sstrip preserve permissions

Unlike "strip", the "sstrip" does not preserve the file permissions of its target. So if you have a package that sets special permissions, such as the setuid bit, sstrip will remove it. This patch adds some minimal lines to preserve the permissions of stripped files.

---
 .../sstrip/sstrip-20154-preserve-permissions.patch |   43 ++++++++++++++++++++
 1 file changed, 43 insertions(+)
 create mode 100644 package/sstrip/sstrip-20154-preserve-permissions.patch

new file mode 100644
index 0000000..0bd9162
@@ -0,0 +1,43 @@
+diff -urN sstrip.ORIG/src/sstrip.c sstrip/src/sstrip.c
+--- sstrip.ORIG/src/sstrip.c	2010-12-11 23:59:51.464374000 -0500
++++ sstrip/src/sstrip.c	2012-02-21 11:58:56.320295827 -0500
+@@ -60,6 +60,10 @@
+ #include	<fcntl.h>
+ #include	<elf.h>
+ 
++#include <sys/types.h>
++#include <sys/stat.h>
++#include <unistd.h>
++
+ #ifndef TRUE
+ #define	TRUE		1
+ #define	FALSE		0
+@@ -427,6 +431,7 @@
+ 
+ 	for (arg = argv + 1 ; *arg != NULL ; ++arg) {
+ 		filename = *arg;
++		struct stat sb;
+ 
+ 		fd = open(*arg, O_RDWR);
+ 		if (fd < 0) {
+@@ -435,6 +440,9 @@
+ 			continue;
+ 		}
+ 
++		/* Get original file's permissions */
++		if (fstat(fd, &sb) == -1) { perror("fstat"); sb.st_mode = 0; }
++
+ 		switch (readelfheaderident(fd, &e.ehdr32)) {
+ 			case ELFCLASS32:
+ 				if (!(readelfheader32(fd, &e.ehdr32)					&&
+@@ -458,6 +466,10 @@
+ 				++failures;
+ 				break;
+ 		}
++
++		/* Set original file's permissions, including setuid */
++		if (sb.st_mode != 0) { fchmod(fd, sb.st_mode & 07777); }
++
+ 		close(fd);
+ 	}
+ 

1.7.9.5

Danomi> Unlike "strip", the "sstrip" does not preserve the file permissions Danomi> of its target. So if you have a package that sets special permissions, Danomi> such as the setuid bit, sstrip will remove it. This patch adds some Danomi> minimal lines to preserve the permissions of stripped files.

What is the use case for this? From a quick look it seems sstrip just truncates the source file, so permissions shouldn't change.

setuid (to root) doesn't make much sense as we're building as non-root and fixing up permissions at the end using fakeroot (E.G. see BUSYBOX_PERMISSIONS).

Danomi> Signed-off-by: Danomi Manchego <[email protected]> Danomi> ---

 Danomi>  .../sstrip/sstrip-20154-preserve-permissions.patch |   43 ++++++++++++++++++++

Danomi> 1 file changed, 43 insertions(+)
Danomi> create mode 100644 package/sstrip/sstrip-20154-preserve-permissions.patch

Danomi> diff --git a/package/sstrip/sstrip-20154-preserve-permissions.patch b/package/sstrip/sstrip-20154-preserve-permissions.patch Danomi> new file mode 100644
Danomi> index 0000000..0bd9162
Danomi> --- /dev/null
Danomi> +++ b/package/sstrip/sstrip-20154-preserve-permissions.patch
Danomi> @@ -0,0 +1,43 @@
Danomi> +diff -urN sstrip.ORIG/src/sstrip.c sstrip/src/sstrip.c
Danomi> +--- sstrip.ORIG/src/sstrip.c 2010-12-11 23:59:51.464374000 -0500
Danomi> ++++ sstrip/src/sstrip.c 2012-02-21 11:58:56.320295827 -0500
Danomi> +@@ -60,6 +60,10 @@
Danomi> + #include <fcntl.h>
Danomi> + #include <elf.h>
Danomi> +
Danomi> ++#include <sys/types.h>
Danomi> ++#include <sys/stat.h>
Danomi> ++#include <unistd.h>
Danomi> ++
Danomi> + #ifndef TRUE
Danomi> + #define TRUE 1
Danomi> + #define FALSE 0
Danomi> +@@ -427,6 +431,7 @@
Danomi> +
Danomi> + for (arg = argv + 1 ; *arg != NULL ; ++arg) {
Danomi> + filename = *arg;
Danomi> ++ struct stat sb;
Danomi> +
Danomi> + fd = open(*arg, O_RDWR);
Danomi> + if (fd < 0) {
Danomi> +@@ -435,6 +440,9 @@
Danomi> + continue;
Danomi> + }
Danomi> +
Danomi> ++ /* Get original file's permissions */
Danomi> ++ if (fstat(fd, &sb) == -1) { perror("fstat"); sb.st_mode = 0; }
Danomi> ++
Danomi> + switch (readelfheaderident(fd, &e.ehdr32)) {
Danomi> + case ELFCLASS32:
Danomi> + if (!(readelfheader32(fd, &e.ehdr32) &&
Danomi> +@@ -458,6 +466,10 @@
Danomi> + ++failures;
Danomi> + break;
Danomi> + }
Danomi> ++
Danomi> ++ /* Set original file's permissions, including setuid */
Danomi> ++ if (sb.st_mode != 0) { fchmod(fd, sb.st_mode & 07777); }
Danomi> ++
Danomi> + close(fd);
Danomi> + }
Danomi> +
Danomi> --
Danomi> 1.7.9.5

Danomi> _______________________________________________ Danomi> buildroot mailing list
Danomi> buildroot at busybox.net
Danomi> http://lists.busybox.net/mailman/listinfo/buildroot

Bye, Peter Korsgaard

Hi Peter,

We have a proprietary CGI app that gets called by lighttpd, so the caller isn't root. I don't quite remember the exact operation that caused us to do a chmod +s. That's when we noticed that the +s perms were disappearing when sstrip was used, but conserved when strip was used.

Nope. Just try it on a file that gets cut down. For example:

	$ ll output/target/usr/bin/i2cdetect
	-rwxr-xr-x 1 dmocelo dmocelo 34726 Nov  7 21:34
output/target/usr/bin/i2cdetect*

	$ chmod a+s output/target/usr/bin/i2cdetect

	$ ll output/target/usr/bin/i2cdetect
	-rwsr-sr-x 1 dmocelo dmocelo 34726 Nov  7 21:34
output/target/usr/bin/i2cdetect*

	$ # now I'll run a make with sstrip enabled
	$ make
	...

	$ ll output/target/usr/bin/i2cdetect
	-rwxr-xr-x 1 dmocelo dmocelo 11630 Nov  7 21:36
output/target/usr/bin/i2cdetect*

We make our packages to be friendly to multiple build systems, so the package makefile's install operation is taking care of the special permission, rather than using a buildroot-specific mechanism.

But regardless, I was thinking that the difference compared to normal strip operation was reason enough to want to preserve permissions. No?

Danomi -