Scripts to program a Kria platform via JTAG
Assets and scripts to support Linux ssh based recovery mode for Kria
Lava integration notes, files and example jobs
- SD card updating is slow but this is the speed of the target to that device
- To reduce wear recovery does read/compare and write if needed for each 64M chuck
- This should speed things up also in the normal case of writing an SD that is almost exactly the same as what is there already
- Read speed for flash based devices should be faster than write speed
- However, right now the read speed is not that fast
- I was seeing ~10 S to read 64MB from SD card
- I am using a PNY 16GB for testing
- The Sandisk Ultra 32 GB cards may be faster
- I was seeing ~3 S to read 64MB from USB stick
- To reduce wear recovery does read/compare and write if needed for each 64M chuck
- fix emmc for production SOM
- hacking DTB in u-boot DID NOT fix this. Needs investigation.
- Isolate boot failure: firmware, board, or test setup
- Better document OE build process for recorded assets (recovery-boot.bin cpio etc)
- Some of this is now done
- add OE builds to mk-recovery-assets.sh??
- lock down SSH to known keys only
- make meta-ts job deploy sdcard also
- run 10 iterations of full meta-ts
- Add more LAVA example jobs
- A firmware only job
- Move some of the device dictionary into a device type (?)
- figure out what is wrong with kpartx -d
- Do better disk image feedback in LAVA
- LAVA logs appear line based.
- Disk image feedback is 1 character per 64M chunk
- This means no feedback for long periods
- Add a mode or option to do line based feedback
- Handle file based overlays? rsync deployment?
- add usb utils? sfdisk? mkfs.{ext4,vfat}