There apparently is no reason to require 3 queued buffers to call
streamon() for the RkISP1 as the driver operates with a scratch buffer
where frames can be directed to if there's no available buffer provided
by userspace.
Reduce the number of required buffers to 1 to allow applications to
operate with a single queued buffer.
Tested with libcamera, by operating with a single capture request. The
same request (and associated capture buffer) gets recycled once
completed. This of course causes a frame rate drop but doesn't hinder
operations.
Signed-off-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://lore.kernel.org/r/20241007124225.63463-1-jacopo.mondi@ideasonboard.com
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
 #define RKISP1_SP_DEV_NAME     RKISP1_DRIVER_NAME "_selfpath"
 #define RKISP1_MP_DEV_NAME     RKISP1_DRIVER_NAME "_mainpath"
 
-#define RKISP1_MIN_BUFFERS_NEEDED 3
-
 enum rkisp1_plane {
        RKISP1_PLANE_Y  = 0,
        RKISP1_PLANE_CB = 1,
        q->ops = &rkisp1_vb2_ops;
        q->mem_ops = &vb2_dma_contig_memops;
        q->buf_struct_size = sizeof(struct rkisp1_buffer);
-       q->min_queued_buffers = RKISP1_MIN_BUFFERS_NEEDED;
+       q->min_queued_buffers = 1;
        q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC;
        q->lock = &node->vlock;
        q->dev = cap->rkisp1->dev;