Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Wed, 29 Jan 2020 16:01:02 +0100
Message-Id: <20200129150102.2122-1-elver@google.com>
Mime-Version: 1.0
Subject: [PATCH] kcsan: Address missing case with KCSAN_REPORT_VALUE_CHANGE_ONLY
From:   Marco Elver <elver@google.com>
To:     elver@google.com
Cc:     paulmck@kernel.org, andreyknvl@google.com, glider@google.com,
        dvyukov@google.com, kasan-dev@googlegroups.com,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

With KCSAN_REPORT_VALUE_CHANGE_ONLY, KCSAN has still been able to report
data races between reads and writes, if a watchpoint was set up on the
write. If the write rewrote the same value we'd still have reported the
data race. We now unconditionally skip reporting on this case.

Signed-off-by: Marco Elver <elver@google.com>
---
 kernel/kcsan/report.c | 27 ++++++++++++++++++++-------
 1 file changed, 20 insertions(+), 7 deletions(-)

diff --git a/kernel/kcsan/report.c b/kernel/kcsan/report.c
index 33bdf8b229b5..7cd34285df74 100644
--- a/kernel/kcsan/report.c
+++ b/kernel/kcsan/report.c
@@ -130,12 +130,25 @@ static bool rate_limit_report(unsigned long frame1, unsigned long frame2)
  * Special rules to skip reporting.
  */
 static bool
-skip_report(int access_type, bool value_change, unsigned long top_frame)
+skip_report(bool value_change, unsigned long top_frame)
 {
-	const bool is_write = (access_type & KCSAN_ACCESS_WRITE) != 0;
-
-	if (IS_ENABLED(CONFIG_KCSAN_REPORT_VALUE_CHANGE_ONLY) && is_write &&
-	    !value_change) {
+	/*
+	 * The first call to skip_report always has value_change==true, since we
+	 * cannot know the value written of an instrumented access. For the 2nd
+	 * call there are 6 cases with CONFIG_KCSAN_REPORT_VALUE_CHANGE_ONLY:
+	 *
+	 * 1. read watchpoint, conflicting write (value_change==true): report;
+	 * 2. read watchpoint, conflicting write (value_change==false): skip;
+	 * 3. write watchpoint, conflicting write (value_change==true): report;
+	 * 4. write watchpoint, conflicting write (value_change==false): skip;
+	 * 5. write watchpoint, conflicting read (value_change==false): skip;
+	 * 6. write watchpoint, conflicting read (value_change==true): impossible;
+	 *
+	 * Cases 1-4 are intuitive and expected; case 5 ensures we do not report
+	 * data races where the write may have rewritten the same value; and
+	 * case 6 is simply impossible.
+	 */
+	if (IS_ENABLED(CONFIG_KCSAN_REPORT_VALUE_CHANGE_ONLY) && !value_change) {
 		/*
 		 * The access is a write, but the data value did not change.
 		 *
@@ -228,7 +241,7 @@ static bool print_report(const volatile void *ptr, size_t size, int access_type,
 	/*
 	 * Must check report filter rules before starting to print.
 	 */
-	if (skip_report(access_type, true, stack_entries[skipnr]))
+	if (skip_report(true, stack_entries[skipnr]))
 		return false;
 
 	if (type == KCSAN_REPORT_RACE_SIGNAL) {
@@ -237,7 +250,7 @@ static bool print_report(const volatile void *ptr, size_t size, int access_type,
 		other_frame = other_info.stack_entries[other_skipnr];
 
 		/* @value_change is only known for the other thread */
-		if (skip_report(other_info.access_type, value_change, other_frame))
+		if (skip_report(value_change, other_frame))
 			return false;
 	}
 
-- 
2.25.0.341.g760bfbb309-goog