Fix trace remotes read with an offline CPU
v2 cover Fix trace remotes read with an offline CPU
This small series is fixing non-consuming read of a trace remote when the trace_buffer is created after a CPU is offline.
It also extends hotplug testing coverage to include this test case.
I have based this series on top of kvmarm/next which contains the hypervisor tracing patches.
Changes in v2:
* Add a trap to restore offlined CPU on test failure in hotplug.tc.
* Fix assert_loaded/assert_unloaded error propagation.
* Ensure we probe for existing events when setting up consuming read iterator.
v1: https://lore.kernel.org/all/[email protected]/
Vincent Donnefort (2):
tracing: Non-consuming read for trace remotes with an offline CPU
tracing: selftests: Extend hotplug testing for trace remotes
kernel/trace/trace_remote.c | 22 ++++- .../selftests/ftrace/test.d/remotes/functions | 19 +++- .../ftrace/test.d/remotes/hotplug.tc | 88 +++++++++++++++++++ .../test.d/remotes/hypervisor/hotplug.tc | 11 +++ .../selftests/ftrace/test.d/remotes/trace.tc | 27 +----- .../ftrace/test.d/remotes/trace_pipe.tc | 25 ------ 6 files changed, 134 insertions(+), 58 deletions(-) create mode 100644 tools/testing/selftests/ftrace/test.d/remotes/hotplug.tc create mode 100644 tools/testing/selftests/ftrace/test.d/remotes/hypervisor/hotplug.tc
base-commit: 5ad2ff071b5980f072a85c8114649218971c586e
Signature · 1 line
2.53.0.1118.gaef5881109-goog
PATCH v2 1/2 tracing: Non-consuming read for trace remotes with an offline CPU
When a trace_buffer is created while a CPU is offline, this CPU is cleared from the trace_buffer CPU mask, preventing the creation of a non-consuming iterator (ring_buffer_iter). For trace remotes, it means the iterator fails to be allocated (-ENOMEM) even though there are available ring buffers in the trace_buffer.
For non-consuming reads of trace remotes, skip missing ring_buffer_iter to allow reading the available ring buffers.
kernel/trace/trace_remote.c +19 -3
@@ -282,6 +282,14 @@ static void trace_remote_put(struct trace_remote *remote)trace_remote_try_unload (remote); } +static bool trace_remote_has_cpu (struct trace_remote *remote,int cpu) +{ +if (cpu ==RING_BUFFER_ALL_CPUS ) +return true; + +return ring_buffer_poll_remote (remote->trace_buffer, cpu) ==0 ; +} +static void __poll_remote (struct work_struct *work) {struct delayed_work *dwork =to_delayed_work (work);
@@ -324,6 +332,10 @@ static int __alloc_ring_buffer_iter(struct trace_remote_iterator *iter, int cpu) iter->rb_iters[cpu] =ring_buffer_read_start (iter->remote->trace_buffer, cpu,GFP_KERNEL );if (!iter->rb_iters[cpu]) { +/* This CPU isn't part of trace_buffer. Skip it */ +if (!trace_remote_has_cpu (iter->remote, cpu)) +continue ; +__free_ring_buffer_iter (iter,RING_BUFFER_ALL_CPUS );return -ENOMEM ; }
@@ -347,10 +359,10 @@ static struct trace_remote_iteratorif (ret)return ERR_PTR (ret); -/* Test the CPU */ - ret =ring_buffer_poll_remote (remote->trace_buffer, cpu); -if (ret) +if (!trace_remote_has_cpu (remote, cpu)) { + ret = -ENODEV ; goto err; + } iter =kzalloc_obj (*iter);if (iter) {
@@ -361,6 +373,7 @@ static struct trace_remote_iteratorswitch (type) {case TRI_CONSUMING : +ring_buffer_poll_remote (remote->trace_buffer, cpu);INIT_DELAYED_WORK (&iter->poll_work, __poll_remote);schedule_delayed_work (&iter->poll_work,msecs_to_jiffies (remote->poll_ms));break ;
@@ -476,6 +489,9 @@ __peek_event(struct trace_remote_iterator *iter, int cpu, u64 *ts, unsigned longreturn ring_buffer_peek (iter->remote->trace_buffer, cpu, ts, lost_events);case TRI_NONCONSUMING : rb_iter =__get_rb_iter (iter, cpu); +if (!rb_iter) +return NULL ; + rb_evt =ring_buffer_iter_peek (rb_iter, ts);if (!rb_evt)return NULL ;
Signature · 1 line
2.53.0.1118.gaef5881109-goog
PATCH v2 2/2 tracing: selftests: Extend hotplug testing for trace remotes
In reply to Vincent DonnefortThe hotplug testing only tries reading a trace remote buffer, loaded before a CPU is offline. Extend this testing to cover:
* A trace remote buffer loaded after a CPU is offline.
* A trace remote buffer loaded before a CPU is online.
Because of these added test cases, move the hotplug testing into a separate hotplug.tc file.
tools/testing/selftests/ftrace/test.d/remotes/functions +15 -4
@@ -24,12 +24,21 @@ setup_remote_test() assert_loaded() { - grep -q "(loaded)" buffer_size_kb + grep -q "(loaded)" buffer_size_kb || return 1 } assert_unloaded() { - grep -q "(unloaded)" buffer_size_kb + grep -q "(unloaded)" buffer_size_kb || return 1 +} + +reload_remote() +{ + echo 0 > tracing_on + clear_trace + assert_unloaded + echo 1 > tracing_on + assert_loaded } dump_trace_pipe()
@@ -79,10 +88,12 @@ get_cpu_ids() sed -n 's/^processor\s*:\s*\([0-9]\+\).*/\1/p' /proc/cpuinfo } -get_page_size() { +get_page_size() +{ sed -ne 's/^.*data.*size:\([0-9][0-9]*\).*/\1/p' events/header_page } -get_selftest_event_size() { +get_selftest_event_size() +{ sed -ne 's/^.*field:.*;.*size:\([0-9][0-9]*\);.*/\1/p' events/*/selftest/format | awk '{s+=$1} END {print s}' }
tools/testing/selftests/ftrace/test.d/remotes/hotplug.tc new file +88 -0
88 lines not on this page · raw message
tools/testing/selftests/ftrace/test.d/remotes/hypervisor/hotplug.tc new file +11 -0
@@ -0,0 +1,11 @@ +#!/bin/sh +# SPDX-License-Identifier: GPL-2.0 +# description: Test hypervisor trace read with an offline CPU +# requires: remotes/hypervisor/write_event + +SOURCE_REMOTE_TEST=1 +. $TEST_DIR/remotes/hotplug.tc + +set -e +setup_remote "hypervisor" +test_hotplug
tools/testing/selftests/ftrace/test.d/remotes/trace.tc +1 -26
@@ -58,11 +58,7 @@ test_trace() # # Ensure the writer is not on the reader page by reloading the buffer - echo 0 > tracing_on - echo 0 > trace - assert_unloaded - echo 1 > tracing_on - assert_loaded + reload_remote # Ensure ring-buffer overflow by emitting events from the same CPU for cpu in $(get_cpu_ids); do
@@ -96,27 +92,6 @@ test_trace() cd - > /dev/null done - - # - # Test with hotplug - # - - [ "$(get_cpu_ids | wc -l)" -ge 2 ] || return 0 - - echo 0 > trace - - for cpu in $(get_cpu_ids); do - echo 0 > /sys/devices/system/cpu/cpu$cpu/online || return 0 - break - done - - for i in $(seq 1 8); do - echo $i > write_event - done - - check_trace 1 8 trace - - echo 1 > /sys/devices/system/cpu/cpu$cpu/online } if [ -z "$SOURCE_REMOTE_TEST" ]; then
tools/testing/selftests/ftrace/test.d/remotes/trace_pipe.tc +0 -25
@@ -92,31 +92,6 @@ test_trace_pipe() rm $output cd - > /dev/null done - - # - # Test interaction with hotplug - # - - [ "$(get_cpu_ids | wc -l)" -ge 2 ] || return 0 - - echo 0 > trace - - for cpu in $(get_cpu_ids); do - echo 0 > /sys/devices/system/cpu/cpu$cpu/online || return 0 - break - done - - for i in $(seq 1 8); do - echo $i > write_event - done - - output=$(dump_trace_pipe) - - check_trace 1 8 $output - - rm $output - - echo 1 > /sys/devices/system/cpu/cpu$cpu/online } if [ -z "$SOURCE_REMOTE_TEST" ]; then
Signature · 1 line
2.53.0.1118.gaef5881109-goog
Marc Zyngier
AppliedQuoted 9 lines · Vincent Donnefort
This small series is fixing non-consuming read of a trace remote when the
trace_bufferis created after a CPU is offline.It also extends hotplug testing coverage to include this test case.
I have based this series on top of kvmarm/next which contains the hypervisor tracing patches.
[...]
Applied to next, thanks!
[1/2] tracing: Non-consuming read for trace remotes with an offline CPU
commit: ce47b798ed1e44a6ae2c2966cdf7cba6b428083e
[2/2] tracing: selftests: Extend hotplug testing for trace remotes
commit: ec07906bdc52848bd7dc93d1d44e642dcdc7a15aCheers,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.