summaryrefslogtreecommitdiff
path: root/Documentation/tools/rtla/common_timerlat_options.rst
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2024-05-18 04:49:18 +0300
committerLinus Torvalds <torvalds@linux-foundation.org>2024-05-18 04:49:18 +0300
commit26aa834ff2eef00a863a64b1a94a5a88a94eb963 (patch)
treeef165f0cbf4272d525d034837569735bfa6d3392 /Documentation/tools/rtla/common_timerlat_options.rst
parentfa3889d970da8f95bdd90fb5a07d521810be5678 (diff)
parent59c22f70b2951d81de410d477ae536ba951b4f37 (diff)
downloadlinux-26aa834ff2eef00a863a64b1a94a5a88a94eb963.tar.xz
Merge tag 'trace-tools-v6.10' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace
Pull tracing tool updates from Steven Rostedt: "Specific for timerlat: - Improve the output of timerlat top by adding a missing \n, and by avoiding printing color-formatting characters where they are translated to regular characters. - Improve timerlat auto-analysis output by replacing '\t' with spaces to avoid copy-and-paste issues when reporting problems. - Make the user-space (-u) option the default, as it is the most complete test. Add a -k option to use the in-kernel workload. - On timerlat top and hist, add a summary with the overall results. For instance, the minimum value for all CPUs, the overall average and the maximum value from all CPUs. - timerlat hist was printing initial values (i.e., 0 as max, and ~0 as min) if the trace stopped before the first Ret-User event. This problem was fixed by printing the " - " no value string to the output if that was the case. For all RTLA tools: - Add a --warm-up <seconds> option, allowing the workload to run for <seconds> before starting to collect results. - Add a --trace-buffer-size option, allowing the user to set the tracing buffer size for -t option. This option is mainly useful for reducing the trace file. Now rtla depends on libtracefs >= 1.6. - Fix the -t [trace_file] parsing, now it does not require the '=' before the option parameter, and better handles the multiple ways a user can pass the trace_file.txt" * tag 'trace-tools-v6.10' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: rtla: Documentation: Fix -t, --trace rtla: Fix -t\--trace[=file] rtla/timerlat: Fix histogram report when a cpu count is 0 rtla: Add --trace-buffer-size option rtla/timerlat: Make user-space threads the default rtla: Add the --warm-up option rtla/timerlat: Add a summary for hist mode rtla/timerlat: Add a summary for top mode rtla/timerlat: Use pretty formatting only on interactive tty rtla/auto-analysis: Replace \t with spaces rtla/timerlat: Simplify "no value" printing on top
Diffstat (limited to 'Documentation/tools/rtla/common_timerlat_options.rst')
-rw-r--r--Documentation/tools/rtla/common_timerlat_options.rst10
1 files changed, 9 insertions, 1 deletions
diff --git a/Documentation/tools/rtla/common_timerlat_options.rst b/Documentation/tools/rtla/common_timerlat_options.rst
index d3255ed70195..cef6651f1435 100644
--- a/Documentation/tools/rtla/common_timerlat_options.rst
+++ b/Documentation/tools/rtla/common_timerlat_options.rst
@@ -22,17 +22,25 @@
Save the stack trace at the *IRQ* if a *Thread* latency is higher than the
argument in us.
+**-t**, **--trace** \[*file*]
+
+ Save the stopped trace to [*file|timerlat_trace.txt*].
+
**--dma-latency** *us*
Set the /dev/cpu_dma_latency to *us*, aiming to bound exit from idle latencies.
*cyclictest* sets this value to *0* by default, use **--dma-latency** *0* to have
similar results.
+**-k**, **--kernel-threads**
+
+ Use timerlat kernel-space threads, in contrast of **-u**.
+
**-u**, **--user-threads**
Set timerlat to run without a workload, and then dispatches user-space workloads
to wait on the timerlat_fd. Once the workload is awakes, it goes to sleep again
adding so the measurement for the kernel-to-user and user-to-kernel to the tracer
- output.
+ output. **--user-threads** will be used unless the user specify **-k**.
**-U**, **--user-load**