Corey Minyard found a race added in commit 271b72c7fa
(udp: RCU handling for Unicast packets.)
"If the socket is moved from one list to another list in-between the
time the hash is calculated and the next field is accessed, and the
socket has moved to the end of the new list, the traversal will not
complete properly on the list it should have, since the socket will
be on the end of the new list and there's not a way to tell it's on a
new list and restart the list traversal. I think that this can be
solved by pre-fetching the "next" field (with proper barriers) before
checking the hash."
This patch corrects this problem, introducing a new
sk_for_each_rcu_safenext() macro.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
UDP sockets are hashed in a 128 slots hash table.
This hash table is protected by *one* rwlock.
This rwlock is readlocked each time an incoming UDP message is handled.
This rwlock is writelocked each time a socket must be inserted in
hash table (bind time), or deleted from this table (close time)
This is not scalable on SMP machines :
1) Even in read mode, lock() and unlock() are atomic operations and
must dirty a contended cache line, shared by all cpus.
2) A writer might be starved if many readers are 'in flight'. This can
happen on a machine with some NIC receiving many UDP messages. User
process can be delayed a long time at socket creation/dismantle time.
This patch prepares RCU migration, by introducing 'struct udp_table
and struct udp_hslot', and using one spinlock per chain, to reduce
contention on central rwlock.
Introducing one spinlock per chain reduces latencies, for port
randomization on heavily loaded UDP servers. This also speedup
bindings to specific ports.
udp_lib_unhash() was uninlined, becoming to big.
Some cleanups were done to ease review of following patch
(RCUification of UDP Unicast lookups)
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Open code NIP6_FMT in the one call inside sscanf and one user
of NIP6() that could use %p6 in the netfilter code.
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add new_mapping() implementation to the netlink xfrm_mgr to notify
address/port changes detected in UDP encapsulated ESP packets.
Signed-off-by: Martin Willi <martin@strongswan.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a patch to provide on demand route cache rebuilding. Currently, our
route cache is rebulid periodically regardless of need. This introduced
unneeded periodic latency. This patch offers a better approach. Using code
provided by Eric Dumazet, we compute the standard deviation of the average hash
bucket chain length while running rt_check_expire. Should any given chain
length grow to larger that average plus 4 standard deviations, we trigger an
emergency hash table rebuild for that net namespace. This allows for the common
case in which chains are well behaved and do not grow unevenly to not incur any
latency at all, while those systems (which may be being maliciously attacked),
only rebuild when the attack is detected. This patch take 2 other factors into
account:
1) chains with multiple entries that differ by attributes that do not affect the
hash value are only counted once, so as not to unduly bias system to rebuilding
if features like QOS are heavily used
2) if rebuilding crosses a certain threshold (which is adjustable via the added
sysctl in this patch), route caching is disabled entirely for that net
namespace, since constant rebuilding is less efficient that no caching at all
Tested successfully by me.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fixes an OOPS in hard_header if a Phonet address is assigned to a
non-Phonet network interface.
Signed-off-by: Remi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'for-linus' of git://neil.brown.name/md:
md: allow extended partitions on md devices.
md: use sysfs_notify_dirent to notify changes to md/dev-xxx/state
md: use sysfs_notify_dirent to notify changes to md/array_state
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
HID: Add support for Sony Vaio VGX-TP1E
HID: fix lock imbalance in hiddev
HID: fix lock imbalance in hidraw
HID: fix hidbus/appletouch device binding regression
HID: add hid_type to general hid struct
HID: quirk for OLED devices present in ASUS G50/G70/G71
HID: Remove "default m" for Thrustmaster and Zeroplus
HID: fix hidraw_exit section mismatch
HID: add support for another Gyration remote control
Revert "HID: Invert HWHEEL mappings for some Logitech mice"
Based on input from Andi Kleen:
share the platform detection code with ioat_dma and disable the channel in
dma engine only for specific platforms.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (29 commits)
tcp: Restore ordering of TCP options for the sake of inter-operability
net: Fix disjunct computation of netdev features
sctp: Fix to handle SHUTDOWN in SHUTDOWN_RECEIVED state
sctp: Fix to handle SHUTDOWN in SHUTDOWN-PENDING state
sctp: Add check for the TSN field of the SHUTDOWN chunk
sctp: Drop ICMP packet too big message with MTU larger than current PMTU
p54: enable 2.4/5GHz spectrum by eeprom bits.
orinoco: reduce stack usage in firmware download path
ath5k: fix suspend-related oops on rmmod
[netdrvr] fec_mpc52xx: Implement polling, to make netconsole work.
qlge: Fix MSI/legacy single interrupt bug.
smc911x: Make the driver safer on SMP
smc911x: Add IRQ polarity configuration
smc911x: Allow Kconfig dependency on ARM
sis190: add identifier for Atheros AR8021 PHY
8139x: reduce message severity on driver overlap
igb: add IGB_DCA instead of selecting INTEL_IOATDMA
igb: fix tx data corruption with transition to L0s on 82575
ehea: Fix memory hotplug support
netdev: DM9000: remove BLACKFIN hacking in DM9000 netdev driver
...
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
math-emu: Fix thinko in _FP_DIV
math-emu: Fix signalling of underflow and inexact while packing result.
sparc: Add checkstack support
sparc: correct section of current_pc()
sparc: correct section of apc_no_idle
sparc64: Fix race in arch/sparc64/kernel/trampoline.S
libata currently has a pretty dumb ATA_MAX_QUEUE loop for finding
a free tag to use. Instead of fixing that up, convert libata to
using block layer tagging - gets rid of code in libata, and is also
much faster.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We're getting a lot of storage drivers blamed for interrupt misrouting
issues. This patch provides a standard way of reporting the problem
... and, if possible, correcting it.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
* 'v28-range-hrtimers-for-linus-v2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (37 commits)
hrtimers: add missing docbook comments to struct hrtimer
hrtimers: simplify hrtimer_peek_ahead_timers()
hrtimers: fix docbook comments
DECLARE_PER_CPU needs linux/percpu.h
hrtimers: fix typo
rangetimers: fix the bug reported by Ingo for real
rangetimer: fix BUG_ON reported by Ingo
rangetimer: fix x86 build failure for the !HRTIMERS case
select: fix alpha OSF wrapper
select: fix alpha OSF wrapper
hrtimer: peek at the timer queue just before going idle
hrtimer: make the futex() system call use the per process slack value
hrtimer: make the nanosleep() syscall use the per process slack
hrtimer: fix signed/unsigned bug in slack estimator
hrtimer: show the timer ranges in /proc/timer_list
hrtimer: incorporate feedback from Peter Zijlstra
hrtimer: add a hrtimer_start_range() function
hrtimer: another build fix
hrtimer: fix build bug found by Ingo
hrtimer: make select() and poll() use the hrtimer range feature
...
* git://git.kernel.org/pub/scm/linux/kernel/git/viro/bdev: (66 commits)
[PATCH] kill the rest of struct file propagation in block ioctls
[PATCH] get rid of struct file use in blkdev_ioctl() BLKBSZSET
[PATCH] get rid of blkdev_locked_ioctl()
[PATCH] get rid of blkdev_driver_ioctl()
[PATCH] sanitize blkdev_get() and friends
[PATCH] remember mode of reiserfs journal
[PATCH] propagate mode through swsusp_close()
[PATCH] propagate mode through open_bdev_excl/close_bdev_excl
[PATCH] pass fmode_t to blkdev_put()
[PATCH] kill the unused bsize on the send side of /dev/loop
[PATCH] trim file propagation in block/compat_ioctl.c
[PATCH] end of methods switch: remove the old ones
[PATCH] switch sr
[PATCH] switch sd
[PATCH] switch ide-scsi
[PATCH] switch tape_block
[PATCH] switch dcssblk
[PATCH] switch dasd
[PATCH] switch mtd_blkdevs
[PATCH] switch mmc
...
