]> git.karo-electronics.de Git - linux-beck.git/log
linux-beck.git
10 years agonet: filter: rework/optimize internal BPF interpreter's instruction set
Alexei Starovoitov [Fri, 28 Mar 2014 17:58:25 +0000 (18:58 +0100)]
net: filter: rework/optimize internal BPF interpreter's instruction set

This patch replaces/reworks the kernel-internal BPF interpreter with
an optimized BPF instruction set format that is modelled closer to
mimic native instruction sets and is designed to be JITed with one to
one mapping. Thus, the new interpreter is noticeably faster than the
current implementation of sk_run_filter(); mainly for two reasons:

1. Fall-through jumps:

  BPF jump instructions are forced to go either 'true' or 'false'
  branch which causes branch-miss penalty. The new BPF jump
  instructions have only one branch and fall-through otherwise,
  which fits the CPU branch predictor logic better. `perf stat`
  shows drastic difference for branch-misses between the old and
  new code.

2. Jump-threaded implementation of interpreter vs switch
   statement:

  Instead of single table-jump at the top of 'switch' statement,
  gcc will now generate multiple table-jump instructions, which
  helps CPU branch predictor logic.

Note that the verification of filters is still being done through
sk_chk_filter() in classical BPF format, so filters from user- or
kernel space are verified in the same way as we do now, and same
restrictions/constraints hold as well.

We reuse current BPF JIT compilers in a way that this upgrade would
even be fine as is, but nevertheless allows for a successive upgrade
of BPF JIT compilers to the new format.

The internal instruction set migration is being done after the
probing for JIT compilation, so in case JIT compilers are able to
create a native opcode image, we're going to use that, and in all
other cases we're doing a follow-up migration of the BPF program's
instruction set, so that it can be transparently run in the new
interpreter.

In short, the *internal* format extends BPF in the following way (more
details can be taken from the appended documentation):

  - Number of registers increase from 2 to 10
  - Register width increases from 32-bit to 64-bit
  - Conditional jt/jf targets replaced with jt/fall-through
  - Adds signed > and >= insns
  - 16 4-byte stack slots for register spill-fill replaced
    with up to 512 bytes of multi-use stack space
  - Introduction of bpf_call insn and register passing convention
    for zero overhead calls from/to other kernel functions
  - Adds arithmetic right shift and endianness conversion insns
  - Adds atomic_add insn
  - Old tax/txa insns are replaced with 'mov dst,src' insn

Performance of two BPF filters generated by libpcap resp. bpf_asm
was measured on x86_64, i386 and arm32 (other libpcap programs
have similar performance differences):

fprog #1 is taken from Documentation/networking/filter.txt:
tcpdump -i eth0 port 22 -dd

fprog #2 is taken from 'man tcpdump':
tcpdump -i eth0 'tcp port 22 and (((ip[2:2] - ((ip[0]&0xf)<<2)) -
   ((tcp[12]&0xf0)>>2)) != 0)' -dd

Raw performance data from BPF micro-benchmark: SK_RUN_FILTER on the
same SKB (cache-hit) or 10k SKBs (cache-miss); time in ns per call,
smaller is better:

--x86_64--
         fprog #1  fprog #1   fprog #2  fprog #2
         cache-hit cache-miss cache-hit cache-miss
old BPF      90       101        192       202
new BPF      31        71         47        97
old BPF jit  12        34         17        44
new BPF jit TBD

--i386--
         fprog #1  fprog #1   fprog #2  fprog #2
         cache-hit cache-miss cache-hit cache-miss
old BPF     107       136        227       252
new BPF      40       119         69       172

--arm32--
         fprog #1  fprog #1   fprog #2  fprog #2
         cache-hit cache-miss cache-hit cache-miss
old BPF     202       300        475       540
new BPF     180       270        330       470
old BPF jit  26       182         37       202
new BPF jit TBD

Thus, without changing any userland BPF filters, applications on
top of AF_PACKET (or other families) such as libpcap/tcpdump, cls_bpf
classifier, netfilter's xt_bpf, team driver's load-balancing mode,
and many more will have better interpreter filtering performance.

While we are replacing the internal BPF interpreter, we also need
to convert seccomp BPF in the same step to make use of the new
internal structure since it makes use of lower-level API details
without being further decoupled through higher-level calls like
sk_unattached_filter_{create,destroy}(), for example.

Just as for normal socket filtering, also seccomp BPF experiences
a time-to-verdict speedup:

05-sim-long_jumps.c of libseccomp was used as micro-benchmark:

  seccomp_rule_add_exact(ctx,...
  seccomp_rule_add_exact(ctx,...

  rc = seccomp_load(ctx);

  for (i = 0; i < 10000000; i++)
     syscall(199, 100);

'short filter' has 2 rules
'large filter' has 200 rules

'short filter' performance is slightly better on x86_64/i386/arm32
'large filter' is much faster on x86_64 and i386 and shows no
               difference on arm32

--x86_64-- short filter
old BPF: 2.7 sec
 39.12%  bench  libc-2.15.so       [.] syscall
  8.10%  bench  [kernel.kallsyms]  [k] sk_run_filter
  6.31%  bench  [kernel.kallsyms]  [k] system_call
  5.59%  bench  [kernel.kallsyms]  [k] trace_hardirqs_on_caller
  4.37%  bench  [kernel.kallsyms]  [k] trace_hardirqs_off_caller
  3.70%  bench  [kernel.kallsyms]  [k] __secure_computing
  3.67%  bench  [kernel.kallsyms]  [k] lock_is_held
  3.03%  bench  [kernel.kallsyms]  [k] seccomp_bpf_load
new BPF: 2.58 sec
 42.05%  bench  libc-2.15.so       [.] syscall
  6.91%  bench  [kernel.kallsyms]  [k] system_call
  6.25%  bench  [kernel.kallsyms]  [k] trace_hardirqs_on_caller
  6.07%  bench  [kernel.kallsyms]  [k] __secure_computing
  5.08%  bench  [kernel.kallsyms]  [k] sk_run_filter_int_seccomp

--arm32-- short filter
old BPF: 4.0 sec
 39.92%  bench  [kernel.kallsyms]  [k] vector_swi
 16.60%  bench  [kernel.kallsyms]  [k] sk_run_filter
 14.66%  bench  libc-2.17.so       [.] syscall
  5.42%  bench  [kernel.kallsyms]  [k] seccomp_bpf_load
  5.10%  bench  [kernel.kallsyms]  [k] __secure_computing
new BPF: 3.7 sec
 35.93%  bench  [kernel.kallsyms]  [k] vector_swi
 21.89%  bench  libc-2.17.so       [.] syscall
 13.45%  bench  [kernel.kallsyms]  [k] sk_run_filter_int_seccomp
  6.25%  bench  [kernel.kallsyms]  [k] __secure_computing
  3.96%  bench  [kernel.kallsyms]  [k] syscall_trace_exit

--x86_64-- large filter
old BPF: 8.6 seconds
    73.38%    bench  [kernel.kallsyms]  [k] sk_run_filter
    10.70%    bench  libc-2.15.so       [.] syscall
     5.09%    bench  [kernel.kallsyms]  [k] seccomp_bpf_load
     1.97%    bench  [kernel.kallsyms]  [k] system_call
new BPF: 5.7 seconds
    66.20%    bench  [kernel.kallsyms]  [k] sk_run_filter_int_seccomp
    16.75%    bench  libc-2.15.so       [.] syscall
     3.31%    bench  [kernel.kallsyms]  [k] system_call
     2.88%    bench  [kernel.kallsyms]  [k] __secure_computing

--i386-- large filter
old BPF: 5.4 sec
new BPF: 3.8 sec

--arm32-- large filter
old BPF: 13.5 sec
 73.88%  bench  [kernel.kallsyms]  [k] sk_run_filter
 10.29%  bench  [kernel.kallsyms]  [k] vector_swi
  6.46%  bench  libc-2.17.so       [.] syscall
  2.94%  bench  [kernel.kallsyms]  [k] seccomp_bpf_load
  1.19%  bench  [kernel.kallsyms]  [k] __secure_computing
  0.87%  bench  [kernel.kallsyms]  [k] sys_getuid
new BPF: 13.5 sec
 76.08%  bench  [kernel.kallsyms]  [k] sk_run_filter_int_seccomp
 10.98%  bench  [kernel.kallsyms]  [k] vector_swi
  5.87%  bench  libc-2.17.so       [.] syscall
  1.77%  bench  [kernel.kallsyms]  [k] __secure_computing
  0.93%  bench  [kernel.kallsyms]  [k] sys_getuid

BPF filters generated by seccomp are very branchy, so the new
internal BPF performance is better than the old one. Performance
gains will be even higher when BPF JIT is committed for the
new structure, which is planned in future work (as successive
JIT migrations).

BPF has also been stress-tested with trinity's BPF fuzzer.

Joint work with Daniel Borkmann.

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Hagen Paul Pfeifer <hagen@jauu.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Paul Moore <pmoore@redhat.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: H. Peter Anvin <hpa@linux.intel.com>
Cc: linux-kernel@vger.kernel.org
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: isdn: use sk_unattached_filter api
Daniel Borkmann [Fri, 28 Mar 2014 17:58:24 +0000 (18:58 +0100)]
net: isdn: use sk_unattached_filter api

Similarly as in ppp, we need to migrate the ISDN/PPP code to make use
of the sk_unattached_filter api in order to decouple having direct
filter structure access. By using sk_unattached_filter_{create,destroy},
we can allow for the possibility to jit compile filters for faster
filter verdicts as well.

Joint work with Alexei Starovoitov.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Cc: Karsten Keil <isdn@linux-pingi.de>
Cc: isdn4linux@listserv.isdn4linux.de
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ppp: use sk_unattached_filter api
Daniel Borkmann [Fri, 28 Mar 2014 17:58:23 +0000 (18:58 +0100)]
net: ppp: use sk_unattached_filter api

For the ppp driver, there are currently two open-coded BPF filters in use,
that is, pass_filter and active_filter. Migrate both to make proper use
of sk_unattached_filter_{create,destroy} API so that the actual BPF code
is decoupled from direct access, and filters can be jited as a side-effect
by the internal filter compiler.

Joint work with Alexei Starovoitov.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: linux-ppp@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ptp: do not reimplement PTP/BPF classifier
Daniel Borkmann [Fri, 28 Mar 2014 17:58:22 +0000 (18:58 +0100)]
net: ptp: do not reimplement PTP/BPF classifier

There are currently pch_gbe, cpts, and ixp4xx_eth drivers that open-code
and reimplement a BPF classifier for the PTP protocol. Since all of them
effectively do the very same thing and load the very same PTP/BPF filter,
we can just consolidate that code by introducing ptp_classify_raw() in
the time-stamping core framework which can be used in drivers.

As drivers get initialized after bootstrapping the core networking
subsystem, they can make use of ptp_insns wrapped through
ptp_classify_raw(), which allows to simplify and remove PTP classifier
setup code in drivers.

Joint work with Alexei Starovoitov.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Cc: Richard Cochran <richard.cochran@omicron.at>
Cc: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ptp: use sk_unattached_filter_create() for BPF
Daniel Borkmann [Fri, 28 Mar 2014 17:58:21 +0000 (18:58 +0100)]
net: ptp: use sk_unattached_filter_create() for BPF

This patch migrates an open-coded sk_run_filter() implementation with
proper use of the BPF API, that is, sk_unattached_filter_create(). This
migration is needed, as we will be internally transforming the filter
to a different representation, and therefore needs to be decoupled.

It is okay to do so as skb_timestamping_init() is called during
initialization of the network stack in core initcall via sock_init().
This would effectively also allow for PTP filters to be jit compiled if
bpf_jit_enable is set.

For better readability, there are also some newlines introduced, also
ptp_classify.h is only in kernel space.

Joint work with Alexei Starovoitov.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Cc: Richard Cochran <richard.cochran@omicron.at>
Cc: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: filter: move filter accounting to filter core
Daniel Borkmann [Fri, 28 Mar 2014 17:58:20 +0000 (18:58 +0100)]
net: filter: move filter accounting to filter core

This patch basically does two things, i) removes the extern keyword
from the include/linux/filter.h file to be more consistent with the
rest of Joe's changes, and ii) moves filter accounting into the filter
core framework.

Filter accounting mainly done through sk_filter_{un,}charge() take
care of the case when sockets are being cloned through sk_clone_lock()
so that removal of the filter on one socket won't result in eviction
as it's still referenced by the other.

These functions actually belong to net/core/filter.c and not
include/net/sock.h as we want to keep all that in a central place.
It's also not in fast-path so uninlining them is fine and even allows
us to get rd of sk_filter_release_rcu()'s EXPORT_SYMBOL and a forward
declaration.

Joint work with Alexei Starovoitov.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: filter: keep original BPF program around
Daniel Borkmann [Fri, 28 Mar 2014 17:58:19 +0000 (18:58 +0100)]
net: filter: keep original BPF program around

In order to open up the possibility to internally transform a BPF program
into an alternative and possibly non-trivial reversible representation, we
need to keep the original BPF program around, so that it can be passed back
to user space w/o the need of a complex decoder.

The reason for that use case resides in commit a8fc92778080 ("sk-filter:
Add ability to get socket filter program (v2)"), that is, the ability
to retrieve the currently attached BPF filter from a given socket used
mainly by the checkpoint-restore project, for example.

Therefore, we add two helpers sk_{store,release}_orig_filter for taking
care of that. In the sk_unattached_filter_create() case, there's no such
possibility/requirement to retrieve a loaded BPF program. Therefore, we
can spare us the work in that case.

This approach will simplify and slightly speed up both, sk_get_filter()
and sock_diag_put_filterinfo() handlers as we won't need to successively
decode filters anymore through sk_decode_filter(). As we still need
sk_decode_filter() later on, we're keeping it around.

Joint work with Alexei Starovoitov.

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: filter: add jited flag to indicate jit compiled filters
Daniel Borkmann [Fri, 28 Mar 2014 17:58:18 +0000 (18:58 +0100)]
net: filter: add jited flag to indicate jit compiled filters

This patch adds a jited flag into sk_filter struct in order to indicate
whether a filter is currently jited or not. The size of sk_filter is
not being expanded as the 32 bit 'len' member allows upper bits to be
reused since a filter can currently only grow as large as BPF_MAXINSNS.

Therefore, there's enough room also for other in future needed flags to
reuse 'len' field if necessary. The jited flag also allows for having
alternative interpreter functions running as currently, we can only
detect jit compiled filters by testing fp->bpf_func to not equal the
address of sk_run_filter().

Joint work with Alexei Starovoitov.

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
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 [Sat, 29 Mar 2014 22:48:54 +0000 (18:48 -0400)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

Conflicts:
drivers/net/ethernet/marvell/mvneta.c

The mvneta.c conflict is a case of overlapping changes,
a conversion to devm_ioremap_resource() vs. a conversion
to netdev_alloc_pcpu_stats.

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agophy/at8031: enable at8031 to work on interrupt mode
Zhao Qiang [Fri, 28 Mar 2014 07:39:41 +0000 (15:39 +0800)]
phy/at8031: enable at8031 to work on interrupt mode

The at8031 can work on polling mode and interrupt mode.
Add ack_interrupt and config intr funcs to enable
interrupt mode for it.

Signed-off-by: Zhao Qiang <B45475@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: fix checkpatch errors of "foo*" and "foo * bar"
Wang Yufen [Fri, 28 Mar 2014 04:07:04 +0000 (12:07 +0800)]
ipv6: fix checkpatch errors of "foo*" and "foo * bar"

ERROR: "(foo*)" should be "(foo *)"
ERROR: "foo * bar" should be "foo *bar"

Suggested-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Wang Yufen <wangyufen@huawei.com>
Acked-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: fix checkpatch errors of brace and trailing statements
Wang Yufen [Fri, 28 Mar 2014 04:07:03 +0000 (12:07 +0800)]
ipv6: fix checkpatch errors of brace and trailing statements

ERROR: open brace '{' following enum go on the same line
ERROR: open brace '{' following struct go on the same line
ERROR: trailing statements should be on next line

Signed-off-by: Wang Yufen <wangyufen@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: fix checkpatch errors comments and space
Wang Yufen [Fri, 28 Mar 2014 04:07:02 +0000 (12:07 +0800)]
ipv6: fix checkpatch errors comments and space

WARNING: please, no space before tabs
WARNING: please, no spaces at the start of a line
ERROR: spaces required around that ':' (ctx:VxW)
ERROR: spaces required around that '>' (ctx:VxV)
ERROR: spaces required around that '>=' (ctx:VxV)

Signed-off-by: Wang Yufen <wangyufen@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'netpoll-next'
David S. Miller [Sat, 29 Mar 2014 22:00:37 +0000 (18:00 -0400)]
Merge branch 'netpoll-next'

Eric W. Biederman says:

====================
netpoll: Cleanups and fixes

This should be a small set of safe cleanups and fixes to netpoll.

The fixes are vlan headers are now always inserted when needed, and
napi polling is always avoided when network devices are closed.

There are a bunch of little cleanups removing unnecessary code, fixing
function naming, not taking unnecessary locks and removing general
silliness.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetpoll: Respect NETIF_F_LLTX
Eric W. Biederman [Thu, 27 Mar 2014 22:42:20 +0000 (15:42 -0700)]
netpoll: Respect NETIF_F_LLTX

Stop taking the transmit lock when a network device has specified
NETIF_F_LLTX.

If no locks needed to trasnmit a packet this is the ideal scenario for
netpoll as all packets can be trasnmitted immediately.

Even if some locks are needed in ndo_start_xmit skipping any unnecessary
serialization is desirable for netpoll as it makes it more likely a
debugging packet may be trasnmitted immediately instead of being
deferred until later.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetpoll: Remove strong unnecessary assumptions about skbs
Eric W. Biederman [Thu, 27 Mar 2014 22:41:04 +0000 (15:41 -0700)]
netpoll: Remove strong unnecessary assumptions about skbs

Remove the assumption that the skbs that make it to
netpoll_send_skb_on_dev are allocated with find_skb, such that
skb->users == 1 and nothing is attached that would prevent the skbs from
being freed from hard irq context.

Remove this assumption by replacing __kfree_skb on error paths with
dev_kfree_skb_irq (in hard irq context) and kfree_skb (in process
context).

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetpoll: Rename netpoll_rx_enable/disable to netpoll_poll_disable/enable
Eric W. Biederman [Thu, 27 Mar 2014 22:39:03 +0000 (15:39 -0700)]
netpoll: Rename netpoll_rx_enable/disable to netpoll_poll_disable/enable

The netpoll_rx_enable and netpoll_rx_disable functions have always
controlled polling the network drivers transmit and receive queues.

Rename them to netpoll_poll_enable and netpoll_poll_disable to make
their functionality clear.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetpoll: Move rx enable/disable into __dev_close_many
Eric W. Biederman [Thu, 27 Mar 2014 22:38:17 +0000 (15:38 -0700)]
netpoll: Move rx enable/disable into __dev_close_many

Today netpoll_rx_enable and netpoll_rx_disable are called from
dev_close and and __dev_close, and not from dev_close_many.

Move the calls into __dev_close_many so that we have a single call
site to maintain, and so that dev_close_many gains this protection as
well.  Which importantly makes batched network device deletes safe.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetpoll: Only call ndo_start_xmit from a single place
Eric W. Biederman [Thu, 27 Mar 2014 22:37:28 +0000 (15:37 -0700)]
netpoll: Only call ndo_start_xmit from a single place

Factor out the code that needs to surround ndo_start_xmit
from netpoll_send_skb_on_dev into netpoll_start_xmit.

It is an unfortunate fact that as the netpoll code has been maintained
the primary call site ndo_start_xmit learned how to handle vlans
and timestamps but the second call of ndo_start_xmit in queue_process
did not.

With the introduction of netpoll_start_xmit this associated logic now
happens at both call sites of ndo_start_xmit and should make it easy
for that to continue into the future.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetpoll: Remove gfp parameter from __netpoll_setup
Eric W. Biederman [Thu, 27 Mar 2014 22:36:38 +0000 (15:36 -0700)]
netpoll: Remove gfp parameter from __netpoll_setup

The gfp parameter was added in:
commit 47be03a28cc6c80e3aa2b3e8ed6d960ff0c5c0af
Author: Amerigo Wang <amwang@redhat.com>
Date:   Fri Aug 10 01:24:37 2012 +0000

    netpoll: use GFP_ATOMIC in slave_enable_netpoll() and __netpoll_setup()

    slave_enable_netpoll() and __netpoll_setup() may be called
    with read_lock() held, so should use GFP_ATOMIC to allocate
    memory. Eric suggested to pass gfp flags to __netpoll_setup().

Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The reason for the gfp parameter was removed in:
commit c4cdef9b7183159c23c7302aaf270d64c549f557
Author: dingtianhong <dingtianhong@huawei.com>
Date:   Tue Jul 23 15:25:27 2013 +0800

    bonding: don't call slave_xxx_netpoll under spinlocks

    The slave_xxx_netpoll will call synchronize_rcu_bh(),
    so the function may schedule and sleep, it should't be
    called under spinlocks.

    bond_netpoll_setup() and bond_netpoll_cleanup() are always
    protected by rtnl lock, it is no need to take the read lock,
    as the slave list couldn't be changed outside rtnl lock.

Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Nothing else that calls __netpoll_setup or ndo_netpoll_setup
requires a gfp paramter, so remove the gfp parameter from both
of these functions making the code clearer.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'skb_cow_head'
David S. Miller [Sat, 29 Mar 2014 21:50:14 +0000 (17:50 -0400)]
Merge branch 'skb_cow_head'

Francois Romieu says:

====================
remove open-coded skb_cow_head.

As per http://marc.info/?l=linux-netdev&m=139440579104701.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agowimax/i2400m: remove open-coded skb_cow_head.
françois romieu [Sat, 29 Mar 2014 11:26:30 +0000 (12:26 +0100)]
wimax/i2400m: remove open-coded skb_cow_head.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Inaky Perez-Gonzalez <inaky.perez-gonzalez@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotg3: remove open-coded skb_cow_head.
françois romieu [Sat, 29 Mar 2014 11:26:29 +0000 (12:26 +0100)]
tg3: remove open-coded skb_cow_head.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Nithin Nayak Sujir <nsujir@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: remove open-coded skb_cow_head.
françois romieu [Sat, 29 Mar 2014 11:26:28 +0000 (12:26 +0100)]
bna: remove open-coded skb_cow_head.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlge: remove open-coded skb_cow_head.
françois romieu [Sat, 29 Mar 2014 11:26:27 +0000 (12:26 +0100)]
qlge: remove open-coded skb_cow_head.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Cc: Shahed Shaikh <shahed.shaikh@qlogic.com>
Cc: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agojme: remove open-coded skb_cow_head.
françois romieu [Sat, 29 Mar 2014 11:26:26 +0000 (12:26 +0100)]
jme: remove open-coded skb_cow_head.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Guo-Fu Tseng <cooldavid@cooldavid.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoatl1e: remove open-coded skb_cow_head.
françois romieu [Sat, 29 Mar 2014 11:26:16 +0000 (12:26 +0100)]
atl1e: remove open-coded skb_cow_head.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Chris Snook <chris.snook@gmail.com>
Cc: Jay Cliburn <jcliburn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoatl1c: remove open-coded skb_cow_head.
françois romieu [Sat, 29 Mar 2014 11:26:15 +0000 (12:26 +0100)]
atl1c: remove open-coded skb_cow_head.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Jay Cliburn <jcliburn@gmail.com>
Cc: Chris Snook <chris.snook@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoatl1: remove open-coded skb_cow_head.
françois romieu [Sat, 29 Mar 2014 11:26:14 +0000 (12:26 +0100)]
atl1: remove open-coded skb_cow_head.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Chris Snook <chris.snook@gmail.com>
Cc: Jay Cliburn <jcliburn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: mvneta: use devm_ioremap_resource() instead of of_iomap()
Thomas Petazzoni [Thu, 27 Mar 2014 10:39:29 +0000 (11:39 +0100)]
net: mvneta: use devm_ioremap_resource() instead of of_iomap()

The mvneta driver currently uses of_iomap(), which has two drawbacks:
it doesn't request the resource, and it isn't devm-style so some error
handling is needed.

This commit switches to use devm_ioremap_resource() instead, which
automatically requests the resource (so the I/O registers region shows
up properly in /proc/iomem), and also is devm-style, which allows to
get rid of some error handling to unmap the I/O registers region.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Fri, 28 Mar 2014 22:09:37 +0000 (15:09 -0700)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) We've discovered a common error in several networking drivers, they
    put VLAN offload features into ->vlan_features, which would suggest
    that they support offloading 2 or more levels of VLAN encapsulation.
    Not only do these devices not do that, but we don't have the
    infrastructure yet to handle that at all.

    Fixes from Vlad Yasevich.

 2) Fix tcpdump crash with bridging and vlans, also from Vlad.

 3) Some MAINTAINERS updates for random32 and bonding.

 4) Fix late reseeds of prandom generator, from Sasha Levin.

 5) Bridge doesn't handle stacked vlans properly, fix from Toshiaki
    Makita.

 6) Fix deadlock in openvswitch, from Flavio Leitner.

 7) get_timewait4_sock() doesn't report delay times correctly, fix from
    Eric Dumazet.

 8) Duplicate address detection and addrconf verification need to run in
    contexts where RTNL can be obtained.  Move them to run from a
    workqueue.  From Hannes Frederic Sowa.

 9) Fix route refcount leaking in ip tunnels, from Pravin B Shelar.

10) Don't return -EINTR from non-blocking recvmsg() on AF_UNIX sockets,
    from Eric Dumazet.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (28 commits)
  vlan: Warn the user if lowerdev has bad vlan features.
  veth: Turn off vlan rx acceleration in vlan_features
  ifb: Remove vlan acceleration from vlan_features
  qlge: Do not propaged vlan tag offloads to vlans
  bridge: Fix crash with vlan filtering and tcpdump
  net: Account for all vlan headers in skb_mac_gso_segment
  MAINTAINERS: bonding: change email address
  MAINTAINERS: bonding: change email address
  ipv6: move DAD and addrconf_verify processing to workqueue
  tcp: fix get_timewait4_sock() delay computation on 64bit
  openvswitch: fix a possible deadlock and lockdep warning
  bridge: Fix handling stacked vlan tags
  bridge: Fix inabillity to retrieve vlan tags when tx offload is disabled
  vhost: validate vhost_get_vq_desc return value
  vhost: fix total length when packets are too short
  random32: avoid attempt to late reseed if in the middle of seeding
  random32: assign to network folks in MAINTAINERS
  net/mlx4_core: pass pci_device_id.driver_data to __mlx4_init_one during reset
  core, nfqueue, openvswitch: Orphan frags in skb_zerocopy and handle errors
  vlan: Set hard_header_len according to available acceleration
  ...

10 years agoMerge branch 'vlan_offloads'
David S. Miller [Fri, 28 Mar 2014 21:17:16 +0000 (17:17 -0400)]
Merge branch 'vlan_offloads'

Vlad Yasevich says:

====================
Audit all drivers for correct vlan_features.

Some drivers set vlan acceleration features in vlan_features.  This causes
issues with Q-in-Q/802.1ad configurations.

Audit all the drivers for correct vlan_features.  Fix broken ones.
Add a warning to vlan code to help catch future offenders.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agovlan: Warn the user if lowerdev has bad vlan features.
Vlad Yasevich [Fri, 28 Mar 2014 02:14:49 +0000 (22:14 -0400)]
vlan: Warn the user if lowerdev has bad vlan features.

Some drivers incorrectly assign vlan acceleration features to
vlan_features thus causing issues for Q-in-Q vlan configurations.
Warn the user of such cases.

Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoveth: Turn off vlan rx acceleration in vlan_features
Vlad Yasevich [Fri, 28 Mar 2014 02:14:48 +0000 (22:14 -0400)]
veth: Turn off vlan rx acceleration in vlan_features

For completeness, turn off vlan rx acceleration in vlan_features so
that it doesn't show up on q-in-q setups.

Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoifb: Remove vlan acceleration from vlan_features
Vlad Yasevich [Fri, 28 Mar 2014 02:14:47 +0000 (22:14 -0400)]
ifb: Remove vlan acceleration from vlan_features

Do not include vlan acceleration features in vlan_features as that
precludes correct Q-in-Q operation.

Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlge: Do not propaged vlan tag offloads to vlans
Vlad Yasevich [Fri, 28 Mar 2014 02:14:46 +0000 (22:14 -0400)]
qlge: Do not propaged vlan tag offloads to vlans

qlge driver turns off NETIF_F_HW_CTAG_FILTER, but forgets to
turn off HW_CTAG_TX and HW_CTAG_RX on vlan devices.  With the
current settings, q-in-q will only generate a single vlan header.
Remember to mask off CTAG_TX and CTAG_RX features in vlan_features.

CC: Shahed Shaikh <shahed.shaikh@qlogic.com>
CC: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
CC: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
Acked-by: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: Fix crash with vlan filtering and tcpdump
Vlad Yasevich [Fri, 28 Mar 2014 01:51:18 +0000 (21:51 -0400)]
bridge: Fix crash with vlan filtering and tcpdump

When the vlan filtering is enabled on the bridge, but
the filter is not configured on the bridge device itself,
running tcpdump on the bridge device will result in a
an Oops with NULL pointer dereference.  The reason
is that br_pass_frame_up() will bypass the vlan
check because promisc flag is set.  It will then try
to get the table pointer and process the packet based
on the table.  Since the table pointer is NULL, we oops.
Catch this special condition in br_handle_vlan().

Reported-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
CC: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
Acked-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Account for all vlan headers in skb_mac_gso_segment
Vlad Yasevich [Thu, 27 Mar 2014 21:26:18 +0000 (17:26 -0400)]
net: Account for all vlan headers in skb_mac_gso_segment

skb_network_protocol() already accounts for multiple vlan
headers that may be present in the skb.  However, skb_mac_gso_segment()
doesn't know anything about it and assumes that skb->mac_len
is set correctly to skip all mac headers.  That may not
always be the case.  If we are simply forwarding the packet (via
bridge or macvtap), all vlan headers may not be accounted for.

A simple solution is to allow skb_network_protocol to return
the vlan depth it has calculated.  This way skb_mac_gso_segment
will correctly skip all mac headers.

Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMAINTAINERS: bonding: change email address
Veaceslav Falico [Thu, 27 Mar 2014 17:43:50 +0000 (18:43 +0100)]
MAINTAINERS: bonding: change email address

Signed-off-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMAINTAINERS: bonding: change email address
Jay Vosburgh [Thu, 27 Mar 2014 17:33:44 +0000 (10:33 -0700)]
MAINTAINERS: bonding: change email address

Update my email address.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'akpm' (patches from Andrew Morton)
Linus Torvalds [Fri, 28 Mar 2014 20:57:13 +0000 (13:57 -0700)]
Merge branch 'akpm' (patches from Andrew Morton)

Merge two fixes from Andrew Morton:
 "The x86 fix should come from x86 guys but they appear to be
  conferencing or otherwise distracted.

  The ocfs2 fix is a bit of a mess - the code runs into an immediate
  NULL deref and we're trying to work out how this got through test and
  review, but we haven't heard from Goldwyn in the past few days.
  Sasha's patch fixes the oops, but the feature as a whole is probably
  broken.  So this is a stopgap for 3.14 - I'll aim to get the real
  fixes into 3.14.x"

* emailed patches from Andrew Morton akpm@linux-foundation.org>:
  x86: fix boot on uniprocessor systems
  ocfs2: check if cluster name exists before deref

10 years agox86: fix boot on uniprocessor systems
Artem Fetishev [Fri, 28 Mar 2014 20:33:39 +0000 (13:33 -0700)]
x86: fix boot on uniprocessor systems

On x86 uniprocessor systems topology_physical_package_id() returns -1
which causes rapl_cpu_prepare() to leave rapl_pmu variable uninitialized
which leads to GPF in rapl_pmu_init().

See arch/x86/kernel/cpu/perf_event_intel_rapl.c.

It turns out that physical_package_id and core_id can actually be
retreived for uniprocessor systems too.  Enabling them also fixes
rapl_pmu code.

Signed-off-by: Artem Fetishev <artem_fetishev@epam.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoocfs2: check if cluster name exists before deref
Sasha Levin [Fri, 28 Mar 2014 20:33:38 +0000 (13:33 -0700)]
ocfs2: check if cluster name exists before deref

Commit c74a3bdd9b52 ("ocfs2: add clustername to cluster connection") is
trying to strlcpy a string which was explicitly passed as NULL in the
very same patch, triggering a NULL ptr deref.

  BUG: unable to handle kernel NULL pointer dereference at           (null)
  IP: strlcpy (lib/string.c:388 lib/string.c:151)
  CPU: 19 PID: 19426 Comm: trinity-c19 Tainted: G        W     3.14.0-rc7-next-20140325-sasha-00014-g9476368-dirty #274
  RIP:  strlcpy (lib/string.c:388 lib/string.c:151)
  Call Trace:
   ocfs2_cluster_connect (fs/ocfs2/stackglue.c:350)
   ocfs2_cluster_connect_agnostic (fs/ocfs2/stackglue.c:396)
   user_dlm_register (fs/ocfs2/dlmfs/userdlm.c:679)
   dlmfs_mkdir (fs/ocfs2/dlmfs/dlmfs.c:503)
   vfs_mkdir (fs/namei.c:3467)
   SyS_mkdirat (fs/namei.c:3488 fs/namei.c:3472)
   tracesys (arch/x86/kernel/entry_64.S:749)

akpm: this patch probably disables the feature.  A temporary thing to
avoid triviel oopses.

Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoipv6: move DAD and addrconf_verify processing to workqueue
Hannes Frederic Sowa [Thu, 27 Mar 2014 17:28:07 +0000 (18:28 +0100)]
ipv6: move DAD and addrconf_verify processing to workqueue

addrconf_join_solict and addrconf_join_anycast may cause actions which
need rtnl locked, especially on first address creation.

A new DAD state is introduced which defers processing of the initial
DAD processing into a workqueue.

To get rtnl lock we need to push the code paths which depend on those
calls up to workqueues, specifically addrconf_verify and the DAD
processing.

(v2)
addrconf_dad_failure needs to be queued up to the workqueue, too. This
patch introduces a new DAD state and stop the DAD processing in the
workqueue (this is because of the possible ipv6_del_addr processing
which removes the solicited multicast address from the device).

addrconf_verify_lock is removed, too. After the transition it is not
needed any more.

As we are not processing in bottom half anymore we need to be a bit more
careful about disabling bottom half out when we lock spin_locks which are also
used in bh.

Relevant backtrace:
[  541.030090] RTNL: assertion failed at net/core/dev.c (4496)
[  541.031143] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G           O 3.10.33-1-amd64-vyatta #1
[  541.031145] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007
[  541.031146]  ffffffff8148a9f0 000000000000002f ffffffff813c98c1 ffff88007c4451f8
[  541.031148]  0000000000000000 0000000000000000 ffffffff813d3540 ffff88007fc03d18
[  541.031150]  0000880000000006 ffff88007c445000 ffffffffa0194160 0000000000000000
[  541.031152] Call Trace:
[  541.031153]  <IRQ>  [<ffffffff8148a9f0>] ? dump_stack+0xd/0x17
[  541.031180]  [<ffffffff813c98c1>] ? __dev_set_promiscuity+0x101/0x180
[  541.031183]  [<ffffffff813d3540>] ? __hw_addr_create_ex+0x60/0xc0
[  541.031185]  [<ffffffff813cfe1a>] ? __dev_set_rx_mode+0xaa/0xc0
[  541.031189]  [<ffffffff813d3a81>] ? __dev_mc_add+0x61/0x90
[  541.031198]  [<ffffffffa01dcf9c>] ? igmp6_group_added+0xfc/0x1a0 [ipv6]
[  541.031208]  [<ffffffff8111237b>] ? kmem_cache_alloc+0xcb/0xd0
[  541.031212]  [<ffffffffa01ddcd7>] ? ipv6_dev_mc_inc+0x267/0x300 [ipv6]
[  541.031216]  [<ffffffffa01c2fae>] ? addrconf_join_solict+0x2e/0x40 [ipv6]
[  541.031219]  [<ffffffffa01ba2e9>] ? ipv6_dev_ac_inc+0x159/0x1f0 [ipv6]
[  541.031223]  [<ffffffffa01c0772>] ? addrconf_join_anycast+0x92/0xa0 [ipv6]
[  541.031226]  [<ffffffffa01c311e>] ? __ipv6_ifa_notify+0x11e/0x1e0 [ipv6]
[  541.031229]  [<ffffffffa01c3213>] ? ipv6_ifa_notify+0x33/0x50 [ipv6]
[  541.031233]  [<ffffffffa01c36c8>] ? addrconf_dad_completed+0x28/0x100 [ipv6]
[  541.031241]  [<ffffffff81075c1d>] ? task_cputime+0x2d/0x50
[  541.031244]  [<ffffffffa01c38d6>] ? addrconf_dad_timer+0x136/0x150 [ipv6]
[  541.031247]  [<ffffffffa01c37a0>] ? addrconf_dad_completed+0x100/0x100 [ipv6]
[  541.031255]  [<ffffffff8105313a>] ? call_timer_fn.isra.22+0x2a/0x90
[  541.031258]  [<ffffffffa01c37a0>] ? addrconf_dad_completed+0x100/0x100 [ipv6]

Hunks and backtrace stolen from a patch by Stephen Hemminger.

Reported-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: net: add a core netdev->tx_dropped counter
Eric Dumazet [Thu, 27 Mar 2014 15:45:56 +0000 (08:45 -0700)]
net: net: add a core netdev->tx_dropped counter

Dropping packets in __dev_queue_xmit() when transmit queue
is stopped (NIC TX ring buffer full or BQL limit reached) currently
outputs a syslog message.

It would be better to get a precise count of such events available in
netdevice stats so that monitoring tools can have a clue.

This extends the work done in caf586e5f23ce
("net: add a core netdev->rx_dropped counter")

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agopacket: respect devices with LLTX flag in direct xmit
Daniel Borkmann [Thu, 27 Mar 2014 15:38:30 +0000 (16:38 +0100)]
packet: respect devices with LLTX flag in direct xmit

Quite often it can be useful to test with dummy or similar
devices as a blackhole sink for skbs. Such devices are only
equipped with a single txq, but marked as NETIF_F_LLTX as
they do not require locking their internal queues on xmit
(or implement locking themselves). Therefore, rather use
HARD_TX_{UN,}LOCK API, so that NETIF_F_LLTX will be respected.

trafgen mmap/TX_RING example against dummy device with config
foo: { fill(0xff, 64) } results in the following performance
improvements for such scenarios on an ordinary Core i7/2.80GHz:

Before:

 Performance counter stats for 'trafgen -i foo -o du0 -n100000000' (10 runs):

   160,975,944,159 instructions:k            #    0.55  insns per cycle          ( +-  0.09% )
   293,319,390,278 cycles:k                  #    0.000 GHz                      ( +-  0.35% )
       192,501,104 branch-misses:k                                               ( +-  1.63% )
               831 context-switches:k                                            ( +-  9.18% )
                 7 cpu-migrations:k                                              ( +-  7.40% )
            69,382 cache-misses:k            #    0.010 % of all cache refs      ( +-  2.18% )
       671,552,021 cache-references:k                                            ( +-  1.29% )

      22.856401569 seconds time elapsed                                          ( +-  0.33% )

After:

 Performance counter stats for 'trafgen -i foo -o du0 -n100000000' (10 runs):

   133,788,739,692 instructions:k            #    0.92  insns per cycle          ( +-  0.06% )
   145,853,213,256 cycles:k                  #    0.000 GHz                      ( +-  0.17% )
        59,867,100 branch-misses:k                                               ( +-  4.72% )
               384 context-switches:k                                            ( +-  3.76% )
                 6 cpu-migrations:k                                              ( +-  6.28% )
            70,304 cache-misses:k            #    0.077 % of all cache refs      ( +-  1.73% )
        90,879,408 cache-references:k                                            ( +-  1.35% )

      11.719372413 seconds time elapsed                                          ( +-  0.24% )

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: nlmon: flag nlmon devs with LLTX/SG
Daniel Borkmann [Thu, 27 Mar 2014 15:34:59 +0000 (16:34 +0100)]
net: nlmon: flag nlmon devs with LLTX/SG

As in xmit path we merely update statistics and free the skb, we
can mark the device with LLTX feature, so that upper layers can
avoid taking the single txq lock on xmit. While at it, also add
missing NETIF_F_SG.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: fix get_timewait4_sock() delay computation on 64bit
Eric Dumazet [Thu, 27 Mar 2014 14:19:19 +0000 (07:19 -0700)]
tcp: fix get_timewait4_sock() delay computation on 64bit

It seems I missed one change in get_timewait4_sock() to compute
the remaining time before deletion of IPV4 timewait socket.

This could result in wrong output in /proc/net/tcp for tm->when field.

Fixes: 96f817fedec4 ("tcp: shrink tcp6_timewait_sock by one cache line")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoopenvswitch: fix a possible deadlock and lockdep warning
Flavio Leitner [Thu, 27 Mar 2014 14:05:34 +0000 (11:05 -0300)]
openvswitch: fix a possible deadlock and lockdep warning

There are two problematic situations.

A deadlock can happen when is_percpu is false because it can get
interrupted while holding the spinlock. Then it executes
ovs_flow_stats_update() in softirq context which tries to get
the same lock.

The second sitation is that when is_percpu is true, the code
correctly disables BH but only for the local CPU, so the
following can happen when locking the remote CPU without
disabling BH:

       CPU#0                            CPU#1
  ovs_flow_stats_get()
   stats_read()
 +->spin_lock remote CPU#1        ovs_flow_stats_get()
 |  <interrupted>                  stats_read()
 |  ...                       +-->  spin_lock remote CPU#0
 |                            |     <interrupted>
 |  ovs_flow_stats_update()   |     ...
 |   spin_lock local CPU#0 <--+     ovs_flow_stats_update()
 +---------------------------------- spin_lock local CPU#1

This patch disables BH for both cases fixing the deadlocks.
Acked-by: Jesse Gross <jesse@nicira.com>
=================================
[ INFO: inconsistent lock state ]
3.14.0-rc8-00007-g632b06a #1 Tainted: G          I
---------------------------------
inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage.
swapper/0/0 [HC0[0]:SC1[5]:HE1:SE0] takes:
(&(&cpu_stats->lock)->rlock){+.?...}, at: [<ffffffffa05dd8a1>] ovs_flow_stats_update+0x51/0xd0 [openvswitch]
{SOFTIRQ-ON-W} state was registered at:
[<ffffffff810f973f>] __lock_acquire+0x68f/0x1c40
[<ffffffff810fb4e2>] lock_acquire+0xa2/0x1d0
[<ffffffff817d8d9e>] _raw_spin_lock+0x3e/0x80
[<ffffffffa05dd9e4>] ovs_flow_stats_get+0xc4/0x1e0 [openvswitch]
[<ffffffffa05da855>] ovs_flow_cmd_fill_info+0x185/0x360 [openvswitch]
[<ffffffffa05daf05>] ovs_flow_cmd_build_info.constprop.27+0x55/0x90 [openvswitch]
[<ffffffffa05db41d>] ovs_flow_cmd_new_or_set+0x4dd/0x570 [openvswitch]
[<ffffffff816c245d>] genl_family_rcv_msg+0x1cd/0x3f0
[<ffffffff816c270e>] genl_rcv_msg+0x8e/0xd0
[<ffffffff816c0239>] netlink_rcv_skb+0xa9/0xc0
[<ffffffff816c0798>] genl_rcv+0x28/0x40
[<ffffffff816bf830>] netlink_unicast+0x100/0x1e0
[<ffffffff816bfc57>] netlink_sendmsg+0x347/0x770
[<ffffffff81668e9c>] sock_sendmsg+0x9c/0xe0
[<ffffffff816692d9>] ___sys_sendmsg+0x3a9/0x3c0
[<ffffffff8166a911>] __sys_sendmsg+0x51/0x90
[<ffffffff8166a962>] SyS_sendmsg+0x12/0x20
[<ffffffff817e3ce9>] system_call_fastpath+0x16/0x1b
irq event stamp: 1740726
hardirqs last  enabled at (1740726): [<ffffffff8175d5e0>] ip6_finish_output2+0x4f0/0x840
hardirqs last disabled at (1740725): [<ffffffff8175d59b>] ip6_finish_output2+0x4ab/0x840
softirqs last  enabled at (1740674): [<ffffffff8109be12>] _local_bh_enable+0x22/0x50
softirqs last disabled at (1740675): [<ffffffff8109db05>] irq_exit+0xc5/0xd0

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(&(&cpu_stats->lock)->rlock);
  <Interrupt>
    lock(&(&cpu_stats->lock)->rlock);

 *** DEADLOCK ***

5 locks held by swapper/0/0:
 #0:  (((&ifa->dad_timer))){+.-...}, at: [<ffffffff810a7155>] call_timer_fn+0x5/0x320
 #1:  (rcu_read_lock){.+.+..}, at: [<ffffffff81788a55>] mld_sendpack+0x5/0x4a0
 #2:  (rcu_read_lock_bh){.+....}, at: [<ffffffff8175d149>] ip6_finish_output2+0x59/0x840
 #3:  (rcu_read_lock_bh){.+....}, at: [<ffffffff8168ba75>] __dev_queue_xmit+0x5/0x9b0
 #4:  (rcu_read_lock){.+.+..}, at: [<ffffffffa05e41b5>] internal_dev_xmit+0x5/0x110 [openvswitch]

stack backtrace:
CPU: 0 PID: 0 Comm: swapper/0 Tainted: G          I  3.14.0-rc8-00007-g632b06a #1
Hardware name:                  /DX58SO, BIOS SOX5810J.86A.5599.2012.0529.2218 05/29/2012
 0000000000000000 0fcf20709903df0c ffff88042d603808 ffffffff817cfe3c
 ffffffff81c134c0 ffff88042d603858 ffffffff817cb6da 0000000000000005
 ffffffff00000001 ffff880400000000 0000000000000006 ffffffff81c134c0
Call Trace:
 <IRQ>  [<ffffffff817cfe3c>] dump_stack+0x4d/0x66
 [<ffffffff817cb6da>] print_usage_bug+0x1f4/0x205
 [<ffffffff810f7f10>] ? check_usage_backwards+0x180/0x180
 [<ffffffff810f8963>] mark_lock+0x223/0x2b0
 [<ffffffff810f96d3>] __lock_acquire+0x623/0x1c40
 [<ffffffff810f5707>] ? __lock_is_held+0x57/0x80
 [<ffffffffa05e26c6>] ? masked_flow_lookup+0x236/0x250 [openvswitch]
 [<ffffffff810fb4e2>] lock_acquire+0xa2/0x1d0
 [<ffffffffa05dd8a1>] ? ovs_flow_stats_update+0x51/0xd0 [openvswitch]
 [<ffffffff817d8d9e>] _raw_spin_lock+0x3e/0x80
 [<ffffffffa05dd8a1>] ? ovs_flow_stats_update+0x51/0xd0 [openvswitch]
 [<ffffffffa05dd8a1>] ovs_flow_stats_update+0x51/0xd0 [openvswitch]
 [<ffffffffa05dcc64>] ovs_dp_process_received_packet+0x84/0x120 [openvswitch]
 [<ffffffff810f93f7>] ? __lock_acquire+0x347/0x1c40
 [<ffffffffa05e3bea>] ovs_vport_receive+0x2a/0x30 [openvswitch]
 [<ffffffffa05e4218>] internal_dev_xmit+0x68/0x110 [openvswitch]
 [<ffffffffa05e41b5>] ? internal_dev_xmit+0x5/0x110 [openvswitch]
 [<ffffffff8168b4a6>] dev_hard_start_xmit+0x2e6/0x8b0
 [<ffffffff8168be87>] __dev_queue_xmit+0x417/0x9b0
 [<ffffffff8168ba75>] ? __dev_queue_xmit+0x5/0x9b0
 [<ffffffff8175d5e0>] ? ip6_finish_output2+0x4f0/0x840
 [<ffffffff8168c430>] dev_queue_xmit+0x10/0x20
 [<ffffffff8175d641>] ip6_finish_output2+0x551/0x840
 [<ffffffff8176128a>] ? ip6_finish_output+0x9a/0x220
 [<ffffffff8176128a>] ip6_finish_output+0x9a/0x220
 [<ffffffff8176145f>] ip6_output+0x4f/0x1f0
 [<ffffffff81788c29>] mld_sendpack+0x1d9/0x4a0
 [<ffffffff817895b8>] mld_send_initial_cr.part.32+0x88/0xa0
 [<ffffffff817691b0>] ? addrconf_dad_completed+0x220/0x220
 [<ffffffff8178e301>] ipv6_mc_dad_complete+0x31/0x50
 [<ffffffff817690d7>] addrconf_dad_completed+0x147/0x220
 [<ffffffff817691b0>] ? addrconf_dad_completed+0x220/0x220
 [<ffffffff8176934f>] addrconf_dad_timer+0x19f/0x1c0
 [<ffffffff810a71e9>] call_timer_fn+0x99/0x320
 [<ffffffff810a7155>] ? call_timer_fn+0x5/0x320
 [<ffffffff817691b0>] ? addrconf_dad_completed+0x220/0x220
 [<ffffffff810a76c4>] run_timer_softirq+0x254/0x3b0
 [<ffffffff8109d47d>] __do_softirq+0x12d/0x480

Signed-off-by: Flavio Leitner <fbl@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocxgb4vf: Adds device Id for few more Chelsio adapters
Hariprasad Shenai [Thu, 27 Mar 2014 12:47:10 +0000 (18:17 +0530)]
cxgb4vf: Adds device Id for few more Chelsio adapters

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocxgb4: Adds device ID for few more Chelsio Adapters
Hariprasad Shenai [Thu, 27 Mar 2014 12:47:09 +0000 (18:17 +0530)]
cxgb4: Adds device ID for few more Chelsio Adapters

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: Fix handling stacked vlan tags
Toshiaki Makita [Thu, 27 Mar 2014 12:46:56 +0000 (21:46 +0900)]
bridge: Fix handling stacked vlan tags

If a bridge with vlan_filtering enabled receives frames with stacked
vlan tags, i.e., they have two vlan tags, br_vlan_untag() strips not
only the outer tag but also the inner tag.

br_vlan_untag() is called only from br_handle_vlan(), and in this case,
it is enough to set skb->vlan_tci to 0 here, because vlan_tci has already
been set before calling br_handle_vlan().

Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Vlad Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: Fix inabillity to retrieve vlan tags when tx offload is disabled
Toshiaki Makita [Thu, 27 Mar 2014 12:46:55 +0000 (21:46 +0900)]
bridge: Fix inabillity to retrieve vlan tags when tx offload is disabled

Bridge vlan code (br_vlan_get_tag()) assumes that all frames have vlan_tci
if they are tagged, but if vlan tx offload is manually disabled on bridge
device and frames are sent from vlan device on the bridge device, the tags
are embedded in skb->data and they break this assumption.
Extract embedded vlan tags and move them to vlan_tci at ingress.

Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Vlad Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'mlx4_vxlan'
David S. Miller [Fri, 28 Mar 2014 20:30:05 +0000 (16:30 -0400)]
Merge branch 'mlx4_vxlan'

Or Gerlitz says:

====================
Implement vxlan ndo calls

This short series adds support for the vxlan ndo calls, the udp
port is programmed to the firmware using a new command we introduce
here which is called "config device".
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4: Implement vxlan ndo calls
Or Gerlitz [Thu, 27 Mar 2014 12:02:04 +0000 (14:02 +0200)]
net/mlx4: Implement vxlan ndo calls

Add implementation for the add/del vxlan port ndo calls, using the
CONFIG_DEV firmware command.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agomlx4: Add support for CONFIG_DEV command
Or Gerlitz [Thu, 27 Mar 2014 12:02:03 +0000 (14:02 +0200)]
mlx4: Add support for CONFIG_DEV command

Introduce the CONFIG_DEV firmware command which we will use to
configure the UDP port assumed by the firmware for the VXLAN offloads.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4: USe one wrapper that returns -EPERM
Or Gerlitz [Thu, 27 Mar 2014 12:02:02 +0000 (14:02 +0200)]
net/mlx4: USe one wrapper that returns -EPERM

When a VF issues a firmware command which is disallowed for them, the PF
rerturns -EPERM from that command wrapper. Move to use one such wrapper
instance, instead of repeating the same code on such commands.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobnx2x: Fix possible memory leak on iov error flow
Yuval Mintz [Thu, 27 Mar 2014 11:46:37 +0000 (13:46 +0200)]
bnx2x: Fix possible memory leak on iov error flow

Commit 2dc33bbc4 "bnx2x: Remove the sriov VFOP mechanism" introduced a possible
memory leak on the error flow during multicast filters configuration.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: Ariel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agovhost: validate vhost_get_vq_desc return value
Michael S. Tsirkin [Thu, 27 Mar 2014 10:53:37 +0000 (12:53 +0200)]
vhost: validate vhost_get_vq_desc return value

vhost fails to validate negative error code
from vhost_get_vq_desc causing
a crash: we are using -EFAULT which is 0xfffffff2
as vector size, which exceeds the allocated size.

The code in question was introduced in commit
8dd014adfea6f173c1ef6378f7e5e7924866c923
    vhost-net: mergeable buffers support

CVE-2014-0055

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agovhost: fix total length when packets are too short
Michael S. Tsirkin [Thu, 27 Mar 2014 10:00:26 +0000 (12:00 +0200)]
vhost: fix total length when packets are too short

When mergeable buffers are disabled, and the
incoming packet is too large for the rx buffer,
get_rx_bufs returns success.

This was intentional in order for make recvmsg
truncate the packet and then handle_rx would
detect err != sock_len and drop it.

Unfortunately we pass the original sock_len to
recvmsg - which means we use parts of iov not fully
validated.

Fix this up by detecting this overrun and doing packet drop
immediately.

CVE-2014-0077

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agorandom32: avoid attempt to late reseed if in the middle of seeding
Sasha Levin [Fri, 28 Mar 2014 16:38:42 +0000 (17:38 +0100)]
random32: avoid attempt to late reseed if in the middle of seeding

Commit 4af712e8df ("random32: add prandom_reseed_late() and call when
nonblocking pool becomes initialized") has added a late reseed stage
that happens as soon as the nonblocking pool is marked as initialized.

This fails in the case that the nonblocking pool gets initialized
during __prandom_reseed()'s call to get_random_bytes(). In that case
we'd double back into __prandom_reseed() in an attempt to do a late
reseed - deadlocking on 'lock' early on in the boot process.

Instead, just avoid even waiting to do a reseed if a reseed is already
occuring.

Fixes: 4af712e8df99 ("random32: add prandom_reseed_late() and call when nonblocking pool becomes initialized")
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Fri, 28 Mar 2014 20:03:00 +0000 (13:03 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input

Pull input subsystem fixes from Dmitry Torokhov:
 "Updates to Synaptics touchpad to better cope with devices in Lenovo
  laptops, and a couple more fixes"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: synaptics - add manual min/max quirk for ThinkPad X240
  Input: synaptics - add manual min/max quirk
  Input: cypress_ps2 - don't report as a button pads
  Input: da9052_onkey - use correct register bit for key status
  Input: adp5588-keys - get value from data out when dir is out

10 years agorandom32: assign to network folks in MAINTAINERS
Sasha Levin [Thu, 27 Mar 2014 06:01:34 +0000 (02:01 -0400)]
random32: assign to network folks in MAINTAINERS

lib/random32.c was split out of the network code and is de-facto
still maintained by the almighty net/ gods.

Make it a bit more official so that people who aren't aware of
that know where to send their patches.

Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agosh_eth: ensure pm_runtime cannot suspend the device during init
Ben Dooks [Fri, 21 Mar 2014 11:09:14 +0000 (12:09 +0100)]
sh_eth: ensure pm_runtime cannot suspend the device during init

The pm_rumtime work queue is causing the device to be suspended during
initialisation, thus the initialisation may not be able to access registers
properly. As the code is called from a work queue, it is possible that this
is not seen from certain configurations/builds due to the asynchronos
nature of the code.

Another issue has also been found where the network device registration
calls back into the driver thus causing further pm_runtime calls that
also caused issues with the MDIO bus code. This has now been checked
and is the only place the MDIO can be called without the device open.

Use pm_runtime_get_sync() and pm_runtime_put() to ensure that the
pm system does not suspend it during the probe() call and remove the
now unnecessary pm_runtime_resume() call. Also add a call in the error
path to call pm_runtime_disable().

This fixes the external abort that can cause /sbin/init or other such
init processed to die.

Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: stmmac: Add SOCFPGA glue driver
Dinh Nguyen [Thu, 27 Mar 2014 03:45:10 +0000 (22:45 -0500)]
net: stmmac: Add SOCFPGA glue driver

Like the STi and sunxi series SOCs, Altera's SOCFPGA also needs a glue layer
on top of the Synopsys gmac IP.

This patch adds the platform driver for the glue layer which configures the IP
before the generic STMMAC driver takes over.

Signed-off-by: Dinh Nguyen <dinguyen@altera.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoRDMA/cxgb4: set error code on kmalloc() failure
Yann Droneaud [Fri, 28 Mar 2014 18:55:21 +0000 (14:55 -0400)]
RDMA/cxgb4: set error code on kmalloc() failure

If kmalloc() fails in c4iw_alloc_ucontext(), the function
leaves but does not set an error code in ret variable:
it will return 0 to the caller.

This patch set ret to -ENOMEM in such case.

Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Steve Wise <swise@chelsio.com>
Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'tipc-next'
David S. Miller [Fri, 28 Mar 2014 18:46:34 +0000 (14:46 -0400)]
Merge branch 'tipc-next'

Erik Hugne says:

====================
tipc: fix handling of NETDEV_CHANGEADDR event

Aside from manual reconfiguration of the netdevice hwaddr, this can also
be changed automatically for an interface bond in active-backup mode
if fail_over_mac is enabled. This patchset fixes the handling of this
event in TIPC by properly updating the l2 media address for the bearer,
followed by a reinitialization of the node discovery mechanism.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: make discovery domain a bearer attribute
Erik Hugne [Fri, 28 Mar 2014 09:32:09 +0000 (10:32 +0100)]
tipc: make discovery domain a bearer attribute

The node discovery domain is assigned when a bearer is enabled.
In the previous commit we reflect this attribute directly in the
bearer structure since it's needed to reinitialize the node
discovery mechanism after a hardware address change.

There's no need to replicate this attribute anywhere else, so we
remove it from the tipc_link_req structure.

Signed-off-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: fix neighbor detection problem after hw address change
Erik Hugne [Fri, 28 Mar 2014 09:32:08 +0000 (10:32 +0100)]
tipc: fix neighbor detection problem after hw address change

If the hardware address of a underlying netdevice is changed, it is
not enough to simply reset the bearer/links over this device. We
also need to reflect this change in the TIPC bearer and node
discovery structures aswell.

This patch adds the necessary reinitialization of the node disovery
mechanism following a hardware address change so that the correct
originating media address is advertised in the discovery messages.

Signed-off-by: Erik Hugne <erik.hugne@ericsson.com>
Reported-by: Dong Liu <dliu.cn@gmail.com>
Reviewed-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Fri, 28 Mar 2014 18:44:06 +0000 (14:44 -0400)]
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates

This series contains updates to e1000e, igb, i40e and i40evf

Anjali provides i40e fix to remove the ATR filter on RST as well as FIN
packets.  Cleans up add_del_fdir() because it was used and implemented
only for the add, so change the name and drop a parameter.  Adds the
ability to drop a flow if we wanted to and adds a flow director
message level to be used for flow director specific messages.

Mitch fixes an issue on i40evf where the Tx watchdog handler was causing
an oops when sending an admin queue message to request a reset because
the admin queue functions use spinlocks.

Greg provides a change to i40e to make the alloc and free queue vector
calls orthogonal.

Shannon fixes i40e to verify the eeprom checksum and firmware CRC status
bits, and shutdown the driver if they fail.  This change stops the
processing of traffic, but does not kill the PF netdev so that the
NVMUpdate process still has a chance at fixing the image.  Also provides
a fix to make sure the VSI has a netdev before trying to use it in
the debugfs netdev_ops commands.

Jakub Kicinski provides patches for e1000e and igb to fix a number issues
found in the PTP code.

v2:
- drop patch 11 "i40e: Add a fallback debug flow for the driver" from the
  series based on feedback from David Miller
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sxgbe: fix potential null dereference
Byungho An [Fri, 28 Mar 2014 17:57:44 +0000 (10:57 -0700)]
net: sxgbe: fix potential null dereference

This fixes following:

drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c:1828 sxgbe_hw_init()
error: potential null dereference 'priv->hw'.  (kmalloc returns null)

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Byungho An <bh74.an@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sxgbe: fix sparse warnings about static declaration
Byungho An [Fri, 28 Mar 2014 17:57:36 +0000 (10:57 -0700)]
net: sxgbe: fix sparse warnings about static declaration

This fixes followings:

sparse warnings: (new ones prefixed by >>)

>> drivers/net/ethernet/samsung/sxgbe/sxgbe_platform.c:197:5:
sparse: symbol 'sxgbe_platform_freeze' was not declared. Should it be static?
>> drivers/net/ethernet/samsung/sxgbe/sxgbe_platform.c:204:5:
sparse: symbol 'sxgbe_platform_restore' was not declared. Should it be static?
>> drivers/net/ethernet/samsung/sxgbe/sxgbe_platform.c:228:24:
sparse: symbol 'sxgbe_platform_driver' was not declared. Should it be static?

>> drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c:1795:6:
sparse: symbol 'sxgbe_get_ops' was not declared. Should it be static?

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Byungho An <bh74.an@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'be2net-next'
David S. Miller [Fri, 28 Mar 2014 18:25:02 +0000 (14:25 -0400)]
Merge branch 'be2net-next'

Sathya Perla says:

====================
be2net: add vxlan offload support

The first patch adds the FW cmds needed to configure the Skyhawk-R
chip for supporting VxLAN offloads. The second patch implements the
ndo_add/del_vxlan_port() methods and the plumbing for supporting
RX/TX csum, TSO and RSS steering offloads for VxLAN traffic.

v2 changes:
NETIF_F_SG need not be set for hw_enc_features by the driver as it is
done by the stack.

v3 changes:
* Defer FW cmds needed for VxLAN offloads to a workqueue
* Reset FW to VxLAN offloads disabled state in the unload path

v4 changes:
* Revert the usage of workqueue (introduced in v3) to implement
ndo_add/del_vxlan_port() as it is currently not needed (none of the
FW cmd calls sleep.) Suggested by David M.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobe2net: csum, tso and rss steering offload support for VxLAN
Sathya Perla [Thu, 27 Mar 2014 05:16:19 +0000 (10:46 +0530)]
be2net: csum, tso and rss steering offload support for VxLAN

This patch mainly implements the add/del_vxlan_port() methods by invoking
the needed FW cmds for supporting VxLAN offloads for Skyhawk-R.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobe2net: add FW cmds needed for VxLAN offloads
Sathya Perla [Thu, 27 Mar 2014 05:16:18 +0000 (10:46 +0530)]
be2net: add FW cmds needed for VxLAN offloads

This patch adds support for the FW cmds needed for VxLAN offloads
on Skyhawk-R:
1) The VxLAN UDP port needs to be configured via the port-desc of
   SET_PROFILE_CONFIG_v1 cmd.
   This patch re-factors the be_set_profile_config() code (used so far
   only for setting VF QoS) to be used to set any type of descriptor.
2) The MANAGE_IFACE_FILTERS cmds is needed to convert a normal interface
   into a tunnel interface. This allows for RSS to work even on the inner
   TCP/UDP headers of VxLAN traffic.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 28 Mar 2014 17:58:10 +0000 (10:58 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "I didn't want these to wait for stable cycle.

  The nouveau and radeon ones are the same problem, where the runtime pm
  stuff broke non-runtime pm managed secondary GPUs.

  The udl fix is for an oops on unplug, and the i915 fix is for a
  regression on Sandybridge even though it may break haswell (regression
  wins)"

Daniel Vetter comments:
 "My apologies for the i915 regression fumble, that thing somehow fell
  through the cracks here for almost half a year :( Imo that's more than
  enough flailing to just go ahead with the revert, and the re-broken
  hsw should get peoples attention ..."

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/i915: Undo gtt scratch pte unmapping again
  drm/radeon: fix runtime suspend breaking secondary GPUs
  drm/nouveau: fail runtime pm properly.
  drm/udl: take reference to device struct for dma-bufs

10 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Fri, 28 Mar 2014 17:55:44 +0000 (10:55 -0700)]
Merge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux

Pull i2c build fix from Wolfram Sang:
 "The build fix from my last request unveiled another build problem
  which is fixed with this patch"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: cpm: Fix build by adding of_address.h and of_irq.h

10 years agoMerge tag 'stable/for-linus-3.14-rc8-tag' of git://git.kernel.org/pub/scm/linux/kerne...
Linus Torvalds [Fri, 28 Mar 2014 17:52:05 +0000 (10:52 -0700)]
Merge tag 'stable/for-linus-3.14-rc8-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip

Pull Xen bugfixes from David Vrabel:
 "Fix two bugs that cause x86 PV guest crashes.

  1. Ballooning a 32-bit guest would eventually crash it.

  2. Revert a broken fix for a regression with NUMA_BALACING.  The bad
     fix caused PV guests to crash after migration.  This is not ideal
     but unpicking the madness that is _PAGE_NUMA == _PAGE_PROTNONE will
     take a while longer"

* tag 'stable/for-linus-3.14-rc8-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  Revert "xen: properly account for _PAGE_NUMA during xen pte translations"
  xen/balloon: flush persistent kmaps in correct position

10 years agoInput: synaptics - add manual min/max quirk for ThinkPad X240
Hans de Goede [Fri, 28 Mar 2014 08:01:38 +0000 (01:01 -0700)]
Input: synaptics - add manual min/max quirk for ThinkPad X240

This extends Benjamin Tissoires manual min/max quirk table with support for
the ThinkPad X240.

Cc: stable@vger.kernel.org
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
10 years agoInput: synaptics - add manual min/max quirk
Benjamin Tissoires [Fri, 28 Mar 2014 07:43:00 +0000 (00:43 -0700)]
Input: synaptics - add manual min/max quirk

The new Lenovo Haswell series (-40's) contains a new Synaptics touchpad.
However, these new Synaptics devices report bad axis ranges.
Under Windows, it is not a problem because the Windows driver uses RMI4
over SMBus to talk to the device. Under Linux, we are using the PS/2
fallback interface and it occurs the reported ranges are wrong.

Of course, it would be too easy to have only one range for the whole
series, each touchpad seems to be calibrated in a different way.

We can not use SMBus to get the actual range because I suspect the firmware
will switch into the SMBus mode and stop talking through PS/2 (this is the
case for hybrid HID over I2C / PS/2 Synaptics touchpads).

So as a temporary solution (until RMI4 land into upstream), start a new
list of quirks with the min/max manually set.

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
CC: stable@vger.kernel.org
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
10 years agoigb: fix race conditions on queuing skb for HW time stamp
Jakub Kicinski [Sat, 15 Mar 2014 14:55:32 +0000 (14:55 +0000)]
igb: fix race conditions on queuing skb for HW time stamp

igb has a single set of TX time stamping resources per NIC.
Use a simple bit lock to avoid race conditions and leaking skbs
when multiple TX rings try to claim time stamping.

Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoigb: never generate both software and hardware timestamps
Jakub Kicinski [Sat, 15 Mar 2014 14:55:26 +0000 (14:55 +0000)]
igb: never generate both software and hardware timestamps

skb_tx_timestamp() does not report software time stamp
if SKBTX_IN_PROGRESS is set. According to timestamping.txt
software time stamps are a fallback and should not be
generated if hardware time stamp is provided.

Move call to skb_tx_timestamp() after setting
SKBTX_IN_PROGRESS.

Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: remove redundant if clause from PTP work
Jakub Kicinski [Sat, 15 Mar 2014 14:55:05 +0000 (14:55 +0000)]
e1000e: remove redundant if clause from PTP work

tx_hwtstamp_skb is always set before work is scheduled,
work is cancelled before tx_hwtstamp_skb is set to NULL.
PTP work cannot ever see tx_hwtstamp_skb set to NULL.

Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: add timeout for TX HW time stamping work
Jakub Kicinski [Sat, 15 Mar 2014 14:55:00 +0000 (14:55 +0000)]
e1000e: add timeout for TX HW time stamping work

Hardware may fail to report time stamp e.g.:
 - when hardware time stamping is not enabled
 - when time stamp is requested shortly after ifup

Timeout time stamp reading work to prevent it from
scheduling itself indefinitely. Report timeout events
via system log and device stats.

Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Use DEBUG_FD message level for an FD message
Anjali Singhai Jain [Thu, 6 Mar 2014 09:00:04 +0000 (09:00 +0000)]
i40e: Use DEBUG_FD message level for an FD message

We don't need to print this info unless at FD message level.

Change-ID: I329efdd8e754a0ea0669ec04d12e03db02e6b76e
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e/i40evf: Add an FD message level
Anjali Singhai Jain [Thu, 6 Mar 2014 09:00:03 +0000 (09:00 +0000)]
i40e/i40evf: Add an FD message level

To use for Flow Director specific messages.

Change-ID: I69e39a410aa2661f8fd1ed6af0126fa4c335cb77
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Acked-by: Kevin Scott <kevin.c.scott@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: check for netdev before debugfs use
Shannon Nelson [Thu, 6 Mar 2014 09:00:02 +0000 (09:00 +0000)]
i40e: check for netdev before debugfs use

Make sure the VSI has a netdev before trying to use it
in the debugfs netdev_ops commands.

Change-ID: I2d744fc0c32b3226534ce2cde171d9675c5440a6
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40evf: remove double space after return
Jesse Brandeburg [Thu, 6 Mar 2014 09:00:01 +0000 (09:00 +0000)]
i40evf: remove double space after return

There were two spaces between return and the value, we only need one.

Change-ID: Iaa42c33f50d8d149cdf1a4c9c1902295bfd991c4
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Kevin Scott <kevin.c.scott@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Add functionality for FD SB to drop packets
Anjali Singhai Jain [Thu, 6 Mar 2014 09:00:00 +0000 (09:00 +0000)]
i40e: Add functionality for FD SB to drop packets

With this change we can drop a flow if we wanted to.

Change-ID: I222b1ae960e61a31965bafe3159a95099e70c7d2
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Cleanup in FDIR SB ethtool code
Anjali Singhai Jain [Thu, 6 Mar 2014 08:59:59 +0000 (08:59 +0000)]
i40e: Cleanup in FDIR SB ethtool code

Function add_del_fdir was used and implemented only for add. So change the name
and drop a parameter.

Change-ID: Icf2c6c3bbd4fd00cf8d9613a3f6d8c08e0f8e288
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: eeprom integrity check on load and empr
Shannon Nelson [Thu, 6 Mar 2014 08:59:58 +0000 (08:59 +0000)]
i40e: eeprom integrity check on load and empr

The driver needs to verify the eeprom checksum and firmware crc status bits,
and shutdown the driver if they fail. This code stops the processing of traffic,
but doesn't kill the PF netdev so that the NVMUpdate process should still have a
chance at fixing the image. The eeprom is checked on driver load and after an
EMP reset, the latter of which should be generated after an NVMUpdate.

Change-ID: I34deef21d2e16bf5a43c603cf8af27e6a29dc9d2
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Make the alloc and free queue vector calls orthogonal
Greg Rose [Thu, 6 Mar 2014 08:59:57 +0000 (08:59 +0000)]
i40e: Make the alloc and free queue vector calls orthogonal

It's annoying to search for a matching alloc and free set of function calls
when they don't use the same framework for the name of the functions.  Fix
that up in the case of alloc and free of vsi queue vectors.

i40e_vsi_free_q*
i40e_vsi_alloc_q*

Change-ID: I510eb863a0fbe405312bebea55c2846c76285e6d
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40evf: fix oops in watchdog handler
Mitch Williams [Thu, 6 Mar 2014 08:59:56 +0000 (08:59 +0000)]
i40evf: fix oops in watchdog handler

The Tx watchdog handler runs in interrupt context, so it would cause an
oops when sending an admin queue message to request a reset, because the
admin queue functions use spinlocks.

Instead, set a flag and let the reset task handle sending the request.

Change-ID: I65879470b72963d9c308edfb8f45ac4fbba2c14f
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Delete ATR filter on RST
Anjali Singhai Jain [Thu, 6 Mar 2014 08:59:54 +0000 (08:59 +0000)]
i40e: Delete ATR filter on RST

We currently delete ATR filter on FIN alone, delete on RST as well.

Change-ID: Ie7cae5d1046b9d1d4a0d6ef5bdbf41224c3dade6
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoMerge branch 'mdio_reset-next'
David S. Miller [Fri, 28 Mar 2014 05:38:10 +0000 (01:38 -0400)]
Merge branch 'mdio_reset-next'

Florian Fainelli says:

====================
net: remove empty MDIO bus reset functions

This patchset removes all empty MDIO bus reset functions that we have in the
tree (hopefully all of them). This function is optional for the MDIO bus code in
drivers/net/phy/mdio_bus.c::mdiobus_register, the documentation is updated to
reflect that.

Having less call sites to update will allow us to implement a generic BMCR_RESET
based PHY reset for MDIO buses that require to reset their PHYs before their
PHY detection succeeds (e.g: HiSilicon hip04 current submission) and provide it
as a helper.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoDocumentation: networking: phy.txt: MDIO bus reset is optional
Florian Fainelli [Thu, 27 Mar 2014 01:07:26 +0000 (18:07 -0700)]
Documentation: networking: phy.txt: MDIO bus reset is optional

Update the MDIO bus documentation to mention that the MDIO bus reset
function is completely optional. It became optional with commit
e13934563db0 ("[PATCH] PHY Layer fixup")

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobfin_mac: remove empty MDIO bus reset function
Florian Fainelli [Thu, 27 Mar 2014 01:07:25 +0000 (18:07 -0700)]
bfin_mac: remove empty MDIO bus reset function

bfin_mdiobus_reset() does nothing useful and is optional for the MDIO
bus code, so let's just remove it.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor6040: remove empty MDIO bus reset function
Florian Fainelli [Thu, 27 Mar 2014 01:07:24 +0000 (18:07 -0700)]
r6040: remove empty MDIO bus reset function

r6040_mdiobus_reset() does nothing useful and is optional for the MDIO
bus code, so let's just remove it.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ftgmac100: remove empty MDIO bus reset function
Florian Fainelli [Thu, 27 Mar 2014 01:07:23 +0000 (18:07 -0700)]
net: ftgmac100: remove empty MDIO bus reset function

ftgmac100_mdiobus_reset() does nothing useful and is optional for the
MDIO bus code, so let's just remove it.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sun4i: remove empty MDIO bus reset function
Florian Fainelli [Thu, 27 Mar 2014 01:07:22 +0000 (18:07 -0700)]
net: sun4i: remove empty MDIO bus reset function

sun4i_mdio_reset() does nothing useful and is optional for the MDIO bus
code, so let's just remove it.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>