transparent field of a socket is either inet_twsk(sk)->tw_transparent
for timewait sockets, or inet_sk(sk)->transparent for other sockets
(TCP/UDP).
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
CAIF sockets should use socket's default send and receive buffers sizes.
Signed-off-by: Sjur Braendeland <sjur.brandeland@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Check that receive function pointer is not null before calling it.
Signed-off-by: Sjur Braendeland <sjur.brandeland@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use pr_debug for flow control printouts, and refine an error printout.
Signed-off-by: Sjur Braendeland <sjur.brandeland@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove debugging quirk redefining pr_debug to pr_warning.
Signed-off-by: Sjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/core/ethtool.c: In function 'ethtool_get_regs':
net/core/ethtool.c:818:2: error: implicit declaration of function 'vmalloc'
net/core/ethtool.c:818:9: warning: assignment makes pointer from integer without a cast
net/core/ethtool.c:833:2: error: implicit declaration of function 'vfree'
Signed-off-by: David S. Miller <davem@davemloft.net>
Special care should be taken when slow path is hit in ip_fragment() :
When walking through frags, we transfert truesize ownership from skb to
frags. Then if we hit a slow_path condition, we must undo this or risk
uncharging frags->truesize twice, and in the end, having negative socket
sk_wmem_alloc counter, or even freeing socket sooner than expected.
Many thanks to Nick Bowler, who provided a very clean bug report and
test program.
Thanks to Jarek for reviewing my first patch and providing a V2
While Nick bisection pointed to commit 2b85a34e91 (net: No more
expensive sock_hold()/sock_put() on each tx), underlying bug is older
(2.6.12-rc5)
A side effect is to extend work done in commit b2722b1c3a
(ip_fragment: also adjust skb->truesize for packets not owned by a
socket) to ipv6 as well.
Reported-and-bisected-by: Nick Bowler <nbowler@elliptictech.com>
Tested-by: Nick Bowler <nbowler@elliptictech.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Jarek Poplawski <jarkao2@gmail.com>
CC: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some NICs have huge register files which exceed the maximum heap
allocation size.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The `options_received' struct is redundant, since it re-duplicates the existing
`p' and `x_recv' fields. This patch removes the sub-struct and migrates the
format conversion operations to ccid3_hc_tx_parse_options().
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This adds a function to take care of the following, separate cases occurring in
the computation of the Loss Rate p:
* 1/(2^32-1) is mapped into 0% as per RFC 4342, 8.5;
* 1/0 is mapped into 100%, the maximum;
* to avoid that p = 1/x is rounded down to 0 when x is very large, since this
means accidentally re-entering slow-start indicated by p == 0, the minimum
resolution value of p is now returned instead;
* a bug in ccid3_hc_rx_getsockopt is fixed: 1/0 was mapped into ~0U.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This patch is thanks to an investigation by Leandro Sales de Melo and his
colleagues. They worked out two state diagrams which highlight the fact that
the xxx_TERM states in CCID-3/4 are in fact not necessary.
And this can be confirmed by in turn looking at the code: the xxx_TERM states
are only ever set in ccid3_hc_{rx,tx}_exit(): when CCID-3 sets the state
to xxx_TERM, it is at a time where no more processing should be going on,
hence it is not necessary to introduce a dedicated exit state - this is already
implied by unloading the CCID.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
The constants DCCPO_{MIN,MAX}_CCID_SPECIFIC are nowhere used in the code, but
instead for the CCID-specific options numbers are used.
This patch unifies the use of CCID-specific option numbers, by adding symbolic
names reflecting the definitions in RFC 4340, 10.3.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This
1. adds packet type information to ccid_hc_{rx,tx}_parse_options(). This is
necessary, since table 3 in RFC 4340, 5.8 leaves it to the CCIDs to state
which options may (not) appear on what packet type.
2. adds such a check for CCID-3's {Loss Event, Receive} Rate as specified in
RFC 4340 8.3 ("Receive Rate options MUST NOT be sent on DCCP-Data packets")
and 8.5 ("Loss Event Rate options MUST NOT be sent on DCCP-Data packets").
3. removes an unused argument `idx' from ccid_hc_{rx,tx}_parse_options(). This
is also no longer necessary, since the CCID-specific option-parsing routines
are passed every single parameter of the type-length-value option encoding.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
If a RST comes in immediately after checking sk->sk_err, tcp_poll will
return POLLIN but not POLLOUT. Fix this by checking sk->sk_err at the end
of tcp_poll. Additionally, ensure the correct order of operations on SMP
machines with memory barriers.
Signed-off-by: Tom Marshall <tdm.code@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Just use explicit casts, since we really can't change the
types of structures exported to userspace which have been
around for 15 years or so.
Reported-by: Dan Rosenberg <dan.j.rosenberg@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The family parameter xfrm_state_find is used to find a state matching a
certain policy. This value is set to the template's family
(encap_family) right before xfrm_state_find is called.
The family parameter is however also used to construct a temporary state
in xfrm_state_find itself which is wrong for inter-family scenarios
because it produces a selector for the wrong family. Since this selector
is included in the xfrm_user_acquire structure, user space programs
misinterpret IPv6 addresses as IPv4 and vice versa.
This patch splits up the original init_tempsel function into a part that
initializes the selector respectively the props and id of the temporary
state, to allow for differing ip address families whithin the state.
Signed-off-by: Thomas Egerer <thomas.egerer@secunet.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a driver doesn't fill the entire buffer, old
heap contents may remain, and if it also doesn't
update the length properly, this old heap content
will be copied back to userspace.
It is very unlikely that this happens in any of
the drivers using private ioctls since it would
show up as junk being reported by iwpriv, but it
seems better to be safe here, so use kzalloc.
Reported-by: Jeff Mahoney <jeffm@suse.com>
Cc: stable@kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Under load, netif_rx() can drop incoming packets but administrators dont
have a chance to spot which device needs some tuning (RPS activation for
example)
This patch adds rx_dropped accounting in vlans and tunnels.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ipv6 can be a module, we should test CONFIG_IPV6 and CONFIG_IPV6_MODULE
to enable ipv6 bits in ip_gre.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Related dicussion here : http://lkml.org/lkml/2010/9/3/16
Introduce a function br_parse_ip_options that will audit the
skb and possibly refill IP options before a packet enters the
IP stack. If no options are present, the function will zero out
the skb cb area so that it is not misinterpreted as options by some
unsuspecting IP layer routine. If packet consistency fails, drop it.
Signed-off-by: Bandan Das <bandan.das@stratus.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is basically just a cleanup. IRQs were disabled on the previous
line so we don't need to do it again here. In the current code IRQs
would get turned on one line earlier than intended.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the original code if the copy_from_user() fails in rds_rdma_pages()
then the error handling fails and we get a stack trace from kmalloc().
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (21 commits)
dca: disable dca on IOAT ver.3.0 multiple-IOH platforms
netpoll: Disable IRQ around RCU dereference in netpoll_rx
sctp: Do not reset the packet during sctp_packet_config().
net/llc: storing negative error codes in unsigned short
MAINTAINERS: move atlx discussions to netdev
drivers/net/cxgb3/cxgb3_main.c: prevent reading uninitialized stack memory
drivers/net/eql.c: prevent reading uninitialized stack memory
drivers/net/usb/hso.c: prevent reading uninitialized memory
xfrm: dont assume rcu_read_lock in xfrm_output_one()
r8169: Handle rxfifo errors on 8168 chips
3c59x: Remove atomic context inside vortex_{set|get}_wol
tcp: Prevent overzealous packetization by SWS logic.
net: RPS needs to depend upon USE_GENERIC_SMP_HELPERS
phylib: fix PAL state machine restart on resume
net: use rcu_barrier() in rollback_registered_many
bonding: correctly process non-linear skbs
ipv4: enable getsockopt() for IP_NODEFRAG
ipv4: force_igmp_version ignored when a IGMPv3 query received
ppp: potential NULL dereference in ppp_mp_explode()
net/llc: make opt unsigned in llc_ui_setsockopt()
...
The ethtool utility does not set masks for flow parameters that are
not specified, so if both value and mask are 0 then this must be
treated as equivalent to a mask with all bits set. Currently that is
done in the only driver that implements RX n-tuple filtering, ixgbe.
Move it to the ethtool core.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sctp_packet_config() is called when getting the packet ready
for appending of chunks. The function should not touch the
current state, since it's possible to ping-pong between two
transports when sending, and that can result packet corruption
followed by skb overlfow crash.
Reported-by: Thomas Dreibholz <dreibh@iem.uni-due.de>
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
previously, if a vlan master device was moved from one network namespace
to another, all 802.1q and macvlan slaves were deleted.
we can use dev->reg_state to figure out whether dev_change_net_namespace
is happening, since that won't set dev->reg_state NETREG_UNREGISTERING.
so, this changes 8021q and macvlan to ignore NETDEV_UNREGISTER when
reg_state is not NETREG_UNREGISTERING.
Signed-off-by: David Lamparter <equinox@diac24.net>
Reviewed-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Daniel Lezcano <daniel.lezcano@free.fr>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
To be able to switch on GRO on a device, ethtool_set_gro() checks this
device provides a get_rx_csum() method.
Some devices dont provide this method, while they do support RX
checksumming.
This patch allows bonding to support GRO :
ethtool -K bond0 gro on
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the alloc_skb() fails then we return 65431 instead of -ENOBUFS
(-105).
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As RTNL is held while doing tunnels inserts and deletes, we can remove
ip6_tnl_lock spinlock. My initial RCU conversion was conservative and
converted the rwlock to spinlock, with no RTNL requirement.
Use appropriate rcu annotations and modern lockdep checks as well.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ip_local_out() is called with rcu_read_lock() held from ip_queue_xmit()
but not from other call sites.
Reported-and-bisected-by: Nick Bowler <nbowler@elliptictech.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rcu_dereference_raw() now needs to know the type of its argument.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some buggy APs do not respond to unicast probe requests
or send unicast probe requests very delayed so in the
worst case we should try to send broadcast probe requests,
otherwise we can get disconnected from these APs.
Even if drivers do not have filters to disregard probe
responses from foreign APs mac80211 will only process
probe responses from our associated AP for re-arming
connection monitoring.
We need to do this since the beacon monitor does not
push back the connection monitor by design so even if we
are getting beacons from these type of APs our connection
monitor currently relies heavily on the way the probe
requests are received on the AP. An example of an AP
affected by this is the Nexus One, but this has also been
observed with random APs.
We can probably optimize this later by using null funcs
instead of probe requests.
For more details refer to:
http://code.google.com/p/chromium-os/issues/detail?id=5715
This patch has fixes for stable kernels [2.6.35+].
Cc: stable@kernel.org
Cc: Paul Stewart <pstew@google.com>
Cc: Amod Bodas <amod.bodas@atheros.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The beacon monitor should be disabled when going off channel
to prevent spurious warnings and triggering connection
deterioration work such as sending probe requests. Re-enable
the beacon monitor once we come back to the home channel.
This patch has fixes for stable kernels [2.6.34+].
Cc: stable@kernel.org
Cc: Paul Stewart <pstew@google.com>
Cc: Amod Bodas <amod.bodas@atheros.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This will be used by other components next. The beacon
monitor was added as of 2.6.34 so these fixes are applicable
only to kernels >= 2.6.34.
Cc: stable@kernel.org
Cc: Paul Stewart <pstew@google.com>
Cc: Amod Bodas <amod.bodas@atheros.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When we go offchannel mac80211 currently leaves alive the
connection idle monitor. This should be instead postponed
until we come back to our home channel, otherwise by the
time we get back to the home channel we could be triggering
unecesary probe requests. For APs that do not respond to
unicast probe requests (Nexus One is a simple example) this
means we essentially get disconnected after the probes
fails.
This patch has stable fixes for kernels [2.6.35+]
Cc: stable@kernel.org
Cc: Paul Stewart <pstew@google.com>
Cc: Amod Bodas <amod.bodas@atheros.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Upon beacon loss we send probe requests after 30 seconds of idle
time and we wait for each probe response 1/2 second. We send a
total of 3 probe requests before giving up on the AP. In the case
that we reset the connection idle monitor we should reset the probe
requests count to 0. Right now this won't help in any way but
the next patch will.
This patch has fixes for stable kernel [2.6.35+].
Cc: stable@kernel.org
Cc: Paul Stewart <pstew@google.com>
Cc: Amod Bodas <amod.bodas@atheros.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This will be used in another place later. The connection
monitor was added as of 2.6.35 so these fixes will be
applicable to >= 2.6.35.
Cc: stable@kernel.org
Cc: Paul Stewart <pstew@google.com>
Cc: Amod Bodas <amod.bodas@atheros.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When a driver advertises p2p device support,
mac80211 will handle it, but internally it will
rewrite the interface type to STA/AP rather than
P2P-STA/GO since otherwise a lot of paths need
to be touched that are otherwise identical. A
p2p boolean tells drivers whether or not a given
interface will be used for p2p or not.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This adds P2P-STA and P2P-GO as device types so
we can distinguish between those and normal STA
or AP (respectively) type interfaces.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When an interface is brought up, the recent changes
to allow changing type-while-up only set the running
bit after everything was done. This broke a number
of things, including idle calculation for monitor
interfaces, and it also broke WDS station insertion
(although nobody noticed yet).
Thus, change the code to set the running bit earlier,
but keep it after the driver's add_interface was
called because otherwise drivers may iterate over
interfaces they haven't fully set up yet.
Reported-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Instead of using a WARN_ON(!mutex_is_locked())
use lockdep_assert_held() which compiles away
completely when lockdep isn't enabled, and
also is a more accurate assertion since it
checks that the current thread is holding the
mutex.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This code is modifying the station flags, and
as such should hold the flags lock so it can
do so atomically vs. other flags modifications
and readers. This issue was introduced when
this code was added in eccb8e8f, as it used
the wrong lock (thus not fixing the race that
was previously documented in a comment.)
Cc: stable@kernel.org [2.6.31+]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The output becomes:
[ 41.261941] ieee80211 phy0: Selected rate control algorithm 'minstrel_ht'
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Currently vlan devices don't have GRO by default as none of the Ethernet
drivers add NETIF_F_GRO to their vlan_features.
As GRO is a software feature add GRO to dev->vlan_features in
register_netdevice() and let vlan_dev_init() take care that it gets
enabled only when dev->features has NETIF_F_GRO too.
Signed-off-by: Brandon Philips <bphilips@suse.de>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
dev->ip_ptr is protected by rtnl and rcu.
Yet some places dont use appropriate primitives and/or locking rules.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/phonet/socket.c: In function ‘pn_res_seq_show’:
net/phonet/socket.c:726: warning: format ‘%02X’ expects type ‘unsigned int’, but argument 3 has type ‘long int’
Signed-off-by: David S. Miller <davem@davemloft.net>
I wish we could use something cleaner, such as bind(). But that would
not work since resource subscription is orthogonal/in addition to the
normal object ID allocated via bind(). This is similar to multicasting
which also uses ioctl()'s.
Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When both destination device and object are nul, Phonet routes the
packet according to the resource field. In fact, this is the most
common pattern when sending Phonet "request" packets. In this case,
the packet is delivered to whichever endpoint (socket) has
registered the resource.
This adds a new table so that Linux processes can register their
Phonet sockets to Phonet resources, if they have adequate privileges.
(Namespace support is not implemented at the moment.)
Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Closing a pipe endpoint is not normally allowed by the Phonet pipe,
other than as a side after-effect of removing the pipe between two
endpoints. But there is no way to prevent Linux userspace processes
from being killed or suffering from bugs, so this can still happen.
We might as well forcefully close Phonet pipe endpoints then.
The cellular modem supports only a few existing pipes at a time. So we
really should not leak them. This change instructs the modem to destroy
the pipe if either of the pipe's endpoint (Linux socket) is closed too
early.
Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There may be applications trying to seek
on the irnet character device, so we should
use noop_llseek to avoid returning an error
when the default llseek changes to no_llseek.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Samuel Ortiz <samuel@sortiz.org>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
As RTNL is held while doing tunnels inserts and deletes, we can remove
ipip6_lock spinlock. My initial RCU conversion was conservative and
converted the rwlock to spinlock, with no RTNL requirement.
Use appropriate rcu annotations and modern lockdep checks as well.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As RTNL is held while doing tunnels inserts and deletes, we can remove
ipgre_lock spinlock. My initial RCU conversion was conservative and
converted the rwlock to spinlock, with no RTNL requirement.
Use appropriate rcu annotations and modern lockdep checks as well.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As RTNL is held while doing tunnels inserts and deletes, we can remove
ipip_lock spinlock. My initial RCU conversion was conservative and
converted the rwlock to spinlock, with no RTNL requirement.
Use appropriate rcu annotations and modern lockdep checks as well.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct ethtool_rawip4_spec and struct ethtool_ether_spec are neither
commented nor used by any driver, so remove them. Adjust padding in
the user-visible unions that included these structures.
Fix references to struct ethtool_rawip4_spec in
ethtool_get_rx_ntuple(), which should use struct ethtool_usrip4_spec.
struct ethtool_usrip4_spec cannot hold IPv6 host addresses and there
is no separate structure that can, so remove ETH_RX_NFC_IP6 and the
reference to it in niu.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This simplifies and consolidates the TX option-parsing code:
1. The Loss Intervals option is not currently used, so dead code related to
this option is removed. I am aware of no plans to support the option, but
if someone wants to implement it (e.g. for inter-op tests), it is better
to start afresh than having to also update currently unused code.
2. The Loss Event and Receive Rate options have a lot of code in common (both
are 32 bit, both have same length etc.), so this is consolidated.
3. The test against GSR is not necessary, because
- on first loading CCID3, ccid_new() zeroes out all fields in the socket;
- ccid3_hc_tx_packet_recv() treats 0 and ~0U equivalently, due to
pinv = opt_recv->ccid3or_loss_event_rate;
if (pinv == ~0U || pinv == 0)
hctx->p = 0;
- as a result, the sequence number field is removed from opt_recv.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This removes the RTT-sampling function tfrc_tx_hist_rtt(), since
1. it suffered from complex passing of return values (the return value both
indicated successful lookup while the value doubled as RTT sample);
2. when for some odd reason the sample value equalled 0, this triggered a bug
warning about "bogus Ack", due to the ambiguity of the return value;
3. on a passive host which has not sent anything the TX history is empty and
thus will lead to unwanted "bogus Ack" warnings such as
ccid3_hc_tx_packet_recv: server(e7b7d518): DATAACK with bogus ACK-28197148
ccid3_hc_tx_packet_recv: server(e7b7d518): DATAACK with bogus ACK-26641606.
The fix is to replace the implicit encoding by performing the steps manually.
Furthermore, the "bogus Ack" warning has been removed, since it can actually be
triggered due to several reasons (network reordering, old packet, (3) above),
hence it is not very useful.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This fixes a subtle bug in the calculation of the inter-packet gap and shows
that t_delta, as it is currently used, is not needed.
The algorithm from RFC 5348, 8.3 below continually computes a send time t_nom,
which is initialised with the current time t_now; t_gran = 1E6 / HZ specifies
the scheduling granularity, s the packet size, and X the sending rate:
t_distance = t_nom - t_now; // in microseconds
t_delta = min(t_ipi, t_gran) / 2; // `delta' parameter in microseconds
if (t_distance >= t_delta) {
reschedule after (t_distance / 1000) milliseconds;
} else {
t_ipi = s / X; // inter-packet interval in usec
t_nom += t_ipi; // compute the next send time
send packet now;
}
Problem:
--------
Rescheduling requires a conversion into milliseconds (sk_reset_timer()). The
highest jiffy resolution with HZ=1000 is 1 millisecond, so using a higher
granularity does not make much sense here.
As a consequence, values of t_distance < 1000 are truncated to 0. This issue
has so far been resolved by using instead
if (t_distance >= t_delta + 1000)
reschedule after (t_distance / 1000) milliseconds;
This is unnecessarily large, a lower bound is t_delta' = max(t_delta, 1000).
And it implies a further simplification:
a) when HZ >= 500, then t_delta <= t_gran/2 = 10^6/(2*HZ) <= 1000, so that
t_delta' = MAX(1000, t_delta) = 1000 (constant value);
b) when HZ < 500, then t_delta = 1/2*MIN(rtt, t_ipi, t_gran) <= t_gran/2,
so that 1000 <= t_delta' <= t_gran/2.
The maximum error of using a constant t_delta in (b) is less than half a jiffy.
Fix:
----
The patch replaces t_delta with a constant, whose value depends on CONFIG_HZ,
changing the above algorithm to:
if (t_distance >= t_delta')
reschedule after (t_distance / 1000) milliseconds;
where t_delta' = 10^6/(2*HZ) if HZ < 500, and t_delta' = 1000 otherwise.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
You cannot invoke __smp_call_function_single() unless the
architecture sets this symbol.
Reported-by: Daniel Hellstrom <daniel@gaisler.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Accept already has socket locking.
[ Extend socket locking over TCP_LISTEN state test. -DaveM ]
Signed-off-by: Andrew Hendry <andrew.hendry@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Accept updates socket values in 3 lines so wrapped with lock_sock.
Signed-off-by: Andrew Hendry <andrew.hendry@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Listen updates socket values and needs lock_sock.
Signed-off-by: Andrew Hendry <andrew.hendry@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'bugfixes' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6:
SUNRPC: Fix the NFSv4 and RPCSEC_GSS Kconfig dependencies
statfs() gives ESTALE error
NFS: Fix a typo in nfs_sockaddr_match_ipaddr6
sunrpc: increase MAX_HASHTABLE_BITS to 14
gss:spkm3 miss returning error to caller when import security context
gss:krb5 miss returning error to caller when import security context
Remove incorrect do_vfs_lock message
SUNRPC: cleanup state-machine ordering
SUNRPC: Fix a race in rpc_info_open
SUNRPC: Fix race corrupting rpc upcall
Fix null dereference in call_allocate
netdev_wait_allrefs() waits that all references to a device vanishes.
It currently uses a _very_ pessimistic 250 ms delay between each probe.
Some users reported that no more than 4 devices can be dismantled per
second, this is a pretty serious problem for some setups.
Most of the time, a refcount is about to be released by an RCU callback,
that is still in flight because rollback_registered_many() uses a
synchronize_rcu() call instead of rcu_barrier(). Problem is visible if
number of online cpus is one, because synchronize_rcu() is then a no op.
time to remove 50 ipip tunnels on a UP machine :
before patch : real 11.910s
after patch : real 1.250s
Reported-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reported-by: Octavian Purdila <opurdila@ixiacom.com>
Reported-by: Benjamin LaHaise <bcrl@kvack.org>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sta_info_get_bss() is used to match STA pointers
for VLAN/AP interfaces, but if the same station
is also added to multiple other interfaces it
will erroneously match because both pointers are
NULL, fix this by ignoring NULL pointers here.
Reported-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This is needed to avoid warning in ieee80211_restart_hw about hardware
scan in progress.
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Acked-by: Wey-Yi W Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
hdr pointer is left dangling after call to ieee80211_skb_resize. This
can cause guards around mesh path selection to fail.
Signed-off-by: Steve deRosier <steve@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
There is a path in nl80211_set_wiphy where result is tested but
uninitialized.
I am hitting this path when I attempt:
sh# iw dev wlan0 set channel 10
command failed: Unknown error 1069727332 (-1069727332)
Signed-off-by: William Jordan <bjordan@rajant.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Replace sizeof(rtap_namespace_sizes) / sizeof(rtap_namespace_sizes[0])
with ARRAY_SIZE(rtap_namespace_sizes) in net/wireless/radiotap.c
Signed-off-by: Nikitas Angelinas <nikitasangelinas@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Allocate hash tables for every online cpus, not every possible ones.
NUMA aware allocations.
Dont use a full page on arches where PAGE_SIZE > 1024*sizeof(void *)
misc:
__percpu , __read_mostly, __cpuinit annotations
flow_compare_t is just an "unsigned long"
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While integrating your man-pages patch for IP_NODEFRAG, I noticed
that this option is settable by setsockopt(), but not gettable by
getsockopt(). I suppose this is not intended. The (untested,
trivial) patch below adds getsockopt() support.
Signed-off-by: Michael kerrisk <mtk.manpages@gmail.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After all these years, it turns out that the
/proc/sys/net/ipv4/conf/*/force_igmp_version
parameter isn't fully implemented.
*Symptom*:
When set force_igmp_version to a value of 2, the kernel should only perform
multicast IGMPv2 operations (IETF rfc2236). An host-initiated Join message
will be sent as a IGMPv2 Join message. But if a IGMPv3 query message is
received, the host responds with a IGMPv3 join message. Per rfc3376 and
rfc2236, a IGMPv2 host should treat a IGMPv3 query as a IGMPv2 query and
respond with an IGMPv2 Join message.
*Consequences*:
This is an issue when a IGMPv3 capable switch is the querier and will only
issue IGMPv3 queries (which double as IGMPv2 querys) and there's an
intermediate switch that is only IGMPv2 capable. The intermediate switch
processes the initial v2 Join, but fails to recognize the IGMPv3 Join responses
to the Query, resulting in a dropped connection when the intermediate v2-only
switch times it out.
*Identifying issue in the kernel source*:
The issue is in this section of code (in net/ipv4/igmp.c), which is called when
an IGMP query is received (from mainline 2.6.36-rc3 gitweb):
...
A IGMPv3 query has a length >= 12 and no sources. This routine will exit after
line 880, setting the general query timer (random timeout between 0 and query
response time). This calls igmp_gq_timer_expire():
...
.. which only sends a v3 response. So if a v3 query is received, the kernel
always sends a v3 response.
IGMP queries happen once every 60 sec (per vlan), so the traffic is low. A
IGMPv3 query *is* a strict superset of a IGMPv2 query, so this patch properly
short circuit's the v3 behaviour.
One issue is that this does not address force_igmp_version=1. Then again, I've
never seen any IGMPv1 multicast equipment in the wild. However there is a lot
of v2-only equipment. If it's necessary to support the IGMPv1 case as well:
837 if (len == 8 || IGMP_V2_SEEN(in_dev) || IGMP_V1_SEEN(in_dev)) {
Signed-off-by: David S. Miller <davem@davemloft.net>
The members of struct llc_sock are unsigned so if we pass a negative
value for "opt" it can cause a sign bug. Also it can cause an integer
overflow when we multiply "opt * HZ".
CC: stable@kernel.org
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The maximum size of the authcache is now set to 1024 (10 bits),
but on our server we need at least 4096 (12 bits). Increase
MAX_HASHTABLE_BITS to 14. This is a maximum of 16384 entries,
each containing a pointer (8 bytes on x86_64). This is
exactly the limit of kmalloc() (128K).
Signed-off-by: Miquel van Smoorenburg <mikevs@xs4all.net>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
spkm3 miss returning error to up layer when import security context,
it may be return ok though it has failed to import security context.
Signed-off-by: Bian Naimeng <biannm@cn.fujitsu.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
krb5 miss returning error to up layer when import security context,
it may be return ok though it has failed to import security context.
Signed-off-by: Bian Naimeng <biannm@cn.fujitsu.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This is just a minor cleanup: net/sunrpc/clnt.c clarifies the rpc client
state machine by commenting each state and by laying out the functions
implementing each state in the order that each state is normally
executed (in the absence of errors).
The previous patch "Fix null dereference in call_allocate" changed the
order of the states. Move the functions and update the comments to
reflect the change.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
There is a race between rpc_info_open and rpc_release_client()
in that nothing stops a process from opening the file after
the clnt->cl_kref goes to zero.
Fix this by using atomic_inc_unless_zero()...
Reported-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org
If rpc_queue_upcall() adds a new upcall to the rpci->pipe list just
after rpc_pipe_release calls rpc_purge_list(), but before it calls
gss_pipe_release (as rpci->ops->release_pipe(inode)), then the latter
will free a message without deleting it from the rpci->pipe list.
We will be left with a freed object on the rpc->pipe list. Most
frequent symptoms are kernel crashes in rpc.gssd system calls on the
pipe in question.
Reported-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org
In call_allocate we need to reach the auth in order to factor au_cslack
into the allocation.
As of a17c2153d2 "SUNRPC: Move the bound
cred to struct rpc_rqst", call_allocate attempts to do this by
dereferencing tk_client->cl_auth, however this is not guaranteed to be
defined--cl_auth can be zero in the case of gss context destruction (see
rpc_free_auth).
Reorder the client state machine to bind credentials before allocating,
so that we can instead reach the auth through the cred.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org
The list_head conversion unearther an unnecessary flow
check. Since flow is always NULL here we don't need to
see if a matching flow exists already.
Reported-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that est_tree_lock is acquired with BH protection, the other
call is unnecessary.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use rcu_dereference_rtnl() helper
Change hard coded constants in fib_flag_trans()
7 -> RTN_UNREACHABLE
8 -> RTN_PROHIBIT
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove code that trimmed excess trailing info from incoming messages
arriving over an Ethernet interface. TIPC now ignores the extra info
while the message is being processed by the node, and only trims it off
if the message is retransmitted to another node. (This latter step is
done to ensure the extra info doesn't cause the sk_buff to exceed the
outgoing interface's MTU limit.) The outgoing buffer is guaranteed to
be linear.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
xfrm4_tunnel_register() & xfrm6_tunnel_register() should
use rcu_assign_pointer() to make sure previous writes
(to handler->next) are committed to memory before chain
insertion.
deregister functions dont need a particular barrier.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
__lock_sock() and __release_sock() releases and regrabs lock but
were missing proper annotations. Add it. This removes following
warning from sparse. (Currently __lock_sock() does not emit any
warning about it but I think it is better to add also.)
net/core/sock.c:1580:17: warning: context imbalance in '__release_sock' - unexpected unlock
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
move_addr_to_kernel() and copy_from_user() requires their argument
as __user pointer but were missing proper markups. Add it.
This removes following warnings from sparse.
net/core/iovec.c:44:52: warning: incorrect type in argument 1 (different address spaces)
net/core/iovec.c:44:52: expected void [noderef] <asn:1>*uaddr
net/core/iovec.c:44:52: got void *msg_name
net/core/iovec.c:55:34: warning: incorrect type in argument 2 (different address spaces)
net/core/iovec.c:55:34: expected void const [noderef] <asn:1>*from
net/core/iovec.c:55:34: got struct iovec *msg_iov
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a list_has_sctp_addr function to simplify loop
Based on a patches by Dan Carpenter and David Miller
Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 30fff923 introduced in linux-2.6.33 (udp: bind() optimisation)
added a secondary hash on UDP, hashed on (local addr, local port).
Problem is that following sequence :
fd = socket(...)
connect(fd, &remote, ...)
not only selects remote end point (address and port), but also sets
local address, while UDP stack stored in secondary hash table the socket
while its local address was INADDR_ANY (or ipv6 equivalent)
Sequence is :
- autobind() : choose a random local port, insert socket in hash tables
[while local address is INADDR_ANY]
- connect() : set remote address and port, change local address to IP
given by a route lookup.
When an incoming UDP frame comes, if more than 10 sockets are found in
primary hash table, we switch to secondary table, and fail to find
socket because its local address changed.
One solution to this problem is to rehash datagram socket if needed.
We add a new rehash(struct socket *) method in "struct proto", and
implement this method for UDP v4 & v6, using a common helper.
This rehashing only takes care of secondary hash table, since primary
hash (based on local port only) is not changed.
Reported-by: Krzysztof Piotr Oledzki <ole@ans.pl>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Tested-by: Krzysztof Piotr Oledzki <ole@ans.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use cmpxchg() to get rid of spinlocks in inet_add_protocol() and
friends.
inet_protos[] & inet6_protos[] are moved to read_mostly section
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add two CMSGs for masked versions of cswp and fadd. args
struct modified to use a union for different atomic op type's
arguments. Change IB to do masked atomic ops. Atomic op type
in rds_message similarly unionized.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
This prints the constant identifier for work completion status and rdma
cm event types, like we already do for IB event types.
A core string array helper is added that each string type uses.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Nothing was canceling the send and receive work that might have been
queued as a conn was being destroyed.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rds_conn_shutdown() can return before the connection is shut down when
it encounters an existing state that it doesn't understand. This lets
rds_conn_destroy() then start tearing down the conn from under paths
that are still using it.
It's more reliable the shutdown work and wait for krdsd to complete the
shutdown callback. This stopped some hangs I was seeing where krdsd was
trying to shut down a freed conn.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Right now there's nothing to stop the various paths that use
rs->rs_transport from racing with rmmod and executing freed transport
code. The simple fix is to have binding to a transport also hold a
reference to the transport's module, removing this class of races.
We already had an unused t_owner field which was set for the modular
transports and which wasn't set for the built-in loop transport.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rs_transport is now also used by the rdma paths once the socket is
bound. We don't need this stale comment to tell us what cscope can.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rds_conn_destroy() can race with all other modifications of the
rds_conn_count but it was modifying the count without locking.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
The RDS IB device list wasn't protected by any locking. Traversal in
both the get_mr and FMR flushing paths could race with additon and
removal.
List manipulation is done with RCU primatives and is protected by the
write side of a rwsem. The list traversal in the get_mr fast path is
protected by a rcu read critical section. The FMR list traversal is
more problematic because it can block while traversing the list. We
protect this with the read side of the rwsem.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
It's nice to not have to go digging in the code to see which event
occurred. It's easy to throw together a quick array that maps the ib
event enums to their strings. I didn't see anything in the stack that
does this translation for us, but I also didn't look very hard.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Flushing FMRs is somewhat expensive, and is currently kicked off when
the interrupt handler notices that we are getting low. The result of
this is that FMR flushing only happens from the interrupt cpus.
This spreads the load more effectively by triggering flushes just before
we allocate a new FMR.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
We're seeing bugs today where IB connection shutdown clears the send
ring while the tasklet is processing completed sends. Implementation
details cause this to dereference a null pointer. Shutdown needs to
wait for send completion to stop before tearing down the connection. We
can't simply wait for the ring to empty because it may contain
unsignaled sends that will never be processed.
This patch tracks the number of signaled sends that we've posted and
waits for them to complete. It also makes sure that the tasklet has
finished executing.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
We are *definitely* counting cycles as closely as DaveM, so
ensure hwcache alignment for our recv ring control structs.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
The recv refill path was leaking fragments because the recv event handler had
marked a ring element as free without freeing its frag. This was happening
because it wasn't processing receives when the conn wasn't marked up or
connecting, as can be the case if it races with rmmod.
Two observations support always processing receives in the callback.
First, buildup should only post receives, thus triggering recv event handler
calls, once it has built up all the state to handle them. Teardown should
destroy the CQ and drain the ring before tearing down the state needed to
process recvs. Both appear to be true today.
Second, this test was fundamentally racy. There is nothing to stop rmmod and
connection destruction from swooping in the moment after the conn state was
sampled but before real receive procesing starts.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
We were seeing very nasty bugs due to fundamental assumption the current code
makes about concurrent work struct processing. The code simpy isn't able to
handle concurrent connection shutdown work function execution today, for
example, which is very much possible once a multi-threaded krdsd was
introduced. The problem compounds as additional work structs are added to the
mix.
krdsd is no longer perforance critical now that send and receive posting and
FMR flushing are done elsewhere, so the safest fix is to move back to the
single threaded krdsd that the current code was built around.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
This patch moves the FMR flushing work in to its own mult-threaded work queue.
This is to maintain performance in preparation for returning the main krdsd
work queue back to a single threaded work queue to avoid deep-rooted
concurrency bugs.
This is also good because it further separates FMRs, which might be removed
some day, from the rest of the code base.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
IB connections were not being destroyed during rmmod.
First, recently IB device removal callback was changed to disconnect
connections that used the removing device rather than destroying them. So
connections with devices during rmmod were not being destroyed.
Second, rds_ib_destroy_nodev_conns() was being called before connections are
disassociated with devices. It would almost never find connections in the
nodev list.
We first get rid of rds_ib_destroy_conns(), which is no longer called, and
refactor the existing caller into the main body of the function and get rid of
the list and lock wrappers.
Then we call rds_ib_destroy_nodev_conns() *after* ib_unregister_client() has
removed the IB device from all the conns and put the conns on the nodev list.
The result is that IB connections are destroyed by rmmod.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
The RDS IB client removal callback can queue work to drop the final reference
to an IB device. We have to make sure that this function has returned before
we complete rmmod or the work threads can try to execute freed code.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Using a delayed work queue helps us make sure a healthy number of FMRs
have queued up over the limit. It makes for a large improvement in RDMA
iops.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
FRM allocation and recycling is performance critical and fairly lock
intensive. The current code has a per connection lock that all
processes bang on and it becomes a major bottleneck on large systems.
This changes things to use a number of cmpxchg based lists instead,
allowing us to go through the whole FMR lifecycle without locking inside
RDS.
Zach Brown pointed out that our usage of cmpxchg for xlist removal is
racey if someone manages to remove and add back an FMR struct into the list
while another CPU can see the FMR's address at the head of the list.
The second CPU might assume the list hasn't changed when in fact any
number of operations might have happened in between the deletion and
reinsertion.
This commit maintains a per cpu count of CPUs that are currently
in xlist removal, and establishes a grace period to make sure that
nobody can see an entry we have just removed from the list.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
rds_send_xmit() was changed to hold an interrupt masking spinlock instead of a
mutex so that it could be called from the IB receive tasklet path. This broke
the TCP transport because its xmit method can block and masks and unmasks
interrupts.
This patch serializes callers to rds_send_xmit() with a simple bit instead of
the current spinlock or previous mutex. This enables rds_send_xmit() to be
called from any context and to call functions which block. Getting rid of the
c_send_lock exposes the bare c_lock acquisitions which are changed to block
interrupts.
A waitqueue is added so that rds_conn_shutdown() can wait for callers to leave
rds_send_xmit() before tearing down partial send state. This lets us get rid
of c_senders.
rds_send_xmit() is changed to check the conn state after acquiring the
RDS_IN_XMIT bit to resolve races with the shutdown path. Previously both
worked with the conn state and then the lock in the same order, allowing them
to race and execute the paths concurrently.
rds_send_reset() isn't racing with rds_send_xmit() now that rds_conn_shutdown()
properly ensures that rds_send_xmit() can't start once the conn state has been
changed. We can remove its previous use of the spinlock.
Finally, c_send_generation is redundant. Callers can race to test the c_flags
bit by simply retrying instead of racing to test the c_send_generation atomic.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
conn->c_lock is acquired in interrupt context. rds_conn_message_info() is
called from user context and was acquiring c_lock without blocking interrupts,
leading to possible deadlocks.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rds_send_acked_before() wasn't blocking interrupts when acquiring c_lock from
user context but nothing calls it. Rather than fix its use of c_lock we just
remove the function.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
When prefilling the rds frags, we end up doing a lot of allocations.
We're not in atomic context here, and so there's no reason to dip into
atomic reserves. This changes the prefills to use masks that allow
waiting.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This patch is based heavily on an initial patch by Chris Mason.
Instead of freeing slab memory and pages, it keeps them, and
funnels them back to be reused.
The lock minimization strategy uses xchg and cmpxchg atomic ops
for manipulation of pointers to list heads. We anchor the lists with a
pointer to a list_head struct instead of a static list_head struct.
We just have to carefully use the existing primitives with
the difference between a pointer and a static head struct.
For example, 'list_empty()' means that our anchor pointer points to a list with
a single item instead of meaning that our static head element doesn't point to
any list items.
Original patch by Chris, with significant mods and fixes by Andy and Zach.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: Zach Brown <zach.brown@oracle.com>
All it does is call unmap_sg(), so just call that directly.
The comment above unmap_page also may be incorrect, so we
shouldn't hold on to it, either.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
refill_one() should never be called on a recv struct that
doesn't need a new r_frag allocated. Add a WARN and remove
conditional around r_frag alloc code.
Also, add a comment to explain why r_ibinc may or may not
need refilling.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Instead of splitting up a page into RDS_FRAG_SIZE chunks
ourselves, ask rds_page_remainder_alloc() to do it. While it
is possible PAGE_SIZE > FRAG_SIZE, on x86en it isn't, so having
duplicate "carve up a page into buffers" code seems excessive.
The other modification this spawns is the use of a single
struct scatterlist in rds_page_frag instead of a bare page ptr.
This causes verbosity to increase in some places, and decrease
in others.
Finally, I decided to unify the lifetimes and alloc/free of
rds_page_frag and its page. This is a nice simplification in itself,
but will be extra-nice once we come to adding cmason's recycling
patch.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Currently IB device removal destroys connections which are associated with the
device. This prevents connections from being re-established when replacement
devices are added.
Instead we'll queue shutdown work on the connections as their devices are
removed. When we see that devices are added we triger connection attempts on
all connections that don't currently have a device.
The result is that RDS sockets can resume device-independent work (bcopy, not
RDMA) across IB device removal and restoration.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
A few paths had the same block of code to queue a connection's connect work if
it was in the right state. Let's move this in to a helper function.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
The RDS IB client .remove callback used to free the rds_ibdev for the given
device unconditionally. This could race other users of the struct. This patch
adds refcounting so that we only free the rds_ibdev once all of its users are
done.
Many rds_ibdev users are tied to connections. We give the connection a
reference and change these users to reference the device in the connection
instead of looking it up in the IB client data. The only user of the IB client
data remaining is the first lookup of the device as connections are built up.
Incrementing the reference count of a device found in the IB client data could
race with final freeing so we use an RCU grace period to make sure that freeing
won't happen until those lookups are done.
MRs need the rds_ibdev to get at the pool that they're freed in to. They exist
outside a connection and many MRs can reference different devices from one
socket, so it was natural to have each MR hold a reference. MR refs can be
dropped from interrupt handlers and final device teardown can block so we push
it off to a work struct. Pool teardown had to be fixed to cancel its pending
work instead of deadlocking waiting for all queued work, including itself, to
finish.
MRs get their reference from the global device list, which gets a reference.
It is left unprotected by locks and remains racy. A simple global lock would
be a significant bottleneck. More scalable (complicated) locking should be
done carefully in a later patch.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rds_ib_xmit_rdma() was calling ib_get_client_data() to get at the rds_ibdevice
just to get the max_sge for the transmit. This patch instead has it get it
directly off the rds_ibdev which is stored on the connection.
The current code won't free the rds_ibdev until all the IB connections that use
it are freed. So it's safe to reference the rds_ibdev this way. In the future
it also makes it easier to support proper reference counting of the rds_ibdev
struct.
As an additional bonus, this gets rid of the performance hit of calling in to
the IB stack to look up the rds_ibdev. The current implementation in the IB
stack acquires an interrupt blocking spinlock to protect the registration of
client callback data.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rds_ib_cm_handle_connect() could return without unlocking the c_conn_lock if
rds_setup_qp() failed. Rather than adding another imbalanced mutex_unlock() to
this error path we only unlock the mutex once as we exit the function, reducing
the likelyhood of making this same mistake in the future. We remove the
previous mulitple return sites, leaving one unambigious return path.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
This makes sure we have the proper number of references in
rds_ib_xmit_atomic and rds_ib_xmit_rdma. We also consistently
drop references the same way for all message types as the IOs end.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The connection hash was almost entirely RCU ready, this
just makes the final couple of changes to use RCU instead
of spinlocks for everything.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The RDS send_xmit code was trying to get fancy with message
counting and was dropping the final reference on the RDMA messages
too early. This resulted in memory corruption and oopsen.
The fix here is to always add a ref as the parts of the message passes
through rds_send_xmit, and always drop a ref as the parts of the message
go through completion handling.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This is the first in a long line of patches that tries to fix races
between RDS connection shutdown and RDS traffic.
Here we are maintaining a count of active senders to make sure
the connection doesn't go away while they are using it.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The RDS bind lookups are somewhat expensive in terms of CPU
time and locking overhead. This commit changes them into a
faster RCU based hash tree instead of the rbtrees they were using
before.
On large NUMA systems it is a significant improvement.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Allocate send/recv rings in memory that is node-local to the HCA.
This significantly helps performance.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_ib_get_device is called very often as we turn an
ip address into a corresponding device structure. It currently
take a global spinlock as it walks different lists to find active
devices.
This commit changes the lists over to RCU, which isn't very complex
because they are not updated very often at all.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This removes a global waitqueue used to wait for rds messages
and replaces it with a waitqueue inside the rds_message struct.
The global waitqueue turns into a global lock and significantly
bottlenecks operations on large machines.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The bind_lock is almost entirely readonly, but it gets
hammered during normal operations and is a major bottleneck.
This commit changes it to an rwlock, which takes it from 80%
of the system time on a big numa machine down to much lower
numbers.
A better fix would involve RCU, which is done in a later commit
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Update comments to reflect changes in previous commit.
Keeping as separate commits due to different authorship.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_send_xmit is required to loop around after it releases the lock
because someone else could done a trylock, found someone working on the
list and backed off.
But, once we drop our lock, it is possible that someone else does come
in and make progress on the list. We should detect this and not loop
around if another process is actually working on the list.
This patch adds a generation counter that is bumped every time we
get the lock and do some send work. If the retry notices someone else
has bumped the generation counter, it does not need to loop around and
continue working.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
The purpose of the send quota was really to give fairness
when different connections were all using the same
workq thread to send backlogged msgs -- they could only send
so many before another connection could make progress.
Now that each connection is pushing the backlog from its
completion handler, they are all guaranteed to make progress
and the quota isn't needed any longer.
A thread *will* have to send all previously queued data, as well
as any further msgs placed on the queue while while c_send_lock
was held. In a pathological case a single process can get
roped into doing this for long periods while other threads
get off free. But, since it can only do this until the transport
reports full, this is a bounded scenario.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Now that rds_send_xmit() does not block, we can call it directly
instead of going through the helper thread.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_sendmsg() is calling the send worker function to
send the just-queued datagrams, presumably because it wants
the behavior where anything not sent will re-call the send
worker. We now ensure all queued datagrams are sent by retrying
from the send completion handler, so this isn't needed any more.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_message_put() cannot be called with irqs off, so move it after
irqs are re-enabled.
Spinlocks throughout the function do not to use _irqsave because
the lock of c_send_lock at top already disabled irqs.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
This change allows us to call rds_send_xmit() from a tasklet,
which is crucial to our new operating model.
* Change c_send_lock to a spinlock
* Update stats fields "sem_" to "_lock"
* Remove unneeded rds_conn_is_sending()
About locking between shutdown and send -- send checks if the
connection is up. Shutdown puts the connection into
DISCONNECTING. After this, all threads entering send will exit
immediately. However, a thread could be *in* send_xmit(), so
shutdown acquires the c_send_lock to ensure everyone is out
before proceeding with connection shutdown.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Performance is better if we use allocations that don't block
to refill the receive ring. Since the whole reason we were
kicking out to the worker thread was so we could do blocking
allocs, we no longer need to do this.
Remove gfp params from rds_ib_recv_refill(); we always use
GFP_NOWAIT.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
We now ask the transport to give us a rm for the congestion
map, and then we handle it normally. Previously, the
transport defined a function that we would call to send
a congestion map.
Convert TCP and loop transports to new cong map method.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Now that we are signaling send completions much less, we are likely
to have dirty entries in the send queue when the connection is
shut down (on rmmod, for example.) These are cleaned up a little
further down in conn_shutdown, but if we wait on the ring_empty_wait
for them, it'll never happen, and we hand on unload.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Previously, RDS would wait until the final send WR had completed
and then handle cleanup. With silent ops, we do not know
if an atomic, rdma, or data op will be last. This patch
handles any of these cases by keeping a pointer to the last
op in the message in m_last_op.
When the TX completion event fires, rds dispatches to per-op-type
cleanup functions, and then does whole-message cleanup, if the
last op equalled m_last_op.
This patch also moves towards having op-specific functions take
the op struct, instead of the overall rm struct.
rds_ib_connection has a pointer to keep track of a a partially-
completed data send operation. This patch changes it from an
rds_message pointer to the narrower rm_data_op pointer, and
modifies places that use this pointer as needed.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
It hasn't cropped up in the field, but this code ensures it is
impossible to issue operations that pass an rdma cookie (DEST, MAP)
in the same sendmsg call that's actually initiating rdma or atomic
ops.
Disallowing this perverse-but-technically-allowed usage makes silent
RDMA heuristics slightly easier.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Add a flag to the API so users can indicate they want
silent operations. This is needed because silent ops
cannot be used with USE_ONCE MRs, so we can't just
assume silent.
Also, change send_xmit to do atomic op before rdma op if
both are present, and centralize the hairy logic to determine if
we want to attempt silent, or not.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
When dropping ops in the send queue, we notify the client
of failed rdma ops they asked for notifications on, but not
atomic ops. It should be for both.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Do not allocate sgs for data for 0-length datagrams
Set data.op_active in rds_sendmsg() instead of
rds_message_copy_from_user().
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Simplify rds_send_xmit().
Send a congestion map (via xmit_cong_map) without
decrementing send_quota.
Move resetting of conn xmit variables to end of loop.
Update comments.
Implement a special case to turn off sending an rds header
when there is an atomic op and no other data.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
A big changeset, but it's all pretty dumb.
struct rds_rdma_op was already embedded in struct rm_rdma_op.
Remove rds_rdma_op and put its members in rm_rdma_op. Rename
members with "op_" prefix instead of "r_", for consistency.
Of course this breaks a lot, so fixup the code accordingly.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
cmsg_rdma_args just calls rdma_prepare and does a little
arg checking -- not quite enough to justify its existence.
Plus, it is the only caller of rdma_prepare().
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Maybe things worked fine with the flow control code running
even in the non-flow-control case, but making it explicitly
conditional helps the non-fc case be easier to read.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Removed unsignaled_bytes sysctl and code to signal
based on it. I believe unsignaled_wrs is more than
sufficient for our purposes.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Now that the header always goes first, it is possible to
simplify rds_ib_xmit. Instead of having a path to handle 0-byte
dgrams and another path to handle >0, these can both be handled
in one path. This lets us eliminate xmit_populate_wr().
Rename sent to bytes_sent, to differentiate better from other
variable named "send".
Signed-off-by: Andy Grover <andy.grover@oracle.com>
These functions were to cope with differently ordered
sg entries depending on RDS 3.0 or 3.1+. Now that
we've dropped 3.0 compatibility we no longer need them.
Also, modify usage sites for these to refer to sge[0] or [1]
directly. Reorder code to initialize header sgs first.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
RDS 3.0 connections (in OFED 1.3 and earlier) put the
header at the end. 3.1 connections put it at the head.
The code has significant added complexity in order to
handle both configurations. In OFED 1.6 we can
drop this and simplify the code by only supporting
"header-first" configuration.
This patch checks the protocol version, and if prior
to 3.1, does not complete the connection.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
both atomics and rdmas need to convert ib-specific completion codes
into RDS status codes. Rename rds_ib_rdma_send_complete to
rds_ib_send_complete, and have it take a pointer to the function to
call with the new error code.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Instead of using a constant for initiator_depth and
responder_resources, read the per-QP values when the
device is enumerated, and then use these values when creating
the connection.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Implement a CMSG-based interface to do FADD and CSWP ops.
Alter send routines to handle atomic ops.
Add atomic counters to stats.
Add xmit_atomic() to struct rds_transport
Inline rds_ib_send_unmap_rdma into unmap_rm
Signed-off-by: Andy Grover <andy.grover@oracle.com>
The previous code was correct, but made the assumption that
if r_notifier was non-NULL then either r_recverr or r_notify
was true. Valid, but fragile. Changed to explicitly check
r_recverr (shows up in greps for recverr now, too.)
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_message_alloc_sgs() now returns correctly-initialized
sg lists, so calleds need not do this themselves.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
This eliminates a separate memory alloc, although
it is now necessary to add an "r_active" flag, since
it is no longer to use the m_rdma_op pointer as an
indicator of if an rdma op is present.
rdma SGs allocated from rm sg pool.
rds_rm_size also gets bigger. It's a little inefficient to
run through CMSGs twice, but it makes later steps a lot smoother.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Eric Dumazet