diff --git a/src/gallium/drivers/v3d/v3d_context.h b/src/gallium/drivers/v3d/v3d_context.h
index a5275c41243..427fea125f4 100644
--- a/src/gallium/drivers/v3d/v3d_context.h
+++ b/src/gallium/drivers/v3d/v3d_context.h
@@ -280,6 +280,7 @@ struct v3d_ssbo_stateobj {
 struct v3d_job_key {
         struct pipe_surface *cbufs[4];
         struct pipe_surface *zsbuf;
+        struct pipe_surface *bbuf;
 };
 
 enum v3d_ez_state {
@@ -336,10 +337,15 @@ struct v3d_job {
         /* Size of the submit.bo_handles array. */
         uint32_t bo_handles_size;
 
-        /** @{ Surfaces to submit rendering for. */
+        /** @{
+         * Surfaces to submit rendering for.
+         * For blit operations, bbuf is the source surface, and cbufs[0] is
+         * the destination surface.
+         */
         uint32_t nr_cbufs;
         struct pipe_surface *cbufs[4];
         struct pipe_surface *zsbuf;
+        struct pipe_surface *bbuf;
         /** @} */
         /** @{
          * Bounding box of the scissor across all queued drawing.
@@ -648,7 +654,8 @@ void v3d_job_free(struct v3d_context *v3d, struct v3d_job *job);
 struct v3d_job *v3d_get_job(struct v3d_context *v3d,
                             uint32_t nr_cbufs,
                             struct pipe_surface **cbufs,
-                            struct pipe_surface *zsbuf);
+                            struct pipe_surface *zsbuf,
+                            struct pipe_surface *bbuf);
 struct v3d_job *v3d_get_job_for_fbo(struct v3d_context *v3d);
 void v3d_job_set_tile_buffer_size(struct v3d_job *job);
 void v3d_job_add_bo(struct v3d_job *job, struct v3d_bo *bo);
diff --git a/src/gallium/drivers/v3d/v3d_job.c b/src/gallium/drivers/v3d/v3d_job.c
index d608684a5a3..744ed0fbfa6 100644
--- a/src/gallium/drivers/v3d/v3d_job.c
+++ b/src/gallium/drivers/v3d/v3d_job.c
@@ -72,6 +72,8 @@ v3d_job_free(struct v3d_context *v3d, struct v3d_job *job)
                                             job->zsbuf->texture);
                 pipe_surface_reference(&job->zsbuf, NULL);
         }
+        if (job->bbuf)
+                pipe_surface_reference(&job->bbuf, NULL);
 
         if (v3d->job == job)
                 v3d->job = NULL;
@@ -300,6 +302,13 @@ v3d_job_set_tile_buffer_size(struct v3d_job *job)
                         max_bpp = MAX2(max_bpp, surf->internal_bpp);
                 }
         }
+
+        if (job->bbuf) {
+                struct v3d_surface *bsurf = v3d_surface(job->bbuf);
+                assert(job->bbuf->texture->nr_samples <= 1 || job->msaa);
+                max_bpp = MAX2(max_bpp, bsurf->internal_bpp);
+        }
+
         job->internal_bpp = max_bpp;
         STATIC_ASSERT(RENDER_TARGET_MAXIMUM_32BPP == 0);
         tile_size_index += max_bpp;
@@ -321,7 +330,8 @@ struct v3d_job *
 v3d_get_job(struct v3d_context *v3d,
             uint32_t nr_cbufs,
             struct pipe_surface **cbufs,
-            struct pipe_surface *zsbuf)
+            struct pipe_surface *zsbuf,
+            struct pipe_surface *bbuf)
 {
         /* Return the existing job for this FBO if we have one */
         struct v3d_job_key local_key = {
@@ -332,6 +342,7 @@ v3d_get_job(struct v3d_context *v3d,
                         cbufs[3],
                 },
                 .zsbuf = zsbuf,
+                .bbuf = bbuf,
         };
         struct hash_entry *entry = _mesa_hash_table_search(v3d->jobs,
                                                            &local_key);
@@ -363,6 +374,11 @@ v3d_get_job(struct v3d_context *v3d,
                 if (zsbuf->texture->nr_samples > 1)
                         job->msaa = true;
         }
+        if (bbuf) {
+                pipe_surface_reference(&job->bbuf, bbuf);
+                if (bbuf->texture->nr_samples > 1)
+                        job->msaa = true;
+        }
 
         for (int i = 0; i < job->nr_cbufs; i++) {
                 if (cbufs[i])
@@ -399,7 +415,7 @@ v3d_get_job_for_fbo(struct v3d_context *v3d)
         uint32_t nr_cbufs = v3d->framebuffer.nr_cbufs;
         struct pipe_surface **cbufs = v3d->framebuffer.cbufs;
         struct pipe_surface *zsbuf = v3d->framebuffer.zsbuf;
-        struct v3d_job *job = v3d_get_job(v3d, nr_cbufs, cbufs, zsbuf);
+        struct v3d_job *job = v3d_get_job(v3d, nr_cbufs, cbufs, zsbuf, NULL);
 
         if (v3d->framebuffer.samples >= 1)
                 job->msaa = true;
diff --git a/src/gallium/drivers/v3d/v3dx_rcl.c b/src/gallium/drivers/v3d/v3dx_rcl.c
index 7cebd3d0fc0..31d7d2b4cd4 100644
--- a/src/gallium/drivers/v3d/v3dx_rcl.c
+++ b/src/gallium/drivers/v3d/v3dx_rcl.c
@@ -114,7 +114,8 @@ static void
 store_general(struct v3d_job *job,
               struct v3d_cl *cl, struct pipe_surface *psurf,
               int layer, int buffer, int pipe_bit,
-              uint32_t *stores_pending, bool general_color_clear)
+              uint32_t *stores_pending, bool general_color_clear,
+              bool is_blit)
 {
         struct v3d_surface *surf = v3d_surface(psurf);
         bool separate_stencil = surf->separate_stencil && buffer == STENCIL;
@@ -158,8 +159,11 @@ store_general(struct v3d_job *job,
                         store.height_in_ub_or_stride = slice->stride;
                 }
 
+                assert(!is_blit || job->bbuf);
                 if (psurf->texture->nr_samples > 1)
                         store.decimate_mode = V3D_DECIMATE_MODE_ALL_SAMPLES;
+                else if (is_blit && job->bbuf->texture->nr_samples > 1)
+                        store.decimate_mode = V3D_DECIMATE_MODE_4X;
                 else
                         store.decimate_mode = V3D_DECIMATE_MODE_SAMPLE_0;
 
@@ -213,6 +217,14 @@ static void
 v3d_rcl_emit_loads(struct v3d_job *job, struct v3d_cl *cl, int layer)
 {
         uint32_t loads_pending = job->load;
+        uint32_t blit_pending = job->bbuf ? PIPE_CLEAR_COLOR0 : 0;
+
+        assert(!job->bbuf || V3D_VERSION >= 40);
+
+        /* When blitting, no color buffer is loaded; instead the blit source
+         * buffer is loaded.
+         */
+        assert(!job->bbuf || job->load == 0);
 
         for (int i = 0; i < job->nr_cbufs; i++) {
                 uint32_t bit = PIPE_CLEAR_COLOR0 << i;
@@ -229,6 +241,12 @@ v3d_rcl_emit_loads(struct v3d_job *job, struct v3d_cl *cl, int layer)
                              bit, &loads_pending);
         }
 
+        if (blit_pending) {
+                load_general(cl, job->bbuf, RENDER_TARGET_0, layer,
+                             PIPE_CLEAR_COLOR0, &blit_pending);
+                assert(blit_pending == 0);
+        }
+
         if ((loads_pending & PIPE_CLEAR_DEPTHSTENCIL) &&
             (V3D_VERSION >= 40 ||
              (job->zsbuf && job->zsbuf->texture->nr_samples > 1))) {
@@ -313,6 +331,7 @@ v3d_rcl_emit_stores(struct v3d_job *job, struct v3d_cl *cl, int layer)
          * perspective.  Non-MSAA surfaces will use
          * STORE_MULTI_SAMPLE_RESOLVED_TILE_COLOR_BUFFER_EXTENDED.
          */
+        assert(!job->bbuf || job->nr_cbufs == 1);
         for (int i = 0; i < job->nr_cbufs; i++) {
                 uint32_t bit = PIPE_CLEAR_COLOR0 << i;
                 if (!(job->store & bit))
@@ -325,7 +344,7 @@ v3d_rcl_emit_stores(struct v3d_job *job, struct v3d_cl *cl, int layer)
                 }
 
                 store_general(job, cl, psurf, layer, RENDER_TARGET_0 + i, bit,
-                              &stores_pending, general_color_clear);
+                              &stores_pending, general_color_clear, job->bbuf);
         }
 
         if (job->store & PIPE_CLEAR_DEPTHSTENCIL && job->zsbuf &&
@@ -336,20 +355,23 @@ v3d_rcl_emit_stores(struct v3d_job *job, struct v3d_cl *cl, int layer)
                                 store_general(job, cl, job->zsbuf, layer,
                                               Z, PIPE_CLEAR_DEPTH,
                                               &stores_pending,
-                                              general_color_clear);
+                                              general_color_clear,
+                                              false);
                         }
 
                         if (job->store & PIPE_CLEAR_STENCIL) {
                                 store_general(job, cl, job->zsbuf, layer,
                                               STENCIL, PIPE_CLEAR_STENCIL,
                                               &stores_pending,
-                                              general_color_clear);
+                                              general_color_clear,
+                                              false);
                         }
                 } else {
                         store_general(job, cl, job->zsbuf, layer,
                                       zs_buffer_from_pipe_bits(job->store),
                                       job->store & PIPE_CLEAR_DEPTHSTENCIL,
-                                      &stores_pending, general_color_clear);
+                                      &stores_pending, general_color_clear,
+                                      false);
                 }
         }
 
@@ -473,6 +495,10 @@ v3d_setup_render_target(struct v3d_job *job, int cbuf,
 
         struct v3d_surface *surf = v3d_surface(job->cbufs[cbuf]);
         *rt_bpp = surf->internal_bpp;
+        if (job->bbuf) {
+           struct v3d_surface *bsurf = v3d_surface(job->bbuf);
+           *rt_bpp = MAX2(*rt_bpp, bsurf->internal_bpp);
+        }
         *rt_type = surf->internal_type;
         *rt_clamp = V3D_RENDER_TARGET_CLAMP_NONE;
 }