On Fri, Jun 07, 2019 at 05:41:56AM +0000, Yoshihiro Shimoda wrote: > > bool blk_can_use_iommu_merging(struct request_queue *q, struct device *dev) > > { > > if (!IOMMU_CAN_MERGE_SEGMENTS(dev)) > > return false; > > As Robin mentioned, all IOMMUs can merge segments so that we don't need > this condition, IIUC. However, this should check whether the device is mapped > on iommu by using device_iommu_mapped(). There are plenty of dma_map_ops based drivers that can't merge segments. Examples: - arch/ia64/sn/pci/pci_dma.c - arch/mips/jazz/jazzdma.c - arch/sparc/mm/io-unit.c - arch/sparc/mm/iommu.c - arch/x86/kernel/pci-calgary_64.c Nevermind the diret mapping, swiotlb and other weirdos. > > > blk_queue_virt_boundary(q, IOMMU_PAGE_SIZE(dev)); > > blk_queue_max_segment_size(q, IOMMU_MAX_SEGMENT_SIZE(dev)); > > By the way, I reported an issue [1] and I'm thinking dima_is_direct() environment > (especially for swiotlb) is also needed such max_segment_size changes somehow. > What do you think? > > [1] > https://marc.info/?l=linux-block&m=155954415603356&w=2 That doesn't seem to be related to the segment merging. I'll take a look, but next time please Cc the author of a suspect commit if you already bisect things.