This fixes a possible kernel oops due to using stack allocated platform
data for the USB PHY driver on DA8XX devices. If the platform device
probe is deferred, then we get a corrupt pointer for the platform data.
We now use a global static struct for the platform data so that the
platform data pointer does not get written over.
Tested on OMAP-L138 LCDK board using legacy board file.
Fixes: 9b504750923cb ("ARM: davinci: da8xx: Add USB PHY platform device")
Signed-off-by: David Lechner <[email protected]>
---
arch/arm/mach-davinci/usb-da8xx.c | 11 +++++++----
1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/arch/arm/mach-davinci/usb-da8xx.c b/arch/arm/mach-davinci/usb-da8xx.c
index b1e53e31..9ff9624 100644
--- a/arch/arm/mach-davinci/usb-da8xx.c
+++ b/arch/arm/mach-davinci/usb-da8xx.c
@@ -23,17 +23,20 @@
#define DA8XX_USB0_BASE 0x01e00000
#define DA8XX_USB1_BASE 0x01e25000
+static struct da8xx_usb_phy_platform_data da8xx_usb_phy_pdata;
+
static struct platform_device da8xx_usb_phy = {
.name = "da8xx-usb-phy",
.id = -1,
+ .dev = {
+ .platform_data = &da8xx_usb_phy_pdata,
+ },
};
+
int __init da8xx_register_usb_phy(void)
{
- struct da8xx_usb_phy_platform_data pdata;
-
- pdata.cfgchip = da8xx_get_cfgchip();
- da8xx_usb_phy.dev.platform_data = &pdata;
+ da8xx_usb_phy_pdata.cfgchip = da8xx_get_cfgchip();
return platform_device_register(&da8xx_usb_phy);
}
--
2.7.4
On 03/15/2018 01:02 PM, David Lechner wrote:
> This fixes a possible kernel oops due to using stack allocated platform
> data for the USB PHY driver on DA8XX devices. If the platform device
> probe is deferred, then we get a corrupt pointer for the platform data.
>
> We now use a global static struct for the platform data so that the
> platform data pointer does not get written over.
>
> Tested on OMAP-L138 LCDK board using legacy board file.
>
> Fixes: 9b504750923cb ("ARM: davinci: da8xx: Add USB PHY platform device")
> Signed-off-by: David Lechner <[email protected]>
> ---
superseded by v2 that applies cleanly before common clock series
On Thursday 15 March 2018 11:32 PM, David Lechner wrote:
> This fixes a possible kernel oops due to using stack allocated platform
> data for the USB PHY driver on DA8XX devices. If the platform device
> probe is deferred, then we get a corrupt pointer for the platform data.
>
> We now use a global static struct for the platform data so that the
> platform data pointer does not get written over.
>
> Tested on OMAP-L138 LCDK board using legacy board file.
Oops. That should have been caught in review.
>
> Fixes: 9b504750923cb ("ARM: davinci: da8xx: Add USB PHY platform device")
The offending commit seems to be bdec5a6b5789 ("ARM: da8xx: use platform
data for CFGCHIP syscon regmap") in my tree which should be in
linux-next. I don't think we have this issue in mainline. Right?
> Signed-off-by: David Lechner <[email protected]>
Thanks,
Sekhar
On 03/16/2018 05:17 AM, Sekhar Nori wrote:
> On Thursday 15 March 2018 11:32 PM, David Lechner wrote:
>> This fixes a possible kernel oops due to using stack allocated platform
>> data for the USB PHY driver on DA8XX devices. If the platform device
>> probe is deferred, then we get a corrupt pointer for the platform data.
>>
>> We now use a global static struct for the platform data so that the
>> platform data pointer does not get written over.
>>
>> Tested on OMAP-L138 LCDK board using legacy board file.
>
> Oops. That should have been caught in review.
>
>>
>> Fixes: 9b504750923cb ("ARM: davinci: da8xx: Add USB PHY platform device")
>
> The offending commit seems to be bdec5a6b5789 ("ARM: da8xx: use platform
> data for CFGCHIP syscon regmap") in my tree which should be in
> linux-next. I don't think we have this issue in mainline. Right?
You are right. I'm not sure how I got that older commit. The issue is not
in mainline yet.
>
>> Signed-off-by: David Lechner <[email protected]>
>
> Thanks,
> Sekhar
>