blktrace: add blktrace ftrace corruption regression test

Add regression test for blktrace ftrace corruption bug that occurs when
sysfs trace is enabled followed by ftrace blk tracer.

When /sys/block/*/trace/enable is enabled and then ftrace's blk tracer
is activated, the trace output becomes corrupted showing "Unknown action"
with invalid hex values instead of proper action codes.

The root cause is that ftrace allocates a blk_io_trace2 buffer (64 bytes)
but calls record_blktrace_event() which writes v1 format (48 bytes),
causing field offset mismatches and corruption.

This test verifies that the trace output is correct and doesn't show
the corruption pattern.

This test confirms the fix by the kernel patch:

commit e48886b9d668d80be24e37345bd0904e9138473c
Author: Chaitanya Kulkarni <[email protected]>
Date:   Mon Oct 27 22:50:42 2025 -0700

    blktrace: for ftrace use correct trace format ver

Link: https://lore.kernel.org/linux-block/[email protected]/
Signed-off-by: Chaitanya Kulkarni <[email protected]>
Reviewed-by: Johannes Thumshirn <[email protected]>
Signed-off-by: Shin'ichiro Kawasaki <[email protected]>

Author: ChaitanayaKulkarni

tests/blktrace/002

@@ -0,0 +1,97 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-3.0+
+# Copyright (C) 2025 Chaitanya Kulkarni <[email protected]>
+#
+# Regression test for blktrace ftrace corruption bug when using sysfs
+# trace enable with ftrace blk tracer.
+#
+# Bug: When sysfs trace is enabled (/sys/block/*/trace/enable) and then
+# ftrace blk tracer is enabled, the trace output becomes corrupted showing
+# "Unknown action" with invalid hex values.
+#
+# Root cause: ftrace allocated blk_io_trace2 buffer (64 bytes) but called
+# record_blktrace_event() which writes v1 format (48 bytes), causing field
+# offset mismatches and corruption.
+
+. tests/blktrace/rc
+. common/null_blk
+
+DESCRIPTION="blktrace ftrace corruption with sysfs trace"
+QUICK=1
+
+requires() {
+	_have_null_blk
+	_have_tracefs
+}
+
+test() {
+	echo "Running ${TEST_NAME}"
+
+	local trace_dir="/sys/kernel/debug/tracing"
+	local device
+
+	# Initialize null_blk with one device
+	if ! _configure_null_blk nullb1 power=1; then
+		return 1
+	fi
+
+	device=/dev/nullb1
+
+	# Verify device exists
+	if [[ ! -b "$device" ]]; then
+		echo "Device $device not found"
+		_exit_null_blk
+		return 1
+	fi
+
+	# Clean up any previous trace state
+	echo 0 > "$trace_dir/tracing_on" 2>/dev/null
+	echo > "$trace_dir/trace" 2>/dev/null
+	echo nop > "$trace_dir/current_tracer" 2>/dev/null
+
+	# Enable sysfs trace for nullb1 (this triggers the bug path)
+	if [[ -f /sys/block/nullb1/trace/enable ]]; then
+		echo 1 > /sys/block/nullb1/trace/enable
+	else
+		echo "No sysfs trace support"
+		_exit_null_blk
+		return 1
+	fi
+
+	# Enable blk ftrace tracer
+	echo blk > "$trace_dir/current_tracer"
+	echo 1 > "$trace_dir/tracing_on"
+
+	# Generate some I/O
+	dd if="$device" of=/dev/null bs=4k count=10 iflag=direct >> "$FULL" 2>&1
+
+	# Stop tracing
+	echo 0 > "$trace_dir/tracing_on"
+
+	# Check trace output for corruption
+	# Get first 10 non-comment lines
+	local trace_output
+	trace_output=$(grep -v "^#" "$trace_dir/trace" | head -10)
+
+	if [[ -z "$trace_output" ]]; then
+		echo "No trace output captured"
+		_exit_null_blk
+		return 1
+	fi
+
+	# Check for "Unknown action" which indicates the bug
+	if echo "$trace_output" | grep -q "Unknown action"; then
+		echo "BUG: Trace corruption detected with 'Unknown action'"
+		echo "$trace_output" | head -5 >> "$FULL"
+	else
+		echo "Trace output looks correct"
+	fi
+
+	# Cleanup: disable sysfs trace
+	echo 0 > /sys/block/nullb1/trace/enable 2>/dev/null
+	echo nop > "$trace_dir/current_tracer" 2>/dev/null
+
+	_exit_null_blk
+
+	echo "Test complete"
+}

tests/blktrace/002.out

@@ -0,0 +1,3 @@
+Running blktrace/002
+Trace output looks correct
+Test complete