offset is zero, it's not necessary to use set function. Reset
function is straightforward, and will remove the unnecessary add
operation in set function.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomac80211: use reset to set header pointer
Zhang Shengju [Thu, 3 Mar 2016 01:16:56 +0000 (01:16 +0000)]
mac80211: use reset to set header pointer

Since offset is zero, it's not necessary to use set function. Reset
function is straightforward, and will remove the unnecessary add
operation in set function.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomac80211_hwsim: use reset to set mac header
Zhang Shengju [Thu, 3 Mar 2016 01:16:55 +0000 (01:16 +0000)]
mac80211_hwsim: use reset to set mac header

Since offset is zero, it's not necessary to use set function. Reset
function is straightforward, and will remove the unnecessary add
operation in set function.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agovxlan: use reset to set header pointers
Zhang Shengju [Thu, 3 Mar 2016 01:16:54 +0000 (01:16 +0000)]
vxlan: use reset to set header pointers

Since offset is zero, it's not necessary to use set function. Reset
function is straightforward, and will remove the unnecessary add operation
in set function.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'rxrpc-rewrite-20160304' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Fri, 4 Mar 2016 22:19:16 +0000 (17:19 -0500)]
Merge tag 'rxrpc-rewrite-20160304' of git://git./linux/kernel/git/dhowells/linux-fs

David Howells says:

====================
RxRPC: Rewrite part 1

Here's the first set of patches from my RxRPC rewrite, aimed at net-next.
These do some clean ups and bug fixes.  Most of the changes are small, but
there are a couple of bigger changes:

 (*) Convert call flag numbers and event numbers into enums.  Then rename
     the event numbers to all have _EV_ in their name to stop confusion.
     Fix one instance of an event bit being used instead of a flag bit.

 (*) A copy of the Rx protocol header is kept in the sk_buff private data.
     Keep this in host byte order rather than network byte order as it
     makes more sense.  A number of other fields then get converted into
     host byte order too.

     Conversion between host and network byte order is then done at the
     packet reception/generation stage.

This is based on net-next/master
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'DIV_ROUND_UP-uapi'
David S. Miller [Fri, 4 Mar 2016 21:10:37 +0000 (16:10 -0500)]
Merge branch 'DIV_ROUND_UP-uapi'

Nicolas Dichtel says:

====================
uapi: consolidate DIV_ROUND_UP definition

The inital goal was to consolidate ethtool.h uapi header. But I took the
opportunity to remove all duplicate definitions of DIV_ROUND_UP.

v3: add patch #2 and #3

v2: split the patch
    define DIV_ROUND_UP in uapi
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoethtool.h: define INT_MAX for userland
Nicolas Dichtel [Fri, 4 Mar 2016 10:52:19 +0000 (11:52 +0100)]
ethtool.h: define INT_MAX for userland

INT_MAX needs limits.h in userland.
When ethtool.h is included by a userland app, we got the following error:

.../usr/include/linux/ethtool.h: In function 'ethtool_validate_speed':
.../usr/include/linux/ethtool.h:1471:18: error: 'INT_MAX' undeclared (first use in this function)
  return speed <= INT_MAX || speed == SPEED_UNKNOWN
                  ^

Fixes: e02564ee334a ("ethtool: make validate_speed accept all speeds between 0 and INT_MAX")
CC: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrm/vmwgfx: remove userland definition of DIV_ROUND_UP
Nicolas Dichtel [Fri, 4 Mar 2016 10:52:18 +0000 (11:52 +0100)]
drm/vmwgfx: remove userland definition of DIV_ROUND_UP

Let's use __KERNEL_DIV_ROUND_UP, which is defined in uapi/linux/kernel.h.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocxgb4i: don't redefine DIV_ROUND_UP
Nicolas Dichtel [Fri, 4 Mar 2016 10:52:17 +0000 (11:52 +0100)]
cxgb4i: don't redefine DIV_ROUND_UP

let's use the common definition to avoid the following warning during the
compilation:

drivers/scsi/cxgbi/cxgb4i/cxgb4i.c:161:0: warning: "DIV_ROUND_UP" redefined
 #define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 ^
In file included from include/linux/list.h:8:0,
                 from include/linux/module.h:9,
                 from drivers/scsi/cxgbi/cxgb4i/cxgb4i.c:16:
include/linux/kernel.h:67:0: note: this is the location of the previous definition
 #define DIV_ROUND_UP __KERNEL_DIV_ROUND_UP
 ^

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agouapi: define DIV_ROUND_UP for userland
Nicolas Dichtel [Fri, 4 Mar 2016 10:52:16 +0000 (11:52 +0100)]
uapi: define DIV_ROUND_UP for userland

DIV_ROUND_UP is defined in linux/kernel.h only for the kernel.
When ethtool.h is included by a userland app, we got the following error:

include/linux/ethtool.h:1218:8: error: variably modified 'queue_mask' at file scope
  __u32 queue_mask[DIV_ROUND_UP(MAX_NUM_QUEUE, 32)];
        ^

Let's add a common definition in uapi and use it everywhere.

Fixes: ac2c7ad0e5d6 ("net/ethtool: introduce a new ioctl for per queue setting")
CC: Kan Liang <kan.liang@intel.com>
Suggested-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonfp: remove Rolf Neugebauer as co-maintainer
Simon Horman [Fri, 4 Mar 2016 06:58:07 +0000 (15:58 +0900)]
nfp: remove Rolf Neugebauer as co-maintainer

Rolf is no longer in his previous role at Netronome and as far as I know no
longer working on the NFP driver. Thus it does not seem appropriate for him
to be a co-maintainer anymore.

Reviewed-by: Dinan Gunawardena <dinan.gunawardena@netronome.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agorxrpc: Don't try to map ICMP to error as the lower layer already did that
David Howells [Fri, 4 Mar 2016 16:02:03 +0000 (16:02 +0000)]
rxrpc: Don't try to map ICMP to error as the lower layer already did that

In the ICMP message processing code, don't try to map ICMP codes to UNIX
error codes as the caller (IPv4/IPv6) already did that for us (ee_errno).

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Clear the unused part of a sockaddr_rxrpc for memcmp() use
David Howells [Fri, 4 Mar 2016 15:59:49 +0000 (15:59 +0000)]
rxrpc: Clear the unused part of a sockaddr_rxrpc for memcmp() use

Clear the unused part of a sockaddr_rxrpc structs so that memcmp() can be
used to compare them.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: rxkad: Casts are needed when comparing be32 values
David Howells [Fri, 4 Mar 2016 15:59:13 +0000 (15:59 +0000)]
rxrpc: rxkad: Casts are needed when comparing be32 values

Forced casts are needed to avoid sparse warning when directly comparing
be32 values.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: rxkad: The version number in the response should be net byte order
David Howells [Fri, 4 Mar 2016 15:59:00 +0000 (15:59 +0000)]
rxrpc: rxkad: The version number in the response should be net byte order

The version number rxkad places in the response should be network byte
order.

Whilst we're at it, rearrange the code to be more readable.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Use ACCESS_ONCE() when accessing circular buffer pointers
David Howells [Fri, 4 Mar 2016 15:58:06 +0000 (15:58 +0000)]
rxrpc: Use ACCESS_ONCE() when accessing circular buffer pointers

Use ACCESS_ONCE() when accessing the other-end pointer into a circular
buffer as it's possible the other-end pointer might change whilst we're
doing this, and if we access it twice, we might get some weird things
happening.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Adjust some whitespace and comments
David Howells [Fri, 4 Mar 2016 15:56:19 +0000 (15:56 +0000)]
rxrpc: Adjust some whitespace and comments

Remove some excess whitespace, insert some missing spaces and adjust a
couple of comments.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Be more selective about the types of received packets we accept
David Howells [Fri, 4 Mar 2016 15:56:06 +0000 (15:56 +0000)]
rxrpc: Be more selective about the types of received packets we accept

Currently, received RxRPC packets outside the range 1-13 are rejected.
There are, however, holes in the range that should also be rejected - plus
at least one type we don't yet support - so reject these also.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Fix defined range for /proc/sys/net/rxrpc/rx_mtu
David Howells [Fri, 4 Mar 2016 15:55:32 +0000 (15:55 +0000)]
rxrpc: Fix defined range for /proc/sys/net/rxrpc/rx_mtu

The upper bound of the defined range for rx_mtu is being set in the same
member as the lower bound (extra1) rather than the correct place (extra2).
I'm not entirely sure why this compiles.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: The protocol family should be set to PF_RXRPC not PF_UNIX
David Howells [Fri, 4 Mar 2016 15:54:27 +0000 (15:54 +0000)]
rxrpc: The protocol family should be set to PF_RXRPC not PF_UNIX

Fix the protocol family set in the proto_ops for rxrpc to be PF_RXRPC not
PF_UNIX.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Keep the skb private record of the Rx header in host byte order
David Howells [Fri, 4 Mar 2016 15:53:46 +0000 (15:53 +0000)]
rxrpc: Keep the skb private record of the Rx header in host byte order

Currently, a copy of the Rx packet header is copied into the the sk_buff
private data so that we can advance the pointer into the buffer,
potentially discarding the original.  At the moment, this copy is held in
network byte order, but this means we're doing a lot of unnecessary
translations.

The reasons it was done this way are that we need the values in network
byte order occasionally and we can use the copy, slightly modified, as part
of an iov array when sending an ack or an abort packet.

However, it seems more reasonable on review that it would be better kept in
host byte order and that we make up a new header when we want to send
another packet.

To this end, rename the original header struct to rxrpc_wire_header (with
BE fields) and institute a variant called rxrpc_host_header that has host
order fields.  Change the struct in the sk_buff private data into an
rxrpc_host_header and translate the values when filling it in.

This further allows us to keep values kept in various structures in host
byte order rather than network byte order and allows removal of some fields
that are byteswapped duplicates.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Rename call events to begin RXRPC_CALL_EV_
David Howells [Fri, 4 Mar 2016 15:53:46 +0000 (15:53 +0000)]
rxrpc: Rename call events to begin RXRPC_CALL_EV_

Rename call event names to begin RXRPC_CALL_EV_ to distinguish them from the
flags.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Convert call flag and event numbers into enums
David Howells [Fri, 4 Mar 2016 15:53:46 +0000 (15:53 +0000)]
rxrpc: Convert call flag and event numbers into enums

Convert call flag and event numbers into enums and move their definitions
outside of the struct.

Also move the call state enum outside of the struct and add an extra
element to count the number of states.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Fix a case where a call event bit is being used as a flag bit
David Howells [Fri, 4 Mar 2016 15:53:46 +0000 (15:53 +0000)]
rxrpc: Fix a case where a call event bit is being used as a flag bit

Fix a case where RXRPC_CALL_RELEASE (an event) is being used to specify a
flag bit.  RXRPC_CALL_RELEASED should be used instead.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agonet: sched: use pfifo_fast for non real queues
Eric Dumazet [Wed, 2 Mar 2016 16:21:43 +0000 (08:21 -0800)]
net: sched: use pfifo_fast for non real queues

Some devices declare a high number of TX queues, then set a much
lower real_num_tx_queues

This cause setups using fq_codel, sfq or fq as the default qdisc to consume
more memory than really needed.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ipv6: Fix refcnt on host routes
David Ahern [Wed, 2 Mar 2016 19:30:07 +0000 (11:30 -0800)]
net: ipv6: Fix refcnt on host routes

Andrew and Ying Huang's test robot both reported usage count problems that
trace back to the 'keep address on ifdown' patch.

>From Andrew:
We execute CRIU test on linux-next. On the current linux-next kernel
they hangs on creating a network namespace.

The kernel log contains many massages like this:
[ 1036.122108] unregister_netdevice: waiting for lo to become free.
Usage count = 2
[ 1046.165156] unregister_netdevice: waiting for lo to become free.
Usage count = 2
[ 1056.210287] unregister_netdevice: waiting for lo to become free.
Usage count = 2

I tried to revert this patch and the bug disappeared.

Here is a set of commands to reproduce this bug:

[root@linux-next-test linux-next]# uname -a
Linux linux-next-test 4.5.0-rc6-next-20160301+ #3 SMP Wed Mar 2
17:32:18 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

[root@linux-next-test ~]# unshare -n
[root@linux-next-test ~]# ip link set up dev lo
[root@linux-next-test ~]# ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
group default qlen 1
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
[root@linux-next-test ~]# logout
[root@linux-next-test ~]# unshare -n

 -----

The problem is a change made to RTM_DELADDR case in __ipv6_ifa_notify that
was added in an early version of the offending patch and is no longer
needed.

Fixes: f1705ec197e7 ("net: ipv6: Make address flushing on ifdown optional")
Cc: Andrey Wagin <avagin@gmail.com>
Cc: Ying Huang <ying.huang@linux.intel.com>
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Tested-by: Jeremiah Mahler <jmmahler@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ezchip: adapt driver to little endian architecture
Lada Trimasova [Thu, 3 Mar 2016 14:07:46 +0000 (17:07 +0300)]
net: ezchip: adapt driver to little endian architecture

Since ezchip network driver is written with big endian EZChip platform it
is necessary to add support for little endian architecture.

The first issue is that the order of the bits in a bit field is
implementation specific. So all the bit fields are removed.
Named constants are used to access necessary fields.

And the second one is that network byte order is big endian.
For example, data on ethernet is transmitted with most-significant
octet (byte) first. So in case of little endian architecture
it is important to swap data byte order when we read it from
register. In case of unaligned access we can use "get_unaligned_be32"
and in other case we can use function "ioread32_rep" which reads all
data from register and works either with little endian or big endian
architecture.

And then when we are going to write data to register we need to restore
byte order using the function "put_unaligned_be32" in case of
unaligned access and in other case "iowrite32_rep".

The last little fix is a space between type and pointer to observe
coding style.

Signed-off-by: Lada Trimasova <ltrimas@synopsys.com>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Noam Camus <noamc@ezchip.com>
Cc: Tal Zilcer <talz@ezchip.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosh_eth, ravb: Use ARCH_RENESAS
Simon Horman [Wed, 2 Mar 2016 01:28:13 +0000 (10:28 +0900)]
sh_eth, ravb: Use ARCH_RENESAS

Make use of ARCH_RENESAS in place of ARCH_SHMOBILE.

This is part of an ongoing process to migrate from ARCH_SHMOBILE to
ARCH_RENESAS the motivation for which being that RENESAS seems to be a more
appropriate name than SHMOBILE for the majority of Renesas ARM based SoCs.

Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: mellanox: add DEVLINK dependencies
Arnd Bergmann [Wed, 2 Mar 2016 09:40:54 +0000 (10:40 +0100)]
net: mellanox: add DEVLINK dependencies

The new NET_DEVLINK infrastructure can be a loadable module, but the drivers
using it might be built-in, which causes link errors like:

drivers/net/built-in.o: In function `mlx4_load_one':
:(.text+0x2fbfda): undefined reference to `devlink_port_register'
:(.text+0x2fc084): undefined reference to `devlink_port_unregister'
drivers/net/built-in.o: In function `mlxsw_sx_port_remove':
:(.text+0x33a03a): undefined reference to `devlink_port_type_clear'
:(.text+0x33a04e): undefined reference to `devlink_port_unregister'

There are multiple ways to avoid this:

a) add 'depends on NET_DEVLINK || !NET_DEVLINK' dependencies
   for each user
b) use 'select NET_DEVLINK' from each driver that uses it
   and hide the symbol in Kconfig.
c) make NET_DEVLINK a 'bool' option so we don't have to
   list it as a dependency, and rely on the APIs to be
   stubbed out when it is disabled
d) use IS_REACHABLE() rather than IS_ENABLED() to check for
   NET_DEVLINK in include/net/devlink.h

This implements a variation of approach a) by adding an
intermediate symbol that drivers can depend on, and changes
the three drivers using it.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 09d4d087cd48 ("mlx4: Implement devlink interface")
Fixes: c4745500e988 ("mlxsw: Implement devlink interface")
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
David S. Miller [Thu, 3 Mar 2016 21:27:23 +0000 (16:27 -0500)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth-next

Johan Hedberg says:

====================
pull request: bluetooth-next 2016-03-01

Here's our main set of Bluetooth & 802.15.4 patches for the 4.6 kernel.

 - New Bluetooth HCI driver for Intel/AG6xx controllers
 - New Broadcom ACPI IDs
 - LED trigger support for indicating Bluetooth powered state
 - Various fixes in mac802154, 6lowpan and related drivers
 - New USB IDs for AR3012 Bluetooth controllers

Please let me know if there are any issues pulling. Thanks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: relax setup_tc ndo op handle restriction
John Fastabend [Mon, 29 Feb 2016 19:26:13 +0000 (11:26 -0800)]
net: relax setup_tc ndo op handle restriction

I added this check in setup_tc to multiple drivers,

 if (handle != TC_H_ROOT || tc->type != TC_SETUP_MQPRIO)

Unfortunately restricting to TC_H_ROOT like this breaks the old
instantiation of mqprio to setup a hardware qdisc. This patch
relaxes the test to only check the type to make it equivalent
to the check before I broke it. With this the old instantiation
continues to work.

A good smoke test is to setup mqprio with,

# tc qdisc add dev eth4 root mqprio num_tc 8 \
  map 0 1 2 3 4 5 6 7 \
  queues 0@0 1@1 2@2 3@3 4@4 5@5 6@6 7@7

Fixes: e4c6734eaab9 ("net: rework ndo tc op to consume additional qdisc handle paramete")
Reported-by: Singh Krishneil <krishneil.k.singh@intel.com>
Reported-by: Jake Keller <jacob.e.keller@intel.com>
CC: Murali Karicheri <m-karicheri2@ti.com>
CC: Shradha Shah <sshah@solarflare.com>
CC: Or Gerlitz <ogerlitz@mellanox.com>
CC: Ariel Elior <ariel.elior@qlogic.com>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Bruce Allan <bruce.w.allan@intel.com>
CC: Jesse Brandeburg <jesse.brandeburg@intel.com>
CC: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoethernet/atl1c: remove left over dead code
Eric Engestrom [Mon, 29 Feb 2016 16:40:23 +0000 (16:40 +0000)]
ethernet/atl1c: remove left over dead code

Left over from c24588afc536a35c924d014f13b669b20ccf8553
("atl1c: using fixed TXQ configuration for l2cb and l1c")

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/ipv4: remove left over dead code
Eric Engestrom [Mon, 29 Feb 2016 16:38:06 +0000 (16:38 +0000)]
net/ipv4: remove left over dead code

8cc785f6f429c2a3fb81745dc142cbd72a462c4a ("net: ipv4: make the ping
/proc code AF-independent") removed the code using it, but renamed this
variable instead of removing it.

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/rtnetlink: remove dead code
Eric Engestrom [Mon, 29 Feb 2016 16:36:38 +0000 (16:36 +0000)]
net/rtnetlink: remove dead code

3b766cd832328fcb87db3507e7b98cf42f21689d ("net/core: Add reading VF
statistics through the PF netdevice") added that variable but it's never
been used.

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'cxgb4-next'
David S. Miller [Wed, 2 Mar 2016 19:46:30 +0000 (14:46 -0500)]
Merge branch 'cxgb4-next'

Hariprasad Shenai says:

====================
cxgb4/cxgb4vf: Cleanup and minor fixes

This series sets FBMIN to 64 bytes for Chelsio's T6 series of adapters,
check to replenish fl is revised, some code cleanup in cxgb4vf sge
initialization code and removes dead code.

This patch series has been created against net-next tree and includes
patches on cxgb4 and cxgb4vf driver.

We have included all the maintainers of respective drivers. Kindly review
the change and let us know in case of any review comments.
====================

8 years agocxgb4vf: Remove dead functions collect_netdev_[um]c_list_addrs
Hariprasad Shenai [Tue, 1 Mar 2016 11:49:36 +0000 (17:19 +0530)]
cxgb4vf: Remove dead functions collect_netdev_[um]c_list_addrs

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocxgb4vf: Remove redundant adapter ready check during probe
Hariprasad Shenai [Tue, 1 Mar 2016 11:49:35 +0000 (17:19 +0530)]
cxgb4vf: Remove redundant adapter ready check during probe

Function t4vf_wait_dev_ready() is already called in t4vf_prep_adapter(),
no need to call it again in adap_init0().

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocxgb4vf: Make sge init code more readable
Hariprasad Shenai [Tue, 1 Mar 2016 11:49:34 +0000 (17:19 +0530)]
cxgb4vf: Make sge init code more readable

Adds a new function t4vf_fl_pkt_align() and use the same in SGE
initialization code to find out freelist packet alignment

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocxgb4/cxgb4vf: For T6 adapter, set FBMIN to 64 bytes
Hariprasad Shenai [Tue, 1 Mar 2016 11:49:33 +0000 (17:19 +0530)]
cxgb4/cxgb4vf: For T6 adapter, set FBMIN to 64 bytes

T4 and T5 hardware will not coalesce Free List PCI-E Fetch Requests if
the Host Driver provides more Free List Pointers than the Fetch Burst
Minimum value.  So if we set FBMIN to 64 bytes and the Host Driver
supplies 128 bytes of Free List Pointer data, the hardware will issue two
64-byte PCI-E Fetch Requests rather than a single coallesced 128-byte
Fetch Request. T6 fixes this. So, for T4/T5 we set the FBMIN value to 128

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocxgb4/cxgb4vf: Use fl capacity to check if fl needs to be replenished
Hariprasad Shenai [Tue, 1 Mar 2016 11:49:32 +0000 (17:19 +0530)]
cxgb4/cxgb4vf: Use fl capacity to check if fl needs to be replenished

Use freelist capacity instead of freelist size while checking, if
freelist needs to be refilled

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'stmmac-optimizations'
David S. Miller [Wed, 2 Mar 2016 19:21:34 +0000 (14:21 -0500)]
Merge branch 'stmmac-optimizations'

Alexandre TORGUE says:

====================
stmmac: enhance driver performances and update the version

According to Giuseppe, I send the v3 series.

This is a subset of patches to rework the driver in order to improve its
performances and make it more robust under stress conditions.

All patches have been ported on STi mainstream kernel branch and
tested on ARM STiH4xx platforms and newer ones.

This series also updates the driver version and prepares it
to include further development to support new chips.

In detail, these patches are:

o to rework and improve the internal DMA bus settings

  Fine tuning is mandatory on some platforms for both
  performance and stability issues.

o to rework and optimize the descriptor management.

  This will help a lot on performance side and preparing
  the inclusion on the GMAC4.x.

o to add a set of optimizations for both xmit and rx functions.

  These will help a lot on performance side and making the driver
  more robust in case of low memory conditions and under some
  stress test, performed for example on IP-STB.

Below some throughput figures obtained on some boxes before and after
the patches.

                       nuttcp (mbps)       iperf (Mbps)
------------------------------------------------------------------
                      tcp     udp          tcp      udp
                   tx   rx   tx  rx      tx   rx   tx  rx
                    ------------------------------------------
   old             680   800 480  506    760  800   600  700
   new             830   880 540  630    840  880   700   800

V2: - rx_copybreak is now managed by using ethtool.
V3: - improve comments on PCIe detailing that there are no regressions
    - rework some APIs to properly define some params as bool as expected
    - rework the formula to get the element inside the ring. Comparing V2,
patches 4 and 13 have been merged because the same formula have been
used. After this rework, no evident benefit has been noticed in terms
of performances so the table above is still valid. Disassembling the
code for SH4 and ARM, with the new formula just an instr is saved
(depending on compiler flags) and this gives us not so relevanti gain,
for example, on SH4 where some instr are executed in the same pipeline
stage.
Ring sizes are now fixed and maybe they can be reworked to be tuned
w/o using stmmaceth= cmdline option. Indeed, nobody change these sizes
and indeed the numbers selected by default respect the budget and
avoid to pass invalid setup. These are the best driver default sizes
for ring and chain.

====================

8 years agostmmac: update version to Oct_2015
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:43 +0000 (14:27 +0100)]
stmmac: update version to Oct_2015

This patch just updates the driver to the version fully
tested on STi platforms. This version is Oct_2015.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: tune rx copy via threshold.
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:42 +0000 (14:27 +0100)]
stmmac: tune rx copy via threshold.

There is a threshold now used to also limit the skb allocation
when use zero-copy. This is to avoid that there are incoherence
in the ring due to a failure on skb allocation under very
aggressive testing and under low memory conditions.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: do not perform zero-copy for rx frames
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:41 +0000 (14:27 +0100)]
stmmac: do not perform zero-copy for rx frames

This patch is to allow this driver to copy tiny frames during the reception
process. This is giving more stability while stressing the driver on STi
embedded systems.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: fix phy init when attached to a phy
Fabrice Gasnier [Mon, 29 Feb 2016 13:27:40 +0000 (14:27 +0100)]
stmmac: fix phy init when attached to a phy

phy_bus_name can be NULL when "fixed-link" property isn't used.
Then, since "stmmac: do not poll phy handler when attach a switch",
phy_bus_name ptr needs to be checked before strcmp is called.

Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: do not poll phy handler when attach a switch
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:39 +0000 (14:27 +0100)]
stmmac: do not poll phy handler when attach a switch

This patch avoids to call the stmmac_adjust_link when
the driver is connected to a switch by using the FIXED_PHY
support. Prior this patch the phydev->irq was set as PHY_POLL
so periodically the phy handler was invoked spending useless
time because the link cannot actually change.
Note that the stmmac_adjust_link will be called just one
time and this guarantees that the ST glue logic will be
setup according to the mode and speed fixed.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: first frame prep at the end of xmit routine
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:38 +0000 (14:27 +0100)]
stmmac: first frame prep at the end of xmit routine

This patch is to fill the first descriptor just before granting
the DMA engine so at the end of the xmit.
The patch takes care about the algorithm adopted to mitigate the
interrupts, then it fixes the last segment in case of no fragments.
Moreover, this new implementation does not pass any "ter" field when
prepare the descriptors because this is not necessary.
The patch also details the memory barrier in the xmit.

As final results, this patch guarantees the same performances
but fixing a case if small datagram are sent. In fact, this
kind of test is impacted if no coalesce is done.

Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: set dirty index out of the loop
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:37 +0000 (14:27 +0100)]
stmmac: set dirty index out of the loop

The dirty index can be updated out of the loop where all the
tx resources are claimed. This will help on performances too.
Also a useless debug printk has been removed from the main loop.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: optimize tx clean function
Fabrice Gasnier [Mon, 29 Feb 2016 13:27:36 +0000 (14:27 +0100)]
stmmac: optimize tx clean function

This patch "inline" get_tx_owner and get_ls routines. It Results in a
unique read to tdes0, instead of three, to check TX_OWN and LS bits,
and other status bits.

It helps improve driver TX path by removing two uncached read/writes
inside TX clean loop for enhanced descriptors but not for normal ones
because the des1 must be read in any case.

Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: optimize tx desc management
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:35 +0000 (14:27 +0100)]
stmmac: optimize tx desc management

This patch is to optimize the way to manage the TDES inside the
xmit function. When prepare the frame, some settings (e.g. OWN
bit) can be merged. This has been reworked to improve the tx
performances.

Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: merge get_rx_owner into rx_status routine.
Fabrice Gasnier [Mon, 29 Feb 2016 13:27:34 +0000 (14:27 +0100)]
stmmac: merge get_rx_owner into rx_status routine.

The RDES0 register can be read several times while doing RX of a
packet.
This patch slightly improves RX path performance by reading rdes0
once for two operation: check rx owner, get rx status bits.

Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: add is_jumbo field to dma data
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:33 +0000 (14:27 +0100)]
stmmac: add is_jumbo field to dma data

Optimize tx_clean by avoiding a des3 read in stmmac_clean_desc3().

In ring mode, TX, des3 seems only used when xmit a jumbo frame.
In case of normal descriptors, it may also be used for time
stamping.
Clean it in the above two case, without reading it.

Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: add last_segment field to dma data
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:32 +0000 (14:27 +0100)]
stmmac: add last_segment field to dma data

last_segment field is read twice from dma descriptors in stmmac_clean().
Add last_segment to dma data so that this flag is from priv
structure in cache instead of memory.
It avoids reading twice from memory for each loop in stmmac_clean().

Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: add length field to dma data
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:31 +0000 (14:27 +0100)]
stmmac: add length field to dma data

Currently, the code pulls out the length field when
unmapping a buffer directly from the descriptor. This will result
in an uncached read to a dma_alloc_coherent() region. There is no
need to do this, so this patch simply puts the value directly into
a data structure which will hit the cache.

Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: review RX/TX ring management
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:30 +0000 (14:27 +0100)]
stmmac: review RX/TX ring management

This patch is to rework the ring management now optimized.
The indexes into the ring buffer are always incremented, and
the entry is accessed via doing a modulo to find the "real"
position in the ring.
It is inefficient, modulo is an expensive operation.

The formula [(entry + 1) & (size - 1)] is now adopted on
a ring that is power-of-2 in size.
Then, the number of elements cannot be set by command line but
it is fixed.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: change descriptor layout
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:29 +0000 (14:27 +0100)]
stmmac: change descriptor layout

This patch completely changes the descriptor layout to improve
the whole performances due to the single read usage of the
descriptors in critical paths.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: rework DMA bus setting and introduce new platform AXI structure
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:28 +0000 (14:27 +0100)]
stmmac: rework DMA bus setting and introduce new platform AXI structure

This patch restructures the DMA bus settings and this is done
by introducing a new platform structure used for programming
the AXI Bus Mode Register inside the DMA module.
This structure can be populated from device-tree as documented in the
binding txt file.

After initializing the DMA, the AXI register can be optionally tuned
for platform drivers based.
This patch also reworks some parameters to make coherent the DMA
configuration now that AXI register is introduced.
For example, the burst_len is managed by using the mentioned axi
support above; so the snps,burst-len parameter has been removed.
It makes sense to provide the AAL parameter from DT to Address-Aligned
Beats inside the Register0 and review the PBL settings when initialize
the engine.

For PCI glue, rebuilding the story of this setting, it
was added to align a configuration so not for fixing some
known problem. No issue raised after this patch.
It is safe to use the default burst length instead of
tuning it to the maximum value

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: share reset function between dwmac100 and dwmac1000
Giuseppe Cavallaro [Mon, 29 Feb 2016 13:27:27 +0000 (14:27 +0100)]
stmmac: share reset function between dwmac100 and dwmac1000

This patch is to share the same reset procedure between dwmac100 and
dwmac1000 chips.
This will also help on enhancing the driver and support new chips.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'rds-support-FRMR-and-cleanups'
David S. Miller [Wed, 2 Mar 2016 19:13:20 +0000 (14:13 -0500)]
Merge branch 'rds-support-FRMR-and-cleanups'

Santosh Shilimkar says:

====================
RDS: Major clean-up with couple of new features for 4.6

v3:
Re-generated the same series by omitting "-D" option from git format-patch
command. Since first patch has file removals, git apply/am can't deal
with it when formated with '-D' option.

v2:
Dropped module parameter from [PATCH 11/13] as suggested by David Miller

Series is generated against net-next but also applies against Linus's tip
cleanly. Entire patchset is available at below git tree:

 git://git.kernel.org/pub/scm/linux/kernel/git/ssantosh/linux.git for_4.6/net-next/rds_v2

The diff-stat looks bit scary since almost ~4K lines of code is
getting removed. Brief summary of the series:

- Drop the stale iWARP support:
RDS iWarp support code has become stale and non testable for
sometime.  As discussed and agreed earlier on list, am dropping
its support for good. If new iWarp user(s) shows up in future,
the plan is to adapt existing IB RDMA with special sink case.
- RDS gets SO_TIMESTAMP support
- Long due RDS maintainer entry gets updated
- Some RDS IB code refactoring towards new FastReg Memory registration (FRMR)
- Lastly the initial support for FRMR

RDS IB RDMA performance with FRMR is not yet as good as FMR and I do have
some patches in progress to address that. But they are not ready for 4.6
so I left them out of this series.

Also am keeping eye on new CQ API adaptations like other ULPs doing and
will try to adapt RDS for the same most likely in 4.7+ timeframe.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: Support Fastreg MR (FRMR) memory registration mode
Avinash Repaka [Tue, 1 Mar 2016 23:20:54 +0000 (15:20 -0800)]
RDS: IB: Support Fastreg MR (FRMR) memory registration mode

Fastreg MR(FRMR) is another method with which one can
register memory to HCA. Some of the newer HCAs supports only fastreg
mr mode, so we need to add support for it to have RDS functional
on them.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Avinash Repaka <avinash.repaka@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: allocate extra space on queues for FRMR support
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:53 +0000 (15:20 -0800)]
RDS: IB: allocate extra space on queues for FRMR support

Fastreg MR(FRMR) memory registration and invalidation makes use
of work request and completion queues for its operation. Patch
allocates extra queue space towards these operation(s).

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: add Fastreg MR (FRMR) detection support
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:52 +0000 (15:20 -0800)]
RDS: IB: add Fastreg MR (FRMR) detection support

Discovere Fast Memmory Registration support using IB device
IB_DEVICE_MEM_MGT_EXTENSIONS. Certain HCA might support just FRMR
or FMR or both FMR and FRWR. In case both mr type are supported,
default FMR is used.

Default MR is still kept as FMR against what everyone else
is following. Default will be changed to FRMR once the
RDS performance with FRMR is comparable with FMR. The
work is in progress for the same.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: add mr reused stats
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:51 +0000 (15:20 -0800)]
RDS: IB: add mr reused stats

Add MR reuse statistics to RDS IB transport.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: handle the RDMA CM time wait event
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:50 +0000 (15:20 -0800)]
RDS: IB: handle the RDMA CM time wait event

Drop the RDS connection on RDMA_CM_EVENT_TIMEWAIT_EXIT so that
it can reconnect and resume.

While testing fastreg, this error happened in couple of tests but
was getting un-noticed.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: add connection info to ibmr
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:49 +0000 (15:20 -0800)]
RDS: IB: add connection info to ibmr

Preperatory patch for FRMR support. From connection info,
we can retrieve cm_id which contains qp handled needed for
work request posting.

We also need to drop the RDS connection on QP error states
where connection handle becomes useful.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: move FMR code to its own file
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:48 +0000 (15:20 -0800)]
RDS: IB: move FMR code to its own file

No functional change.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: create struct rds_ib_fmr
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:47 +0000 (15:20 -0800)]
RDS: IB: create struct rds_ib_fmr

Keep fmr related filed in its own struct. Fastreg MR structure
will be added to the union.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: Re-organise ibmr code
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:46 +0000 (15:20 -0800)]
RDS: IB: Re-organise ibmr code

No functional changes. This is in preperation towards adding
fastreg memory resgitration support.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: IB: Remove the RDS_IB_SEND_OP dependency
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:45 +0000 (15:20 -0800)]
RDS: IB: Remove the RDS_IB_SEND_OP dependency

This helps to combine asynchronous fastreg MR completion handler
with send completion handler.

No functional change.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMAINTAINERS: update RDS entry
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:44 +0000 (15:20 -0800)]
MAINTAINERS: update RDS entry

Acked-by: Chien Yen <chien.yen@oracle.com>
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: Add support for SO_TIMESTAMP for incoming messages
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:43 +0000 (15:20 -0800)]
RDS: Add support for SO_TIMESTAMP for incoming messages

The SO_TIMESTAMP generates time stamp for each incoming RDS messages
User app can enable it by using SO_TIMESTAMP setsocketopt() at
SOL_SOCKET level. CMSG data of cmsg type SO_TIMESTAMP contains the
time stamp in struct timeval format.

Reviewed-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: Drop stale iWARP RDMA transport
santosh.shilimkar@oracle.com [Tue, 1 Mar 2016 23:20:42 +0000 (15:20 -0800)]
RDS: Drop stale iWARP RDMA transport

RDS iWarp support code has become stale and non testable. As
indicated earlier, am dropping the support for it.

If new iWarp user(s) shows up in future, we can adapat the RDS IB
transprt for the special RDMA READ sink case. iWarp needs an MR
for the RDMA READ sink.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'qed-next'
David S. Miller [Wed, 2 Mar 2016 19:04:19 +0000 (14:04 -0500)]
Merge branch 'qed-next'

Yuval Mintz says:

====================
qed: update series

This patch series tries to improve general configuration by changing
configuration to better suit B0 boards and allow more available
resources to each physical function.
In additition, it contains some small fixes and semantic changes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Remove unused NVM vendor ID
Yuval Mintz [Wed, 2 Mar 2016 18:26:04 +0000 (20:26 +0200)]
qed: Remove unused NVM vendor ID

Remove 2 unused fields from driver code.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Fix error flow on slowpath start
Yuval Mintz [Wed, 2 Mar 2016 18:26:03 +0000 (20:26 +0200)]
qed: Fix error flow on slowpath start

In case of problems when initializing the chip, the error flows aren't
being properly done. Specifically, it's possible that the chip would be
left in a configuration allowing it [internally] to access the host
memory, causing fatal problems in the device that would require power
cycle to overcome.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Move statistics to L2 code
Yuval Mintz [Wed, 2 Mar 2016 18:26:02 +0000 (20:26 +0200)]
qed: Move statistics to L2 code

Current statistics logic is meant for L2, not for all future protocols.
Move this content to the proper designated file.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Support B0 instead of A0
Yuval Mintz [Wed, 2 Mar 2016 18:26:01 +0000 (20:26 +0200)]
qed: Support B0 instead of A0

BB_A0 is a development model that is will not reach actual clients.
In fact, future firmware would simply fail to initialize such chip.

This changes the configuration into B0 instead of A0, and adds a safeguard
against the slim chance someone would actually try this with an A0 adapter
in which case probe would gracefully fail.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Correct BAR sizes for older MFW
Ram Amrani [Wed, 2 Mar 2016 18:26:00 +0000 (20:26 +0200)]
qed: Correct BAR sizes for older MFW

Driver learns the inner bar sized from a register configured by management
firmware, but older versions are not setting this register.
But since we know which values were configured back then, use them instead.

Signed-off-by: Ram Amrani <Ram.Amrani@qlogic.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobatman-adv: clarify CFG80211 dependency
Arnd Bergmann [Wed, 2 Mar 2016 13:54:35 +0000 (14:54 +0100)]
batman-adv: clarify CFG80211 dependency

The driver calls cfg80211_get_station, which may be part of a
module, so we must not enable BATMAN_ADV_BATMAN_V if
BATMAN_ADV=y and CFG80211=m:

net/built-in.o: In function `batadv_v_elp_get_throughput':
(text+0x5c62c): undefined reference to `cfg80211_get_station'

This clarifies the dependency to cover all combinations.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: c833484e5f38 ("batman-adv: ELP - compute the metric based on the estimated throughput")
Acked-by: Antonio Quartulli <a@unstable.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agowan: lmc: Switch to using managed resources
Amitoj Kaur Chawla [Sat, 27 Feb 2016 17:04:16 +0000 (22:34 +0530)]
wan: lmc: Switch to using managed resources

Use managed resource functions devm_kzalloc and pcim_enable_device
to simplify error handling. Subsequently, remove unnecessary
kfree, pci_disable_device and pci_release_regions.

To be compatible with the change, various gotos are replaced with
direct returns and unneeded labels are dropped.

Also, `sc` was only being freed in the probe function and not the
remove function before the change. By using devm_kzalloc this patch
also fixes this memory leak.

Signed-off-by: Amitoj Kaur Chawla <amitoj1606@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge
David S. Miller [Tue, 1 Mar 2016 22:48:00 +0000 (17:48 -0500)]
Merge tag 'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge

Antonio Quartulli says:

====================
batman-adv 20160229

this is our (hopefully) latest batch of patches intended for net-next.

With this patchset we finally introduce B.A.T.M.A.N. V: the latest
version of our routing protocol.
Technical documentation describing the protocol in more detail can
be found in our wiki[1][2][3][4].

For what concerns this pull request, you can find the high level
description right below.

[1] https://www.open-mesh.org/projects/batman-adv/wiki/BATMAN_V
[2] https://www.open-mesh.org/projects/batman-adv/wiki/OGMv2
[3] https://www.open-mesh.org/projects/batman-adv/wiki/ELP
[4] https://www.open-mesh.org/projects/batman-adv/wiki/BATMAN_V_Tests

 ...

With this patchset we finally introduce our new routing protocol:
B.A.T.M.A.N. V. Its implementation started quite some years ago,
but due to the big changes being introduced it took a while to be
discussed, designed, worked, re-worked, tested and debugged (well,
we're never done with the latest). The entire operation has
basically been a team work involving all the core contributors
together with other people interested in the project.

The new protocol is divided into two main subcomponents, called
respectively ELP and OGMv2. The former is in charge of
dealing with the neighbour discovery and link quality estimation,
while the latter implements the algorithm that spreads the
metrics around the network and computes optimal paths.

The biggest change introduced with B.A.T.M.A.N. V is the new
metric: the protocol won't rely on packet loss anymore, but it
will use the estimated throughput extracted directly from the
wifi driver (when available) by querying cfg80211.
Batman-adv will also send some unicast probing packets when
an interface is not used for payload traffic to make sure that
such values are current.

The new protocol can be compiled-in or not like other
features we have and when selected will pull in CFG80211 as
dependency for the reason described above.

Thanks to the big work brought up in the past by Marek Lindner,
batman-adv can easily deal several protocol implementations,
therefore compiling in this new version does not exclude the
older.
This means that the user is offered the option to choose
the protocol when creating the mesh interface (default is the
old one to keep backward compatibility).

Along with the protocol there are some sysfs knobs that are
introduced to fine tune some of its behaviours, but users
are recommended to keep the default values unless they know
what they are doing.

The last patch is about advertising our own patchwork platform
(thanks to Sven Eckelmann for having set that up!) in the
MAINTAINERS file.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: pktgen: use reset to set mac header
Zhang Shengju [Mon, 29 Feb 2016 08:21:30 +0000 (08:21 +0000)]
net: pktgen: use reset to set mac header

Since offset is zero, it's not necessary to use set function.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosch_mqprio: Fix build with older gcc.
David S. Miller [Tue, 1 Mar 2016 22:44:59 +0000 (17:44 -0500)]
sch_mqprio: Fix build with older gcc.

  CC [M]  net/sched/sch_mqprio.o
net/sched/sch_mqprio.c: In function ?mqprio_init?:
net/sched/sch_mqprio.c:145: error: unknown field ?tc? specified in initializer
net/sched/sch_mqprio.c:145: warning: missing braces around initializer
net/sched/sch_mqprio.c:145: warning: (near initialization for ?tc.<anonymous>?)
make[2]: *** [net/sched/sch_mqprio.o] Error 1
make[1]: *** [net/sched] Error 2
make: *** [net] Error 2

Several people reported this, surround the unnamed union
member initialization with braces to fix.

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoof_mdio: kill useless variable in of_mdiobus_register()
Sergei Shtylyov [Sun, 28 Feb 2016 18:56:34 +0000 (21:56 +0300)]
of_mdio: kill useless variable in of_mdiobus_register()

of_mdiobus_register()  declares the 'paddr' variable to hold the result of
the of_get_property()  but only uses it once after that while the function
can be called directly from the *if* statement. Remove that variable and
switch to calling of_find_property() instead since  we don't care about
the "reg" property's value anyway...

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'qed-next'
David S. Miller [Tue, 1 Mar 2016 22:39:51 +0000 (17:39 -0500)]
Merge branch 'qed-next'

Yuval Mintz says:

====================
qed: Attention support patch series

Until now we've only enabled attention generation for the sake of
management firmware indications [required for link notifications].

This series enables [almost] all the attention sources of the HW,
currently for the sake of logging information relating to issues
experienced by HW. In future, infrastructure laid here would also be used
for the sake of the recovery process.

The first patch in the series is a semantic alignemnt of the code.
The later 3 patches incremently create said infrastructure and enrich
the logged information.
Notice #3 contains quite a bit of structures [consisting of ~1K lines]
that will eventually be removed and incorporated in the binary fw file.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Print additional HW attention info
Yuval Mintz [Sun, 28 Feb 2016 10:26:55 +0000 (12:26 +0200)]
qed: Print additional HW attention info

This patch utilizes the attention infrastructure to log additional
information that relates only to specific HW blocks.
For some of those HW blocks, it also stops automatically disabling the
attention generation as the attention is considered benign and thus
should only be logged; No fear of it flooding the system.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Print HW attention reasons
Yuval Mintz [Sun, 28 Feb 2016 10:26:54 +0000 (12:26 +0200)]
qed: Print HW attention reasons

Each HW block contains common information about attention reasons,
raising a bit for each one of the different sub-reasons that caused it
to raise an attention.

This patch extends the infrastructure by allowing logging of the various
reasons causing the HW blocks to generate an attention.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Add support for HW attentions
Yuval Mintz [Sun, 28 Feb 2016 10:26:53 +0000 (12:26 +0200)]
qed: Add support for HW attentions

HW is capable of generating attentnions for a multitude of reasons,
but current driver is enabling attention generation only for management
firmware [required for link notifications].

This patch enables almost all of the possible reasons for HW attentions,
logging the HW block generating the attention and preventing further
attentions from that source [to prevent possible attention flood].
It also lays the infrastructure for additional exploration of the various
attentions.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Semantic refactoring of interrupt code
Yuval Mintz [Sun, 28 Feb 2016 10:26:52 +0000 (12:26 +0200)]
qed: Semantic refactoring of interrupt code

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: remove skb_sender_cpu_clear()
WANG Cong [Sun, 28 Feb 2016 04:19:54 +0000 (20:19 -0800)]
net: remove skb_sender_cpu_clear()

After commit 52bd2d62ce67 ("net: better skb->sender_cpu and skb->napi_id cohabitation")
skb_sender_cpu_clear() becomes empty and can be removed.

Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mlx5-next'
David S. Miller [Tue, 1 Mar 2016 22:28:01 +0000 (17:28 -0500)]
Merge branch 'mlx5-next'

Saeed Mahameed says:

====================
mlx5 driver updates

This series includes some bug fixes and updates for the mlx5 core
and ethernet driver.

From Gal, two fixes that protects the update CQ moderation flows
when it is not allowed.

From Moshe, two fixes for the core and ethernet driver in
non-cached(NC) and write combining(WC) buffers mappings,
which prevents the driver from double memory mappings.

From Or, reduce the firmware command completion timeout.

From Tariq, several small trivial fixes.

Changes from v0:
- "Fix global UAR mapping" commit messages updated to explain ARCH_HAS_IOREMAP_WC usage.
- rebased to commit 8d3f2806f8fb 'Merge branch ethtool-ksettings'

Changes from v1:
- Removed ARCH_HAS_IOREMAP_WC config flag from "Fix global UAR mapping" commit, as it was not accurate to use it.
- Squashed "Fix global UAR mapping" and "net/mlx5: Avoid double mapping of io mapped memory"
- Added more info for "Fix global UAR mapping" in commit message

Changes from v2:
- None. resubmission per Dave's request due to two parallel submissions to mlx5 driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5: Fix global UAR mapping
Moshe Lazer [Tue, 1 Mar 2016 22:13:40 +0000 (00:13 +0200)]
net/mlx5: Fix global UAR mapping

Avoid double mapping of io mapped memory, Device page may be
mapped to non-cached(NC) or to write-combining(WC).
The code before this fix tries to map it both to WC and NC
contrary to what stated in Intel's software developer manual.

Here we remove the global WC mapping of all UARS
"dev->priv.bf_mapping", since UAR mapping should be decided
per UAR (e.g we want different mappings for EQs, CQs vs QPs).

Caller will now have to choose whether to map via
write-combining API or not.

mlx5e SQs will choose write-combining in order to perform
BlueFlame writes.

Fixes: 88a85f99e51f ('TX latency optimization to save DMA reads')
Signed-off-by: Moshe Lazer <moshel@mellanox.com>
Reviewed-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5: Make command timeout way shorter
Or Gerlitz [Tue, 1 Mar 2016 22:13:39 +0000 (00:13 +0200)]
net/mlx5: Make command timeout way shorter

The command timeout is terribly long, whole two hours. Make it 60s so if
things do go wrong, the user gets feedback in relatively short time, so
they can take corrective actions and/or investigate using tools and such.

Fixes: e126ba97dba9 ('mlx5: Add driver for Mellanox Connect-IB adapters')
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Don't modify CQ before it was created
Gal Pressman [Tue, 1 Mar 2016 22:13:38 +0000 (00:13 +0200)]
net/mlx5e: Don't modify CQ before it was created

Calling mlx5e_set_coalesce while the interface is down will result in
modifying CQs that don't exist.

Fixes: f62b8bb8f2d3 ('net/mlx5: Extend mlx5_core to support ConnectX-4
Ethernet functionality')
Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Don't try to modify CQ moderation if it is not supported
Gal Pressman [Tue, 1 Mar 2016 22:13:37 +0000 (00:13 +0200)]
net/mlx5e: Don't try to modify CQ moderation if it is not supported

If CQ moderation is not supported by the device, print a warning on
netdevice load, and return error when trying to modify/query cq
moderation via ethtool.

Fixes: f62b8bb8f2d3 ('net/mlx5: Extend mlx5_core to support ConnectX-4
Ethernet functionality')
Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Set drop RQ's necessary parameters only
Tariq Toukan [Tue, 1 Mar 2016 22:13:36 +0000 (00:13 +0200)]
net/mlx5e: Set drop RQ's necessary parameters only

By its role, there is no need to set all the other parameters
for the drop RQ.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Move common case counters within sq_stats struct
Tariq Toukan [Tue, 1 Mar 2016 22:13:35 +0000 (00:13 +0200)]
net/mlx5e: Move common case counters within sq_stats struct

For data cache locality considerations, we moved the nop and
csum_offload_inner within sq_stats struct as they are more
commonly accessed in xmit path.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Changed naming convention of tx queues in ethtool stats
Tariq Toukan [Tue, 1 Mar 2016 22:13:34 +0000 (00:13 +0200)]
net/mlx5e: Changed naming convention of tx queues in ethtool stats

Instead of the pair (channel, tc), we now use a single number that
goes over all tx queues of a TC, for all TCs.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Placement changed for carrier state updates
Tariq Toukan [Tue, 1 Mar 2016 22:13:33 +0000 (00:13 +0200)]
net/mlx5e: Placement changed for carrier state updates

More proper to declare carrier state UP only after the channels
are ready for traffic.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>