[PATCH v3 RESEND] drm/mediatek: dsi: Store driver data before invoking mipi_dsi_host_register

[PATCH v3 RESEND] drm/mediatek: dsi: Store driver data before invoking mipi_dsi_host_register

[Luca Leonardo Scorcia]

The call to mipi_dsi_host_register triggers a callback to mtk_dsi_bind,
which uses dev_get_drvdata to retrieve the mtk_dsi struct, so this
structure needs to be stored inside the driver data before invoking it.

As drvdata is currently uninitialized it leads to a crash when
registering the DSI DRM encoder right after acquiring
the mode_config.idr_mutex, blocking all subsequent DRM operations.

Fixes the following crash during mediatek-drm probe (tested on Xiaomi
Smart Clock x04g):

Unable to handle kernel NULL pointer dereference at virtual address
 0000000000000040
[...]
Modules linked in: mediatek_drm(+) drm_display_helper cec drm_client_lib
 drm_dma_helper drm_kms_helper panel_simple
[...]
Call trace:
 drm_mode_object_add+0x58/0x98 (P)
 __drm_encoder_init+0x48/0x140
 drm_encoder_init+0x6c/0xa0
 drm_simple_encoder_init+0x20/0x34 [drm_kms_helper]
 mtk_dsi_bind+0x34/0x13c [mediatek_drm]
 component_bind_all+0x120/0x280
 mtk_drm_bind+0x284/0x67c [mediatek_drm]
 try_to_bring_up_aggregate_device+0x23c/0x320
 __component_add+0xa4/0x198
 component_add+0x14/0x20
 mtk_dsi_host_attach+0x78/0x100 [mediatek_drm]
 mipi_dsi_attach+0x2c/0x50
 panel_simple_dsi_probe+0x4c/0x9c [panel_simple]
 mipi_dsi_drv_probe+0x1c/0x28
 really_probe+0xc0/0x3dc
 __driver_probe_device+0x80/0x160
 driver_probe_device+0x40/0x120
 __device_attach_driver+0xbc/0x17c
 bus_for_each_drv+0x88/0xf0
 __device_attach+0x9c/0x1cc
 device_initial_probe+0x54/0x60
 bus_probe_device+0x34/0xa0
 device_add+0x5b0/0x800
 mipi_dsi_device_register_full+0xdc/0x16c
 mipi_dsi_host_register+0xc4/0x17c
 mtk_dsi_probe+0x10c/0x260 [mediatek_drm]
 platform_probe+0x5c/0xa4
 really_probe+0xc0/0x3dc
 __driver_probe_device+0x80/0x160
 driver_probe_device+0x40/0x120
 __driver_attach+0xc8/0x1f8
 bus_for_each_dev+0x7c/0xe0
 driver_attach+0x24/0x30
 bus_add_driver+0x11c/0x240
 driver_register+0x68/0x130
 __platform_register_drivers+0x64/0x160
 mtk_drm_init+0x24/0x1000 [mediatek_drm]
 do_one_initcall+0x60/0x1d0
 do_init_module+0x54/0x240
 load_module+0x1838/0x1dc0
 init_module_from_file+0xd8/0xf0
 __arm64_sys_finit_module+0x1b4/0x428
 invoke_syscall.constprop.0+0x48/0xc8
 do_el0_svc+0x3c/0xb8
 el0_svc+0x34/0xe8
 el0t_64_sync_handler+0xa0/0xe4
 el0t_64_sync+0x198/0x19c
Code: 52800022 941004ab 2a0003f3 37f80040 (29005a80)
---[ end trace 0000000000000000 ]---

Fixes: e4732b590a77 ("drm/mediatek: dsi: Register DSI host after acquiring clocks and PHY")
Signed-off-by: Luca Leonardo Scorcia <l.scorcia@gmail.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: CK Hu <ck.hu@mediatek.com>
---
v3: Also initialize irq_wait_queue before calling platform_set_drvdata
v2: Added Fixes tag

 drivers/gpu/drm/mediatek/mtk_dsi.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/mediatek/mtk_dsi.c b/drivers/gpu/drm/mediatek/mtk_dsi.c
index d7726091819c..acee2227275b 100644
--- a/drivers/gpu/drm/mediatek/mtk_dsi.c
+++ b/drivers/gpu/drm/mediatek/mtk_dsi.c
@@ -1232,6 +1232,11 @@ static int mtk_dsi_probe(struct platform_device *pdev)
 
 	dsi->host.ops = &mtk_dsi_ops;
 	dsi->host.dev = dev;
+
+	init_waitqueue_head(&dsi->irq_wait_queue);
+
+	platform_set_drvdata(pdev, dsi);
+
 	ret = mipi_dsi_host_register(&dsi->host);
 	if (ret < 0)
 		return dev_err_probe(dev, ret, "Failed to register DSI host\n");
@@ -1243,10 +1248,6 @@ static int mtk_dsi_probe(struct platform_device *pdev)
 		return dev_err_probe(&pdev->dev, ret, "Failed to request DSI irq\n");
 	}
 
-	init_waitqueue_head(&dsi->irq_wait_queue);
-
-	platform_set_drvdata(pdev, dsi);
-
 	dsi->bridge.of_node = dev->of_node;
 	dsi->bridge.type = DRM_MODE_CONNECTOR_DSI;
 
-- 
2.43.0

Claude review: drm/mediatek: dsi: Store driver data before invoking mipi_dsi_host_register

[Claude Code Review Bot]

Overall Series Review

Subject: drm/mediatek: dsi: Store driver data before invoking mipi_dsi_host_register
Author: Luca Leonardo Scorcia <l.scorcia@gmail.com>
Patches: 1
Reviewed: 2026-03-22T03:36:31.537150

---

This is a single-patch fix for a real NULL pointer dereference crash in the MediaTek DSI driver during probe. The analysis is correct and the fix is straightforward. The commit `e4732b590a77` moved `mipi_dsi_host_register` earlier in the probe sequence but left `platform_set_drvdata` after it, causing `dev_get_drvdata` in the `mtk_dsi_bind` callback path to return NULL.

The patch is well-motivated with a clear crash trace, has already been reviewed by the MediaTek DRM maintainers (CK Hu) and Collabora's Angelo Del Regno, and is on its v3 iteration.

**Verdict: Looks good. Ready to merge.**

---
Generated by Claude Code Patch Reviewer

Claude review: drm/mediatek: dsi: Store driver data before invoking mipi_dsi_host_register

[Claude Code Review Bot]

Patch Review

**Problem analysis:** Correct. The call chain is:
`mipi_dsi_host_register` → `panel_simple_dsi_probe` → `mipi_dsi_attach` → `mtk_dsi_host_attach` → `component_add` → `mtk_dsi_bind` → `dev_get_drvdata` (returns NULL) → crash in `drm_simple_encoder_init`.

**Fix correctness:**

1. **Moving `platform_set_drvdata(pdev, dsi)` before `mipi_dsi_host_register`** — correct and necessary. This is the core fix.

2. **Moving `init_waitqueue_head(&dsi->irq_wait_queue)` before `mipi_dsi_host_register`** — also correct. Since the bind path can now execute during `mipi_dsi_host_register`, and the IRQ handler (`mtk_dsi_irq`) uses `dsi->irq_wait_queue` via `wake_up_interruptible`, the waitqueue must be initialized before any code path could potentially trigger it. While the IRQ is requested *after* `mipi_dsi_host_register`, initializing the waitqueue early is defensive and costs nothing.

3. **Error path consideration:** After this patch, if `mipi_dsi_host_register` fails, `platform_set_drvdata` will have been called but the function returns an error and the device will be torn down by devm, so no issue. If `devm_request_irq` fails, the existing cleanup (`mipi_dsi_host_unregister`) is still correct — `drvdata` being set is harmless.

**No issues found.** The Fixes tag correctly references the commit that introduced the regression. The patch is minimal and focused.

---
Generated by Claude Code Patch Reviewer