On Wed, 2008-11-05 at 14:02 +0000, Mel Gorman wrote:
> Currently the caller of get_huge_pages() is expected to align the length to
> the hugepage boundary. This requires the caller to know alignment is required,
> have an ALIGN() macro and know which pagesize is in use. As well as being
> a useless and unnecessary burden, it prevents anything useful being done
> with the wasted bytes. This patch relieves the caller of some responsibility.
>
> Signed-off-by: Mel Gorman <[EMAIL PROTECTED]>
Acked-by: Adam Litke <[EMAIL PROTECTED]>
> ---
> alloc.c | 32 ++++++++++++++++++++------------
> 1 files changed, 20 insertions(+), 12 deletions(-)
>
> diff --git a/alloc.c b/alloc.c
> index 750b2cb..b87a60d 100644
> --- a/alloc.c
> +++ b/alloc.c
> @@ -69,16 +69,17 @@ static void *fallback_base_pages(size_t len, ghp_t flags)
> * flags: Flags specifying the behaviour of the function
> *
> * This function allocates a region of memory backed by huge pages and
> - * at least hugepage-aligned. This is not a suitable drop-in for malloc()
> - * and is only suitable in the event the length is expected to be
> - * hugepage-aligned. However, a malloc-like library could use this function
> - * to create additional heap similar in principal to what morecore does for
> - * glibc malloc.
> + * at least hugepage-aligned. This is not a suitable drop-in for malloc().
> + * As the length is always aligned to a hugepage-boundary, on average
> + * half a hugepage will be wasted unless care is taken. The intention is that
> + * a malloc-like library uses this function to create additional heap similar
> + * in principal to what morecore does for glibc malloc.
> */
> void *get_huge_pages(size_t len, ghp_t flags)
> {
> void *buf;
> int heap_fd;
> + size_t aligned_len, wasteage;
>
> /* Create a file descriptor for the new region */
> heap_fd = hugetlbfs_unlinked_fd();
> @@ -87,15 +88,22 @@ void *get_huge_pages(size_t len, ghp_t flags)
> return NULL;
> }
>
> + /* Align the len parameter */
> + aligned_len = ALIGN(len, gethugepagesize());
> + wasteage = aligned_len - len;
> + if (wasteage != 0)
> + DEBUG("get_huge_pages: Wasted %zd bytes due to alignment\n",
> + wasteage);
> +
> /* Map the requested region */
> - buf = mmap(NULL, len, PROT_READ|PROT_WRITE,
> - MAP_PRIVATE, heap_fd, 0);
> + buf = mmap(NULL, aligned_len, PROT_READ|PROT_WRITE,
> + MAP_PRIVATE, heap_fd, 0);
> if (buf == MAP_FAILED) {
> close(heap_fd);
>
> /* Try falling back to base pages if allowed */
> if (flags & GHP_FALLBACK)
> - return fallback_base_pages(len, flags);
> + return fallback_base_pages(aligned_len, flags);
>
> WARNING("get_huge_pages: New region mapping failed (flags:
> 0x%lX): %s\n",
> flags, strerror(errno));
> @@ -103,19 +111,19 @@ void *get_huge_pages(size_t len, ghp_t flags)
> }
>
> /* Fault the region to ensure accesses succeed */
> - if (hugetlbfs_prefault(heap_fd, buf, len) != 0) {
> - munmap(buf, len);
> + if (hugetlbfs_prefault(heap_fd, buf, aligned_len) != 0) {
> + munmap(buf, aligned_len);
> close(heap_fd);
>
> /* Try falling back to base pages if allowed */
> if (flags & GHP_FALLBACK)
> - return fallback_base_pages(len, flags);
> + return fallback_base_pages(aligned_len, flags);
> }
>
> /* Close the file so we do not have to track the descriptor */
> if (close(heap_fd) != 0) {
> WARNING("Failed to close new heap fd: %s\n", strerror(errno));
> - munmap(buf, len);
> + munmap(buf, aligned_len);
> return NULL;
> }
>
--
Adam Litke - (agl at us.ibm.com)
IBM Linux Technology Center
-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Libhugetlbfs-devel mailing list
Libhugetlbfs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libhugetlbfs-devel
