This patch introduces a constant limit of the fragment queue hash
table bucket list lengths. Currently the limit 128 is choosen somewhat
arbitrary and just ensures that we can fill up the fragment cache with
empty packets up to the default ip_frag_high_thresh limits. It should
just protect from list iteration eating considerable amounts of cpu.
If we reach the maximum length in one hash bucket a warning is printed.
This is implemented on the caller side of inet_frag_find to distinguish
between the different users of inet_fragment.c.
I dropped the out of memory warning in the ipv4 fragment lookup path,
because we already get a warning by the slab allocator.
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Jesper Dangaard Brouer <jbrouer@redhat.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Dave Jones reported a lockdep splat occurring in IP defrag code.
commit 6d7b857d54 (net: use lib/percpu_counter API for
fragmentation mem accounting) added a possible deadlock.
Because percpu_counter_sum_positive() needs to acquire
a lock that can be used from softirq, we need to disable BH
in sum_frag_mem_limit()
Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Updating the fragmentation queues LRU (Least-Recently-Used) list,
required taking the hash writer lock. However, the LRU list isn't
tied to the hash at all, so we can use a separate lock for it.
Original-idea-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace the per network namespace shared atomic "mem" accounting
variable, in the fragmentation code, with a lib/percpu_counter.
Getting percpu_counter to scale to the fragmentation code usage
requires some tweaks.
At first view, percpu_counter looks superfast, but it does not
scale on multi-CPU/NUMA machines, because the default batch size
is too small, for frag code usage. Thus, I have adjusted the
batch size by using __percpu_counter_add() directly, instead of
percpu_counter_sub() and percpu_counter_add().
The batch size is increased to 130.000, based on the largest 64K
fragment memory usage. This does introduce some imprecise
memory accounting, but its does not need to be strict for this
use-case.
It is also essential, that the percpu_counter, does not
share cacheline with other writers, to make this scale.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This change is primarily a preparation to ease the extension of memory
limit tracking.
The change does reduce the number atomic operation, during freeing of
a frag queue. This does introduce a some performance improvement, as
these atomic operations are at the core of the performance problems
seen on NUMA systems.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fragmentation code cacheline adjusting of struct inet_frag_queue.
Take advantage of the size of struct timer_list, and move all but
spinlock_t lock, below the timer struct. On 64-bit 'lru_list',
'list' and 'refcnt', fits exactly into the next cacheline, and a
new cacheline starts at 'fragments'.
The netns_frags *net pointer is moved to the end of the struct,
because its used in a compare, with "next/close-by" elements of
which this struct is embedded into.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The globally shared rwlock, of struct inet_frags, shares
cacheline with the 'rnd' number, which is used by the hash
calculations. Fix this, as this obviously is a bad idea, as
unnecessary cache-misses will occur when accessing the 'rnd'
number.
Also small note that, moving function ptr (*match) up in struct,
is to avoid it lands on the next cacheline (on 64-bit).
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This small cacheline adjustment of struct netns_frags improves
performance significantly for the fragmentation code.
Struct members 'lru_list' and 'mem' are both hot elements, and it
hurts performance, due to cacheline bouncing at every call point,
when they share a cacheline. Also notice, how mem is placed
together with 'high_thresh' and 'low_thresh', as they are used in
the compare operations together.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Michal Kubeček <mkubecek@suse.cz>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
IPv4 conntrack defragments incoming packet at the PRE_ROUTING hook and
(in case of forwarded packets) refragments them at POST_ROUTING
independent of the IP_DF flag. Refragmentation uses the dst_mtu() of
the local route without caring about the original fragment sizes,
thereby breaking PMTUD.
This patch fixes this by keeping track of the largest received fragment
with IP_DF set and generates an ICMP fragmentation required error during
refragmentation if that size exceeds the MTU.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: David S. Miller <davem@davemloft.net>
- match() method returns a boolean
- return (A && B && C && D) -> return A && B && C && D
- fix indentation
Signed-off-by: Eric Dumazet <edumazet@google.com>
add fast path for in-order fragments
As the fragments are sent in order in most of OSes, such as Windows, Darwin and
FreeBSD, it is likely the new fragments are at the end of the inet_frag_queue.
In the fast path, we check if the skb at the end of the inet_frag_queue is the
prev we expect.
Signed-off-by: Changli Gao <xiaosuo@gmail.com>
----
include/net/inet_frag.h | 1 +
net/ipv4/ip_fragment.c | 12 ++++++++++++
net/ipv6/reassembly.c | 11 +++++++++++
3 files changed, 24 insertions(+)
Signed-off-by: David S. Miller <davem@davemloft.net>
Impact: Attribute function with __releases(...)
Fix this sparse warning:
net/ipv4/inet_fragment.c:276:35: warning: context imbalance in 'inet_frag_find' - unexpected unlock
Signed-off-by: Hannes Eder <hannes@hanneseder.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
On Fri, 2008-03-28 at 03:24 -0700, Andrew Morton wrote:
> they should all be renamed.
Done for include/net and net
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On namespace start we mainly prepare the ctl variables.
When the namespace is stopped we have to kill all the fragments that
point to this namespace. The inet_frags_exit_net() handles it.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The inet_frags.lru_list is used for evicting only, so we have
to make it per-namespace, to evict only those fragments, who's
namespace exceeded its high threshold, but not the whole hash.
Besides, this helps to avoid long loops in evictor.
The spinlock is not per-namespace because it protects the
hash table as well, which is global.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since we have one hashtable to lookup the fragment, having
different secret_interval-s for hash rebuild doesn't make
sense, so move this one to inet_frags.
The inet_frags_ctl becomes empty after this, so remove it.
The appropriate ctl table is kept read-only in namespaces.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is the same as with the timeout variable.
Currently, after exceeding the high threshold _all_
the fragments are evicted, but it will be fixed in
later patch.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move it to the netns_frags, adjust the usage and
make the appropriate ctl table writable.
Now fragment, that live in different namespaces can
live for different times.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is also simple, but introduces more changes, since
then mem counter is altered in more places.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is simple - just move the variable from struct inet_frags
to struct netns_frags and adjust the usage appropriately.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since fragment management code is consolidated, we cannot have the
pointer from inet_frag_queue to struct net, since we must know what
king of fragment this is.
So, I introduce the netns_frags structure. This one is currently
empty, but will be eventually filled with per-namespace
attributes. Each inet_frag_queue is tagged with this one.
The conntrack_reasm is not "netns-izated", so it has one static
netns_frags instance to keep working in init namespace.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since this callback is used to check for conflicts in
hashtable when inserting a newly created frag queue, we can
do the same by checking for matching the queue with the
argument, used to create one.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Here we need another callback ->match to check whether the
entry found in hash matches the key passed. The key used
is the same as the creation argument for inet_frag_create.
Yet again, this ->match is the same for netfilter and ipv6.
Running a frew steps forward - this callback will later
replace the ->equal one.
Since the inet_frag_find() uses the already consolidated
inet_frag_create() remove the xxx_frag_create from protocol
codes.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This one uses the xxx_frag_intern() and xxx_frag_alloc()
routines, which are already consolidated, so remove them
from protocol code (as promised).
The ->constructor callback is used to init the rest of
the frag queue and it is the same for netfilter and ipv6.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Just perform the kzalloc() allocation and setup common
fields in the inet_frag_queue(). Then return the result
to the caller to initialize the rest.
The inet_frag_alloc() may return NULL, so check the
return value before doing the container_of(). This looks
ugly, but the xxx_frag_alloc() will be removed soon.
The xxx_expire() timer callbacks are patches,
because the argument is now the inet_frag_queue, not
the protocol specific queue.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This routine checks for the existence of a given entry
in the hash table and inserts the new one if needed.
The ->equal callback is used to compare two frag_queue-s
together, but this one is temporary and will be removed
later. The netfilter code and the ipv6 one use the same
routine to compare frags.
The inet_frag_intern() always returns non-NULL pointer,
so convert the inet_frag_queue into protocol specific
one (with the container_of) without any checks.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
These ones use the generic data types too, so move
them in one place.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The evictors collect some statistics for ipv4 and ipv6,
so make it return the number of evicted queues and account
them all at once in the caller.
The XXX_ADD_STATS_BH() macros are just for this case,
but maybe there are places in code, that can make use of
them as well.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
To make in possible we need to know the exact frag queue
size for inet_frags->mem management and two callbacks:
* to destoy the skb (optional, used in conntracks only)
* to free the queue itself (mandatory, but later I plan to
move the allocation and the destruction of frag_queues
into the common place, so this callback will most likely
be optional too).
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This code works with the generic data types as well, so
move this into inet_fragment.c
This move makes it possible to hide the secret_timer
management and the secret_rebuild routine completely in
the inet_fragment.c
Introduce the ->hashfn() callback in inet_frags() to get
the hashfun for a given inet_frag_queue() object.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since now all the xxx_frag_kill functions now work
with the generic inet_frag_queue data type, this can
be moved into a common place.
The xxx_unlink() code is moved as well.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some sysctl variables are used to tune the frag queues
management and it will be useful to work with them in
a common way in the future, so move them into one
structure, moreover they are the same for all the frag
management codes.
I don't place them in the existing inet_frags object,
introduced in the previous patch for two reasons:
1. to keep them in the __read_mostly section;
2. not to export the whole inet_frags objects outside.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are some objects that are common in all the places
which are used to keep track of frag queues, they are:
* hash table
* LRU list
* rw lock
* rnd number for hash function
* the number of queues
* the amount of memory occupied by queues
* secret timer
Move all this stuff into one structure (struct inet_frags)
to make it possible use them uniformly in the future. Like
with the previous patch this mostly consists of hunks like
- write_lock(&ipfrag_lock);
+ write_lock(&ip4_frags.lock);
To address the issue with exporting the number of queues and
the amount of memory occupied by queues outside the .c file
they are declared in, I introduce a couple of helpers.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce the struct inet_frag_queue in include/net/inet_frag.h
file and place there all the common fields from three structs:
* struct ipq in ipv4/ip_fragment.c
* struct nf_ct_frag6_queue in nf_conntrack_reasm.c
* struct frag_queue in ipv6/reassembly.c
After this, replace these fields on appropriate structures with
this structure instance and fix the users to use correct names
i.e. hunks like
- atomic_dec(&fq->refcnt);
+ atomic_dec(&fq->q.refcnt);
(these occupy most of the patch)
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hannes Frederic Sowa