I've seen what appears to be livelocks in the stress_inorder_work() function, and looking at the code it is clear we can have a case where we continually retry acquiring the locks and never check to see if we have passed the specified timeout. This patch reworks that function so we always check the timeout before iterating through the loop again. I believe others may have hit this previously here: https://lore.kernel.org/lkml/895ef450-4fb3-5d29-a6ad-790657106a5a@intel.com/
Unified · Split
| @@ -465,17 +465,18 @@ retry: · 464 unmodified lines | |||
| 465 | ww_mutex_unlock(&locks[order[n]]); | 465 | ww_mutex_unlock(&locks[order[n]]); |
| 466 | 466 | ||
| 467 | if (err == -EDEADLK) { | 467 | if (err == -EDEADLK) { |
| 468 | ww_mutex_lock_slow(&locks[order[contended]], &ctx); | 468 | if (!time_after(jiffies, stress->timeout)) { |
| 469 | goto retry; | 469 | ww_mutex_lock_slow(&locks[order[contended]], &ctx); |
| 470 | goto retry; | ||
| 471 | } | ||
| 470 | } | 472 | } |
| 471 | 473 | ||
| 474 | ww_acquire_fini(&ctx); | ||
| 472 | if (err) { | 475 | if (err) { |
| 473 | pr_err_once("stress (%s) failed with %d\n", | 476 | pr_err_once("stress (%s) failed with %d\n", |
| 474 | __func__, err); | 477 | __func__, err); |
| 475 | break; | 478 | break; |
| 476 | } | 479 | } |
| 477 | |||
| 478 | ww_acquire_fini(&ctx); | ||
| 479 | } while (!time_after(jiffies, stress->timeout)); | 480 | } while (!time_after(jiffies, stress->timeout)); |
| 480 | 481 | ||
| 481 | kfree(order); | 482 | kfree(order); |