On Wed, Jul 30, 2025 at 05:45:27PM +0530, Kanchan Joshi wrote:

On 7/29/2025 8:04 PM, Keith Busch wrote:

quoted

@@ -151,6 +146,11 @@ bool blk_rq_dma_map_iter_start(struct request *req, struct device *dma_dev,
  	memset(&iter->p2pdma, 0, sizeof(iter->p2pdma));
  	iter->status = BLK_STS_OK;
  
+	if (req->rq_flags & RQF_SPECIAL_PAYLOAD)
+		iter->iter.bvec = &req->special_vec;

I am not certain yet, but thinking whether this is enough to handle 
RQF_SPECIAL_PAYLOAD correctly.
Maybe "req->special_vec.bv_len" also need to be included here to 
initialize the iter.

Yeah, I think you're right. I had only tested 'discard' for the special
payload, and it appeared to be okay, but I suspect for the wrong
reasons. I'll fix it up.