[PATCH v2 0/2] iio: buffer-dmaengine: fix dma_vec building for coalesced sg tables

Patch 1 fixes iio_dmaengine_buffer_submit_block() mixing up the CPU and the
DMA view of a DMABUF's scatterlist: it counted entries with
sg_nents_for_len() (CPU lengths) but consumed sg_dma_address()/sg_dma_len(),
which are only valid for the first sgt->nents entries. With an IOMMU
coalescing the mapping, the loop walks past the mapped set and hands a
garbage vec to the DMA engine, which wedges the buffer.

Patch 2 is the cleanup Jonathan spotted while reviewing v1: the max_size
computation at the top of the same function is dead, as only the fileio
branch consumes it and that branch computes it again where it is used.

---
Changes in v2:
- Patch 1:
  - Size the vec array with sgt->nents instead of sg_nents_for_dma().
  - Cut the new comment down to the invariant a reader needs.
  - Moved 'nents = i' up to right after the fill loop.
  - MOved `sgl = block->sg_table->sgl` to the place where we need sgl.
- Patch 2:
  - New patch dropping the dead max_size computation at the top of
    iio_dmaengine_buffer_submit_block() (Jonathan).
- Link to v1: https://patch.msgid.link/20260818-iio-buffer-dmabuf-iommu-fic-v1-1-4ff1e44a5073@analog.com

---
Michael Hennerich (1):
      iio: buffer-dmaengine: fix sg entry iteration when building dma_vecs

Nuno Sá (1):
      iio: buffer-dmaengine: drop dead max_size computation

 drivers/iio/buffer/industrialio-buffer-dmaengine.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)
---
base-commit: b756b143e5391151e577ae645b1378a43f93c2f5
change-id: 20260818-iio-buffer-dmabuf-iommu-fic-1b281f15e5a4
--

Thanks!
- Nuno Sá