5.15-stable review patch. If anyone has any objections, please let me know.
------------------
From: Ivaylo Dimitrov <redacted>
commit e81250ec6b69248b00d38c523dc6a13efaf38aab upstream.
The OMAP SSI driver uses a synthetic HSI controller device allocated via
hsi_alloc_controller(), which does not go through the normal OF/platform
device initialization path.
As a result, the embedded struct device does not have a DMA mask
initialized by default.
After recent DMA API hardening changes, dma_map_sg() and related helpers
now require a valid dma_mask to be present, otherwise the driver may
crash or trigger warnings when attempting DMA mapping operations.
Fix this by explicitly initializing the DMA mask for the SSI controller
device and setting a 32-bit DMA mask, which matches the hardware
capabilities.
Cc: <redacted>
Fixes: f959dcd6ddfd ("dma-direct: Fix potential NULL pointer dereference")
Reported-by: Merlijn Wajer <redacted>
Closes: https://lore.kernel.org/linux-omap/4ed95c71-2066-6b4c-ad1b-53ef02d79d53@wizzup.org/ (local)
Signed-off-by: Ivaylo Dimitrov <redacted>
Link: https://patch.msgid.link/20260724130522.706480-1-ivo.g.dimitrov.75@gmail.com (local)
Signed-off-by: Sebastian Reichel <redacted>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
drivers/hsi/controllers/omap_ssi_core.c | 6 ++++++
1 file changed, 6 insertions(+)
--- a/drivers/hsi/controllers/omap_ssi_core.c
+++ b/drivers/hsi/controllers/omap_ssi_core.c
@@ -509,6 +509,12 @@ static int ssi_probe(struct platform_dev
pm_runtime_enable(&pd->dev);
+ ssi->device.dma_mask = &ssi->device.coherent_dma_mask;
+
+ err = dma_set_mask_and_coherent(&ssi->device, DMA_BIT_MASK(32));
+ if (err)
+ goto out2;
+
err = ssi_hw_init(ssi);
if (err < 0)
goto out2;