openwrt/staging/blogic.git
10 years agopkt_sched: fq: do not hold qdisc lock while allocating memory
Eric Dumazet [Fri, 7 Mar 2014 06:57:52 +0000 (22:57 -0800)]
pkt_sched: fq: do not hold qdisc lock while allocating memory

Resizing fq hash table allocates memory while holding qdisc spinlock,
with BH disabled.

This is definitely not good, as allocation might sleep.

We can drop the lock and get it when needed, we hold RTNL so no other
changes can happen at the same time.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: afe4fd062416 ("pkt_sched: fq: Fair Queue packet scheduler")
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Sat, 8 Mar 2014 23:49:29 +0000 (18:49 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates

This series contains updates to e1000e, ixgbevf and igb.

Majority of this series contains fixes and cleanups to e1000e,
most notably are:

Todd provides a fix to PTP in e1000e which adds a lock in e1000e_phc_adjfreq
to prevent concurrent changes to TIMINCA and SYSTIMH/L.  Then provides an
igb fix to use ARRAY_SIZE for array size calculation.

David provides the remaining e1000e which contain:
 - cleanup of pointer references that are no longer used
 - fix an issue on systems with Management Engine enabled with the
   ethernet cable unplugged
 - fix an issue on 82579 where enabling EEE LPI sooner than one second
   after link up causes link issues on some switches
 - refactor the power management flows to prevent the suspend path from
   being executed twice when hibernating
 - refactor the runtime power management to fix interfering with the
   functionality of Energy Efficient Ethernet when enabled and to fix
   the device from repeatedly flip between suspend and resume with the
   interface administratively downed
 - enable the feature PHY Ultra Low Power Mode which is a power saving
   feature that reduces the power consumption of the PHY when a cable is
   not connected
 - fix the ethtool offline tests for 82579 parts
 - fix SHRA register access for 82579 parts which was introduced by
   previous commit c3a0dce35af0 "e1000e: fix overrun of PHY RAR array"

Florian provides a fix for ixgbevf where skb->pkt_type was being checked
like a bitmask, but it is not a bitmask.

Fix an issue reported by Stephen Hemminger where there was a warning
about code defined but never used if IGB_HWMON is not defined.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoigb: fix warning if !CONFIG_IGB_HWMON
Jeff Kirsher [Thu, 6 Mar 2014 05:28:06 +0000 (05:28 +0000)]
igb: fix warning if !CONFIG_IGB_HWMON

Fix warning about code defined but never used if IGB_HWMON not defined.

Reported-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoigb: fix array size calculation
Todd Fujinaka [Tue, 4 Mar 2014 02:25:22 +0000 (02:25 +0000)]
igb: fix array size calculation

Use ARRAY_SIZE for array size calculation.

Signed-off-by: Todd Fujinaka <todd.fujinaka@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoixgbevf: fix skb->pkt_type checks
Florian Fainelli [Fri, 28 Feb 2014 23:46:49 +0000 (15:46 -0800)]
ixgbevf: fix skb->pkt_type checks

skb->pkt_type is not a bitmask, but contains only value at a time from
the range defined in include/uapi/linux/if_packet.h.

Checking it like if it was a bitmask of values would also cause
PACKET_OTHERHOST, PACKET_LOOPBACK and PACKET_FASTROUTE to be matched by
this check since their lower 2 bits are also set, although that does not
fix a real bug, it is still potentially confusing.

This bogus check was introduced in commit 815cccbf ("ixgbe: add setlink,
getlink support to ixgbe and ixgbevf").

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Fix SHRA register access for 82579
David Ertman [Wed, 5 Mar 2014 07:50:46 +0000 (07:50 +0000)]
e1000e: Fix SHRA register access for 82579

Previous commit c3a0dce35af0 fixed an overrun for the RAR on i218 devices.
This commit also attempted to homogenize the RAR/SHRA access for all parts
accessed by the e1000e driver.  This change introduced an error for
assigning MAC addresses to guest OS's for 82579 devices.

Only RAR[0] is accessible to the driver for 82579 parts, and additional
addresses must be placed into the SHRA[L|H] registers.  The rar_entry_count
was changed in the previous commit to an inaccurate value that accounted
for all RAR and SHRA registers, not just the ones usable by the driver.

This patch fixes the count to the correct value and adjusts the
e1000_rar_set_pch2lan() function to user the correct index.

Cc: John Greene <jogreene@redhat.com>
Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Fix ethtool offline tests for 82579 parts
David Ertman [Wed, 5 Mar 2014 07:54:19 +0000 (07:54 +0000)]
e1000e: Fix ethtool offline tests for 82579 parts

Changes to the rar_entry_count value require a change to the indexing
used to access the SHRA[H|L] registers when testing them with
'ethtool -t <iface> offline'

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Fix not generating an error on invalid load parameter
David Ertman [Wed, 26 Feb 2014 00:02:24 +0000 (00:02 +0000)]
e1000e: Fix not generating an error on invalid load parameter

Valid values for InterruptThrottleRate are 10-100000, or one of
0, 1, 3, 4.  '2' is not valid.  This is a legacy from the branching
from the e1000 driver code that e1000e was based from.

Prior to this patch, if the e1000e driver  was loaded with a forced
invalid InterruptThrottleRate of '2', then no throttle rate would be
set and no error message generated.

Now, a message will be generated that an invalid value was used and the
value for InterruptThrottleRate will be set to the default value.

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Feature Enable PHY Ultra Low Power Mode (ULP)
David Ertman [Sat, 22 Feb 2014 03:15:17 +0000 (03:15 +0000)]
e1000e: Feature Enable PHY Ultra Low Power Mode (ULP)

ULP is a power saving feature that reduces the power consumption of the
PHY when a cable is not connected.

ULP is gated on the following conditions:
1) The hardware must support ULP.  Currently this is only I218
   devices from Intel
2) ULP is initiated by the driver, so, no driver results in no ULP.
3) ULP's implementation utilizes Runtime Power Management to toggle its
   execution.  ULP is enabled/disabled based on the state of Runtime PM.
4) ULP is not active when wake-on-unicast, multicast or broadcast is active
   as these features are mutually-exclusive.

Since the PHY is in an unavailable state while ULP is active, any access
of the PHY registers will fail.  This is resolved by utilizing kernel
calls that cause the device to exit Runtime PM (e.g. pm_runtime_get_sync)
and then, after PHY access is complete,  allow the device to resume
Runtime PM (e.g. pm_runtime_put_sync).

Under certain conditions, toggling the LANPHYPC is necessary to disable
ULP mode.  Break out existing code to toggle LANPHYPC to a new function
to avoid code duplication.

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Cc: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e Refactor of Runtime Power Management
David Ertman [Fri, 14 Feb 2014 07:16:46 +0000 (07:16 +0000)]
e1000e Refactor of Runtime Power Management

Fix issues with:
RuntimePM causing the device to repeatedly flip between suspend and resume
with the interface administratively downed.
Having RuntimePM enabled interfering with the functionality of Energy
Efficient Ethernet.

Added checks to disallow functions that should not be executed if the
device is currently runtime suspended

Make runtime_idle callback to use same deterministic behavior as the igb
driver.

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Refactor PM flows
David Ertman [Fri, 14 Feb 2014 07:16:41 +0000 (07:16 +0000)]
e1000e: Refactor PM flows

Refactor the system power management flows to prevent the suspend path from
being executed twice when hibernating since both the freeze and
poweroff callbacks were set to e1000_suspend() via SET_SYSTEM_SLEEP_PM_OPS.
There are HW workarounds that are performed during this flow and calling
them twice was causing erroneous behavior.

Re-arrange the code to take advantage of common code paths and explicitly
set the individual dev_pm_ops callbacks for suspend, resume, freeze,
thaw, poweroff and restore.

Add a boolean parameter (reset) to the e1000e_down function to allow
for cases when the HW should not be reset when downed during a PM event.

Now that all suspend/shutdown paths result in a call to __e1000_shutdown()
that checks Wake on Lan status, removing redundant check for WoL in
e1000_power_down_phy().

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Add missing branding strings in ich8lan.c
David Ertman [Wed, 5 Feb 2014 01:09:54 +0000 (01:09 +0000)]
e1000e: Add missing branding strings in ich8lan.c

Branding strings from recently released and soon to be released
hardware configurations that are supported by e1000e.

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Cleanup - Update GPL header and Copyright
David Ertman [Tue, 4 Feb 2014 01:56:06 +0000 (01:56 +0000)]
e1000e: Cleanup - Update GPL header and Copyright

This patch is to update the GPL header by removing the portion that
refers to the Free Software Foundation address.

Change the copyright date for 2014.

Reformat the header comments to conform to kernel networking coding norms

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Fix 82579 sets LPI too early.
David Ertman [Fri, 24 Jan 2014 23:07:48 +0000 (23:07 +0000)]
e1000e: Fix 82579 sets LPI too early.

Enabling EEE LPI sooner than one second after link up on 82579 causes link
issues with some switches.

Remove EEE enablement for 82579 parts from the link initialization flow to
avoid initializing too early.  EEE initialization for 82579 will be done
in e1000e_update_phy_task.

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Acked-by: Bruce W Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Resolve issues with Management Engine (ME) briefly blocking PHY resets
David Ertman [Thu, 23 Jan 2014 06:29:13 +0000 (06:29 +0000)]
e1000e: Resolve issues with Management Engine (ME) briefly blocking PHY resets

On a ME enabled system with the cable out, the driver init flow would
generate an erroneous message indicating that resets were being blocked
by an active ME session.  Cause was ME clearing the semaphore bit to
block further PHY resets for up to 50 msec during power-on/cycle.  After
this interval, ME would re-set the bit and allow PHY resets.

To resolve this, change the flow of e1000e_phy_hw_reset_generic() to
utilize a delay and retry method.  Poll the FWSM register to minimize
any extra time added to the flow.  If the delay times out at 100ms
(checked in 10msec increments), then return the value E1000_BLK_PHY_RESET,
as this is the accurate state of the PHY.  Attempting to alter just the
call to e1000e_phy_hw_reset_generic() in e1000_init_phy_workarounds_pchlan()
just caused the problem to move further down the flow.

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Acked-by: Bruce W. Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Cleanup unecessary references
David Ertman [Wed, 22 Jan 2014 00:21:41 +0000 (00:21 +0000)]
e1000e: Cleanup unecessary references

Cleaning up some pointer references that are no longer necessary

Signed-off-by: Dave Ertman <davidx.m.ertman@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: PTP lock in e1000e_phc_adjustfreq
Todd Fujinaka [Sat, 18 Jan 2014 06:09:33 +0000 (06:09 +0000)]
e1000e: PTP lock in e1000e_phc_adjustfreq

Add lock in e1000e_phc_adjfreq to prevent concurrent changes to TIMINCA
and SYSTIMH/L.

Signed-off-by: Todd Fujinaka <todd.fujinaka@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years ago6lowpan: reassembly: fix return of init function
Alexander Aring [Fri, 7 Mar 2014 10:06:54 +0000 (11:06 +0100)]
6lowpan: reassembly: fix return of init function

This patch adds a missing return after fragmentation init. Otherwise we
register a sysctl interface and deregister it afterwards which makes no
sense.

Signed-off-by: Alexander Aring <alex.aring@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge tag 'linux-can-next-for-3.15-20140307' of git://gitorious.org/linux-can/linux...
David S. Miller [Fri, 7 Mar 2014 22:08:57 +0000 (17:08 -0500)]
Merge tag 'linux-can-next-for-3.15-20140307' of git://gitorious.org/linux-can/linux-can-next

Marc Kleine-Budde says:

====================
pull-request: can-next 2014-02-12

this is a pull request of twelve patches for net-next/master.

Alexander Shiyan contributes two patches for the mcp251x, one making
the driver more quiet and the other one improves the compile time
coverage by removing the #ifdef CONFIG_PM_SLEEP. Then two patches for
the flexcan driver by me, one removing the #ifdef CONFIG_PM_SLEEP, too,
the other one making use of platform_get_device_id(). Another patch by
me which converts the janz-ican3 driver to use netdev_<level>(). The
remaining 7 patches are by Oliver Hartkopp, they add CAN FD support to
the netlink configuration interface.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'r8152'
David S. Miller [Fri, 7 Mar 2014 21:24:54 +0000 (16:24 -0500)]
Merge branch 'r8152'

Hayes Wang says:

====================
r8152: tx/rx improvement

 - Select the suitable spin lock for each function.
 - Add additional check to reduce the spin lock.
 - Up the priority of the tx to avoid interrupted by rx.
 - Support rx checksum, large send, and IPv6 hw checksum.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: support IPv6
hayeswang [Fri, 7 Mar 2014 03:04:40 +0000 (11:04 +0800)]
r8152: support IPv6

Support hw IPv6 checksum for TCP and UDP packets.

Note that the hw has the limitation of the range of the transport
offset. Besides, the TCP Pseudo Header of the IPv6 TSO of the hw
bases on the Microsoft document which excludes the packet length.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: support TSO
hayeswang [Fri, 7 Mar 2014 03:04:39 +0000 (11:04 +0800)]
r8152: support TSO

Support scatter gather and TSO.

Adjust the tx checksum function and set the max gso size to fix the
size of the tx aggregation buffer.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: support rx checksum
hayeswang [Fri, 7 Mar 2014 03:04:38 +0000 (11:04 +0800)]
r8152: support rx checksum

Support hw rx checksum for TCP and UDP packets.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: calculate the dropped packets for rx
hayeswang [Fri, 7 Mar 2014 03:04:37 +0000 (11:04 +0800)]
r8152: calculate the dropped packets for rx

Continue dealing with the remain rx packets, even though the allocation
of the skb fail. This could calculate the correct dropped packets.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: up the priority of the transmission
hayeswang [Fri, 7 Mar 2014 03:04:36 +0000 (11:04 +0800)]
r8152: up the priority of the transmission

move the tx_bottom() from delayed_work to tasklet. It makes the rx
and tx balanced. If the device is in runtime suspend when getting
the tx packet, wakeup the device before trasmitting.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: check tx agg list before spin lock
hayeswang [Fri, 7 Mar 2014 03:04:35 +0000 (11:04 +0800)]
r8152: check tx agg list before spin lock

Check tx agg list before spin lock to avoid doing spin lock every
times.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: replace spin_lock_irqsave and spin_unlock_irqrestore
hayeswang [Fri, 7 Mar 2014 03:04:34 +0000 (11:04 +0800)]
r8152: replace spin_lock_irqsave and spin_unlock_irqrestore

Use spin_lock and spin_unlock in interrupt context.

The ndo_start_xmit would not be called in interrupt context, so
replace the relative spin_lock_irqsave and spin_unlock_irqrestore
with spin_lock_bh and spin_unlock_bh.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Fri, 7 Mar 2014 21:12:43 +0000 (16:12 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates

This series contains updates to i40e and i40evf.

Most notable are:
Joseph completes the implementation of the ethtool ntuple rule
management interface by adding the get, update and delete interface
reset.

Akeem provides a fix to prevent a possible overflow due to multiplication
of number and size by using kzalloc, so use kcalloc.

Jesse provides an implementation for skb_set_hash() and adds the L4 type
return when we know it is an L4 hash.  He also adds a counter to
statistics for Tx timeouts to help users.  Lastly he provides a change
to stay away from the cache line where the done bit may be getting
written back for the transmit ring since the hardware may be writing the
whole cache line for a partial update.

Shannon cleans up code comments.

Anjali removes a firmware workaround for newer firmware since the number
of MSIx vectors are being reported correctly.

v2:
 -  dropped patch 01 of the series based on feedback from the author
    Joe Perches and Shannon Nelson.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge tag 'rxrpc-devel-20140304' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Fri, 7 Mar 2014 21:08:59 +0000 (16:08 -0500)]
Merge tag 'rxrpc-devel-20140304' of git://git./linux/kernel/git/dhowells/linux-fs

David Howells says:

====================
net-next: AF_RXRPC fixes and development

Here are some AF_RXRPC fixes:

 (1) Fix to remove incorrect checksum calculation made during recvmsg().  It's
     unnecessary to try to do this there since we check the checksum before
     reading the RxRPC header from the packet.

 (2) Fix to prevent the sending of an ABORT packet in response to another
     ABORT packet and inducing a storm.

 (3) Fix UDP MTU calculation from parsing ICMP_FRAG_NEEDED packets where we
     don't handle the ICMP packet not specifying an MTU size.

And development patches:

 (4) Add sysctls for configuring RxRPC parameters, specifically various delays
     pertaining to ACK generation, the time before we resend a packet for
     which we don't receive an ACK, the maximum time a call is permitted to
     live and the amount of time transport, connection and dead call
     information is cached.

 (5) Improve ACK packet production by adjusting the handling of ACK_REQUESTED
     packets, ignoring the MORE_PACKETS flag, delaying the production of
     otherwise immediate ACK_IDLE packets and delaying all ACK_IDLE production
     (barring the call termination) to half a second.

 (6) Add more sysctl parameters to expose the Rx window size, the maximum
     packet size that we're willing to receive and the number of jumbo rxrpc
     packets we're willing to handle in a single UDP packet.

 (7) Request ACKs on alternate DATA packets so that the other side doesn't
     wait till we fill up the Tx window.

 (8) Use a RCU hash table to look up the rxrpc_call for an incoming packet
     rather than stepping through a hierarchy involving several spinlocks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'xen-netback-next'
David S. Miller [Fri, 7 Mar 2014 20:57:47 +0000 (15:57 -0500)]
Merge branch 'xen-netback-next'

Zoltan Kiss says:

====================
xen-netback: TX grant mapping with SKBTX_DEV_ZEROCOPY instead of copy

A long known problem of the upstream netback implementation that on the TX
path (from guest to Dom0) it copies the whole packet from guest memory into
Dom0. That simply became a bottleneck with 10Gb NICs, and generally it's a
huge perfomance penalty. The classic kernel version of netback used grant
mapping, and to get notified when the page can be unmapped, it used page
destructors. Unfortunately that destructor is not an upstreamable solution.
Ian Campbell's skb fragment destructor patch series [1] tried to solve this
problem, however it seems to be very invasive on the network stack's code,
and therefore haven't progressed very well.
This patch series use SKBTX_DEV_ZEROCOPY flags to tell the stack it needs to
know when the skb is freed up. That is the way KVM solved the same problem,
and based on my initial tests it can do the same for us. Avoiding the extra
copy boosted up TX throughput from 6.8 Gbps to 7.9 (I used a slower AMD
Interlagos box, both Dom0 and guest on upstream kernel, on the same NUMA node,
running iperf 2.0.5, and the remote end was a bare metal box on the same 10Gb
switch)
Based on my investigations the packet get only copied if it is delivered to
Dom0 IP stack through deliver_skb, which is due to this [2] patch. This affects
DomU->Dom0 IP traffic and when Dom0 does routing/NAT for the guest. That's a bit
unfortunate, but luckily it doesn't cause a major regression for this usecase.
In the future we should try to eliminate that copy somehow.
There are a few spinoff tasks which will be addressed in separate patches:
- grant copy the header directly instead of map and memcpy. This should help
  us avoiding TLB flushing
- use something else than ballooned pages
- fix grant map to use page->index properly
I've tried to broke it down to smaller patches, with mixed results, so I
welcome suggestions on that part as well:
1: Use skb->cb to store pending_idx
2: Some refactoring
3: Change RX path for mapped SKB fragments (moved here to keep bisectability,
review it after #4)
4: Introduce TX grant mapping
5: Remove old TX grant copy definitons and fix indentations
6: Add stat counters for zerocopy
7: Handle guests with too many frags
8: Timeout packets in RX path
9: Aggregate TX unmap operations

v2: I've fixed some smaller things, see the individual patches. I've added a
few new stat counters, and handling the important use case when an older guest
sends lots of slots. Instead of delayed copy now we timeout packets on the RX
path, based on the assumption that otherwise packets should get stucked
anywhere else. Finally some unmap batching to avoid too much TLB flush

v3: Apart from fixing a few things mentioned in responses the important change
is the use the hypercall directly for grant [un]mapping, therefore we can
avoid m2p override.

v4: Now we are using a new grant mapping API to avoid m2p_override. The RX queue
timeout logic changed also.

v5: Only minor fixes based on Wei's comments

v6: Important bugfixes for xenvif_poll exit path and zerocopy callback, see
first 2 patches. Also rework of handling packets with too many slots, and
reorder the series a bit.

v7: Small fixes in comments/log messages/error paths, and merging the frag
overflow stats patch into its parent.

[1] http://lwn.net/Articles/491522/
[2] https://lkml.org/lkml/2012/7/20/363
====================

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Aggregate TX unmap operations
Zoltan Kiss [Thu, 6 Mar 2014 21:48:31 +0000 (21:48 +0000)]
xen-netback: Aggregate TX unmap operations

Unmapping causes TLB flushing, therefore we should make it in the largest
possible batches. However we shouldn't starve the guest for too long. So if
the guest has space for at least two big packets and we don't have at least a
quarter ring to unmap, delay it for at most 1 milisec.

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Timeout packets in RX path
Zoltan Kiss [Thu, 6 Mar 2014 21:48:30 +0000 (21:48 +0000)]
xen-netback: Timeout packets in RX path

A malicious or buggy guest can leave its queue filled indefinitely, in which
case qdisc start to queue packets for that VIF. If those packets came from an
another guest, it can block its slots and prevent shutdown. To avoid that, we
make sure the queue is drained in every 10 seconds.
The QDisc queue in worst case takes 3 round to flush usually.

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Handle guests with too many frags
Zoltan Kiss [Thu, 6 Mar 2014 21:48:29 +0000 (21:48 +0000)]
xen-netback: Handle guests with too many frags

Xen network protocol had implicit dependency on MAX_SKB_FRAGS. Netback has to
handle guests sending up to XEN_NETBK_LEGACY_SLOTS_MAX slots. To achieve that:
- create a new skb
- map the leftover slots to its frags (no linear buffer here!)
- chain it to the previous through skb_shinfo(skb)->frag_list
- map them
- copy and coalesce the frags into a brand new one and send it to the stack
- unmap the 2 old skb's pages

It's also introduces new stat counters, which help determine how often the guest
sends a packet with more than MAX_SKB_FRAGS frags.

NOTE: if bisect brought you here, you should apply the series up until
"xen-netback: Timeout packets in RX path", otherwise malicious guests can block
other guests by not releasing their sent packets.

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Add stat counters for zerocopy
Zoltan Kiss [Thu, 6 Mar 2014 21:48:28 +0000 (21:48 +0000)]
xen-netback: Add stat counters for zerocopy

These counters help determine how often the buffers had to be copied. Also
they help find out if packets are leaked, as if "sent != success + fail",
there are probably packets never freed up properly.

NOTE: if bisect brought you here, you should apply the series up until
"xen-netback: Timeout packets in RX path", otherwise Windows guests can't work
properly and malicious guests can block other guests by not releasing their sent
packets.

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Remove old TX grant copy definitons and fix indentations
Zoltan Kiss [Thu, 6 Mar 2014 21:48:27 +0000 (21:48 +0000)]
xen-netback: Remove old TX grant copy definitons and fix indentations

These became obsolete with grant mapping. I've left intentionally the
indentations in this way, to improve readability of previous patches.

NOTE: if bisect brought you here, you should apply the series up until
"xen-netback: Timeout packets in RX path", otherwise Windows guests can't work
properly and malicious guests can block other guests by not releasing their sent
packets.

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Introduce TX grant mapping
Zoltan Kiss [Thu, 6 Mar 2014 21:48:26 +0000 (21:48 +0000)]
xen-netback: Introduce TX grant mapping

This patch introduces grant mapping on netback TX path. It replaces grant copy
operations, ditching grant copy coalescing along the way. Another solution for
copy coalescing is introduced in "xen-netback: Handle guests with too many
frags", older guests and Windows can broke before that patch applies.
There is a callback (xenvif_zerocopy_callback) from core stack to release the
slots back to the guests when kfree_skb or skb_orphan_frags called. It feeds a
separate dealloc thread, as scheduling NAPI instance from there is inefficient,
therefore we can't do dealloc from the instance.

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Handle foreign mapped pages on the guest RX path
Zoltan Kiss [Thu, 6 Mar 2014 21:48:25 +0000 (21:48 +0000)]
xen-netback: Handle foreign mapped pages on the guest RX path

RX path need to know if the SKB fragments are stored on pages from another
domain.
Logically this patch should be after introducing the grant mapping itself, as
it makes sense only after that. But to keep bisectability, I moved it here. It
shouldn't change any functionality here. xenvif_zerocopy_callback and
ubuf_to_vif are just stubs here, they will be introduced properly later on.

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Minor refactoring of netback code
Zoltan Kiss [Thu, 6 Mar 2014 21:48:24 +0000 (21:48 +0000)]
xen-netback: Minor refactoring of netback code

This patch contains a few bits of refactoring before introducing the grant
mapping changes:
- introducing xenvif_tx_pending_slots_available(), as this is used several
  times, and will be used more often
- rename the thread to vifX.Y-guest-rx, to signify it does RX work from the
  guest point of view

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Use skb->cb for pending_idx
Zoltan Kiss [Thu, 6 Mar 2014 21:48:23 +0000 (21:48 +0000)]
xen-netback: Use skb->cb for pending_idx

Storing the pending_idx at the first byte of the linear buffer never looked
good, skb->cb is a more proper place for this. It also prevents the header to
be directly grant copied there, and we don't have the pending_idx after we
copied the header here, so it's time to change it.
It also introduces helpers for the RX side

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agol2tp: keep original skb ownership
Eric Dumazet [Thu, 6 Mar 2014 02:19:34 +0000 (18:19 -0800)]
l2tp: keep original skb ownership

There is no reason to orphan skb in l2tp.

This breaks things like per socket memory limits, TCP Small queues...

Fix this before more people copy/paste it.

This is very similar to commit 8f646c922d550
("vxlan: keep original skb ownership")

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: do not leak non zero tstamp in output packets
Eric Dumazet [Wed, 5 Mar 2014 22:08:38 +0000 (14:08 -0800)]
tcp: do not leak non zero tstamp in output packets

Usage of skb->tstamp should remain private to TCP stack
(only set on packets on write queue, not on cloned ones)

Otherwise, packets given to loopback interface with a non null tstamp
can confuse netif_rx() / net_timestamp_check()

Other possibility would be to clear tstamp in loopback_xmit(),
as done in skb_scrub_packet()

Fixes: 740b0f1841f6 ("tcp: switch rtt estimations to usec resolution")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocan: add bittiming check at interface open for CAN FD
Oliver Hartkopp [Fri, 28 Feb 2014 15:36:25 +0000 (16:36 +0100)]
can: add bittiming check at interface open for CAN FD

Additionally to have the second (data) bitrate available the data bitrate
has to be greater or equal to the arbitration bitrate in CAN FD.

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Stephane Grosjean <s.grosjean@peak-system.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: allow to change the device mtu for CAN FD capable devices
Oliver Hartkopp [Fri, 28 Feb 2014 15:36:24 +0000 (16:36 +0100)]
can: allow to change the device mtu for CAN FD capable devices

The configuration for CAN FD depends on CAN_CTRLMODE_FD enabled in the driver
specific ctrlmode_supported capabilities.

The configuration can be done either with the 'fd { on | off }' option in the
'ip' tool from iproute2 or by setting the CAN netdevice MTU to CAN_MTU (16) or
to CANFD_MTU (72).

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Stephane Grosjean <s.grosjean@peak-system.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: introduce the data bitrate configuration for CAN FD
Oliver Hartkopp [Fri, 28 Feb 2014 15:36:23 +0000 (16:36 +0100)]
can: introduce the data bitrate configuration for CAN FD

As CAN FD offers a second bitrate for the data section of the CAN frame the
infrastructure for storing and configuring this second bitrate is introduced.
Improved the readability of the if-statement by inserting some newlines.

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Stephane Grosjean <s.grosjean@peak-system.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: provide a separate bittiming_const parameter to bittiming functions
Oliver Hartkopp [Fri, 28 Feb 2014 15:36:22 +0000 (16:36 +0100)]
can: provide a separate bittiming_const parameter to bittiming functions

As the bittiming calculation functions are to be used with different
bittiming_const structures for CAN and CAN FD the direct reference to
priv->bittiming_const inside these functions has to be removed.

Also moved the check for existing bittiming const to one place.

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Stephane Grosjean <s.grosjean@peak-system.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: move sanity check for bitrate and tq into can_get_bittiming
Oliver Hartkopp [Fri, 28 Feb 2014 15:36:21 +0000 (16:36 +0100)]
can: move sanity check for bitrate and tq into can_get_bittiming

This patch moves a sanity check in order to have a second user for CAN FD.
Also simplify the return value generation in can_get_bittiming() as only
correct return values of can_[calc|fixup]_bittiming() lead to a return value of
zero.

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Stephane Grosjean <s.grosjean@peak-system.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: only send bitrate data via netlink when available
Oliver Hartkopp [Fri, 28 Feb 2014 15:36:20 +0000 (16:36 +0100)]
can: only send bitrate data via netlink when available

When setting the bitrate both can_calc_bittiming() and can_fixup_bittiming()
lead to the bitrate variable to be set, when a proper bit timing is available.
Only then the bitrate configuration is stored for the device, so checking for
priv->bittiming.bitrate is always sufficient.

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Stephane Grosjean <s.grosjean@peak-system.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: preserve skbuff protocol in can_put_echo_skb
Oliver Hartkopp [Fri, 28 Feb 2014 15:36:19 +0000 (16:36 +0100)]
can: preserve skbuff protocol in can_put_echo_skb

The skbuff protocol value was formerly fixed/sanitized to ETH_P_CAN in
can_put_echo_skb(). With CAN FD this value has to be preserved.
This patch changes the hard assignment of the protocol value to a check of
valid protocol values for CAN and CAN FD.

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Stephane Grosjean <s.grosjean@peak-system.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: janz-ican3: convert dev_<level> printing to netdev_<level>
Marc Kleine-Budde [Tue, 4 Mar 2014 21:00:47 +0000 (22:00 +0100)]
can: janz-ican3: convert dev_<level> printing to netdev_<level>

This patch converts the dev_<level> printing to netdev_<level>, this makes it
possible to remove the "struct device *dev" pointer from the "struct
ican3_dev".

Cc: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agoi40e/i40evf: Bump pf&vf build versions
Catherine Sullivan [Thu, 6 Feb 2014 05:51:14 +0000 (05:51 +0000)]
i40e/i40evf: Bump pf&vf build versions

Bump i40e to 0.3.34 and i40evf to 0.9.14.

Change-ID: I6b3fb8ccf55b128d2baa4bdc20d3911ec81d4a5b
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e/i40evf: carefully fill tx ring
Jesse Brandeburg [Thu, 6 Feb 2014 05:51:13 +0000 (05:51 +0000)]
i40e/i40evf: carefully fill tx ring

We need to make sure that we stay away from the cache line
where the DD bit (done) may be getting written back for
the transmit ring since the hardware may be writing the
whole cache line for a partial update.

Change-ID: Id0b6dfc01f654def6a2a021af185803be1915d7e
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: fix nvm version and remove firmware report
Jesse Brandeburg [Thu, 6 Feb 2014 05:51:12 +0000 (05:51 +0000)]
i40e: fix nvm version and remove firmware report

The driver needs to use the format that the current NVM
uses when printing the version of the NVM.  It should remain
this way from now on forward.

The driver was reporting when firmware was less than
an expected version number, but this is not a requirement
for the product and we print the firmware number at
init and in ethtool -i output.  Just remove the print.

Change-ID: Ide0b856cd454ebf867610ef9a0d639bb358a4a60
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Fix static checker warning
Neerav Parikh [Thu, 6 Feb 2014 05:51:11 +0000 (05:51 +0000)]
i40e: Fix static checker warning

This patch fixes the following static checker warning:

  drivers/net/ethernet/intel/i40e/i40e_dcb.c:342
  i40e_lldp_to_dcb_config() warn: 'tlv' can't be NULL.

Exit criteria from the while loop is encountering LLDP END
LV or if the TLV length goes beyond the buffer length.

Change-ID: I7548b16db90230ec2ba0fa791b0343ca8b7dd5bb
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Neerav Parikh <Neerav.Parikh@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Kevin Scott <kevin.c.scott@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Tested-By: Jack Morgan<jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Remove a redundant filter addition
Anjali Singhai Jain [Thu, 6 Feb 2014 05:51:10 +0000 (05:51 +0000)]
i40e: Remove a redundant filter addition

Remove a redundant filter addition to stop FW complaints about a redundant
filter removal.

Change-ID: I22bef6b682bd8d43432557e6e2b3e73ffb27b985
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: count timeout events
Jesse Brandeburg [Thu, 6 Feb 2014 05:51:09 +0000 (05:51 +0000)]
i40e: count timeout events

The ethtool -S statistics should have a counter for
tx timeouts in order to better help inform the masses.

Change-ID: Ice4b20ed4a151509f366719ab105be49c9e7b2b4
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Remove a FW workaround for Number of MSIX vectors
Anjali Singhai Jain [Thu, 6 Feb 2014 05:51:08 +0000 (05:51 +0000)]
i40e: Remove a FW workaround for Number of MSIX vectors

The Number of MSIX vectors being reported is correct and hence
we need a check to do the right thing for FWs before and after.

Change-ID: I50902d1c848adcb960ea49ac73f7865ca871a1c3
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: clean up comment style
Shannon Nelson [Thu, 6 Feb 2014 05:51:06 +0000 (05:51 +0000)]
i40e: clean up comment style

Lots of trivial changes to remove double spaces in function headers,
unnecessary periods in short comments, and adjust the English usage here
and there.

No actual code was harmed in the making of this patch.

Change-ID: I6e756c500756945e81a61ffb10221753eb7923ea
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Kevin Scott <kevin.c.scott@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e/i40evf: i40e implementation for skb_set_hash
Jesse Brandeburg [Wed, 12 Feb 2014 01:45:33 +0000 (01:45 +0000)]
i40e/i40evf: i40e implementation for skb_set_hash

Original comment from Tom Herbert <therbert@google.com>

   Drivers should call skb_set_hash to set the hash and its type
   in an skbuff.

This patch builds upon Tom's original implementation and adds
the L4 type return when we know it is an L4 hash.
This requires use of the ptype decoder ring, so enable it.

Change-ID: I2f9fa86d1a6add58cff13386f7f4238b1abcc468
CC: Tom Herbert <therbert@google.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Acked-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Prevent overflow due to kzalloc
Akeem G Abodunrin [Thu, 6 Feb 2014 05:51:02 +0000 (05:51 +0000)]
i40e: Prevent overflow due to kzalloc

To prevent the possibility of overflow due multiplication of number and size
use kcalloc instead of kzalloc.

Change-ID: Ibe4d81ed7d9738d3bbe66ee4844ff9be817e8080
Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Flow Director sideband accounting
Joseph Gasparakis [Wed, 12 Feb 2014 01:45:30 +0000 (01:45 +0000)]
i40e: Flow Director sideband accounting

This patch completes implementation of the ethtool ntuple
rule management interface.  It adds the get, update and delete
interface reset.

Change-ID: Ida7f481d9ee4e405ed91340b858eabb18a52fdb5
Signed-off-by: Joseph Gasparakis <joseph.gasparakis@intel.com>
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40evf: Enable the ndo_set_features netdev op
Greg Rose [Thu, 30 Jan 2014 12:40:27 +0000 (12:40 +0000)]
i40evf: Enable the ndo_set_features netdev op

Set netdev->hw_features to enable the ndo_set_features netdev op.

Change-Id: I5a086fbfa5a089de5adba2800c4d0b3a73747b11
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agobonding: fix const in options processing
stephen hemminger [Thu, 6 Mar 2014 22:20:17 +0000 (14:20 -0800)]
bonding: fix const in options processing

This is a fixup patch to resolve issues with const from my earlier patch.
Make all the setter functions use const on input parameter.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet_sched: htb: do not acquire qdisc lock in dump operations
Eric Dumazet [Wed, 5 Mar 2014 18:14:34 +0000 (10:14 -0800)]
net_sched: htb: do not acquire qdisc lock in dump operations

htb_dump() and htb_dump_class() do not strictly need to acquire
qdisc lock to fetch qdisc and/or class parameters.

We hold RTNL and no changes can occur.

This reduces by 50% qdisc lock pressure while doing tc qdisc|class dump
operations.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch '6lowpan'
David S. Miller [Thu, 6 Mar 2014 22:21:43 +0000 (17:21 -0500)]
Merge branch '6lowpan'

Alexander Aring says:

====================
6lowpan: header cleanup

this patch series fix a missing include of 6LoWPAN header and move it
into the include/net directory. Since we did some code sharing with
bluetooth 6LoWPAN the header turns into a generic header for 6LoWPAN.
Instead to use a relative path in bluetooth 6LoWPAN we can now use
include <net/6lowpan.h>.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years ago6lowpan: move 6lowpan header to include/net
Alexander Aring [Wed, 5 Mar 2014 13:29:05 +0000 (14:29 +0100)]
6lowpan: move 6lowpan header to include/net

This header is used by bluetooth and ieee802154 branch. This patch
move this header to the include/net directory to avoid a use of a
relative path in include.

Signed-off-by: Alexander Aring <alex.aring@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years ago6lowpan: add missing include of net/ipv6.h
Alexander Aring [Wed, 5 Mar 2014 13:29:04 +0000 (14:29 +0100)]
6lowpan: add missing include of net/ipv6.h

The 6lowpan.h file contains some static inline function which use
internal ipv6 api structs. Add a include of ipv6.h to be sure that it's
known before.

Signed-off-by: Alexander Aring <alex.aring@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobe2net: do external loopback test only when it is requested
Ivan Vecera [Wed, 5 Mar 2014 10:54:05 +0000 (11:54 +0100)]
be2net: do external loopback test only when it is requested

v2: remove unnecessary braces from all 'loopback' if-blocks (thx Sergei)

Cc: sathya.perla@emulex.com
Cc: subbu.seetharaman@emulex.com
Cc: ajit.khaparde@emulex.com
Cc: sergei.shtylyov@cogentembedded.com
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Acked-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: mlx4_en_verify_params() can be static
Fengguang Wu [Wed, 5 Mar 2014 09:57:23 +0000 (11:57 +0200)]
net/mlx4_en: mlx4_en_verify_params() can be static

Fix static error introduced by commit:
b97b33a3df0439401f80f041eda507d4fffa0dbf [645/653] net/mlx4_en: Verify
mlx4_en module parameters

sparse warnings:
drivers/net/ethernet/mellanox/mlx4/en_main.c:335:6: sparse: symbol
'mlx4_en_verify_params' was not declared. Should it be static?

CC: netdev@vger.kernel.org
CC: linux-kernel@vger.kernel.org
CC: Eugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocan: flexcan: make use of platform_get_device_id()
Marc Kleine-Budde [Tue, 4 Mar 2014 21:04:22 +0000 (22:04 +0100)]
can: flexcan: make use of platform_get_device_id()

This patch replaces an open coded pdev->id_entry by platform_get_device_id().

Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: flexcan: Remove #ifdef CONFIG_PM_SLEEP
Marc Kleine-Budde [Wed, 5 Mar 2014 18:10:44 +0000 (19:10 +0100)]
can: flexcan: Remove #ifdef CONFIG_PM_SLEEP

This patch removes #ifdef CONFIG_PM_SLEEP to improve compile coverage.

Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: mcp251x: Remove #ifdef CONFIG_PM_SLEEP
Alexander Shiyan [Sat, 22 Feb 2014 05:51:14 +0000 (09:51 +0400)]
can: mcp251x: Remove #ifdef CONFIG_PM_SLEEP

This patch removes #ifdef CONFIG_PM_SLEEP to improve compile coverage.

Signed-off-by: Alexander Shiyan <shc_work@mail.ru>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agocan: mcp251x: Make driver more quiet
Alexander Shiyan [Wed, 5 Mar 2014 17:31:56 +0000 (21:31 +0400)]
can: mcp251x: Make driver more quiet

This patch moves one diagnostic message used for debugging purposes
to dev_dbg() and removes one useless message.

Signed-off-by: Alexander Shiyan <shc_work@mail.ru>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agobonding: options handling cleanup
stephen hemminger [Wed, 5 Mar 2014 00:36:44 +0000 (16:36 -0800)]
bonding: options handling cleanup

Make local functions static (ie. only used in bond_options.c)
Make bond options parsing tables constant.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: remove dead code
stephen hemminger [Wed, 5 Mar 2014 00:34:25 +0000 (16:34 -0800)]
bonding: remove dead code

These functions are defined but no longer used.
Compile tested only.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: Use NET_ADD_STATS instead of NET_ADD_STATS_BH in tcp_event_new_data_sent()
David S. Miller [Thu, 6 Mar 2014 20:03:17 +0000 (15:03 -0500)]
tcp: Use NET_ADD_STATS instead of NET_ADD_STATS_BH in tcp_event_new_data_sent()

Can be invoked from non-BH context.

Based upon a patch by Eric Dumazet.

Fixes: f19c29e3e391 ("tcp: snmp stats for Fast Open, SYN rtx, and data pkts")
Reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: make slave status notifications GFP_ATOMIC
Veaceslav Falico [Thu, 6 Mar 2014 14:33:22 +0000 (15:33 +0100)]
bonding: make slave status notifications GFP_ATOMIC

Currently we're using GFP_KERNEL, however there are some path(s) where we
can hold some spinlocks, specifically bond->curr_slave_lock:

[    4.722916] BUG: sleeping function called from invalid context at mm/slub.c:965
[    4.724438] in_atomic(): 1, irqs_disabled(): 0, pid: 940, name: ifup-eth
[    4.726034] 5 locks held by ifup-eth/940:
...snip...
[    4.734646]  #4:  (&bond->curr_slave_lock){+...+.}, at: [<ffffffffa00badc6>] bond_enslave+0xda6/0xdd0 [bonding]
...snip...
[    4.759081]  [<ffffffffa00b6f11>] bond_change_active_slave+0x191/0x3b0 [bonding]
[    4.760917]  [<ffffffffa00b7227>] bond_select_active_slave+0xf7/0x1d0 [bonding]
[    4.762751]  [<ffffffffa00badce>] bond_enslave+0xdae/0xdd0 [bonding]
...snip...

As it's out of hot path and is a really rare event - change the gfp_t flags
to GFP_ATOMIC to avoid sleeping under spinlock.

v2: convert new notify calls to GFP_ATOMIC.

CC: Thomas Glanzmann <thomas@glanzmann.de>
CC: Ding Tianhong <dingtianhong@huawei.com>
CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoinet: remove now unused flag DST_NOPEER
Hannes Frederic Sowa [Thu, 6 Mar 2014 08:11:07 +0000 (09:11 +0100)]
inet: remove now unused flag DST_NOPEER

Commit e688a604807647 ("net: introduce DST_NOPEER dst flag") introduced
DST_NOPEER because because of crashes in ipv6_select_ident called from
udp6_ufo_fragment.

Since commit 916e4cf46d0204 ("ipv6: reuse ip6_frag_id from
ip6_ufo_append_data") we don't call ipv6_select_ident any more from
ip6_ufo_append_data, thus this flag lost its purpose and can be removed.

Cc: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'r8152'
David S. Miller [Thu, 6 Mar 2014 18:15:17 +0000 (13:15 -0500)]
Merge branch 'r8152'

Hayes Wang says:

====================
r8152: cleanups

Deal with some empty lines and spaces, replace some tp->netdev with netdev,
and remove the unnecessary function.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: remove rtl8152_get_stats
hayeswang [Thu, 6 Mar 2014 07:07:18 +0000 (15:07 +0800)]
r8152: remove rtl8152_get_stats

The rtl8152_get_stats() returns the point address of the struct
net_device_stats. This could be got from struct net_device directly.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: replace tp->netdev with netdev
hayeswang [Thu, 6 Mar 2014 07:07:17 +0000 (15:07 +0800)]
r8152: replace tp->netdev with netdev

Replace some tp->netdev with netdev.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: deal with the empty line and space
hayeswang [Thu, 6 Mar 2014 07:07:16 +0000 (15:07 +0800)]
r8152: deal with the empty line and space

Add or remove some empty lines. Replace the spaces with the tabs.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Thu, 6 Mar 2014 01:32:02 +0000 (20:32 -0500)]
Merge git://git./linux/kernel/git/davem/net

Conflicts:
drivers/net/wireless/ath/ath9k/recv.c
drivers/net/wireless/mwifiex/pcie.c
net/ipv6/sit.c

The SIT driver conflict consists of a bug fix being done by hand
in 'net' (missing u64_stats_init()) whilst in 'net-next' a helper
was created (netdev_alloc_pcpu_stats()) which takes care of this.

The two wireless conflicts were overlapping changes.

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoieee802154: fix whitespace issues in Kconfig
Alexander Aring [Tue, 4 Mar 2014 18:01:18 +0000 (19:01 +0100)]
ieee802154: fix whitespace issues in Kconfig

This patch fixes some whitespace issues in Kconfig files of IEEE
802.15.4 subsytem.

Signed-off-by: Alexander Aring <alex.aring@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoat86rf230: add help and 212 to Kconfig menu entry
Alexander Aring [Tue, 4 Mar 2014 18:01:17 +0000 (19:01 +0100)]
at86rf230: add help and 212 to Kconfig menu entry

Since commit 8fad346f366a72978ea942abd06bd501ebd39c22
(ieee802154: add basic support for RF212 to at86rf230 driver)

we support at86rf212 as well.

Signed-off-by: Alexander Aring <alex.aring@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobe2net: dma_sync each RX frag before passing it to the stack
Sathya Perla [Tue, 4 Mar 2014 06:44:38 +0000 (12:14 +0530)]
be2net: dma_sync each RX frag before passing it to the stack

The driver currently maps a page for DMA, divides the page into multiple
frags and posts them to the HW. It un-maps the page after data is received
on all the frags of the page. This scheme doesn't work when bounce buffers
are used for DMA (swiotlb=force kernel param).

This patch fixes this problem by calling dma_sync_single_for_cpu() for each
frag (excepting the last one) so that the data is copied from the bounce
buffers. The page is un-mapped only when DMA finishes on the last frag of
the page.
(Thanks Ben H. for suggesting the dma_sync API!)

This patch also renames the "last_page_user" field of be_rx_page_info{}
struct to "last_frag" to improve readability of the fixed code.

Reported-by: Li Fengmao <li.fengmao@zte.com.cn>
Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'mpls_tc'
David S. Miller [Tue, 4 Mar 2014 18:51:13 +0000 (13:51 -0500)]
Merge branch 'mpls_tc'

Simon Wunderlich says:

====================
this series contains a header file proposal for MPLS labels. These
labels do not seem to be properly defined in the kernel so far. We are
developing a wired/wireless 802.21/MPLS switch and need to check the
MPLS labels to use the traffic control info for transmissions over
802.11 networks.

Changes to third version:

 * rename mpls_label_stack to mpls_label (thanks Neil)
 * fix over-indendented closing brac (thanks Sergei)
 * add Johannes' Ack
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocfg80211: add MPLS and 802.21 classification
Simon Wunderlich [Mon, 3 Mar 2014 16:23:12 +0000 (17:23 +0100)]
cfg80211: add MPLS and 802.21 classification

MPLS labels may contain traffic control information, which should be
evaluated and used by the wireless subsystem if present.

Also check for IEEE 802.21 which is always network control traffic.

Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de>
Signed-off-by: Mathias Kretschmer <mathias.kretschmer@fokus.fraunhofer.de>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoUAPI: add MPLS label stack definition
Simon Wunderlich [Mon, 3 Mar 2014 16:23:11 +0000 (17:23 +0100)]
UAPI: add MPLS label stack definition

Labels for the Multiprotocol Label Switching are defined in RFC 3032
which was superseded by RFC 5462. Add the definition to UAPI and a stub
header for include/linux.

Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de>
Signed-off-by: Mathias Kretschmer <mathias.kretschmer@fokus.fraunhofer.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoif_ether.h: add IEEE 802.21 Ethertype
Simon Wunderlich [Mon, 3 Mar 2014 16:23:10 +0000 (17:23 +0100)]
if_ether.h: add IEEE 802.21 Ethertype

Add the Ethertype for IEEE Std 802.21 - Media Independent Handover
Protocol. This Ethertype is used for network control messages.

Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de>
Signed-off-by: Mathias Kretschmer <mathias.kretschmer@fokus.fraunhofer.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Tue, 4 Mar 2014 16:44:32 +0000 (08:44 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Fix memory leak in ieee80211_prep_connection(), sta_info leaked on
    error.  From Eytan Lifshitz.

 2) Unintentional switch case fallthrough in nft_reject_inet_eval(),
    from Patrick McHardy.

 3) Must check if payload lenth is a power of 2 in
    nft_payload_select_ops(), from Nikolay Aleksandrov.

 4) Fix mis-checksumming in xen-netfront driver, ip_hdr() is not in the
    correct place when we invoke skb_checksum_setup().  From Wei Liu.

 5) TUN driver should not advertise HW vlan offload features in
    vlan_features.  Fix from Fernando Luis Vazquez Cao.

 6) IPV6_VTI needs to select NET_IPV_TUNNEL to avoid build errors, fix
    from Steffen Klassert.

 7) Add missing locking in xfrm_migrade_state_find(), we must hold the
    per-namespace xfrm_state_lock while traversing the lists.  Fix from
    Steffen Klassert.

 8) Missing locking in ath9k driver, access to tid->sched must be done
    under ath_txq_lock().  Fix from Stanislaw Gruszka.

 9) Fix two bugs in TCP fastopen.  First respect the size argument given
    to tcp_sendmsg() in the fastopen path, and secondly prevent
    tcp_send_syn_data() from potentially using order-5 allocations.
    From Eric Dumazet.

10) Fix handling of default neigh garbage collection params, from Jiri
    Pirko.

11) Fix cwnd bloat and over-inflation of RTT when transmit segmentation
    is in use.  From Eric Dumazet.

12) Missing initialization of Realtek r8169 driver's statistics
    seqlocks.  Fix from Kyle McMartin.

13) Fix RTNL assertion failures in 802.3ad and AB ARP monitor of bonding
    driver, from Ding Tianhong.

14) Bonding slave release race can cause divide by zero, fix from
    Nikolay Aleksandrov.

15) Overzealous return from neigh_periodic_work() causes reachability
    time to not be computed.  Fix from Duain Jiong.

16) Fix regression in ipv6_find_hdr(), it should not return -ENOENT when
    a specific target is specified and found.  From Hans Schillstrom.

17) Fix VLAN tag stripping regression in BNA driver, from Ivan Vecera.

18) Tail loss probe can calculate bogus RTTs due to missing packet
    marking on retransmit.  Fix from Yuchung Cheng.

19) We cannot do skb_dst_drop() in iptunnel_pull_header() because
    multicast loopback detection in later code paths need access to
    skb_rtable().  Fix from Xin Long.

20) The macvlan driver regresses in that it propagates lower device
    offload support disables into itself, causing severe slowdowns when
    running over a bridge.  Provide the software offloads always on
    macvlan devices to deal with this and the regression is gone.  From
    Vlad Yasevich.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (103 commits)
  macvlan: Add support for 'always_on' offload features
  net: sctp: fix sctp_sf_do_5_1D_ce to verify if we/peer is AUTH capable
  ip_tunnel:multicast process cause panic due to skb->_skb_refdst NULL pointer
  net: cpsw: fix cpdma rx descriptor leak on down interface
  be2net: isolate TX workarounds not applicable to Skyhawk-R
  be2net: Fix skb double free in be_xmit_wrokarounds() failure path
  be2net: clear promiscuous bits in adapter->flags while disabling promiscuous mode
  be2net: Fix to reset transparent vlan tagging
  qlcnic: dcb: a couple off by one bugs
  tcp: fix bogus RTT on special retransmission
  hsr: off by one sanity check in hsr_register_frame_in()
  can: remove CAN FD compatibility for CAN 2.0 sockets
  can: flexcan: factor out soft reset into seperate funtion
  can: flexcan: flexcan_remove(): add missing netif_napi_del()
  can: flexcan: fix transition from and to freeze mode in chip_{,un}freeze
  can: flexcan: factor out transceiver {en,dis}able into seperate functions
  can: flexcan: fix transition from and to low power mode in chip_{en,dis}able
  can: flexcan: flexcan_open(): fix error path if flexcan_chip_start() fails
  can: flexcan: fix shutdown: first disable chip, then all interrupts
  USB AX88179/178A: Support D-Link DUB-1312
  ...

10 years agoMerge tag 'regulator-v3.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 4 Mar 2014 16:41:42 +0000 (08:41 -0800)]
Merge tag 'regulator-v3.14-rc5' of git://git./linux/kernel/git/broonie/regulator

Pull regulator fixes from Mark Brown:
 "A couple of fixes here which ensure that regulators using the core
  support for GPIO enables work in all cases by ensuring that helpers
  are used consistently rather than open coding in places and hence not
  having GPIO support in some of them"

* tag 'regulator-v3.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: core: Replace direct ops->disable usage
  regulator: core: Replace direct ops->enable usage

10 years agoMerge branch 'akpm' (patches from Andrew Morton)
Linus Torvalds [Tue, 4 Mar 2014 16:29:39 +0000 (08:29 -0800)]
Merge branch 'akpm' (patches from Andrew Morton)

Merge misc fixes from Andrew Morton.

* emailed patches from Andrew Morton akpm@linux-foundation.org>:
  mm: page_alloc: exempt GFP_THISNODE allocations from zone fairness
  mm: numa: bugfix for LAST_CPUPID_NOT_IN_PAGE_FLAGS
  MAINTAINERS: add and correct types of some "T:" entries
  MAINTAINERS: use tab for separator
  rapidio/tsi721: fix tasklet termination in dma channel release
  hfsplus: fix remount issue
  zram: avoid null access when fail to alloc meta
  sh: prefix sh-specific "CCR" and "CCR2" by "SH_"
  ocfs2: fix quota file corruption
  drivers/rtc/rtc-s3c.c: fix incorrect way of save/restore of S3C2410_TICNT for TYPE_S3C64XX
  kallsyms: fix absolute addresses for kASLR
  scripts/gen_initramfs_list.sh: fix flags for initramfs LZ4 compression
  mm: include VM_MIXEDMAP flag in the VM_SPECIAL list to avoid m(un)locking
  memcg: reparent charges of children before processing parent
  memcg: fix endless loop in __mem_cgroup_iter_next()
  lib/radix-tree.c: swapoff tmpfs radix_tree: remember to rcu_read_unlock
  dma debug: account for cachelines and read-only mappings in overlap tracking
  mm: close PageTail race
  MAINTAINERS: EDAC: add Mauro and Borislav as interim patch collectors

10 years agomm: page_alloc: exempt GFP_THISNODE allocations from zone fairness
Johannes Weiner [Mon, 3 Mar 2014 23:38:41 +0000 (15:38 -0800)]
mm: page_alloc: exempt GFP_THISNODE allocations from zone fairness

Jan Stancek reports manual page migration encountering allocation
failures after some pages when there is still plenty of memory free, and
bisected the problem down to commit 81c0a2bb515f ("mm: page_alloc: fair
zone allocator policy").

The problem is that GFP_THISNODE obeys the zone fairness allocation
batches on one hand, but doesn't reset them and wake kswapd on the other
hand.  After a few of those allocations, the batches are exhausted and
the allocations fail.

Fixing this means either having GFP_THISNODE wake up kswapd, or
GFP_THISNODE not participating in zone fairness at all.  The latter
seems safer as an acute bugfix, we can clean up later.

Reported-by: Jan Stancek <jstancek@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: <stable@kernel.org> [3.12+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agomm: numa: bugfix for LAST_CPUPID_NOT_IN_PAGE_FLAGS
Liu Ping Fan [Mon, 3 Mar 2014 23:38:39 +0000 (15:38 -0800)]
mm: numa: bugfix for LAST_CPUPID_NOT_IN_PAGE_FLAGS

When doing some numa tests on powerpc, I triggered an oops bug.  I find
it is caused by using page->_last_cpupid.  It should be initialized as
"-1 & LAST_CPUPID_MASK", but not "-1".  Otherwise, in task_numa_fault(),
we will miss the checking (last_cpupid == (-1 & LAST_CPUPID_MASK)).  And
finally cause an oops bug in task_numa_group(), since the online cpu is
less than possible cpu.  This happen with CONFIG_SPARSE_VMEMMAP disabled

Call trace:

  SMP NR_CPUS=64 NUMA PowerNV
  Modules linked in:
  CPU: 24 PID: 804 Comm: systemd-udevd Not tainted3.13.0-rc1+ #32
  task: c000001e2746aa80 ti: c000001e32c50000 task.ti:c000001e32c50000
  REGS: c000001e32c53510 TRAP: 0300   Not tainted(3.13.0-rc1+)
  MSR: 9000000000009032 <SF,HV,EE,ME,IR,DR,RI>  CR:28024424  XER: 20000000
  CFAR: c000000000009324 DAR: 7265717569726857 DSISR:40000000 SOFTE: 1
  NIP  .task_numa_fault+0x1470/0x2370
  LR  .task_numa_fault+0x1468/0x2370
  Call Trace:
   .task_numa_fault+0x1468/0x2370 (unreliable)
   .do_numa_page+0x480/0x4a0
   .handle_mm_fault+0x4ec/0xc90
   .do_page_fault+0x3a8/0x890
   handle_page_fault+0x10/0x30
  Instruction dump:
  3c82fefb 3884b138 48d9cff1 60000000 48000574 3c62fefb3863af78 3c82fefb
  3884b138 48d9cfd5 60000000 e93f0100 <812902e47d2907b45529063e 7d2a07b4
  ---[ end trace 15f2510da5ae07cf ]---

Signed-off-by: Liu Ping Fan <pingfank@linux.vnet.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoMAINTAINERS: add and correct types of some "T:" entries
Joe Perches [Mon, 3 Mar 2014 23:38:38 +0000 (15:38 -0800)]
MAINTAINERS: add and correct types of some "T:" entries

Tree location entries should start with the appropriate type.

Add git to some, hg to another.

Neaten tree type description.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoMAINTAINERS: use tab for separator
Joe Perches [Mon, 3 Mar 2014 23:38:37 +0000 (15:38 -0800)]
MAINTAINERS: use tab for separator

Convert whitespace to single tab for separators.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agorapidio/tsi721: fix tasklet termination in dma channel release
Alexandre Bounine [Mon, 3 Mar 2014 23:38:36 +0000 (15:38 -0800)]
rapidio/tsi721: fix tasklet termination in dma channel release

This patch is a modification of the patch originally proposed by
Xiaotian Feng <xtfeng@gmail.com>: https://lkml.org/lkml/2012/11/5/413
This new version disables DMA channel interrupts and ensures that the
tasklet wil not be scheduled again before calling tasklet_kill().

Unfortunately the updated patch was not released at that time due to
planned rework of Tsi721 mport driver to use threaded interrupts (which
has yet to happen).  Recently the issue was reported again:
https://lkml.org/lkml/2014/2/19/762.

Description from the original Xiaotian's patch:

 "Some drivers use tasklet_disable in device remove/release process,
  tasklet_disable will inc tasklet->count and return.  If the tasklet is
  not handled yet under some softirq pressure, the tasklet will be
  placed on the tasklet_vec, never have a chance to be excuted.  This
  might lead to a heavy loaded ksoftirqd, wakeup with pending_softirq,
  but tasklet is disabled.  tasklet_kill should be used in this case."

This patch is applicable to kernel versions starting from v3.5.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Xiaotian Feng <xtfeng@gmail.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Mike Galbraith <bitbucket@online.de>
Cc: <stable@vger.kernel.org> [3.5+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agohfsplus: fix remount issue
Vyacheslav Dubeyko [Mon, 3 Mar 2014 23:38:35 +0000 (15:38 -0800)]
hfsplus: fix remount issue

Current implementation of HFS+ driver has small issue with remount
option.  Namely, for example, you are unable to remount from RO mode
into RW mode by means of command "mount -o remount,rw /dev/loop0
/mnt/hfsplus".  Trying to execute sequence of commands results in an
error message:

  mount /dev/loop0 /mnt/hfsplus
  mount -o remount,ro /dev/loop0 /mnt/hfsplus
  mount -o remount,rw /dev/loop0 /mnt/hfsplus

  mount: you must specify the filesystem type

  mount -t hfsplus -o remount,rw /dev/loop0 /mnt/hfsplus

  mount: /mnt/hfsplus not mounted or bad option

The reason of such issue is failure of mount syscall:

  mount("/dev/loop0", "/mnt/hfsplus", 0x2282a60, MS_MGC_VAL|MS_REMOUNT, NULL) = -1 EINVAL (Invalid argument)

Namely, hfsplus_parse_options_remount() method receives empty "input"
argument and return false in such case.  As a result, hfsplus_remount()
returns -EINVAL error code.

This patch fixes the issue by means of return true for the case of empty
"input" argument in hfsplus_parse_options_remount() method.

Signed-off-by: Vyacheslav Dubeyko <slava@dubeyko.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agozram: avoid null access when fail to alloc meta
Minchan Kim [Mon, 3 Mar 2014 23:38:34 +0000 (15:38 -0800)]
zram: avoid null access when fail to alloc meta

zram_meta_alloc could fail so caller should check it.  Otherwise, your
system will hang.

Signed-off-by: Minchan Kim <minchan@kernel.org>
Acked-by: Jerome Marchand <jmarchan@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agosh: prefix sh-specific "CCR" and "CCR2" by "SH_"
Geert Uytterhoeven [Mon, 3 Mar 2014 23:38:33 +0000 (15:38 -0800)]
sh: prefix sh-specific "CCR" and "CCR2" by "SH_"

Commit bcf24e1daa94 ("mmc: omap_hsmmc: use the generic config for
omap2plus devices"), enabled the build for other platforms for compile
testing.

sh-allmodconfig now fails with:

    include/linux/omap-dma.h:171:8: error: expected identifier before numeric constant
    make[4]: *** [drivers/mmc/host/omap_hsmmc.o] Error 1

This happens because SuperH #defines "CCR", which is one of the enum
values in include/linux/omap-dma.h.  There's a similar issue with "CCR2"
on sh2a.

As "CCR" and "CCR2" are too generic names for global #defines, prefix
them with "SH_" to fix this.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>