linux/kernel/events/core.c, branch v4.19

perf/ring_buffer: Prevent concurent ring buffer access

2018-10-02T07:37:59Z

Some of the scheduling tracepoints allow the perf_tp_event code to write to ring buffer under different cpu than the code is running on. This results in corrupted ring buffer data demonstrated in following perf commands: # perf record -e 'sched:sched_switch,sched:sched_wakeup' perf bench sched messaging # Running 'sched/messaging' benchmark: # 20 sender and receiver processes per group # 10 groups == 400 processes run Total time: 0.383 [sec] [ perf record: Woken up 8 times to write data ] 0x42b890 [0]: failed to process type: -1765585640 [ perf record: Captured and wrote 4.825 MB perf.data (29669 samples) ] # perf report --stdio 0x42b890 [0]: failed to process type: -1765585640 The reason for the corruption are some of the scheduling tracepoints, that have __perf_task dfined and thus allow to store data to another cpu ring buffer: sched_waking sched_wakeup sched_wakeup_new sched_stat_wait sched_stat_sleep sched_stat_iowait sched_stat_blocked The perf_tp_event function first store samples for current cpu related events defined for tracepoint: hlist_for_each_entry_rcu(event, head, hlist_entry) perf_swevent_event(event, count, &data, regs); And then iterates events of the 'task' and store the sample for any task's event that passes tracepoint checks: ctx = rcu_dereference(task->perf_event_ctxp[perf_sw_context]); list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { if (event->attr.type != PERF_TYPE_TRACEPOINT) continue; if (event->attr.config != entry->type) continue; perf_swevent_event(event, count, &data, regs); } Above code can race with same code running on another cpu, ending up with 2 cpus trying to store under the same ring buffer, which is specifically not allowed. This patch prevents the problem, by allowing only events with the same current cpu to receive the event. NOTE: this requires the use of (per-task-)per-cpu buffers for this feature to work; perf-record does this. Signed-off-by: Jiri Olsa [peterz: small edits to Changelog] Signed-off-by: Peter Zijlstra (Intel) Cc: Alexander Shishkin Cc: Andrew Vagin Cc: Arnaldo Carvalho de Melo Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Fixes: e6dab5ffab59 ("perf/trace: Add ability to set a target task for events") Link: http://lkml.kernel.org/r/20180923161343.GB15054@krava Signed-off-by: Ingo Molnar

perf/core: Fix perf_pmu_unregister() locking

2018-10-02T07:37:56Z

When we unregister a PMU, we fail to serialize the @pmu_idr properly. Fix that by doing the entire thing under pmu_lock. Signed-off-by: Peter Zijlstra (Intel) Cc: Alexander Shishkin Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Fixes: 2e80a82a49c4 ("perf: Dynamic pmu types") Signed-off-by: Ingo Molnar

perf/core: Add sanity check to deal with pinned event failure

2018-09-28T20:44:53Z

It is possible that a failure can occur during the scheduling of a pinned event. The initial portion of perf_event_read_local() contains the various error checks an event should pass before it can be considered valid. Ensure that the potential scheduling failure of a pinned event is checked for and have a credible error. Suggested-by: Peter Zijlstra Signed-off-by: Reinette Chatre Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra (Intel) Cc: fenghua.yu@intel.com Cc: tony.luck@intel.com Cc: acme@kernel.org Cc: gavin.hindman@intel.com Cc: jithu.joseph@intel.com Cc: dave.hansen@intel.com Cc: hpa@zytor.com Cc: stable@vger.kernel.org Link: https://lkml.kernel.org/r/6486385d1f30336e9973b24c8c65f5079543d3d3.1537377064.git.reinette.chatre@intel.com

perf/core: Force USER_DS when recording user stack data

2018-09-10T12:01:46Z

Perf can record user stack data in response to a synchronous request, such as a tracepoint firing. If this happens under set_fs(KERNEL_DS), then we end up reading user stack data using __copy_from_user_inatomic() under set_fs(KERNEL_DS). I think this conflicts with the intention of using set_fs(KERNEL_DS). And it is explicitly forbidden by hardware on ARM64 when both CONFIG_ARM64_UAO and CONFIG_ARM64_PAN are used. So fix this by forcing USER_DS when recording user stack data. Signed-off-by: Yabin Cui Acked-by: Peter Zijlstra (Intel) Cc: Cc: Alexander Shishkin Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Thomas Gleixner Fixes: 88b0193d9418 ("perf/callchain: Force USER_DS when invoking perf_callchain_user()") Link: http://lkml.kernel.org/r/20180823225935.27035-1-yabinc@google.com Signed-off-by: Ingo Molnar

Merge tag 'perf-urgent-for-mingo-4.19-20180903' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent

2018-09-09T19:36:31Z

Pull perf/urgent fixes from Arnaldo Carvalho de Melo: Kernel: - Modify breakpoint fixes (Jiri Olsa) perf annotate: - Fix parsing aarch64 branch instructions after objdump update (Kim Phillips) - Fix parsing indirect calls in 'perf annotate' (Martin Liška) perf probe: - Ignore SyS symbols irrespective of endianness on PowerPC (Sandipan Das) perf trace: - Fix include path for asm-generic/unistd.h on arm64 (Kim Phillips) Core libraries: - Fix potential null pointer dereference in perf_evsel__new_idx() (Hisao Tanabe) - Use fixed size string for comms instead of scanf("%m"), that is not present in the bionic libc and leads to a crash (Chris Phlipot) - Fix bad memory access in trace info on 32-bit systems, we were reading 8 bytes from a 4-byte long variable when saving the command line in the perf.data file. (Chris Phlipot) Build system: - Streamline bpf examples and headers installation, clarifying some install messages. (Arnaldo Carvalho de Melo) Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: Ingo Molnar

perf/hw_breakpoint: Simplify breakpoint enable in perf_event_modify_breakpoint

2018-08-30T17:49:24Z

We can safely enable the breakpoint back for both the fail and success paths by checking only the bp->attr.disabled, which either holds the new 'requested' disabled state or the original breakpoint state. Committer testing: At the end of the series, the 'perf test' entry introduced as the first patch now runs to completion without finding the fixed issues: # perf test "bp modify" 62: x86 bp modify : Ok # In verbose mode: # perf test -v "bp modify" 62: x86 bp modify : --- start --- test child forked, pid 5161 rip 5950a0, bp_1 0x5950a0 in bp_1 rip 5950a0, bp_1 0x5950a0 in bp_1 test child finished with 0 ---- end ---- x86 bp modify: Ok Suggested-by: Oleg Nesterov Acked-by: Oleg Nesterov Signed-off-by: Jiri Olsa Tested-by: Arnaldo Carvalho de Melo Cc: Alexander Shishkin Cc: David Ahern Cc: Milind Chabbi Cc: Namhyung Kim Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/20180827091228.2878-6-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo

Merge branch 'siginfo-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace

2018-08-21T20:47:29Z

Pull core signal handling updates from Eric Biederman: "It was observed that a periodic timer in combination with a sufficiently expensive fork could prevent fork from every completing. This contains the changes to remove the need for that restart. This set of changes is split into several parts: - The first part makes PIDTYPE_TGID a proper pid type instead something only for very special cases. The part starts using PIDTYPE_TGID enough so that in __send_signal where signals are actually delivered we know if the signal is being sent to a a group of processes or just a single process. - With that prep work out of the way the logic in fork is modified so that fork logically makes signals received while it is running appear to be received after the fork completes" * 'siginfo-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: (22 commits) signal: Don't send signals to tasks that don't exist signal: Don't restart fork when signals come in. fork: Have new threads join on-going signal group stops fork: Skip setting TIF_SIGPENDING in ptrace_init_task signal: Add calculate_sigpending() fork: Unconditionally exit if a fatal signal is pending fork: Move and describe why the code examines PIDNS_ADDING signal: Push pid type down into complete_signal. signal: Push pid type down into __send_signal signal: Push pid type down into send_signal signal: Pass pid type into do_send_sig_info signal: Pass pid type into send_sigio_to_task & send_sigurg_to_task signal: Pass pid type into group_send_sig_info signal: Pass pid and pid type into send_sigqueue posix-timers: Noralize good_sigevent signal: Use PIDTYPE_TGID to clearly store where file signals will be sent pid: Implement PIDTYPE_TGID pids: Move the pgrp and session pid pointers from task_struct to signal_struct kvm: Don't open code task_pid in kvm_vcpu_ioctl pids: Compute task_tgid using signal->leader_pid ...

Merge tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux

2018-08-14T23:39:13Z

Pull arm64 updates from Will Deacon: "A bunch of good stuff in here. Worth noting is that we've pulled in the x86/mm branch from -tip so that we can make use of the core ioremap changes which allow us to put down huge mappings in the vmalloc area without screwing up the TLB. Much of the positive diffstat is because of the rseq selftest for arm64. Summary: - Wire up support for qspinlock, replacing our trusty ticket lock code - Add an IPI to flush_icache_range() to ensure that stale instructions fetched into the pipeline are discarded along with the I-cache lines - Support for the GCC "stackleak" plugin - Support for restartable sequences, plus an arm64 port for the selftest - Kexec/kdump support on systems booting with ACPI - Rewrite of our syscall entry code in C, which allows us to zero the GPRs on entry from userspace - Support for chained PMU counters, allowing 64-bit event counters to be constructed on current CPUs - Ensure scheduler topology information is kept up-to-date with CPU hotplug events - Re-enable support for huge vmalloc/IO mappings now that the core code has the correct hooks to use break-before-make sequences - Miscellaneous, non-critical fixes and cleanups" * tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux: (90 commits) arm64: alternative: Use true and false for boolean values arm64: kexec: Add comment to explain use of __flush_icache_range() arm64: sdei: Mark sdei stack helper functions as static arm64, kaslr: export offset in VMCOREINFO ELF notes arm64: perf: Add cap_user_time aarch64 efi/libstub: Only disable stackleak plugin for arm64 arm64: drop unused kernel_neon_begin_partial() macro arm64: kexec: machine_kexec should call __flush_icache_range arm64: svc: Ensure hardirq tracing is updated before return arm64: mm: Export __sync_icache_dcache() for xen-privcmd drivers/perf: arm-ccn: Use devm_ioremap_resource() to map memory arm64: Add support for STACKLEAK gcc plugin arm64: Add stack information to on_accessible_stack drivers/perf: hisi: update the sccl_id/ccl_id when MT is supported arm64: fix ACPI dependencies rseq/selftests: Add support for arm64 arm64: acpi: fix alignment fault in accessing ACPI efi/arm: map UEFI memory map even w/o runtime services enabled efi/arm: preserve early mapping of UEFI memory map longer for BGRT drivers: acpi: add dependency of EFI for arm64 ...

arm64: perf: Add cap_user_time aarch64

2018-07-31T09:14:00Z

It is useful to get the running time of a thread. Doing so in an efficient manner can be important for performance of user applications. Avoiding system calls in `clock_gettime` when handling CLOCK_THREAD_CPUTIME_ID is important. Other clocks are handled in the VDSO, but CLOCK_THREAD_CPUTIME_ID falls back on the system call. CLOCK_THREAD_CPUTIME_ID is not handled in the VDSO since it would have costs associated with maintaining updated user space accessible time offsets. These offsets have to be updated everytime the a thread is scheduled/descheduled. However, for programs regularly checking the running time of a thread, this is a performance improvement. This patch takes a middle ground, and adds support for cap_user_time an optional feature of the perf_event API. This way costs are only incurred when the perf_event api is enabled. This is done the same way as it is in x86. Ultimately this allows calculating the thread running time in userspace on aarch64 as follows (adapted from perf_event_open manpage): u32 seq, time_mult, time_shift; u64 running, count, time_offset, quot, rem, delta; struct perf_event_mmap_page *pc; pc = buf; // buf is the perf event mmaped page as documented in the API. if (pc->cap_usr_time) { do { seq = pc->lock; barrier(); running = pc->time_running; count = readCNTVCT_EL0(); // Read ARM hardware clock. time_offset = pc->time_offset; time_mult = pc->time_mult; time_shift = pc->time_shift; barrier(); } while (pc->lock != seq); quot = (count >> time_shift); rem = count & (((u64)1 << time_shift) - 1); delta = time_offset + quot * time_mult + ((rem * time_mult) >> time_shift); running += delta; // running now has the current nanosecond level thread time. } Summary of changes in the patch: For aarch64 systems, make arch_perf_update_userpage update the timing information stored in the perf_event page. Requiring the following calculations: - Calculate the appropriate time_mult, and time_shift factors to convert ticks to nano seconds for the current clock frequency. - Adjust the mult and shift factors to avoid shift factors of 32 bits. (possibly unnecessary) - The time_offset userspace should apply when doing calculations: negative the current sched time (now), because time_running and time_enabled fields of the perf_event page have just been updated. Toggle bits to appropriate values: - Enable cap_user_time Signed-off-by: Michael O'Farrell Signed-off-by: Will Deacon

Merge branch 'perf/urgent' into perf/core, to pick up fixes

2018-07-25T09:47:02Z

Signed-off-by: Ingo Molnar