Fixes the following behavior: for connections that had no RTT sample
at the time of initializing congestion control, BBR was initializing
the pacing rate to a high nominal rate (based an a guess of RTT=1ms,
in case this is LAN traffic). Then BBR never adjusted the pacing rate
downward upon obtaining an actual RTT sample, if the connection never
filled the pipe (e.g. all sends were small app-limited writes()).
This fix adjusts the pacing rate upon obtaining the first RTT sample.
Fixes: 0f8782ea14 ("tcp_bbr: add BBR congestion control")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix a corner case noticed by Eric Dumazet, where BBR's setting
sk->sk_pacing_rate to 0 during initialization could theoretically
cause packets in the sending host to hang if there were packets "in
flight" in the pacing infrastructure at the time the BBR congestion
control state is initialized. This could occur if the pacing
infrastructure happened to race with bbr_init() in a way such that the
pacer read the 0 rather than the immediately following non-zero pacing
rate.
Fixes: 0f8782ea14 ("tcp_bbr: add BBR congestion control")
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce a helper to initialize the BBR pacing rate unconditionally,
based on the current cwnd and RTT estimate. This is a pure refactor,
but is needed for two following fixes.
Fixes: 0f8782ea14 ("tcp_bbr: add BBR congestion control")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce a helper to convert a BBR bandwidth and gain factor to a
pacing rate in bytes per second. This is a pure refactor, but is
needed for two following fixes.
Fixes: 0f8782ea14 ("tcp_bbr: add BBR congestion control")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In bbr_set_pacing_rate(), which decides whether to cut the pacing
rate, there was some code that considered exiting STARTUP to be
equivalent to the notion of filling the pipe (i.e.,
bbr_full_bw_reached()). Specifically, as the code was structured,
exiting STARTUP and going into PROBE_RTT could cause us to cut the
pacing rate down to something silly and low, based on whatever
bandwidth samples we've had so far, when it's possible that all of
them have been small app-limited bandwidth samples that are not
representative of the bandwidth available in the path. (The code was
correct at the time it was written, but the state machine changed
without this spot being adjusted correspondingly.)
Fixes: 0f8782ea14 ("tcp_bbr: add BBR congestion control")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When there is an established connection in direction A->B, it is
possible to receive a packet on port B which then executes
ct(commit,force) without first performing ct() - ie, a lookup.
In this case, we would expect that this packet can delete the existing
entry so that we can commit a connection with direction B->A. However,
currently we only perform a check in skb_nfct_cached() for whether
OVS_CS_F_TRACKED is set and OVS_CS_F_INVALID is not set, ie that a
lookup previously occurred. In the above scenario, a lookup has not
occurred but we should still be able to statelessly look up the
existing entry and potentially delete the entry if it is in the
opposite direction.
This patch extends the check to also hint that if the action has the
force flag set, then we will lookup the existing entry so that the
force check at the end of skb_nfct_cached has the ability to delete
the connection.
Fixes: dd41d330b03 ("openvswitch: Add force commit.")
CC: Pravin Shelar <pshelar@nicira.com>
CC: dev@openvswitch.org
Signed-off-by: Joe Stringer <joe@ovn.org>
Signed-off-by: Greg Rose <gvrose8192@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some time ago David Woodhouse reported skb_under_panic
when we try to push ethernet header to fragmented ipv6 skbs.
It was fixed for ipv6 by Florian Westphal in
commit 1d325d217c ("ipv6: ip6_fragment: fix headroom tests and skb leak")
However similar problem still exist in ipv4.
It does not trigger skb_under_panic due paranoid check
in ip_finish_output2, however according to Alexey Kuznetsov
current state is abnormal and ip_fragment should be fixed too.
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Marcelo noticed an array overflow caused by commit c28445c3cb
("sctp: add reconf_enable in asoc ep and netns"), in which sctp
would add SCTP_CID_RECONF into extensions when reconf_enable is
set in sctp_make_init and sctp_make_init_ack.
Then now when all ext chunks are set, 4 ext chunk ids can be put
into extensions array while extensions array size is 3. It would
cause a kernel panic because of this overflow.
This patch is to fix it by defining extensions array size is 4 in
both sctp_make_init and sctp_make_init_ack.
Fixes: c28445c3cb ("sctp: add reconf_enable in asoc ep and netns")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make name consistent with other TC event notification routines, such as
tcf_add_notify() and tcf_del_notify()
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When PACKET_QDISC_BYPASS is not used, Tx queue selection will be done
before the packet is enqueued, taking into account any mappings set by
a queuing discipline such as mqprio without hardware offloading. This
selection may be affected by a previously saved queue_mapping, either on
the Rx path, or done before the packet reaches the device, as it's
currently the case for AF_PACKET.
In order for queue selection to work as expected when using traffic
control, there can't be another selection done before that point is
reached, so move the call to packet_pick_tx_queue to
packet_direct_xmit, leaving the default xmit path as it was before
PACKET_QDISC_BYPASS was introduced.
A forward declaration of packet_pick_tx_queue() is introduced to avoid
the need to reorder the functions within the file.
Fixes: d346a3fae3 ("packet: introduce PACKET_QDISC_BYPASS socket option")
Signed-off-by: Iván Briano <ivan.briano@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With 802.1ad support the vlan_ingress code started checking for vlan
protocol mismatch which causes the current tag to be inserted and the
bridge vlan protocol & pvid to be set. The vlan tag insertion changes
the skb mac_header and thus the lookup mac dest pointer which was loaded
prior to calling br_allowed_ingress in br_handle_frame_finish is VLAN_HLEN
bytes off now, pointing to the last two bytes of the destination mac and
the first four of the source mac causing lookups to always fail and
broadcasting all such packets to all ports. Same thing happens for locally
originated packets when passing via br_dev_xmit. So load the dest pointer
after the vlan checks and possible skb change.
Fixes: 8580e2117c ("bridge: Prepare for 802.1ad vlan filtering support")
Reported-by: Anitha Narasimha Murthy <anitha@cumulusnetworks.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Acked-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
When we convert atomic_t to refcount_t, a new kernel warning
on "increment on 0" is introduced in the netpoll code,
zap_completion_queue(). In fact for this special case, we know
the refcount is 0 and we just have to set it to 1 to satisfy
the following dev_kfree_skb_any(), so we can just use
refcount_set(..., 1) instead.
Fixes: 633547973f ("net: convert sk_buff.users from atomic_t to refcount_t")
Reported-by: Dave Jones <davej@codemonkey.org.uk>
Cc: Reshetova, Elena <elena.reshetova@intel.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The configure callback of fib_rules_ops can change the refcnt of a
fib rule. For instance, mlxsw takes a refcnt when adding the processing
of the rule to a work queue. Thus the rule refcnt can not be reset to
to 1 afterwards. Move the refcnt setting to after the allocation.
Fixes: 5361e209dd ("net: avoid one splat in fib_nl_delrule()")
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Don't populate array error_code on the stack but make it static. Makes
the object code smaller by almost 250 bytes:
Before:
text data bss dec hex filename
10366 983 0 11349 2c55 net/dccp/input.o
After:
text data bss dec hex filename
10161 1039 0 11200 2bc0 net/dccp/input.o
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The bpf_skb_adjust_net() ignores the return value of bpf_skb_net_shrink/grow,
and always return 0, fix it by return 'ret'.
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking fixes from David Miller:
1) Fix 64-bit division in mlx5 IPSEC offload support, from Ilan Tayari
and Arnd Bergmann.
2) Fix race in statistics gathering in bnxt_en driver, from Michael
Chan.
3) Can't use a mutex in RCU reader protected section on tap driver, from
Cong WANG.
4) Fix mdb leak in bridging code, from Eduardo Valentin.
5) Fix free of wrong pointer variable in nfp driver, from Dan Carpenter.
6) Buffer overflow in brcmfmac driver, from Arend van SPriel.
7) ioremap_nocache() return value needs to be checked in smsc911x
driver, from Alexey Khoroshilov.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (34 commits)
net: stmmac: revert "support future possible different internal phy mode"
sfc: don't read beyond unicast address list
datagram: fix kernel-doc comments
socket: add documentation for missing elements
smsc911x: Add check for ioremap_nocache() return code
brcmfmac: fix possible buffer overflow in brcmf_cfg80211_mgmt_tx()
net: hns: Bugfix for Tx timeout handling in hns driver
net: ipmr: ipmr_get_table() returns NULL
nfp: freeing the wrong variable
mlxsw: spectrum_switchdev: Check status of memory allocation
mlxsw: spectrum_switchdev: Remove unused variable
mlxsw: spectrum_router: Fix use-after-free in route replace
mlxsw: spectrum_router: Add missing rollback
samples/bpf: fix a build issue
bridge: mdb: fix leak on complete_info ptr on fail path
tap: convert a mutex to a spinlock
cxgb4: fix BUG() on interrupt deallocating path of ULD
qed: Fix printk option passed when printing ipv6 addresses
net: Fix minor code bug in timestamping.txt
net: stmmac: Make 'alloc_dma_[rt]x_desc_resources()' look even closer
...
An underscore in the kernel-doc comment section has special meaning
and mis-use generates an errors.
./net/core/datagram.c:207: ERROR: Unknown target name: "msg".
./net/core/datagram.c:379: ERROR: Unknown target name: "msg".
./net/core/datagram.c:816: ERROR: Unknown target name: "t".
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ipmr_get_table() function doesn't return error pointers it returns
NULL on error.
Fixes: 4f75ba6982 ("net: ipmr: Add ipmr_rtm_getroute")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RESEND_ON_SPLIT, RADOS_BACKOFF, OSDMAP_PG_UPMAP and CRUSH_CHOOSE_ARGS
feature bits, and various other changes in the RADOS client protocol.
On top of that we have a new fsc mount option to allow supplying
fscache uniquifier (similar to NFS) and the usual pile of filesystem
fixes from Zheng.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iQEcBAABCAAGBQJZZQT+AAoJEEp/3jgCEfOLSsMH/i8ZdSzp7ocX00oLMlIxzFEk
5BUXZ086mEPAE4fjJFPO7+qYk6y26MzAhJL+bj8r5E0GvBEpQkoAoSQZ19Mj5ApC
nZnllzQ2C8kYvM4hp4Z2pLrF/OYACj/WJJgbTxubBET1zRq1iPj4EgbzBEraPvma
K76W9ILKNUjIoSDlNR5qvykXXfvi2dxRpi/8nvfMCOcjlw/7orjXVLa05fKmmOoX
OvpOjicWOrc8NlacGK+j1j1aaKlmLvZb9Ff+45hfC/L5PPQblM0dypFCVfq3MFFq
nUxKgTCAQDPrndzCdURCtdovjFKbskRGKmhnd0EZkdDCcnUmg6nLxqta6g2Dbs0=
=ioKM
-----END PGP SIGNATURE-----
Merge tag 'ceph-for-4.13-rc1' of git://github.com/ceph/ceph-client
Pull ceph updates from Ilya Dryomov:
"The main item here is support for v12.y.z ("Luminous") clusters:
RESEND_ON_SPLIT, RADOS_BACKOFF, OSDMAP_PG_UPMAP and CRUSH_CHOOSE_ARGS
feature bits, and various other changes in the RADOS client protocol.
On top of that we have a new fsc mount option to allow supplying
fscache uniquifier (similar to NFS) and the usual pile of filesystem
fixes from Zheng"
* tag 'ceph-for-4.13-rc1' of git://github.com/ceph/ceph-client: (44 commits)
libceph: advertise support for NEW_OSDOP_ENCODING and SERVER_LUMINOUS
libceph: osd_state is 32 bits wide in luminous
crush: remove an obsolete comment
crush: crush_init_workspace starts with struct crush_work
libceph, crush: per-pool crush_choose_arg_map for crush_do_rule()
crush: implement weight and id overrides for straw2
libceph: apply_upmap()
libceph: compute actual pgid in ceph_pg_to_up_acting_osds()
libceph: pg_upmap[_items] infrastructure
libceph: ceph_decode_skip_* helpers
libceph: kill __{insert,lookup,remove}_pg_mapping()
libceph: introduce and switch to decode_pg_mapping()
libceph: don't pass pgid by value
libceph: respect RADOS_BACKOFF backoffs
libceph: make DEFINE_RB_* helpers more general
libceph: avoid unnecessary pi lookups in calc_target()
libceph: use target pi for calc_target() calculations
libceph: always populate t->target_{oid,oloc} in calc_target()
libceph: make sure need_resend targets reflect latest map
libceph: delete from need_resend_linger before check_linger_pool_dne()
...
Fix the below warning generated by static checker:
net/mpls/af_mpls.c:2111 mpls_getroute()
error: uninitialized symbol 'in_label'."
Fixes: 397fc9e5ce ("mpls: route get support")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As Hongjun/Nicolas summarized in their original patch:
"
When a device changes from one netns to another, it's first unregistered,
then the netns reference is updated and the dev is registered in the new
netns. Thus, when a slave moves to another netns, it is first
unregistered. This triggers a NETDEV_UNREGISTER event which is caught by
the bonding driver. The driver calls bond_release(), which calls
dev_set_mtu() and thus triggers NETDEV_CHANGEMTU (the device is still in
the old netns).
"
This is a very special case, because the device is being unregistered
no one should still care about the NETDEV_CHANGEMTU event triggered
at this point, we can avoid broadcasting this event on this path,
and avoid touching inetdev_event()/addrconf_notify() path.
It requires to export __dev_set_mtu() to bonding driver.
Reported-by: Hongjun Li <hongjun.li@6wind.com>
Reported-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Cc: Jay Vosburgh <j.vosburgh@gmail.com>
Cc: Veaceslav Falico <vfalico@gmail.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are two problems with calling sock_create_kern() from
rds_tcp_accept_one()
1. it sets up a new_sock->sk that is wasteful, because this ->sk
is going to get replaced by inet_accept() in the subsequent ->accept()
2. The new_sock->sk is a leaked reference in sock_graft() which
expects to find a null parent->sk
Avoid these problems by calling sock_create_lite().
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is not just a pointer to crush_work, it is the whole structure.
That is not a problem since it only contains a pointer. But it will
be a problem if new data members are added to crush_work.
Reflects ceph.git commit ee957dd431bfbeb6dadaf77764db8e0757417328.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If there is no crush_choose_arg_map for a given pool, a NULL pointer is
passed to preserve existing crush_do_rule() behavior.
Reflects ceph.git commits 55fb91d64071552ea1bc65ab4ea84d3c8b73ab4b,
dbe36e08be00c6519a8c89718dd47b0219c20516.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
bucket_straw2_choose needs to use weights that may be different from
weight_items. For instance to compensate for an uneven distribution
caused by a low number of values. Or to fix the probability biais
introduced by conditional probabilities (see
http://tracker.ceph.com/issues/15653 for more information).
We introduce a weight_set for each straw2 bucket to set the desired
weight for a given item at a given position. The weight of a given item
when picking the first replica (first position) may be different from
the weight the second replica (second position). For instance the weight
matrix for a given bucket containing items 3, 7 and 13 could be as
follows:
position 0 position 1
item 3 0x10000 0x100000
item 7 0x40000 0x10000
item 13 0x40000 0x10000
When crush_do_rule picks the first of two replicas (position 0), item 7,
3 are four times more likely to be choosen by bucket_straw2_choose than
item 13. When choosing the second replica (position 1), item 3 is ten
times more likely to be choosen than item 7, 13.
By default the weight_set of each bucket exactly matches the content of
item_weights for each position to ensure backward compatibility.
bucket_straw2_choose compares items by using their id. The same ids are
also used to index buckets and they must be unique. For each item in a
bucket an array of ids can be provided for placement purposes and they
are used instead of the ids. If no replacement ids are provided, the
legacy behavior is preserved.
Reflects ceph.git commit 19537a450fd5c5a0bb8b7830947507a76db2ceca.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Previously, pg_to_raw_osds() didn't filter for existent OSDs because
raw_to_up_osds() would filter for "up" ("up" is predicated on "exists")
and raw_to_up_osds() was called directly after pg_to_raw_osds(). Now,
with apply_upmap() call in there, nonexistent OSDs in pg_to_raw_osds()
output can affect apply_upmap(). Introduce remove_nonexistent_osds()
to deal with that.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Move raw_pg_to_pg() call out of get_temp_osds() and into
ceph_pg_to_up_acting_osds(), for upcoming apply_upmap().
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
pg_temp and pg_upmap encodings are the same (PG -> array of osds),
except for the incremental remove: it's an empty mapping in new_pg_temp
for pg_temp and a separate old_pg_upmap set for pg_upmap. (This isn't
to allow for empty pg_upmap mappings -- apparently, pg_temp just wasn't
looked at as an example for pg_upmap encoding.)
Reuse __decode_pg_temp() for decoding pg_upmap and new_pg_upmap.
__decode_pg_temp() stores into pg_temp union member, but since pg_upmap
union member is identical, reading through pg_upmap later is OK.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Some of these won't be as efficient as they could be (e.g.
ceph_decode_skip_set(... 32 ...) could advance by len * sizeof(u32)
once instead of advancing by sizeof(u32) len times), but that's fine
and not worth a bunch of extra macro code.
Replace skip_name_map() with ceph_decode_skip_map as an example.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Make __{lookup,remove}_pg_mapping() look like their ceph_spg_mapping
counterparts: take const struct ceph_pg *.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
For luminous and beyond we are encoding the actual spgid, which
requires operating with the correct pg_num, i.e. that of the target
pool.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
need_check_tiering logic doesn't make a whole lot of sense. Drop it
and apply tiering unconditionally on every calc_target() call instead.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Otherwise we may miss events like PG splits, pool deletions, etc when
we get multiple incremental maps at once. Because check_pool_dne() can
now be fed an unlinked request, finish_request() needed to be taught to
handle unlinked requests.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When processing a map update consisting of multiple incrementals, we
may end up running check_linger_pool_dne() on a lingering request that
was previously added to need_resend_linger list. If it is concluded
that the target pool doesn't exist, the request is killed off while
still on need_resend_linger list, which leads to a crash on a NULL
lreq->osd in kick_requests():
libceph: linger_id 18446462598732840961 pool does not exist
BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
IP: ceph_osdc_handle_map+0x4ae/0x870
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Replace it with more fine-grained bools to separate updating
ceph_osd_request_target fields and the decision to resend.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Give upper layers a chance to reencode the message after the connection
is negotiated and ->peer_features is set. OSD client will use this to
support both luminous and pre-luminous OSDs (in a single cluster): the
former need MOSDOp v8; the latter will continue to be sent MOSDOp v4.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The old (v15) pi->last_force_request_resend has been repurposed to
make pre-RESEND_ON_SPLIT clients that don't check for PG splits but do
obey pi->last_force_request_resend resend on splits. See ceph.git
commit 189ca7ec6420 ("mon/OSDMonitor: make pre-luminous clients resend
ops on split").
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
