forked from mirror/openwrt
de59fc4540
All NETGEAR EX6150v2 validate the rootfs for which OpenWrt places a fakeheader at the position, where the bootloader expects it. Some EX6150v2 bootloaders do however make a broken assumption about where the rootfs starts. This is due to them calculating the rootfs start not based upon the kernel-length but the string-offset of the FIT-image. We have to be compatible with both this broken as well as the valid calculation. So we do relocate the FDT string section to a block-boundary and enlarge the FIT image to end at this boundary + BLOCKSIZE / 2. This way, both the broken as well as correct calculations do expect the rootfs-header at the same position. It is worth noting, that this is a rare edge-case in which only happens if the image-length as well as the start of the string-section are not placed in the same erase-block. This is an edge-case which happens very rarely (thus it was not spotted prior). Affected: - U-Boot 2012.07 (Jun 16 2016 - 11:59:37) Signed-off-by: David Bauer <mail@david-bauer.net> (cherry picked from commit 8f9546f7b0a14f3afa813e39ed45c968ece24464) |
||
|---|---|---|
| .. | ||
| airoha | ||
| apm821xx | ||
| archs38 | ||
| armsr | ||
| at91 | ||
| ath79 | ||
| bcm27xx | ||
| bcm47xx | ||
| bcm53xx | ||
| bcm4908 | ||
| bmips | ||
| d1 | ||
| gemini | ||
| generic | ||
| imx | ||
| ipq40xx | ||
| ipq806x | ||
| ixp4xx | ||
| kirkwood | ||
| lantiq | ||
| layerscape | ||
| loongarch64 | ||
| malta | ||
| mediatek | ||
| mpc85xx | ||
| mvebu | ||
| mxs | ||
| octeon | ||
| octeontx | ||
| omap | ||
| pistachio | ||
| qoriq | ||
| qualcommax | ||
| ramips | ||
| realtek | ||
| rockchip | ||
| sifiveu | ||
| starfive | ||
| sunxi | ||
| tegra | ||
| uml | ||
| x86 | ||
| zynq | ||
| Makefile | ||