tree: dd9c3716a80121dc50fc2f4187b381591f01764c [path history] [tgz]
  1. blackout-host/
  2. blackout-target/
  3. blobfs-fsck-soft/
  4. minfs-fsck/
  5. BUILD.gn
  6. OWNERS
  7. README.md
tools/blackout/README.md

blackout - power failure testing for the filesystems

TODO(fxbug.dev/34494): write a guide for writing additional power failure tests using this framework.

Running power failure tests

The tests are not built by default. They need to be included in your fx set line -

fx set core.x64 --with //tools/blackout:all
fx build

If you are using real hardware, you will likely want to be using a netbooted environment, in which case you will add the --netboot flag to your fx set line.

fx set core.x64 --with //tools/blackout:all --netboot

The tests require a spare partition to operate on, because they format it (and partitions can't be mounted twice anyway). There are two ways to achieve this, depending on the device being used to test.

qemu

Create a local file to be used as the test partition.

truncate --size=300M blackout.bin

Run the virtual machine with the new block device attached. The additional options disable the qemu writeback cache that is normally automatically used for attached block devices, because it can hide errors.

# shell 1
fx qemu -Nk -m 16384 -- -drive file=$PWD/blackout.bin,index=0,media=disk,cache=directsync

Set up the rest of the environment in the standard way. For example (in separate shells) -

# shell 2
fx serve
# shell 3 (optional)
fx log
# shell 4 (optional)
fx shell

real hardware

Make sure you have --netboot on your fx set line. Set up your test device with the bootloader and zedboot, but don't pave it. When your device is booted into zedboot, run the following commands -

# shell 1
fx netboot
# shell 2
fx serve-updates

As of January 2020, the fx netboot script quits after one serve. This isn't desirable when doing many test runs in a row. A quick edit in the //tools/devshell/netboot script, removing the -1 from the call on the last line, will stop this behavior.

running the test

Next we collect a bit of information we need to run the test. Use lsblk on the device to find the block ID for the spare partition you created. If you are doing this on real hardware, have a relay, and would like to do a hard reboot as part of the test, find the path to the relay device on the host machine.

Run the test with the following command on the host machine -

./out/default/host_x64/<test_name_with_underscores>_host "/block/device" "$(fx get-device-addr)"

If you have a power relay to cut power to real hardware, add --relay /dev/<relay-device>.

This will run the test once. Blackout tests also have the ability to run tests multiple times in a row by adding the -i <iterations> flag. This will run the test iterations times, aggregating statistics on the test runs, particularly the number of failures. The -f flag can also be added when -i is provided, which will cause blackout to exit on the first verification failure it encounters. This is useful for manual investigation of failures - blackout won‘t reformat the drive again, leaving it in it’s corrputed state.