diff mbox

[LEDE-DEV] build: reintroduce md5 checksums for generated firmware images

Message ID 20160818030035.5044ae5a@mir
State Not Applicable
Headers show

Commit Message

Stefan Lippers-Hollmann Aug. 18, 2016, 1 a.m. UTC 
Hi

On 2016-08-18, Karl Palsson wrote:
> Why not just add sha256 sum rather than going back to md5?
[...]

Generating md5sums in addition to sha256sums files only affects build
time slightly, but doesn't increase the firmware size itself. Given 
that apparently nothing currently requires a sha256sum binary at firmware
runtime, I'd personally consider it to be cheaper not bumping busybox'
size needlessly.

While it's obvious that MD5 can't be considered cryptographically 
secure anymore, it does still work nicely for checking file integrity
in regards to potential transmission errors.

$ ./scripts/diffconfig.sh 
CONFIG_TARGET_ar71xx=y
CONFIG_TARGET_ar71xx_generic=y
CONFIG_TARGET_ar71xx_generic_Default=y

$ ls -gG no-sha256sum/busybox with-sha256sum/busybox
-rwxr-xr-x 1 300437 Aug 17 23:38 no-sha256sum/busybox
-rwxr-xr-x 1 304405 Aug 17 23:38 with-sha256sum/busybox

Giving it a quick/ rough compressed size estimate (-6, with a dictionary 
size of 8 MB, would need 9 MB RAM for decompressing, which would be too 
much for 32 MB RAM devices, so the actual compressed size delta of the 
squashfs image is probably slightly larger):

$ xz -6 no-sha256sum/busybox
$ xz -6 with-sha256sum/busybox
$ ls -gG no-sha256sum/busybox.xz with-sha256sum/busybox.xz 
-rwxr-xr-x 1 173188 Aug 17 23:38 no-sha256sum/busybox.xz
-rwxr-xr-x 1 174704 Aug 17 23:38 with-sha256sum/busybox.xz

The difference between "no-sha256sum/busybox" and
"with-sha256sum/busybox":


Regards
	Stefan Lippers-Hollmann
diff mbox

Patch

--- a/package/utils/busybox/Config-defaults.in
+++ b/package/utils/busybox/Config-defaults.in
@@ -705,7 +705,7 @@  config BUSYBOX_DEFAULT_SHA1SUM
 	default n
 config BUSYBOX_DEFAULT_SHA256SUM
 	bool
-	default n
+	default y
 config BUSYBOX_DEFAULT_SHA512SUM
 	bool
 	default n