| Message ID | 20230515183045.654199-1-leobras@redhat.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | trace,smp: Add tracepoints for csd | expand |
On Mon, 2023-05-15 at 15:30 -0300, Leonardo Bras wrote: > Changes since RFCv3: > - Split the patch in 2: entry/exit and queue > - Fix 'struct __call_single_data' & call_single_data_t alignment issue > - Made all TP_printk follow the same pattern > > Changes since RFCv2: > - Fixed some spacing issues and trace calls > > Changes since RFCv1: > - Implemented trace_csd_queue_cpu() as suggested by Valentin Schneider > - Using EVENT_CLASS in order to avoid duplication > - Introduced new helper: csd_do_func() > - Name change from smp_call_function_* to csd_function_* > - Rebased on top of torvalds/master > > Leonardo Bras (2): > trace,smp: Add tracepoints around remotelly called functions > trace,smp: Add tracepoints for scheduling remotelly called functions > > include/trace/events/smp.h | 72 ++++++++++++++++++++++++++++++++++++++ > kernel/smp.c | 41 +++++++++++++--------- > 2 files changed, 96 insertions(+), 17 deletions(-) > create mode 100644 include/trace/events/smp.h > argh, almost sure some llvm build will output some alignment warning because csd_do_func() has an 'struct __call_single_data' argument while the tracepoints have call_single_data_t arguments. Other than that, it should be fine. I will collect other suggestions before sending a v5 with this fix. Leo
On 15/05/23 15:30, Leonardo Bras wrote: > Changes since RFCv3: > - Split the patch in 2: entry/exit and queue > - Fix 'struct __call_single_data' & call_single_data_t alignment issue > - Made all TP_printk follow the same pattern > One bikeshedding nit I have is I'd prefer to see the csd_* events in a csd trace system, to have a sane nomenclature. Other than that and the other comment on csd_do_func(), this LGTM. > Changes since RFCv2: > - Fixed some spacing issues and trace calls > > Changes since RFCv1: > - Implemented trace_csd_queue_cpu() as suggested by Valentin Schneider > - Using EVENT_CLASS in order to avoid duplication > - Introduced new helper: csd_do_func() > - Name change from smp_call_function_* to csd_function_* > - Rebased on top of torvalds/master > > Leonardo Bras (2): > trace,smp: Add tracepoints around remotelly called functions > trace,smp: Add tracepoints for scheduling remotelly called functions > > include/trace/events/smp.h | 72 ++++++++++++++++++++++++++++++++++++++ > kernel/smp.c | 41 +++++++++++++--------- > 2 files changed, 96 insertions(+), 17 deletions(-) > create mode 100644 include/trace/events/smp.h > > -- > 2.40.1
On Tue, May 30, 2023 at 7:36 AM Valentin Schneider <vschneid@redhat.com> wrote: > > On 15/05/23 15:30, Leonardo Bras wrote: > > Changes since RFCv3: > > - Split the patch in 2: entry/exit and queue > > - Fix 'struct __call_single_data' & call_single_data_t alignment issue > > - Made all TP_printk follow the same pattern > > > > One bikeshedding nit I have is I'd prefer to see the csd_* events in a csd > trace system, to have a sane nomenclature. Oh, here you mean the file name? (ie: trace/events/smp.h -> trace/events/csd.h) > > Other than that and the other comment on csd_do_func(), this LGTM. > Thanks! Other than the above question, it should be fast for me to send a v5. Best regards, Leo > > Changes since RFCv2: > > - Fixed some spacing issues and trace calls > > > > Changes since RFCv1: > > - Implemented trace_csd_queue_cpu() as suggested by Valentin Schneider > > - Using EVENT_CLASS in order to avoid duplication > > - Introduced new helper: csd_do_func() > > - Name change from smp_call_function_* to csd_function_* > > - Rebased on top of torvalds/master > > > > Leonardo Bras (2): > > trace,smp: Add tracepoints around remotelly called functions > > trace,smp: Add tracepoints for scheduling remotelly called functions > > > > include/trace/events/smp.h | 72 ++++++++++++++++++++++++++++++++++++++ > > kernel/smp.c | 41 +++++++++++++--------- > > 2 files changed, 96 insertions(+), 17 deletions(-) > > create mode 100644 include/trace/events/smp.h > > > > -- > > 2.40.1 >
On Wed, Jun 14, 2023 at 1:14 AM Leonardo Bras Soares Passos <leobras@redhat.com> wrote: > > On Tue, May 30, 2023 at 7:36 AM Valentin Schneider <vschneid@redhat.com> wrote: > > > > On 15/05/23 15:30, Leonardo Bras wrote: > > > Changes since RFCv3: > > > - Split the patch in 2: entry/exit and queue > > > - Fix 'struct __call_single_data' & call_single_data_t alignment issue > > > - Made all TP_printk follow the same pattern > > > > > > > One bikeshedding nit I have is I'd prefer to see the csd_* events in a csd > > trace system, to have a sane nomenclature. > > Oh, here you mean the file name? (ie: trace/events/smp.h -> trace/events/csd.h) Or just -#define TRACE_SYSTEM smp +#define TRACE_SYSTEM csd ? > > > > > Other than that and the other comment on csd_do_func(), this LGTM. > > > > Thanks! > Other than the above question, it should be fast for me to send a v5. > > Best regards, > Leo > > > > > Changes since RFCv2: > > > - Fixed some spacing issues and trace calls > > > > > > Changes since RFCv1: > > > - Implemented trace_csd_queue_cpu() as suggested by Valentin Schneider > > > - Using EVENT_CLASS in order to avoid duplication > > > - Introduced new helper: csd_do_func() > > > - Name change from smp_call_function_* to csd_function_* > > > - Rebased on top of torvalds/master > > > > > > Leonardo Bras (2): > > > trace,smp: Add tracepoints around remotelly called functions > > > trace,smp: Add tracepoints for scheduling remotelly called functions > > > > > > include/trace/events/smp.h | 72 ++++++++++++++++++++++++++++++++++++++ > > > kernel/smp.c | 41 +++++++++++++--------- > > > 2 files changed, 96 insertions(+), 17 deletions(-) > > > create mode 100644 include/trace/events/smp.h > > > > > > -- > > > 2.40.1 > >
On 14/06/23 02:16, Leonardo Bras Soares Passos wrote: > On Wed, Jun 14, 2023 at 1:14 AM Leonardo Bras Soares Passos > <leobras@redhat.com> wrote: >> >> On Tue, May 30, 2023 at 7:36 AM Valentin Schneider <vschneid@redhat.com> wrote: >> > >> > On 15/05/23 15:30, Leonardo Bras wrote: >> > > Changes since RFCv3: >> > > - Split the patch in 2: entry/exit and queue >> > > - Fix 'struct __call_single_data' & call_single_data_t alignment issue >> > > - Made all TP_printk follow the same pattern >> > > >> > >> > One bikeshedding nit I have is I'd prefer to see the csd_* events in a csd >> > trace system, to have a sane nomenclature. >> >> Oh, here you mean the file name? (ie: trace/events/smp.h -> trace/events/csd.h) > > Or just > -#define TRACE_SYSTEM smp > +#define TRACE_SYSTEM csd > ? > At the very least this yes, and making the file name match that probably wouldn't hurt.