[openwrt/openwrt] mvebu: puzzle-m90x: wipe rootfs_data on sysupgrade
LEDE Commits
lede-commits at lists.infradead.org
Tue Apr 9 07:02:10 PDT 2024
dangole pushed a commit to openwrt/openwrt.git, branch main:
https://git.openwrt.org/d4e8653e1e9ec90404d12ea24fb0f8e507b337d5
commit d4e8653e1e9ec90404d12ea24fb0f8e507b337d5
Author: Daniel Golle <daniel at makrotopia.org>
AuthorDate: Tue Jan 9 19:54:51 2024 +0000
mvebu: puzzle-m90x: wipe rootfs_data on sysupgrade
The sysupgrade formware of the Puzzle series is a slightly strange
dual-boot approach while remaining compatible with Marvell's SDK
firmware upgrade binary format -- which happens to be a full-disk
image with GPT partition table. Hence that /lib/upgrade/emmc-puzzle.sh
script is like an exotic disease which results from those decisions,
and as we also want to somehow stay compatible with the IEI-World
stock firmware we got to use it in that same way (we are not
compatible with the QNAP-branded identical hardware device anyway).
Currently, on sysupgrade the result is that one ends up with the old
content of rootfs_data (a GPT partition on those devices) as nothing
ever wipes or in any way re-creates the filesystem there. As a simple
work-around, let's kill the filesystem on rootfs_data so fstools
re-formats it on the next boot.
Signed-off-by: Daniel Golle <daniel at makrotopia.org>
---
target/linux/mvebu/cortexa72/base-files/lib/upgrade/emmc-puzzle.sh | 2 ++
1 file changed, 2 insertions(+)
diff --git a/target/linux/mvebu/cortexa72/base-files/lib/upgrade/emmc-puzzle.sh b/target/linux/mvebu/cortexa72/base-files/lib/upgrade/emmc-puzzle.sh
index 5e5c356ed6..9a33b7714c 100755
--- a/target/linux/mvebu/cortexa72/base-files/lib/upgrade/emmc-puzzle.sh
+++ b/target/linux/mvebu/cortexa72/base-files/lib/upgrade/emmc-puzzle.sh
@@ -32,5 +32,7 @@ platform_do_upgrade_emmc() {
v "Writing new UUID to /dev/$diskdev..."
get_image_dd "$1" of="/dev/$diskdev" bs=1 skip=440 count=4 seek=440 conv=fsync
+ dd if=/dev/zero of=$(find_mmc_part rootfs_data) bs=512 count=8
+
sleep 1
}
More information about the lede-commits
mailing list