lkml.org 
[lkml]   [2019]   [Aug]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH V4 1/3] riscv: Add perf callchain support
Hi Mao,

Mao Han <han_mao@c-sky.com> 於 2019年8月20日 週二 下午4:57寫道:
>
> This patch add support for perf callchain sampling on riscv platform.
> The return address of leaf function is retrieved from pt_regs as
> it is not saved in the outmost frame.
>
> Signed-off-by: Mao Han <han_mao@c-sky.com>
> Cc: Paul Walmsley <paul.walmsley@sifive.com>
> Cc: Greentime Hu <green.hu@gmail.com>
> Cc: Palmer Dabbelt <palmer@sifive.com>
> Cc: linux-riscv <linux-riscv@lists.infradead.org>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: Guo Ren <guoren@kernel.org>
> ---
> arch/riscv/Makefile | 3 +
> arch/riscv/kernel/Makefile | 3 +-
> arch/riscv/kernel/perf_callchain.c | 115 +++++++++++++++++++++++++++++++++++++
> 3 files changed, 120 insertions(+), 1 deletion(-)
> create mode 100644 arch/riscv/kernel/perf_callchain.c
>
> diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
> index 7a117be..946565b 100644
> --- a/arch/riscv/Makefile
> +++ b/arch/riscv/Makefile
> @@ -54,6 +54,9 @@ endif
> ifeq ($(CONFIG_MODULE_SECTIONS),y)
> KBUILD_LDFLAGS_MODULE += -T $(srctree)/arch/riscv/kernel/module.lds
> endif
> +ifeq ($(CONFIG_PERF_EVENTS),y)
> + KBUILD_CFLAGS += -fno-omit-frame-pointer
> +endif
>
> KBUILD_CFLAGS_MODULE += $(call cc-option,-mno-relax)
>
> diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
> index 2420d37..b1bea89 100644
> --- a/arch/riscv/kernel/Makefile
> +++ b/arch/riscv/kernel/Makefile
> @@ -38,6 +38,7 @@ obj-$(CONFIG_MODULE_SECTIONS) += module-sections.o
> obj-$(CONFIG_FUNCTION_TRACER) += mcount.o ftrace.o
> obj-$(CONFIG_DYNAMIC_FTRACE) += mcount-dyn.o
>
> -obj-$(CONFIG_PERF_EVENTS) += perf_event.o
> +obj-$(CONFIG_PERF_EVENTS) += perf_event.o
> +obj-$(CONFIG_PERF_EVENTS) += perf_callchain.o
>
> clean:
> diff --git a/arch/riscv/kernel/perf_callchain.c b/arch/riscv/kernel/perf_callchain.c
> new file mode 100644
> index 0000000..d75d15c
> --- /dev/null
> +++ b/arch/riscv/kernel/perf_callchain.c
> @@ -0,0 +1,115 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/* Copyright (C) 2019 Hangzhou C-SKY Microsystems co.,ltd. */
> +
> +#include <linux/perf_event.h>
> +#include <linux/uaccess.h>
> +
> +/* Kernel callchain */
> +struct stackframe {
> + unsigned long fp;
> + unsigned long ra;
> +};
> +
> +static int unwind_frame_kernel(struct stackframe *frame)
> +{
> + if (kstack_end((void *)frame->fp))
> + return -EPERM;
> + if (frame->fp & 0x3 || frame->fp < TASK_SIZE)
> + return -EPERM;
> + if (frame->fp < CONFIG_PAGE_OFFSET)
> + return -EPERM;
> +
> + *frame = *((struct stackframe *)frame->fp - 1);
> + if (__kernel_text_address(frame->ra)) {
> + int graph = 0;
> +
> + frame->ra = ftrace_graph_ret_addr(NULL, &graph, frame->ra,
> + NULL);
> + }
> + return 0;
> +}
> +
> +static void notrace walk_stackframe(struct stackframe *fr,
> + struct perf_callchain_entry_ctx *entry)
> +{
> + do {
> + perf_callchain_store(entry, fr->ra);
> + } while (unwind_frame_kernel(fr) >= 0);
> +}
> +
> +/*
> + * Get the return address for a single stackframe and return a pointer to the
> + * next frame tail.
> + */
> +static unsigned long user_backtrace(struct perf_callchain_entry_ctx *entry,
> + unsigned long fp, unsigned long reg_ra)
> +{
> + struct stackframe buftail;
> + unsigned long ra = 0;
> + unsigned long *user_frame_tail =
> + (unsigned long *)(fp - sizeof(struct stackframe));
> +
> + /* Check accessibility of one struct frame_tail beyond */
> + if (!access_ok(user_frame_tail, sizeof(buftail)))
> + return 0;
> + if (__copy_from_user_inatomic(&buftail, user_frame_tail,
> + sizeof(buftail)))
> + return 0;
> +
> + if (reg_ra != 0)
> + ra = reg_ra;
> + else
> + ra = buftail.ra;
> +
> + fp = buftail.fp;
> + perf_callchain_store(entry, ra);
> +
> + return fp;
> +}
> +
> +/*
> + * This will be called when the target is in user mode
> + * This function will only be called when we use
> + * "PERF_SAMPLE_CALLCHAIN" in
> + * kernel/events/core.c:perf_prepare_sample()
> + *
> + * How to trigger perf_callchain_[user/kernel] :
> + * $ perf record -e cpu-clock --call-graph fp ./program
> + * $ perf report --call-graph
> + *
> + * On RISC-V platform, the program being sampled and the C library
> + * need to be compiled with -fno-omit-frame-pointer, otherwise
> + * the user stack will not contain function frame.
> + */
> +void perf_callchain_user(struct perf_callchain_entry_ctx *entry,
> + struct pt_regs *regs)
> +{
> + unsigned long fp = 0;
> +
> + /* RISC-V does not support virtualization. */
> + if (perf_guest_cbs && perf_guest_cbs->is_in_guest())
> + return;
> +
> + fp = regs->s0;
> + perf_callchain_store(entry, regs->sepc);
> +
> + fp = user_backtrace(entry, fp, regs->ra);
> + while (fp && !(fp & 0x3) && entry->nr < entry->max_stack)
> + fp = user_backtrace(entry, fp, 0);
> +}
> +
> +void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry,
> + struct pt_regs *regs)
> +{
> + struct stackframe fr;
> +
> + /* RISC-V does not support virtualization. */
> + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) {
> + pr_warn("RISC-V does not support perf in guest mode!");
> + return;
> + }
> +
> + fr.fp = regs->s0;
> + fr.ra = regs->ra;
> + walk_stackframe(&fr, entry);
> +}
> --
> 2.7.4
>

Not sure if I did something wrong. I encounter a build error when I
try to build tools/perf/tests

CC arch/riscv/util/dwarf-regs.o
arch/riscv/util/dwarf-regs.c:64:5: error: no previous prototype for
‘regs_query_register_offset’ [-Werror=missing-prototypes]

I simply add its prototype and it could be built pass.
This is my testing results.
# ./perf test
1: vmlinux symtab matches kallsyms : Skip
2: Detect openat syscall event : FAILED!
3: Detect openat syscall event on all cpus : FAILED!
4: Read samples using the mmap interface : FAILED!
5: Test data source output : Ok
6: Parse event definition strings : FAILED!
7: Simple expression parser : Ok
8: PERF_RECORD_* events & perf_sample fields : FAILED!
9: Parse perf pmu format : Ok
10: DSO data read : Ok
11: DSO data cache : Ok
12: DSO data reopen : Ok
13: Roundtrip evsel->name : Ok
14: Parse sched tracepoints fields : FAILED!
15: syscalls:sys_enter_openat event fields : FAILED!
16: Setup struct perf_event_attr : FAILED!
17: Match and link multiple hists : Ok
18: 'import perf' in python : FAILED!

19: Breakpoint overflow signal handler : FAILED!
20: Breakpoint overflow sampling : FAILED!
21: Breakpoint accounting : Skip
22: Watchpoint :
22.1: Read Only Watchpoint : FAILED!
22.2: Write Only Watchpoint : FAILED!
22.3: Read / Write Watchpoint : FAILED!
22.4: Modify Watchpoint : FAILED!
23: Number of exit events of a simple workload : Ok
24: Software clock events period values : Ok
25: Object code reading : Ok
26: Sample parsing : Ok
27: Use a dummy software event to keep tracking : Ok
28: Parse with no sample_id_all bit set : Ok
29: Filter hist entries : Ok
30: Lookup mmap thread : Ok
31: Share thread mg : Ok
32: Sort output of hist entries : Ok
33: Cumulate child hist entries : Ok
34: Track with sched_switch : Ok
35: Filter fds with revents mask in a fdarray : Ok
36: Add fd to a fdarray, making it autogrow : Ok
37: kmod_path__parse : Ok
38: Thread map : Ok
39: LLVM search and compile :
39.1: Basic BPF llvm compile : Skip
39.2: kbuild searching : Skip
39.3: Compile source for BPF prologue generation : Skip
39.4: Compile source for BPF relocation : Skip
40: Session topology : FAILED!
41: BPF filter :
41.1: Basic BPF filtering : Skip
41.2: BPF pinning : Skip
41.3: BPF relocation checker : Skip
42: Synthesize thread map : Ok
43: Remove thread map : Ok
44: Synthesize cpu map : Ok
45: Synthesize stat config : Ok
46: Synthesize stat : Ok
47: Synthesize stat round : Ok
48: Synthesize attr update : Ok
49: Event times : Ok
50: Read backward ring buffer : Skip
51: Print cpu map : Ok
52: Probe SDT events : Skip
53: is_printable_array : Ok
54: Print bitmap : Ok
55: perf hooks : Ok
56: builtin clang support : Skip (not
compiled in)
57: unit_number__scnprintf : Ok
58: mem2node : Ok
59: time utils : Ok
60: map_groups__merge_in : Ok
61: probe libc's inet_pton & backtrace it with ping : FAILED!
62: Add vfs_getname probe to get syscall args filenames : FAILED!
63: Check open filename arg using perf trace + vfs_getname: Skip
64: Use vfs_getname probe to get syscall args filenames : FAILED!
65: Zstd perf.data compression/decompression : Skip

I also try the command that Paul pointed out.
./perf record -e cpu-clock --call-graph fp /bin/ls
It works fine now. It can generate a perf.data now.

\
 
 \ /
  Last update: 2019-08-21 11:17    [W:0.066 / U:4.596 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site