explain why we use mmap instead of malloc

diff --git a/libbb/copyfd.c b/libbb/copyfd.c
index ed383ae289173ec945e528066b6bb87af8085b6a..08bc6f8dfd7cea500550685bd548bd4151940f41 100644 (file)
--- a/libbb/copyfd.c
+++ b/libbb/copyfd.c
@@ -22,6 +22,8 @@ static off_t bb_full_fd_action(int src_fd, int dst_fd, off_t size)
        char *buffer;
        int buffer_size;
 
+       /* We want page-aligned buffer, just in case kernel is clever
+        * and can do page-aligned io more efficiently */
        buffer = mmap(NULL, CONFIG_FEATURE_COPYBUF_KB * 1024,
                        PROT_READ | PROT_WRITE,
                        MAP_PRIVATE | MAP_ANON,