A key with the ".buckets=N" modifier is stored as the start of its bucket,
and hist_trigger_print_key() prints the range it covers as start to
start + N - 1. That addition overflows once the start is within N of
U64_MAX:
# echo 'hist:keys=offset.buckets=10' >> \
events/syscalls/sys_enter_lseek/trigger
# ./seeker 5 -1 # lseek(0, 5) and lseek(0, -1)
# cat events/syscalls/sys_enter_lseek/hist
...
{ offset: ~ 18446744073709551610-3 } hitcount: 1
{ offset: ~ 0-9 } hitcount: 1
...
18446744073709551610 + 10 - 1 does not fit in 64 bits, so the range ends
below where it starts.
The bucket does end at U64_MAX: it holds every value hist_field_bucket()
maps onto its start, and there is no value above U64_MAX. Clamp the printed
end there. Printed honestly the group is shorter than the size asked for,
which histogram.rst does not mention, so say it there.
Fixes: de9a48a360b7 ("tracing: Add linear buckets to histogram logic")
Signed-off-by: Donggeun Yoo <[email protected]>
Assisted-by: Claude:claude-fable-5
---
Tested under QEMU/KVM on 704340f1cd0d, x86_64, unfixed arm first. Keys come
from lseek(2) offsets.
Nine cases over buckets=1, 10 and 16. One printed range differs:
buckets=10, key 18446744073709551610
before ~ 18446744073709551610-3
after ~ 18446744073709551610-18446744073709551615
The other eight are identical on both arms.
ftracetest test.d/trigger: 42 pass, 2 xfail, 1 fail, per-test verdicts
identical on both arms. The failure is trigger-synthetic-event-dynstring.tc,
which needs a ping(8) that is not a busybox applet. No case in the suite
uses .buckets.
Documentation/trace/histogram.rst | 3 +++
kernel/trace/trace_events_hist.c | 7 ++++++-
2 files changed, 9 insertions(+), 1 deletion(-)
diff --git a/Documentation/trace/histogram.rst
b/Documentation/trace/histogram.rst
index 340bcb5099e7..97b67f05f5ea 100644
--- a/Documentation/trace/histogram.rst
+++ b/Documentation/trace/histogram.rst
@@ -1943,6 +1943,9 @@ the ".buckets" modifier and specify a size (in this case
groups of 10)::
Entries: 16
Dropped: 0
+Keys are 64-bit, so the group holding the largest values is truncated where
+that range ends and can be smaller than the size asked for.
+
To save stacktraces, create a synthetic event with a field of type "unsigned
long[]"
or even just "long[]". For example, to see how long a task is blocked in an
uninterruptible state::
diff --git a/kernel/trace/trace_events_hist.c b/kernel/trace/trace_events_hist.c
index 8af97fd4ee2d..86df1a29d8d5 100644
--- a/kernel/trace/trace_events_hist.c
+++ b/kernel/trace/trace_events_hist.c
@@ -5560,9 +5560,14 @@ static void hist_trigger_print_key(struct seq_file *m,
*(u64 *)(key + key_field->offset));
} else if (key_field->flags & HIST_FIELD_FL_BUCKET) {
unsigned long buckets = key_field->buckets;
+ u64 end;
+
uval = *(u64 *)(key + key_field->offset);
+ end = uval + buckets - 1;
+ if (end < uval)
+ end = U64_MAX;
seq_printf(m, "%s: ~ %llu-%llu", field_name,
- uval, uval + buckets -1);
+ uval, end);
} else if (key_field->flags & HIST_FIELD_FL_STRING) {
seq_printf(m, "%s: %-50s", field_name,
(char *)(key + key_field->offset));
--
2.53.0
