perf evsel: Make evsel__env always return a valid env
PATCH 1/2 perf evsel: Make evsel__env always return a valid env
It's possible to have an evsel and evsel->evlist populated without an evsel->evlist->env, when, e.g., cmd_record is in its error path.
Future patches will add support for evsel__open_strerror to be able to customize error messaging based on perf_env__{arch,cpuid}, so let's have evsel__env return &perf_env instead of NULL in that case.
--- tools/perf/util/evsel.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/evsel.c +1 -1
@@ -2857,7 +2857,7 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,struct perf_env *evsel__env (struct evsel *evsel) { -if (evsel && evsel->evlist) +if (evsel && evsel->evlist && evsel->evlist->env)return evsel->evlist->env;return &perf_env; }
Signature · 1 line
2.31.1
PATCH 2/2 perf tools: Improve IBS error handling
This patch improves the error message returned on failed perf_event_open() on AMD when using IBS.
Output of executing 'perf record -e ibs_op// true' BEFORE this patch:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (ibs_op//u). /bin/dmesg | grep -i perf may provide additional information.
Output after:
AMD IBS can't exclude kernel events. Try running at a higher privilege level.
Output of executing 'sudo perf record -e ibs_op// true' BEFORE this patch:
Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (ibs_op//). /bin/dmesg | grep -i perf may provide additional information.
Output after:
Error:
AMD IBS may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity
--- tools/perf/util/evsel.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+)
tools/perf/util/evsel.c +24 -0
@@ -2743,9 +2743,22 @@ static bool find_process(const char *name)return ret ? false : true; } +static bool is_amd (const char *arch,const char *cpuid) +{ +return arch && !strcmp ("x86" , arch) && cpuid &&strstarts (cpuid,"AuthenticAMD" ); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ +return evsel->core.attr.precise_ip || !strncmp (evsel->pmu_name,"ibs" ,3 ); +} +int evsel__open_strerror (struct evsel *evsel,struct target *target,int err,char *msg,size_t size) { +struct perf_env *env =evsel__env (evsel); +const char *arch =perf_env__arch (env); +const char *cpuid =perf_env__cpuid (env);char sbuf[STRERR_BUFSIZE ];int printed =0 , enforced =0 ;
@@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,return scnprintf (msg, size,"wrong clockid (%d)." , clockid);if (perf_missing_features.aux_output)return scnprintf (msg, size,"The 'aux_output' feature is not supported, update the kernel." ); +if (is_amd (arch, cpuid)) { +if (is_amd_ibs (evsel)) { +if (evsel->core.attr.exclude_kernel) +return scnprintf (msg, size, +"AMD IBS can't exclude kernel events. Try running at a higher privilege level." ); +if (!evsel->core.system_wide) +return scnprintf (msg, size, +"AMD IBS may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity" ); + } + } +break ;case ENODATA : return scnprintf(msg, size, "Cannot collect data source with the load latency event alone. "
Signature · 1 line
2.31.1
Re: [PATCH 2/2] perf tools: Improve IBS error handling
On Mon, Oct 04, 2021 at 04:41:14PM -0500, Kim Phillips wrote:
SNIP
Quoted 32 lines
--- tools/perf/util/evsel.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+)tools/perf/util/evsel.c +13
-0@@ -2743,9 +2743,22 @@ static bool find_process(const char *name)return ret ? false : true; } +static bool is_amd (const char *arch,const char *cpuid) +{ +return arch && !strcmp ("x86" , arch) && cpuid &&strstarts (cpuid,"AuthenticAMD" ); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ +return evsel->core.attr.precise_ip || !strncmp (evsel->pmu_name,"ibs" ,3 ); +} +int evsel__open_strerror (struct evsel *evsel,struct target *target,int err,char *msg,size_t size) { +struct perf_env *env =evsel__env (evsel); +const char *arch =perf_env__arch (env); +const char *cpuid =perf_env__cpuid (env);char sbuf[STRERR_BUFSIZE ];int printed =0 , enforced =0 ;
tools/perf/util/evsel.c int evsel__open_strerror(struct evsel *evsel, struct target *target, · line 2858
Quoted 5 lines
tools/perf/util/evsel.c +2
-0@@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,return scnprintf (msg, size,"wrong clockid (%d)." , clockid);if (perf_missing_features.aux_output)return scnprintf (msg, size,"The 'aux_output' feature is not supported, update the kernel." ); +if (is_amd (arch, cpuid)) { +if (is_amd_ibs (evsel)) {
would single 'is_amd_ibs' call be better? checking on both amd and ibs
jirka
Quoted 15 lines
tools/perf/util/evsel.c +9
-0+if (evsel->core.attr.exclude_kernel) +return scnprintf (msg, size, +"AMD IBS can't exclude kernel events. Try running at a higher privilege level." ); +if (!evsel->core.system_wide) +return scnprintf (msg, size, +"AMD IBS may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity" ); + } + } +break ;case ENODATA : return scnprintf(msg, size, "Cannot collect data source with the load latency event alone. "Signature · 1 line
2.31.1
Re: [PATCH 2/2] perf tools: Improve IBS error handling
Quoted 32 lines · Jiri Olsa
Quoted 31 lines · Kim Phillips
--- tools/perf/util/evsel.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+)tools/perf/util/evsel.c +13
-0@@ -2743,9 +2743,22 @@ static bool find_process(const char *name)return ret ? false : true; } +static bool is_amd (const char *arch,const char *cpuid) +{ +return arch && !strcmp ("x86" , arch) && cpuid &&strstarts (cpuid,"AuthenticAMD" ); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ +return evsel->core.attr.precise_ip || !strncmp (evsel->pmu_name,"ibs" ,3 ); +} +int evsel__open_strerror (struct evsel *evsel,struct target *target,int err,char *msg,size_t size) { +struct perf_env *env =evsel__env (evsel); +const char *arch =perf_env__arch (env); +const char *cpuid =perf_env__cpuid (env);char sbuf[STRERR_BUFSIZE ];int printed =0 , enforced =0 ;
Quoted 8 lines
tools/perf/util/evsel.c int evsel__open_strerror(struct evsel *evsel, struct target *target, · line 2858
Quoted 6 lines
tools/perf/util/evsel.c +2
-0@@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,return scnprintf (msg, size,"wrong clockid (%d)." , clockid);if (perf_missing_features.aux_output)return scnprintf (msg, size,"The 'aux_output' feature is not supported, update the kernel." ); +if (is_amd (arch, cpuid)) { +if (is_amd_ibs (evsel)) {would single '
is_amd_ibs' call be better? checking on both amd and ibs
Good suggestion. If you look at the later patch in the BRS series, I have rewritten it to add the new AMD PMU like so:
if (is_amd()) {
if (is_amd_ibs()) {
if (evsel->this)
return
if (evsel->that)
return
}
+ if (is_amd_brs()) {
+ if (evsel->this)
+ return
+ if (evsel->that)
+ return
+ }
}
Below is the full proposed replacement patch for patch 12 of 13 of the BRS series.
Another option is to have the is_amd_{ibs,brs} functions call is_amd() themselves, so the if (evsel->) code could be unindented by one tab, would that be better?
Thanks,
Kim
From a4cbab762719b30bddec2e278cf8b8eb82e83865 Mon Sep 17 00:00:00 2001
From: Stephane Eranian <[email protected]>
Date: Thu, 9 Sep 2021 00:56:59 -0700
Subject: [PATCH] perf tools: Improve error handling of AMD Branch Sampling
This patch improves the error message printed by perf when perf_event_open() fails on AMD Zen3 when using the branch sampling feature. In the case of EINVAL, there are two main reasons: frequency mode or period is smaller than the depth of the branch sampling buffer (16). The patch checks the parameters of the call and tries to print a relevant message to explain the error:
$ perf record -b -e cpu/branch-brs/ -c 10 ls Error: AMD Branch Sampling does not support sampling period smaller than what is reported in /sys/devices/cpu/caps/branches.
$ perf record -b -e cpu/branch-brs/ ls Error: AMD Branch Sampling does not support frequency mode sampling, must pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.
--- tools/perf/util/evsel.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
tools/perf/util/evsel.c +14 -0
@@ -2753,6 +2753,12 @@ static bool is_amd_ibs(struct evsel *evsel)return evsel->core.attr.precise_ip || !strncmp (evsel->pmu_name,"ibs" ,3 ); } +static bool is_amd_brs (struct evsel *evsel) +{ +return ((evsel->core.attr.config &0xff ) ==0xc4 ) && + (evsel->core.attr.sample_type &PERF_SAMPLE_BRANCH_STACK ); +} +int evsel__open_strerror (struct evsel *evsel,struct target *target,int err,char *msg,size_t size) {
@@ -2863,6 +2869,14 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,return scnprintf (msg, size,"AMD IBS may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity" ); } +if (is_amd_brs (evsel)) { +if (evsel->core.attr.freq) +return scnprintf (msg, size, +"AMD Branch Sampling does not support frequency mode sampling, must pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/." ); +/* another reason is that the period is too small */ +return scnprintf (msg, size, +"AMD Branch Sampling does not support sampling period smaller than what is reported in /sys/devices/cpu/caps/branches." ); + } }break ;
Signature · 1 line
2.31.1
Re: [PATCH 2/2] perf tools: Improve IBS error handling
Quoted 59 lines · Kim Phillips
Quoted 30 lines · Jiri Olsa
Quoted 29 lines · Kim Phillips
---
tools/perf/util/evsel.c| 24 ++++++++++++++++++++++++1 file changed, 24 insertions(+)tools/perf/util/evsel.c +13
-0@@ -2743,9 +2743,22 @@ static bool find_process(const char *name) return ret ? false : true; } +static bool is_amd (const char *arch,const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid, "AuthenticAMD"); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ + return evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name, "ibs", 3); +} + int evsel__open_strerror(struct evsel *evsel, struct target *target, int err, char *msg, size_t size) { + struct perf_env *env = evsel__env(evsel); + const char *arch = perf_env__arch(env); + const char *cpuid = perf_env__cpuid(env); char sbuf[STRERR_BUFSIZE]; int printed = 0, enforced = 0;Quoted 8 lines
tools/perf/util/evsel.c int evsel__open_strerror(struct evsel *evsel, struct target *target, · line 2858
Quoted 6 lines
tools/perf/util/evsel.c +2
-0@@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target, return scnprintf(msg, size, "wrong clockid (%d).", clockid); if (perf_missing_features.aux_output) return scnprintf(msg, size, "The 'aux_output' feature is not supported, update the kernel."); + if (is_amd(arch, cpuid)) { + if (is_amd_ibs(evsel)) {would single '
is_amd_ibs' call be better? checking on both amd and ibsGood suggestion. If you look at the later patch in the BRS series, I have rewritten it to add the new AMD PMU like so:
if (
is_amd()) {
if (is_amd_ibs()) {
if (evsel->this)
return
if (evsel->that)
return
}
+ if (is_amd_brs()) {
+ if (evsel->this)
+ return
+ if (evsel->that)
+ return
+ }
}
Quoted 6 lines
Below is the full proposed replacement patch for patch 12 of 13 of the BRS series.
Another option is to have the is_amd_{ibs,brs} functions call
is_amd()themselves, so the if (evsel->) code could be unindented by one tab, would that be better?
No reply, so guessing not.
Arnaldo, can these two patches be applied please?
Thanks,
Kim
Re: [PATCH 2/2] perf tools: Improve IBS error handling
In reply to Kim PhillipsQuoted 49 lines · Kim Phillips
Quoted 35 lines · Jiri Olsa
Quoted 34 lines · Kim Phillips
---
tools/perf/util/evsel.c| 24 ++++++++++++++++++++++++1 file changed, 24 insertions(+)tools/perf/util/evsel.c +12
-0@@ -2743,9 +2743,22 @@ static bool find_process(const char *name) return ret ? false : true; } +static bool is_amd(const char *arch, const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid,"AuthenticAMD" ); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ + return evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name,"ibs" ,3 ); +} + int evsel__open_strerror(struct evsel *evsel, struct target *target, int err, char *msg, size_t size) { + struct perf_env *env = evsel__env(evsel); + const char *arch = perf_env__arch(env); + const char *cpuid = perf_env__cpuid(env); char sbuf[STRERR_BUFSIZE];int printed = 0, enforced = 0;
@@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
return scnprintf(msg, size, "wrong clockid (%d).",Quoted 8 lines
Quoted 6 lines
clockid);
if (perf_missing_features.aux_output)
returnscnprintf(msg, size, "The 'aux_output' feature is not supported, update the kernel.");
+ if (is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {would single '
is_amd_ibs' call be better? checking on both amd and ibsGood suggestion. If you look at the later patch in the BRS series, I have rewritten it to add the new AMD PMU like so:
Quoted 14 lines
if (
is_amd()) {
if (is_amd_ibs()) {
if (evsel->this)
return
if (evsel->that)
return
}
+ if (is_amd_brs()) {
+ if (evsel->this)
+ return
+ if (evsel->that)
+ return
+ }
}
Hi Kim,
From my point of view, it won't be a good idea of adding so many
checks in common function definition itself.
Can you just create a check to see if its amd machine and then add a function call which will handle all four conditions together?
which is basically for:
+ if (is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {
+ if (evsel->core.attr.exclude_kernel)
+ return scnprintf(msg, size,
+ "AMD IBS can't exclude kernel events. Try running at a higher
privilege level.");
+ if (!evsel->core.system_wide)
+ return scnprintf(msg, size,
+ "AMD IBS may only be available in system-wide/per-cpu mode. Try using
-a, or -C and workload affinity");
+ }
and this:
+ if (is_amd_brs(evsel)) {
+ if (evsel->core.attr.freq)
+ return scnprintf(msg, size,
+ "AMD Branch Sampling does not support frequency mode sampling, must
pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.");
+ /* another reason is that the period is too small */
+ return scnprintf(msg, size,
+ "AMD Branch Sampling does not support sampling period smaller than
what is reported in /sys/devices/cpu/caps/branches.");
+ }
So, incase we are in amd machine, common function evsel__open_strerror will call function may be something like amd_evesel_open_strerror_check which will look for both ibs and brs conditions and return corresponding error statement.
Thanks,
Kajol Jain
Quoted 80 lines
Below is the full proposed replacement patch for patch 12 of 13 of the BRS series.
Another option is to have the is_amd_{ibs,brs} functions call
is_amd()themselves, so the if (evsel->) code could be unindented by one tab, would that be better?Thanks,
Kim
From
a4cbab762719b30bddec2e278cf8b8eb82e83865Mon Sep 17 00:00:00 2001
From: Stephane Eranian <[email protected]>
Date: Thu, 9 Sep 2021 00:56:59 -0700
Subject: [PATCH] perf tools: Improve error handling of AMD Branch SamplingThis patch improves the error message printed by perf when
perf_event_open()fails on AMD Zen3 when using the branch sampling feature. In the case of EINVAL, there are two main reasons: frequency mode or period is smaller than the depth of the branch sampling buffer (16). The patch checks the parameters of the call and tries to print a relevant message to explain the error:$ perf record -b -e cpu/branch-brs/ -c 10 ls Error: AMD Branch Sampling does not support sampling period smaller than what is reported in /sys/devices/cpu/caps/branches.$ perf record -b -e cpu/branch-brs/ ls Error: AMD Branch Sampling does not support frequency mode sampling, must pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.Signed-off-by Stephane Eranian <[email protected]>[Rebased on commit
9fe8895a27a84("perf env: Addperf_env__cpuid, perf_env__{nr_}pmu_mappings")]Signed-off-by Kim Phillips <[email protected]>---
tools/perf/util/evsel.c | 14 ++++++++++++++1 file changed, 14 insertions(+)tools/perf/util/evsel.c +6
-0@@ -2753,6 +2753,12 @@ static bool is_amd_ibs(struct evsel *evsel) return evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name,"ibs" ,3 ); } +static bool is_amd_brs (struct evsel *evsel) +{ + return ((evsel->core.attr.config & 0xff) == 0xc4) && + (evsel->core.attr.sample_type & PERF_SAMPLE_BRANCH_STACK); +} + int evsel__open_strerror(struct evsel *evsel, struct target *target, int err, char *msg, size_t size){
tools/perf/util/evsel.c +0
-0@@ -2863,6 +2869,14 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target, return scnprintf(msg, size,"AMD IBS may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity");
}
+ if (is_amd_brs(evsel)) {
+ if (evsel->core.attr.freq)
+ return scnprintf(msg, size,
+ "AMD Branch Sampling does not support frequency mode sampling, must pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.");
+ /* another reason is that the period is too small */ + return scnprintf(msg, size,
+ "AMD Branch Sampling does not support sampling period smaller than what is reported in/sys/devices/cpu/caps/branches.");
+ }
}break;
Re: [PATCH 2/2] perf tools: Improve IBS error handling
Quoted 96 lines · kajoljain
Quoted 49 lines · Kim Phillips
Quoted 35 lines · Jiri Olsa
Quoted 34 lines · Kim Phillips
--- tools/perf/util/evsel.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+)tools/perf/util/evsel.c +12
-0@@ -2743,9 +2743,22 @@ static bool find_process(const char *name) return ret ? false : true; } +static bool is_amd(const char *arch, const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid,"AuthenticAMD" ); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ + return evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name,"ibs" ,3 ); +} + int evsel__open_strerror(struct evsel *evsel, struct target *target, int err, char *msg, size_t size) { + struct perf_env *env = evsel__env(evsel); + const char *arch = perf_env__arch(env); + const char *cpuid = perf_env__cpuid(env); char sbuf[STRERR_BUFSIZE]; int printed = 0, enforced = 0; @@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel*evsel, struct target *target,
return scnprintf(msg, size, "wrong clockid (%d).",Quoted 8 lines
Quoted 6 lines
clockid);
if (perf_missing_features.aux_output)
return scnprintf(msg, size, "The 'aux_output' feature
is not supported, update the kernel.");
+ if (is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {would single '
is_amd_ibs' call be better? checking on both amd and ibsGood suggestion. If you look at the later patch in the BRS series, I have rewritten it to add the new AMD PMU like so:
Quoted 14 lines
if (
is_amd()) {
if (is_amd_ibs()) {
if (evsel->this)
return
if (evsel->that)
return
}
+ if (is_amd_brs()) {
+ if (evsel->this)
+ return
+ if (evsel->that)
+ return
+ }
}Hi Kim,
From my point of view, it won't be a good idea of adding so many
checks in common function definition itself.
Can you just create a check to see if its amd machine and then add a function call which will handle all four conditions together?which is basically for:
(quoted diff) +5
-0+ if (is_amd(arch, cpuid)) { + if (is_amd_ibs(evsel)) { + if (evsel->core.attr.exclude_kernel) + return scnprintf(msg, size, + "AMD IBS can't exclude kernel events. Try running at a higherprivilege level.");
(quoted diff) +4
-1+ if (!evsel->core.system_wide) + return scnprintf(msg, size, + "AMD IBS may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity"); + }and this:
tools/perf/util/evsel.c +4
-0+if (is_amd_brs (evsel)) { +if (evsel->core.attr.freq) + return scnprintf(msg, size, + "AMD Branch Sampling does not support frequency mode sampling, mustpass a fixed sampling period via -c option or
cpu/branch-brs,period=xxxx/.");tools/perf/util/evsel.c +3
-0+/* another reason is that the period is too small */ + return scnprintf(msg, size, + "AMD Branch Sampling does not support sampling period smaller than
Quoted 2 lines
what is reported in
/sys/devices/cpu/caps/branches.");
+ }
IIRC, I tried something like that but carrying the
struct target *target, int err, char *msg, size_t size
parameters made things worse.
Quoted 4 lines
So, incase we are in amd machine, common function
evsel__open_strerrorwill call function may be something likeamd_evesel_open_strerror_checkwhich will look for both ibs and brs conditions and return corresponding error statement.
The vast majority of decisions made by evsel__open_strerror are going to be common across most arch/uarches. AMD has only these two pesky exceptions to the rule and therefore IMO it's ok to have them inline with the common function, since the decisions are so deeply intertwined. A new amd_evsel_open_strerror_check sounds like it'd duplicate too much of the common function code in order to handle the common error cases.
Kim
Re: [PATCH 2/2] perf tools: Improve IBS error handling
Quoted 114 lines · Kim Phillips
Quoted 89 lines · kajoljain
Quoted 50 lines · Kim Phillips
Quoted 36 lines · Jiri Olsa
Quoted 35 lines · Kim Phillips
---
tools/perf/util/evsel.c| 24 ++++++++++++++++++++++++1 file changed, 24 insertions(+)tools/perf/util/evsel.c +9
-0@@ -2743,9 +2743,22 @@ static bool find_process(const char *name) return ret ? false : true; } +static bool is_amd(const char *arch, const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid,"AuthenticAMD" ); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ + return evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name,"ibs" ,3 ); +} +int evsel__open_strerror(struct evsel *evsel, struct target *target,
int err, char *msg,size_tsize)
{
+ structperf_env*env =evsel__env(evsel);
+ const char *arch =perf_env__arch(env);
+ const char *cpuid =perf_env__cpuid(env);
char sbuf[STRERR_BUFSIZE];
int printed = 0, enforced = 0;
@@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
return scnprintf(msg, size, "wrong clockid (%d).",Quoted 8 lines
Quoted 6 lines
clockid);
if (perf_missing_features.aux_output)
returnscnprintf(msg, size, "The 'aux_output' feature is not supported, update the kernel.");
+ if (is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {would single '
is_amd_ibs' call be better? checking on both amd and ibsGood suggestion. If you look at the later patch in the BRS series, I have rewritten it to add the new AMD PMU like so:
Quoted 14 lines
if (
is_amd()) {
if (is_amd_ibs()) {
if (evsel->this)
return
if (evsel->that)
return
}
+ if (is_amd_brs()) {
+ if (evsel->this)
+ return
+ if (evsel->that)
+ return
+ }
}Hi Kim,
From my point of view, it won't be a good idea of adding so many checks in common function definition itself.
Can you just create a check to see if its amd machine and then add a function call which will handle all four conditions together?which is basically for:
+ if (
is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {
+ if (evsel->core.attr.exclude_kernel)
+ return scnprintf(msg, size,
+ "AMD IBS can't exclude kernel events. Try running at a higher privilege level.");
+ if (!evsel->core.system_wide)
+ return scnprintf(msg, size,
+ "AMD IBS may only be available in system-wide/per-cpu mode. Try using
-a, or -C and workload affinity");
+ }and this:
Quoted 11 lines
+ if (
is_amd_brs(evsel)) {
+ if (evsel->core.attr.freq)
+ return scnprintf(msg, size,
+ "AMD Branch Sampling does not support frequency mode sampling, must pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.");
+ /* another reason is that the period is too small */ + return scnprintf(msg, size,
+ "AMD Branch Sampling does not support sampling period smaller than what is reported in/sys/devices/cpu/caps/branches.");
+ }IIRC, I tried something like that but carrying the
struct target *target, int err, char *msg,
size_tsizeparameters made things worse.
Quoted 4 lines
So, incase we are in amd machine, common function
evsel__open_strerrorwill call function may be something likeamd_evesel_open_strerror_checkwhich will look for both ibs and brs conditions and return corresponding error statement.
Quoted 7 lines
The vast majority of decisions made by
evsel__open_strerrorare going to be common across most arch/uarches. AMD has only these two pesky exceptions to the rule and therefore IMO it's ok to have them inline with the common function, since the decisions are so deeply intertwined. A newamd_evsel_open_strerror_checksounds like it'd duplicate too much of the common function code in order to handle the common error cases.
Hi Kim,
Sorry for the confusion, what I meant by adding new function is just
to handle these corner error cases and not duplicating whole evsel__open_strerror code.
Maybe something like below code, Its just prototype of code to show you the flow, you can refine it and check for any build or indentation issues using checkpatch.pl script.
So basically, in common function we can just have 2 calls, first to check if we are in amd system and second to return corresponding error message, rather then adding whole chunk of if's which are specific to amd.
tools/perf/util/evsel.c +3 -0
@@ -2852,9 +2852,40 @@ static bool find_process(const char *name)return ret ? false : true; } +static bool is_amd (const char *arch,const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid,
"AuthenticAMD");
+}
+
+static int error_amd_ibs_brs(struct evsel *evsel, char *msg, size_t size) +{
+ if (evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name,
"ibs", 3)) {
+ if (evsel->core.attr.exclude_kernel)
+ return scnprintf(msg, size,
+ "AMD IBS can't exclude kernel events. Try running at a higher
privilege level.");
+ if (!evsel->core.system_wide)
+ return scnprintf(msg, size,
+ "AMD IBS may only be available in system-wide/per-cpu mode. Try
using -a, or -C and workload affinity");
+ }
+
+ if (((evsel->core.attr.config & 0xff) == 0xc4) &&
(evsel->core.attr.sample_type & PERF_SAMPLE_BRANCH_STACK)) {
+ if (evsel->core.attr.freq) {
+ return scnprintf(msg, size,
+ "AMD Branch Sampling does not support frequency mode sampling,
must pass a fixed sampling
+ period via -c option or cpu/branch-brs,period=xxxx/.");
+ /* another reason is that the period is too small */
+ return scnprintf(msg, size,
+ "AMD Branch Sampling does not support sampling period smaller
than what is reported in /sys/devices/cpu/caps/branches.");
+ }
+ }
+}
+
int evsel__open_strerror(struct evsel *evsel, struct target *target,
int err, char *msg, size_t size)
{
+ struct perf_env *env = evsel__env(evsel);
+ const char *arch = perf_env__arch(env);
+ const char *cpuid = perf_env__cpuid(env);
char sbuf[STRERR_BUFSIZE];
int printed = 0, enforced = 0;@@ -2950,6 +2981,8 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
return scnprintf(msg, size, "wrong clockid
(%d).", clockid);
if (perf_missing_features.aux_output)
return scnprintf(msg, size, "The 'aux_output'
feature is not supported, update the kernel.");
+ if (is_amd(arch, cpuid))
+ return error_amd_ibs_brs(evsel, msg, size);
break;
case ENODATA:
return scnprintf(msg, size, "Cannot collect data source
with the load latency event alone. "
Thanks,
Kajol Jain
Quoted 2 lines
Kim
Re: [PATCH 2/2] perf tools: Improve IBS error handling
Quoted 197 lines · kajoljain
Quoted 114 lines · Kim Phillips
Quoted 89 lines · kajoljain
Quoted 50 lines · Kim Phillips
Quoted 36 lines · Jiri Olsa
Quoted 35 lines · Kim Phillips
--- tools/perf/util/evsel.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+)tools/perf/util/evsel.c +12
-0@@ -2743,9 +2743,22 @@ static bool find_process(const char *name) return ret ? false : true; } +static bool is_amd(const char *arch, const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid,"AuthenticAMD" ); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ + return evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name,"ibs" ,3 ); +} + int evsel__open_strerror(struct evsel *evsel, struct target *target, int err, char *msg, size_t size) { + struct perf_env *env = evsel__env(evsel); + const char *arch = perf_env__arch(env); + const char *cpuid = perf_env__cpuid(env); char sbuf[STRERR_BUFSIZE]; int printed = 0, enforced = 0; @@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel*evsel, struct target *target,
return scnprintf(msg, size, "wrong clockid (%d).",Quoted 8 lines
Quoted 6 lines
clockid);
if (perf_missing_features.aux_output)
return scnprintf(msg, size, "The 'aux_output' feature
is not supported, update the kernel.");
+ if (is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {would single '
is_amd_ibs' call be better? checking on both amd and ibsGood suggestion. If you look at the later patch in the BRS series, I have rewritten it to add the new AMD PMU like so:
Quoted 14 lines
if (
is_amd()) {
if (is_amd_ibs()) {
if (evsel->this)
return
if (evsel->that)
return
}
+ if (is_amd_brs()) {
+ if (evsel->this)
+ return
+ if (evsel->that)
+ return
+ }
}Hi Kim,
From my point of view, it won't be a good idea of adding so many
checks in common function definition itself.
Can you just create a check to see if its amd machine and then add a function call which will handle all four conditions together?which is basically for:
+ if (
is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {
+ if (evsel->core.attr.exclude_kernel)
+ return scnprintf(msg, size,
+ "AMD IBS can't exclude kernel events. Try running at a higher privilege level.");
+ if (!evsel->core.system_wide)
+ return scnprintf(msg, size,
+ "AMD IBS may only be available in system-wide/per-cpu mode. Try using
-a, or -C and workload affinity");
+ }and this:
Quoted 11 lines
+ if (
is_amd_brs(evsel)) {
+ if (evsel->core.attr.freq)
+ return scnprintf(msg, size,
+ "AMD Branch Sampling does not support frequency mode sampling, must pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.");
+ /* another reason is that the period is too small */ + return scnprintf(msg, size,
+ "AMD Branch Sampling does not support sampling period smaller than what is reported in/sys/devices/cpu/caps/branches.");
+ }IIRC, I tried something like that but carrying the
struct target *target, int err, char *msg,
size_tsizeparameters made things worse.
Quoted 4 lines
So, incase we are in amd machine, common function
evsel__open_strerrorwill call function may be something likeamd_evesel_open_strerror_checkwhich will look for both ibs and brs conditions and return corresponding error statement.Quoted 7 lines
The vast majority of decisions made by
evsel__open_strerrorare going to be common across most arch/uarches. AMD has only these two pesky exceptions to the rule and therefore IMO it's ok to have them inline with the common function, since the decisions are so deeply intertwined. A newamd_evsel_open_strerror_checksounds like it'd duplicate too much of the common function code in order to handle the common error cases.Hi Kim,
Sorry for the confusion, what I meant by adding new function is just
to handle these corner error cases and not duplicating wholeevsel__open_strerrorcode.Maybe something like below code, Its just prototype of code to show you the flow, you can refine it and check for any build or indentation issues using checkpatch.pl script.
So basically, in common function we can just have 2 calls, first to check if we are in amd system and second to return corresponding error message, rather then adding whole chunk of if's which are specific to amd.
tools/perf/util/evsel.c +20
-0@@ -2852,9 +2852,40 @@ static bool find_process(const char *name)return ret ? false : true; } +static bool is_amd (const char *arch,const char *cpuid) +{ +return arch && !strcmp ("x86" , arch) && cpuid &&strstarts (cpuid,"AuthenticAMD" ); +} + +static int error_amd_ibs_brs(struct evsel *evsel, char *msg, size_t size) +{ + if (evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name, "ibs", 3)) { + if (evsel->core.attr.exclude_kernel) + return scnprintf(msg, size, + "AMD IBS can't exclude kernel events. Try running at a higher privilege level."); + if (!evsel->core.system_wide) + return scnprintf(msg, size, + "AMD IBS may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity"); + } + + if (((evsel->core.attr.config & 0xff) == 0xc4) && (evsel->core.attr.sample_type & PERF_SAMPLE_BRANCH_STACK)) { + if (evsel->core.attr.freq) { + return scnprintf(msg, size, + "AMD Branch Sampling does not support frequency mode sampling,must pass a fixed sampling
tools/perf/util/evsel.c +4
-0+ period via -c option or cpu/branch-brs,period=xxxx/."); + /* another reason is that the period is too small */ + return scnprintf(msg, size, + "AMD Branch Sampling does not support sampling period smallerthan what is reported in
/sys/devices/cpu/caps/branches.");tools/perf/util/evsel.c +9
-0+ } + } +} +int evsel__open_strerror (struct evsel *evsel,struct target *target,int err,char *msg,size_t size) { +struct perf_env *env =evsel__env (evsel); +const char *arch =perf_env__arch (env); +const char *cpuid =perf_env__cpuid (env);char sbuf[STRERR_BUFSIZE ];int printed =0 , enforced =0 ;@@ -2950,6 +2981,8 @@ int evsel__open_strerror(struct evsel *evsel,struct target *target, return scnprintf(msg, size, "wrong clockid (%d).", clockid);if (perf_missing_features.aux_output) return scnprintf(msg, size, "The 'aux_output' feature is not supported, update the kernel."); +if (is_amd (arch, cpuid)) +return error_amd_ibs_brs (evsel, msg, size);
Quoted 4 lines
break;
case ENODATA:
return scnprintf(msg, size, "Cannot collect data source
with the load latency event alone. "
That change will makes AMD machines fail to fall back to the default "The sys_perf_event_open() syscall returned with..." error string in case it's not those AMD IBS and BRS sub-conditions.
Is having the AMD error code checking in the main evsel__open_strerror() so bad? Other arches and their PMU implementations may find error conditions that they have in common with AMD's, therefore opening up the code for opposite types of refactoring and reuse than what is being requested here. E.g., I've seen other hardware configurations - not specific to one architecture - that could also use this message:
{"AMD IBS"->"%s",pmu_name} may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity");
Thanks,
Kim
Re: [PATCH 2/2] perf tools: Improve IBS error handling
Quoted 205 lines · Kim Phillips
Quoted 204 lines · kajoljain
Quoted 119 lines · Kim Phillips
Quoted 92 lines · kajoljain
Quoted 53 lines · Kim Phillips
Quoted 39 lines · Jiri Olsa
Quoted 38 lines · Kim Phillips
---
tools/perf/util/evsel.c| 24 ++++++++++++++++++++++++1 file changed, 24 insertions(+)tools/perf/util/evsel.c +9
-0@@ -2743,9 +2743,22 @@ static bool find_process(const char *name) return ret ? false : true; } +static bool is_amd(const char *arch, const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid &&strstarts (cpuid,"AuthenticAMD" ); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ + return evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name,"ibs" ,3 ); +} +int evsel__open_strerror(struct evsel *evsel, struct target *target,
int err, char *msg,size_tsize)
{
+ structperf_env*env =evsel__env(evsel);
+ const char *arch =perf_env__arch(env);
+ const char *cpuid =perf_env__cpuid(env);
char sbuf[STRERR_BUFSIZE];
int printed = 0, enforced = 0;
@@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
returnscnprintf(msg, size, "wrong clockid (%d).", clockid);
if (perf_missing_features.aux_output)Quoted 8 lines
Quoted 5 lines
return scnprintf(msg, size, "The '
aux_output' feature
is not supported, update the kernel.");
+ if (is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {would single '
is_amd_ibs' call be better? checking on both amd and ibsGood suggestion. If you look at the later patch in the BRS series, I have rewritten it to add the new AMD PMU like so:
Quoted 14 lines
if (
is_amd()) {
if (is_amd_ibs()) {
if (evsel->this)
return
if (evsel->that)
return
}
+ if (is_amd_brs()) {
+ if (evsel->this)
+ return
+ if (evsel->that)
+ return
+ }
}Hi Kim,
From my point of view, it won't be a good idea of adding so many checks in common function definition itself.
Can you just create a check to see if its amd machine and then add a function call which will handle all four conditions together?which is basically for:
+ if (
is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {
+ if (evsel->core.attr.exclude_kernel)
+ return scnprintf(msg, size,
+ "AMD IBS can't exclude kernel events. Try running at a higher privilege level.");
+ if (!evsel->core.system_wide)
+ return scnprintf(msg, size,
+ "AMD IBS may only be available in system-wide/per-cpu mode. Try using
-a, or -C and workload affinity");
+ }and this:
Quoted 12 lines
+ if (
is_amd_brs(evsel)) {
+ if (evsel->core.attr.freq)
+ return scnprintf(msg, size,
+ "AMD Branch Sampling does not support frequency mode sampling, must
pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.");
+ /* another reason is that the period is too small */ + return scnprintf(msg, size,
+ "AMD Branch Sampling does not support sampling period smaller than what is reported in/sys/devices/cpu/caps/branches.");
+ }IIRC, I tried something like that but carrying the
struct target *target, int err, char *msg,
size_tsizeparameters made things worse.
Quoted 5 lines
So, incase we are in amd machine, common function
evsel__open_strerrorwill call function may be something likeamd_evesel_open_strerror_checkwhich will look for both ibs and brs conditions and return corresponding
error statement.Quoted 7 lines
The vast majority of decisions made by
evsel__open_strerrorare going to be common across most arch/uarches. AMD has only these two pesky exceptions to the rule and therefore IMO it's ok to have them inline with the common function, since the decisions are so deeply intertwined. A newamd_evsel_open_strerror_checksounds like it'd duplicate too much of the common function code in order to handle the common error cases.Hi Kim,
Sorry for the confusion, what I meant by adding new function is just to handle these corner error cases and not duplicating wholeevsel__open_strerrorcode.Maybe something like below code, Its just prototype of code to show you the flow, you can refine it and check for any build or indentation issues using checkpatch.pl script.
So basically, in common function we can just have 2 calls, first to check if we are in amd system and second to return corresponding error message, rather then adding whole chunk of if's which are specific to amd.
tools/perf/util/evsel.c +20
-0@@ -2852,9 +2852,40 @@ static bool find_process(const char *name) return ret ? false : true; } +static bool is_amd(const char *arch, const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid, "AuthenticAMD"); +} + +static int error_amd_ibs_brs(struct evsel *evsel, char *msg, size_t size) +{ + if (evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name, "ibs", 3)) { + if (evsel->core.attr.exclude_kernel) + return scnprintf(msg, size, + "AMD IBS can't exclude kernel events. Try running at a higher privilege level."); + if (!evsel->core.system_wide) + return scnprintf(msg, size, + "AMD IBS may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity"); + } + + if (((evsel->core.attr.config & 0xff) == 0xc4) && (evsel->core.attr.sample_type & PERF_SAMPLE_BRANCH_STACK)) { + if (evsel->core.attr.freq) { + return scnprintf(msg, size, + "AMD Branch Sampling does not support frequency mode sampling,must pass a fixed sampling
+ period via -c option or cpu/branch-brs,period=xxxx/."); + /* another reason is that the period is too small */ + return scnprintf(msg, size,
+ "AMD Branch Sampling does not support sampling period smaller than what is reported in/sys/devices/cpu/caps/branches."); + }
+ }tools/perf/util/evsel.c +2
-0+} +int
evsel__open_strerror(struct evsel *evsel, struct target *target, int err, char *msg, size_t size){
+ structperf_env*env =evsel__env(evsel);
+ const char *arch =perf_env__arch(env);
+ const char *cpuid =perf_env__cpuid(env);
char sbuf[STRERR_BUFSIZE];
int printed = 0, enforced = 0;tools/perf/util/evsel.c +0
-0@@ -2950,6 +2981,8 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,return
scnprintf(msg, size, "wrong clockid (%d).", clockid);
if (perf_missing_features.aux_output) returnscnprintf(msg, size, "The 'aux_output' feature is not supported, update the kernel.");
+ if (is_amd(arch, cpuid))
+ returnerror_amd_ibs_brs(evsel, msg, size);
Quoted 8 lines
Quoted 4 lines
break;
case ENODATA:
return scnprintf(msg, size, "Cannot collect data source with the load latency event alone. "That change will makes AMD machines fail to fall back to the default "The
sys_perf_event_open()syscall returned with..." error string in case it's not those AMD IBS and BRS sub-conditions.
Yes right, as I mentioned before, the code I pointed was just a prototype to show you the flow, these corner cases can be handled on top of it.
Quoted 9 lines
Is having the AMD error code checking in the main
evsel__open_strerror()so bad? Other arches and their PMU implementations may find error conditions that they have in common with AMD's, therefore opening up the code for opposite types of refactoring and reuse than what is being requested here. E.g., I've seen other hardware configurations - not specific to one architecture - that could also use this message:
From my understanding, adding too many checks in common function for a specific arch is not a good practice. Since you already adding multiple functions to get information like ,if current system is amd/ibs/brs. Can't we rather just add a single function and handled all these checks there?
That's just my thoughts, if maintainers are ok with it, then its fine for me too.
Thanks,
Kajol Jain
Quoted 6 lines
{"AMD IBS"->"%s",pmu_name} may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity");
Thanks,
Kim
Re: [PATCH 2/2] perf tools: Improve IBS error handling
Quoted 215 lines · kajoljain
Quoted 205 lines · Kim Phillips
Quoted 204 lines · kajoljain
Quoted 119 lines · Kim Phillips
Quoted 92 lines · kajoljain
Quoted 53 lines · Kim Phillips
Quoted 39 lines · Jiri Olsa
Quoted 38 lines · Kim Phillips
--- tools/perf/util/evsel.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+)tools/perf/util/evsel.c +12
-0@@ -2743,9 +2743,22 @@ static bool find_process(const char *name) return ret ? false : true; } +static bool is_amd(const char *arch, const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid &&strstarts (cpuid,"AuthenticAMD" ); +} + +static bool is_amd_ibs (struct evsel *evsel) +{ + return evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name,"ibs" ,3 ); +} + int evsel__open_strerror(struct evsel *evsel, struct target *target, int err, char *msg, size_t size) { + struct perf_env *env = evsel__env(evsel); + const char *arch = perf_env__arch(env); + const char *cpuid = perf_env__cpuid(env); char sbuf[STRERR_BUFSIZE]; int printed = 0, enforced = 0; @@ -2841,6 +2854,17 @@ int evsel__open_strerror(struct evsel*evsel, struct target *target,
return scnprintf(msg, size, "wrong clockid (%d).",
clockid);
if (perf_missing_features.aux_output)Quoted 8 lines
Quoted 5 lines
return scnprintf(msg, size, "The '
aux_output'
feature
is not supported, update the kernel.");
+ if (is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {would single '
is_amd_ibs' call be better? checking on both amd and ibsGood suggestion. If you look at the later patch in the BRS series, I have rewritten it to add the new AMD PMU like so:
Quoted 14 lines
if (
is_amd()) {
if (is_amd_ibs()) {
if (evsel->this)
return
if (evsel->that)
return
}
+ if (is_amd_brs()) {
+ if (evsel->this)
+ return
+ if (evsel->that)
+ return
+ }
}Hi Kim,
From my point of view, it won't be a good idea of adding so many
checks in common function definition itself.
Can you just create a check to see if its amd machine and then add a function call which will handle all four conditions together?which is basically for:
+ if (
is_amd(arch, cpuid)) {
+ if (is_amd_ibs(evsel)) {
+ if (evsel->core.attr.exclude_kernel)
+ return scnprintf(msg, size,
+ "AMD IBS can't exclude kernel events. Try running at a higher privilege level.");
+ if (!evsel->core.system_wide)
+ return scnprintf(msg, size,
+ "AMD IBS may only be available in system-wide/per-cpu mode. Try using
-a, or -C and workload affinity");
+ }and this:
Quoted 12 lines
+ if (
is_amd_brs(evsel)) {
+ if (evsel->core.attr.freq)
+ return scnprintf(msg, size,
+ "AMD Branch Sampling does not support frequency mode sampling, must
pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.");
+ /* another reason is that the period is too small */ + return scnprintf(msg, size,
+ "AMD Branch Sampling does not support sampling period smaller than what is reported in/sys/devices/cpu/caps/branches.");
+ }IIRC, I tried something like that but carrying the
struct target *target, int err, char *msg,
size_tsizeparameters made things worse.
Quoted 5 lines
So, incase we are in amd machine, common function
evsel__open_strerrorwill call function may be something likeamd_evesel_open_strerror_checkwhich will look for both ibs and brs conditions and return corresponding
error statement.Quoted 7 lines
The vast majority of decisions made by
evsel__open_strerrorare going to be common across most arch/uarches. AMD has only these two pesky exceptions to the rule and therefore IMO it's ok to have them inline with the common function, since the decisions are so deeply intertwined. A newamd_evsel_open_strerror_checksounds like it'd duplicate too much of the common function code in order to handle the common error cases.Hi Kim,
Sorry for the confusion, what I meant by adding new function is just
to handle these corner error cases and not duplicating wholeevsel__open_strerrorcode.Maybe something like below code, Its just prototype of code to show you the flow, you can refine it and check for any build or indentation issues using checkpatch.pl script.
So basically, in common function we can just have 2 calls, first to check if we are in amd system and second to return corresponding error message, rather then adding whole chunk of if's which are specific to amd.
tools/perf/util/evsel.c +20
-0@@ -2852,9 +2852,40 @@ static bool find_process(const char *name) return ret ? false : true; } +static bool is_amd(const char *arch, const char *cpuid) +{ + return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid, "AuthenticAMD"); +} + +static int error_amd_ibs_brs(struct evsel *evsel, char *msg, size_t size) +{ + if (evsel->core.attr.precise_ip || !strncmp(evsel->pmu_name, "ibs", 3)) { + if (evsel->core.attr.exclude_kernel) + return scnprintf(msg, size, + "AMD IBS can't exclude kernel events. Try running at a higher privilege level."); + if (!evsel->core.system_wide) + return scnprintf(msg, size, + "AMD IBS may only be available in system-wide/per-cpu mode. Try using -a, or -C and workload affinity"); + } + + if (((evsel->core.attr.config & 0xff) == 0xc4) && (evsel->core.attr.sample_type & PERF_SAMPLE_BRANCH_STACK)) { + if (evsel->core.attr.freq) { + return scnprintf(msg, size, + "AMD Branch Sampling does not support frequency mode sampling,must pass a fixed sampling
+ period via -c option or cpu/branch-brs,period=xxxx/."); + /* another reason is that the period is too small */ + return scnprintf(msg, size,
+ "AMD Branch Sampling does not support sampling period smaller than what is reported in/sys/devices/cpu/caps/branches."); + }
+ }tools/perf/util/evsel.c +7
-0+} + int evsel__open_strerror(struct evsel *evsel, struct target *target, int err, char *msg, size_t size) { + struct perf_env *env = evsel__env(evsel); + const char *arch = perf_env__arch(env); + const char *cpuid = perf_env__cpuid(env); char sbuf[STRERR_BUFSIZE]; int printed = 0, enforced = 0;@@ -2950,6 +2981,8 @@ int evsel__open_strerror(struct evsel *evsel,struct target *target, return scnprintf(msg, size, "wrong clockid (%d).", clockid); if (perf_missing_features.aux_output) return scnprintf(msg, size, "The 'aux_output' feature is not supported, update the kernel."); + if (is_amd(arch, cpuid)) + return error_amd_ibs_brs(evsel, msg, size);Quoted 8 lines
Quoted 4 lines
break;
case ENODATA:
return scnprintf(msg, size, "Cannot collect data source
with the load latency event alone. "That change will makes AMD machines fail to fall back to the default "The
sys_perf_event_open()syscall returned with..." error string in case it's not those AMD IBS and BRS sub-conditions.
Quoted 3 lines
Yes right, as I mentioned before, the code I pointed was just a prototype to show you the flow, these corner cases can be handled on top of it.
Right but these corner cases disrupt the existing flow: adding int ret; ret = foo(); if (ret) goto report_generic_einval doesn't go with the flow.
Quoted 11 lines
Quoted 8 lines
Is having the AMD error code checking in the main
evsel__open_strerror()so bad? Other arches and their PMU implementations may find error conditions that they have in common with AMD's, therefore opening up the code for opposite types of refactoring and reuse than what is being requested here. E.g., I've seen other hardware configurations - not specific to one architecture - that could also use this message:From my understanding, adding too many checks in common function
for a specific arch is not a good practice. Since you already adding
My point above is that other arches can come in and adopt the same error conditions and text.
Quoted 3 lines
multiple functions to get information like ,if current system is amd/ibs/brs. Can't we rather just add a single function and handled all these checks there?
That will remove the code from the common path. Code that's possible (and in some cases likely) that will be adopted by other arches.
Thanks,
Kim
Quoted 7 lines · Kim Phillips
It's possible to have an evsel and
evsel->evlistpopulated without anevsel->evlist->env, when, e.g.,cmd_recordis in its error path.Future patches will add support for
evsel__open_strerrorto be able to customize error messaging based on perf_env__{arch,cpuid}, so let's haveevsel__envreturn&perf_envinstead of NULL in that case.
Patch looks good to me.
Thanks,
Kajol Jain
Quoted 33 lines
Signed-off-by Kim Phillips <[email protected]>Cc Arnaldo Carvalho de Melo <[email protected]>Cc Alexander Shishkin <[email protected]>Cc Boris Ostrovsky <[email protected]>Cc Ian Rogers <[email protected]>Cc Ingo Molnar <[email protected]>Cc Jiri Olsa <[email protected]>Cc Joao Martins <[email protected]>Cc Konrad Rzeszutek Wilk <[email protected]>Cc Mark Rutland <[email protected]>Cc Michael Petlan <[email protected]>Cc Namhyung Kim <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Robert Richter <[email protected]>Cc Stephane Eranian <[email protected]>--- tools/perf/util/evsel.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)tools/perf/util/evsel.c +1
-1@@ -2857,7 +2857,7 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,struct perf_env *evsel__env (struct evsel *evsel) { -if (evsel && evsel->evlist) +if (evsel && evsel->evlist && evsel->evlist->env)return evsel->evlist->env;return &perf_env; }
Arnaldo Carvalho de Melo
maintainerEm Tue, Nov 23, 2021 at 01:56:38PM +0530, kajoljain escreveu:
Quoted 14 lines · Kim Phillips
Quoted 7 lines
It's possible to have an evsel and
evsel->evlistpopulated without anevsel->evlist->env, when, e.g.,cmd_recordis in its error path.Future patches will add support for
evsel__open_strerrorto be able to customize error messaging based on perf_env__{arch,cpuid}, so let's haveevsel__envreturn&perf_envinstead of NULL in that case.Patch looks good to me.
Reviewed-by Kajol Jain<[email protected]>
Collecting this one now, found it thru the cracks.
- Arnaldo
Quoted 36 lines
Thanks,
Kajol JainQuoted 33 lines
Signed-off-by Kim Phillips <[email protected]>Cc Arnaldo Carvalho de Melo <[email protected]>Cc Alexander Shishkin <[email protected]>Cc Boris Ostrovsky <[email protected]>Cc Ian Rogers <[email protected]>Cc Ingo Molnar <[email protected]>Cc Jiri Olsa <[email protected]>Cc Joao Martins <[email protected]>Cc Konrad Rzeszutek Wilk <[email protected]>Cc Mark Rutland <[email protected]>Cc Michael Petlan <[email protected]>Cc Namhyung Kim <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Robert Richter <[email protected]>Cc Stephane Eranian <[email protected]>--- tools/perf/util/evsel.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)tools/perf/util/evsel.c +1
-1@@ -2857,7 +2857,7 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,struct perf_env *evsel__env (struct evsel *evsel) { -if (evsel && evsel->evlist) +if (evsel && evsel->evlist && evsel->evlist->env)return evsel->evlist->env;return &perf_env; }
Signature · 1 line
- Arnaldo