Hi Pali, On Sun, Jan 29, 2023 at 3:20 AM Pali Rohár <p...@kernel.org> wrote: > > On Saturday 28 January 2023 18:07:29 Tony Dinh wrote: > > > > > > +#define BOOT_TARGET_DEVICES(func) \ > > > > > > + func(SCSI, scsi, 0) \ > > > > > > + func(USB, usb, 0) \ > > > > > > + func(PXE, pxe, na) \ > > > > > > + func(DHCP, dhcp, na) > > > > > > > > > > I see that you have also NAND with UBIFS in DTS file. What is stored > > > > > on > > > > > NAND? Does not it have bootable system and should be in default boot > > > > > target list? > > > > > > > > The UBIFS in stock FW was the rootfs. But it is not bootable any more > > > > with the new u-boot. > > > > > > Hm... Why it is not bootable by new u-boot? Is there some unknown issue > > > that it suddenly stopped working? Because u-boot as a bootloader is > > > there for booting system and it is its primary functionality. And if new > > > version cannot boot something which old version was able then it is a > > > bug on which u-boot should focus. If you have more details about this > > > issue, I can look at it. > > > > IMO, I don't think we need to waste time with stock FW on these NAS > > boards. They are no longer supported by the manufacturers for a while, > > and therefore have some inherent security problems (most if not all > > Kirkwood boards suffer the same fate). Usually people just modernize > > these boxes with a new u-boot and boot them with a modern Linux > > distro. It's a great thing that modern boards like Turris Omnia are > > supported by the manufacturer and other FOSS contributors. With > > Thecus, Netgear, Seagate... even a lot of Synology boxes, people just > > run really old kernel and rootfs and never see an update. > > Well, my point here is that this is ubifs stuff and I have one arm > device on which u-boot ubifs code is unable to parse and read data from > it. I thought that ubifs is completely broken in u-boot but then I > enabled ubifs on another powerpc device and it worked perfectly.
I had a similar experience on a few ARM devices. But when I wiped the partition and created a brand new UBIFS on the same devices with the latest u-boot code, it worked perfectly too. So perhaps it was something in the old filesystem in stock FW that caused the problem. I have on these ARM boards the latest u-boot (2017-2023). So if you would like me to help testing some ideas or some problems, I'd be glad to do what I can. Perhaps we can do it on a different thread, I might be able to find my backup mtd dumps for an old device such as Pogo V4 and try mounting its partition with the new u-boot. Thanks, Tony > I tried to debug that issue on arm device without any success. > Having broken code is u-boot is not something which should be present in > project and I wanted to fix it...