Le mercredi 09 novembre 2011 à 16:21 -0500, David Miller a écrit :
> From: David Miller <davem@davemloft.net>
> Date: Wed, 09 Nov 2011 16:16:44 -0500 (EST)
>
> > From: Eric Dumazet <eric.dumazet@gmail.com>
> > Date: Wed, 09 Nov 2011 12:14:09 +0100
> >
> >> unres_qlen is the number of frames we are able to queue per unresolved
> >> neighbour. Its default value (3) was never changed and is responsible
> >> for strange drops, especially if IP fragments are used, or multiple
> >> sessions start in parallel. Even a single tcp flow can hit this limit.
> >  ...
> >
> > Ok, I've applied this, let's see what happens :-)
>
> Early answer, build fails.
>
> Please test build this patch with DECNET enabled and resubmit.  The
> decnet neigh layer still refers to the removed ->queue_len member.
>
> Thanks.

Ouch, this was fixed on one machine yesterday, but not the other one I
used this morning, sorry.

[PATCH V5 net-next] neigh: new unresolved queue limits

unres_qlen is the number of frames we are able to queue per unresolved
neighbour. Its default value (3) was never changed and is responsible
for strange drops, especially if IP fragments are used, or multiple
sessions start in parallel. Even a single tcp flow can hit this limit.

$ arp -d 192.168.20.108 ; ping -c 2 -s 8000 192.168.20.108
PING 192.168.20.108 (192.168.20.108) 8000(8028) bytes of data.
8008 bytes from 192.168.20.108: icmp_seq=2 ttl=64 time=0.322 ms
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds support for frame fragmentation.
Signed-off-by: NAlexander Smirnov <alex.bluesman.smirnov@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Reading /proc/net/snmp6 on a machine with a lot of cpus is very
expensive (can be ~88000 us).

This is because ICMPV6MSG MIB uses 4096 bytes per cpu, and folding
values for all possible cpus can read 16 Mbytes of memory (32MBytes on
non x86 arches)

ICMP messages are not considered as fast path on a typical server, and
eventually few cpus handle them anyway. We can afford an atomic
operation instead of using percpu data.

This saves 4096 bytes per cpu and per network namespace.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Le lundi 07 novembre 2011 à 15:33 +0100, Eric Dumazet a écrit :

> At least, in recent kernels we dont change dst->refcnt in forwarding
> patch (usinf NOREF skb->dst)
>
> One particular point is the atomic_inc(dst->refcnt) we have to perform
> when queuing an UDP packet if socket asked PKTINFO stuff (for example a
> typical DNS server has to setup this option)
>
> I have one patch somewhere that stores the information in skb->cb[] and
> avoid the atomic_{inc|dec}(dst->refcnt).
>

OK I found it, I did some extra tests and believe its ready.

[PATCH net-next] ipv4: IP_PKTINFO doesnt need dst reference

When a socket uses IP_PKTINFO notifications, we currently force a dst
reference for each received skb. Reader has to access dst to get needed
information (rt_iif & rt_spec_dst) and must release dst reference.

We also forced a dst reference if skb was put in socket backlog, even
without IP_PKTINFO handling. This happens under stress/load.

We can instead store the needed information in skb->cb[], so that only
softirq handler really access dst, improving cache hit ratios.

This removes two atomic operations per packet, and false sharing as
well.

On a benchmark using a mono threaded receiver (doing only recvmsg()
calls), I can reach 720.000 pps instead of 570.000 pps.

IP_PKTINFO is typically used by DNS servers, and any multihomed aware
UDP application.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Reading /proc/net/snmp on a machine with a lot of cpus is very expensive
(can be ~88000 us).

This is because ICMPMSG MIB uses 4096 bytes per cpu, and folding values
for all possible cpus can read 16 Mbytes of memory.

ICMP messages are not considered as fast path on a typical server, and
eventually few cpus handle them anyway. We can afford an atomic
operation instead of using percpu data.

This saves 4096 bytes per cpu and per network namespace.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make clear that sk_clone() and inet_csk_clone() return a locked socket.

Add _lock() prefix and kerneldoc.
Suggested-by: NLinus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fast retransmission after changing the last address
with ASCONF negotiation
Signed-off-by: NMichio Honda <micchie@sfc.wide.ad.jp>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The documentation for how the length of attributes
is checked is wrong ("Exact length" isn't true, the
policy checks are for "minimum length") and a bit
misleading. Make it more complete and explain what
really happens.

Cc: Thomas Graf <tgraf@suug.ch>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The warning really shouldn't happen, but until we
find the reason why it does don't spew it all the
time, just once is enough to know we've hit it.
Reported-by: NLinus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

the tcp and udp code creates a set of struct file_operations at runtime
while it can also be done at compile time, with the added benefit of then
having these file operations be const.

the trickiest part was to get the "THIS_MODULE" reference right; the naive
method of declaring a struct in the place of registration would not work
for this reason.
Signed-off-by: NArjan van de Ven <arjan@linux.intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This is to address the following error during the compilation:

  In file included from kernel/sysctl_binary.c:6:
  include/net/ip_vs.h:1406: error: expected identifier or ‘(’ before ‘{’ token
  make[1]: *** [kernel/sysctl_binary.o] Error 1
  make[1]: *** Waiting for unfinished jobs....

That manifests itself when CONFIG_IP_VS_NFCT is undefined in .config file.
Signed-off-by: NKrzysztof Wilczynski <krzysztof.wilczynski@linux.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

This patch exports several definitions that used to live under
include/net/netfilter/nf_nat.h. These definitions, although not
exported, have been used by iptables and other userspace
applications like miniupnpd since long time. Basically, these
userspace tools included some internal definition of the required
structures and they assume no changes in the binary representation
(which is OK indeed).

To resolve this situation, this patch makes public the required
structure and install them in INSTALL_HDR_PATH.

See: https://bugs.gentoo.org/376873, for more information.

This patch is heavily based on the initial patch sent by:

Anthony G. Basile <blueness@gentoo.org>

Which was entitled:

netfilter: export sanitized nf_nat.h to INSTALL_HDR_PATH
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

Acked-by: NJulian Anastasov <ja@ssi.bg>
Acked-by Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

Acked-by: NJulian Anastasov <ja@ssi.bg>
Acked-by Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

Standardize the style for compiler based printf format verification.
Standardized the location of __printf too.

Done via script and a little typing.

$ grep -rPl --include=*.[ch] -w "__attribute__" * | \
  grep -vP "^(tools|scripts|include/linux/compiler-gcc.h)" | \
  xargs perl -n -i -e 'local $/; while (<>) { s/\b__attribute__\s*\(\s*\(\s*format\s*\(\s*printf\s*,\s*(.+)\s*,\s*(.+)\s*\)\s*\)\s*\)/__printf($1, $2)/g ; print; }'

[akpm@linux-foundation.org: revert arch bits]
Signed-off-by: NJoe Perches <joe@perches.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

These two small inlines make calls to try_module_get() and
module_put() which would force us to keep module.h present
within yet another common include header.  We can avoid this
by turning them into macros.  The hci_dev_hold construct
is patterned off of raw_spin_trylock_irqsave() in spinlock.h
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

This file was using the module get/put functions in two simple inline
functions.  But module_get/put were only within scope because of
the implicit presence of module.h being everywhere.

Rather than add module.h to another file in include/  -- which is
exactly the thing we are trying to avoid, simply convert these
one-line functions into a define, as per what was done for the
device_schedule_callback() in commit 523ded71.
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

The implicit presence of module.h everywhere meant that this header
also was getting moduleparam.h which defines struct kernel_param.

Since it only needs to know that kernel_param is a struct, call that
out instead of adding an include of moduleparam.h -- to get rid of this:

include/net/netfilter/nf_conntrack.h:316: warning: 'struct kernel_param' declared inside parameter list
include/net/netfilter/nf_conntrack.h:316: warning: its scope is only this definition or declaration, which is probably not what you want
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

The <linux/module.h> pretty much brings in the kitchen sink along
with it, so it should be avoided wherever reasonably possible in
terms of being included from other commonly used <linux/something.h>
files, as it results in a measureable increase on compile times.

The worst culprit was probably device.h since it is used everywhere.
This file also had an implicit dependency/usage of mutex.h which was
masked by module.h, and is also fixed here at the same time.

There are over a dozen other headers that simply declare the
struct instead of pulling in the whole file, so follow their lead
and simply make it a few more.

Most of the implicit dependencies on module.h being present by
these headers pulling it in have been now weeded out, so we can
finally make this change with hopefully minimal breakage.
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

This file has modular references, but they are limited to
those which are covered by the simple "struct module;"
declaration used in dozens of other places.  In fact that
declaration is already there (just outside of the context
of this commit) so simply remove the include line.
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

There is nothing module specific in this header, and removing
it doesn't seem to uncover any implicit dependencies either.
Must be simply a vestige of an ancient legacy.
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

commit 66b13d99 (ipv4: tcp: fix TOS value in ACK messages sent from
TIME_WAIT) fixed IPv4 only.

This part is for the IPv6 side, adding a tclass param to ip6_xmit()

We alias tw_tclass and tw_tos, if socket family is INET6.

[ if sockets is ipv4-mapped, only IP_TOS socket option is used to fill
TOS field, TCLASS is not taken into account ]
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It was enabled by default and the messages guarded
by the define are useful.
Signed-off-by: NFlavio Leitner <fbl@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This helps in more control over debugging.
root@qemu-img-64:~# ls /pass/123
ls: cannot access /pass/123: No such file or directory
root@qemu-img-64:~# cat /sys/kernel/debug/tracing/trace
# tracer: nop
#
#           TASK-PID    CPU#    TIMESTAMP  FUNCTION
#              | |       |          |         |
              ls-1536  [001]    70.928584: 9p_protocol_dump: clnt 18446612132784021504 P9_TWALK(tag = 1)
000: 16 00 00 00 6e 01 00 01 00 00 00 02 00 00 00 01
010: 00 03 00 31 32 33 00 00 00 ff ff ff ff 00 00 00

              ls-1536  [001]    70.928587: <stack trace>
 => trace_9p_protocol_dump
 => p9pdu_finalize
 => p9_client_rpc
 => p9_client_walk
 => v9fs_vfs_lookup
 => d_alloc_and_lookup
 => walk_component
 => path_lookupat
              ls-1536  [000]    70.929696: 9p_protocol_dump: clnt 18446612132784021504 P9_RLERROR(tag = 1)
000: 0b 00 00 00 07 01 00 02 00 00 00 4e 03 00 02 00
010: 00 00 00 00 03 00 02 00 00 00 00 00 ff 43 00 00

              ls-1536  [000]    70.929697: <stack trace>
 => trace_9p_protocol_dump
 => p9_client_rpc
 => p9_client_walk
 => v9fs_vfs_lookup
 => d_alloc_and_lookup
 => walk_component
 => path_lookupat
 => do_path_lookup
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NEric Van Hensbergen <ericvh@gmail.com>

Without this msize=4294967295 will result in a crash
Signed-off-by: NDan Carpenter <error27@gmail.com>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NEric Van Hensbergen <ericvh@gmail.com>

* remove lot of update to different data structure
* add a seperate callback for zero copy request.
* above makes non zero copy code path simpler
* remove conditionalizing TREAD/TREADDIR/TWRITE in the zero copy path
* Fix the dotu p9_check_errors with zero copy. Add sufficient doc around
* Add support for both in and output buffers in zero copy callback
* pin and unpin pages in the same context
* use helpers instead of defining page offset and rest of page ourself
* Fix mem leak in p9_check_errors
* Remove 'E' and 'F' in p9pdu_vwritef
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NEric Van Hensbergen <ericvh@gmail.com>

There is a long standing bug in linux tcp stack, about ACK messages sent
on behalf of TIME_WAIT sockets.

In the IP header of the ACK message, we choose to reflect TOS field of
incoming message, and this might break some setups.

Example of things that were broken :
  - Routing using TOS as a selector
  - Firewalls
  - Trafic classification / shaping

We now remember in timewait structure the inet tos field and use it in
ACK generation, and route lookup.

Notes :
 - We still reflect incoming TOS in RST messages.
 - We could extend MuraliRaja Muniraju patch to report TOS value in
netlink messages for TIME_WAIT sockets.
 - A patch is needed for IPv6
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Now tcp_md5_hash_header() has a const tcphdr argument, we can add more
const attributes to callers.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

tcp_md5_hash_header() writes into skb header a temporary zero value,
this might confuse other users of this area.

Since tcphdr is small (20 bytes), copy it in a temporary variable and
make the change in the copy.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

INET_ECN_encapsulate() is better understood if we can read the official
statement.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Adding const qualifiers to pointers can ease code review, and spot some
bugs. It might allow compiler to optimize code further.

For example, is it legal to temporary write a null cksum into tcphdr
in tcp_md5_hash_header() ? I am afraid a sniffer could catch the
temporary null value...
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Preliminary patch before tcp constification
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Added recovery check of CA wake status in case of wake up timeout.
Added check of CA wake status in case of wake down timeout.
Signed-off-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Added sanity check for length of CAIF frames, and tear down of
CAIF link-layer device upon protocol error.
Signed-off-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

CAIF HSI uses a timer for inactivity. Upon timeout HSI-wake signaling
is initiated to allow power-down of the HSI block.
Signed-off-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Some platforms do not allow to put HSI block into low-power
mode when FIFO is not empty. The patch flushes (by reading)
FIFO at wake down sequence. Asynchronous read and write is
implemented for that. As a side effect this will also greatly
improve performance.
Signed-off-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fragmented multicast frames are delivered to a single macvlan port,
because ip defrag logic considers other samples are redundant.

Implement a defrag step before trying to send the multicast frame.
Reported-by: NBen Greear <greearb@candelatech.com>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 903ab86d of 1 March this year ("udp: Add
lockless transmit path") introduced a new fast TX path that broke the checksum
coverage computation of UDP-lite, which so far depended on up->len (only set
if the socket is locked and 0 in the fast path).

Fixed by providing both fast- and slow-path computation of checksum coverage.
The latter can be removed when UDP(-lite)v6 also uses a lockless transmit path.
Reported-by: NThomas Volkert <thomas@homer-conferencing.com>
Signed-off-by: NGerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Reuse the already existing struct nl80211_sta_flag_update to specify
both, a flag mask and the flag set itself. This means
nl80211_sta_flag_update is now used for setting station flags and also
for getting station flags.
Signed-off-by: NHelmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

mac80211 already filled in the MCS rate info for rx'ed frames but tx'ed
frames that are sent to a monitor interface during the status callback
lack this information.

Add the radiotap fields for MCS info to ieee80211_tx_status_rtap_hdr
and populate them when sending tx'ed frames to the monitors.

The needed headroom is only extended by one byte since we don't include
legacy rate information in the rtap header for HT frames.
Signed-off-by: NHelmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>