[PATCH v3 4/4] arm/dts: Add tps65217 regulator DT data to am335x-bone.dts

[mporter@ti.com (Matt Porter)]

On Wed, Sep 05, 2012 at 03:29:30PM +0200, Koen Kooi wrote:
> 
> Op 28 aug. 2012, om 07:34 heeft "AnilKumar, Chimata" <anilkumar@ti.com> het volgende geschreven:
> 
> > Hi Koen,
> > 
> > On Fri, Aug 24, 2012 at 13:32:17, Koen Kooi wrote:
> >> 
> >> Op 24 aug. 2012, om 09:56 heeft Koen Kooi <koen@dominion.thruhere.net> het volgende geschreven:
> >> 
> >>> 
> >>> Op 24 aug. 2012, om 09:26 heeft "AnilKumar, Chimata" <anilkumar@ti.com> het volgende geschreven:
> >>> 
> >>>> Hi Koen,
> >>>> 
> >>>> On Fri, Aug 24, 2012 at 11:58:34, Koen Kooi wrote:
> >>>>> 
> >>>>> Op 24 aug. 2012, om 07:50 heeft "AnilKumar, Chimata" <anilkumar@ti.com> het volgende geschreven:
> >>>>> 
> >>>>>> Hi Koen,
> >>>>>> 
> >>>>>> On Thu, Aug 23, 2012 at 19:43:48, Koen Kooi wrote:
> >>>>>>> 
> >>>>>>> Op 21 aug. 2012, om 13:17 heeft AnilKumar Ch <anilkumar@ti.com> het volgende geschreven:
> >>>>>>> 
> >>>>>>>> Add tps65217 regulator device tree data to AM335x-Bone by adding
> >>>>>>>> regulator consumers with tightened constraints and regulator-name.
> >>>>>>>> TPS65217 regulator handle can be obtained by using this regulator
> >>>>>>>> name.
> >>>>>>>> 
> >>>>>>>> This patch also add I2C node with I2C frequency and tps65217 PMIC
> >>>>>>>> I2C slave address.
> >>>>>>>> 
> >>>>>>>> Signed-off-by: AnilKumar Ch <anilkumar@ti.com>
> >>>>>>> 
> >>>>>>> I tried this and the kernel immediately crashes on my beaglebone. Could you upload the complete git tree and .config you used to test this to somewhere public please?
> >>>>>> 
> >>>>>> Use this repo to test on beaglebone
> >>>>>> https://github.com/hvaibhav/am335x-linux/commits/am335x-upstream-staging-pinctrl
> >>>>>> 
> >>>>>> This wiki talks about how to build and use?
> >>>>>> https://github.com/hvaibhav/am335x-linux/wiki/How-To-Use-Upstream-Tree
> >>>>>> 
> >>>>>> Note: Enable tps65217 regulator in kernel config.
> >>>>> 
> >>>>> I used that repo and as a seperate test I rebased that to latest mainline, same thing: as soon as I turn on the TPS in the .config it crashes on boot. Is the pinctrl repo the *exact* repo you used to test the patches on beaglebone?
> >>>> 
> >>>> I tested on latest mainline after merging to
> >>>> am335x-upstream-staging-pinctrl (voltage also changing)
> >>>> 
> >>>> Can you share your .config and uImage?
> >>> 
> >>> Config: https://github.com/beagleboard/kernel/blob/beaglebone-3.6/patches/configs/beaglebone
> >>> 
> >>>> My config details:- (After merge)
> >>>> 1. omap2plus_defconfig
> >>>> 2. Enable tps65217 MFD driver
> >>>> 3. Enable tps65217 regulator driver
> >>> 
> >>> 
> >>> I rebased onto latest mainline and refreshed the base patches from Vaibhav and I now get: 
> >>> 
> >>> [    0.246796] tps65217 0-0024: TPS65217 ID 0xf version 1.1
> >>> 
> >>> So it boots! I don't know what made it break before, but it's working now :)
> >> 
> >> *sigh* I'm an idiot:
> >> 
> >> root at beaglebone:~# uname -a
> >> Linux beaglebone 3.6.0-rc3-00103-gfd02083 #86 SMP Fri Aug 24 09:45:54 CEST 2012 armv7l GNU/Linux
> >> root at beaglebone:~# zcat /proc/config.gz | grep 217
> >> CONFIG_MFD_TPS65217=y
> >> # CONFIG_REGULATOR_TPS65217 is not set
> >> 
> >> Will retry with regulator driver actually turned on in a bit.
> > 
> > Is it working after enabling the regulator?
> 
> It took me a while to get back to this problem, but it still isn't working for me. I did manage to get more info on the error:
> 
> root at bone-mainline:~# insmod tps65217-regulator.ko
> [   32.754419] Unable to handle kernel NULL pointer dereference at virtual address 000000c8
> [   32.763087] pgd = cea60000
> [   32.765969] [000000c8] *pgd=8fbed831, *pte=00000000, *ppte=00000000
> [   32.772617] Internal error: Oops: 17 [#1] SMP THUMB2
> [   32.777827] Modules linked in: tps65217_regulator(+) ip_tables x_tables snd_soc_omap snd_soc_core regmap_spi snd_pcm snd_timer snd soundcore snd_page_alloc ipv6
> [   32.792976] CPU: 0    Not tainted  (3.6.0-rc4 #109)
> [   32.798106] PC is at regmap_read+0x8/0x38
> [   32.802315] LR is at regulator_get_voltage_sel_regmap+0x15/0x38

I just got to this same point last night as I needed this working to
test omap_hsmmc with edma dmaengine...

The problem is that the tps65217-regulator driver is handing the wrong
device to regulator_register() and it contains a null regmap. This patch
passes in the the parent dev to fix it. Maybe I missed another patch
that addresses this in a different way, such as the regulator devices
being stuffed with the regmap devres?

-Matt

>From 40d118bebc5eaf2a8df4f8b5e113b892a3210f96 Mon Sep 17 00:00:00 2001
From: Matt Porter <mporter@ti.com>
Date: Wed, 5 Sep 2012 10:09:50 -0400
Subject: [PATCH] regulator: tps65217: fix crash during registration

The struct device for each platform device bound to this
driver is a child to the parent mfd device. The parent device
is the one that actually contains the regmap devres so fix the
null pointer dereference during the first regmap access in
registration by passing in the parent device.

Signed-off-by: Matt Porter <mporter@ti.com>
---
 drivers/regulator/tps65217-regulator.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/regulator/tps65217-regulator.c b/drivers/regulator/tps65217-regulator.c
index 6caa222..af4916c 100644
--- a/drivers/regulator/tps65217-regulator.c
+++ b/drivers/regulator/tps65217-regulator.c
@@ -292,7 +292,7 @@ static int __devinit tps65217_regulator_probe(struct platform_device *pdev)
 	tps = dev_to_tps65217(pdev->dev.parent);
 	tps->info[pdev->id] = info;
 
-	config.dev = &pdev->dev;
+	config.dev = pdev->dev.parent;
 	config.of_node = pdev->dev.of_node;
 	config.init_data = pdev->dev.platform_data;
 	config.driver_data = tps;
-- 
1.7.9.5