Re: [PATCH v1 3/3] ASoC: amd: acp: Add legacy audio driver support for Rembrandt platform

From: Christophe JAILLET
Date: Sun Jul 31 2022 - 11:41:39 EST


Hi,

this patch has already reached -next, but a few nit below.

Le 07/07/2022 à 18:11, V sujith kumar Reddy a écrit :
Add i2s and dmic support for Rembrandt platform,
Add machine support for nau8825, max98360 and rt5682s,rt1019 codec
in legacy driver for rembrandt platform.
Here codec is in a slave mode.

Signed-off-by: V sujith kumar Reddy <Vsujithkumar.Reddy@xxxxxxx>
---
sound/soc/amd/acp/Kconfig | 11 +
sound/soc/amd/acp/Makefile | 2 +
sound/soc/amd/acp/acp-i2s.c | 137 ++++++++-
sound/soc/amd/acp/acp-legacy-mach.c | 32 +++
sound/soc/amd/acp/acp-mach-common.c | 86 +++++-
sound/soc/amd/acp/acp-mach.h | 6 +
sound/soc/amd/acp/acp-pci.c | 6 +
sound/soc/amd/acp/acp-platform.c | 16 +-
sound/soc/amd/acp/acp-rembrandt.c | 401 +++++++++++++++++++++++++++
sound/soc/amd/acp/amd.h | 62 ++++-
sound/soc/amd/acp/chip_offset_byte.h | 28 ++
11 files changed, 781 insertions(+), 6 deletions(-)
create mode 100644 sound/soc/amd/acp/acp-rembrandt.c


[...]

diff --git a/sound/soc/amd/acp/acp-rembrandt.c b/sound/soc/amd/acp/acp-rembrandt.c
new file mode 100644
index 000000000000..2b57c0ca4e99
--- /dev/null
+++ b/sound/soc/amd/acp/acp-rembrandt.c
@@ -0,0 +1,401 @@
+// SPDX-License-Identifier: (GPL-2.0-only OR BSD-3-Clause)
+//
+// This file is provided under a dual BSD/GPLv2 license. When using or
+// redistributing this file, you may do so under either license.

These lines are useless. There is already a SPDX-License-Identifier just above.

+//
+// Copyright(c) 2022 Advanced Micro Devices, Inc.
+//
+// Authors: Ajit Kumar Pandey <AjitKumar.Pandey@xxxxxxx>
+// V sujith kumar Reddy <Vsujithkumar.Reddy@xxxxxxx>
+/*
+ * Hardware interface for Renoir ACP block
+ */
+

[...]

+static int rembrandt_audio_probe(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ struct acp_chip_info *chip;
+ struct acp_dev_data *adata;
+ struct resource *res;
+
+ chip = dev_get_platdata(&pdev->dev);
+ if (!chip || !chip->base) {
+ dev_err(&pdev->dev, "ACP chip data is NULL\n");
+ return -ENODEV;
+ }
+
+ if (chip->acp_rev != ACP6X_DEV) {
+ dev_err(&pdev->dev, "Un-supported ACP Revision %d\n", chip->acp_rev);
+ return -ENODEV;
+ }
+
+ rmb_acp_init(chip->base);

Should rmb_acp_deinit() be called if an error occurs below?
Or a devm_add_action_or_reset() + .remove() simplification?

(this is called in the .remove() function)

+
+ adata = devm_kzalloc(dev, sizeof(struct acp_dev_data), GFP_KERNEL);
+ if (!adata)
+ return -ENOMEM;
+
+ res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "acp_mem");
+ if (!res) {
+ dev_err(&pdev->dev, "IORESOURCE_MEM FAILED\n");
+ return -ENODEV;
+ }
+
+ adata->acp_base = devm_ioremap(&pdev->dev, res->start, resource_size(res));
+ if (!adata->acp_base)
+ return -ENOMEM;
+
+ res = platform_get_resource_byname(pdev, IORESOURCE_IRQ, "acp_dai_irq");
+ if (!res) {
+ dev_err(&pdev->dev, "IORESOURCE_IRQ FAILED\n");
+ return -ENODEV;
+ }
+
+ adata->i2s_irq = res->start;
+ adata->dev = dev;
+ adata->dai_driver = acp_rmb_dai;
+ adata->num_dai = ARRAY_SIZE(acp_rmb_dai);
+ adata->rsrc = &rsrc;
+
+ adata->machines = snd_soc_acpi_amd_rmb_acp_machines;
+ acp_machine_select(adata);
+
+ dev_set_drvdata(dev, adata);
+ acp6x_enable_interrupts(adata);
+ acp_platform_register(dev);
+
+ return 0;
+}
+
+static int rembrandt_audio_remove(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ struct acp_dev_data *adata = dev_get_drvdata(dev);
+ struct acp_chip_info *chip;
+
+ chip = dev_get_platdata(&pdev->dev);
+ if (!chip || !chip->base) {
+ dev_err(&pdev->dev, "ACP chip data is NULL\n");
+ return -ENODEV;
+ }

These tests and dev_err and return look useless.
The same is already tested at the biginning of the probe and if it fails, the probe will fail, right?

+
+ rmb_acp_deinit(chip->base);
+
+ acp6x_disable_interrupts(adata);
+ acp_platform_unregister(dev);
+ return 0;
+}

[...]