Thanks, I have edited it.
1 Like
Your issue is that packages that are defined localy as iwinfo are too old for those that are pulled from git.
Correct me if I'm wrong, but he should be able to pull appropriate snapshots of packages from the time of your latest branch.
Yes, if he find the commit hash from around that time and sets that in feeds
I'm doing a test build at the moment, so I'll post the hashes if/once it completes without errors.
Okay, so anyone looking to build @robimarko's RB450Gx4-rebase-sdhci branch should use the following feed snapshots:
src-git packages https://git.openwrt.org/feed/packages.git^c967c23124e4073a2615cbc5b37eef6a8c47a69
src-git luci https://git.openwrt.org/project/luci.git^acc2c8d27781ce24e9d91353bfb211ff960e166c
src-git routing https://git.openwrt.org/feed/routing.git^c687bd3055d08c6914f53f2042d1aed316dbaae51
src-git telephony https://git.openwrt.org/feed/telephony.git^60266dc66579928df8e30baea5ccc4e8c6558d69
Just edit feeds.conf.default before updating/installing feeds and you should be able to get a successful build.
3 Likes
Guys, that branch is a dead one and there is no RB450Gx4.
Use RB450Gx4-rebase-sdhci branch, thats the last one and has full SD card support
It works, thank you!
Is there any progress on when rb450Gx4 will be officially supported?
Thanks
1 Like
Hopefully soon, I should have some time in February to respin the whole patch series.
I got the SD card support stuff along with LDO driver for it merged in upstream Linux, so its awaiting merge in another of my PR-s in OpenWrt.
Hi, I have recently received a few new RB450Gx4 boards and installed OpenWRT via sysupgrade (after doing ubidetach and ubiformat).
Unfortunately I am not able to boot some of them from NAND (others are working fine).
An error like this occurs:
OpenWrt kernel loader for Qualcomm IPQ-4XXX/IPQ-806X
Copyright (C) 2019 Sergey Sergeev <adron@mstnt.com>
Extracting LZMA kernel...LzmaDecode error 1 at 84802191, osize:8504384 ip:0 op:0
00 0f 06 bc 8e b0 02 73 92 d1 62 7f 6e 73 d1 33
failed, data error!
After further investigation I found that the boards are equipped with different NAND chips (same size like on working boards, but different erase size, page size and OOB size).
On the boards not booting from NAND i have found a Winbond W29N04GV91AA chip:
nand: device found, Manufacturer ID: 0xef, Chip ID: 0xdc
nand: Winbond W29N04GV
nand: 512 MiB, SLC, erase size: 128 KiB, page size: 2048, OOB size: 64
Partitions reported via /proc/mtd are:
dev: size erasesize name
mtd0: 01000000 00020000 "kernel"
mtd1: 1f000000 00020000 "ubi"
mtd2: 00023000 00001000 "Qualcomm-boot1"
mtd3: 0005d000 00001000 "Qualcomm-boot2"
mtd4: 0000e000 00001000 "RouterBoot"
mtd5: 00001000 00001000 "hard_config"
mtd6: 00010000 00001000 "dts_config"
mtd7: 00001000 00001000 "soft_config"
On boards working well, the erase size for kernel and ubi partition is twice, everything else is the same.
I see the same problem with pre-built binary image from google drive as well as image compiled from RB450Gx4-rebase-sdhci branch - slighly different error message but basically the same problem.
Also I noted the RouterBOOT version is slightly newer on the boards not booting from NAND (6.42 vs. 6.41.3).
Is there anybody, who managed to get OpenWRT to boot from NAND with this type
of NAND chip?
Could the problem be related to the different erase size on the NAND chip or different RooterBOOT version?
Any hint how to get OpenWRT booting via these NAND chips would be very welcome - thank you!
I am currently sick, and unable to test anything but its most likely becase this Winbond uses 128k blocks and 2048 pages instead of double that of Toshiba that is on mine.
So those parameters need to be changed in OpenWrt and the aux loader
I just meet with the same problem. Could you give more hints about this? Thank you!
Hi, it seems thart I meet the same problem with bm42.
I search sourcecodes with keywords BLOCKSIZE and PAGESIZE, and find two places.
the first is target/linux/ipq40xx/image/ipq-aux-loader/src/Makefile:
#ubi image vars
BLOCKSIZE := 256k
PAGESIZE := 4096
.........
$(ldr-ubifs): $(ldr-elf)
rm -Rf ubi/ubifs-rootdir
mkdir ubi/ubifs-rootdir
cp bin/loader.elf ubi/ubifs-rootdir/kernel
$(STAGING_DIR_HOST)/bin/mkfs.ubifs \
-m $(PAGESIZE) -e 126976 -c 120 --compr=none \
-r ubi/ubifs-rootdir $(ldr-ubifs)
rm -Rf ubi/ubifs-rootdir
the second is /target/linux/ipq40xx/image/Makefile, line 310 define Device/MikrotikNAND
I try to change BLOCKSIZE,PAGESIZE to 128k,2048, and change mkfs.ubifs parameters -e to 126976, -c to 120
I compiled it succesfully, but it still didn't work. My router cannot startup and restart repeatedly.
I really know little about how to correct these NAND parameters, hope you could help me ,thank you!
Aux loader has a hardcoded UBI parameters in one of the files.
That needs to be changed as well
I am rebasing the original PR to current master and then I will reopen the PR.
Guys, I am preparing a new image.
Whole SD card/LDO situation was upstream to Linux and then backported into OpenWrt recently for another board by me.
So, I should have a new image today/tommorow.
I am gonna need testers with W29N04GV NAND instead of Toshiba one
Thank you for your efforts! I would be happy to test with W29N04GV NAND.
Can someone with the Winbond NAND make a dump of the kernel partition so I can take a look at UBIFS parameters they used?
Thanks
I have uploaded a dump of the kernel partition to Google Drive:
https://drive.google.com/drive/folders/10G51s4FMimYu1AbIAvT8pyvacT4QoUpZ?usp=sharing
Image was built with changed BLOCKSIZE to 128k and PAGESIZE to 2048 for Device/MikrotikNAND in target/linux/ipq40xx/image/Makefile and also changed the hardcoded parameters for mkfs.ubifs in target/linux/ipq40xx/image/ipq-aux-loader/src/Makefile to -e 126976 -c 120
1 Like