I'm working on supporting unloading the HCA kernel driver on the Linux NFS client while there are active NFS/RDMA mounts. Happily, the Linux NFS server already supports driver unloading. I have prototyped adequate infrastructure to clean up RDMA- related resources (eg, QP and MRs) as part of handling a DEVICE_REMOVAL event. However I'm wondering about long-term DMA-mapped memory. These would be Send and Receive buffers that are mapped and then re-used for each RPC, using the ib_dma_sync_* API. I believe ULPs will have to map these buffers again once a new device is available. But as part of DEVICE_REMOVAL, does the ULP have to explicitly unmap this memory? I'm not seeing an answer in DMA-API-HOWTO.txt. -- Chuck Lever -- 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
