Hi, all, While working on a bugfix on RTRS[1], I noticed there are quite a few other drivers have the same problem, due to the fact dma_map_sg return 0 on error, not like most of the cases, return negative value for error. I "grep -A 5 dma_map_sg' in kernel tree, and audit/fix the one I feel is buggy, hence this patchset. As suggested by Christoph Hellwig, I now send the patches per subsystem, this is for infiniband subsystem. The first one for mthca, leon mentioned it's too old to change, I still keep it in the hope others have a different opinion. The second patch change the return value follow the patch from Christoph Hellwig, we might want to change the type for in the drivers which calls dma_map_sg or ib_dma_map_sg to unsigned int to. Thanks! [1] https://lore.kernel.org/linux-rdma/20220818105355.110344-1-haris.iqbal@xxxxxxxxx/T/#t Jack Wang (2): infiniband/mthca: Fix dma_map_sg error check RDMA: dma-mapping: Return an unsigned int from ib_dma_map_sg{,_attrs} drivers/infiniband/core/device.c | 2 +- drivers/infiniband/hw/mthca/mthca_memfree.c | 7 ++++--- include/rdma/ib_verbs.h | 6 +++--- 3 files changed, 8 insertions(+), 7 deletions(-) -- 2.34.1
