[LU-6128] LNet: Problem handling error when lnet_check_routes() fail.

• Status: Resolved
• Project: Lustre
• Fix Version/s: Lustre 2.7.0
• Type: Bug
• Priority: Critical
• Reporter: Amir Shehata (Inactive)
• Assignee: Amir Shehata (Inactive)
• Resolution: Fixed
• Labels: HB
• Severity: 3

Description

In LNetCtl():

                LNET_MUTEX_LOCK(&the_lnet.ln_api_mutex);
                rc = lnet_add_route(config->cfg_net,
                                    config->cfg_config_u.cfg_route.rtr_hop,
                                    config->cfg_nid,
                                    config->cfg_config_u.cfg_route.
                                        rtr_priority);
                LNET_MUTEX_UNLOCK(&the_lnet.ln_api_mutex);
                return (rc != 0) ? rc : lnet_check_routes();

If lnet_check_routes() returns error, it means the route just added was invalid and must be removed, see LNetNIInit(). It's wrong to just return the error to user space:
1. User space would treat the error as a failure to add the route, but the route was indeed added already.
2. Two conflicting routes now live in the routing table.

Comments

[Isaac Huang (Inactive)]

I think the correct action would be to remove the route just added and return the error from lnet_check_routes().

[Gerrit Updater]

Amir Shehata ([e-mail]) uploaded a new patch: http://review.whamcloud.com/13445
Subject: LU-6128 lnet: handle lnet_check_routes() errors
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: f628bdd8d240b135c92996a910ba200e0c7279e1

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/13445/
Subject: LU-6128 lnet: handle lnet_check_routes() errors
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: b53998fdcc70c2f58bf436c318a0165f0f5ce224

[Peter Jones]

Landed for 2.7