This patch adds support for non-linear data on raw records. It
extends raw records to have one or multiple fragments that will
be written linearly into the ring slot, where each fragment can
optionally have a custom callback handler to walk and extract
complex, possibly non-linear data.
If a callback handler is provided for a fragment, then the new
__output_custom() will be used instead of __output_copy() for
the perf_output_sample() part. perf_prepare_sample() does all
the size calculation only once, so perf_output_sample() doesn't
need to redo the same work anymore, meaning real_size and padding
will be cached in the raw record. The raw record becomes 32 bytes
in size without holes; to not increase it further and to avoid
doing unnecessary recalculations in fast-path, we can reuse
next pointer of the last fragment, idea here is borrowed from
ZERO_OR_NULL_PTR(), which should keep the perf_output_sample()
path for PERF_SAMPLE_RAW minimal.
This facility is needed for BPF's event output helper as a first
user that will, in a follow-up, add an additional perf_raw_frag
to its perf_raw_record in order to be able to more efficiently
dump skb context after a linear head meta data related to it.
skbs can be non-linear and thus need a custom output function to
dump buffers. Currently, the skb data needs to be copied twice;
with the help of __output_custom() this work only needs to be
done once. Future users could be things like XDP/BPF programs
that work on different context though and would thus also have
a different callback function.
The few users of raw records are adapted to initialize their frag
data from the raw record itself, no change in behavior for them.
The code is based upon a PoC diff provided by Peter Zijlstra [1].
[1] http://thread.gmane.org/gmane.linux.network/421294
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jann Horn reported following analysis that could potentially result
in a very hard to trigger (if not impossible) UAF race, to quote his
event timeline:
- Set up a process with threads T1, T2 and T3
- Let T1 set up a socket filter F1 that invokes another filter F2
through a BPF map [tail call]
- Let T1 trigger the socket filter via a unix domain socket write,
don't wait for completion
- Let T2 call PERF_EVENT_IOC_SET_BPF with F2, don't wait for completion
- Now T2 should be behind bpf_prog_get(), but before bpf_prog_put()
- Let T3 close the file descriptor for F2, dropping the reference
count of F2 to 2
- At this point, T1 should have looked up F2 from the map, but not
finished executing it
- Let T3 remove F2 from the BPF map, dropping the reference count of
F2 to 1
- Now T2 should call bpf_prog_put() (wrong BPF program type), dropping
the reference count of F2 to 0 and scheduling bpf_prog_free_deferred()
via schedule_work()
- At this point, the BPF program could be freed
- BPF execution is still running in a freed BPF program
While at PERF_EVENT_IOC_SET_BPF time it's only guaranteed that the perf
event fd we're doing the syscall on doesn't disappear from underneath us
for whole syscall time, it may not be the case for the bpf fd used as
an argument only after we did the put. It needs to be a valid fd pointing
to a BPF program at the time of the call to make the bpf_prog_get() and
while T2 gets preempted, F2 must have dropped reference to 1 on the other
CPU. The fput() from the close() in T3 should also add additionally delay
to the reference drop via exit_task_work() when bpf_prog_release() gets
called as well as scheduling bpf_prog_free_deferred().
That said, it makes nevertheless sense to move the BPF prog destruction
generally after RCU grace period to guarantee that such scenario above,
but also others as recently fixed in ceb5607035 ("bpf, perf: delay release
of BPF prog after grace period") with regards to tail calls won't happen.
Integrating bpf_prog_free_deferred() directly into the RCU callback is
not allowed since the invocation might happen from either softirq or
process context, so we're not permitted to block. Reviewing all bpf_prog_put()
invocations from eBPF side (note, cBPF -> eBPF progs don't use this for
their destruction) with call_rcu() look good to me.
Since we don't know whether at the time of attaching the program, we're
already part of a tail call map, we need to use RCU variant. However, due
to this, there won't be severely more stress on the RCU callback queue:
situations with above bpf_prog_get() and bpf_prog_put() combo in practice
normally won't lead to releases, but even if they would, enough effort/
cycles have to be put into loading a BPF program into the kernel already.
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking fixes from David Miller:
"I've been traveling so this accumulates more than week or so of bug
fixing. It perhaps looks a little worse than it really is.
1) Fix deadlock in ath10k driver, from Ben Greear.
2) Increase scan timeout in iwlwifi, from Luca Coelho.
3) Unbreak STP by properly reinjecting STP packets back into the
stack. Regression fix from Ido Schimmel.
4) Mediatek driver fixes (missing malloc failure checks, leaking of
scratch memory, wrong indexing when mapping TX buffers, etc.) from
John Crispin.
5) Fix endianness bug in icmpv6_err() handler, from Hannes Frederic
Sowa.
6) Fix hashing of flows in UDP in the ruseport case, from Xuemin Su.
7) Fix netlink notifications in ovs for tunnels, delete link messages
are never emitted because of how the device registry state is
handled. From Nicolas Dichtel.
8) Conntrack module leaks kmemcache on unload, from Florian Westphal.
9) Prevent endless jump loops in nft rules, from Liping Zhang and
Pablo Neira Ayuso.
10) Not early enough spinlock initialization in mlx4, from Eric
Dumazet.
11) Bind refcount leak in act_ipt, from Cong WANG.
12) Missing RCU locking in HTB scheduler, from Florian Westphal.
13) Several small MACSEC bug fixes from Sabrina Dubroca (missing RCU
barrier, using heap for SG and IV, and erroneous use of async flag
when allocating AEAD conext.)
14) RCU handling fix in TIPC, from Ying Xue.
15) Pass correct protocol down into ipv4_{update_pmtu,redirect}() in
SIT driver, from Simon Horman.
16) Socket timer deadlock fix in TIPC from Jon Paul Maloy.
17) Fix potential deadlock in team enslave, from Ido Schimmel.
18) Memory leak in KCM procfs handling, from Jiri Slaby.
19) ESN generation fix in ipv4 ESP, from Herbert Xu.
20) Fix GFP_KERNEL allocations with locks held in act_ife, from Cong
WANG.
21) Use after free in netem, from Eric Dumazet.
22) Uninitialized last assert time in multicast router code, from Tom
Goff.
23) Skip raw sockets in sock_diag destruction broadcast, from Willem
de Bruijn.
24) Fix link status reporting in thunderx, from Sunil Goutham.
25) Limit resegmentation of retransmit queue so that we do not
retransmit too large GSO frames. From Eric Dumazet.
26) Delay bpf program release after grace period, from Daniel
Borkmann"
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (141 commits)
openvswitch: fix conntrack netlink event delivery
qed: Protect the doorbell BAR with the write barriers.
neigh: Explicitly declare RCU-bh read side critical section in neigh_xmit()
e1000e: keep VLAN interfaces functional after rxvlan off
cfg80211: fix proto in ieee80211_data_to_8023 for frames without LLC header
qlcnic: use the correct ring in qlcnic_83xx_process_rcv_ring_diag()
bpf, perf: delay release of BPF prog after grace period
net: bridge: fix vlan stats continue counter
tcp: do not send too big packets at retransmit time
ibmvnic: fix to use list_for_each_safe() when delete items
net: thunderx: Fix TL4 configuration for secondary Qsets
net: thunderx: Fix link status reporting
net/mlx5e: Reorganize ethtool statistics
net/mlx5e: Fix number of PFC counters reported to ethtool
net/mlx5e: Prevent adding the same vxlan port
net/mlx5e: Check for BlueFlame capability before allocating SQ uar
net/mlx5e: Change enum to better reflect usage
net/mlx5: Add ConnectX-5 PCIe 4.0 to list of supported devices
net/mlx5: Update command strings
net: marvell: Add separate config ANEG function for Marvell 88E1111
...
Commit dead9f29dd ("perf: Fix race in BPF program unregister") moved
destruction of BPF program from free_event_rcu() callback to __free_event(),
which is problematic if used with tail calls: if prog A is attached as
trace event directly, but at the same time present in a tail call map used
by another trace event program elsewhere, then we need to delay destruction
via RCU grace period since it can still be in use by the program doing the
tail call (the prog first needs to be dropped from the tail call map, then
trace event with prog A attached destroyed, so we get immediate destruction).
Fixes: dead9f29dd ("perf: Fix race in BPF program unregister")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Cc: Jann Horn <jann@thejh.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull perf fixes from Ingo Molnar:
"A handful of tooling fixes, two PMU driver fixes and a cleanup of
redundant code that addresses a security analyzer false positive"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf/core: Remove a redundant check
perf/x86/intel/uncore: Remove SBOX support for Broadwell server
perf ctf: Convert invalid chars in a string before set value
perf record: Fix crash when kptr is restricted
perf symbols: Check kptr_restrict for root
perf/x86/intel/rapl: Fix pmus free during cleanup
There is no way to end up in _free_event() with event::pmu being NULL.
The latter is initialized in event allocation path and remains set
forever. In case of allocation failure, the error path doesn't use
_free_event().
Having the check, however, suggests that it is possible to have a
event::pmu==NULL situation in _free_event() and confuses the robots.
This patch gets rid of the check.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: eranian@google.com
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1465303455-26032-1-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull networking updates from David Miller:
"Highlights:
1) Support SPI based w5100 devices, from Akinobu Mita.
2) Partial Segmentation Offload, from Alexander Duyck.
3) Add GMAC4 support to stmmac driver, from Alexandre TORGUE.
4) Allow cls_flower stats offload, from Amir Vadai.
5) Implement bpf blinding, from Daniel Borkmann.
6) Optimize _ASYNC_ bit twiddling on sockets, unless the socket is
actually using FASYNC these atomics are superfluous. From Eric
Dumazet.
7) Run TCP more preemptibly, also from Eric Dumazet.
8) Support LED blinking, EEPROM dumps, and rxvlan offloading in mlx5e
driver, from Gal Pressman.
9) Allow creating ppp devices via rtnetlink, from Guillaume Nault.
10) Improve BPF usage documentation, from Jesper Dangaard Brouer.
11) Support tunneling offloads in qed, from Manish Chopra.
12) aRFS offloading in mlx5e, from Maor Gottlieb.
13) Add RFS and RPS support to SCTP protocol, from Marcelo Ricardo
Leitner.
14) Add MSG_EOR support to TCP, this allows controlling packet
coalescing on application record boundaries for more accurate
socket timestamp sampling. From Martin KaFai Lau.
15) Fix alignment of 64-bit netlink attributes across the board, from
Nicolas Dichtel.
16) Per-vlan stats in bridging, from Nikolay Aleksandrov.
17) Several conversions of drivers to ethtool ksettings, from Philippe
Reynes.
18) Checksum neutral ILA in ipv6, from Tom Herbert.
19) Factorize all of the various marvell dsa drivers into one, from
Vivien Didelot
20) Add VF support to qed driver, from Yuval Mintz"
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1649 commits)
Revert "phy dp83867: Fix compilation with CONFIG_OF_MDIO=m"
Revert "phy dp83867: Make rgmii parameters optional"
r8169: default to 64-bit DMA on recent PCIe chips
phy dp83867: Make rgmii parameters optional
phy dp83867: Fix compilation with CONFIG_OF_MDIO=m
bpf: arm64: remove callee-save registers use for tmp registers
asix: Fix offset calculation in asix_rx_fixup() causing slow transmissions
switchdev: pass pointer to fib_info instead of copy
net_sched: close another race condition in tcf_mirred_release()
tipc: fix nametable publication field in nl compat
drivers: net: Don't print unpopulated net_device name
qed: add support for dcbx.
ravb: Add missing free_irq() calls to ravb_close()
qed: Remove a stray tab
net: ethernet: fec-mpc52xx: use phy_ethtool_{get|set}_link_ksettings
net: ethernet: fec-mpc52xx: use phydev from struct net_device
bpf, doc: fix typo on bpf_asm descriptions
stmmac: hardware TX COE doesn't work when force_thresh_dma_mode is set
net: ethernet: fs-enet: use phy_ethtool_{get|set}_link_ksettings
net: ethernet: fs-enet: use phydev from struct net_device
...
The nf_conntrack_core.c fix in 'net' is not relevant in 'net-next'
because we no longer have a per-netns conntrack hash.
The ip_gre.c conflict as well as the iwlwifi ones were cases of
overlapping changes.
Conflicts:
drivers/net/wireless/intel/iwlwifi/mvm/tx.c
net/ipv4/ip_gre.c
net/netfilter/nf_conntrack_core.c
Signed-off-by: David S. Miller <davem@davemloft.net>
Allowing unprivileged kernel profiling lets any user dump follow kernel
control flow and dump kernel registers. This most likely allows trivial
kASLR bypassing, and it may allow other mischief as well. (Off the top
of my head, the PERF_SAMPLE_REGS_INTR output during /dev/urandom reads
could be quite interesting.)
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit:
2665784850 ("perf/core: Verify we have a single perf_hw_context PMU")
forcefully prevents multiple PMUs from sharing perf_hw_context, as this
generally doesn't make sense. It is a common bug for uncore PMUs to
use perf_hw_context rather than perf_invalid_context, which this detects.
However, systems exist with heterogeneous CPUs (and hence heterogeneous
HW PMUs), for which sharing perf_hw_context is necessary, and possible
in some limited cases.
To make this work we have to perform some gymnastics, as we did in these
commits:
66eb579e66 ("perf: allow for PMU-specific event filtering")
c904e32a69 ("arm: perf: filter unschedulable events")
To allow those systems to work, we must allow PMUs for heterogeneous
CPUs to share perf_hw_context, though we must still disallow sharing
otherwise to detect the common misuse of perf_hw_context.
This patch adds a new PERF_PMU_CAP_HETEROGENEOUS_CPUS for this, updates
the core logic to account for this, and makes use of it in the arm_pmu
code that is used for systems with heterogeneous CPUs. Comments are
added to make the rationale clear and hopefully avoid accidental abuse.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20160426103346.GA20836@leverpostej
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Export an additional common attribute for PMUs that support address range
filtering to let the perf userspace identify such PMUs in a uniform way.
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1461771888-10409-8-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Many instruction tracing PMUs out there support address range-based
filtering, which would, for example, generate trace data only for a
given range of instruction addresses, which is useful for tracing
individual functions, modules or libraries. Other PMUs may also
utilize this functionality to allow filtering to or filtering out
code at certain address ranges.
This patch introduces the interface for userspace to specify these
filters and for the PMU drivers to apply these filters to hardware
configuration.
The user interface is an ASCII string that is passed via an ioctl()
and specifies (in the form of an ASCII string) address ranges within
certain object files or within kernel. There is no special treatment
for kernel modules yet, but it might be a worthy pursuit.
The PMU driver interface basically adds two extra callbacks to the
PMU driver structure, one of which validates the filter configuration
proposed by the user against what the hardware is actually capable of
doing and the other one translates hardware-independent filter
configuration into something that can be programmed into the
hardware.
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1461771888-10409-6-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Trace filtering code needs an iterator that can go through all events in
a context, including inactive and filtered, to be able to update their
filters' address ranges based on mmap or exec events.
This patch changes perf_event_aux_ctx() to optionally do this.
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1461771888-10409-5-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
For instruction trace filtering, namely, for communicating filter
definitions from userspace, I'd like to re-use the SET_FILTER code
that the tracepoints are using currently.
To that end, move the relevant code out from behind the
CONFIG_EVENT_TRACING dependency.
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1461771888-10409-2-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Conflicts:
net/ipv4/ip_gre.c
Minor conflicts between tunnel bug fixes in net and
ipv6 tunnel cleanups in net-next.
Signed-off-by: David S. Miller <davem@davemloft.net>
Jann reported that the ptrace_may_access() check in
find_lively_task_by_vpid() is racy against exec().
Specifically:
perf_event_open() execve()
ptrace_may_access()
commit_creds()
... if (get_dumpable() != SUID_DUMP_USER)
perf_event_exit_task();
perf_install_in_context()
would result in installing a counter across the creds boundary.
Fix this by wrapping lots of perf_event_open() in cred_guard_mutex.
This should be fine as perf_event_exit_task() is already called with
cred_guard_mutex held, so all perf locks already nest inside it.
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This patch introduces 'write_backward' bit to perf_event_attr, which
controls the direction of a ring buffer. After set, the corresponding
ring buffer is written from end to beginning. This feature is design to
support reading from overwritable ring buffer.
Ring buffer can be created by mapping a perf event fd. Kernel puts event
records into ring buffer, user tooling like perf fetch them from
address returned by mmap(). To prevent racing between kernel and tooling,
they communicate to each other through 'head' and 'tail' pointers.
Kernel maintains 'head' pointer, points it to the next free area (tail
of the last record). Tooling maintains 'tail' pointer, points it to the
tail of last consumed record (record has already been fetched). Kernel
determines the available space in a ring buffer using these two
pointers to avoid overwrite unfetched records.
By mapping without 'PROT_WRITE', an overwritable ring buffer is created.
Different from normal ring buffer, tooling is unable to maintain 'tail'
pointer because writing is forbidden. Therefore, for this type of ring
buffers, kernel overwrite old records unconditionally, works like flight
recorder. This feature would be useful if reading from overwritable ring
buffer were as easy as reading from normal ring buffer. However,
there's an obscure problem.
The following figure demonstrates a full overwritable ring buffer. In
this figure, the 'head' pointer points to the end of last record, and a
long record 'E' is pending. For a normal ring buffer, a 'tail' pointer
would have pointed to position (X), so kernel knows there's no more
space in the ring buffer. However, for an overwritable ring buffer,
kernel ignore the 'tail' pointer.
(X) head
. |
. V
+------+-------+----------+------+---+
|A....A|B.....B|C........C|D....D| |
+------+-------+----------+------+---+
Record 'A' is overwritten by event 'E':
head
|
V
+--+---+-------+----------+------+---+
|.E|..A|B.....B|C........C|D....D|E..|
+--+---+-------+----------+------+---+
Now tooling decides to read from this ring buffer. However, none of these
two natural positions, 'head' and the start of this ring buffer, are
pointing to the head of a record. Even the full ring buffer can be
accessed by tooling, it is unable to find a position to start decoding.
The first attempt tries to solve this problem AFAIK can be found from
[1]. It makes kernel to maintain 'tail' pointer: updates it when ring
buffer is half full. However, this approach introduces overhead to
fast path. Test result shows a 1% overhead [2]. In addition, this method
utilizes no more tham 50% records.
Another attempt can be found from [3], which allows putting the size of
an event at the end of each record. This approach allows tooling to find
records in a backward manner from 'head' pointer by reading size of a
record from its tail. However, because of alignment requirement, it
needs 8 bytes to record the size of a record, which is a huge waste. Its
performance is also not good, because more data need to be written.
This approach also introduces some extra branch instructions to fast
path.
'write_backward' is a better solution to this problem.
Following figure demonstrates the state of the overwritable ring buffer
when 'write_backward' is set before overwriting:
head
|
V
+---+------+----------+-------+------+
| |D....D|C........C|B.....B|A....A|
+---+------+----------+-------+------+
and after overwriting:
head
|
V
+---+------+----------+-------+---+--+
|..E|D....D|C........C|B.....B|A..|E.|
+---+------+----------+-------+---+--+
In each situation, 'head' points to the beginning of the newest record.
From this record, tooling can iterate over the full ring buffer and fetch
records one by one.
The only limitation that needs to be considered is back-to-back reading.
Due to the non-deterministic of user programs, it is impossible to ensure
the ring buffer keeps stable during reading. Consider an extreme situation:
tooling is scheduled out after reading record 'D', then a burst of events
come, eat up the whole ring buffer (one or multiple rounds). When the
tooling process comes back, reading after 'D' is incorrect now.
To prevent this problem, we need to find a way to ensure the ring buffer
is stable during reading. ioctl(PERF_EVENT_IOC_PAUSE_OUTPUT) is
suggested because its overhead is lower than
ioctl(PERF_EVENT_IOC_ENABLE).
By carefully verifying 'header' pointer, reader can avoid pausing the
ring-buffer. For example:
/* A union of all possible events */
union perf_event event;
p = head = perf_mmap__read_head();
while (true) {
/* copy header of next event */
fetch(&event.header, p, sizeof(event.header));
/* read 'head' pointer */
head = perf_mmap__read_head();
/* check overwritten: is the header good? */
if (!verify(sizeof(event.header), p, head))
break;
/* copy the whole event */
fetch(&event, p, event.header.size);
/* read 'head' pointer again */
head = perf_mmap__read_head();
/* is the whole event good? */
if (!verify(event.header.size, p, head))
break;
p += event.header.size;
}
However, the overhead is high because:
a) In-place decoding is not safe.
Copying-verifying-decoding is required.
b) Fetching 'head' pointer requires additional synchronization.
(From Alexei Starovoitov:
Even when this trick works, pause is needed for more than stability of
reading. When we collect the events into overwrite buffer we're waiting
for some other trigger (like all cpu utilization spike or just one cpu
running and all others are idle) and when it happens the buffer has
valuable info from the past. At this point new events are no longer
interesting and buffer should be paused, events read and unpaused until
next trigger comes.)
This patch utilizes event's default overflow_handler introduced
previously. perf_event_output_backward() is created as the default
overflow handler for backward ring buffers. To avoid extra overhead to
fast path, original perf_event_output() becomes __perf_event_output()
and marked '__always_inline'. In theory, there's no extra overhead
introduced to fast path.
Performance testing:
Calling 3000000 times of 'close(-1)', use gettimeofday() to check
duration. Use 'perf record -o /dev/null -e raw_syscalls:*' to capture
system calls. In ns.
Testing environment:
CPU : Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz
Kernel : v4.5.0
MEAN STDVAR
BASE 800214.950 2853.083
PRE1 2253846.700 9997.014
PRE2 2257495.540 8516.293
POST 2250896.100 8933.921
Where 'BASE' is pure performance without capturing. 'PRE1' is test
result of pure 'v4.5.0' kernel. 'PRE2' is test result before this
patch. 'POST' is test result after this patch. See [4] for the detailed
experimental setup.
Considering the stdvar, this patch doesn't introduce performance
overhead to the fast path.
[1] http://lkml.iu.edu/hypermail/linux/kernel/1304.1/04584.html
[2] http://lkml.iu.edu/hypermail/linux/kernel/1307.1/00535.html
[3] http://lkml.iu.edu/hypermail/linux/kernel/1512.0/01265.html
[4] http://lkml.kernel.org/g/56F89DCD.1040202@huawei.com
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Cc: <acme@kernel.org>
Cc: <pi3orama@163.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/r/1459865478-53413-1-git-send-email-wangnan0@huawei.com
[ Fixed the changelog some more. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Markus reported that 0 should also disable the throttling we per
Documentation/sysctl/kernel.txt.
Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Fixes: 91a612eea9 ("perf/core: Fix dynamic interrupt throttle")
Signed-off-by: Ingo Molnar <mingo@kernel.org>
move trace_call_bpf() into helper function to minimize the size
of perf_trace_*() tracepoint handlers.
text data bss dec hex filename
10541679 5526646 2945024 19013349 1221ee5 vmlinux_before
10509422 5526646 2945024 18981092 121a0e4 vmlinux_after
It may seem that perf_fetch_caller_regs() can also be moved,
but that is incorrect, since ip/sp will be wrong.
bpf+tracepoint performance is not affected, since
perf_swevent_put_recursion_context() is now inlined.
export_symbol_gpl can also be dropped.
No measurable change in normal perf tracepoints.
Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
during bpf program loading remember the last byte of ctx access
and at the time of attaching the program to tracepoint check that
the program doesn't access bytes beyond defined in tracepoint fields
This also disallows access to __dynamic_array fields, but can be
relaxed in the future.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
introduce BPF_PROG_TYPE_TRACEPOINT program type and allow it to be attached
to the perf tracepoint handler, which will copy the arguments into
the per-cpu buffer and pass it to the bpf program as its first argument.
The layout of the fields can be discovered by doing
'cat /sys/kernel/debug/tracing/events/sched/sched_switch/format'
prior to the compilation of the program with exception that first 8 bytes
are reserved and not accessible to the program. This area is used to store
the pointer to 'struct pt_regs' which some of the bpf helpers will use:
+---------+
| 8 bytes | hidden 'struct pt_regs *' (inaccessible to bpf program)
+---------+
| N bytes | static tracepoint fields defined in tracepoint/format (bpf readonly)
+---------+
| dynamic | __dynamic_array bytes of tracepoint (inaccessible to bpf yet)
+---------+
Not that all of the fields are already dumped to user space via perf ring buffer
and broken application access it directly without consulting tracepoint/format.
Same rule applies here: static tracepoint fields should only be accessed
in a format defined in tracepoint/format. The order of fields and
field sizes are not an ABI.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
split allows to move expensive update of 'struct trace_entry' to later phase.
Repurpose unused 1st argument of perf_tp_event() to indicate event type.
While splitting use temp variable 'rctx' instead of '*rctx' to avoid
unnecessary loads done by the compiler due to -fno-strict-aliasing
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set a default event->overflow_handler in perf_event_alloc() so don't
need to check event->overflow_handler in __perf_event_overflow().
Following commits can give a different default overflow_handler.
Initial idea comes from Peter:
http://lkml.kernel.org/r/20130708121557.GA17211@twins.programming.kicks-ass.net
Since the default value of event->overflow_handler is not NULL, existing
'if (!overflow_handler)' checks need to be changed.
is_default_overflow_handler() is introduced for this.
No extra performance overhead is introduced into the hot path because in the
original code we still need to read this handler from memory. A conditional
branch is avoided so actually we remove some instructions.
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <pi3orama@163.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/r/1459147292-239310-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Add new ioctl() to pause/resume ring-buffer output.
In some situations we want to read from the ring-buffer only when we
ensure nothing can write to the ring-buffer during reading. Without
this patch we have to turn off all events attached to this ring-buffer
to achieve this.
This patch is a prerequisite to enable overwrite support for the
perf ring-buffer support. Following commits will introduce new methods
support reading from overwrite ring buffer. Before reading, caller
must ensure the ring buffer is frozen, or the reading is unreliable.
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <pi3orama@163.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/r/1459147292-239310-2-git-send-email-wangnan0@huawei.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Now that we can ensure that when ring buffer's AUX area is on the way
to getting unmapped new transactions won't start, we only need to stop
all events that can potentially be writing aux data to our ring buffer.
Having done that, we can safely free the AUX pages and corresponding
PMU data, as this time it is guaranteed to be the last aux reference
holder.
This partially reverts:
57ffc5ca67 ("perf: Fix AUX buffer refcounting")
... which was made to defer deallocation that was otherwise possible
from an NMI context. Now it is no longer the case; the last call to
rb_free_aux() that drops the last AUX reference has to happen in
perf_mmap_close() on that AUX area.
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/87d1qtz23d.fsf@ashishki-desk.ger.corp.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
There should (and can) only be a single PMU for perf_hw_context
events.
This is because of how we schedule events: once a hardware event fails to
schedule (the PMU is 'full') we stop trying to add more. The trivial
'fix' would break the Round-Robin scheduling we do.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
In the error path, event_file not being NULL is used to determine
whether the event itself still needs to be free'd, so fix it up to
avoid leaking.
Reported-by: Leon Yu <chianglungyu@gmail.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Fixes: 130056275a ("perf: Do not double free")
Link: http://lkml.kernel.org/r/87twk06yxp.fsf@ashishki-desk.ger.corp.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Stephane reported that commit:
3cbaa59069 ("perf: Fix ctx time tracking by introducing EVENT_TIME")
introduced a regression wrt. time tracking, as easily observed by:
> This patch introduce a bug in the time tracking of events when
> multiplexing is used.
>
> The issue is easily reproducible with the following perf run:
>
> $ perf stat -a -C 0 -e branches,branches,branches,branches,branches,branches -I 1000
> 1.000730239 652,394 branches (66.41%)
> 1.000730239 597,809 branches (66.41%)
> 1.000730239 593,870 branches (66.63%)
> 1.000730239 651,440 branches (67.03%)
> 1.000730239 656,725 branches (66.96%)
> 1.000730239 <not counted> branches
>
> One branches event is shown as not having run. Yet, with
> multiplexing, all events should run especially with a 1s (-I 1000)
> interval. The delta for time_running comes out to 0. Yet, the event
> has run because the kernel is actually multiplexing the events. The
> problem is that the time tracking is the kernel and especially in
> ctx_sched_out() is wrong now.
>
> The problem is that in case that the kernel enters ctx_sched_out() with the
> following state:
> ctx->is_active=0x7 event_type=0x1
> Call Trace:
> [<ffffffff813ddd41>] dump_stack+0x63/0x82
> [<ffffffff81182bdc>] ctx_sched_out+0x2bc/0x2d0
> [<ffffffff81183896>] perf_mux_hrtimer_handler+0xf6/0x2c0
> [<ffffffff811837a0>] ? __perf_install_in_context+0x130/0x130
> [<ffffffff810f5818>] __hrtimer_run_queues+0xf8/0x2f0
> [<ffffffff810f6097>] hrtimer_interrupt+0xb7/0x1d0
> [<ffffffff810509a8>] local_apic_timer_interrupt+0x38/0x60
> [<ffffffff8175ca9d>] smp_apic_timer_interrupt+0x3d/0x50
> [<ffffffff8175ac7c>] apic_timer_interrupt+0x8c/0xa0
>
> In that case, the test:
> if (is_active & EVENT_TIME)
>
> will be false and the time will not be updated. Time must always be updated on
> sched out.
Fix this by always updating time if EVENT_TIME was set, as opposed to
only updating time when EVENT_TIME changed.
Reported-by: Stephane Eranian <eranian@google.com>
Tested-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: kan.liang@intel.com
Cc: namhyung@kernel.org
Fixes: 3cbaa59069 ("perf: Fix ctx time tracking by introducing EVENT_TIME")
Link: http://lkml.kernel.org/r/20160329072644.GB3408@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
There were two problems with the dynamic interrupt throttle mechanism,
both triggered by the same action.
When you (or perf_fuzzer) write a huge value into
/proc/sys/kernel/perf_event_max_sample_rate the computed
perf_sample_allowed_ns becomes 0. This effectively disables the whole
dynamic throttle.
This is fixed by ensuring update_perf_cpu_limits() never sets the
value to 0. However, we allow disabling of the dynamic throttle by
writing 100 to /proc/sys/kernel/perf_cpu_time_max_percent. This will
generate a warning in dmesg.
The second problem is that by setting the max_sample_rate to a huge
number, the adaptive process can take a few tries, since it halfs the
limit each time. Change that to directly compute a new value based on
the observed duration.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Its possible to IOC_PERIOD while the event is throttled, this would
re-start the event and the next tick would then try to unthrottle it,
and find the event wasn't actually stopped anymore.
This would tickle a WARN in the x86-pmu code which isn't expecting to
start a !stopped event.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: dvyukov@google.com
Cc: oleg@redhat.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160310143924.GR6356@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull NOHZ updates from Ingo Molnar:
"NOHZ enhancements, by Frederic Weisbecker, which reorganizes/refactors
the NOHZ 'can the tick be stopped?' infrastructure and related code to
be data driven, and harmonizes the naming and handling of all the
various properties"
[ This makes the ugly "fetch_or()" macro that the scheduler used
internally a new generic helper, and does a bad job at it.
I'm pulling it, but I've asked Ingo and Frederic to get this
fixed up ]
* 'timers-nohz-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched-clock: Migrate to use new tick dependency mask model
posix-cpu-timers: Migrate to use new tick dependency mask model
sched: Migrate sched to use new tick dependency mask model
sched: Account rr tasks
perf: Migrate perf to use new tick dependency mask model
nohz: Use enum code for tick stop failure tracing message
nohz: New tick dependency mask
nohz: Implement wide kick on top of irq work
atomic: Export fetch_or()
Pull nohz enhancements from Frederic Weisbecker:
"Currently in nohz full configs, the tick dependency is checked
asynchronously by nohz code from interrupt and context switch for each
concerned subsystem with a set of function provided by these. Such
functions are made of many conditions and details that can be heavyweight
as they are called on fastpath: sched_can_stop_tick(),
posix_cpu_timer_can_stop_tick(), perf_event_can_stop_tick()...
Thomas suggested a few months ago to make that tick dependency check
synchronous. Instead of checking subsystems details from each interrupt
to guess if the tick can be stopped, every subsystem that may have a tick
dependency should set itself a flag specifying the state of that
dependency. This way we can verify if we can stop the tick with a single
lightweight mask check on fast path.
This conversion from a pull to a push model to implement tick dependency
is the core feature of this patchset that is split into:
* Nohz wide kick simplification
* Improve nohz tracing
* Introduce tick dependency mask
* Migrate scheduler, posix timers, perf events and sched clock tick
dependencies to the tick dependency mask."
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The error path in perf_event_open() is such that asking for a sampling
event on a PMU that doesn't generate interrupts will end up in dropping
the perf_sched_count even though it hasn't been incremented for this
event yet.
Given a sufficient amount of these calls, we'll end up disabling
scheduler's jump label even though we'd still have active events in the
system, thereby facilitating the arrival of the infernal regions upon us.
I'm fixing this by moving account_event() inside perf_event_alloc().
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <stable@vger.kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1456917854-29427-1-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Instead of providing asynchronous checks for the nohz subsystem to verify
perf event tick dependency, migrate perf to the new mask.
Perf needs the tick for two situations:
1) Freq events. We could set the tick dependency when those are
installed on a CPU context. But setting a global dependency on top of
the global freq events accounting is much easier. If people want that
to be optimized, we can still refine that on the per-CPU tick dependency
level. This patch dooesn't change the current behaviour anyway.
2) Throttled events: this is a per-cpu dependency.
Reviewed-by: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Since there is no serialization between task_function_call() doing
task_curr() and the other CPU doing context switches, we could end
up not sending an IPI even if we had to.
And I'm not sure I still buy my own argument we're OK.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dvyukov@google.com
Cc: eranian@google.com
Cc: oleg@redhat.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160224174948.340031200@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Completely reworks perf_install_in_context() (again!) in order to
ensure that there will be no ctx time hole between add_event_to_ctx()
and any potential ctx_sched_in().
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dvyukov@google.com
Cc: eranian@google.com
Cc: oleg@redhat.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160224174948.279399438@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The recent commit 3e349507d1 ("perf: Fix perf_enable_on_exec() event
scheduling") caused this by moving task_ctx_sched_out() from before
__perf_event_mask_enable() to after it.
The overlooked consequence of that change is that task_ctx_sched_out()
would update the ctx time fields, and now __perf_event_mask_enable()
uses stale time.
In order to fix this, explicitly stop our context's time before
enabling the event(s).
Reported-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dvyukov@google.com
Cc: eranian@google.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Cc: vince@deater.net
Fixes: 3e349507d1 ("perf: Fix perf_enable_on_exec() event scheduling")
Link: http://lkml.kernel.org/r/20160224174948.159242158@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Currently any ctx_sched_in() call will re-start the ctx time tracking,
this means that calls like:
ctx_sched_in(.event_type = EVENT_PINNED);
ctx_sched_in(.event_type = EVENT_FLEXIBLE);
will have a hole in their ctx time tracking. This is likely harmless
but can confuse things a little. By adding EVENT_TIME, we can have the
first ctx_sched_in() (is_active: 0 -> !0) start the time and any
further ctx_sched_in() will leave the timestamps alone.
Secondly, this allows for an early disable like:
ctx_sched_out(.event_type = EVENT_TIME);
which would update the ctx time (if the ctx is active) and any further
calls to ctx_sched_out() would not further modify the ctx time.
For ctx_sched_in() any 0 -> !0 transition will automatically include
EVENT_TIME.
For ctx_sched_out(), any transition that clears EVENT_ALL will
automatically clear EVENT_TIME.
These two rules ensure that under normal circumstances we need not
bother with EVENT_TIME and get natural ctx time behaviour.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dvyukov@google.com
Cc: eranian@google.com
Cc: oleg@redhat.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160224174948.100446561@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Because event_sched_out() checks event->pending_disable _before_
actually disabling the event, it can happen that the event fires after
it checks but before it gets disabled.
This would leave event->pending_disable set and the queued irq_work
will try and process it.
However, if the event trigger was during schedule(), the event might
have been de-scheduled by the time the irq_work runs, and
perf_event_disable_local() will fail.
Fix this by checking event->pending_disable _after_ we call
event->pmu->del(). This depends on the latter being a compiler
barrier, such that the compiler does not lift the load and re-creates
the problem.
Tested-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dvyukov@google.com
Cc: eranian@google.com
Cc: oleg@redhat.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160224174948.040469884@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
perf_install_in_context() relies upon the context switch hooks to have
scheduled in events when the IPI misses its target -- after all, if
the task has moved from the CPU (or wasn't running at all), it will
have to context switch to run elsewhere.
This however doesn't appear to be happening.
It is possible for the IPI to not happen (task wasn't running) only to
later observe the task running with an inactive context.
The only possible explanation is that the context switch hooks are not
called. Therefore put in a sync_sched() after toggling the jump_label
to guarantee all CPUs will have them enabled before we install an
event.
A simple if (0->1) sync_sched() will not in fact work, because any
further increment can race and complete before the sync_sched().
Therefore we must jump through some hoops.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dvyukov@google.com
Cc: eranian@google.com
Cc: oleg@redhat.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160224174947.980211985@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Alexander reported that when the 'original' context gets destroyed, no
new clones happen.
This can happen irrespective of the ctx switch optimization, any task
can die, even the parent, and we want to continue monitoring the task
hierarchy until we either close the event or no tasks are left in the
hierarchy.
perf_event_init_context() will attempt to pin the 'parent' context
during clone(). At that point current is the parent, and since current
cannot have exited while executing clone(), its context cannot have
passed through perf_event_exit_task_context(). Therefore
perf_pin_task_context() cannot observe ctx->task == TASK_TOMBSTONE.
However, since inherit_event() does:
if (parent_event->parent)
parent_event = parent_event->parent;
it looks at the 'original' event when it does: is_orphaned_event().
This can return true if the context that contains the this event has
passed through perf_event_exit_task_context(). And thus we'll fail to
clone the perf context.
Fix this by adding a new state: STATE_DEAD, which is set by
perf_release() to indicate that the filedesc (or kernel reference) is
dead and there are no observers for our data left.
Only for STATE_DEAD will is_orphaned_event() be true and inhibit
cloning.
STATE_EXIT is otherwise preserved such that is_event_hup() remains
functional and will report when the observed task hierarchy becomes
empty.
Reported-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Tested-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dvyukov@google.com
Cc: eranian@google.com
Cc: oleg@redhat.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Cc: vince@deater.net
Fixes: c6e5b73242 ("perf: Synchronously clean up child events")
Link: http://lkml.kernel.org/r/20160224174947.919845295@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Daniel Borkmann