[PATCH v2 1/2] iio: buffer-dmaengine: fix sg entry iteration when building dma_vecs

From: Michael Hennerich <[email protected]>

iio_dmaengine_buffer_submit_block() counts scatterlist entries with
sg_nents_for_len(), which walks the CPU-side lengths (sg->length), but
then consumes the DMA-side fields (sg_dma_address()/sg_dma_len()).
After dma_map_sgtable() the two views may differ: an IOMMU can coalesce
the mapping so that only the first sgt->nents entries carry valid DMA
addresses, with nents < orig_nents.

On x86 with an IOMMU enabled, a DMABUF block backed by two 1 MiB
system-heap chunks maps to a single 2 MiB IOVA range. The CPU-side
count is 2, so the loop reads one entry past the mapped set and emits a
garbage vec ({addr = ~0, len = 0}). The DMA engine driver rejects the
vec array (prep returns NULL), the fence is signalled with -ENOMEM,
which a userspace poller cannot observe, and the block is left in
ACTIVE state so every further enqueue of it fails with -EBUSY. The
visible symptom is a stream of zero-filled blocks followed by a wedged
buffer.

Platforms without an IOMMU never hit this because nents == orig_nents.

Size the vec array with sgt->nents, i.e. the DMA-mapped view, and stop
the fill loop once bytes_used is covered - which is allowed to be
smaller than the block size - passing the number of vecs actually
filled to dmaengine_prep_peripheral_dma_vec().

One vec per mapped entry is enough since coalescing can only ever
reduce the number of entries. A single mapped entry longer than the
device's maximum segment size would need more than one, but the DMA API
already assumes no single segment exceeds it [1], and splitting a vec
down to the hardware descriptor size is the DMA engine driver's job -
which both current .device_prep_peripheral_dma_vec() implementations
do.

[1]: commit ab2cbeb0ed30 ("iommu/dma: Handle SG length overflow better")

Assisted-by: Claude:claude-fable-5
Fixes: 7a86d469983a ("iio: buffer-dmaengine: Support new DMABUF based userspace API")
Signed-off-by: Michael Hennerich <[email protected]>
Signed-off-by: Nuno Sá <[email protected]>

---

Note the Signed-off-by is just because I'm carrying Michael's patch!
---
 drivers/iio/buffer/industrialio-buffer-dmaengine.c | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/iio/buffer/industrialio-buffer-dmaengine.c b/drivers/iio/buffer/industrialio-buffer-dmaengine.c
index ecc02a427b92..376486be3f55 100644
--- a/drivers/iio/buffer/industrialio-buffer-dmaengine.c
+++ b/drivers/iio/buffer/industrialio-buffer-dmaengine.c
@@ -104,10 +104,13 @@ static int iio_dmaengine_buffer_submit_block(struct iio_dma_buffer_queue *queue,
 	if (block->sg_table) {
 		unsigned long flags;
 
-		sgl = block->sg_table->sgl;
-		nents = sg_nents_for_len(sgl, block->bytes_used);
-		if (nents < 0)
-			return nents;
+		/*
+		 * Only the first sgt->nents entries carry a valid
+		 * sg_dma_address()/sg_dma_len() pair as mapping the table may
+		 * have coalesced entries, in which case nents is smaller than
+		 * orig_nents.
+		 */
+		nents = block->sg_table->nents;
 
 		vecs = kmalloc_array(nents, sizeof(*vecs), GFP_ATOMIC);
 		if (!vecs)
@@ -115,7 +118,8 @@ static int iio_dmaengine_buffer_submit_block(struct iio_dma_buffer_queue *queue,
 
 		len_total = block->bytes_used;
 
-		for (i = 0; i < nents; i++) {
+		sgl = block->sg_table->sgl;
+		for (i = 0; i < nents && len_total; i++) {
 			vecs[i].addr = sg_dma_address(sgl);
 			vecs[i].len = min(sg_dma_len(sgl), len_total);
 			len_total -= vecs[i].len;
@@ -123,6 +127,8 @@ static int iio_dmaengine_buffer_submit_block(struct iio_dma_buffer_queue *queue,
 			sgl = sg_next(sgl);
 		}
 
+		nents = i;
+
 		if (block->cyclic)
 			flags = DMA_PREP_REPEAT;
 		else

-- 
2.55.0