6 Replies Latest reply on Dec 7, 2014 3:40 AM by AlexT_Intel

    flash to build_16 issue

    deium

      I attempted a reboot ota to upgrade an Edison w/mini breakout board windows from build_68

       

      1) is build_16 not a complete flash - device not cleared, residual files were left behind

      2) Windows Edison: drive was not cleared - image files still remained.

      3) what is the implications of edison filesystem when edison drive not cleared.

       

      is this intentional - or is there an error.

      I would assume an official image should be a complete device replacement.

      If this is happening from build_56 as well, then build_16 may not be consistent depending on the prior image before upgrade

        • 1. Re: flash to build_16 issue
          AlexT_Intel

          Not sure what you mean by residual files, but OTA process preserves the contents of /home directory.

           

          OTA drive isn't cleared during OTA update, so that's normal.

           

          Not I got the third one, please elaborate.

          • 2. Re: flash to build_16 issue
            KurtE

            Thanks Alex,

             

            On the previous version of the firmware download when we did the reboot ota, the complete ota drive was cleared and the home directory was cleared as well.  So when we went flashed up to this version, was not sure it succeeded.  Probably should be some note, saying that the functionality has changed.   I believe 3. Is suppose you are re-flashing because you have corrupted the contents.  Maybe I built my own image which is not working well, or I downloaded several packages that were not working well, or I ran out of system space...

             

            Example yesterday when I received my new Edison and mini breakout board (to replace the one I am having issues with),  I was not sure it succeeded.

             

            So I did it the other way.  I downloaded the image to my linux machine and did a sudo flashall

            • 3. Re: flash to build_16 issue
              deium

              AlexT_Intel,

              When flashing from factory build_56 to build_68 performing a reboot ota the ota drive is most definitely cleared.

              When flashing from build_68 to build_16 the ota drive is not cleared.

              This leaves you with the feeling that it is not complete ...

              If it is only the /home directory not being cleared in the flash to build_16, I can get that - was never stated anywhere.

               

              with regard to 3), IF the OTA drive's space is separate from the 4GB on Edison, okay I can understand and have no issue.

              but it the space for the OTA drive is not separate, which was my assumption as the only reason you fill it is to wipe out the 4GB, then at some point shared file space would eventually be corrupted by the by either the OTA drive or the Linux 4GB over writing.

               

              I haven't noticed any documentation that showed 4GB MMC + 600M OTA - so if shared, as the 4GB gets filled, one corrupts the other

               

              Can we confirm:

              1) it is only the /home dir that is not being cleared - I can certainly write a routine for that.

              2) is the OTA drive separate from the 4GB MMC, otherwise what section of the Linux filesystem is it sharing (that might corrupt the ota contents, or the ota contents might corrupt the Linux files)

              • 4. Re: flash to build_16 issue
                AlexT_Intel

                I've just checked the BSP sources and indeed on the previous image you would have OTA partition cleared - good catch (I've moved off that one quite a while ago and haven't used OTA to do that). Home would be cleared as well,only /home/root would be preserved.

                 

                Now on ww42 image (current one in the Downloads section, you call it build_16), they've changed the boot target for the OTA case and it's now ota-update (which doesn't touch home and OTA partitions at all) instead of the first-install one (which should preserve only /home/root and reformats that partition + reformats the OTA partition).

                 

                You can check all of this by looking at the BSP sources, namely these files:

                1. edison-src/device-software/meta-edison-distro/recipes-core/first-install/files/first-install.sh
                2. edison-src/device-software/meta-edison-distro/recipes-core/ota-update/files/first-install.sh
                3. edison-src/device-software/utils/flash/ota_update.cmd (this is what becomes ota_update.scr during postBuild.sh run and then is executed during OTA)

                 

                So it's /home and the OTA partitions which aren't being cleared and the OTA partition is separate from all other ones (but is on the same MMC flash), so you can't corrupt it by just overfilling the rootfs with data, you'd need to mess up with fdisk or parted or write to the disk directly to accomplish that. Neither can OTA data corrupt the rootfs.

                • 5. Re: flash to build_16 issue
                  deium

                  AlexT_Intel,

                  Thanks, getting my answer piece by piece.  So the 766MB ota drive is separate from the Linux system but still part of a 4GB MMC.  Not that I am asking the same question but still need clarification.  By looking at the df output

                  Filesystem          1K-blocks      Used Available Use% Mounted on
                  /dev/root              475080    315108    123276  72% /
                  devtmpfs                491736        0    491736  0% /dev
                  tmpfs                  492048        0    492048  0% /dev/shm
                  tmpfs                  492048      484    491564  0% /run
                  tmpfs                  492048        0    492048  0% /sys/fs/cgroup
                  tmpfs                  492048      484    491564  0% /etc/machine-id
                  systemd-1                5638      5184      454  92% /boot
                  tmpfs                  492048        4    492044  0% /tmp
                  systemd-1              2337308      3760  2317164  0% /home
                  tmpfs                  492048        0    492048  0% /var/volatile
                  /dev/mmcblk0p5            1003        21      911  2% /factory
                  /dev/mmcblk0p10        2337308      3760  2317164  0% /home
                  /dev/mmcblk0p7            5638      5184      454  92% /boot

                  So from the Yocto image Linux side, over filling the Edison Linux filesystem at some point will overwrite/corrupt the OTA drive (if in someway shared) or is the OTA a completely separate partition from what we see here in the df output.

                  Reasoning is that if the OTA is absolutely completely separated, it could house the image on the module and a reboot ota could reset to start without fear of the ota ever becoming corrupted from operational usage.

                  • 6. Re: flash to build_16 issue
                    AlexT_Intel

                    It's a separate partition and it's not normally mounted (so you won't see it in the mount or df output), so you can't corrupt it by overfilling any of the mounted partitions. It's not like a hidden directory or something on one of the normally mounted partitions, it's a separate one.

                     

                    So yes,

                    it could house the image on the module and a reboot ota could reset to start without fear of the ota ever becoming corrupted from operational usage

                    and I've actually did something like that a couple of times when testing new packages. I've written the image to the OTA partition, then ran "reboot ota", it reflashed itself and I've got a fresh system. Then I've tested several packages and during the process I overfilled the rootfs partition, so I just ran "reboot ota" and it was fine again. And a couple of similar cases, so yes, it works in this scenario.