| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Similar to support for ipv4 erspan, this patch adds
erspan v2 to ip6erspan tunnel.
Signed-off-by: William Tu <u9012063@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The patch adds support for erspan version 2. Not all features are
supported in this patch. The SGT (security group tag), GRA (timestamp
granularity), FT (frame type) are set to fixed value. Only hardware
ID and direction are configurable. Optional subheader is also not
supported.
Signed-off-by: William Tu <u9012063@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The patch refactors the existing erspan implementation in order
to support erspan version 2, which has additional metadata. So, in
stead of having one 'struct erspanhdr' holding erspan version 1,
breaks it into 'struct erspan_base_hdr' and 'struct erspan_metadata'.
Signed-off-by: William Tu <u9012063@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Steffen Klassert says:
====================
pull request (net-next): ipsec-next 2017-12-15
1) Currently we can add or update socket policies, but
not clear them. Support clearing of socket policies
too. From Lorenzo Colitti.
2) Add documentation for the xfrm device offload api.
From Shannon Nelson.
3) Fix IPsec extended sequence numbers (ESN) for
IPsec offloading. From Yossef Efraim.
4) xfrm_dev_state_add function returns success even for
unsupported options, fix this to fail in such cases.
From Yossef Efraim.
5) Remove a redundant xfrm_state assignment.
From Aviv Heller.
Please pull or let me know if there are problems.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In order to let MT7530 switch can recognize well those egress packets
having both special tag and VLAN tag, the information about the special
tag should be carried on the existing VLAN tag. On the other hand, it's
unnecessary for extra handling for ingress packets when VLAN tag is
present since it is able to put the VLAN tag after the special tag and
then follow the existing way to parse.
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Prior to this patch, active Fast Open is paused on a specific
destination IP address if the previous connections to the
IP address have experienced recurring timeouts . But recent
experiments by Microsoft (https://goo.gl/cykmn7) and Mozilla
browsers indicate the isssue is often caused by broken middle-boxes
sitting close to the client. Therefore it is much better user
experience if Fast Open is disabled out-right globally to avoid
experiencing further timeouts on connections toward other
destinations.
This patch changes the destination-IP disablement to global
disablement if a connection experiencing recurring timeouts
or aborts due to timeout. Repeated incidents would still
exponentially increase the pause time, starting from an hour.
This is extremely conservative but an unfortunate compromise to
minimize bad experience due to broken middle-boxes.
Reported-by: Dragana Damjanovic <ddamjanovic@mozilla.com>
Reported-by: Patrick McManus <mcmanus@ducksong.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Wei Wang <weiwan@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
skb_warn_bad_offload warns when packets enter the GSO stack that
require skb_checksum_help or vice versa. Do not warn on arbitrary
bad packets. Packet sockets can craft many. Syzkaller was able to
demonstrate another one with eth_type games.
In particular, suppress the warning when segmentation returns an
error, which is for reasons other than checksum offload.
See also commit 36c92474498a ("net: WARN if skb_checksum_help() is
called on skb requiring segmentation") for context on this warning.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Before this patch the bridge used a fixed 256 element hash table which
was fine for small use cases (in my tests it starts to degrade
above 1000 entries), but it wasn't enough for medium or large
scale deployments. Modern setups have thousands of participants in a
single bridge, even only enabling vlans and adding a few thousand vlan
entries will cause a few thousand fdbs to be automatically inserted per
participating port. So we need to scale the fdb table considerably to
cope with modern workloads, and this patch converts it to use a
rhashtable for its operations thus improving the bridge scalability.
Tests show the following results (10 runs each), at up to 1000 entries
rhashtable is ~3% slower, at 2000 rhashtable is 30% faster, at 3000 it
is 2 times faster and at 30000 it is 50 times faster.
Obviously this happens because of the properties of the two constructs
and is expected, rhashtable keeps pretty much a constant time even with
10000000 entries (tested), while the fixed hash table struggles
considerably even above 10000.
As a side effect this also reduces the net_bridge struct size from 3248
bytes to 1344 bytes. Also note that the key struct is 8 bytes.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
First, rename __inet_twsk_hashdance() to inet_twsk_hashdance()
Then, remove one inet_twsk_put() by setting tw_refcnt to 3 instead
of 4, but adding a fat warning that we do not have the right to access
tw anymore after inet_twsk_hashdance()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch enables tail loss probe in cwnd reduction (CWR) state
to detect potential losses. Prior to this patch, since the sender
uses PRR to determine the cwnd in CWR state, the combination of
CWR+PRR plus tcp_tso_should_defer() could cause unnecessary stalls
upon losses: PRR makes cwnd so gentle that tcp_tso_should_defer()
defers sending wait for more ACKs. The ACKs may not come due to
packet losses.
Disallowing TLP when there is unused cwnd had the primary effect
of disallowing TLP when there is TSO deferral, Nagle deferral,
or we hit the rwin limit. Because basically every application
write() or incoming ACK will cause us to run tcp_write_xmit()
to see if we can send more, and then if we sent something we call
tcp_schedule_loss_probe() to see if we should schedule a TLP. At
that point, there are a few common reasons why some cwnd budget
could still be unused:
(a) rwin limit
(b) nagle check
(c) TSO deferral
(d) TSQ
For (d), after the next packet tx completion the TSQ mechanism
will allow us to send more packets, so we don't really need a
TLP (in practice it shouldn't matter whether we schedule one
or not). But for (a), (b), (c) the sender won't send any more
packets until it gets another ACK. But if the whole flight was
lost, or all the ACKs were lost, then we won't get any more ACKs,
and ideally we should schedule and send a TLP to get more feedback.
In particular for a long time we have wanted some kind of timer for
TSO deferral, and at least this would give us some kind of timer
Reported-by: Steve Ibanez <sibanez@stanford.edu>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Nandita Dukkipati <nanditad@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Since we now hold RTNL lock in tc_action_net_exit(), it is good to
batch them to speedup tc action dismantle.
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Back in linux-3.13 (commit b0983d3c9b13 ("tcp: fix dynamic right sizing"))
I addressed the pressing issues we had with receiver autotuning.
But DRS suffers from extra latencies caused by rcv_rtt_est.rtt_us
drifts. One common problem happens during slow start, since the
apparent RTT measured by the receiver can be inflated by ~50%,
at the end of one packet train.
Also, a single drop can delay read() calls by one RTT, meaning
tcp_rcv_space_adjust() can be called one RTT too late.
By replacing the tri-modal heuristic with a continuous function,
we can offset the effects of not growing 'at the optimal time'.
The curve of the function matches prior behavior if the space
increased by 25% and 50% exactly.
Cost of added multiply/divide is small, considering a TCP flow
typically would run this part of the code few times in its life.
I tested this patch with 100 ms RTT / 1% loss link, 100 runs
of (netperf -l 5), and got an average throughput of 4600 Mbit
instead of 1700 Mbit.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Wei Wang <weiwan@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When using large tcp_rmem[2] values (I did tests with 500 MB),
I noticed overflows while computing rcvwin.
Lets fix this before the following patch.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Wei Wang <weiwan@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
While rcvbuf is properly clamped by tcp_rmem[2], rcvwin
is left to a potentially too big value.
It has no serious effect, since :
1) tcp_grow_window() has very strict checks.
2) window_clamp can be mangled by user space to any value anyway.
tcp_init_buffer_space() and companions use tcp_full_space(),
we use tcp_win_from_space() to avoid reloading sk->sk_rcvbuf
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Wei Wang <weiwan@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Unordered idata process is more complicated than unordered data:
- It has to add mid into sctp_stream_out to save the next mid value,
which is separated from ordered idata's.
- To support pd for unordered idata, another mid and pd_mode need to
be added to save the message id and pd state in sctp_stream_in.
- To make unordered idata reasm easier, it adds a new event queue
to save frags for idata.
The patch mostly adds the samilar reasm functions for unordered idata
as ordered idata's, and also adjusts some other codes on assign_mid,
abort_pd and ulpevent_data for idata.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
abort_pd is added as a member of sctp_stream_interleave, used to abort
partial delivery for data or idata, called in sctp_cmd_assoc_failed.
Since stream interleave allows to do partial delivery for each stream
at the same time, sctp_intl_abort_pd for idata would be very different
from the old function sctp_ulpq_abort_pd for data.
Note that sctp_ulpevent_make_pdapi will support per stream in this
patch by adding pdapi_stream and pdapi_seq in sctp_pdapi_event, as
described in section 6.1.7 of RFC6458.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
start_pd is added as a member of sctp_stream_interleave, used to
do partial_delivery for data or idata when datalen >= asoc->rwnd
in sctp_eat_data. The codes have been done in last patches, but
they need to be extracted into start_pd, so that it could be used
for SCTP_CMD_PART_DELIVER cmd as well.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
renege_events is added as a member of sctp_stream_interleave, used to
renege some old data or idata in reasm or lobby queue properly to free
some memory for the new data when there's memory stress.
It defines sctp_renege_events for idata, and leaves sctp_ulpq_renege
as it is for data.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
enqueue_event is added as a member of sctp_stream_interleave, used to
enqueue either data, idata or notification events into user socket rx
queue.
It replaces sctp_ulpq_tail_event used in the other places with
enqueue_event.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
ulpevent_data is added as a member of sctp_stream_interleave, used to
do the most process in ulpq, including to convert data or idata chunk
to event, reasm them in reasm queue and put them in lobby queue in
right order, and deliver them up to user sk rx queue.
This procedure is described in section 2.2.3 of RFC8260.
It adds most functions for idata here to do the similar process as
the old functions for data. But since the details are very different
between them, the old functions can not be reused for idata.
event->ssn and event->ppid settings are moved to ulpevent_data from
sctp_ulpevent_make_rcvmsg, so that sctp_ulpevent_make_rcvmsg could
work for both data and idata.
Note that mid is added in sctp_ulpevent for idata, __packed has to
be used for defining sctp_ulpevent, or it would exceeds the skb cb
that saves a sctp_ulpevent variable for ulp layer process.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
validate_data is added as a member of sctp_stream_interleave, used
to validate ssn/chunk type for data or mid (message id)/chunk type
for idata, called in sctp_eat_data.
If this check fails, an abort packet will be sent, as said in
section 2.2.3 of RFC8260.
It also adds the process for idata in rx path. As Marcelo pointed
out, there's no need to add event table for idata, but just share
chunk_event_table with data's. It would drop data chunk for idata
and drop idata chunk for data by calling validate_data in
sctp_eat_data.
As last patch did, it also replaces sizeof(struct sctp_data_chunk)
with sctp_datachk_len for rx path.
After this patch, the idata can be accepted and delivered to ulp
layer.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
assign_number is added as a member of sctp_stream_interleave, used
to assign ssn for data or mid (message id) for idata, called in
sctp_packet_append_data. sctp_chunk_assign_ssn is left as it is,
and sctp_chunk_assign_mid is added for sctp_stream_interleave_1.
This procedure is described in section 2.2.2 of RFC8260.
All sizeof(struct sctp_data_chunk) in tx path is replaced with
sctp_datachk_len, to make it right for idata as well. And also
adjust sctp_chunk_is_data for SCTP_CID_I_DATA.
After this patch, idata can be built and sent in tx path.
Note that if sp strm_interleave is set, it has to wait_connect in
sctp_sendmsg, as asoc intl_enable need to be known after 4 shake-
hands, to decide if it should use data or idata later. data and
idata can't be mixed to send in one asoc.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
To avoid hundreds of checks for the different process on I-DATA chunk,
struct sctp_stream_interleave is defined as a group of functions used
to replace the codes in some place where it needs to do different job
according to if the asoc intl_enabled is set.
With these ops, it only needs to initialize asoc->stream.si with
sctp_stream_interleave_0 for normal data if asoc intl_enable is 0,
or sctp_stream_interleave_1 for idata if asoc intl_enable is set in
sctp_stream_init.
After that, the members in asoc->stream.si can be used directly in
some special places without checking asoc intl_enable.
make_datafrag is the first member for sctp_stream_interleave, it's
used to make data or idata frags, called in sctp_datamsg_from_user.
The old function sctp_make_datafrag_empty needs to be adjust some
to fit in this ops.
Note that as idata and data chunks have different length, it also
defines data_chunk_len for sctp_stream_interleave to describe the
chunk size.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
sctp_idatahdr and sctp_idata_chunk are used to define and parse
I-DATA chunk format, and sctp_make_idata is a function to build
the chunk.
The I-DATA Chunk Format is defined in section 2.1 of RFC8260.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
asoc intl_enable will be set when local sp strm_interleave is set
and there's I-DATA chunk in init and init_ack extensions, as said
in section 2.2.1 of RFC8260.
asoc intl_enable indicates all data will be sent as I-DATA chunks.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds intl_enable in asoc and netns, and strm_interleave in
sctp_sock to indicate if stream interleave is enabled and supported.
netns intl_enable would be set via procfs, but that is not added yet
until all stream interleave codes are completely implemented; asoc
intl_enable will be set when doing 4-shakehands.
sp strm_interleave can be set by sockopt SCTP_INTERLEAVING_SUPPORTED
which is also added in this patch. This socket option is defined in
section 4.3.1 of RFC8260.
Note that strm_interleave can only be set by sockopt when both netns
intl_enable and sp frag_interleave are set.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Both netlink_add_tap() and netlink_remove_tap() are
called in process context, no need to bother spinlock.
Note, in fact, currently we always hold RTNL when calling
these two functions, so we don't need any other lock at
all, but keeping this lock doesn't harm anything.
Cc: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
nlmon device is not supposed to capture netlink events from
other netns, so instead of filtering events, we can simply
make netlink tap itself per netns.
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Kevin Cernekee <cernekee@chromium.org>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Most callers of rhashtable_walk_start don't care about a resize event
which is indicated by a return value of -EAGAIN. So calls to
rhashtable_walk_start are wrapped wih code to ignore -EAGAIN. Something
like this is common:
ret = rhashtable_walk_start(rhiter);
if (ret && ret != -EAGAIN)
goto out;
Since zero and -EAGAIN are the only possible return values from the
function this check is pointless. The condition never evaluates to true.
This patch changes rhashtable_walk_start to return void. This simplifies
code for the callers that ignore -EAGAIN. For the few cases where the
caller cares about the resize event, particularly where the table can be
walked in mulitple parts for netlink or seq file dump, the function
rhashtable_walk_start_check has been added that returns -EAGAIN on a
resize event.
Signed-off-by: Tom Herbert <tom@quantonium.net>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Fixes: 46e6b992c250 ("rtnetlink: allow GSO maximums to be set on device creation")
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Conflict was two parallel additions of include files to sch_generic.c,
no biggie.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Netlink device already allows changing GSO sizes with
ip set command. The part that is missing is allowing overriding
GSO settings on device creation.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
RACK skips an ACK unless it advances the most recently delivered
TX timestamp (rack.mstamp). Since RACK also uses the most recent
RTT to decide if a packet is lost, RACK should still run the
loss detection whenever the most recent RTT changes. For example,
an ACK that does not advance the timestamp but triggers the cwnd
undo due to reordering, would then use the most recent (higher)
RTT measurement to detect further losses.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Priyaranjan Jha <priyarjha@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
RACK should mark a packet lost when remaining wait time is zero.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Priyaranjan Jha <priyarjha@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When sender detects spurious retransmission, all packets
marked lost are remarked to be in-flight. However some may
be considered lost based on its timestamps in RACK. This patch
forces RACK to re-evaluate, which may be skipped previously if
the ACK does not advance RACK timestamp.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Priyaranjan Jha <priyarjha@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
RACK does not test the loss recovery state correctly to compute
the reordering window. It assumes if lost_out is zero then TCP is
not in loss recovery. But it can be zero during recovery before
calling tcp_rack_detect_loss(): when an ACK acknowledges all
packets marked lost before receiving this ACK, but has not yet
to discover new ones by tcp_rack_detect_loss(). The fix is to
simply test the congestion state directly.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Priyaranjan Jha <priyarjha@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Since the block is freed with last chain being put, once we reach the
end of iteration of list_for_each_entry_safe, the block may be
already freed. I'm hitting this only by creating and deleting clsact:
[ 202.171952] ==================================================================
[ 202.180182] BUG: KASAN: use-after-free in tcf_block_put_ext+0x240/0x390
[ 202.187590] Read of size 8 at addr ffff880225539a80 by task tc/796
[ 202.194508]
[ 202.196185] CPU: 0 PID: 796 Comm: tc Not tainted 4.15.0-rc2jiri+ #5
[ 202.203200] Hardware name: Mellanox Technologies Ltd. "MSN2100-CB2F"/"SA001017", BIOS 5.6.5 06/07/2016
[ 202.213613] Call Trace:
[ 202.216369] dump_stack+0xda/0x169
[ 202.220192] ? dma_virt_map_sg+0x147/0x147
[ 202.224790] ? show_regs_print_info+0x54/0x54
[ 202.229691] ? tcf_chain_destroy+0x1dc/0x250
[ 202.234494] print_address_description+0x83/0x3d0
[ 202.239781] ? tcf_block_put_ext+0x240/0x390
[ 202.244575] kasan_report+0x1ba/0x460
[ 202.248707] ? tcf_block_put_ext+0x240/0x390
[ 202.253518] tcf_block_put_ext+0x240/0x390
[ 202.258117] ? tcf_chain_flush+0x290/0x290
[ 202.262708] ? qdisc_hash_del+0x82/0x1a0
[ 202.267111] ? qdisc_hash_add+0x50/0x50
[ 202.271411] ? __lock_is_held+0x5f/0x1a0
[ 202.275843] clsact_destroy+0x3d/0x80 [sch_ingress]
[ 202.281323] qdisc_destroy+0xcb/0x240
[ 202.285445] qdisc_graft+0x216/0x7b0
[ 202.289497] tc_get_qdisc+0x260/0x560
Fix this by holding the block also by chain 0 and put chain 0
explicitly, out of the list_for_each_entry_safe loop at the very
end of tcf_block_put_ext.
Fixes: efbf78973978 ("net_sched: get rid of rcu_barrier() in tcf_block_put_ext()")
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This converts the pfifo_fast qdisc to use the skb_array data structure
and set the lockless qdisc bit. pfifo_fast is the first qdisc to support
the lockless bit that can be a child of a qdisc requiring locking. So
we add logic to clear the lock bit on initialization in these cases when
the qdisc graft operation occurs.
This also removes the logic used to pick the next band to dequeue from
and instead just checks a per priority array for packets from top priority
to lowest. This might need to be a bit more clever but seems to work
for now.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The sch_mqprio qdisc creates a sub-qdisc per tx queue which are then
called independently for enqueue and dequeue operations. However
statistics are aggregated and pushed up to the "master" qdisc.
This patch adds support for any of the sub-qdiscs to be per cpu
statistic qdiscs. To handle this case add a check when calculating
stats and aggregate the per cpu stats if needed.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The sch_mq qdisc creates a sub-qdisc per tx queue which are then
called independently for enqueue and dequeue operations. However
statistics are aggregated and pushed up to the "master" qdisc.
This patch adds support for any of the sub-qdiscs to be per cpu
statistic qdiscs. To handle this case add a check when calculating
stats and aggregate the per cpu stats if needed.
Also exports __gnet_stats_copy_queue() to use as a helper function.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add qdisc qlen helper routines for lockless qdiscs to use.
The qdisc qlen is no longer used in the hotpath but it is reported
via stats query on the qdisc so it still needs to be tracked. This
adds the per cpu operations needed along with a helper to return
the summation of per cpu stats.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
I can not think of any reason to pull the bad txq skb off the qdisc if
the txq we plan to send this on is still frozen. So check for frozen
queue first and abort before dequeuing either skb_bad_txq skb or
normal qdisc dequeue() skb.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Similar to how gso is handled use skb list for skb_bad_tx this is
required with lockless qdiscs because we may have multiple cores
attempting to push skbs into skb_bad_tx concurrently
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In qdisc_graft_qdisc a "new" qdisc is attached and the 'qdisc_destroy'
operation is called on the old qdisc. The destroy operation will wait
a rcu grace period and call qdisc_rcu_free(). At which point
gso_cpu_skb is free'd along with all stats so no need to zero stats
and gso_cpu_skb from the graft operation itself.
Further after dropping the qdisc locks we can not continue to call
qdisc_reset before waiting an rcu grace period so that the qdisc is
detached from all cpus. By removing the qdisc_reset() here we get
the correct property of waiting an rcu grace period and letting the
qdisc_destroy operation clean up the qdisc correctly.
Note, a refcnt greater than 1 would cause the destroy operation to
be aborted however if this ever happened the reference to the qdisc
would be lost and we would have a memory leak.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This work is preparing the qdisc layer to support egress lockless
qdiscs. If we are running the egress qdisc lockless in the case we
overrun the netdev, for whatever reason, the netdev returns a busy
error code and the skb is parked on the gso_skb pointer. With many
cores all hitting this case at once its possible to have multiple
sk_buffs here so we turn gso_skb into a queue.
This should be the edge case and if we see this frequently then
the netdev/qdisc layer needs to back off.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Enable dflt qdisc support for per cpu stats before this patch a dflt
qdisc was required to use the global statistics qstats and bstats.
This adds a static flags field to qdisc_ops that is propagated
into qdisc->flags in qdisc allocate call. This allows the allocation
block to completely allocate the qdisc object so we don't have
dangling allocations after qdisc init.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
sch_direct_xmit() uses qdisc_qlen as a return value but all call sites
of the routine only check if it is zero or not. Simplify the logic so
that we don't need to return an actual queue length value.
This introduces a case now where sch_direct_xmit would have returned
a qlen of zero but now it returns true. However in this case all
call sites of sch_direct_xmit will implement a dequeue() and get
a null skb and abort. This trades tracking qlen in the hotpath for
an extra dequeue operation. Overall this seems to be good for
performance.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds a flag for queueing disciplines to indicate the stack
does not need to use the qdisc lock to protect operations. This can
be used to build lockless scheduling algorithms and improving
performance.
The flag is checked in the tx path and the qdisc lock is only taken
if it is not set. For now use a conditional if statement. Later we
could be more aggressive if it proves worthwhile and use a static key
or wrap this in a likely().
Also the lockless case drops the TCQ_F_CAN_BYPASS logic. The reason
for this is synchronizing a qlen counter across threads proves to
cost more than doing the enqueue/dequeue operations when tested with
pktgen.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Currently __qdisc_run calls qdisc_run_end() but does not call
qdisc_run_begin(). This makes it hard to track pairs of
qdisc_run_{begin,end} across function calls.
To simplify reading these code paths this patch moves begin/end calls
into qdisc_run().
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Fix BBR so that upon notification of a loss recovery undo BBR resets
long-term bandwidth sampling.
Under high reordering, reordering events can be interpreted as loss.
If the reordering and spurious loss estimates are high enough, this
can cause BBR to spuriously estimate that we are seeing loss rates
high enough to trigger long-term bandwidth estimation. To avoid that
problem, this commit resets long-term bandwidth sampling on loss
recovery undo events.
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
William Tu
David S. Miller
Sean Wang
Yuchung Cheng
Willem de Bruijn
Nikolay Aleksandrov
Eric Dumazet
Neal Cardwell
Cong Wang
Xin Long
Tom Herbert
Stephen Hemminger
Jiri Pirko
John Fastabend