Re: [PATCH V1 06/19] perf tools: Fix kernel_start for PTI on x86

[Adrian Hunter]

On 16/05/18 16:00, Arnaldo Carvalho de Melo wrote:
> Em Tue, May 15, 2018 at 03:43:20PM +0300, Adrian Hunter escreveu:
>> On x86_64, PTI entry trampolines are less than the start of kernel text,
>> but still above 2^63. So leave kernel_start = 1ULL << 63 for x86_64.
>>
>> Signed-off-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
>> ---
>>  tools/perf/util/machine.c | 16 +++++++++++++++-
>>  tools/perf/util/machine.h |  2 ++
>>  2 files changed, 17 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
>> index 72a351613d85..2f39a4d014de 100644
>> --- a/tools/perf/util/machine.c
>> +++ b/tools/perf/util/machine.c
>> @@ -2296,6 +2296,15 @@ int machine__set_current_tid(struct machine *machine, int cpu, pid_t pid,
>>  	return 0;
>>  }
>>  
>> +/*
>> + * Compares the raw arch string. N.B. see instead perf_env__arch() if a
>> + * normalized arch is needed.
>> + */
>> +bool machine__is(struct machine *machine, const char *arch)
>> +{
>> +	return machine && machine->env && !strcmp(machine->env->arch, arch);
>> +}
>> +
> 
> Broke 'perf top':
> 
> gdb perf
> bt
> bt
> #0  __strcmp_sse2_unaligned () at ../sysdeps/x86_64/multiarch/strcmp-sse2-unaligned.S:31
> #1  0x0000000000528b1e in machine__is (machine=0xc55548, arch=0x7d9058 "x86_64") at util/machine.c:2305
> #2  0x0000000000528bbb in machine__get_kernel_start (machine=0xc55548) at util/machine.c:2329
> #3  0x00000000004df894 in machine__kernel_start (machine=0xc55548) at util/machine.h:90
> #4  0x00000000004df8e0 in machine__kernel_ip (machine=0xc55548, ip=140352692842292) at util/machine.h:96
> #5  0x00000000004e3a9c in thread__find_map (thread=0x21f9de0, cpumode=2 '\002', addr=140352692842292, al=0x7fffffff8160)
>     at util/event.c:1550
> #6  0x00000000004e3c51 in machine__resolve (machine=0xc55548, al=0x7fffffff8160, sample=0x7fffffff8360) at util/event.c:1591
> #7  0x000000000045a274 in perf_event__process_sample (tool=0x7fffffffa720, event=0x7fffea662878, evsel=0xc120e0, 
>     sample=0x7fffffff8360, machine=0xc55548) at builtin-top.c:734
> #8  0x000000000045a91c in perf_top__mmap_read_idx (top=0x7fffffffa720, idx=3) at builtin-top.c:872
> #9  0x000000000045aa6b in perf_top__mmap_read (top=0x7fffffffa720) at builtin-top.c:898
> #10 0x000000000045b460 in __cmd_top (top=0x7fffffffa720) at builtin-top.c:1167
> #11 0x000000000045c973 in cmd_top (argc=0, argv=0x7fffffffe0e0) at builtin-top.c:1509
> #12 0x00000000004ce403 in run_builtin (p=0xac5670 <commands+336>, argc=1, argv=0x7fffffffe0e0) at perf.c:303
> #13 0x00000000004ce670 in handle_internal_command (argc=1, argv=0x7fffffffe0e0) at perf.c:355
> #14 0x00000000004ce7c2 in run_argv (argcp=0x7fffffffdf3c, argv=0x7fffffffdf30) at perf.c:399
> #15 0x00000000004ceb90 in main (argc=1, argv=0x7fffffffe0e0) at perf.c:543
> (gdb) fr 1
> #1  0x0000000000528b1e in machine__is (machine=0xc55548, arch=0x7d9058 "x86_64") at util/machine.c:2305
> 2305		return machine && machine->env && !strcmp(machine->env->arch, arch);
> (gdb) list
> 2300	 * Compares the raw arch string. N.B. see instead perf_env__arch() if a
> 2301	 * normalized arch is needed.
> 2302	 */
> 2303	bool machine__is(struct machine *machine, const char *arch)
> 2304	{
> 2305		return machine && machine->env && !strcmp(machine->env->arch, arch);
> 2306	}
> 2307	
> 2308	int machine__get_kernel_start(struct machine *machine)
> 2309	{
> (gdb) p machine
> $1 = (struct machine *) 0xc55548
> (gdb) p machine->env
> $2 = (struct perf_env *) 0xc06400 <perf_env>
> (gdb) p machine-env->arch
> No symbol "env" in current context.
> (gdb) p machine->env->arch
> $3 = 0x0
> (gdb)
> 

If there is no perf_data then perf_session__new() uses perf_env but it seems
perf_env.arch is not initialized.  Would it be OK to initialize
perf_env.arch and perf_env.nr_cpus_avail?