Skip to contentSkip to conversation

current-speed property in serial devices causes kernel panic

DISCUSSION Timur Tabi 2 participants 3 messages Jun 30, 2007 – Jul 2, 2007

lore
Share & tools
lore

Timur Tabi

author
#0 lore raw

I see this code in function of_platform_serial_setup():

static int __devinit of_platform_serial_setup(struct of_device *ofdev,
int type, struct uart_port *port)

{
       struct resource resource;
       struct device_node *np = ofdev->node;
       const unsigned int *clk, *spd;
       int ret;
                    
       memset(port, 0, sizeof *port);
       spd = of_get_property(np, "current-speed", NULL);
...
       port->custom_divisor = *clk / (16 * (*spd));
       
       return 0;
}

There is no check in this code to make sure spd is not null. And sure enough, in most DTS files, current-speed does not exist. So whenever this function is called on a node like this, the kernel panics.

I'm adding support for a new 86xx platform, and I'm also creating a driver for a new SOC device. In my platform driver, I have this code:

static struct of_device_id mpc86xx_ids[] = {
       { .type = "soc", },
       {}
};
static int __init mpc86xx_declare_of_platform_devices(void)  {
       printk(KERN_ALERT "%s\n", __FUNCTION__);

       if (!machine_is(mpc86xx_hpcn))
               return 0;

       of_platform_bus_probe(NULL, mpc86xx_ids, NULL);

       return 0;
}

device_initcall(mpc86xx_declare_of_platform_devices);

The kernel panic occurs only if I call of_platform_bus_probe().

If you look at the code for the 836x platform, you'll see that it also has serial SOC devices and it also calls of_platform_bus_probe(), but it doesn't experience kernel panics.

Is the call to of_platform_bus_probe() effectively trying to probe the serial devices twice? I just don't understand why this code isn't working.

#1 lore raw
Quoted 19 lines · Timur Tabi

I see this code in function of_platform_serial_setup():

static int __devinit of_platform_serial_setup(struct of_device *ofdev,
int type, struct uart_port
*port)

{
        struct resource resource;
        struct device_node *np = ofdev->node;
        const unsigned int *clk, *spd;
        int ret;

        memset(port, 0, sizeof *port);
        spd = of_get_property(np, "current-speed", NULL);
...
        port->custom_divisor = *clk / (16 * (*spd));

        return 0;
}
Quoted 1 line

There is no check in this code to make sure spd is not null. And sure enough, in most DTS files, current-speed does not exist. So whenever this function is called on a node like this, the kernel panics.

I'm wondering why we set port->custom_divisor instead of setting port->speed. (plus the null check).

- k

author
#2 lore raw
In reply to Timur Tabi
Quoted 1 line · Timur Tabi

The kernel panic occurs only if I call of_platform_bus_probe().

FYI, thanks to Kumar, I found the problem: I had CONFIG_SERIAL_OF_PLATFORM enabled. This causes of_serial.c to be compiled and loaded. On my platform, that wasn't doing anything because initially I wasn't calling of_platform_bus_probe(). When I added that call to my platform code, of_serial.c sprang to life and tried to probe the serial nodes in my device tree. Apparently, of_serial.c is not compatible with fsl_soc.c.

Signature · 2 lines
Timur Tabi
Linux Kernel Developer @ Freescale