A customer has reported a soft-lockup when running an intensive
memory stress test, where the trace on multiple CPU's looks like this:
RIP: 0010:[<ffffffff810c53fe>]
[<ffffffff810c53fe>] native_queued_spin_lock_slowpath+0x10e/0x190
...
Call Trace:
[<ffffffff81182d07>] queued_spin_lock_slowpath+0x7/0xa
[<ffffffff811bc331>] change_protection_range+0x3b1/0x930
[<ffffffff811d4be8>] change_prot_numa+0x18/0x30
[<ffffffff810adefe>] task_numa_work+0x1fe/0x310
[<ffffffff81098322>] task_work_run+0x72/0x90
Further investigation showed that the lock contention here is pmd_lock().
The task_numa_work() function makes sure that only one thread is let to perform
the work in a single scan period (via cmpxchg), but if there's a thread with
mmap_sem locked for writing for several periods, multiple threads in
task_numa_work() can build up a convoy waiting for mmap_sem for read and then
all get unblocked at once.
This patch changes the down_read() to the trylock version, which prevents the
build up. For a workload experiencing mmap_sem contention, it's probably better
to postpone the NUMA balancing work anyway. This seems to have fixed the soft
lockups involving pmd_lock(), which is in line with the convoy theory.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20170515131316.21909-1-vbabka@suse.cz
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When priority inheritance was added back in 2.6.18 to sched_setscheduler(), it
added a path to taking an rt-mutex wait_lock, which is not IRQ safe. As PI
is not a common occurrence, lockdep will likely never trigger if
sched_setscheduler was called from interrupt context. A BUG_ON() was added
to trigger if __sched_setscheduler() was ever called from interrupt context
because there was a possibility to take the wait_lock.
Today the wait_lock is irq safe, but the path to taking it in
sched_setscheduler() is the same as the path to taking it from normal
context. The wait_lock is taken with raw_spin_lock_irq() and released with
raw_spin_unlock_irq() which will indiscriminately enable interrupts,
which would be bad in interrupt context.
The problem is that normalize_rt_tasks, which is called by triggering the
sysrq nice-all-RT-tasks was changed to call __sched_setscheduler(), and this
is done from interrupt context!
Now __sched_setscheduler() takes a "pi" parameter that is used to know if
the priority inheritance should be called or not. As the BUG_ON() only cares
about calling the PI code, it should only bug if called from interrupt
context with the "pi" parameter set to true.
Reported-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Tested-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@osdl.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: dbc7f069b9 ("sched: Use replace normalize_task() with __sched_setscheduler()")
Link: http://lkml.kernel.org/r/20170308124654.10e598f2@gandalf.local.home
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Inter-Processor-Interrupt(IPI) is needed when a page is unmapped and the
process' mm_cpumask() shows the process has ever run on other CPUs. page
migration, page reclaim all need IPIs. The number of IPI needed to send
to different CPUs is especially large for multi-threaded workload since
mm_cpumask() is per process.
For smp_call_function_many(), whenever a CPU queues a CSD to a target
CPU, it will send an IPI to let the target CPU to handle the work.
This isn't necessary - we need only send IPI when queueing a CSD
to an empty call_single_queue.
The reason:
flush_smp_call_function_queue() that is called upon a CPU receiving an
IPI will empty the queue and then handle all of the CSDs there. So if
the target CPU's call_single_queue is not empty, we know that:
i. An IPI for the target CPU has already been sent by 'previous queuers';
ii. flush_smp_call_function_queue() hasn't emptied that CPU's queue yet.
Thus, it's safe for us to just queue our CSD there without sending an
addtional IPI. And for the 'previous queuers', we can limit it to the
first queuer.
To demonstrate the effect of this patch, a multi-thread workload that
spawns 80 threads to equally consume 100G memory is used. This is tested
on a 2 node broadwell-EP which has 44cores/88threads and 32G memory. So
after 32G memory is used up, page reclaiming starts to happen a lot.
With this patch, IPI number dropped 88% and throughput increased about
15% for the above workload.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Link: http://lkml.kernel.org/r/20170519075331.GE2084@aaronlu.sh.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull power management fixes from Rafael Wysocki:
"These fix RTC wakeup from suspend-to-idle broken recently, fix CPU
idleness detection condition in the schedutil cpufreq governor, fix a
cpufreq driver build failure, fix an error code path in the power
capping framework, clean up the hibernate core and update the
intel_pstate documentation.
Specifics:
- Fix RTC wakeup from suspend-to-idle broken by the recent rework of
ACPI wakeup handling (Rafael Wysocki).
- Update intel_pstate driver documentation to reflect the current
code and explain how it works in more detail (Rafael Wysocki).
- Fix an issue related to CPU idleness detection on systems with
shared cpufreq policies in the schedutil governor (Juri Lelli).
- Fix a possible build issue in the dbx500 cpufreq driver (Arnd
Bergmann).
- Fix a function in the power capping framework core to return an
error code instead of 0 when there's an error (Dan Carpenter).
- Clean up variable definition in the hibernation core (Pushkar
Jambhlekar)"
* tag 'pm-4.12-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
cpufreq: dbx500: add a Kconfig symbol
PM / hibernate: Declare variables as static
PowerCap: Fix an error code in powercap_register_zone()
RTC: rtc-cmos: Fix wakeup from suspend-to-idle
PM / wakeup: Fix up wakeup_source_report_event()
cpufreq: intel_pstate: Document the current behavior and user interface
cpufreq: schedutil: use now as reference when aggregating shared policy requests
Pull networking fixes from David Miller:
"Mostly netfilter bug fixes in here, but we have some bits elsewhere as
well.
1) Don't do SNAT replies for non-NATed connections in IPVS, from
Julian Anastasov.
2) Don't delete conntrack helpers while they are still in use, from
Liping Zhang.
3) Fix zero padding in xtables's xt_data_to_user(), from Willem de
Bruijn.
4) Add proper RCU protection to nf_tables_dump_set() because we
cannot guarantee that we hold the NFNL_SUBSYS_NFTABLES lock. From
Liping Zhang.
5) Initialize rcv_mss in tcp_disconnect(), from Wei Wang.
6) smsc95xx devices can't handle IPV6 checksums fully, so don't
advertise support for offloading them. From Nisar Sayed.
7) Fix out-of-bounds access in __ip6_append_data(), from Eric
Dumazet.
8) Make atl2_probe() propagate the error code properly on failures,
from Alexey Khoroshilov.
9) arp_target[] in bond_check_params() is used uninitialized. This
got changes from a global static to a local variable, which is how
this mistake happened. Fix from Jarod Wilson.
10) Fix fallout from unnecessary NULL check removal in cls_matchall,
from Jiri Pirko. This is definitely brown paper bag territory..."
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (26 commits)
net: sched: cls_matchall: fix null pointer dereference
vsock: use new wait API for vsock_stream_sendmsg()
bonding: fix randomly populated arp target array
net: Make IP alignment calulations clearer.
bonding: fix accounting of active ports in 3ad
net: atheros: atl2: don't return zero on failure path in atl2_probe()
ipv6: fix out of bound writes in __ip6_append_data()
bridge: start hello_timer when enabling KERNEL_STP in br_stp_start
smsc95xx: Support only IPv4 TCP/UDP csum offload
arp: always override existing neigh entries with gratuitous ARP
arp: postpone addr_type calculation to as late as possible
arp: decompose is_garp logic into a separate function
arp: fixed error in a comment
tcp: initialize rcv_mss to TCP_MIN_MSS instead of 0
netfilter: xtables: fix build failure from COMPAT_XT_ALIGN outside CONFIG_COMPAT
ebtables: arpreply: Add the standard target sanity check
netfilter: nf_tables: revisit chain/object refcounting from elements
netfilter: nf_tables: missing sanitization in data from userspace
netfilter: nf_tables: can't assume lock is acquired when dumping set elems
netfilter: synproxy: fix conntrackd interaction
...
* intel_pstate:
cpufreq: intel_pstate: Document the current behavior and user interface
* pm-cpufreq:
cpufreq: dbx500: add a Kconfig symbol
* pm-cpufreq-sched:
cpufreq: schedutil: use now as reference when aggregating shared policy requests
The assignmnet:
ip_align = strict ? 2 : NET_IP_ALIGN;
in compare_pkt_ptr_alignment() trips up Coverity because we can only
get to this code when strict is true, therefore ip_align will always
be 2 regardless of NET_IP_ALIGN's value.
So just assign directly to '2' and explain the situation in the
comment above.
Reported-by: "Gustavo A. R. Silva" <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull scheduler fix from Thomas Gleixner:
"A single scheduler fix:
Prevent idle task from ever being preempted. That makes sure that
synchronize_rcu_tasks() which is ignoring idle task does not pretend
that no task is stuck in preempted state. If that happens and idle was
preempted on a ftrace trampoline the machine crashes due to
inconsistent state"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/core: Call __schedule() from do_idle() without enabling preemption
Pull irq fixes from Thomas Gleixner:
"A set of small fixes for the irq subsystem:
- Cure a data ordering problem with chained interrupts
- Three small fixlets for the mbigen irq chip"
* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
genirq: Fix chained interrupt data ordering
irqchip/mbigen: Fix the clear register offset calculation
irqchip/mbigen: Fix potential NULL dereferencing
irqchip/mbigen: Fix memory mapping code
Pull tracing fixes from Steven Rostedt:
- Fix a bug caused by not cleaning up the new instance unique triggers
when deleting an instance. It also creates a selftest that triggers
that bug.
- Fix the delayed optimization happening after kprobes boot up self
tests being removed by freeing of init memory.
- Comment kprobes on why the delay optimization is not a problem for
removal of modules, to keep other developers from searching that
riddle.
- Fix another case of rcu not watching in stack trace tracing.
* tag 'trace-v4.12-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Make sure RCU is watching before calling a stack trace
kprobes: Document how optimized kprobes are removed from module unload
selftests/ftrace: Add test to remove instance with active event triggers
selftests/ftrace: Fix bashisms
ftrace: Remove #ifdef from code and add clear_ftrace_function_probes() stub
ftrace/instances: Clear function triggers when removing instances
ftrace: Simplify glob handling in unregister_ftrace_function_probe_func()
tracing/kprobes: Enforce kprobes teardown after testing
tracing: Move postpone selftests to core from early_initcall
Pull block fixes from Jens Axboe:
"A small collection of fixes that should go into this cycle.
- a pull request from Christoph for NVMe, which ended up being
manually applied to avoid pulling in newer bits in master. Mostly
fibre channel fixes from James, but also a few fixes from Jon and
Vijay
- a pull request from Konrad, with just a single fix for xen-blkback
from Gustavo.
- a fuseblk bdi fix from Jan, fixing a regression in this series with
the dynamic backing devices.
- a blktrace fix from Shaohua, replacing sscanf() with kstrtoull().
- a request leak fix for drbd from Lars, fixing a regression in the
last series with the kref changes. This will go to stable as well"
* 'for-linus' of git://git.kernel.dk/linux-block:
nvmet: release the sq ref on rdma read errors
nvmet-fc: remove target cpu scheduling flag
nvme-fc: stop queues on error detection
nvme-fc: require target or discovery role for fc-nvme targets
nvme-fc: correct port role bits
nvme: unmap CMB and remove sysfs file in reset path
blktrace: fix integer parse
fuseblk: Fix warning in super_setup_bdi_name()
block: xen-blkback: add null check to avoid null pointer dereference
drbd: fix request leak introduced by locking/atomic, kref: Kill kref_sub()
sscanf is a very poor way to parse integer. For example, I input
"discard" for act_mask, it gets 0xd and completely messes up. Using
correct API to do integer parse.
This patch also makes attributes accept any base of integer.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Pull networking fixes from David Miller:
1) Don't allow negative TCP reordering values, from Soheil Hassas
Yeganeh.
2) Don't overflow while parsing ipv6 header options, from Craig Gallek.
3) Handle more cleanly the case where an individual route entry during
a dump will not fit into the allocated netlink SKB, from David
Ahern.
4) Add missing CONFIG_INET dependency for mlx5e, from Arnd Bergmann.
5) Allow neighbour updates to converge more quickly via gratuitous
ARPs, from Ihar Hrachyshka.
6) Fix compile error from CONFIG_INET is disabled, from Eric Dumazet.
7) Fix use after free in x25 protocol init, from Lin Zhang.
8) Valid VLAN pvid ranges passed into br_validate(), from Tobias
Jungel.
9) NULL out address lists in child sockets in SCTP, this is similar to
the fix we made for inet connection sockets last week. From Eric
Dumazet.
10) Fix NULL deref in mlxsw driver, from Ido Schimmel.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (27 commits)
mlxsw: spectrum: Avoid possible NULL pointer dereference
sh_eth: Do not print an error message for probe deferral
sh_eth: Use platform device for printing before register_netdev()
mlxsw: spectrum_router: Fix rif counter freeing routine
mlxsw: spectrum_dpipe: Fix incorrect entry index
cxgb4: update latest firmware version supported
qmi_wwan: add another Lenovo EM74xx device ID
sctp: do not inherit ipv6_{mc|ac|fl}_list from parent
udp: make *udp*_queue_rcv_skb() functions static
bridge: netlink: check vlan_default_pvid range
net: ethernet: faraday: To support device tree usage.
net: x25: fix one potential use-after-free issue
bpf: adjust verifier heuristics
ipv6: Check ip6_find_1stfragopt() return value properly.
selftests/bpf: fix broken build due to types.h
bnxt_en: Check status of firmware DCBX agent before setting DCB_CAP_DCBX_HOST.
bnxt_en: Call bnxt_dcb_init() after getting firmware DCBX configuration.
net: fix compile error in skb_orphan_partial()
ipv6: Prevent overrun when parsing v6 header options
neighbour: update neigh timestamps iff update is effective
...
Pull pid namespace fixes from Eric Biederman:
"These are two bugs that turn out to have simple fixes that were
reported during the merge window. Both of these issues have existed
for a while and it just happens that they both were reported at almost
the same time"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
pid_ns: Fix race between setns'ed fork() and zap_pid_ns_processes()
pid_ns: Sleep in TASK_INTERRUPTIBLE in zap_pid_ns_processes
Current limits with regards to processing program paths do not
really reflect today's needs anymore due to programs becoming
more complex and verifier smarter, keeping track of more data
such as const ALU operations, alignment tracking, spilling of
PTR_TO_MAP_VALUE_ADJ registers, and other features allowing for
smarter matching of what LLVM generates.
This also comes with the side-effect that we result in fewer
opportunities to prune search states and thus often need to do
more work to prove safety than in the past due to different
register states and stack layout where we mismatch. Generally,
it's quite hard to determine what caused a sudden increase in
complexity, it could be caused by something as trivial as a
single branch somewhere at the beginning of the program where
LLVM assigned a stack slot that is marked differently throughout
other branches and thus causing a mismatch, where verifier
then needs to prove safety for the whole rest of the program.
Subsequently, programs with even less than half the insn size
limit can get rejected. We noticed that while some programs
load fine under pre 4.11, they get rejected due to hitting
limits on more recent kernels. We saw that in the vast majority
of cases (90+%) pruning failed due to register mismatches. In
case of stack mismatches, majority of cases failed due to
different stack slot types (invalid, spill, misc) rather than
differences in spilled registers.
This patch makes pruning more aggressive by also adding markers
that sit at conditional jumps as well. Currently, we only mark
jump targets for pruning. For example in direct packet access,
these are usually error paths where we bail out. We found that
adding these markers, it can reduce number of processed insns
by up to 30%. Another option is to ignore reg->id in probing
PTR_TO_MAP_VALUE_OR_NULL registers, which can help pruning
slightly as well by up to 7% observed complexity reduction as
stand-alone. Meaning, if a previous path with register type
PTR_TO_MAP_VALUE_OR_NULL for map X was found to be safe, then
in the current state a PTR_TO_MAP_VALUE_OR_NULL register for
the same map X must be safe as well. Last but not least the
patch also adds a scheduling point and bumps the current limit
for instructions to be processed to a more adequate value.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Thomas discovered a bug where the kprobe trace tests had a race
condition where the kprobe_optimizer called from a delayed work queue
that does the optimizing and "unoptimizing" of a kprobe, can try to
modify the text after it has been freed by the init code.
The kprobe trace selftest is a special case, and Thomas and myself
investigated to see if there's a chance that this could also be a bug
with module unloading, as the code is not obvious to how it handles
this. After adding lots of printks, I figured it out. Thomas suggested
that this should be commented so that others will not have to go
through this exercise again.
Link: http://lkml.kernel.org/r/20170516145835.3827d3aa@gandalf.local.home
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
