On 6 July 2014 18:08, Felix Fietkau <[email protected]> wrote:
> On 2014-07-06 13:17, Roman Yeryomin wrote:
>> On 6 July 2014 01:20, Roman Yeryomin <[email protected]> wrote:
>>> On 5 July 2014 18:03, Felix Fietkau <[email protected]> wrote:
>>>> On 2014-07-05 00:35, Roman Yeryomin wrote:
>>>>> This patch set fully fixes rb532 support and removes broken mark.
>>>>>
>>>>> Roman Yeryomin (4):
>>>>>   [rb532] fix jffs2 images
>>>>>   [rb532] change default package set
>>>>>   [rb532] add wan interface
>>>>>   [rb532] remove broken mark
>>>> What about squashfs images? It looks to me like one of your patches
>>>> which was applied in r40717 tried to fix them but actually left them in
>>>> a broken state - the cmdline mounts squashfs directly on sda2 instead of
>>>> going through block2mtd.
>>>
>>> Yes, that's sounds weird to me also now but they work like that for me.
>>> I will check that again tomorrow.
>>
>> So, squashfs works both ways. But the older squashfs cmdline was
>> broken anyway as it was using mtdblock0 (like jffs2).
>> I can add a patch to change squashfs to using block2mtd again, if you
>> want, but I don't see any reasons to do that.
> Are you sure that it actually mounts the jffs2 overlay? I'd expect it to
> come up with tmpfs the way it's implemented right now.

You, were right, I didn't look at that before (I just used ext4 images).
Please see the patch attached. I don't know if this hack is acceptable
but I don't see any other way to fix it.
The key point is that we have to rename the onboard flash partition to
make splitting work (without introducing other split methods?). This,
potentially, can break something elsewhere but I don't see where right
now.

Regards,
Roman
commit 06fd753c3e119240cf4003f9e88e7292ddf9973d
Author: Roman Yeryomin <[email protected]>
Date:   Mon Jul 7 01:29:52 2014 +0300

    [rb532] really fix squashfx images
    
    The previous attempt (r40717) to fix squashfx made the system bootable but overlayfs remained broken.
    This enables block2mtd again using the right partition number and naming and adds MTD_ROOTFS_SPLIT option.
    However onboard rootfs partition name had to be renamed to place overlay on CF instead of onboard NAND.
    
    Signed-off-by: Roman Yeryomin <[email protected]>

diff --git a/target/linux/rb532/config-3.10 b/target/linux/rb532/config-3.10
index 3e7c9e4..5e7dd3d 100644
--- a/target/linux/rb532/config-3.10
+++ b/target/linux/rb532/config-3.10
@@ -103,7 +103,7 @@ CONFIG_MTD_NAND_ECC=y
 CONFIG_MTD_NAND_PLATFORM=y
 CONFIG_MTD_PHYSMAP=y
 # CONFIG_MTD_ROOTFS_ROOT_DEV is not set
-# CONFIG_MTD_ROOTFS_SPLIT is not set
+CONFIG_MTD_ROOTFS_SPLIT=y
 # CONFIG_MTD_SM_COMMON is not set
 # CONFIG_MTD_SPLIT is not set
 CONFIG_NEED_DMA_MAP_STATE=y
diff --git a/target/linux/rb532/image/Makefile b/target/linux/rb532/image/Makefile
index 3cc7bce..706c768 100644
--- a/target/linux/rb532/image/Makefile
+++ b/target/linux/rb532/image/Makefile
@@ -56,7 +56,7 @@ root=/dev/sda2 rootfstype=ext4
 endef
 
 define Image/cmdline/squashfs
-root=/dev/sda2 rootfstype=squashfs
+block2mtd.block2mtd=/dev/sda2,65536,rootfs root=/dev/mtdblock2 rootfstype=squashfs
 endef
 
 define Image/cmdline/yaffs2
diff --git a/target/linux/rb532/patches-3.10/004-rb532_partition_info-rename-rootfs-to-rootfs_onboard.patch b/target/linux/rb532/patches-3.10/004-rb532_partition_info-rename-rootfs-to-rootfs_onboard.patch
new file mode 100644
index 0000000..e3f6c41
--- /dev/null
+++ b/target/linux/rb532/patches-3.10/004-rb532_partition_info-rename-rootfs-to-rootfs_onboard.patch
@@ -0,0 +1,11 @@
+--- a/arch/mips/rb532/devices.c	2014-07-07 01:05:10.000000000 +0300
++++ b/arch/mips/rb532/devices.c	2014-07-07 01:05:27.268505016 +0300
+@@ -202,7 +202,7 @@ static struct mtd_partition rb532_partit
+ 		.offset = 0,
+ 		.size = 4 * 1024 * 1024,
+ 	}, {
+-		.name = "rootfs",
++		.name = "rootfs_onboard",
+ 		.offset = MTDPART_OFS_NXTBLK,
+ 		.size = MTDPART_SIZ_FULL,
+ 	}
_______________________________________________
openwrt-devel mailing list
[email protected]
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel

Reply via email to