Re: Yet another updated entropy patch

[olafBuddenhagen]

Hi,

On Fri, Jul 13, 2007 at 01:04:01AM -0400, Michael Casadevall wrote:

> + void entropy_putchar(int c) {
[...]
> +   if (entropy_write_offset == ENTROPYBUFSIZE) {
> +     entropy_write_offset = 0;
> +   }
> +
> +   if (entropy_write_offset == ENTROPYBUFSIZE) {
> +     entropy_write_offset = 0;
> +   }// Deep magic (tells any read functiosn more entropy is available)

Surely you don't really want to check twice :-)

> + void entropy_putdata(void *data, int size) {
> +   // This function is used to copy data from memory pointers directly
> +   // into the bufffer
> +   io_req_t ior; // Used to tell reads that we got entropy
> +
> +   // See if we are initalized; if not, initalize the driver
> +   if (!entropy_init_done) {
> +     entropyinit();
> +   }
> +
> +   /* See if the buffer is full, if it is, bail out */
> +   if (entropy_write_offset+1 == entropy_read_offset) {
> +     return;
> +   }

You only check whether a single byte fits in, although you might write
more than one...

> +  // Try to grab a lock on the device, otherwise bail
> +   if (!simple_lock_try(&entropy_lock)) {
> +     return; // Bailing out
> +   }
> +
> +   // Lets advance the pointer on
> +   entropy_write_offset++;

Still incrementing before the actual write?

-antrik-