There is no point using RCU for dst we allocate for a very short time
(used once).
Change dst_release() to take DST_NOCACHE into account, but also change
skb_dst_set_noref() to force a refcount increment for such dst.
This is a _huge_ gain, because we dont waste memory to store xx thousand
of dsts. Instead of queueing them to RCU, we can free them instantly.
CPU caches can stay hot, re-using same memory blocks to hold temporary
dsts.
Note : remove unneeded smp_mb__before_atomic_dec(); in dst_release(),
since atomic_dec_return() implies a full memory barrier.
Stress test, 160.000.000 udp frames sent, IP route cache disabled
(DDOS).
Before:
real 0m38.091s
user 0m13.189s
sys 7m53.018s
After:
real 0m29.946s
user 0m12.157s
sys 7m40.605s
For reference, if IP route cache was enabled :
real 0m32.030s
user 0m10.521s
sys 8m15.243s
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces netif_alloc_netdev_queues which is called from
register_device instead of alloc_netdev_mq. This makes TX queue
allocation symmetric with RX allocation. Also, queue locks allocation
is done in netdev_init_one_queue. Change set_real_num_tx_queues to
fail if requested number < 1 or greater than number of allocated
queues.
Signed-off-by: Tom Herbert <therbert@google.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The bonding driver currently modifies the netpoll structure in its xmit path
while sending frames from netpoll. This is racy, as other cpus can access the
netpoll structure in parallel. Since the bonding driver points np->dev to a
slave device, other cpus can inadvertently attempt to send data directly to
slave devices, leading to improper locking with the bonding master, lost frames,
and deadlocks. This patch fixes that up.
This patch also removes the real_dev pointer from the netpoll structure as that
data is really only used by bonding in the poll_controller, and we can emulate
its behavior by check each slave for IS_UP.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit e446630c96, i.e. v2.6.35-rc1,
the mcp251x chip model can be selected via the modalias member in the
struct spi_board_info. The driver stores the actual model in the
struct mcp251x_platform_data.
From the driver point of view the platform_data should be read only.
Since all in-tree users of the mcp251x have already been converted to
the modalias method, this patch moves the "model" member from the
struct mcp251x_platform_data to the driver's private data structure.
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Cc: Christian Pellegrin <chripell@fsfe.org>
Cc: Marc Zyngier <maz@misterjones.org>
In a network bench, I noticed an unfortunate false sharing between
'loopback_dev' and 'count' fields in "struct net".
'count' is written each time a socket is created or destroyed, while
loopback_dev might be often read in routing code.
Move loopback_dev in a read mostly section of "struct net"
Note: struct netns_xfrm is cache line aligned on SMP.
(It contains a "struct dst_ops")
Move it at the end to avoid holes, and reduce sizeof(struct net) by 128
bytes on ia32.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Do some cleanups of TIPC based on make namespacecheck
1. Don't export unused symbols
2. Eliminate dead code
3. Make functions and variables local
4. Rename buf_acquire to tipc_buf_acquire since it is used in several files
Compile tested only.
This make break out of tree kernel modules that depend on TIPC routines.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Acked-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit b30973f877 (node-aware skb allocation) spread a wrong habit of
allocating net drivers skbs on a given memory node : The one closest to
the NIC hardware. This is wrong because as soon as we try to scale
network stack, we need to use many cpus to handle traffic and hit
slub/slab management on cross-node allocations/frees when these cpus
have to alloc/free skbs bound to a central node.
skb allocated in RX path are ephemeral, they have a very short
lifetime : Extra cost to maintain NUMA affinity is too expensive. What
appeared as a nice idea four years ago is in fact a bad one.
In 2010, NIC hardwares are multiqueue, or we use RPS to spread the load,
and two 10Gb NIC might deliver more than 28 million packets per second,
needing all the available cpus.
Cost of cross-node handling in network and vm stacks outperforms the
small benefit hardware had when doing its DMA transfert in its 'local'
memory node at RX time. Even trying to differentiate the two allocations
done for one skb (the sk_buff on local node, the data part on NIC
hardware node) is not enough to bring good performance.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based on suggestion by Rémi Denis-Courmont to implement 'connect'
for Pipe controller logic, this patch implements 'connect' socket
call for the Pipe controller logic.
The patch does following:-
- Removes setsockopts for PNPIPE_CREATE and PNPIPE_DESTROY
- Adds setsockopt for setting the Pipe handle value
- Implements connect socket call
- Updates the Pipe controller logic
User-space should now follow below sequence with Pipe controller:-
-socket
-bind
-setsockopt for PNPIPE_PIPE_HANDLE
-connect
-setsockopt for PNPIPE_ENCAP_IP
-setsockopt for PNPIPE_ENABLE
GPRS/3G data has been tested working fine with this.
Signed-off-by: Kumar Sanghvi <kumar.sanghvi@stericsson.com>
Acked-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using the frame registration notification, we
can see when probe requests are requested and
notify the low-level driver via filtering. The
flag is also set in AP and IBSS modes.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Drivers may need to adjust their filters according
to frame registrations, so notify them about them.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
wext: fix alignment problem in serializing 'struct iw_point'
This fixes a typo in the definition of the serialized length of struct iw_point:
a) wireless.h is exported to userspace, the typo causes IW_EV_POINT_PK_LEN
to be 12 on 64-bit, and 8 on 32-bit systems (causing misalignment);
b) in compat-64 mode iwe_stream_add_point() memcpys overlap (see below).
The second case in in compat-64 mode looks like (variable names are as in
include/net/iw_handler.h:iwe_stream_add_point()):
point_len = IW_EV_COMPAT_POINT_LEN = 8
lcp_len = IW_EV_COMPAT_LCP_LEN = 4
2nd memcpy: IW_EV_POINT_PK_LEN - IW_EV_LCP_PK_LEN = 12 - 4 = 8
IW_EV_LCP_PK_LEN
<--------------> *---> 'extra' data area
+-------+-------+-------+-------+---------------+------- ...-+
| len | cmd |length | flags | (empty) -> extra ... |
+-------+-------+-------+-------+---------------+------- ...-+
2 2 2 2 4
lcp_len
<--------------> <-!! OVERLAP !!>
<--1st memcpy--><------- 2nd memcpy ----------->
<---- 3rd memcpy ------- ... >
<--------- point_len ---------->
This case could cause overrun whenever iw_point.length < 4.
The other two cases are -
* 32-bit systems: IW_EV_POINT_PK_LEN - IW_EV_LCP_PK_LEN = 8 - 4 = 4,
the second memcpy copies exactly the 4 required bytes;
* 64-bit systems: IW_EV_POINT_PK_LEN - IW_EV_LCP_PK_LEN = 12 - 4 = 8,
the second memcpy copies a superfluous (but non overlapping) 4 bytes.
The patch changes IW_EV_POINT_PK_LEN to be 8, so that in all 3 cases always only
the requested iw_point.{length,flags} (both __u16) are copied, avoiding overrrun
(compat-64) and superfluous copy (64-bit). In addition, the userspace header is
sanitized (in agreement with version 30 of the wireless tools).
Many thanks to Johannes Berg for help and review with this patch.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
We tried very hard to remove all possible dev_hold()/dev_put() pairs in
network stack, using RCU conversions.
There is still an unavoidable device refcount change for every dst we
create/destroy, and this can slow down some workloads (routers or some
app servers, mmap af_packet)
We can switch to a percpu refcount implementation, now dynamic per_cpu
infrastructure is mature. On a 64 cpus machine, this consumes 256 bytes
per device.
On x86, dev_hold(dev) code :
before
lock incl 0x280(%ebx)
after:
movl 0x260(%ebx),%eax
incl fs:(%eax)
Stress bench :
(Sending 160.000.000 UDP frames,
IP route cache disabled, dual E5540 @2.53GHz,
32bit kernel, FIB_TRIE)
Before:
real 1m1.662s
user 0m14.373s
sys 12m55.960s
After:
real 0m51.179s
user 0m15.329s
sys 10m15.942s
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit adds a bt_sock_stream_recvmsg() function for use by any
Bluetooth code that uses SOCK_STREAM sockets. This code is copied
from rfcomm_sock_recvmsg() with minimal modifications to remove
RFCOMM-specific functionality and improve readability.
L2CAP (with the SOCK_STREAM socket type) and RFCOMM have common needs
when it comes to reading data. Proper stream read semantics require
that applications can read from a stream one byte at a time and not
lose any data. The RFCOMM code already operated on and pulled data
from the underlying L2CAP socket, so very few changes were required to
make the code more generic for use with non-RFCOMM data over L2CAP.
Applications that need more awareness of L2CAP frame boundaries are
still free to use SOCK_SEQPACKET sockets, and may verify that they
connection did not fall back to basic mode by calling getsockopt().
Signed-off-by: Mat Martineau <mathewm@codeaurora.org>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
HCI transport drivers may not know what type of radio an AMP device has
so only say whether they're BR/EDR or AMP devices.
Signed-off-by: David Vrabel <david.vrabel@csr.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
Le mardi 12 octobre 2010 à 00:02 +0200, Eric Dumazet a écrit :
> Here is the followup patch.
>
> Thanks !
>
Oops, this was an old version, the up2date ones also took care of "used"
field.
I guess its time for a sleep, sorry again.
[PATCH net-next V2] neigh: reorder struct neighbour fields
(refcnt) and (ha_lock, ha, used, dev, output, ops, primary_key) should
be placed on a separate cache lines.
refcnt can be often written, while other fields are mostly read.
This gave me good result on stress test :
before:
real 0m45.570s
user 0m15.525s
sys 9m56.669s
After:
real 0m41.841s
user 0m15.261s
sys 8m45.949s
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct dst_ops tracks number of allocated dst in an atomic_t field,
subject to high cache line contention in stress workload.
Switch to a percpu_counter, to reduce number of time we need to dirty a
central location. Place it on a separate cache line to avoid dirtying
read only fields.
Stress test :
(Sending 160.000.000 UDP frames,
IP route cache disabled, dual E5540 @2.53GHz,
32bit kernel, FIB_TRIE, SLUB/NUMA)
Before:
real 0m51.179s
user 0m15.329s
sys 10m15.942s
After:
real 0m45.570s
user 0m15.525s
sys 9m56.669s
With a small reordering of struct neighbour fields, subject of a
following patch, (to separate refcnt from other read mostly fields)
real 0m41.841s
user 0m15.261s
sys 8m45.949s
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a seqlock in struct neighbour to protect neigh->ha[], and avoid
dirtying neighbour in stress situation (many different flows / dsts)
Dirtying takes place because of read_lock(&n->lock) and n->used writes.
Switching to a seqlock, and writing n->used only on jiffies changes
permits less dirtying.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch updates IEEE802.11 mesh constants to be consistent with newly
approved values. It modifies some values, as well as adds many new constants
in preparation for updating mesh code to the current 802.11s drafts. ANA
numbers were taken from:
https://mentor.ieee.org/802.11/dcn/09/11-09-0031-12-0000-ana-database-assigned-number-authority.xls
A few notes are in order:
1. This will break backwards compatibility with existing Linux kernels as
over-the-air constants have changed.
2. Some old and obsolete constants have been retained for now as the mesh code
itself hasn't been updated yet to the new 802.11s draft. This was desired to
keep the existing mesh scheme working until it can be updated. Adding the
approved values is the first step in updating the mesh code.
3. Obsolete constants have been clearly marked.
4. All ANA approved 802.11s constants have been added.
Signed-off-by: Steve deRosier <steve@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Using these, user space can calculate a relative channel utilization
with arbitrary intervals by regularly taking snapshots of the survey
results.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Some stats for /proc/net/wireless (and wext in general) are not
being set. This patch addresses a few of those with values easily
obtained from mac80211 core.
Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When a new dst is used to send a frame, neigh_resolve_output() tries to
associate an struct hh_cache to this dst, calling neigh_hh_init() with
the neigh rwlock write locked.
Most of the time, hh_cache is already known and linked into neighbour,
so we find it and increment its refcount.
This patch changes the logic so that we call neigh_hh_init() with
neighbour lock read locked only, so that fast path can be run in
parallel by concurrent cpus.
This brings part of the speedup we got with commit c7d4426a98
(introduce DST_NOCACHE flag) for non cached dsts, even for cached ones,
removing one of the contention point that routers hit on multiqueue
enabled machines.
Further improvements would need to use a seqlock instead of an rwlock to
protect neigh->ha[], to not dirty neigh too often and remove two atomic
ops.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current code works like this:
int garbage, status;
socklen_t len = sizeof(status);
/* enable pipe */
setsockopt(fd, SOL_PNPIPE, PNPIPE_ENABLE, &garbage, sizeof(garbage));
/* disable pipe */
setsockopt(fd, SOL_PNPIPE, PNPIPE_DISABLE, &garbage, sizeof(garbage));
/* get status */
getsockopt(fd, SOL_PNPIPE, PNPIPE_INQ, &status, &len);
...which does not follow the usual socket option pattern. This patch
merges all three "options" into a single gettable&settable option,
before Linux 2.6.37 gets out:
int status;
socklen_t len = sizeof(status);
/* enable pipe */
status = 1;
setsockopt(fd, SOL_PNPIPE, PNPIPE_ENABLE, &status, sizeof(status));
/* disable pipe */
status = 0;
setsockopt(fd, SOL_PNPIPE, PNPIPE_ENABLE, &status, sizeof(status));
/* get status */
getsockopt(fd, SOL_PNPIPE, PNPIPE_ENABLE, &status, &len);
This also fixes the error code from EFAULT to ENOTCONN.
Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Cc: Kumar Sanghvi <kumar.sanghvi@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There's no need for the WDS peer address
to not be const, so make it const.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This is the second step for neighbour RCU conversion.
(first was commit d6bf7817 : RCU conversion of neigh hash table)
neigh_lookup() becomes lockless, but still take a reference on found
neighbour. (no more read_lock()/read_unlock() on tbl->lock)
struct neighbour gets an additional rcu_head field and is freed after an
RCU grace period.
Future work would need to eventually not take a reference on neighbour
for temporary dst (DST_NOCACHE), but this would need dst->_neighbour to
use a noref bit like we did for skb->_dst.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This information is already available in mac80211, we just need to export it
via cfg80211 and nl80211.
Signed-off-by: Bruno Randolf <br1@einfach.org>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This adds API to allow adding per-station GTKs,
updates mac80211 to support it, and also allows
drivers to remove a key from hwaccel again when
this may be necessary due to multiple GTKs.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Allow sysadmins to configure the number of multicast
membership report sent on a link failure event.
Signed-off-by: Flavio Leitner <fleitner@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David
This is the first step for RCU conversion of neigh code.
Next patches will convert hash_buckets[] and "struct neighbour" to RCU
protected objects.
Thanks
[PATCH net-next] net neigh: RCU conversion of neigh hash table
Instead of storing hash_buckets, hash_mask and hash_rnd in "struct
neigh_table", a new structure is defined :
struct neigh_hash_table {
struct neighbour **hash_buckets;
unsigned int hash_mask;
__u32 hash_rnd;
struct rcu_head rcu;
};
And "struct neigh_table" has an RCU protected pointer to such a
neigh_hash_table.
This means the signature of (*hash)() function changed: We need to add a
third parameter with the actual hash_rnd value, since this is not
anymore a neigh_table field.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In various situations, a device provides a packet to our stack and we
drop it before it enters protocol stack :
- softnet backlog full (accounted in /proc/net/softnet_stat)
- bad vlan tag (not accounted)
- unknown/unregistered protocol (not accounted)
We can handle a per-device counter of such dropped frames at core level,
and automatically adds it to the device provided stats (rx_dropped), so
that standard tools can be used (ifconfig, ip link, cat /proc/net/dev)
This is a generalization of commit 8990f468a (net: rx_dropped
accounting), thus reverting it.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The "flags" member of "struct wait_queue_t" is used in several places in
the kernel code without beeing initialized by init_wait(). "flags" is
used in bitwise operations.
If "flags" not initialized then unexpected behaviour may take place.
Incorrect flags might used later in code.
Added initialization of "wait_queue_t.flags" with zero value into
"init_wait".
Signed-off-by: Evgeny Kuznetsov <EXT-Eugeny.Kuznetsov@nokia.com>
[ The bit we care about does end up being initialized by both
prepare_to_wait() and add_to_wait_queue(), so this doesn't seem to
cause actual bugs, but is definitely the right thing to do -Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
With all the recent module loading cleanups, we've minimized the code
that sits under module_mutex, fixing various deadlocks and making it
possible to do most of the module loading in parallel.
However, that whole conversion totally missed the rather obscure code
that adds a new module to the list for BUG() handling. That code was
doubly obscure because (a) the code itself lives in lib/bugs.c (for
dubious reasons) and (b) it gets called from the architecture-specific
"module_finalize()" rather than from generic code.
Calling it from arch-specific code makes no sense what-so-ever to begin
with, and is now actively wrong since that code isn't protected by the
module loading lock any more.
So this commit moves the "module_bug_{finalize,cleanup}()" calls away
from the arch-specific code, and into the generic code - and in the
process protects it with the module_mutex so that the list operations
are now safe.
Future fixups:
- move the module list handling code into kernel/module.c where it
belongs.
- get rid of 'module_bug_list' and just use the regular list of modules
(called 'modules' - imagine that) that we already create and maintain
for other reasons.
Reported-and-tested-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Adrian Bunk <bunk@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The documentation for NL80211_CMD_REMAIN_ON_CHANNEL
isn't accurate, an interface index is required by
the command. Update it accordingly.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Each family may have some amount of boilerplate
locking code that applies to most, or even all,
commands.
This allows a family to handle such things in
a more generic way, by allowing it to
a) include private flags in each operation
b) specify a pre_doit hook that is called,
before an operation's doit() callback and
may return an error directly,
c) specify a post_doit hook that can undo
locking or similar things done by pre_doit,
and finally
d) include two private pointers in each info
struct passed between all these operations
including doit(). (It's two because I'll
need two in nl80211 -- can be extended.)
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Some drivers cannot handle multiple retry rates specified by the rc
algorithm but instead use their own retry table (for example rt2800).
However, if such a device registers itself with a max_rates value of 1
the rc algorithm cannot make use of the extended information the device
can provide about retried rates. On the other hand, if a device
registers itself with a max_rates value > 1 the rc algorithm assumes
that the device can handle multi rate retries.
Fix this issue by introducing another hw parameter max_report_rates that
can be set to a different value then max_rates to indicate if a device
is capable of reporting more rates then specified in max_rates.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Added a nl interface to set the peer bssid of a WDS interface.
Signed-off-by: Bill Jordan <bjordan@rajant.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The net/cfg80211.h header file isn't exported to
userspace, so there's no need for any kind of
__KERNEL__ protection in it. If it was exported,
everything else in it would need protection as
well, not just the logging stuff ...
Cc:Joe Perches <joe@perches.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Some user space applications only want to display survey data for
the operating channel, however there is no API to get that yet.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
