linux-kernel - [PATCH] RFC: mmc: dw_mmc: Always go to STATE_DATA_BUSY from STATE_DATA

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]

Date:	Fri, 15 Mar 2013 14:29:16 -0700
From:	Doug Anderson <dianders@...omium.org>
To:	Chris Ball <cjb@...top.org>
Cc:	Will Newton <will.newton@...il.com>,
	Seungwon Jeon <tgih.jun@...sung.com>,
	Bing Zhao <bzhao@...vell.com>,
	Jaehoon Chung <jh80.chung@...sung.com>,
	Ashok Nagarajan <asnagarajan@...omium.org>,
	Paul Stewart <pstew@...omium.org>,
	Olof Johansson <olof@...om.net>,
	Doug Anderson <dianders@...omium.org>,
	linux-mmc@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: [PATCH] RFC: mmc: dw_mmc: Always go to STATE_DATA_BUSY from STATE_DATA_ERROR

On a flaky piece of hardware that seems good at generating CRC errors,
we have found that often times the CRC errors don't get reported
properly when using CONFIG_MMC_DW_IDMAC (they get reported OK when
using pio).

The flow that happens is:
1. dw_mci_interrupt() fires and status=80b8, pending=8088 so that
   we hit (pending & DW_MCI_DATA_ERROR_FLAGS).  We store 8088 in
   data_status and set EVENT_DATA_ERROR in host->pending_events
2. We schedule the tasklet and it runs.
3. We're in STATE_SENDING_DATA in the tasklet and see
   EVENT_DATA_ERROR so we dw_mci_stop_dma().
4. dw_mci_stop_dma() calls dw_mci_idmac_stop_dma() and
   dw_mci_dma_cleanup().  These stop dma but _don't_ set
   EVENT_XFER_COMPLETE (since we're host->using_dma).
5. data->stop is NULL so we don't send a stop command.
6. We move onto STATE_DATA_ERROR and loop again in the tasklet.
7. We hit STATE_DATA_ERROR but the transfer isn't done, so the tasklet
   stops.

We never seem to get any additional DMA interrupts that cause
EVENT_XFER_COMPLETE and restart the tasklet so we just hang.  That
doesn't seem surprising given that we've stopped DMA.

We did put a print at the end of dw_mci_interrupt() to show the result
of the "mci_readl(host, IDSTS)" and saw 0xa000 in the case of the
above CRC error.

A proposed fix for this is to ignore (but still clear) the
EVENT_XFER_COMPLETE in STATE_DATA_ERROR in the tasklet.

Reported-by: Bing Zhao <bzhao@...vell.com>
Signed-off-by: Doug Anderson <dianders@...omium.org>
---
 drivers/mmc/host/dw_mmc.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
index 9834221..696b3bb 100644
--- a/drivers/mmc/host/dw_mmc.c
+++ b/drivers/mmc/host/dw_mmc.c
@@ -1137,10 +1137,7 @@ static void dw_mci_tasklet_func(unsigned long priv)
 			goto unlock;

 		case STATE_DATA_ERROR:
-			if (!test_and_clear_bit(EVENT_XFER_COMPLETE,
-						&host->pending_events))
-				break;
-
+			clear_bit(EVENT_XFER_COMPLETE, &host->pending_events);
 			state = STATE_DATA_BUSY;
 			break;
 		}
-- 
1.8.1.3

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/