On 11/18/2014 08:56 AM, Lina Iyer wrote:

On Fri, Nov 14 2014 at 15:46 -0700, Stephen Boyd wrote:

quoted

On 10/24, Lina Iyer wrote:

quoted

quoted

+{
+    struct spm_driver_data *drv = this_cpu_ptr(&cpu_spm_drv);
+    u32 start_index;
+    u32 ctl_val;
+
+    if (!drv->available)
+        return -ENXIO;

It would be nice if we didn't need this by only registering the
cpuidle device for this CPU once we've initialized the SPM
hardware.

I did explore it. It strays our cpuidle code away from the standard code
that we are trying to go towards with idle-states framework.

So fix the framework?

quoted

quoted

+
+    /* Write the SPM sequences, first.. */
+    addr = drv->reg_base +
drv->reg_data->reg_offset[SPM_REG_SEQ_ENTRY];
+    seq_data = (const u32 *)drv->reg_data->seq;

Why do we need a cast?

Compiler warns otherwise.

How?

$ cat main.c
extern int magic(const void *d);

struct m {
        unsigned int data[2];
};

struct s {
        const struct m *m;
};

static const struct m m = {
        .data = { 0x345, 0x34},
};

static const struct s s = {
        .m = &m,
};

int main()
{
        const unsigned int *d;
        d = s.m->data;
        return magic(d);
}

$ gcc -c main.c

quoted

quoted

+        .of_match_table = spm_match_table,
+    },
+};
+
+module_platform_driver(spm_driver);

MODULE_LICENSE()?
MODULE_ALIAS()?

MODULE_DESCRIPTION() would work?

Sure, add them all please.

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project