linux-kernel - [PATCH] staging: comedi: ni_mio_common: fix AO timer off-by-one regression

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>] [day] [month] [year] [list]

Message-Id: <20170630110218.10678-1-abbotti@mev.co.uk>
Date:   Fri, 30 Jun 2017 12:02:18 +0100
From:   Ian Abbott <abbotti@....co.uk>
To:     devel@...verdev.osuosl.org
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Ian Abbott <abbotti@....co.uk>,
        H Hartley Sweeten <hsweeten@...ionengravers.com>,
        linux-kernel@...r.kernel.org,
        Éric Piel <piel@...mic.com>,
        "Spencer E . Olson" <olsonse@...ch.edu>,
        "# 4 . 7+" <stable@...r.kernel.org>
Subject: [PATCH] staging: comedi: ni_mio_common: fix AO timer off-by-one regression

As reported by Éric Piel on the Comedi mailing list (see
<https://groups.google.com/forum/#!topic/comedi_list/ueZiR7vTLOU/discussion>),
the analog output asynchronous commands are running too fast with a
period 50 ns shorter than it should be.  This affects all boards with AO
command support that are supported by the "ni_pcimio", "ni_atmio", and
"ni_mio_cs" drivers.

This is a regression bug introduced by commit 080e6795cba3 ("staging:
comedi: ni_mio_common: Cleans up/clarifies ni_ao_cmd"), specifically,
this line in `ni_ao_cmd_set_update()`:

		/* following line: N-1 per STC */
		ni_stc_writel(dev, trigvar - 1, NISTC_AO_UI_LOADA_REG);

The `trigvar` variable value comes from a call to `ni_ns_to_timer()`
which converts a timer period in nanoseconds to a hardware divisor
value. The function already reduces the divisor by 1 as required by the
hardware, so the above line should not reduce it further by 1.  Fix it
by replacing `trigvar` by `trigvar - 1` in the above line, and remove
the misleading comment.

Reported-by: Éric Piel <piel@...mic.com>
Fixes: 080e6795cba3 ("staging: comedi: ni_mio_common: Cleans up/clarifies ni_ao_cmd")
Cc: Éric Piel <piel@...mic.com>
Cc: Spencer E. Olson <olsonse@...ch.edu>
Cc: <stable@...r.kernel.org> # 4.7+
Signed-off-by: Ian Abbott <abbotti@....co.uk>
---
The bug is also in 4.6.x, but this patch does not apply to it cleanly.
---
 drivers/staging/comedi/drivers/ni_mio_common.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/staging/comedi/drivers/ni_mio_common.c b/drivers/staging/comedi/drivers/ni_mio_common.c
index b2e382888981..2f7bfc1c59e5 100644
--- a/drivers/staging/comedi/drivers/ni_mio_common.c
+++ b/drivers/staging/comedi/drivers/ni_mio_common.c
@@ -3116,8 +3116,7 @@ static void ni_ao_cmd_set_update(struct comedi_device *dev,
 		/* following line: 2-1 per STC */
 		ni_stc_writel(dev, 1, NISTC_AO_UI_LOADA_REG);
 		ni_stc_writew(dev, NISTC_AO_CMD1_UI_LOAD, NISTC_AO_CMD1_REG);
-		/* following line: N-1 per STC */
-		ni_stc_writel(dev, trigvar - 1, NISTC_AO_UI_LOADA_REG);
+		ni_stc_writel(dev, trigvar, NISTC_AO_UI_LOADA_REG);
 	} else { /* TRIG_EXT */
 		/* FIXME:  assert scan_begin_arg != 0, ret failure otherwise */
 		devpriv->ao_cmd2  |= NISTC_AO_CMD2_BC_GATE_ENA;
-- 
2.11.0