Hi, With the help of a static bug finder (EBA - http://www.iagoabal.eu/eba) I have found a potential double-lock in drivers/infiniband/core/umem_odp.c: there seems to be a recursive lock on `umem->odp_data->umem_mutex' in function `ib_umem_odp_map_dma_pages'. This problem may have been introduced by commit c1d383b5785b ("IB/core: dma map/unmap locking optimizations"). If this is indeed a bug, I could try to help with a patch... although it doesn't look trivial to fix at first sight. The trace is as follows: 1. Function `ib_umem_odp_map_dma_pages' takes the lock for the first time in line 589: // see https://github.com/torvalds/linux/blob/v4.9-rc7/drivers/infiniband/core/umem_odp.c#L589 mutex_lock(&umem->odp_data->umem_mutex); 2. Call to function `ib_umem_odp_map_dma_single_page' in line 591; 3. Call to function `invalidate_page_trampoline' in line 486; 4. Call to function `ib_umem_notifier_start_account' in line 160; 5. Lock is taken for the second time in line 46: // see https://github.com/torvalds/linux/blob/v4.9-rc7/drivers/infiniband/core/umem_odp.c#L46 mutex_lock(&item->odp_data->umem_mutex); Hope it helps! -- iago -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
