Linux kernel now supports statistics for BPF programs, and bpftool is
able to dump them. However, these statistics are not enabled by default,
and administrators may not know how to access them.
Add a paragraph in bpftool documentation, under the description of the
"bpftool prog show" command, to explain that such statistics are
available and that their collection is controlled via a dedicated sysctl
knob.
Signed-off-by: Quentin Monnet <quentin.monnet@xxxxxxxxxxxxx>
Reviewed-by: Jakub Kicinski <jakub.kicinski@xxxxxxxxxxxxx>
---
tools/bpf/bpftool/Documentation/bpftool-prog.rst | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/tools/bpf/bpftool/Documentation/bpftool-prog.rst b/tools/bpf/bpftool/Documentation/bpftool-prog.rst
index f2beb79a77e7..bb9bb00c0c2c 100644
--- a/tools/bpf/bpftool/Documentation/bpftool-prog.rst
+++ b/tools/bpf/bpftool/Documentation/bpftool-prog.rst
@@ -56,6 +56,14 @@ DESCRIPTION
Output will start with program ID followed by program type and
zero or more named attributes (depending on kernel version).
+ Since Linux 5.1 the kernel can collect statistics on BPF
+ programs (such as the total time spent running the program,
+ and the number of times it was run). If available, bpftool
+ shows such statistics. However, the kernel does not collect
+ them by defaults, as it slightly impacts performance on each
+ program run. Activation or deactivation of the feature is
+ performed via the **kernel.bpf_stats_enabled** sysctl knob.
+
**bpftool prog dump xlated** *PROG* [{ **file** *FILE* | **opcodes** | **visual** | **linum** }]
Dump eBPF instructions of the program from the kernel. By
default, eBPF will be disassembled and printed to standard
--
2.17.1
