On 6/13/24 09:40, Cyril Hrubis wrote:
> From: Cyril Hrubis <chrubis@xxxxxxx>
>
> Signed-off-by: Cyril Hrubis <chrubis@xxxxxxx>
thanks for the test, little detail about the issue in the commit
log is always helpful in long run ...
[...]
> +test() {
> + local loop_dev;
> + echo "Running ${TEST_NAME}"
> +
> + mkdir "$TMPDIR/tmpfs"
> + mount -t tmpfs testfs "$TMPDIR/tmpfs"
> + dd if=/dev/zero of="$TMPDIR/tmpfs/disk.img" bs=1M count=100 &> /dev/null
> +
> + if ! loop_dev="$(losetup -f --show "$TMPDIR/tmpfs/disk.img")"; then
> + return 1
> + fi
> +
> + mkfs.ext2 /dev/loop0 &> /dev/null
> +
> + errors=$(_dmesg_since_test_start |grep "operation not supported error, dev" |wc -l)
nit:- can we use grep -c by any chance instead of wc -l ?
errors=$(_dmesg_since_test_start | grep -c "operation not supported
error, dev")
if not please ignore this comment ...
> +
> + losetup -d "$loop_dev"
> + umount "$TMPDIR/tmpfs"
> +
> + echo "Found $errors error(s) in dmesg"
> +
> + echo "Test complete"
> +}
>
we are looking for REQ_OP_WRITE_ZEORES failure i.e. string
"WRITE_ZEROES" on the
loop device "loop0" that will come from :
blk_complete_reqs()
blk_mq_end_request()
blk_update_request()
blk_print_req_error()
blk_op_str() <---
else we might take random errors into consideration that are not related to
write-zeroes and loop0 ?
other than this it looks good to me ...
-ck
