This patch fixes a bug which would trigger the BUG_ON() at
net/core/dev.c:4156. It was found that this was due to continuing
processing in the current poll call even when the call to
napi_reschedule failed, indicating the device was already on the
polling list. This resulted in an extra call to napi_complete which
triggered the BUG_ON().
This patch ensures that we only contine processing rotting packets in
the current mal_poll call if we are not already on the polling list.
Signed-off-by: Alistair Popple <alistair@popple.id.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Checking if MAC address is valid using is_valid_ether_addr() is already done in
of_get_mac_address(). While at it, reorganize checking so it matches checks in
other drivers.
Signed-off-by: Luka Perkov <luka@openwrt.org>
CC: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
CC: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Checking if MAC address is valid using is_valid_ether_addr() is already done in
of_get_mac_address().
Signed-off-by: Luka Perkov <luka@openwrt.org>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
CC: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Checking if MAC address is valid using is_valid_ether_addr() is already done in
of_get_mac_address().
Signed-off-by: Luka Perkov <luka@openwrt.org>
Acked-by: David Daney <david.daney@cavium.com>
CC: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fast Open currently has a fall back feature to address SYN-data being
dropped but it requires the middle-box to pass on regular SYN retry
after SYN-data. This is implemented in commit aab487435 ("net-tcp:
Fast Open client - detecting SYN-data drops")
However some NAT boxes will drop all subsequent packets after first
SYN-data and blackholes the entire connections. An example is in
commit 356d7d8 "netfilter: nf_conntrack: fix tcp_in_window for Fast
Open".
The sender should note such incidents and fall back to use the regular
TCP handshake on subsequent attempts temporarily as well: after the
second SYN timeouts the original Fast Open SYN is most likely lost.
When such an event recurs Fast Open is disabled based on the number of
recurrences exponentially.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Regression and warn fixes for i915.
* tag 'drm-intel-fixes-2013-10-29' of git://people.freedesktop.org/~danvet/drm-intel:
drm/i915: Fix the PPT fdi lane bifurcate state handling on ivb
drm/i915: No LVDS hardware on Intel D410PT and D425KT
drm/i915/dp: workaround BIOS eDP bpp clamping issue
drm/i915: Add HSW CRT output readout support
drm/i915: Add support for pipe_bpp readout
Unlike UDP or TCP, we do not take the pseudo-header into
account in SCTP checksums. So in case port mapping is the
very same, we do not need to recalculate the whole SCTP
checksum in software, which is very expensive.
Also, similarly as in TCP, take into account when a private
helper mangled the packet. In that case, we also need to
recalculate the checksum even if ports might be same.
Thanks for feedback regarding skb->ip_summed checks from
Julian Anastasov; here's a discussion on these checks for
snat and dnat:
* For snat_handler(), we can see CHECKSUM_PARTIAL from
virtual devices, and from LOCAL_OUT, otherwise it
should be CHECKSUM_UNNECESSARY. In general, in snat it
is more complex. skb contains the original route and
ip_vs_route_me_harder() can change the route after
snat_handler. So, for locally generated replies from
local server we can not preserve the CHECKSUM_PARTIAL
mode. It is an chicken or egg dilemma: snat_handler
needs the device after rerouting (to check for
NETIF_F_SCTP_CSUM), while ip_route_me_harder() wants
the snat_handler() to put the new saddr for proper
rerouting.
* For dnat_handler(), we should not see CHECKSUM_COMPLETE
for SCTP, in fact the small set of drivers that support
SCTP offloading return CHECKSUM_UNNECESSARY on correctly
received SCTP csum. We can see CHECKSUM_PARTIAL from
local stack or received from virtual drivers. The idea is
that SCTP decides to avoid csum calculation if hardware
supports offloading. IPVS can change the device after
rerouting to real server but we can preserve the
CHECKSUM_PARTIAL mode if the new device supports
offloading too. This works because skb dst is changed
before dnat_handler and we see the new device. So, checks
in the 'if' part will decide whether it is ok to keep
CHECKSUM_PARTIAL for the output. If the packet was with
CHECKSUM_NONE, hence we deal with unknown checksum. As we
recalculate the sum for IP header in all cases, it should
be safe to use CHECKSUM_UNNECESSARY. We can forward wrong
checksum in this case (without cp->app). In case of
CHECKSUM_UNNECESSARY, the csum was valid on receive.
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Jeff Kirsher says:
====================
This series contains updates to vxlan, net, ixgbe, ixgbevf, and i40e.
Joseph provides a single patch against vxlan which removes the burden
from the NIC drivers to check if the vxlan driver is enabled in the
kernel and also makes available the vxlan headrooms to the drivers.
Jacob provides majority of the patches, with patches against net, ixgbe
and ixgbevf. His net patch adds might_sleep() call to napi_disable so
that every use of napi_disable during atomic context will be visible.
Then Jacob provides a patch to fix qv_lock_napi call in
ixgbe_napi_disable_all. The other ixgbe patches cleanup
ixgbe_check_minimum_link function to correctly show that there are some
minor loss of encoding, even though we don't calculate it and remove
unnecessary duplication of PCIe bandwidth display. Lastly, Jacob
provides 4 patches against ixgbevf to add ixgbevf_rx_skb in line with
how ixgbe handles the variations on how packets can be received, adds
support in order to track how many packets were cleaned during busy poll
as part of the extended statistics.
Wei Yongjun provides a fix for i40e to return -ENOMEN in the memory
allocation error handling case instead of returning 0, as done
elsewhere in this function.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Amend the documentation in the mvmdio driver to note the fact
that it is now used by both the mvneta and mv643xx_eth drivers.
Signed-off-by: Leigh Brown <leigh@solinno.co.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make only a single call to mutex_unlock in orion_mdio_write.
Signed-off-by: Leigh Brown <leigh@solinno.co.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace manual poll of MVMDIO_SMI_READ_VALID with a call to
orion_mdio_wait_ready. This ensures a consistent timeout,
eliminates a busy loop, and allows for use of interrupts on
systems that support them.
Signed-off-by: Leigh Brown <leigh@solinno.co.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Amend orion_mdio_wait_ready so that the same timeout is used when
polling or using wait_event_timeout. Set the timeout to 1ms.
Replace udelay with usleep_range to avoid a busy loop, and set the
polling interval range as 45us to 55us, so that the first sleep
will be enough in almost all cases.
Generate the same log message at timeout when polling or using
wait_event_timeout.
Signed-off-by: Leigh Brown <leigh@solinno.co.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
The function needn't to be public, so to make it as static.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The interface type, which is being traced by "struct be_adapter::
if_type", isn't used currently. So we can remove that safely
according to Sathya's comments.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use a more current logging style.
Convert printks to pr_<level>.
Consolidate multiple printks into a single printk to avoid
any possible dmesg interleaving. Add a default "event" msg
in case the listed types are ever expanded.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently multicast code attempts to extrace the vlan id from
the skb even when vlan filtering is disabled. This can lead
to mdb entries being created with the wrong vlan id.
Pass the already extracted vlan id to the multicast
filtering code to make the correct id is used in
creation as well as lookup.
Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
Acked-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jesse Gross says:
====================
One patch for net/3.12 fixing an issue where devices could be in an
invalid state they are removed while still attached to OVS.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Update the URLs in the Kconfig file to the new pages at sangoma.com and cisco.com
Signed-off-by: Michael Drüing <michael@drueing.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
This work contains a lightweight BPF-based traffic classifier that can
serve as a flexible alternative to ematch-based tree classification, i.e.
now that BPF filter engine can also be JITed in the kernel. Naturally, tc
actions and policies are supported as well with cls_bpf. Multiple BPF
programs/filter can be attached for a class, or they can just as well be
written within a single BPF program, that's really up to the user how he
wishes to run/optimize the code, e.g. also for inversion of verdicts etc.
The notion of a BPF program's return/exit codes is being kept as follows:
0: No match
-1: Select classid given in "tc filter ..." command
else: flowid, overwrite the default one
As a minimal usage example with iproute2, we use a 3 band prio root qdisc
on a router with sfq each as leave, and assign ssh and icmp bpf-based
filters to band 1, http traffic to band 2 and the rest to band 3. For the
first two bands we load the bytecode from a file, in the 2nd we load it
inline as an example:
echo 1 > /proc/sys/net/core/bpf_jit_enable
tc qdisc del dev em1 root
tc qdisc add dev em1 root handle 1: prio bands 3 priomap 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
tc qdisc add dev em1 parent 1:1 sfq perturb 16
tc qdisc add dev em1 parent 1:2 sfq perturb 16
tc qdisc add dev em1 parent 1:3 sfq perturb 16
tc filter add dev em1 parent 1: bpf run bytecode-file /etc/tc/ssh.bpf flowid 1:1
tc filter add dev em1 parent 1: bpf run bytecode-file /etc/tc/icmp.bpf flowid 1:1
tc filter add dev em1 parent 1: bpf run bytecode-file /etc/tc/http.bpf flowid 1:2
tc filter add dev em1 parent 1: bpf run bytecode "`bpfc -f tc -i misc.ops`" flowid 1:3
BPF programs can be easily created and passed to tc, either as inline
'bytecode' or 'bytecode-file'. There are a couple of front-ends that can
compile opcodes, for example:
1) People familiar with tcpdump-like filters:
tcpdump -iem1 -ddd port 22 | tr '\n' ',' > /etc/tc/ssh.bpf
2) People that want to low-level program their filters or use BPF
extensions that lack support by libpcap's compiler:
bpfc -f tc -i ssh.ops > /etc/tc/ssh.bpf
ssh.ops example code:
ldh [12]
jne #0x800, drop
ldb [23]
jneq #6, drop
ldh [20]
jset #0x1fff, drop
ldxb 4 * ([14] & 0xf)
ldh [%x + 14]
jeq #0x16, pass
ldh [%x + 16]
jne #0x16, drop
pass: ret #-1
drop: ret #0
It was chosen to load bytecode into tc, since the reverse operation,
tc filter list dev em1, is then able to show the exact commands again.
Possible follow-up work could also include a small expression compiler
for iproute2. Tested with the help of bmon. This idea came up during
the Netfilter Workshop 2013 in Copenhagen. Also thanks to feedback from
Eric Dumazet!
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
This cleans code a bit and will be useful when allocating buffers in
other places (like RX path, to avoid skb_copy_from_linear_data_offset).
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso says:
====================
This pull request contains the following netfilter fix:
* fix --queue-bypass in xt_NFQUEUE revision 3. While adding the
revision 3 of this target, the bypass flags were not correctly
handled anymore, thus, breaking packet bypassing if no application
is listening from userspace, patch from Holger Eitzenberger,
reported by Florian Westphal.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
According to Software User's Manual, the event of last-level-cache
read/write misses is mapped to even counters. Odd counters of that
event number count miss cycles.
Signed-off-by: Deng-Cheng Zhu <dengcheng.zhu@imgtec.com>
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/6036/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Nico Golde reports a few straggling uses of [io_]remap_pfn_range() that
really should use the vm_iomap_memory() helper. This trivially converts
two of them to the helper, and comments about why the third one really
needs to continue to use remap_pfn_range(), and adds the missing size
check.
Reported-by: Nico Golde <nico@ngolde.de>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org.
Pull perf tooling fixes from Ingo Molnar:
"This contains five tooling fixes:
- fix a remaining mmap2 assumption which resulted in perf top output
breakage
- fix mmap ring-buffer processing bug that corrupts data
- fix for a severe python scripting memory leak
- fix broken (and user-visible) -g option handling
- fix stdio output
The diffstat size is larger than what we'd like to see this late :-/"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf tools: Fixup mmap event consumption
perf top: Split -G and --call-graph
perf record: Split -g and --call-graph
perf hists: Add color overhead for stdio output buffer
perf tools: Fix up /proc/PID/maps parsing
perf script python: Fix mem leak due to missing Py_DECREFs on dict entries
Without the timer debugging, the delayed kobject release will just
result in undebuggable oopses if it triggers any latent bugs. That
doesn't actually help debugging at all.
So make DEBUG_KOBJECT_RELEASE depend on DEBUG_OBJECTS_TIMERS to avoid
having people enable one without the other.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Originally I've thought that this is leftover hw state dirt from the
BIOS. But after way too much helpless flailing around on my part I've
noticed that the actual bug is when we change the state of an already
active pipe.
For example when we change the fdi lines from 2 to 3 without switching
off outputs in-between we'll never see the crucial on->off transition
in the ->modeset_global_resources hook the current logic relies on.
Patch version 2 got this right by instead also checking whether the
pipe is indeed active. But that in turn broke things when pipes have
been turned off through dpms since the bifurcate enabling is done in
the ->crtc_mode_set callback.
To address this issues discussed with Ville in the patch review move
the setting of the bifurcate bit into the ->crtc_enable hook. That way
we won't wreak havoc with this state when userspace puts all other
outputs into dpms off state. This also moves us forward with our
overall goal to unify the modeset and dpms on paths (which we need to
have to allow runtime pm in the dpms off state).
Unfortunately this requires us to move the bifurcate helpers around a
bit.
Also update the commit message, I've misanalyzed the bug rather badly.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=70507
Tested-by: Jan-Michael Brummer <jan.brummer@tabos.org>
Cc: stable@vger.kernel.org
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
V3 of the NFQUEUE target ignores the --queue-bypass flag,
causing packets to be dropped when the userspace listener
isn't running.
Regression is in since 8746ddcf12 ("netfilter: xt_NFQUEUE:
introduce CPU fanout").
Reported-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Fix to return -ENOMEM in the memory alloc error handling
case instead of 0, as done elsewhere in this function.
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch removes the need to keep a zero_base variable in the adapter
structure. Now we just use two different macros to set the non-zero and
zero base. This adds to readability and shortens some of the structure
initialization under 80 columns. The gathering of status for ethtool was
slightly modified to again better fit into 80 columns and become a bit
more readable.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds the extended statistics similar to the ixgbe driver. These
statistics keep track of how often the busy polling yields, as well as how many
packets are cleaned or missed by the polling routine.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch enables CONFIG_NET_RX_BUSY_POLL support in the VF code. This enables
sockets which have enabled the SO_BUSY_POLL socket option to use the
ndo_busy_poll_recv operation which could result in lower latency, at the cost
of higher CPU utilization, and increased power usage. This support is similar
to how the ixgbe driver works.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
OK, so what I'm actually seeing on my WSM is that sched/clock.c is
'broken' for the purpose we're using it for.
What triggered it is that my WSM-EP is broken :-(
[ 0.001000] tsc: Fast TSC calibration using PIT
[ 0.002000] tsc: Detected 2533.715 MHz processor
[ 0.500180] TSC synchronization [CPU#0 -> CPU#6]:
[ 0.505197] Measured 3 cycles TSC warp between CPUs, turning off TSC clock.
[ 0.004000] tsc: Marking TSC unstable due to check_tsc_sync_source failed
For some reason it consistently detects TSC skew, even though NHM+
should have a single clock domain for 'reasonable' systems.
This marks sched_clock_stable=0, which means that we do fancy stuff to
try and get a 'sane' clock. Part of this fancy stuff relies on the tick,
clearly that's gone when NOHZ=y. So for idle cpus time gets stuck, until
it either wakes up or gets kicked by another cpu.
While this is perfectly fine for the scheduler -- it only cares about
actually running stuff, and when we're running stuff we're obviously not
idle. This does somewhat break down for perf which can trigger events
just fine on an otherwise idle cpu.
So I've got NMIs get get 'measured' as taking ~1ms, which actually
don't last nearly that long:
<idle>-0 [013] d.h. 886.311970: rcu_nmi_enter <-do_nmi
...
<idle>-0 [013] d.h. 886.311997: perf_sample_event_took: HERE!!! : 1040990
So ftrace (which uses sched_clock(), not the fancy bits) only sees
~27us, but we measure ~1ms !!
Now since all this measurement stuff lives in x86 code, we can actually
fix it.
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: mingo@kernel.org
Cc: dave.hansen@linux.intel.com
Cc: eranian@google.com
Cc: Don Zickus <dzickus@redhat.com>
Cc: jmario@redhat.com
Cc: acme@infradead.org
Link: http://lkml.kernel.org/r/20131017133350.GG3364@laptop.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The PPC64 people noticed a missing memory barrier and crufty old
comments in the perf ring buffer code. So update all the comments and
add the missing barrier.
When the architecture implements local_t using atomic_long_t there
will be double barriers issued; but short of introducing more
conditional barrier primitives this is the best we can do.
Reported-by: Victor Kaplansky <victork@il.ibm.com>
Tested-by: Victor Kaplansky <victork@il.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: michael@ellerman.id.au
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Michael Neuling <mikey@neuling.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: anton@samba.org
Cc: benh@kernel.crashing.org
Link: http://lkml.kernel.org/r/20131025173749.GG19466@laptop.lan
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Rather than return true/false indicating whether there was budget left, return
the total packets cleaned. This currently has no use, but will be used in a
following patch which enables CONFIG_NET_RX_BUSY_POLL support in order to track
how many packets were cleaned during the busy poll as part of the extended
statistics.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds ixgbevf_rx_skb in line with how ixgbe handles the variations on
how packets can be received. It will be extended in a following patch for
CONFIG_NET_RX_BUSY_POLL support.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch removes the unnecessary display of PCIe bandwidth twice. Since the
ixgbe_check_minimum_link does a better job, and ensures accurate detection on
even complex chains, this older check is no longer necessary.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch updates the ixgbe_check_minimum_link function to correctly show that
there is some minor loss of encoding, even though we don't calculate it in the
max GT/s equation. It is small enough to not bother, but is better to report it
than not.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
A THP PMD update is accounted for as 512 pages updated in vmstat. This is
large difference when estimating the cost of automatic NUMA balancing and
can be misleading when comparing results that had collapsed versus split
THP. This patch addresses the accounting issue.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-10-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
THP migration uses the page lock to guard against parallel allocations
but there are cases like this still open
Task A Task B
--------------------- ---------------------
do_huge_pmd_numa_page do_huge_pmd_numa_page
lock_page
mpol_misplaced == -1
unlock_page
goto clear_pmdnuma
lock_page
mpol_misplaced == 2
migrate_misplaced_transhuge
pmd = pmd_mknonnuma
set_pmd_at
During hours of testing, one crashed with weird errors and while I have
no direct evidence, I suspect something like the race above happened.
This patch extends the page lock to being held until the pmd_numa is
cleared to prevent migration starting in parallel while the pmd_numa is
being cleared. It also flushes the old pmd entry and orders pagetable
insertion before rmap insertion.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-9-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
There are three callers of task_numa_fault():
- do_huge_pmd_numa_page():
Accounts against the current node, not the node where the
page resides, unless we migrated, in which case it accounts
against the node we migrated to.
- do_numa_page():
Accounts against the current node, not the node where the
page resides, unless we migrated, in which case it accounts
against the node we migrated to.
- do_pmd_numa_page():
Accounts not at all when the page isn't migrated, otherwise
accounts against the node we migrated towards.
This seems wrong to me; all three sites should have the same
sementaics, furthermore we should accounts against where the page
really is, we already know where the task is.
So modify all three sites to always account; we did after all receive
the fault; and always account to where the page is after migration,
regardless of success.
They all still differ on when they clear the PTE/PMD; ideally that
would get sorted too.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-8-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
THP migrations are serialised by the page lock but on its own that does
not prevent THP splits. If the page is split during THP migration then
the pmd_same checks will prevent page table corruption but the unlock page
and other fix-ups potentially will cause corruption. This patch takes the
anon_vma lock to prevent parallel splits during migration.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-7-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The locking for migrating THP is unusual. While normal page migration
prevents parallel accesses using a migration PTE, THP migration relies on
a combination of the page_table_lock, the page lock and the existance of
the NUMA hinting PTE to guarantee safety but there is a bug in the scheme.
If a THP page is currently being migrated and another thread traps a
fault on the same page it checks if the page is misplaced. If it is not,
then pmd_numa is cleared. The problem is that it checks if the page is
misplaced without holding the page lock meaning that the racing thread
can be migrating the THP when the second thread clears the NUMA bit
and faults a stale page.
This patch checks if the page is potentially being migrated and stalls
using the lock_page if it is potentially being migrated before checking
if the page is misplaced or not.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-6-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
If another task handled a hinting fault in parallel then do not double
account for it.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-5-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
ixgbe_napi_disable_all calls napi_disable on each queue, however the busy
polling code introduced a local_bh_disable()d context around the napi_disable.
The original author did not realize that napi_disable might sleep, which would
cause a sleep while atomic BUG. In addition, on a single processor system, the
ixgbe_qv_lock_napi loop shouldn't have to mdelay. This patch adds an
ixgbe_qv_disable along with a new IXGBE_QV_STATE_DISABLED bit, which it uses to
indicate to the poll and napi routines that the q_vector has been disabled. Now
the ixgbe_napi_disable_all function will wait until all pending work has been
finished and prevent any future work from being started.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Cc: Eliezer Tamir <eliezer.tamir@linux.intel.com>
Cc: Alexander Duyck <alexander.duyck@intel.com>
Cc: Hyong-Youb Kim <hykim@myri.com>
Cc: Amir Vadai <amirv@mellanox.com>
Cc: Dmitry Kravkov <dmitry@broadcom.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
napi_disable uses an msleep() call to wait for outstanding napi work to be
finished after setting the disable bit. It does not always sleep incase there
was no outstanding work. This resulted in a rare bug in ixgbe_down operation
where a napi_disable call took place inside of a local_bh_disable()d context.
In order to enable easier detection of future sleep while atomic BUGs, this
patch adds a might_sleep() call, so that every use of napi_disable during
atomic context will be visible.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Cc: Eliezer Tamir <eliezer.tamir@linux.intel.com>
Cc: Alexander Duyck <alexander.duyck@intel.com>
Cc: Hyong-Youb Kim <hykim@myri.com>
Cc: Amir Vadai <amirv@mellanox.com>
Cc: Dmitry Kravkov <dmitry@broadcom.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch removes the burden from the NIC drivers to check if the
vxlan driver is enabled in the kernel and also makes available
the vxlan headrooms to them.
Signed-off-by: Joseph Gasparakis <joseph.gasparakis@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
. Add color overhead for stdio output buffer, which fixes
--stdio output being chopped up on the hot (red) entries,
fix from Jiri Olsa.
. Get 'perf record -g -a sleep 1' working again, removing the
need for -- separating perf options from the workload, restoring
ages old behaviour, fix from Jiri Olsa.
More patches allowing ~/.perfconfig setting up of default
callchain collecting method ("fp" or "dwarf") left for next
merge window.
. Fixup mmap event consumption, where we were acking the
consumption by writing the tail before actually accessing
the event, which could lead to using overwritten records
in things like 'perf record --call-graph'. from Zhouyi Zhou.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.14 (GNU/Linux)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=U+a8
-----END PGP SIGNATURE-----
Merge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent
Pull perf/urgent fixes from Arnaldo Carvalho de Melo:
* Add color overhead for stdio output buffer, which fixes
--stdio output being chopped up on the hot (red) entries,
fix from Jiri Olsa.
* Get 'perf record -g -a sleep 1' working again, removing the
need for -- separating perf options from the workload, restoring
ages old behaviour, fix from Jiri Olsa.
More patches allowing ~/.perfconfig setting up of default
callchain collecting method ("fp" or "dwarf") left for next
merge window.
* Fixup mmap event consumption, where we were acking the
consumption by writing the tail before actually accessing
the event, which could lead to using overwritten records
in things like 'perf record --call-graph'. From Zhouyi Zhou.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
struct esp_data consists of a single pointer, vanishing the need for it
to be a structure. Fold the pointer into 'data' direcly, removing one
level of pointer indirection.
Signed-off-by: Mathias Krause <mathias.krause@secunet.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
The padlen member of struct esp_data is always zero. Get rid of it.
Signed-off-by: Mathias Krause <mathias.krause@secunet.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
