mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Maxime Jayat <maxime.jayat@mobile-devices.fr>
To: Ludovic Desroches <ludovic.desroches@microchip.com>,
	Vinod Koul <vinod.koul@intel.com>
Cc: Nicolas Ferre <nicolas.ferre@microchip.com>,
	Dan Williams <dan.j.williams@intel.com>,
	dmaengine@vger.kernel.org, linux-kernel@vger.kernel.org,
	Maxime Jayat <maxime.jayat@mobile-devices.fr>,
	stable@vger.kernel.org
Subject: [PATCH] dmaengine: at_xdmac: fix rare residue corruption
Date: Thu, 22 Feb 2018 12:39:55 +0100	[thread overview]
Message-ID: <20180222113955.4043-1-maxime.jayat@mobile-devices.fr> (raw)

Despite the efforts made to correctly read the NDA and CUBC registers,
the order in which the registers are read could sometimes lead to an
inconsistent state.

Re-using the timeline from the comments, this following timing of
registers reads could lead to reading NDA with value "@desc2" and
CUBC with value "MAX desc1":

 INITD --------                    ------------
              |____________________|
       _______________________  _______________
 NDA       @desc2             \/   @desc3
       _______________________/\_______________
       __________  ___________  _______________
 CUBC       0    \/ MAX desc1 \/  MAX desc2
       __________/\___________/\_______________
        |  |          |  |
Events:(1)(2)        (3)(4)

(1) check_nda = @desc2
(2) initd = 1
(3) cur_ubc = MAX desc1
(4) cur_nda = @desc2

This is allowed by the condition ((check_nda == cur_nda) && initd),
despite cur_ubc and cur_nda being in the precise state we don't want.

This error leads to incorrect residue computation.

Fix it by inversing the order in which CUBC and INITD are read. This
makes sure that NDA and CUBC are always read together either _before_
INITD goes to 0 or _after_ it is back at 1.
The case where NDA is read before INITD is at 0 and CUBC is read after
INITD is back at 1 will be rejected by check_nda and cur_nda being
different.

Fixes: 53398f488821 ("dmaengine: at_xdmac: fix residue corruption")
Cc: stable@vger.kernel.org
Signed-off-by: Maxime Jayat <maxime.jayat@mobile-devices.fr>
---
Hi,

I had a bug where the serial ports on the Atmel SAMA5D2 were sometimes
returning the same data twice, for up to 4096 bytes.

After investigation, I noticed that the ring buffer used in
atmel_serial (in rx dma mode) had sometimes a incorrect "head" value,
which made the ring buffer do a complete extraneous loop of data
pushed to the tty layer.

I tracked it down to the residue of the dma being wrong, and after
more head scratching, I found this bug in the reading of the
registers.

Before fixing this, I was able to reproduce the bug reliably in a few
minutes. With this patch applied, the bug did not reappear after
several hours in testing.


 drivers/dma/at_xdmac.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/dma/at_xdmac.c b/drivers/dma/at_xdmac.c
index c00e3923d7d8..94236ec9d410 100644
--- a/drivers/dma/at_xdmac.c
+++ b/drivers/dma/at_xdmac.c
@@ -1471,10 +1471,10 @@ at_xdmac_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
 	for (retry = 0; retry < AT_XDMAC_RESIDUE_MAX_RETRIES; retry++) {
 		check_nda = at_xdmac_chan_read(atchan, AT_XDMAC_CNDA) & 0xfffffffc;
 		rmb();
-		initd = !!(at_xdmac_chan_read(atchan, AT_XDMAC_CC) & AT_XDMAC_CC_INITD);
-		rmb();
 		cur_ubc = at_xdmac_chan_read(atchan, AT_XDMAC_CUBC);
 		rmb();
+		initd = !!(at_xdmac_chan_read(atchan, AT_XDMAC_CC) & AT_XDMAC_CC_INITD);
+		rmb();
 		cur_nda = at_xdmac_chan_read(atchan, AT_XDMAC_CNDA) & 0xfffffffc;
 		rmb();
 
-- 
2.14.1

             reply	other threads:[~2018-02-22 11:40 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-02-22 11:39 Maxime Jayat [this message]
2018-03-01  8:25 ` Ludovic Desroches
2018-03-19  7:56   ` Vinod Koul
2018-03-19  7:59     ` Ludovic Desroches
2018-03-27 12:04 ` Vinod Koul

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=20180222113955.4043-1-maxime.jayat@mobile-devices.fr \
    --to=maxime.jayat@mobile-devices.fr \
    --cc=dan.j.williams@intel.com \
    --cc=dmaengine@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=ludovic.desroches@microchip.com \
    --cc=nicolas.ferre@microchip.com \
    --cc=stable@vger.kernel.org \
    --cc=vinod.koul@intel.com \
    /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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®