df80e8ffc6
The raid bdev may receive IO requests after the raid layer invokes spdk_bdev_unregister function. This patch moves the raid_bdev->module->stop to the raid_bdev_destruct function. Then there will be no IO after the raid_bdev->module->stop is invoked. Below is a way to reproduce this issue: (1) create a malloc bdev sudo ./scripts/rpc.py bdev_malloc_create --name malloc0 128 4096 (2) create a concat bdev base on the malloc bdev sudo ./scripts/rpc.py bdev_raid_create --name concat0 \ --raid-level concat --base-bdevs malloc0 --strip-size-kb 4 (3) create a lvstore base on the concat bdev sudo ./scripts/rpc.py bdev_lvol_create_lvstore \ --cluster-sz 4194304 --clear-method unmap concat0 lvs0 (4) remove the concat bdev sudo ./scripts/rpc.py bdev_raid_delete concat0 In the step(4), the spdk app will crash because an IO request is sent to the concat bdev after the concat_stop is invoked. Signed-off-by: Peng Yu <yupeng0921@gmail.com> Change-Id: I76d6964ee9528d4590ed86e9c5c28d53e85da32f Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12221 Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com> Community-CI: Mellanox Build Bot Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com> Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> |
||
|---|---|---|
| .. | ||
| accel | ||
| bdev | ||
| blob | ||
| blobfs | ||
| env_dpdk | ||
| event | ||
| scheduler | ||
| sock | ||
| Makefile | ||