[Forwarding to on-discuss as I doubt the right people are on osol-bugs]
Begin forwarded message:
> From: Chee Wai Yeung <cheewai.yeung at ap.natixis.com>
> Date: June 21, 2009 10:56:27 PM CDT
> To: opensolaris-bugs at opensolaris.org
> Subject: [osol-bugs] Is this a possible bug in vntsdvcc.c?
>
> 180 for (; ; ) {
> 181 /* get the console to be deleted */
> 182 (void) mutex_lock(&groupp->lock);
> 183
> 184 /* clean up any deleted console in the group */
> 185 if (groupp->conspq != NULL) {
> 186 consp = vntsd_que_walk(groupp->conspq,
> 187 (el_func_t)find_clean_cons);
> 188 if (consp == NULL) {
> 189 /* no more cons to delete */
> 190 (void)
> mutex_unlock(&groupp->lock);
> 191 break;
> 192 }
> 193
> 194 /* remove console from the group */
> 195 (void) vntsd_que_rm(&groupp->conspq,
> consp);
> 196 (void) mutex_unlock(&groupp->lock);
> 197
> 198 /* clean up the console */
> 199 cleanup_cons(consp);
> 200 }
> 201
> 202 /* delete group? */
> 203 if (groupp->conspq == NULL) {
> 204 /* no more console in the group delete
> group */
> 205 assert(groupp->vntsd);
> 206
> 207 (void) mutex_lock(&groupp->vntsd->lock);
> 208 (void)
> vntsd_que_rm(&groupp->vntsd->grouppq,
> 209 groupp);
> 210 (void)
> mutex_unlock(&groupp->vntsd->lock);
> 211
> 212 /* clean up the group */
> 213 vntsd_clean_group(groupp);
> 214 break;
> 215 }
> 216 }
>
> Line 182 has call to mutex_lock(&groupp->lock).
>
> If then line 185 returns NULL right away it goes to line 203. The
> block does not call mutex_unlock for the lock &groupp->lock.
>
> Will this later result in a deadlock?
--
Shawn Walker
