This reverts commit c191a836a9.
It causes known regressions for programs that expect to be able to use
SO_REUSEADDR to shutdown a socket, then successfully rebind another
socket to the same ID.
Programs such as haproxy and amavisd expect this to work.
This should fix kernel bugzilla 32832.
Signed-off-by: David S. Miller <davem@davemloft.net>
Machines are getting deadlock in four node cluster environment.
All nodes are accessing (find /gfs2 -depth -print|cpio -ocv > /dev/null)
200 GB storage on a GFS2 filesystem.
This result in memory fragmentation and driver receives 18 frags for
1448 byte packets.
For non tso packet, fw drops the tx request, if it has >14 frags.
Fixing it by pulling extra frags.
Cc: stable@kernel.org
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 74888760d4
"dt/net: Eliminate users of of_platform_{,un}register_driver"
broke building mscan driver. Fix it.
Signed-off-by: Anatolij Gustschin <agust@denx.de>
Cc: Grant Likely <grant.likely@secretlab.ca>
Acked-by: Wolfgang Grandegger <wg@grandegger.com>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ethtool support to configure RX, TX and other channels. combined field
in struct ethtool_channels to reflect set of channel (RX, TX or other).
Other channel can be link interrupts, SR-IOV coordination etc.
ETHTOOL_GCHANNELS will report max and current number of RX channels,
max and current number of TX channels, max and current number of other channel
or max and current number of combined channel.
Number of channel can be modify upto max number of channel through
ETHTOOL_SCHANNELS command.
Ben Hutchings:
o define 'combined' and 'other' types. Most multiqueue drivers pair up RX and TX
queues so that most channels combine RX and TX work.
o Please could you use a kernel-doc comment to describe the structure.
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Reviewed-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SCTP and UDPLITE port support added to the hash:*port* set types.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Since we can not update retran path to unconfirmed transports,
when we remove a peer, the retran path may not be update if the
other transports are all unconfirmed, and we will still using
the removed transport as the retran path. This may cause panic
if retrasnmit happen.
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit fbdf501c93
sctp: Do no select unconfirmed transports for retransmissions
Introduced the initial falt.
commit d598b166ce
sctp: Make sure we always return valid retransmit path
Solved the problem, but forgot to change the DEBUG statement.
Thus it was still possible to dereference a NULL pointer.
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
NETIF_F_TSO_ECN has no effect when TSO is disabled; this just means
that feature state will be accurately reported to user-space.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The feature flags NETIF_F_TSO and NETIF_F_TSO6 independently enable
TSO for IPv4 and IPv6 respectively. However, the test in
netdev_fix_features() and its predecessor functions was never updated
to check for NETIF_F_TSO6, possibly because it was originally proposed
that TSO for IPv6 would be dependent on both feature flags.
Now that these feature flags can be changed independently from
user-space and we depend on netdev_fix_features() to fix invalid
feature combinations, it's important to disable them both if
scatter-gather is disabled. Also disable NETIF_F_TSO_ECN so
user-space sees all TSO features as disabled.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The "iavcc" and "iadev" cases are obvious.
The intr_status and frmr_intr cases are reading a register to clear
the chip status. This driver is pretty old and creaky, and uses
volatile pointer dereferences to do register I/O when it should be
using readl() and friends. However that it outside of the scope of
these changes.
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes the following warnings:
net/9p/client.c:1305:18: warning: variable ‘total’ set but not used [-Wunused-but-set-variable]
net/9p/client.c:1370:18: warning: variable ‘total’ set but not used [-Wunused-but-set-variable]
net/9p/client.c:1769:18: warning: variable ‘total’ set but not used [-Wunused-but-set-variable]
Signed-off-by: David S. Miller <davem@davemloft.net>
Enabling write-combining may also enable read reordering. The BIU is
only guaranteed to read from a 128-bit CSR or 64-bit SRAM word when
the host reads from its lowest address; otherwise the BIU may use the
latched value. Therefore we need to reinstate the read memory
barriers after the first read operation for each CSR or SRAM word.
Signed-off-by; Ben Hutchings <bhutchings@solarflare.com>
It adds -Wall (which the kernel carefully controls already) and of all
things -DDEBUG (which should be set by other means if desired, please
we have dynamic-debug these days).
Kill this noise.
Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5b40964ead ("irda: Remove BKL instances
from af_irda.c") introduced a path where we have a locking unbalance.
If we pass invalid flags, we unlock a socket we never locked,
resulting in this...
=====================================
[ BUG: bad unlock balance detected! ]
-------------------------------------
trinity/20101 is trying to release lock (sk_lock-AF_IRDA) at:
[<ffffffffa057f001>] irda_sendmsg+0x207/0x21d [irda]
but there are no more locks to release!
other info that might help us debug this:
no locks held by trinity/20101.
stack backtrace:
Pid: 20101, comm: trinity Not tainted 2.6.39-rc3+ #3
Call Trace:
[<ffffffffa057f001>] ? irda_sendmsg+0x207/0x21d [irda]
[<ffffffff81085041>] print_unlock_inbalance_bug+0xc7/0xd2
[<ffffffffa057f001>] ? irda_sendmsg+0x207/0x21d [irda]
[<ffffffff81086aca>] lock_release+0xcf/0x18e
[<ffffffff813ed190>] release_sock+0x2d/0x155
[<ffffffffa057f001>] irda_sendmsg+0x207/0x21d [irda]
[<ffffffff813e9f8c>] __sock_sendmsg+0x69/0x75
[<ffffffff813ea105>] sock_sendmsg+0xa1/0xb6
[<ffffffff81100ca3>] ? might_fault+0x5c/0xac
[<ffffffff81086b7c>] ? lock_release+0x181/0x18e
[<ffffffff81100cec>] ? might_fault+0xa5/0xac
[<ffffffff81100ca3>] ? might_fault+0x5c/0xac
[<ffffffff81133b94>] ? fcheck_files+0xb9/0xf0
[<ffffffff813f387a>] ? copy_from_user+0x2f/0x31
[<ffffffff813f3b70>] ? verify_iovec+0x52/0xa6
[<ffffffff813eb4e3>] sys_sendmsg+0x23a/0x2b8
[<ffffffff81086b7c>] ? lock_release+0x181/0x18e
[<ffffffff810773c6>] ? up_read+0x28/0x2c
[<ffffffff814bec3d>] ? do_page_fault+0x360/0x3b4
[<ffffffff81087043>] ? trace_hardirqs_on_caller+0x10b/0x12f
[<ffffffff810458aa>] ? finish_task_switch+0xb2/0xe3
[<ffffffff8104583e>] ? finish_task_switch+0x46/0xe3
[<ffffffff8108364a>] ? trace_hardirqs_off_caller+0x33/0x90
[<ffffffff814bbaf9>] ? retint_swapgs+0x13/0x1b
[<ffffffff81087043>] ? trace_hardirqs_on_caller+0x10b/0x12f
[<ffffffff810a9dd3>] ? audit_syscall_entry+0x11c/0x148
[<ffffffff8125609e>] ? trace_hardirqs_on_thunk+0x3a/0x3f
[<ffffffff814c22c2>] system_call_fastpath+0x16/0x1b
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since ndo_fix_features callback is postponing features change when
bp->recovery_state != BNX2X_RECOVERY_DONE, netdev_update_features()
has to be called again when this condition changes. Previously,
ethtool_ops->set_flags callback returned -EBUSY in that case
(it's not possible in the new model).
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
v5: - don't delay set_features, as it's rtnl_locked - same as recovery process
v4: - complete bp->rx_csum -> NETIF_F_RXCSUM conversion
- add check for failed ndo_set_features in ndo_open callback
v3: - include NETIF_F_LRO in hw_features
- don't call netdev_update_features() if bnx2x_nic_load() failed
v2: - comment in ndo_fix_features callback
Signed-off-by: David S. Miller <davem@davemloft.net>
Note: looks like bnad->conf_mutex is duplicating rtnl_lock.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
Caveats:
- driver modifies vlan_features on HW VLAN TX changes
- broken RX checksum will be reenabled on features change
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a skb is delivered to a registered callback, cn_call_callback()
incorrectly returns -ENODEV after freeing the skb, causing cn_rx_skb()
to free the skb a second time.
Reported-by: Eric B Munson <emunson@mgebm.net>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Tested-by: Eric B Munson <emunson@mgebm.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now there are 2 paths for rx vlan frames. When rx-vlan-hw-accel is
enabled, skb is untagged by NIC, vlan_tci is set and the skb gets into
vlan code in __netif_receive_skb - vlan_hwaccel_do_receive.
For non-rx-vlan-hw-accel however, tagged skb goes thru whole
__netif_receive_skb, it's untagged in ptype_base hander and reinjected
This incosistency is fixed by this patch. Vlan untagging happens early in
__netif_receive_skb so the rest of code (ptype_all handlers, rx_handlers)
see the skb like it was untagged by hw.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
v1->v2:
remove "inline" from vlan_core.c functions
Signed-off-by: David S. Miller <davem@davemloft.net>
controlling igmp_max_membership is useful even when IP_MULTICAST
is off.
Quagga(an OSPF deamon) uses multicast addresses for all interfaces
using a single socket and hits igmp_max_membership limit when
there are 20 interfaces or more.
Always export sysctl igmp_max_memberships in proc, just like
igmp_max_msf
Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
When blinking for a duration set by the user, the value specified is in
seconds but it is used as the number of jiffies in the timeout after which
the Physical ID indicator is deactivated. Fix by converting the timeout
to seconds.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Acked-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pass the correct module name and device interface so that
ethtool can display the proper values.
The firmware version will be fixed later on when the FW
can actually report a version. :)
Reported-by: Richard Farina <sidhayn@gmail.com>
Signed-off-by: Sujith Manoharan <Sujith.Manoharan@atheros.com>
Tested-by: Richard Farina <sidhayn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
During PHY errors, the MAC can sometimes fail to enter an idle state on older
hardware (before AR9380) after an rx stop has been requested.
This typically shows up in the kernel log with messages like these:
ath: Could not stop RX, we could be confusing the DMA engine when we start RX up
------------[ cut here ]------------
WARNING: at drivers/net/wireless/ath/ath9k/recv.c:504 ath_stoprecv+0xcc/0xf0 [ath9k]()
Call Trace:
[<8023f0e8>] dump_stack+0x8/0x34
[<80075050>] warn_slowpath_common+0x78/0xa4
[<80075094>] warn_slowpath_null+0x18/0x24
[<80d66d60>] ath_stoprecv+0xcc/0xf0 [ath9k]
[<80d642cc>] ath_set_channel+0xbc/0x270 [ath9k]
[<80d65254>] ath_radio_disable+0x4a4/0x7fc [ath9k]
When this happens, the state that the MAC enters is easy to identify and
does not result in bogus DMA traffic, however to ensure a working state
after a channel change, the hardware should still be reset.
This patch adds detection for this specific MAC state, after which the above
warnings completely disappear in my tests.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@kernel.org
Cc: Kyungwan Nam <Kyungwan.Nam@Atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Commit 462fb2af97 (bridge : Sanitize skb before it enters the IP
stack), missed one IPCB init before calling ip_options_compile()
Thanks to Scot Doyle for his tests and bug reports.
Reported-by: Scot Doyle <lkml@scotdoyle.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Hiroaki SHIMODA <shimoda.hiroaki@gmail.com>
Acked-by: Bandan Das <bandan.das@stratus.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Cc: Jan Lübbe <jluebbe@debian.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
In vxge_hw_vpath_close, __vxge_hw_vp_terminate memsets the vpath which
clobbers the spin lock state, then the driver attempts to acquire the
spin lock. Resolve this by not zeroing the lock part of vpath struct,
clean-up vpath locking in init, close, and fix locking hole in fw_api
call.
Issue found by Bob Picco <bpicco@redhat.com>
Signed-off-by: Jon Mason <jdmason@kudzu.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hardware time stamp calculation can only be enabled by the privileged
function. Enable it always by default and simply use the ethtool
interface to set a flag to indicate whether or not the respective
function driver should indicate the timestamp along with the received
packet.
Also, make certain fields in vxge_hw_device_config bit-fields to reduce
the size of the struct.
Signed-off-by: Jon Mason <jdmason@kudzu.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most of these are cases where we are trying to read back a register
after a write to ensure completion.
Simply pre-fixing the readl() or readq() with "(void)" is sufficient
because these are volatile operations and the compiler cannot eliminate
them just because no real assignment takes place.
The case of free_rxd_blk()'s assignments to "struct buffAdd *ba" is a
real spurious assignment as this variable is completely otherwise
unused.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Jon Mason <jdmason@kudzu.us>
During self-tests we use efx_process_channel_now() to handle
completion and other events synchronously. This disables interrupts
and NAPI processing for the channel in question, but it may still be
interrupted by another channel. A single socket may receive packets
from multiple net devices or even multiple channels of the same net
device, so this can result in deadlock on a socket lock.
Receiving packets in process context will also result in incorrect
classification by the network cgroup classifier.
Therefore, we must only use efx_process_channel_now() in the offline
loopback tests (which never deliver packets up the stack) and not for
the online interrupt and event tests.
For the interrupt test, there is no reason to process events. We
only care that an interrupt is raised.
For the event test, we want to know whether events have been received,
and there may be many events ahead of the one we inject. Therefore
remove efx_channel::magic_count and instead test whether
efx_channel::eventq_read_ptr advances. This is currently an event
queue index and might wrap around to exactly the same value, resulting
in a false negative. Therefore move the masking to efx_event() and
efx_nic_eventq_read_ack() so that it cannot wrap within the time of
the test.
The event test also tries to diagnose failures by checking whether an
event was delivered without causing an interrupt. Add and use a
helper function that only does this.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
If the TX queues are running during loopback self tests, host
traffic gets looped back which causes the test to fail. Avoid
restarting the TX queues after the port reset so that any packets
sent by the host get held back until after the tests have completed.
[bwh: Also wake all TX queues at the end of self-tests.]
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Fixes bugzilla #32872
The LLC stack pretends to support non-linear skbs but there is a
direct use of skb_tail_pointer() in llc_fixup_skb().
Use pskb_may_pull() to see if data_size bytes remain and can be
accessed linearly in the packet, instead of direct pointer checks.
Signed-off-by: David S. Miller <davem@davemloft.net>
Most of these were legitimate, and once case was a real bug
(not propagating errors from ->xcvr_init() methods).
Signed-off-by: David S. Miller <davem@davemloft.net>
If CAIF Link Layer returns an error, we no longer try to re-build the
CAIF packet and resend it. Instead, we simply return any transmission
errors to the socket client.
Signed-off-by: Sjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cleanup of new CAIF code.
* make local functions static
* remove code that is never used
* expand get_caif_conf() since wrapper is no longer needed
* make args to comparison functions const
* rename connect_req_to_link_param to keep exported names
consistent
Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Sjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the RX network flow classification interface was originally
defined for reporting and controlling of flow hashing, AH and ESP were
not given distinct flow class codes (apparently because the Sun
Neptune hardware treats them very similarly).
For flow steering, they must be distinguished, so new and separate
flow class codes were added for AH and ESP. But for backward-
compatibility, flow hash operations should continue to support the
original class codes.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
