I was recently flashing a Debian image from https://rcn-ee.net to internal eMMC using an SD card. I was flashing 6 BeagleBone Green boards and 2 of those showed errors in the console during the process:
[ 336.911666] omap_hsmmc 481d8000.mmc: prep_slave_sg() failed
[ 336.917278] omap_hsmmc 481d8000.mmc: MMC start dma failure
473.56M 84% 9.12MB/s 0:00:08 [ 336.952874] mmcblk1: unknown error -1 sending read/write command, card status 0x900
[ 336.960803] blk_update_request: I/O error, dev mmcblk1, sector 1087424
[ 336.967438] blk_update_request: I/O error, dev mmcblk1, sector 1087432
The copying completed, the boards rebooted and I couldn’t detect any immediate problems running the mal-flashed boards. It does look rather scary, however, so I gave it a quick googling and ended up with this thread:
Adding the line below to the beginning of “/opt/scripts/tools/eMMC/init-eMMC-flasher-v3.sh” resolves the I/O error.
sysctl -w vm.min_free_kbytes=16384
Any guesses as to whether there’s an actual problem with rsync’s memory usage, causing errors in the flashing process? Would it be a good idea to update the upsream script used in rcn-ee.net?