lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <1484478758-13133-1-git-send-email-hofrat@osadl.org>
Date:   Sun, 15 Jan 2017 12:12:38 +0100
From:   Nicholas Mc Guire <hofrat@...dl.org>
To:     Dmitry Torokhov <dmitry.torokhov@...il.com>
Cc:     Henrik Rydberg <rydberg@...math.org>, Dudley Du <dudl@...ress.com>,
        linux-input@...r.kernel.org, linux-kernel@...r.kernel.org,
        Nicholas Mc Guire <hofrat@...dl.org>
Subject: [PATCH RFC V2] Input: cyapa: use time based retry loop

Using counter based retry loops for peripherals results in the delay
being significantly overrun during high-load situations where delay
functions tend to be vary imprecise and overrun there timeouts. So
condition the termination on the actual condition of 2s for the 
re-calibration to have been successful.

As this is a very long delay there is no advantage in using 
high-resolution timers thus switching this to msleep().

Signed-off-by: Nicholas Mc Guire <hofrat@...dl.org>
---

V2: The maximum recalibration timeout should have been 2 * HZ as
    Dmitry Torokhov <dmitry.torokhov@...il.com> noted - V2 sets
    the timeout boundary to 2s now.

Problem detected with coccinelle script (usleep_range for long delay)

The basic model for the fix up here is to use

  timeout = jiffies + (retry_time)
  do {
      check()
      mleep()
  } while (time_is_after_jiffies(timeout));

rather than using a counter and have very long retry loops under load
conditions. In the specific case it is sleep()/check() though to wait
on the previous write to complete. The initial intent was to move the
usleep_range() to msleep() as the delays are very long - if this
refactoring of the timeout loop is desirable or not is not clear (both
time and count based models seem to be common).

Patch was compile tested with: x86_64_defconfig + CONFIG_MOUSE_CYAPA=m

Patch is against 4.10-rc3 (localversion-next is next-20170113)

 drivers/input/mouse/cyapa_gen3.c | 27 ++++++++++++---------------
 1 file changed, 12 insertions(+), 15 deletions(-)

diff --git a/drivers/input/mouse/cyapa_gen3.c b/drivers/input/mouse/cyapa_gen3.c
index f9600753..0c544de 100644
--- a/drivers/input/mouse/cyapa_gen3.c
+++ b/drivers/input/mouse/cyapa_gen3.c
@@ -789,7 +789,7 @@ static ssize_t cyapa_gen3_do_calibrate(struct device *dev,
 				     const char *buf, size_t count)
 {
 	struct cyapa *cyapa = dev_get_drvdata(dev);
-	int tries;
+	unsigned long timeout;
 	int ret;
 
 	ret = cyapa_read_byte(cyapa, CYAPA_CMD_DEV_STATUS);
@@ -812,31 +812,28 @@ static ssize_t cyapa_gen3_do_calibrate(struct device *dev,
 		goto out;
 	}
 
-	tries = 20;  /* max recalibration timeout 2s. */
+	/* max recalibration timeout 2s. */
+	timeout = jiffies + 2 * HZ;
 	do {
 		/*
 		 * For this recalibration, the max time will not exceed 2s.
 		 * The average time is approximately 500 - 700 ms, and we
 		 * will check the status every 100 - 200ms.
 		 */
-		usleep_range(100000, 200000);
-
+		msleep(100);
 		ret = cyapa_read_byte(cyapa, CYAPA_CMD_DEV_STATUS);
 		if (ret < 0) {
-			dev_err(dev, "Error reading dev status: %d\n",
-				ret);
+			dev_err(dev, "Error reading dev status: %d\n", ret);
 			goto out;
 		}
-		if ((ret & CYAPA_DEV_NORMAL) == CYAPA_DEV_NORMAL)
-			break;
-	} while (--tries);
+		if ((ret & CYAPA_DEV_NORMAL) == CYAPA_DEV_NORMAL) {
+			dev_dbg(dev, "Calibration successful.\n");
+			goto out;
+		}
+	} while (time_is_after_jiffies(timeout));
 
-	if (tries == 0) {
-		dev_err(dev, "Failed to calibrate. Timeout.\n");
-		ret = -ETIMEDOUT;
-		goto out;
-	}
-	dev_dbg(dev, "Calibration successful.\n");
+	dev_err(dev, "Failed to calibrate. Timeout.\n");
+	ret = -ETIMEDOUT;
 
 out:
 	return ret < 0 ? ret : count;
-- 
2.1.4

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ