On Tue, 2009-02-17 at 15:20 +0800, Peter Teoh wrote:
> looking into kernel source, it seemed quite simple:
>
> ./sound/pci/emu10k1/memory.c:
> #define get_aligned_page(offset) ((offset) >> PAGE_SHIFT)
> blk->first_page = get_aligned_page(blk->mem.offset);
> blk->last_page = get_aligned_page(blk->mem.offset + blk->mem.size
> - 1);
> psize = get_aligned_page(size + PAGE_SIZE -1);
> page = get_aligned_page(offset);
>
> ./sound/pci/trident/trident_memory.c:
> #define get_aligned_page(offset) ((offset) >> 12)
> #define get_aligned_page(offset) ((offset) >> 13)
> #define get_aligned_page(offset) ((offset) / ALIGN_PAGE_SIZE)
> ptr = page_to_ptr(trident, get_aligned_page(offset));
> psize = get_aligned_page(size + ALIGN_PAGE_SIZE -1);
>
> ie, just AND with the alignement requirement.
>
> And the trident_memory.c:is_valid_page() function worked like this:
>
> /*
> * check if the given pointer is valid for pages
> */
> static int is_valid_page(unsigned long ptr)
> {
> if (ptr & ~0x3fffffffUL) {
> snd_printk(KERN_ERR "max memory size is 1GB!!\n");
> return 0;
> }
> if (ptr & (SNDRV_TRIDENT_PAGE_SIZE-1)) {
> snd_printk(KERN_ERR "page is not aligned\n");
> return 0;
> }
> return 1;
> }
>
> So, as the last part shows, just AND the address with the alignment
> requirement will do.
>
> But I have not looked else where.
Please look at the snippet below and correct me if I am wrong
#define ALIGN_SIZE 16
#define ALIGN_MASK ~(ALIGN_SIZE - 1)
#define GET_ALIGNED_ADDR(x) (x & ALIGN_MASK)
Wouldn't GET_ALIGNED_ADDR return aligned address?
Regards
- Himanshu
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
