Rollover has to call packet_rcv_has_room on sockets in the fanout
group to find a socket to migrate to. This operation is expensive
especially if the packet sockets use rings, when a lock has to be
acquired.
Avoid pounding on the lock by all sockets by temporarily marking a
socket as "under memory pressure" when such pressure is detected.
While set, only the socket owner may call packet_rcv_has_room on the
socket. Once it detects normal conditions, it clears the flag. The
socket is not used as a victim by any other socket in the meantime.
Under reasonably balanced load, each socket writer frequently calls
packet_rcv_has_room and clears its own pressure field. As a backup
for when the socket is rarely written to, also clear the flag on
reading (packet_recvmsg, packet_poll) if this can be done cheaply
(i.e., without calling packet_rcv_has_room). This is only for
edge cases.
Tested:
Ran bench_rollover: a process with 8 sockets in a single fanout
group, each pinned to a single cpu that receives one nic recv
interrupt. RPS and RFS are disabled. The benchmark uses packet
rx_ring, which has to take a lock when determining whether a
socket has room.
Sent 3.5 Mpps of UDP traffic with sufficient entropy to spread
uniformly across the packet sockets (and inserted an iptables
rule to drop in PREROUTING to avoid protocol stack processing).
Without this patch, all sockets try to migrate traffic to
neighbors, causing lock contention when searching for a non-
empty neighbor. The lock is the top 9 entries.
perf record -a -g sleep 5
- 17.82% bench_rollover [kernel.kallsyms] [k] _raw_spin_lock
- _raw_spin_lock
- 99.00% spin_lock
+ 81.77% packet_rcv_has_room.isra.41
+ 18.23% tpacket_rcv
+ 0.84% packet_rcv_has_room.isra.41
+ 5.20% ksoftirqd/6 [kernel.kallsyms] [k] _raw_spin_lock
+ 5.15% ksoftirqd/1 [kernel.kallsyms] [k] _raw_spin_lock
+ 5.14% ksoftirqd/2 [kernel.kallsyms] [k] _raw_spin_lock
+ 5.12% ksoftirqd/7 [kernel.kallsyms] [k] _raw_spin_lock
+ 5.12% ksoftirqd/5 [kernel.kallsyms] [k] _raw_spin_lock
+ 5.10% ksoftirqd/4 [kernel.kallsyms] [k] _raw_spin_lock
+ 4.66% ksoftirqd/0 [kernel.kallsyms] [k] _raw_spin_lock
+ 4.45% ksoftirqd/3 [kernel.kallsyms] [k] _raw_spin_lock
+ 1.55% bench_rollover [kernel.kallsyms] [k] packet_rcv_has_room.isra.41
On net-next with this patch, this lock contention is no longer a
top entry. Most time is spent in the actual read function. Next up
are other locks:
+ 15.52% bench_rollover bench_rollover [.] reader
+ 4.68% swapper [kernel.kallsyms] [k] memcpy_erms
+ 2.77% swapper [kernel.kallsyms] [k] packet_lookup_frame.isra.51
+ 2.56% ksoftirqd/1 [kernel.kallsyms] [k] memcpy_erms
+ 2.16% swapper [kernel.kallsyms] [k] tpacket_rcv
+ 1.93% swapper [kernel.kallsyms] [k] mlx4_en_process_rx_cq
Looking closer at the remaining _raw_spin_lock, the cost of probing
in rollover is now comparable to the cost of taking the lock later
in tpacket_rcv.
- 1.51% swapper [kernel.kallsyms] [k] _raw_spin_lock
- _raw_spin_lock
+ 33.41% packet_rcv_has_room
+ 28.15% tpacket_rcv
+ 19.54% enqueue_to_backlog
+ 6.45% __free_pages_ok
+ 2.78% packet_rcv_fanout
+ 2.13% fanout_demux_rollover
+ 2.01% netif_receive_skb_internal
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only migrate flows to sockets that have sufficient headroom, where
sufficient is defined as having at least 25% empty space.
The kernel has three different buffer types: a regular socket, a ring
with frames (TPACKET_V[12]) or a ring with blocks (TPACKET_V3). The
latter two do not expose a read pointer to the kernel, so headroom is
not computed easily. All three needs a different implementation to
estimate free space.
Tested:
Ran bench_rollover for 10 sec with 1.5 Mpps of single flow input.
bench_rollover has as many sockets as there are NIC receive queues
in the system. Each socket is owned by a process that is pinned to
one of the receive cpus. RFS is disabled. RPS is enabled with an
identity mapping (cpu x -> cpu x), to count drops with softnettop.
lpbb5:/export/hda3/willemb# ./bench_rollover -r -l 1000 -s
Press [Enter] to exit
cpu rx rx.k drop.k rollover r.huge r.failed
0 16 16 0 0 0 0
1 21 21 0 0 0 0
2 5227502 5227502 0 0 0 0
3 18 18 0 0 0 0
4 6083289 6083289 0 5227496 0 0
5 22 22 0 0 0 0
6 21 21 0 0 0 0
7 9 9 0 0 0 0
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace rollover state per fanout group with state per socket. Future
patches will add fields to the new structure.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
packet_rcv_fanout calls fanout_demux_rollover twice. Move all rollover
logic into the callee to simplify these callsites, especially with
upcoming changes.
The main differences between the two callsites is that the FLAG
variant tests whether the socket previously selected by another
mode (RR, RND, HASH, ..) has room before migrating flows, whereas the
rollover mode has no original socket to test.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Four minor merge conflicts:
1) qca_spi.c renamed the local variable used for the SPI device
from spi_device to spi, meanwhile the spi_set_drvdata() call
got moved further up in the probe function.
2) Two changes were both adding new members to codel params
structure, and thus we had overlapping changes to the
initializer function.
3) 'net' was making a fix to sk_release_kernel() which is
completely removed in 'net-next'.
4) In net_namespace.c, the rtnl_net_fill() call for GET operations
had the command value fixed, meanwhile 'net-next' adjusted the
argument signature a bit.
This also matches example merge resolutions posted by Stephen
Rothwell over the past two days.
Signed-off-by: David S. Miller <davem@davemloft.net>
In preparation for changing how struct net is refcounted
on kernel sockets pass the knowledge that we are creating
a kernel socket from sock_create_kern through to sk_alloc.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes an issue where the send(MSG_DONTWAIT) call
on a TX_RING is not fully non-blocking in cases where the device's sndBuf is
full. We pass nonblock=true to sock_alloc_send_skb() and return any possibly
occuring error code (most likely EGAIN) to the caller. As the fast-path stays
as it is, we keep the unlikely() around skb == NULL.
Signed-off-by: Mathias Kretschmer <mathias.kretschmer@fokus.fraunhofer.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce TP_STATUS_CSUM_VALID tp_status flag to tell the
af_packet user that at least the transport header checksum
has been already validated.
For now, the flag may be set for incoming packets only.
Signed-off-by: Alexander Drozdov <al.drozdov@gmail.com>
Cc: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is just an optimization. We don't need the value of status variable
if the packet is filtered.
Signed-off-by: Alexander Drozdov <al.drozdov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Having to say
> #ifdef CONFIG_NET_NS
> struct net *net;
> #endif
in structures is a little bit wordy and a little bit error prone.
Instead it is possible to say:
> typedef struct {
> #ifdef CONFIG_NET_NS
> struct net *net;
> #endif
> } possible_net_t;
And then in a header say:
> possible_net_t net;
Which is cleaner and easier to use and easier to test, as the
possible_net_t is always there no matter what the compile options.
Further this allows read_pnet and write_pnet to be functions in all
cases which is better at catching typos.
This change adds possible_net_t, updates the definitions of read_pnet
and write_pnet, updates optional struct net * variables that
write_pnet uses on to have the type possible_net_t, and finally fixes
up the b0rked users of read_pnet and write_pnet.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/ethernet/cadence/macb.c
Overlapping changes in macb driver, mostly fixes and cleanups
in 'net' overlapping with the integration of at91_ether into
macb in 'net-next'.
Signed-off-by: David S. Miller <davem@davemloft.net>
When an interface is deleted from a net namespace the ifindex in the
corresponding entries in PF_PACKET sockets' mclists becomes stale.
This can create inconsistencies if later an interface with the same ifindex
is moved from a different namespace (not that unlikely since ifindexes are
per-namespace).
In particular we saw problems with dev->promiscuity, resulting
in "promiscuity touches roof, set promiscuity failed. promiscuity
feature of device might be broken" warnings and EOVERFLOW failures of
setsockopt(PACKET_ADD_MEMBERSHIP).
This patch deletes the mclist entries for interfaces that are deleted.
Since this now causes setsockopt(PACKET_DROP_MEMBERSHIP) to fail with
EADDRNOTAVAIL if called after the interface is deleted, also make
packet_mc_drop not fail.
Signed-off-by: Francesco Ruggeri <fruggeri@arista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/ethernet/rocker/rocker.c
The rocker commit was two overlapping changes, one to rename
the ->vport member to ->pport, and another making the bitmask
expression use '1ULL' instead of plain '1'.
Signed-off-by: David S. Miller <davem@davemloft.net>
After TIPC doesn't depend on iocb argument in its internal
implementations of sendmsg() and recvmsg() hooks defined in proto
structure, no any user is using iocb argument in them at all now.
Then we can drop the redundant iocb argument completely from kinds of
implementations of both sendmsg() and recvmsg() in the entire
networking stack.
Cc: Christoph Hellwig <hch@lst.de>
Suggested-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As part of an effort to move skb->dropcount to skb->cb[], use
a common function in order to set dropcount in struct sk_buff.
Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As part of an effort to move skb->dropcount to skb->cb[] use a common
macro in protocol families using skb->cb[] for ancillary data to
validate available room in skb->cb[].
Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As part of an effort to move skb->dropcount to skb->cb[], 4 bytes
of additional room are needed in skb->cb[] in packet sockets.
Store the skb original length in the first two fields of sockaddr_ll
(sll_family and sll_protocol) as they can be derived from the skb when
needed.
Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Before da413eec72 ("packet: Fixed TPACKET V3 to signal poll when block is
closed rather than every packet") poll listening for an af_packet socket was
not signaled if there was no packets to process. After the patch poll is
signaled evety time when block retire timer expires. That happens because
af_packet closes the current block on timeout even if the block is empty.
Passing empty blocks to the user not only wastes CPU but also wastes ring
buffer space increasing probability of packets dropping on small timeouts.
Signed-off-by: Alexander Drozdov <al.drozdov@gmail.com>
Cc: Dan Collins <dan@dcollins.co.nz>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Guy Harris <guy@alum.mit.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
Packets defragmentation was introduced for PACKET_FANOUT_HASH only,
see 7736d33f42 ("packet: Add pre-defragmentation support for ipv4
fanouts")
It may be useful to have defragmentation enabled regardless of
fanout type. Without that, the AF_PACKET user may have to:
1. Collect fragments from different rings
2. Defragment by itself
Signed-off-by: Alexander Drozdov <al.drozdov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Contrary to common expectations for an "int" return, these functions
return only a positive value -- if used correctly they cannot even
return 0 because the message header will necessarily be in the skb.
This makes the very common pattern of
if (genlmsg_end(...) < 0) { ... }
be a whole bunch of dead code. Many places also simply do
return nlmsg_end(...);
and the caller is expected to deal with it.
This also commonly (at least for me) causes errors, because it is very
common to write
if (my_function(...))
/* error condition */
and if my_function() does "return nlmsg_end()" this is of course wrong.
Additionally, there's not a single place in the kernel that actually
needs the message length returned, and if anyone needs it later then
it'll be very easy to just use skb->len there.
Remove this, and make the functions void. This removes a bunch of dead
code as described above. The patch adds lines because I did
- return nlmsg_end(...);
+ nlmsg_end(...);
+ return 0;
I could have preserved all the function's return values by returning
skb->len, but instead I've audited all the places calling the affected
functions and found that none cared. A few places actually compared
the return value with <= 0 in dump functionality, but that could just
be changed to < 0 with no change in behaviour, so I opted for the more
efficient version.
One instance of the error I've made numerous times now is also present
in net/phonet/pn_netlink.c in the route_dumpit() function - it didn't
check for <0 or <=0 and thus broke out of the loop every single time.
I've preserved this since it will (I think) have caused the messages to
userspace to be formatted differently with just a single message for
every SKB returned to userspace. It's possible that this isn't needed
for the tools that actually use this, but I don't even know what they
are so couldn't test that changing this behaviour would be acceptable.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/xen-netfront.c
Minor overlapping changes in xen-netfront.c, mostly to do
with some buffer management changes alongside the split
of stats into TX and RX.
Signed-off-by: David S. Miller <davem@davemloft.net>
The same macros are used for rx as well. So rename it.
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
The expression in ll_header_truncated() tests less than or equal, but
the warning prints less than. Update the warning.
Reported-by: Jouni Malinen <jkmalinen@gmail.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Due to a misplaced parenthesis, the expression
(unlikely(offset) < 0),
which expands to
(__builtin_expect(!!(offset), 0) < 0),
never evaluates to true. Therefore, when sending packets with
PF_PACKET/SOCK_DGRAM, packet_snd() does not abort as intended
if the creation of the layer 2 header fails.
Spotted by Coverity - CID 1259975 ("Operands don't affect result").
Fixes: 9c7077622d ("packet: make packet_snd fail on len smaller than l2 header")
Signed-off-by: Christoph Jaeger <cj@linux.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make TPACKET_V3 signal poll when block is closed rather than for every
packet. Side effect is that poll will be signaled when block retire
timer expires which didn't previously happen. Issue was visible when
sending packets at a very low frequency such that all blocks are retired
before packets are received by TPACKET_V3. This caused avoidable packet
loss. The fix ensures that the signal is sent when blocks are closed
which covers the normal path where the block is filled as well as the
path where the timer expires. The case where a block is filled without
moving to the next block (ie. all blocks are full) will still cause poll
to be signaled.
Signed-off-by: Dan Collins <dan@dcollins.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
Willem de Bruijn
David S. Miller
Eric W. Biederman
Kretschmer, Mathias
Alexander Drozdov
Francesco Ruggeri
Ying Xue
Eyal Birger
Johannes Berg
Jiri Pirko
Christoph Jaeger
Dan Collins