RDS over IB does not use multipath RDS, so the array
of additional rds_conn_path structures is always superfluous
in this case. Reduce the memory footprint of the rds module
by making this a dynamic allocation predicated on whether
the transport is mp_capable.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Tested-by: Efrain Galaviz <efrain.galaviz@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We could end up executing rds_conn_shutdown before the rds_recv_worker
thread, then rds_conn_shutdown -> rds_tcp_conn_shutdown can do a
sock_release and set sock->sk to null, which may interleave in bad
ways with rds_recv_worker, e.g., it could result in:
"BUG: unable to handle kernel NULL pointer dereference at 0000000000000078"
[ffff881769f6fd70] release_sock at ffffffff815f337b
[ffff881769f6fd90] rds_tcp_recv at ffffffffa043c888 [rds_tcp]
[ffff881769f6fdb0] rds_recv_worker at ffffffffa04a4810 [rds]
[ffff881769f6fde0] process_one_work at ffffffff810a14c1
[ffff881769f6fe40] worker_thread at ffffffff810a1940
[ffff881769f6fec0] kthread at ffffffff810a6b1e
Also, do not enqueue any new shutdown workq items when the connection is
shutting down (this may happen for rds-tcp in softirq mode, if a FIN
or CLOSE is received while the modules is in the middle of an unload)
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
refcount_t type and corresponding API should be
used instead of atomic_t when the variable is used as
a reference counter. This allows to avoid accidental
refcounter overflows that might lead to use-after-free
situations.
Signed-off-by: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Hans Liljestrand <ishkamiel@gmail.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David Windsor <dwindsor@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
refcount_t type and corresponding API should be
used instead of atomic_t when the variable is used as
a reference counter. This allows to avoid accidental
refcounter overflows that might lead to use-after-free
situations.
Signed-off-by: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Hans Liljestrand <ishkamiel@gmail.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David Windsor <dwindsor@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
refcount_t type and corresponding API should be
used instead of atomic_t when the variable is used as
a reference counter. This allows to avoid accidental
refcounter overflows that might lead to use-after-free
situations.
Signed-off-by: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Hans Liljestrand <ishkamiel@gmail.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David Windsor <dwindsor@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If we are unloading the rds_tcp module, we can set linger to 1
and drop pending packets to accelerate reconnect. The peer will
end up resetting the connection based on new generation numbers
of the new incarnation, so hanging on to unsent TCP packets via
linger is mostly pointless in this case.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Tested-by: Jenny Xu <jenny.x.xu@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RDS handshake ping probe added by commit 5916e2c155
("RDS: TCP: Enable multipath RDS for TCP") is sent from rds_sendmsg()
before the first data packet is sent to a peer. If the conversation
is not bidirectional (i.e., one side is always passive and never
invokes rds_sendmsg()) and the passive side restarts its rds_tcp
module, a new HS ping probe needs to be sent, so that the number
of paths can be re-established.
This patch achieves that by sending a HS ping probe from
rds_tcp_accept_one() when c_npaths is 0 (i.e., we have not done
a handshake probe with this peer yet).
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Tested-by: Jenny Xu <jenny.x.xu@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Found when testing between sparc and x86 machines on different
subnets, so the address comparison patterns hit the corner cases and
brought out some bugs fixed by this patch.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Tested-by: Imanti Mendez <imanti.mendez@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After commit 1a0e100fb2 ("RDS: TCP: Force every connection to be
initiated by numerically smaller IP address") we no longer need
the logic associated with cp_outgoing, so clean up usage of this
field.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Tested-by: Imanti Mendez <imanti.mendez@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is incorrect for the rds_connection to piggyback on the
sock_net() refcount for the netns because this gives rise to
a chicken-and-egg problem during rds_conn_destroy. Instead explicitly
take a ref on the net, and hold the netns down till the connection
tear-down is complete.
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking fixes from David Miller:
1) Fix double-free in batman-adv, from Sven Eckelmann.
2) Fix packet stats for fast-RX path, from Joannes Berg.
3) Netfilter's ip_route_me_harder() doesn't handle request sockets
properly, fix from Florian Westphal.
4) Fix sendmsg deadlock in rxrpc, from David Howells.
5) Add missing RCU locking to transport hashtable scan, from Xin Long.
6) Fix potential packet loss in mlxsw driver, from Ido Schimmel.
7) Fix race in NAPI handling between poll handlers and busy polling,
from Eric Dumazet.
8) TX path in vxlan and geneve need proper RCU locking, from Jakub
Kicinski.
9) SYN processing in DCCP and TCP need to disable BH, from Eric
Dumazet.
10) Properly handle net_enable_timestamp() being invoked from IRQ
context, also from Eric Dumazet.
11) Fix crash on device-tree systems in xgene driver, from Alban Bedel.
12) Do not call sk_free() on a locked socket, from Arnaldo Carvalho de
Melo.
13) Fix use-after-free in netvsc driver, from Dexuan Cui.
14) Fix max MTU setting in bonding driver, from WANG Cong.
15) xen-netback hash table can be allocated from softirq context, so use
GFP_ATOMIC. From Anoob Soman.
16) Fix MAC address change bug in bgmac driver, from Hari Vyas.
17) strparser needs to destroy strp_wq on module exit, from WANG Cong.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (69 commits)
strparser: destroy workqueue on module exit
sfc: fix IPID endianness in TSOv2
sfc: avoid max() in array size
rds: remove unnecessary returned value check
rxrpc: Fix potential NULL-pointer exception
nfp: correct DMA direction in XDP DMA sync
nfp: don't tell FW about the reserved buffer space
net: ethernet: bgmac: mac address change bug
net: ethernet: bgmac: init sequence bug
xen-netback: don't vfree() queues under spinlock
xen-netback: keep a local pointer for vif in backend_disconnect()
netfilter: nf_tables: don't call nfnetlink_set_err() if nfnetlink_send() fails
netfilter: nft_set_rbtree: incorrect assumption on lower interval lookups
netfilter: nf_conntrack_sip: fix wrong memory initialisation
can: flexcan: fix typo in comment
can: usb_8dev: Fix memory leak of priv->cmd_msg_buffer
can: gs_usb: fix coding style
can: gs_usb: Don't use stack memory for USB transfers
ixgbe: Limit use of 2K buffers on architectures with 256B or larger cache lines
ixgbe: update the rss key on h/w, when ethtool ask for it
...
The function rds_trans_register always returns 0. As such, it is not
necessary to check the returned value.
Cc: Joe Jin <joe.jin@oracle.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull vfs sendmsg updates from Al Viro:
"More sendmsg work.
This is a fairly separate isolated stuff (there's a continuation
around lustre, but that one was too late to soak in -next), thus the
separate pull request"
* 'work.sendmsg' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
ncpfs: switch to sock_sendmsg()
ncpfs: don't mess with manually advancing iovec on send
ncpfs: sendmsg does *not* bugger iovec these days
ceph_tcp_sendpage(): use ITER_BVEC sendmsg
afs_send_pages(): use ITER_BVEC
rds: remove dead code
ceph: switch to sock_recvmsg()
usbip_recv(): switch to sock_recvmsg()
iscsi_target: deal with short writes on the tx side
[nbd] pass iov_iter to nbd_xmit()
[nbd] switch sock_xmit() to sock_{send,recv}msg()
[drbd] use sock_sendmsg()
Socket option to tap receive path latency in various stages
in nano seconds. It can be enabled on selective sockets using
using SO_RDS_MSG_RXPATH_LATENCY socket option. RDS will return
the data to application with RDS_CMSG_RXPATH_LATENCY in defined
format. Scope is left to add more trace points for future
without need of change in the interface.
Reviewed-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
RDS support max message size as 1M but the code doesn't check this
in all cases. Patch fixes it for RDMA & non-RDMA and RDS MR size
and its enforced irrespective of underlying transport.
Signed-off-by: Avinash Repaka <avinash.repaka@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Tracks the receive side memory added to scokets and removed from sockets.
Signed-off-by: Venkat Venkatsubra <venkat.x.venkatsubra@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
When application sends an RDS RDMA composite message consist of
RDMA transfer to be followed up by non RDMA payload, it expect to
be notified *only* when the full message gets delivered. RDS RDMA
notification doesn't behave this way though.
Thanks to Venkat for debug and root casuing the issue
where only first part of the message(RDMA) was
successfully delivered but remainder payload delivery failed.
In that case, application should not be notified with
a false positive of message delivery success.
Fix this case by making sure the user gets notified only after
the full message delivery.
Reviewed-by: Venkat Venkatsubra <venkat.x.venkatsubra@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
The RDS transport has to be able to distinguish between
two types of failure events:
(a) when the transport fails (e.g., TCP connection reset)
but the RDS socket/connection layer on both sides stays
the same
(b) when the peer's RDS layer itself resets (e.g., due to module
reload or machine reboot at the peer)
In case (a) both sides must reconnect and continue the RDS messaging
without any message loss or disruption to the message sequence numbers,
and this is achieved by rds_send_path_reset().
In case (b) we should reset all rds_connection state to the
new incarnation of the peer. Examples of state that needs to
be reset are next expected rx sequence number from, or messages to be
retransmitted to, the new incarnation of the peer.
To achieve this, the RDS handshake probe added as part of
commit 5916e2c155 ("RDS: TCP: Enable multipath RDS for TCP")
is enhanced so that sender and receiver of the RDS ping-probe
will add a generation number as part of the RDS_EXTHDR_GEN_NUM
extension header. Each peer stores local and remote generation
numbers as part of each rds_connection. Changes in generation
number will be detected via incoming handshake probe ping
request or response and will allow the receiver to reset rds_connection
state.
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>
Mostly simple overlapping changes.
For example, David Ahern's adjacency list revamp in 'net-next'
conflicted with an adjacency list traversal bug fix in 'net'.
Signed-off-by: David S. Miller <davem@davemloft.net>
rds use Kconfig option called "RDS_DEBUG" to enable rds debug messages.
This option cause the rds Makefile to add -DDEBUG to the rds gcc command
line.
When CONFIG_DYNAMIC_DEBUG is enabled, the "DEBUG" macro is used by
include/linux/dynamic_debug.h to decide if dynamic debug prints should
be sent by default to the kernel log.
rds should not enable this macro for production builds. rds dynamic
debug work as expected follow this fix.
Signed-off-by: Shamir Rabinovitch <shamir.rabinovitch@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Reviewed-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This macro's last use was removed in commit d769ef81d5
("RDS: Update rds_conn_shutdown to work with rds_conn_path")
so make the macro and the __rds_conn_error function definition
and declaration disappear.
Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is helpful to detect at compile-time errors related to format
strings.
Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use RDS probe-ping to compute how many paths may be used with
the peer, and to synchronously start the multiple paths. If mprds is
supported, hash outgoing traffic to one of multiple paths in rds_sendmsg()
when multipath RDS is supported by the transport.
CC: Santosh Shilimkar <santosh.shilimkar@oracle.com>
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>
This patch adds ->conn_path_connect callbacks in the rds_transport
that are used to set up a single connection path.
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ->sk_user_data contains a pointer to the rds_conn_path
for the socket. Use this consistently in the rds_tcp_data_ready
callbacks to get the rds_conn_path for rds_recv_incoming.
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Refactor code to avoid separate indirections for single-path
and multipath transports. All transports (both single and mp-capable)
will get a pointer to the rds_conn_path, and can trivially derive
the rds_connection from the ->cp_conn.
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit changes rds_conn_shutdown to take a rds_conn_path *
argument, allowing it to shutdown paths other than c_path[0] for
MP-capable transports.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rds_conn_path_error() is the MP-aware analog of rds_conn_error,
to be used by multipath-capable callers.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rds_conn_path_connect_if_down() works on the rds_conn_path
that it is passed. Callers who are not t_m_capable may continue
calling rds_conn_connect_if_down, which will invoke
rds_conn_path_connect_if_down() with the default c_path[0].
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit allows rds_send_pong() callers to send back
the rds pong message on some path other than c_path[0] by
passing in a struct rds_conn_path * argument. It also
removes the last dependency on the #defines in rds_single.h
from send.c
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pass a struct rds_conn_path to rds_send_xmit so that MP capable
transports can transmit packets on something other than c_path[0].
The eventual goal for MP capable transports is to hash the rds
socket to a path based on the bound local address/port, and use
this path as the argument to rds_send_xmit()
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The only caller of rds_send_get_message() was
rds_iw_send_cq_comp_handler() which was removed as part of
commit dcdede0406 ("RDS: Drop stale iWARP RDMA transport"),
so remove rds_send_get_message() for the same reason.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rds_send_path_drop_acked() is the path-specific version of
rds_send_drop_acked() to be invoked by MP capable callers.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
t_mp_capable transports can use rds_inc_path_init to initialize
all fields in struct rds_incoming, including the i_conn_path.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Transports that are t_mp_capable should set the rds_conn_path
on which the datagram was recived in the ->i_conn_path field
of struct rds_incoming.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The t_mp_capable bit will be used in the core rds module
to support multipathing logic when the transport supports it.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In preparation for multipath RDS, split the rds_connection
structure into a base structure, and a per-path struct rds_conn_path.
The base structure tracks information and locks common to all
paths. The workqs for send/recv/shutdown etc are tracked per
rds_conn_path. Thus the workq callbacks now work with rds_conn_path.
This commit allows for one rds_conn_path per rds_connection, and will
be extended into multiple conn_paths in subsequent commits.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The send path needs to be quiesced before resetting callbacks from
rds_tcp_accept_one(), and commit eb19284026 ("RDS:TCP: Synchronize
rds_tcp_accept_one with rds_send_xmit when resetting t_sock") achieves
this using the c_state and RDS_IN_XMIT bit following the pattern
used by rds_conn_shutdown(). However this leaves the possibility
of a race window as shown in the sequence below
take t_conn_lock in rds_tcp_conn_connect
send outgoing syn to peer
drop t_conn_lock in rds_tcp_conn_connect
incoming from peer triggers rds_tcp_accept_one, conn is
marked CONNECTING
wait for RDS_IN_XMIT to quiesce any rds_send_xmit threads
call rds_tcp_reset_callbacks
[.. race-window where incoming syn-ack can cause the conn
to be marked UP from rds_tcp_state_change ..]
lock_sock called from rds_tcp_reset_callbacks, and we set
t_sock to null
As soon as the conn is marked UP in the race-window above, rds_send_xmit()
threads will proceed to rds_tcp_xmit and may encounter a null-pointer
deref on the t_sock.
Given that rds_tcp_state_change() is invoked in softirq context, whereas
rds_tcp_reset_callbacks() is in workq context, and testing for RDS_IN_XMIT
after lock_sock could result in a deadlock with tcp_sendmsg, this
commit fixes the race by using a new c_state, RDS_TCP_RESETTING, which
will prevent a transition to RDS_CONN_UP from rds_tcp_state_change().
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>
The SO_TIMESTAMP generates time stamp for each incoming RDS messages
User app can enable it by using SO_TIMESTAMP setsocketopt() at
SOL_SOCKET level. CMSG data of cmsg type SO_TIMESTAMP contains the
time stamp in struct timeval format.
Reviewed-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To further improve the RDS connection scalabilty on massive systems
where number of sockets grows into tens of thousands of sockets, there
is a need of larger bind hashtable. Pre-allocated 8K or 16K table is
not very flexible in terms of memory utilisation. The rhashtable
infrastructure gives us the flexibility to grow the hashtbable based
on use and also comes up with inbuilt efficient bucket(chain) handling.
Reviewed-by: David Miller <davem@davemloft.net>
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Santosh Shilimkar says:
====================
RDS: connection scalability and performance improvements
[v4]
Re-sending the same patches from v3 again since my repost of
patch 05/14 from v3 was whitespace damaged.
[v3]
Updated patch "[PATCH v2 05/14] RDS: defer the over_batch work to
send worker" as per David Miller's comment [4] to avoid the magic
value usage. Patch now makes use of already available but unused
send_batch_count module parameter. Rest of the patches are same as
earlier version v2 [3]
[v2]:
Dropped "[PATCH 05/15] RDS: increase size of hash-table to 8K" from
earlier version [1]. I plan to address the hash table scalability using
re-sizable hash tables as suggested by David Laight and David Miller [2]
This series addresses RDS connection bottlenecks on massive workloads and
improve the RDMA performance almost by 3X. RDS TCP also gets a small gain
of about 12%.
RDS is being used in massive systems with high scalability where several
hundred thousand end points and tens of thousands of local processes
are operating in tens of thousand sockets. Being RC(reliable connection),
socket bind and release happens very often and any inefficiencies in
bind hash look ups hurts the overall system performance. RDS bin hash-table
uses global spin-lock which is the biggest bottleneck. To make matter worst,
it uses rcu inside global lock for hash buckets.
This is being addressed by simply using per bucket rw lock which makes the
locking simple and very efficient. The hash table size is still an issue and
I plan to address it by using re-sizable hash tables as suggested on the list.
For RDS RDMA improvement, the completion handling is revamped so that we
can do batch completions. Both send and receive completion handlers are
split logically to achieve the same. RDS 8K messages being one of the
key usecase, mr pool is adapted to have the 8K mrs along with default 1M
mrs. And while doing this, few fixes and couple of bottlenecks seen with
rds_sendmsg() are addressed.
Series applies against 4.3-rc1 as well net-next. Its tested on Oracle
hardware with IB fabric for both bcopy as well as RDMA mode. RDS TCP is
tested with iXGB NIC. Like last time, iWARP transport is untested with
these changes. The patchset is also available at below git repo:
git://git.kernel.org/pub/scm/linux/kernel/git/ssantosh/linux.git net/rds/4.3-v3
As a side note, the IB HCA driver I used for testing misses at least 3
important patches in upstream to see the full blown IB performance and
am hoping to get that in mainline with help of them.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit f711a6ae06 ("net/rds: RDS-TCP: Always create a new rds_sock
for an incoming connection.") modified rds-tcp so that an incoming SYN
would ignore an existing "client" TCP connection which had the local
port set to the transient port. The motivation for ignoring the existing
"client" connection in f711a6ae was to avoid race conditions and an
endless duel of reconnect attempts triggered by a restart/abort of one
of the nodes in the TCP connection.
However, having separate sockets for active and passive sides
is avoidable, and the simpler model of a single TCP socket for
both send and receives of all RDS connections associated with
that tcp socket makes for easier observability. We avoid the race
conditions from f711a6ae by attempting reconnects in rds_conn_shutdown
if, and only if, the (new) c_outgoing bit is set for RDS_TRANS_TCP.
The c_outgoing bit is initialized in __rds_conn_create().
A side-effect of re-using the client rds_connection for an incoming
SYN is the potential of encountering duelling SYNs, i.e., we
have an outgoing RDS_CONN_CONNECTING socket when we get the incoming
SYN. The logic to arbitrate this criss-crossing SYN exchange in
rds_tcp_accept_one() has been modified to emulate the BGP state
machine: the smaller IP address should back off from the connection attempt.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
One global lock protecting hash-tables with 1024 buckets isn't
efficient and it shows up in a massive systems with truck
loads of RDS sockets serving multiple databases. The
perf data clearly highlights the contention on the rw
lock in these massive workloads.
When the contention gets worse, the code gets into a state where
it decides to back off on the lock. So while it has disabled interrupts,
it sits and backs off on this lock get. This causes the system to
become sluggish and eventually all sorts of bad things happen.
The simple fix is to move the lock into the hash bucket and
use per-bucket lock to improve the scalability.
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
If we get an ENOMEM during rds_ib_recv_refill, we might never come
back and refill again later. Patch makes sure to kick krdsd into
helping out.
To achieve this we add RDS_RECV_REFILL flag and update in the refill
path based on that so that at least some therad will keep posting
receive buffers.
Since krdsd and softirq both might race for refill, we decide to
schedule on work queue based on ring_low instead of ring_empty.
Reviewed-by: Ajaykumar Hotchandani <ajaykumar.hotchandani@oracle.com>
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Open the sockets calling sock_create_kern() with the correct struct net
pointer, and use that struct net pointer when verifying the
address passed to rds_bind().
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking updates from David Miller:
1) Add TX fast path in mac80211, from Johannes Berg.
2) Add TSO/GRO support to ibmveth, from Thomas Falcon
3) Move away from cached routes in ipv6, just like ipv4, from Martin
KaFai Lau.
4) Lots of new rhashtable tests, from Thomas Graf.
5) Run ingress qdisc lockless, from Alexei Starovoitov.
6) Allow servers to fetch TCP packet headers for SYN packets of new
connections, for fingerprinting. From Eric Dumazet.
7) Add mode parameter to pktgen, for testing receive. From Alexei
Starovoitov.
8) Cache access optimizations via simplifications of build_skb(), from
Alexander Duyck.
9) Move page frag allocator under mm/, also from Alexander.
10) Add xmit_more support to hv_netvsc, from KY Srinivasan.
11) Add a counter guard in case we try to perform endless reclassify
loops in the packet scheduler.
12) Extern flow dissector to be programmable and use it in new "Flower"
classifier. From Jiri Pirko.
13) AF_PACKET fanout rollover fixes, performance improvements, and new
statistics. From Willem de Bruijn.
14) Add netdev driver for GENEVE tunnels, from John W Linville.
15) Add ingress netfilter hooks and filtering, from Pablo Neira Ayuso.
16) Fix handling of epoll edge triggers in TCP, from Eric Dumazet.
17) Add an ECN retry fallback for the initial TCP handshake, from Daniel
Borkmann.
18) Add tail call support to BPF, from Alexei Starovoitov.
19) Add several pktgen helper scripts, from Jesper Dangaard Brouer.
20) Add zerocopy support to AF_UNIX, from Hannes Frederic Sowa.
21) Favor even port numbers for allocation to connect() requests, and
odd port numbers for bind(0), in an effort to help avoid
ip_local_port_range exhaustion. From Eric Dumazet.
22) Add Cavium ThunderX driver, from Sunil Goutham.
23) Allow bpf programs to access skb_iif and dev->ifindex SKB metadata,
from Alexei Starovoitov.
24) Add support for T6 chips in cxgb4vf driver, from Hariprasad Shenai.
25) Double TCP Small Queues default to 256K to accomodate situations
like the XEN driver and wireless aggregation. From Wei Liu.
26) Add more entropy inputs to flow dissector, from Tom Herbert.
27) Add CDG congestion control algorithm to TCP, from Kenneth Klette
Jonassen.
28) Convert ipset over to RCU locking, from Jozsef Kadlecsik.
29) Track and act upon link status of ipv4 route nexthops, from Andy
Gospodarek.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1670 commits)
bridge: vlan: flush the dynamically learned entries on port vlan delete
bridge: multicast: add a comment to br_port_state_selection about blocking state
net: inet_diag: export IPV6_V6ONLY sockopt
stmmac: troubleshoot unexpected bits in des0 & des1
net: ipv4 sysctl option to ignore routes when nexthop link is down
net: track link-status of ipv4 nexthops
net: switchdev: ignore unsupported bridge flags
net: Cavium: Fix MAC address setting in shutdown state
drivers: net: xgene: fix for ACPI support without ACPI
ip: report the original address of ICMP messages
net/mlx5e: Prefetch skb data on RX
net/mlx5e: Pop cq outside mlx5e_get_cqe
net/mlx5e: Remove mlx5e_cq.sqrq back-pointer
net/mlx5e: Remove extra spaces
net/mlx5e: Avoid TX CQE generation if more xmit packets expected
net/mlx5e: Avoid redundant dev_kfree_skb() upon NOP completion
net/mlx5e: Remove re-assignment of wq type in mlx5e_enable_rq()
net/mlx5e: Use skb_shinfo(skb)->gso_segs rather than counting them
net/mlx5e: Static mapping of netdev priv resources to/from netdev TX queues
net/mlx4_en: Use HW counters for rx/tx bytes/packets in PF device
...
The BUG_ON at line 452/453 is triggered in function rds_send_xmit.
441 while (ret) {
442 tmp = min_t(int, ret, sg->length -
443 conn->c_xmit_data_off);
444 conn->c_xmit_data_off += tmp;
445 ret -= tmp;
446 if (conn->c_xmit_data_off == sg->length) {
447 conn->c_xmit_data_off = 0;
448 sg++;
449 conn->c_xmit_sg++;
450 if (ret != 0 && conn->c_xmit_sg == rm->data.op_nents)
451 printk(KERN_ERR "conn %p rm %p sg %p ret %d\n", conn, rm, sg, ret);
452 BUG_ON(ret != 0 &&
453 conn->c_xmit_sg == rm->data.op_nents);
454 }
455 }
it is complaining the total sent length is bigger that we want to send.
rds_ib_xmit() is wrong for the second entry for the same rds_message returning
wrong value.
the sg and off passed by rds_send_xmit to rds_ib_xmit is based on
scatterlist.offset/length, but the rds_ib_xmit action is based on
scatterlist.dma_address/dma_length. in case dma_length is larger than length
there is problem. for the 2nd and later entries of rds_ib_xmit for same
rds_message, at least one of the following two is wrong:
1) the scatterlist to start with, the choosen one can far beyond the correct
one.
2) the offset to start with within the scatterlist.
fix:
add op_dmasg and op_dmaoff to rm_data_op structure indicating the scatterlist
and offset within the it to start with for rds_ib_xmit respectively. op_dmasg
and op_dmaoff are initialized to zero when doing dma mapping for the first see
of the message and are changed when filling send slots.
the same applies to rds_iw_xmit too.
Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
An application may deterministically attach the underlying transport for
a PF_RDS socket by invoking setsockopt(2) with the SO_RDS_TRANSPORT
option at the SOL_RDS level. The integer argument to setsockopt must be
one of the RDS_TRANS_* transport types, e.g., RDS_TRANS_TCP. The option
must be specified before invoking bind(2) on the socket, and may only
be used once on the socket. An attempt to set the option on a bound
socket, or to invoke the option after a successful SO_RDS_TRANSPORT
attachment, will return EOPNOTSUPP.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
