| Message ID | 20230620133711.22840-6-dwagner@suse.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | nvme-fc: Fix blktests hangers | expand |
diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c index aa2911f07c6c..6f5cfa47fee5 100644 --- a/drivers/nvme/host/fc.c +++ b/drivers/nvme/host/fc.c @@ -2413,8 +2413,6 @@ nvme_fc_ctrl_free(struct kref *ref) nvme_fc_rport_put(ctrl->rport); ida_free(&nvme_fc_ctrl_cnt, ctrl->cnum); - if (ctrl->ctrl.opts) - nvmf_free_options(ctrl->ctrl.opts); kfree(ctrl); } @@ -3568,8 +3566,6 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts, cancel_work_sync(&ctrl->ctrl.reset_work); cancel_delayed_work_sync(&ctrl->connect_work); - ctrl->ctrl.opts = NULL; - /* initiate nvme ctrl ref counting teardown */ nvme_uninit_ctrl(&ctrl->ctrl);
Since the initial additional of the FC transport e399441de911 ("nvme-fabrics: Add host support for FC transport"), the transport also freed the options. Since nvme_free_ctrl() is freeing the options too commit de41447aac03 ("nvme-fc: avoid memory corruption caused by calling nvmf_free_options() twice") was added to avoid double frees. With the change to make the initial connection attempt synchronous again, the life time of all object is known also in the error case. All resources will be freed in the same context. The FC transport should not free the options as the generic auth code is relying to be able to read the options even in the shutdown path (see nvme_auth_free is calling ctrl_max_dhchaps which relies on opts being a valid pointer). TCP and RDMA also avoid freeing the options, so make the FC transport behave the same. Signed-off-by: Daniel Wagner <dwagner@suse.de> --- drivers/nvme/host/fc.c | 4 ---- 1 file changed, 4 deletions(-)