Re: [PATCH] tracing: Add "grouping" to histogram logic

From: Tom Zanussi
Date: Tue Jul 06 2021 - 11:04:45 EST

Next message: Miklos Szeredi: "[GIT PULL] fuse update for 5.14"
Previous message: Paolo Bonzini: "Re: [RFC PATCH v2 51/69] KVM: x86/mmu: Allow per-VM override of the TDP max page level"
In reply to: Tom Zanussi: "Re: [PATCH] tracing: Add "grouping" to histogram logic"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi Steve,

On Fri, 2021-07-02 at 17:56 -0400, Steven Rostedt wrote:
> From: "Steven Rostedt (VMware)" <rostedt@xxxxxxxxxxx>
>
> There's been several times I wished the histogram logic had a
> "grouping"
> feature for the buckets. Currently, each bucket has a size of one.
> That
> is, if you trace the amount of requested allocations, each allocation
> is
> its own bucket, even if you are interested in what allocates 100
> bytes or
> less, 100 to 200, 200 to 300, etc.
>
> Also, without grouping, it fills up the allocated histogram buckets
> quickly. If you are tracking latency, and don't care if something is
> 200
> microseconds off, or 201 microseconds off, but want to track them by
> say
> 10 microseconds each. This can not currently be done.
>
> Introduce a "grouping" command to each field where it will record in
> a
> rounded number. For example:
>
> ># echo 'hist:keys=bytes_req-100:sort=bytes_req' > events/kmem/kmalloc/trigger

Very nice idea, makes a lot of sense to add this.

As for the syntax, it kinds of reads to me like subtraction. If it
were possible to change it to 'x100' that would seem more like what
it's getting at - grouping things 'by 100'. But since that would be
indistinguishable as to whether or not the 'x' was part of the variable
name, how about using '*100'?

> ># cat events/kmem/kmalloc/hist
> # event histogram
> #
> # trigger info:
> hist:keys=bytes_req-100:vals=hitcount:sort=bytes_req-100:size=2048
> [active]
> #
>
> { bytes_req: 0-99 } hitcount: 7207
> { bytes_req: 100-199 } hitcount: 1030
> { bytes_req: 200-299 } hitcount: 602
> { bytes_req: 300-399 } hitcount: 414
> { bytes_req: 400-499 } hitcount: 584
> { bytes_req: 500-599 } hitcount: 35
> { bytes_req: 600-699 } hitcount: 361
> { bytes_req: 700-799 } hitcount: 37
> { bytes_req: 800-899 } hitcount: 26
> { bytes_req: 900-999 } hitcount: 15
> { bytes_req: 1100-1199 } hitcount: 4
> { bytes_req: 1200-1299 } hitcount: 102
> { bytes_req: 1300-1399 } hitcount: 2
> { bytes_req: 1400-1499 } hitcount: 15
> { bytes_req: 1500-1599 } hitcount: 100
> { bytes_req: 1600-1699 } hitcount: 1
> { bytes_req: 1700-1799 } hitcount: 3
> { bytes_req: 1800-1899 } hitcount: 10
> { bytes_req: 2000-2099 } hitcount: 19
> { bytes_req: 2100-2199 } hitcount: 3
> { bytes_req: 2500-2599 } hitcount: 72
> { bytes_req: 2900-2999 } hitcount: 1
> { bytes_req: 3300-3399 } hitcount: 4
> { bytes_req: 3900-3999 } hitcount: 2
> { bytes_req: 4000-4099 } hitcount: 1761
> { bytes_req: 4600-4699 } hitcount: 65
> { bytes_req: 5000-5099 } hitcount: 3
> { bytes_req: 6500-6599 } hitcount: 1
> { bytes_req: 8100-8199 } hitcount: 1
> { bytes_req: 9800-9899 } hitcount: 1
>
> Totals:
> Hits: 12481
> Entries: 30
> Dropped: 0
>
> This finally makes the histograms actual histograms!
>
> Signed-off-by: Steven Rostedt (VMware) <rostedt@xxxxxxxxxxx>
> ---
> kernel/trace/trace_events_hist.c | 34 ++++++++++++++++++++++++++++
> ----
> 1 file changed, 30 insertions(+), 4 deletions(-)
>
> diff --git a/kernel/trace/trace_events_hist.c
> b/kernel/trace/trace_events_hist.c
> index ba03b7d84fc2..d6e644df9506 100644
> --- a/kernel/trace/trace_events_hist.c
> +++ b/kernel/trace/trace_events_hist.c
> @@ -120,6 +120,7 @@ struct hist_field {
> unsigned int size;
> unsigned int offset;
> unsigned int is_signed;
> + unsigned long grouping;
> const char *type;
> struct hist_field *operands[HIST_FIELD_OPERANDS_MAX];
> struct hist_trigger_data *hist_data;
> @@ -3761,6 +3762,7 @@ static int create_key_field(struct
> hist_trigger_data *hist_data,
> {
> struct trace_array *tr = hist_data->event_file->tr;
> struct hist_field *hist_field = NULL;
> + char *field;
> unsigned long flags = 0;
> unsigned int key_size;
> int ret = 0;
> @@ -3768,14 +3770,16 @@ static int create_key_field(struct
> hist_trigger_data *hist_data,
> if (WARN_ON(key_idx >= HIST_FIELDS_MAX))
> return -EINVAL;
>
> + field = strsep(&field_str, "-");
> +
> flags |= HIST_FIELD_FL_KEY;
>
> - if (strcmp(field_str, "stacktrace") == 0) {
> + if (strcmp(field, "stacktrace") == 0) {
> flags |= HIST_FIELD_FL_STACKTRACE;
> key_size = sizeof(unsigned long) *
> HIST_STACKTRACE_DEPTH;
> hist_field = create_hist_field(hist_data, NULL, flags,
> NULL);
> } else {
> - hist_field = parse_expr(hist_data, file, field_str,
> flags,
> + hist_field = parse_expr(hist_data, file, field, flags,
> NULL, 0);
> if (IS_ERR(hist_field)) {
> ret = PTR_ERR(hist_field);
> @@ -3783,7 +3787,7 @@ static int create_key_field(struct
> hist_trigger_data *hist_data,
> }
>
> if (field_has_hist_vars(hist_field, 0)) {
> - hist_err(tr, HIST_ERR_INVALID_REF_KEY,
> errpos(field_str));
> + hist_err(tr, HIST_ERR_INVALID_REF_KEY,
> errpos(field));
> destroy_hist_field(hist_field, 0);
> ret = -EINVAL;
> goto out;
> @@ -3792,6 +3796,14 @@ static int create_key_field(struct
> hist_trigger_data *hist_data,
> key_size = hist_field->size;
> }
>
> + if (field_str) {
> + unsigned long grouping;
> +
> + ret = kstrtoul(field_str, 0, &grouping);
> + if (!ret)
> + hist_field->grouping = grouping;
> + }
> +
> hist_data->fields[key_idx] = hist_field;
>
> key_size = ALIGN(key_size, sizeof(u64));
> @@ -4548,8 +4560,18 @@ static void event_hist_trigger(struct
> event_trigger_data *data,
> if (key_field->flags & HIST_FIELD_FL_STRING) {
> key = (void *)(unsigned
> long)field_contents;
> use_compound_key = true;
> - } else
> + } else {
> + if (key_field->grouping) {
> + unsigned long grouping =
> key_field->grouping;
> +
> + if (field_contents >= LONG_MAX)
> + field_contents =
> div64_ul(field_contents, grouping);
> + else
> + field_contents =
> (u64)((unsigned long)field_contents / grouping);
> + field_contents *= grouping;
> + }
> key = (void *)&field_contents;
> + }

As far as implementation, there's already something similar to how it
could operate, the log2 modifier. If you followed that example, you
could put the above code into a hist_field_fn_t:

static u64 hist_field_grouped(struct hist_field *hist_field,
struct tracing_map_elt *elt,
struct ring_buffer_event *rbe,
void *event)
{
..
}

and also add a new HIST_FIELD_FL_GROUPED flag to enum hist_field_flags.

I know the code handling all this is kind of complicated - if you want
I can rework the patch to do this, just let me know if so.

Thanks,

Tom

> }
>
> if (use_compound_key)
> @@ -4663,6 +4685,8 @@ static void hist_trigger_print_key(struct
> seq_file *m,
> } else {
> uval = *(u64 *)(key + key_field->offset);
> seq_printf(m, "%s: %10llu", field_name, uval);
> + if (key_field->grouping)
> + seq_printf(m, "-%llu", uval +
> key_field->grouping - 1);
> }
> }
>
> @@ -5096,6 +5120,8 @@ static void hist_field_print(struct seq_file
> *m, struct hist_field *hist_field)
> seq_printf(m, ".%s", flags);
> }
> }
> + if (hist_field->grouping)
> + seq_printf(m, "-%ld", hist_field->grouping);
> }
>
> static int event_hist_trigger_print(struct seq_file *m,

Next message: Miklos Szeredi: "[GIT PULL] fuse update for 5.14"
Previous message: Paolo Bonzini: "Re: [RFC PATCH v2 51/69] KVM: x86/mmu: Allow per-VM override of the TDP max page level"
In reply to: Tom Zanussi: "Re: [PATCH] tracing: Add "grouping" to histogram logic"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]