zbd/012: avoid fio stop by I/O scheduler set failure

The test case zbd/012 fails occasionally due to a sudden fio stop. At
the fio stop, fio outputs the following error message:

 fio: unable to set io scheduler to none
 fio: pid=119786, err=22/file:backend.c:1485, func=iosched_switch, error=Invalid argument

The test case specifies --scheduler=none option to the fio command. At
the workload start, fio sets I/O scheduler of the test target device to
"none" by writing to the sysfs "queue/scheduler" attribute. Subsequently,
fio verifies this action by reading the attribute, expecting to find the
string "[none]". However, it instead finds "[mq-deadline]", leading to
the error.

The test case runs another process to switch the I/O scheduler of the
test target device between "none" and "mq-deadline" every 0.1 seconds.
When the switch to "mq-deadline" occurs in the interim between the sysfs
attribute write and read by fio, fio encounters the "[mq-deadline]"
value, resulting in the error.

To avoid the failure, drop the --scheduler=none option from the fio
command in the test case zbd/012. I confirmed that the test case still
can recreate the hang with this fix, using the kernel v6.12.

Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
1 file changed
tree: bf17207748dc556fbc3873cd3fc86afd1d2b6b9d
  1. .github/
  2. common/
  3. contrib/
  4. Documentation/
  5. LICENSES/
  6. src/
  7. tests/
  8. .dir-locals.el
  9. .gitignore
  10. check
  11. CONTRIBUTING.md
  12. Makefile
  13. new
  14. README.md
README.md

blktests

Build Status

blktests is a test framework for the Linux kernel block layer and storage stack. It is inspired by the xfstests filesystem testing framework. It was originally written by Omar Sandoval and announced in 2017.

Getting Started

The dependencies are minimal, but make sure you have them installed:

  • bash (>= 4.2)
  • GNU coreutils
  • GNU awk
  • util-linux
  • fio
  • gcc
  • make
  • systemd-udev (udevadm)

Some tests require the following:

  • e2fsprogs, xfsprogs, f2fs-tools and btrfs-progs
  • nvme-cli
  • multipath-tools (Debian, openSUSE, Arch Linux) or device-mapper-multipath (Fedora)
  • nbd-client and nbd-server (Debian) or nbd (Fedora, openSUSE, Arch Linux)
  • dmsetup (Debian) or device-mapper (Fedora, openSUSE, Arch Linux)
  • rublk (cargo install --version=^0.1 rublk) for ublk test
  • python3, ethtool, iproute2 for nvme-tcp zero-copy offload test

Build blktests with make. Optionally, install it to a known location with make install (/usr/local/blktests by default, but this can be changed by passing DESTDIR and/or prefix).

Add the list of block devices you want to test on in a file named config (note: these tests are potentially destructive):

TEST_DEVS=(/dev/nvme0n1 /dev/sdb)

And as root, run the default set of tests with ./check.

Do not add anything to the TEST_DEVS array containing data that you want to keep.

See here for more detailed information on configuration and running tests.

Adding Tests

New test cases are welcomed when,

  • a bug in block layer or storage stack is found and the new test case confirms fix of the bug,
  • a new feature is introduced in block layer or storage stack, and the new test cases confirm that the feature is working well, or,
  • the new test cases extend coverage of block layer and storage stack code.

The ./new script creates a new test from a template. The generated template contains more detailed documentation. The ./new script itself can be referred to as a document. It describes variables and functions that test cases should implement, global variables that test cases can refer and coding guidelines.

Patches to linux-block@vger.kernel.org and pull requests on GitHub are both accepted. See here for more information on contributing.