On 15.09.2017 11:47, peterh wrote:
> From: Kuanling Huang <peterh@xxxxxxxxxxxx>
>
> By analyzing the perf on btrfs send, we found it take large
> amount of cpu time on page_cache_sync_readahead. This effort
> can be reduced after switching to asynchronous one. Overall
> performance gain on HDD and SSD were 9 and 15 percent if
> simply send a large file.
>
> Signed-off-by: Kuanling Huang <peterh@xxxxxxxxxxxx>
Reviewed-by: Nikolay Borisov <nborisov@xxxxxxxx>
> ---
> fs/btrfs/send.c | 21 ++++++++++++++++-----
> 1 file changed, 16 insertions(+), 5 deletions(-)
>
> diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c
> index 63a6152..7a5eb66 100644
> --- a/fs/btrfs/send.c
> +++ b/fs/btrfs/send.c
> @@ -4475,16 +4475,27 @@ static ssize_t fill_read_buf(struct send_ctx *sctx, u64 offset, u32 len)
> /* initial readahead */
> memset(&sctx->ra, 0, sizeof(struct file_ra_state));
> file_ra_state_init(&sctx->ra, inode->i_mapping);
> - btrfs_force_ra(inode->i_mapping, &sctx->ra, NULL, index,
> - last_index - index + 1);
>
> while (index <= last_index) {
> unsigned cur_len = min_t(unsigned, len,
> PAGE_CACHE_SIZE - pg_offset);
> - page = find_or_create_page(inode->i_mapping, index, GFP_NOFS);
> + page = find_lock_page(inode->i_mapping, index);
> if (!page) {
> - ret = -ENOMEM;
> - break;
> + page_cache_sync_readahead(inode->i_mapping,
> + &sctx->ra, NULL, index,
> + last_index + 1 - index);
> +
> + page = find_or_create_page(inode->i_mapping, index, GFP_KERNEL);
> + if (!page) {
> + ret = -ENOMEM;
> + break;
> + }
> + }
> +
> + if (PageReadahead(page)) {
> + page_cache_async_readahead(inode->i_mapping,
> + &sctx->ra, NULL, page, index,
> + last_index + 1 - index);
> }
>
> if (!PageUptodate(page)) {
>
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
