linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Marco Elver <elver@google.com>
To: elver@google.com, paulmck@kernel.org
Cc: boqun.feng@gmail.com, mark.rutland@arm.com, will@kernel.org,
	glider@google.com, dvyukov@google.com,
	kasan-dev@googlegroups.com, linux-kernel@vger.kernel.org
Subject: [PATCH 4/7] kcsan: Reduce get_ctx() uses in kcsan_found_watchpoint()
Date: Mon,  7 Jun 2021 14:56:50 +0200	[thread overview]
Message-ID: <20210607125653.1388091-5-elver@google.com> (raw)
In-Reply-To: <20210607125653.1388091-1-elver@google.com>

There are a number get_ctx() calls that are close to each other, which
results in poor codegen (repeated preempt_count loads).

Specifically in kcsan_found_watchpoint() (even though it's a slow-path)
it is beneficial to keep the race-window small until the watchpoint has
actually been consumed to avoid missed opportunities to report a race.

Let's clean it up a bit before we add more code in
kcsan_found_watchpoint().

Signed-off-by: Marco Elver <elver@google.com>
---
 kernel/kcsan/core.c | 26 ++++++++++++++++----------
 1 file changed, 16 insertions(+), 10 deletions(-)

diff --git a/kernel/kcsan/core.c b/kernel/kcsan/core.c
index d92977ede7e1..906100923b88 100644
--- a/kernel/kcsan/core.c
+++ b/kernel/kcsan/core.c
@@ -301,9 +301,9 @@ static inline void reset_kcsan_skip(void)
 	this_cpu_write(kcsan_skip, skip_count);
 }
 
-static __always_inline bool kcsan_is_enabled(void)
+static __always_inline bool kcsan_is_enabled(struct kcsan_ctx *ctx)
 {
-	return READ_ONCE(kcsan_enabled) && get_ctx()->disable_count == 0;
+	return READ_ONCE(kcsan_enabled) && !ctx->disable_count;
 }
 
 /* Introduce delay depending on context and configuration. */
@@ -353,10 +353,17 @@ static noinline void kcsan_found_watchpoint(const volatile void *ptr,
 					    atomic_long_t *watchpoint,
 					    long encoded_watchpoint)
 {
+	struct kcsan_ctx *ctx = get_ctx();
 	unsigned long flags;
 	bool consumed;
 
-	if (!kcsan_is_enabled())
+	/*
+	 * We know a watchpoint exists. Let's try to keep the race-window
+	 * between here and finally consuming the watchpoint below as small as
+	 * possible -- avoid unneccessarily complex code until consumed.
+	 */
+
+	if (!kcsan_is_enabled(ctx))
 		return;
 
 	/*
@@ -364,14 +371,12 @@ static noinline void kcsan_found_watchpoint(const volatile void *ptr,
 	 * reporting a race where e.g. the writer set up the watchpoint, but the
 	 * reader has access_mask!=0, we have to ignore the found watchpoint.
 	 */
-	if (get_ctx()->access_mask != 0)
+	if (ctx->access_mask)
 		return;
 
 	/*
-	 * Consume the watchpoint as soon as possible, to minimize the chances
-	 * of !consumed. Consuming the watchpoint must always be guarded by
-	 * kcsan_is_enabled() check, as otherwise we might erroneously
-	 * triggering reports when disabled.
+	 * Consuming the watchpoint must be guarded by kcsan_is_enabled() to
+	 * avoid erroneously triggering reports if the context is disabled.
 	 */
 	consumed = try_consume_watchpoint(watchpoint, encoded_watchpoint);
 
@@ -409,6 +414,7 @@ kcsan_setup_watchpoint(const volatile void *ptr, size_t size, int type)
 	unsigned long access_mask;
 	enum kcsan_value_change value_change = KCSAN_VALUE_CHANGE_MAYBE;
 	unsigned long ua_flags = user_access_save();
+	struct kcsan_ctx *ctx = get_ctx();
 	unsigned long irq_flags = 0;
 
 	/*
@@ -417,7 +423,7 @@ kcsan_setup_watchpoint(const volatile void *ptr, size_t size, int type)
 	 */
 	reset_kcsan_skip();
 
-	if (!kcsan_is_enabled())
+	if (!kcsan_is_enabled(ctx))
 		goto out;
 
 	/*
@@ -489,7 +495,7 @@ kcsan_setup_watchpoint(const volatile void *ptr, size_t size, int type)
 	 * Re-read value, and check if it is as expected; if not, we infer a
 	 * racy access.
 	 */
-	access_mask = get_ctx()->access_mask;
+	access_mask = ctx->access_mask;
 	new = 0;
 	switch (size) {
 	case 1:
-- 
2.32.0.rc1.229.g3e70b5a671-goog


  parent reply	other threads:[~2021-06-07 12:57 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-06-07 12:56 [PATCH 0/7] kcsan: Introduce CONFIG_KCSAN_PERMISSIVE Marco Elver
2021-06-07 12:56 ` [PATCH 1/7] kcsan: Improve some Kconfig comments Marco Elver
2021-06-07 12:56 ` [PATCH 2/7] kcsan: Remove CONFIG_KCSAN_DEBUG Marco Elver
2021-06-07 12:56 ` [PATCH 3/7] kcsan: Introduce CONFIG_KCSAN_STRICT Marco Elver
2021-06-07 12:56 ` Marco Elver [this message]
2021-06-07 12:56 ` [PATCH 5/7] kcsan: Rework atomic.h into permissive.h Marco Elver
2021-06-07 12:56 ` [PATCH 6/7] kcsan: Print if strict or non-strict during init Marco Elver
2021-06-07 12:56 ` [PATCH 7/7] kcsan: permissive: Ignore data-racy 1-bit value changes Marco Elver
2021-06-09 12:38 ` [PATCH 0/7] kcsan: Introduce CONFIG_KCSAN_PERMISSIVE Mark Rutland
2021-06-09 14:48   ` Marco Elver
2021-06-15 18:19   ` Paul E. McKenney
2021-06-15 18:51     ` Marco Elver
2021-06-15 20:39       ` Paul E. McKenney

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210607125653.1388091-5-elver@google.com \
    --to=elver@google.com \
    --cc=boqun.feng@gmail.com \
    --cc=dvyukov@google.com \
    --cc=glider@google.com \
    --cc=kasan-dev@googlegroups.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mark.rutland@arm.com \
    --cc=paulmck@kernel.org \
    --cc=will@kernel.org \
    --subject='Re: [PATCH 4/7] kcsan: Reduce get_ctx() uses in kcsan_found_watchpoint()' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
on how to clone and mirror all data and code used for this inbox