Re: [PATCH] bitmap: bitmap_parse takes a kernel buffer instead of a user buffer

[inaky@linux.intel.com]

> Joe wrote:
>> I guess I am a sucker for no-transient-buffer (bufferless?)
>
> Ah - that explains Joe's preference for putting the actual implementing
> code in the user version - it gets to pull in the user string one
> char at a time, avoiding a malloc'd buffer.

I tend to gree w/ Joe there.

I wonder if a hybrid would be ok, although I the pseudo impl
I propose below is kind of dirty, but some people might find it
justified enough:

static
__bitmap_parse(const void *_buf, size_t size, enum { KERNEL, USER } type,
               unsigned long *dst, int nbits)
{
        const char __user *ubuf = _buf;
        const char *buf = _buf;
        ...
        switch(type) {
        case USER:
                if (get_user(c, ubuf++))
                        return -EFAULT;
                break;
        case KERNEL:
                c = *buf++;
                break;
        default:
                BUG();
        ...
}


int bitmap_parse(const char *buf, unsigned int buflen,
                 unsigned long *maskp, int nmaskbits) {
    return __bitmap_parse(buf, buflen, KERNEL, maskp, nmaskbits);
}


int bitmap_parse_user(const char __user *buf, unsigned int buflen,
                      unsigned long *maskp, int nmaskbits) {
    return __bitmap_parse(buf, buflen, USER, maskp, nmaskbits);
}


[that or exposing __bitmap_user() as a extern / EXPORT and putting
 bitmap_parse{,_user}() as an inline in the header file]

It's nitty-gritty, but it removes the kmalloc from the equation...

-- 
Inaky