libbb: use sendfile() to copy data between file descriptors

Busybox already uses sendfile in httpd. This patch proposes to use it
globally to copy data between file descriptors.

It speeds up the copying on slow systems a lot - below are the times needed
to copy a 450Mb file with and without this option enabled on a BeagleBone
Black:

sendfile:
user    0m0.000s
sys     0m8.170s

read/write 4k:
user    0m0.470s
sys     0m16.300s

function                                             old     new   delta
bb_full_fd_action                                    394     474     +80

Signed-off-by: Bartosz Golaszewski <bartekgola@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

4 files changed, 71 insertions, 42 deletions

diff --git a/Config.in b/Config.in
index b83beb52d..285fe0a19 100644
--- a/Config.in
+++ b/Config.in
@@ -264,6 +264,18 @@ config PAM
 	  Use PAM in some busybox applets (currently login and httpd) instead
 	  of direct access to password database.
 
+config FEATURE_USE_SENDFILE
+	bool "Use sendfile system call"
+	default y
+	help
+	  When enabled, busybox will use the kernel sendfile() function
+	  instead of read/write loops to copy data between file descriptors
+	  (for example, cp command does this a lot).
+	  If sendfile() doesn't work, copying code falls back to read/write
+	  loop. sendfile() was originally implemented for faster I/O
+	  from files to sockets, but since Linux 2.6.33 it was extended
+	  to work for many more file types.
+
 config LONG_OPTS
 	bool "Support for --long-options"
 	default y
diff --git a/libbb/copyfd.c b/libbb/copyfd.c
index eda2747f9..7e3531903 100644
--- a/libbb/copyfd.c
+++ b/libbb/copyfd.c
@@ -8,6 +8,20 @@
  */
 
 #include "libbb.h"
+#if ENABLE_FEATURE_USE_SENDFILE
+# include <sys/sendfile.h>
+#else
+# define sendfile(a,b,c,d) (-1)
+#endif
+
+/*
+ * We were using 0x7fff0000 as sendfile chunk size, but it
+ * was seen to cause largish delays when user tries to ^C a file copy.
+ * Let's use a saner size.
+ * Note: needs to be >= max(CONFIG_FEATURE_COPYBUF_KB),
+ * or else "copy to eof" code will use neddlesly short reads.
+ */
+#define SENDFILE_BIGBUF (16*1024*1024)
 
 /* Used by NOFORK applets (e.g. cat) - must not use xmalloc.
  * size < 0 means "ignore write errors", used by tar --to-command
@@ -18,12 +32,13 @@ static off_t bb_full_fd_action(int src_fd, int dst_fd, off_t size)
 	int status = -1;
 	off_t total = 0;
 	bool continue_on_write_error = 0;
-#if CONFIG_FEATURE_COPYBUF_KB <= 4
+	ssize_t sendfile_sz;
+#if CONFIG_FEATURE_COPYBUF_KB > 4
+	char *buffer = buffer; /* for compiler */
+	int buffer_size = 0;
+#else
 	char buffer[CONFIG_FEATURE_COPYBUF_KB * 1024];
 	enum { buffer_size = sizeof(buffer) };
-#else
-	char *buffer;
-	int buffer_size;
 #endif
 
 	if (size < 0) {
@@ -31,46 +46,58 @@ static off_t bb_full_fd_action(int src_fd, int dst_fd, off_t size)
 		continue_on_write_error = 1;
 	}
 
-#if CONFIG_FEATURE_COPYBUF_KB > 4
-	if (size > 0 && size <= 4 * 1024)
-		goto use_small_buf;
-	/* 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,
-			/* ignored: */ -1, 0);
-	buffer_size = CONFIG_FEATURE_COPYBUF_KB * 1024;
-	if (buffer == MAP_FAILED) {
- use_small_buf:
-		buffer = alloca(4 * 1024);
-		buffer_size = 4 * 1024;
-	}
-#endif
-
 	if (src_fd < 0)
 		goto out;
 
+	sendfile_sz = !ENABLE_FEATURE_USE_SENDFILE
+		? 0
+		: SENDFILE_BIGBUF;
 	if (!size) {
-		size = buffer_size;
+		size = SENDFILE_BIGBUF;
 		status = 1; /* copy until eof */
 	}
 
 	while (1) {
 		ssize_t rd;
 
-		rd = safe_read(src_fd, buffer, size > buffer_size ? buffer_size : size);
-
-		if (!rd) { /* eof - all done */
-			status = 0;
-			break;
+		if (sendfile_sz) {
+			rd = sendfile(dst_fd, src_fd, NULL,
+				size > sendfile_sz ? sendfile_sz : size);
+			if (rd >= 0)
+				goto read_ok;
+			sendfile_sz = 0; /* do not try sendfile anymore */
+		}
+#if CONFIG_FEATURE_COPYBUF_KB > 4
+		if (buffer_size == 0) {
+			if (size > 0 && size <= 4 * 1024)
+				goto use_small_buf;
+			/* 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,
+					/* ignored: */ -1, 0);
+			buffer_size = CONFIG_FEATURE_COPYBUF_KB * 1024;
+			if (buffer == MAP_FAILED) {
+ use_small_buf:
+				buffer = alloca(4 * 1024);
+				buffer_size = 4 * 1024;
+			}
 		}
+#endif
+		rd = safe_read(src_fd, buffer,
+			size > buffer_size ? buffer_size : size);
 		if (rd < 0) {
 			bb_perror_msg(bb_msg_read_error);
 			break;
 		}
+ read_ok:
+		if (!rd) { /* eof - all done */
+			status = 0;
+			break;
+		}
 		/* dst_fd == -1 is a fake, else... */
-		if (dst_fd >= 0) {
+		if (dst_fd >= 0 && !sendfile_sz) {
 			ssize_t wr = full_write(dst_fd, buffer, rd);
 			if (wr < rd) {
 				if (!continue_on_write_error) {
@@ -92,10 +119,8 @@ static off_t bb_full_fd_action(int src_fd, int dst_fd, off_t size)
 	}
  out:
 
-#if CONFIG_FEATURE_COPYBUF_KB > 4
-	if (buffer_size != 4 * 1024)
+	if (buffer_size > 4 * 1024)
 		munmap(buffer, buffer_size);
-#endif
 	return status ? -1 : total;
 }
 
diff --git a/networking/Config.src b/networking/Config.src
index e56646917..15a696876 100644
--- a/networking/Config.src
+++ b/networking/Config.src
@@ -181,14 +181,6 @@ config FEATURE_HTTPD_RANGES
 	  "Range: bytes=NNN-[MMM]" header. Allows for resuming interrupted
 	  downloads, seeking in multimedia players etc.
 
-config FEATURE_HTTPD_USE_SENDFILE
-	bool "Use sendfile system call"
-	default y
-	depends on HTTPD
-	help
-	  When enabled, httpd will use the kernel sendfile() function
-	  instead of read/write loop.
-
 config FEATURE_HTTPD_SETUID
 	bool "Enable -u <user> option"
 	default y
diff --git a/networking/httpd.c b/networking/httpd.c
index 621d9cddc..9cf080401 100644
--- a/networking/httpd.c
+++ b/networking/httpd.c
@@ -133,7 +133,7 @@
 # include <security/pam_appl.h>
 # include <security/pam_misc.h>
 #endif
-#if ENABLE_FEATURE_HTTPD_USE_SENDFILE
+#if ENABLE_FEATURE_USE_SENDFILE
 # include <sys/sendfile.h>
 #endif
 /* amount of buffering in a pipe */
@@ -1624,7 +1624,7 @@ static NOINLINE void send_file_and_exit(const char *url, int what)
 #endif
 	if (what & SEND_HEADERS)
 		send_headers(HTTP_OK);
-#if ENABLE_FEATURE_HTTPD_USE_SENDFILE
+#if ENABLE_FEATURE_USE_SENDFILE
 	{
 		off_t offset = range_start;
 		while (1) {
@@ -1654,7 +1654,7 @@ static NOINLINE void send_file_and_exit(const char *url, int what)
 			break;
 	}
 	if (count < 0) {
- IF_FEATURE_HTTPD_USE_SENDFILE(fin:)
+ IF_FEATURE_USE_SENDFILE(fin:)
 		if (verbose > 1)
 			bb_perror_msg("error");
 	}