From: Yunseong Kim <[email protected]>
According to perf-kvm.txt, the default operation is --guest, meaning
perf_host should be set to false and perf_guest to true. Currently,
the incorrect usage of --no-host and --no-guest still results in
the default --guest operation.
This patch updates the usage messages to include the --no-guest option
for those who only want host data, as well as the --no-host and
--no-guest case.
Signed-off-by: Yunseong Kim <[email protected]>
---
tools/perf/builtin-kvm.c | 19 +++++++++++++------
1 file changed, 13 insertions(+), 6 deletions(-)
diff --git a/tools/perf/builtin-kvm.c b/tools/perf/builtin-kvm.c
index 71165036e4ca..654a07658cdd 100644
--- a/tools/perf/builtin-kvm.c
+++ b/tools/perf/builtin-kvm.c
@@ -2129,7 +2129,7 @@ int cmd_kvm(int argc, const char **argv)
OPT_BOOLEAN(0, "guest", &perf_guest,
"Collect guest os data"),
OPT_BOOLEAN(0, "host", &perf_host,
- "Collect host os data"),
+ "Collect host os data. Host only with --host --no-guest"),
OPT_STRING(0, "guestmount", &symbol_conf.guestmount, "directory",
"guest mount directory under which every guest os"
" instance has a subdir"),
@@ -2150,16 +2150,23 @@ int cmd_kvm(int argc, const char **argv)
"buildid-list", "stat", NULL };
const char *kvm_usage[] = { NULL, NULL };
- perf_host = 0;
- perf_guest = 1;
+ /*
+ * tools/perf/Documentation/perf-kvm.txt
+ *
+ * Default('') -> perf.data.guest
+ * --host -> perf.data.kvm
+ * --guest -> perf.data.guest
+ * --host --guest -> perf.data.kvm
+ * --host --no-guest -> perf.data.host
+ */
+ perf_host = false;
+ perf_guest = true;
argc = parse_options_subcommand(argc, argv, kvm_options, kvm_subcommands, kvm_usage,
PARSE_OPT_STOP_AT_NON_OPTION);
- if (!argc)
+ if (!argc || (!perf_host && !perf_guest))
usage_with_options(kvm_usage, kvm_options);
- if (!perf_host)
- perf_guest = 1;
if (!file_name) {
file_name = get_filename_for_perf_kvm();
--
2.34.1
Hello,
On Wed, May 15, 2024 at 12:54 PM <[email protected]> wrote:
>
> From: Yunseong Kim <[email protected]>
>
> According to perf-kvm.txt, the default operation is --guest, meaning
> perf_host should be set to false and perf_guest to true. Currently,
> the incorrect usage of --no-host and --no-guest still results in
> the default --guest operation.
Hmm, right.. This is not good.
>
> This patch updates the usage messages to include the --no-guest option
> for those who only want host data, as well as the --no-host and
> --no-guest case.
I'm not sure if it's worth mentioning the host-only use case which is
not expected for this command. I think it's enough to warn the
no-host and no-guest use cases.
>
> Signed-off-by: Yunseong Kim <[email protected]>
> ---
> tools/perf/builtin-kvm.c | 19 +++++++++++++------
> 1 file changed, 13 insertions(+), 6 deletions(-)
>
> diff --git a/tools/perf/builtin-kvm.c b/tools/perf/builtin-kvm.c
> index 71165036e4ca..654a07658cdd 100644
> --- a/tools/perf/builtin-kvm.c
> +++ b/tools/perf/builtin-kvm.c
> @@ -2129,7 +2129,7 @@ int cmd_kvm(int argc, const char **argv)
> OPT_BOOLEAN(0, "guest", &perf_guest,
> "Collect guest os data"),
> OPT_BOOLEAN(0, "host", &perf_host,
> - "Collect host os data"),
> + "Collect host os data. Host only with --host --no-guest"),
> OPT_STRING(0, "guestmount", &symbol_conf.guestmount, "directory",
> "guest mount directory under which every guest os"
> " instance has a subdir"),
> @@ -2150,16 +2150,23 @@ int cmd_kvm(int argc, const char **argv)
> "buildid-list", "stat", NULL };
> const char *kvm_usage[] = { NULL, NULL };
>
> - perf_host = 0;
> - perf_guest = 1;
> + /*
> + * tools/perf/Documentation/perf-kvm.txt
> + *
> + * Default('') -> perf.data.guest
> + * --host -> perf.data.kvm
> + * --guest -> perf.data.guest
> + * --host --guest -> perf.data.kvm
> + * --host --no-guest -> perf.data.host
> + */
> + perf_host = false;
> + perf_guest = true;
Probably not needed.
>
> argc = parse_options_subcommand(argc, argv, kvm_options, kvm_subcommands, kvm_usage,
> PARSE_OPT_STOP_AT_NON_OPTION);
> - if (!argc)
> + if (!argc || (!perf_host && !perf_guest))
> usage_with_options(kvm_usage, kvm_options);
>
> - if (!perf_host)
> - perf_guest = 1;
Instead of showing the whole usage, I think it's better to
print a specific message for this case.
Oh, I've realized that we have 'h' modifier for hypervisors
(on ARM64?) Then maybe no-host + no-guest is a valid
combination. Probably we need to add --hypervisor option
then.
Thanks,
Namhyung
>
> if (!file_name) {
> file_name = get_filename_for_perf_kvm();
> --
> 2.34.1
>
On 5/23/24 1:23 오후, Namhyung Kim wrote:
> Hello,
>
> On Wed, May 15, 2024 at 12:54 PM <[email protected]> wrote:
>>
>> From: Yunseong Kim <[email protected]>
>>
>> According to perf-kvm.txt, the default operation is --guest, meaning
>> perf_host should be set to false and perf_guest to true. Currently,
>> the incorrect usage of --no-host and --no-guest still results in
>> the default --guest operation.
>
> Hmm, right.. This is not good.
>
>>
>> This patch updates the usage messages to include the --no-guest option
>> for those who only want host data, as well as the --no-host and
>> --no-guest case.
>
> I'm not sure if it's worth mentioning the host-only use case which is
> not expected for this command. I think it's enough to warn the
> no-host and no-guest use cases.
Thank you Namhyung for the code review!
I wasn't sure if it was necessary to mention that either.
I should have asked in the patch mail first.
>> Signed-off-by: Yunseong Kim <[email protected]>
>> ---
>> tools/perf/builtin-kvm.c | 19 +++++++++++++------
>> 1 file changed, 13 insertions(+), 6 deletions(-)
>>
>> diff --git a/tools/perf/builtin-kvm.c b/tools/perf/builtin-kvm.c
>> index 71165036e4ca..654a07658cdd 100644
>> --- a/tools/perf/builtin-kvm.c
>> +++ b/tools/perf/builtin-kvm.c
>> @@ -2129,7 +2129,7 @@ int cmd_kvm(int argc, const char **argv)
>> OPT_BOOLEAN(0, "guest", &perf_guest,
>> "Collect guest os data"),
>> OPT_BOOLEAN(0, "host", &perf_host,
>> - "Collect host os data"),
>> + "Collect host os data. Host only with --host --no-guest"),
>> OPT_STRING(0, "guestmount", &symbol_conf.guestmount, "directory",
>> "guest mount directory under which every guest os"
>> " instance has a subdir"),
>> @@ -2150,16 +2150,23 @@ int cmd_kvm(int argc, const char **argv)
>> "buildid-list", "stat", NULL };
>> const char *kvm_usage[] = { NULL, NULL };
>>
>> - perf_host = 0;
>> - perf_guest = 1;
>> + /*
>> + * tools/perf/Documentation/perf-kvm.txt
>> + *
>> + * Default('') -> perf.data.guest
>> + * --host -> perf.data.kvm
>> + * --guest -> perf.data.guest
>> + * --host --guest -> perf.data.kvm
>> + * --host --no-guest -> perf.data.host
>> + */
>> + perf_host = false;
>> + perf_guest = true;
>
> Probably not needed.
I agree.
>>
>> argc = parse_options_subcommand(argc, argv, kvm_options, kvm_subcommands, kvm_usage,
>> PARSE_OPT_STOP_AT_NON_OPTION);
>> - if (!argc)
>> + if (!argc || (!perf_host && !perf_guest))
>> usage_with_options(kvm_usage, kvm_options);
>>
>> - if (!perf_host)
>> - perf_guest = 1;
>
> Instead of showing the whole usage, I think it's better to
> print a specific message for this case.
>
> Oh, I've realized that we have 'h' modifier for hypervisors
> (on ARM64?) Then maybe no-host + no-guest is a valid
> combination. Probably we need to add --hypervisor option
> then.
>
> Thanks,
> Namhyung
Yes, I have a native Arm64 devlopment environment.
So I'll implement and test it and send the patch.
Thank you so much for the detailed review!
Warm Regards,
Yunseong Kim
>>
>> if (!file_name) {
>> file_name = get_filename_for_perf_kvm();
>> --
>> 2.34.1
>>