i was thinking of submitting the following as a new "chapter" for the
doc -- it would address *some* of these issues:
Chapter XX: Page-related memory management
The following functions and macro definitions are available via
include/linux/gfp.h for page-based memory management:
struct page *alloc_pages(gfp_mask, order);
unsigned long __get_free_pages(gfp_mask, order);
unsigned long get_zeroed_page(gfp_mask);
void __free_pages(struct page *page, order);
void free_pages(unsigned long addr, order);
#define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
#define __get_free_page(gfp_mask) __get_free_pages((gfp_mask),0)
#define __free_page(page) __free_pages((page), 0)
#define free_page(addr) free_pages((addr),0)
#define __get_dma_pages(gfp_mask, order) \
__get_free_pages((gfp_mask) | GFP_DMA,(order))
Given the above, some basic suggestions for page-based memory management:
(a) If you need to allocate or free a single page, use the single page
version of the routine/macro, rather than calling the multi-page
version with an order value of zero, such as:
alloc_pages(gfp_mask, 0); /* no */
alloc_page(gfp_mask); /* better */
(b) If you need to allocate a single zeroed page by logical address,
use get_zeroed_page(), rather than __get_free_page() followed
by a call to memset() to clear that page.
(c) If you need to specifically allocate some DMA pages, use the
__get_dma_pages() macro, as in:
__get_free_pages(GFP_KERNEL|GFP_DMA, order) /* no */
__get_dma_pages(GFP_KERNEL, order) /* better */
(d) If you need to clear (zero) a page, be aware that every
architecture defines a clear_page() routine, either as a macro
in include/<arch>/page.h or as an assembler routine.
You should check if it's appropriate to use that routine/macro,
rather than making an explicit call to memset(...,0, PAGE_SIZE).
-
