Regression for non-eMMC cards with commit fa550189?

PATCH New Tony Lindgren 4 participants 8 messages May 28, 2012 – Jun 6, 2012

Subsystems (6)

Arm (32-bit) Arm sub-architectures MMC, SD and SDIO Multifunction devices (MFD) +2 more subsystems

lore

Share & tools

lore

Tony Lindgren

author maintainer

2012-05-28 13:21 UTC #0 lore raw

Hi Ulf & Chris,

Looks like commit fa550189 (mmc: core: Prevent eMMC VCC supply to be cut from late init) causes MMC card to stop working at least with n8x0 using drivers/mmc/host/omap.c. The card(s) on it are not eMMC.

Reverting fa550189 makes things work again. Any ideas what could be causing this?

No debug output from MMC_DEBUG, looks like commands won't even get started?

Regards,

Tony

Ulf Hansson

2012-05-29 06:52 UTC #1 lore raw

Hi Tony,

This patch changes the sequence of how the host drivers set_ios function gets called during probe.

Before:
1. mmc_power_off
2. mmc_power_up

Now:
1. mmc_power_up

My guess is then; the omap driver set_ios function requires that mmc_power_off is called before a mmc_power_up.
Hopefully this requirement can be removed and fixed in the host driver somehow.

Please get back to me if you need some more assistance around this matter.

Kind regards
Ulf Hansson

Quoted 15 lines · Tony Lindgren

Hi Ulf & Chris,
Looks like commit fa550189 (mmc: core: Prevent eMMC VCC supply to be cut from late init) causes MMC card to stop working at least with n8x0 using drivers/mmc/host/omap.c. The card(s) on it are not eMMC.
Reverting fa550189 makes things work again. Any ideas what could be causing this?
No debug output from MMC_DEBUG, looks like commands won't even get started?
Regards,
Tony

Felipe Balbi

2012-05-29 07:40 UTC #2 lore raw

Hi,

Quoted 16 lines · Ulf Hansson

Hi Tony,
This patch changes the sequence of how the host drivers set_ios function gets called during probe.
Before:
1. mmc_power_off
2. mmc_power_up
Now:
1. mmc_power_up
My guess is then; the omap driver set_ios function requires that mmc_power_off is called before a mmc_power_up.
Hopefully this requirement can be removed and fixed in the host driver somehow.

Quoted 18 lines

Please get back to me if you need some more assistance around this matter.
Kind regards
Ulf Hansson
Quoted 11 lines · Tony Lindgren
Hi Ulf & Chris,
Looks like commit fa550189 (mmc: core: Prevent eMMC VCC supply to be cut from late init) causes MMC card to stop working at least with n8x0 using drivers/mmc/host/omap.c. The card(s) on it are not eMMC.
Reverting fa550189 makes things work again. Any ideas what could be causing this?
No debug output from MMC_DEBUG, looks like commands won't even get started?

Can you enable debugging on menelaus so we see what the driver is doing ?

I believe that to change the VMMC's supply voltage you need to turn off the regulator, meaning LDO_CTRL7[1:0] = 0b00.

That could be one thing. It would be cool what menelaus is writing to the registers on working and failing case. Also what are the contents of LDO_CTRL1 and LDO_CTRL7 before any operation. Something like this will help:

drivers/mfd/menelaus.c +6 -0

index cb4910a..8cab234 100644
@@ -31,6 +31,8 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  */
 
+#define DEBUG
+
 #include <linux/module.h>
 #include <linux/i2c.h>
 #include <linux/interrupt.h>

@@ -469,6 +471,10 @@ static int menelaus_set_voltage(const struct menelaus_vtg *vtg, int mV,
 	if (vtg == 0)
 		goto set_voltage;
 
+	dev_dbg(&c->dev, "VTG_REG %02x MODE_REG %02x\n",
+			menelaus_read_reg(vtg->vtg_reg),
+			menelaus_read_reg(mode_reg));
+
 	ret = menelaus_read_reg(vtg->vtg_reg);
 	if (ret < 0)
 		goto out;

Signature · 8 lines

balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20120529/cd8579c6/attachment.sig>

Tony Lindgren

author maintainer

2012-05-29 12:47 UTC #3 lore raw

Quoted 5 lines · Felipe Balbi

Can you enable debugging on menelaus so we see what the driver is doing ?
I believe that to change the VMMC's supply voltage you need to turn off the regulator, meaning LDO_CTRL7[1:0] = 0b00.

Looks like it's some kind of race between the slots that now gets triggered in the omap driver.

Commenting out the second slot in the board file fixes things. I tried adding a mutex to protect the sequence between POWER_UP and POWER_ON, but no luck yet.

Also adding printk statements here and there makes it work again.

Regards,

Tony

Felipe Balbi

2012-05-29 12:50 UTC #4 lore raw

Hi,

Quoted 7 lines · Tony Lindgren

Quoted 5 lines · Felipe Balbi
Can you enable debugging on menelaus so we see what the driver is doing ?
I believe that to change the VMMC's supply voltage you need to turn off the regulator, meaning LDO_CTRL7[1:0] = 0b00.

Quoted 7 lines

Looks like it's some kind of race between the slots that now gets triggered in the omap driver.
Commenting out the second slot in the board file fixes things. I tried adding a mutex to protect the sequence between POWER_UP and POWER_ON, but no luck yet.
Also adding printk statements here and there makes it work again.

does this help ?

arch/arm/mach-omap2/board-n8x0.c +2 -2

index 518091c..5a224fc 100644
@@ -397,8 +397,8 @@ static int n8x0_mmc_late_init(struct device *dev)
 	if (r < 0)
 		return r;
 
-	n8x0_mmc_set_power(dev, 0, MMC_POWER_ON, 16); /* MMC_VDD_28_29 */
-	n8x0_mmc_set_power(dev, 1, MMC_POWER_ON, 16);
+//	n8x0_mmc_set_power(dev, 0, MMC_POWER_ON, 16); /* MMC_VDD_28_29 */
+//	n8x0_mmc_set_power(dev, 1, MMC_POWER_ON, 16);
 
 	r = menelaus_set_mmc_slot(1, 1, 0, 1);
 	if (r < 0)

Signature · 8 lines

balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20120529/3cde1dc1/attachment.sig>

Tony Lindgren

author maintainer

2012-05-29 13:03 UTC #5 lore raw

Quoted 13 lines · Felipe Balbi

Hi,
Quoted 10 lines · Tony Lindgren
Quoted 5 lines · Felipe Balbi
Can you enable debugging on menelaus so we see what the driver is doing ?
I believe that to change the VMMC's supply voltage you need to turn off the regulator, meaning LDO_CTRL7[1:0] = 0b00.
Looks like it's some kind of race between the slots that now gets triggered in the omap driver.

Quoted 6 lines

Quoted 4 lines
Commenting out the second slot in the board file fixes things. I tried adding a mutex to protect the sequence between POWER_UP and POWER_ON, but no luck yet.
Also adding printk statements here and there makes it work again.
does this help ?

Thanks does not seem to help.

Tony

Tony Lindgren

author maintainer

2012-05-30 07:13 UTC #6 lore raw

PATCH 1/1 mmc: omap: Fix broken reg_shift initialization

Commit fa550189 (mmc: core: Prevent eMMC VCC supply to be cut from late init) slightly affected timings for how things are done for the omap MMC driver causing the MMC cards not getting detected any longer.

Turns out this was caused by buggy reg_shift initialization in the omap MMC driver that was happening after mmc_add_host() was being called.

Fix this by initializing reg_shift before mmc_add_host() is called.

Signed-off-by Tony Lindgren <[email protected]>

---

Heh found it! Chris, maybe Cc stable on this one?

drivers/mmc/host/omap.c +1 -2

@@ -1485,6 +1485,7 @@ static int __devinit mmc_omap_probe(struct platform_device *pdev)
 	}
 
 	host->nr_slots = pdata->nr_slots;
+	host->reg_shift = (cpu_is_omap7xx() ? 1 : 2);
 
 	host->mmc_omap_wq = alloc_workqueue("mmc_omap", 0, 0);
 	if (!host->mmc_omap_wq)

@@ -1500,8 +1501,6 @@ static int __devinit mmc_omap_probe(struct platform_device *pdev)
 		}
 	}
 
-	host->reg_shift = (cpu_is_omap7xx() ? 1 : 2);
-
 	return 0;
 
 err_destroy_wq:

Chris Ball

2012-06-06 13:48 UTC #7 lore raw

PATCH 1/1 mmc: omap: Fix broken reg_shift initialization

Hi,

Quoted 7 lines · Tony Lindgren

Commit fa550189 (mmc: core: Prevent eMMC VCC supply to be cut from late init) slightly affected timings for how things are done for the omap MMC driver causing the MMC cards not getting detected any longer.
Turns out this was caused by buggy reg_shift initialization in the omap MMC driver that was happening after mmc_add_host() was being called.

Quoted 7 lines

Fix this by initializing reg_shift before mmc_add_host() is called.
Signed-off-by Tony Lindgren <[email protected]>
---
Heh found it! Chris, maybe Cc stable on this one?

Thanks, pushed to mmc-next for 3.5 with a stable@ tag added.

- Chris.

Signature · 2 lines

Chris Ball   <[email protected]>   <http://printf.net/>
One Laptop Per Child