[PATCH 0/2] MMC: disable multiblock reads on controllers that don't support them

[PATCH 0/2] MMC: disable multiblock reads on controllers that don't support them

[Paul Walmsley]

Hello,

Some MMC controller instances integrated on older OMAP34xx/35xx chips
don't work correctly with multiple-block reads, due to a hardware bug.
This series fixes the problem by adding an MMC capability flag for this
case and making the appropriate changes to the MMC core code and OMAP HSMMC
driver to support this.

Once this series is merged, a subsequent patch will be needed by us OMAP
folks to pass the appropriate device data via hwmod - this will be posted
separately.


- Paul

---
mmc_errata_2.1.1.128_fix_3.2
   text	   data	    bss	    dec	    hex	filename
6325283	 655708	5591124	12572115	 bfd5d3	vmlinux.omap2plus_defconfig.orig
6329423	 655700	5591124	12576247	 bfe5f7	vmlinux.omap2plus_defconfig.patched

Paul Walmsley (2):
      MMC: add workaround for controllers with broken multiblock reads
      MMC: omap_hsmmc: disable multiblock reads when platform_data indicates that they are broken


 arch/arm/plat-omap/include/plat/mmc.h |   19 ++++++++++++++++++-
 drivers/mmc/card/block.c              |   21 ++++++++++++++-------
 drivers/mmc/host/omap_hsmmc.c         |    4 ++++
 include/linux/mmc/host.h              |    1 +
 4 files changed, 37 insertions(+), 8 deletions(-)

[PATCH 1/2] MMC: add workaround for controllers with broken multiblock reads

[Paul Walmsley]

Due to hardware bugs, some MMC host controllers don't support
multiple-block reads[1].  To resolve, add a new MMC capability flag,
MMC_CAP_NO_MULTI_READ, which can be set by affected host controller
drivers.  When this capability is set, all reads will be issued one
sector at a time.

1. See for example Advisory 2.1.1.128 "MMC: Multiple Block Read
Operation Issue" in _OMAP3530/3525/3515/3503 Silicon Errata_
Revision F (October 2010) (SPRZ278F), available from
http://focus.ti.com/lit/er/sprz278f/sprz278f.pdf

Signed-off-by: Paul Walmsley <paul@pwsan.com>
Cc: Dave Hylands <dhylands@gmail.com>
Cc: Steve Sakoman <sakoman@gmail.com>
Cc: Chris Ball <cjb@laptop.org>
---
 drivers/mmc/card/block.c |   21 ++++++++++++++-------
 include/linux/mmc/host.h |    1 +
 2 files changed, 15 insertions(+), 7 deletions(-)

diff --git a/drivers/mmc/card/block.c b/drivers/mmc/card/block.c
index 1ff5486..927f9d5 100644
--- a/drivers/mmc/card/block.c
+++ b/drivers/mmc/card/block.c
@@ -954,13 +954,20 @@ static void mmc_blk_rw_rq_prep(struct mmc_queue_req *mqrq,
 	if (brq->data.blocks > card->host->max_blk_count)
 		brq->data.blocks = card->host->max_blk_count;
 
-	/*
-	 * After a read error, we redo the request one sector at a time
-	 * in order to accurately determine which sectors can be read
-	 * successfully.
-	 */
-	if (disable_multi && brq->data.blocks > 1)
-		brq->data.blocks = 1;
+	if (brq->data.blocks > 1) {
+		/*
+		 * After a read error, we redo the request one sector
+		 * at a time in order to accurately determine which
+		 * sectors can be read successfully.
+		 */
+		if (disable_multi)
+			brq->data.blocks = 1;
+
+		/* Some controllers can't do multiblock reads due to hw bugs */
+		if (card->host->caps & MMC_CAP_NO_MULTI_READ &&
+		    rq_data_dir(req) == READ)
+			brq->data.blocks = 1;
+	}
 
 	if (brq->data.blocks > 1 || do_rel_wr) {
 		/* SPI multiblock writes terminate using a special
diff --git a/include/linux/mmc/host.h b/include/linux/mmc/host.h
index 1d09562..745f506 100644
--- a/include/linux/mmc/host.h
+++ b/include/linux/mmc/host.h
@@ -229,6 +229,7 @@ struct mmc_host {
 #define MMC_CAP_MAX_CURRENT_600	(1 << 28)	/* Host max current limit is 600mA */
 #define MMC_CAP_MAX_CURRENT_800	(1 << 29)	/* Host max current limit is 800mA */
 #define MMC_CAP_CMD23		(1 << 30)	/* CMD23 supported. */
+#define MMC_CAP_NO_MULTI_READ	(1 << 31)	/* Multiblock reads don't work */
 
 	mmc_pm_flag_t		pm_caps;	/* supported pm features */
 

[PATCH 2/2] MMC: omap_hsmmc: disable multiblock reads when platform_data indicates that they are broken

[Paul Walmsley]

When device data indicates that multiple block reads are not supported on
a given HSMMC controller instance, log a message to the console, and
pass the appropriate MMC capability flag to the MMC core.

Signed-off-by: Paul Walmsley <paul@pwsan.com>
Cc: Dave Hylands <dhylands@gmail.com>
Cc: Steve Sakoman <sakoman@gmail.com>
Cc: Chris Ball <cjb@laptop.org>
---
 arch/arm/plat-omap/include/plat/mmc.h |   19 ++++++++++++++++++-
 drivers/mmc/host/omap_hsmmc.c         |    4 ++++
 2 files changed, 22 insertions(+), 1 deletions(-)

diff --git a/arch/arm/plat-omap/include/plat/mmc.h b/arch/arm/plat-omap/include/plat/mmc.h
index c7b8741..94cf70a 100644
--- a/arch/arm/plat-omap/include/plat/mmc.h
+++ b/arch/arm/plat-omap/include/plat/mmc.h
@@ -31,7 +31,24 @@
 
 #define OMAP_MMC_MAX_SLOTS	2
 
-#define OMAP_HSMMC_SUPPORTS_DUAL_VOLT	BIT(1)
+/*
+ * struct omap_mmc_dev_attr.flags possibilities
+ *
+ * OMAP_HSMMC_SUPPORTS_DUAL_VOLT: Some HSMMC controller instances can
+ *    operate with either 1.8Vdc or 3.0Vdc card voltages; this flag
+ *    should be set if this is the case.  See for example Section 22.5.3
+ *    "MMC/SD/SDIO1 Bus Voltage Selection" of the OMAP34xx Multimedia
+ *    Device Silicon Revision 3.1.x Revision ZR (July 2011) (SWPU223R).
+ *
+ * OMAP_HSMMC_BROKEN_MULTIBLOCK_READ: Multiple-block read transfers
+ *    don't work correctly on some MMC controller instances on some
+ *    OMAP3 SoCs; this flag should be set if this is the case.  See
+ *    for example Advisory 2.1.1.128 "MMC: Multiple Block Read
+ *    Operation Issue" in _OMAP3530/3525/3515/3503 Silicon Errata_
+ *    Revision F (October 2010) (SPRZ278F).
+ */
+#define OMAP_HSMMC_SUPPORTS_DUAL_VOLT		BIT(0)
+#define OMAP_HSMMC_BROKEN_MULTIBLOCK_READ	BIT(1)
 
 struct omap_mmc_dev_attr {
 	u8 flags;
diff --git a/drivers/mmc/host/omap_hsmmc.c b/drivers/mmc/host/omap_hsmmc.c
index 21e4a79..9e8fc75 100644
--- a/drivers/mmc/host/omap_hsmmc.c
+++ b/drivers/mmc/host/omap_hsmmc.c
@@ -1943,6 +1943,10 @@ static int __init omap_hsmmc_probe(struct platform_device *pdev)
 	omap_hsmmc_context_save(host);
 
 	mmc->caps |= MMC_CAP_DISABLE;
+	if (host->pdata->controller_flags & OMAP_HSMMC_BROKEN_MULTIBLOCK_READ) {
+		dev_info(&pdev->dev, "multiblock reads disabled due to 35xx erratum 2.1.1.128; MMC read performance may suffer\n");
+		mmc->caps |= MMC_CAP_NO_MULTI_READ;
+	}
 
 	pm_runtime_enable(host->dev);
 	pm_runtime_get_sync(host->dev);

[PATCH 0/2] MMC: disable multiblock reads on controllers that don't support them

[Steve Sakoman]

On Thu, Oct 6, 2011 at 1:50 PM, Paul Walmsley <paul@pwsan.com> wrote:
> Hello,
>
> Some MMC controller instances integrated on older OMAP34xx/35xx chips
> don't work correctly with multiple-block reads, due to a hardware bug.
> This series fixes the problem by adding an MMC capability flag for this
> case and making the appropriate changes to the MMC core code and OMAP HSMMC
> driver to support this.
>
> Once this series is merged, a subsequent patch will be needed by us OMAP
> folks to pass the appropriate device data via hwmod - this will be posted
> separately.

I tested these patches with my own version of the above mentioned
hwmod patch on:

3503 - ES2.1 and ES3.1
3530 - ES3.1
3703 - ES1.2
3730 - ES1.0

All worked successfully.

Tested-by: Steve Sakoman <steve@sakoman.com>

[PATCH 0/2] MMC: disable multiblock reads on controllers that don't support them

[Chris Ball]

Hi Paul,

On Thu, Oct 06 2011, Paul Walmsley wrote:
> Some MMC controller instances integrated on older OMAP34xx/35xx chips
> don't work correctly with multiple-block reads, due to a hardware bug.
> This series fixes the problem by adding an MMC capability flag for this
> case and making the appropriate changes to the MMC core code and OMAP HSMMC
> driver to support this.
>
> Once this series is merged, a subsequent patch will be needed by us OMAP
> folks to pass the appropriate device data via hwmod - this will be posted
> separately.

Thanks, both pushed to mmc-next for 3.2.

We've run out of MMC caps nad moved onto CAPS2, so I modified these patches
to define and test MMC_CAP2_NO_MULTI_READ instead of MMC_CAP_NO_MULTI_READ:

  http://dev.laptop.org/git/users/cjb/mmc/commit/?h=mmc-next&id=a4b35b6acf2e9edc52132d34015f0a90b02e2df2
  http://dev.laptop.org/git/users/cjb/mmc/commit/?h=mmc-next&id=6d621423128909f09072835445ce36dd357a758a

(This shouldn't affect your hwmod patch, since that just uses an
OMAP-internal flag.)

- Chris.
-- 
Chris Ball   <cjb@laptop.org>   <http://printf.net/>
One Laptop Per Child