[PATCH v3] drm: renesas: rz-du: mipi_dsi: Fix return path on error

[PATCH v3] drm: renesas: rz-du: mipi_dsi: Fix return path on error

[Chris Brandt]

In case of error, we should unwind correctly.
Switching to using dmam_ instead of dma_ and moving the code earlier
fixes the issue.

Fixes: 6f392f371650 ("drm: renesas: rz-du: Implement MIPI DSI host transfers")
Suggested-by: Pavel Machek <pavel@nabladev.com>
Signed-off-by: Chris Brandt <chris.brandt@renesas.com>
Reviewed-by: Biju Das <biju.das.jz@bp.renesas.com>

---
V2->V3:
 - Removed Acked-by because it was for V1 (Hugo)
 - Added Reviewed-by (Biju)

V1->V2:
 - Need to take into account mipi_dsi_host_register was already called (Geert)
 - Switch to using dmam_alloc_coherent to automate unwinding (Chris)
 - Added Acked-by (hugo)
---
 drivers/gpu/drm/renesas/rz-du/rzg2l_mipi_dsi.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/renesas/rz-du/rzg2l_mipi_dsi.c b/drivers/gpu/drm/renesas/rz-du/rzg2l_mipi_dsi.c
index 29f2b7d24fe5..b908e702b607 100644
--- a/drivers/gpu/drm/renesas/rz-du/rzg2l_mipi_dsi.c
+++ b/drivers/gpu/drm/renesas/rz-du/rzg2l_mipi_dsi.c
@@ -1442,6 +1442,11 @@ static int rzg2l_mipi_dsi_probe(struct platform_device *pdev)
 		return dev_err_probe(dsi->dev, PTR_ERR(dsi->prstc),
 				     "failed to get prst\n");
 
+	dsi->dcs_buf_virt = dmam_alloc_coherent(dsi->dev, RZG2L_DCS_BUF_SIZE,
+						&dsi->dcs_buf_phys, GFP_KERNEL);
+	if (!dsi->dcs_buf_virt)
+		return -ENOMEM;
+
 	platform_set_drvdata(pdev, dsi);
 
 	pm_runtime_enable(dsi->dev);
@@ -1474,11 +1479,6 @@ static int rzg2l_mipi_dsi_probe(struct platform_device *pdev)
 	if (ret < 0)
 		goto err_pm_disable;
 
-	dsi->dcs_buf_virt = dma_alloc_coherent(dsi->host.dev, RZG2L_DCS_BUF_SIZE,
-					       &dsi->dcs_buf_phys, GFP_KERNEL);
-	if (!dsi->dcs_buf_virt)
-		return -ENOMEM;
-
 	return 0;
 
 err_phy:
@@ -1493,8 +1493,6 @@ static void rzg2l_mipi_dsi_remove(struct platform_device *pdev)
 {
 	struct rzg2l_mipi_dsi *dsi = platform_get_drvdata(pdev);
 
-	dma_free_coherent(dsi->host.dev, RZG2L_DCS_BUF_SIZE, dsi->dcs_buf_virt,
-			  dsi->dcs_buf_phys);
 	mipi_dsi_host_unregister(&dsi->host);
 	pm_runtime_disable(&pdev->dev);
 }
-- 
2.52.0

Claude review: drm: renesas: rz-du: mipi_dsi: Fix return path on error

[Claude Code Review Bot]

Overall Series Review

Subject: drm: renesas: rz-du: mipi_dsi: Fix return path on error
Author: Chris Brandt <chris.brandt@renesas.com>
Patches: 1
Reviewed: 2026-05-05T09:27:22.127017

---

This is a single-patch series (v3) that fixes an error-handling bug in the Renesas RZ MIPI DSI probe path. The original code had two problems:

1. **Missing error unwind**: The `dma_alloc_coherent()` call was placed after `mipi_dsi_host_register()`, and on failure it did `return -ENOMEM` directly — skipping the `err_pm_disable` cleanup label, which means `pm_runtime_disable()` and `mipi_dsi_host_unregister()` were never called on that error path.

2. **Leaked DMA allocation**: Since plain `dma_alloc_coherent()` was used, any other error path that went through the `err_phy` / `err_pm_disable` labels would also leak the DMA buffer (though in practice this was unreachable since the alloc came last).

The fix is clean and correct: switch to `dmam_alloc_coherent()` (device-managed) and move it earlier in the probe function — before `platform_set_drvdata()` and `pm_runtime_enable()` — so that on failure, a simple `return -ENOMEM` is sufficient (no cleanup needed at that point), and the buffer is automatically freed on device removal. The corresponding `dma_free_coherent()` in `remove()` is correctly dropped.

The patch has a `Fixes:` tag, a `Reviewed-by` from Biju Das, is on its third revision addressing prior review feedback, and the changelog is well documented. **Recommend: Accept.**

---
Generated by Claude Code Patch Reviewer

Claude review: drm: renesas: rz-du: mipi_dsi: Fix return path on error

[Claude Code Review Bot]

Patch Review

**Bug analysis**: In the original code, `dma_alloc_coherent()` was the last operation in `rzg2l_mipi_dsi_probe()`, placed after `mipi_dsi_host_register()`. On allocation failure it executed:
```c
return -ENOMEM;
```
This bypassed the error cleanup labels (`err_phy`, `err_pm_disable`), leaking the `pm_runtime_enable()` state and the registered MIPI DSI host.

**Fix correctness**:

The allocation is moved to before `platform_set_drvdata()` and `pm_runtime_enable()`:
```c
dsi->dcs_buf_virt = dmam_alloc_coherent(dsi->dev, RZG2L_DCS_BUF_SIZE,
                                        &dsi->dcs_buf_phys, GFP_KERNEL);
if (!dsi->dcs_buf_virt)
    return -ENOMEM;
```
At this point in the probe function, only `devm_*` managed resources have been acquired (clocks, resets, ioremap), so a plain `return` is safe — `devm` teardown handles everything. The switch from `dma_alloc_coherent` to `dmam_alloc_coherent` is the right choice: it's the device-managed variant, so the buffer is automatically freed when the device is unbound, eliminating the need for the explicit `dma_free_coherent()` in `remove()`.

**Remove path**: The `dma_free_coherent()` call is correctly removed from `rzg2l_mipi_dsi_remove()` since `dmam_alloc_coherent()` handles cleanup automatically.

**Device argument consistency**: The original code used `dsi->host.dev` as the device for `dma_alloc_coherent()` / `dma_free_coherent()`, while the new code uses `dsi->dev`. Looking at the probe function, `dsi->host.dev` is assigned to `dsi->dev` later in the function (line 1475: `dsi->host.dev = dsi->dev`), so using `dsi->dev` directly is correct and actually necessary since `dsi->host.dev` isn't set yet at the new call site.

**No issues found.** The patch is minimal, well-targeted, and correctly fixes both the immediate error-path bug and the broader resource management concern by using the managed API.

---
Generated by Claude Code Patch Reviewer