[PATCH -next 0/9] drivers: fix some module autoloading

[PATCH -next 0/9] drivers: fix some module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Yuntao Liu (9):
  usb: ehci-mv: fix module autoloading
  soc: pxa: ssp: fix module autoloading
  misc: atmel-ssc: fix module autoloading
  i2c: at91: fix module autoloading
  mpc85xx_edac: fix module autoloading
  dmaengine: pxa: fix module autoloading
  dmaengine: mmp_pdma: fix module autoloading
  dmaengine: at_hdmac: fix module autoloading
  ipmi: ipmi_ssif: fix module autoloading

 drivers/char/ipmi/ipmi_ssif.c      | 1 +
 drivers/dma/at_hdmac.c             | 1 +
 drivers/dma/mmp_pdma.c             | 1 +
 drivers/dma/pxa_dma.c              | 1 +
 drivers/edac/mpc85xx_edac.c        | 1 +
 drivers/i2c/busses/i2c-at91-core.c | 1 +
 drivers/misc/atmel-ssc.c           | 1 +
 drivers/soc/pxa/ssp.c              | 1 +
 drivers/usb/host/ehci-mv.c         | 1 +
 9 files changed, 9 insertions(+)

-- 
2.34.1

[PATCH -next 1/9] usb: ehci-mv: fix module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
---
 drivers/usb/host/ehci-mv.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/usb/host/ehci-mv.c b/drivers/usb/host/ehci-mv.c
index 2f1fc7eb8b72..33d925316eec 100644
--- a/drivers/usb/host/ehci-mv.c
+++ b/drivers/usb/host/ehci-mv.c
@@ -260,6 +260,7 @@ static const struct platform_device_id ehci_id_table[] = {
 	{"pxa-sph", 0},
 	{},
 };
+MODULE_DEVICE_TABLE(platform, ehci_id_table);
 
 static void mv_ehci_shutdown(struct platform_device *pdev)
 {
-- 
2.34.1

Re: [PATCH -next 1/9] usb: ehci-mv: fix module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
> ---
>  drivers/usb/host/ehci-mv.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/usb/host/ehci-mv.c b/drivers/usb/host/ehci-mv.c
> index 2f1fc7eb8b72..33d925316eec 100644
> --- a/drivers/usb/host/ehci-mv.c
> +++ b/drivers/usb/host/ehci-mv.c
> @@ -260,6 +260,7 @@ static const struct platform_device_id ehci_id_table[] = {
>  	{"pxa-sph", 0},
>  	{},
>  };
> +MODULE_DEVICE_TABLE(platform, ehci_id_table);

Neither of the two entries is used any more, so a better
fix would be to remove the platform_device_id table.

     Arnd

[PATCH -next 2/9] soc: pxa: ssp: fix module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
---
 drivers/soc/pxa/ssp.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/soc/pxa/ssp.c b/drivers/soc/pxa/ssp.c
index 854d32e04558..6ac3f376d030 100644
--- a/drivers/soc/pxa/ssp.c
+++ b/drivers/soc/pxa/ssp.c
@@ -194,6 +194,7 @@ static const struct platform_device_id ssp_id_table[] = {
 	{ "pxa910-ssp",		PXA910_SSP },
 	{ },
 };
+MODULE_DEVICE_TABLE(platform, ssp_id_table);
 
 static struct platform_driver pxa_ssp_driver = {
 	.probe		= pxa_ssp_probe,
-- 
2.34.1

Re: [PATCH -next 2/9] soc: pxa: ssp: fix module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
> ---
>  drivers/soc/pxa/ssp.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/soc/pxa/ssp.c b/drivers/soc/pxa/ssp.c
> index 854d32e04558..6ac3f376d030 100644
> --- a/drivers/soc/pxa/ssp.c
> +++ b/drivers/soc/pxa/ssp.c
> @@ -194,6 +194,7 @@ static const struct platform_device_id ssp_id_table[] = {
>  	{ "pxa910-ssp",		PXA910_SSP },
>  	{ },
>  };
> +MODULE_DEVICE_TABLE(platform, ssp_id_table);
> 

I think we can drop support for legacy probing early next
year when the last pxa board files are gone, so this is 
no longer needed.

     Arnd

[PATCH -next 3/9] misc: atmel-ssc: fix module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
---
 drivers/misc/atmel-ssc.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/misc/atmel-ssc.c b/drivers/misc/atmel-ssc.c
index 6eac0f335915..e7a87183bfbb 100644
--- a/drivers/misc/atmel-ssc.c
+++ b/drivers/misc/atmel-ssc.c
@@ -110,6 +110,7 @@ static const struct platform_device_id atmel_ssc_devtypes[] = {
 		/* sentinel */
 	}
 };
+MODULE_DEVICE_TABLE(platform, atmel_ssc_devtypes);
 
 #ifdef CONFIG_OF
 static const struct of_device_id atmel_ssc_dt_ids[] = {
-- 
2.34.1

Re: [PATCH -next 3/9] misc: atmel-ssc: fix module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
> ---
>  drivers/misc/atmel-ssc.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/misc/atmel-ssc.c b/drivers/misc/atmel-ssc.c
> index 6eac0f335915..e7a87183bfbb 100644
> --- a/drivers/misc/atmel-ssc.c
> +++ b/drivers/misc/atmel-ssc.c
> @@ -110,6 +110,7 @@ static const struct platform_device_id 
> atmel_ssc_devtypes[] = {
>  		/* sentinel */
>  	}
>  };
> +MODULE_DEVICE_TABLE(platform, atmel_ssc_devtypes);

I think this driver is autoloaded by the drivers using it,
so this entry is not needed. there is also an of_device_id
table that will load the driver based on the DT information. 

     Arnd

[PATCH -next 4/9] i2c: at91: fix module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
---
 drivers/i2c/busses/i2c-at91-core.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/i2c/busses/i2c-at91-core.c b/drivers/i2c/busses/i2c-at91-core.c
index dc52b3530725..bc3636f90712 100644
--- a/drivers/i2c/busses/i2c-at91-core.c
+++ b/drivers/i2c/busses/i2c-at91-core.c
@@ -107,6 +107,7 @@ static const struct platform_device_id at91_twi_devtypes[] = {
 		/* sentinel */
 	}
 };
+MODULE_DEVICE_TABLE(platform, at91_twi_devtypes);
 
 #if defined(CONFIG_OF)
 static struct at91_twi_pdata at91sam9x5_config = {
-- 
2.34.1

Re: [PATCH -next 4/9] i2c: at91: fix module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
> ---
>  drivers/i2c/busses/i2c-at91-core.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/i2c/busses/i2c-at91-core.c 
> b/drivers/i2c/busses/i2c-at91-core.c
> index dc52b3530725..bc3636f90712 100644
> --- a/drivers/i2c/busses/i2c-at91-core.c
> +++ b/drivers/i2c/busses/i2c-at91-core.c
> @@ -107,6 +107,7 @@ static const struct platform_device_id 
> at91_twi_devtypes[] = {
>  		/* sentinel */
>  	}
>  };
> +MODULE_DEVICE_TABLE(platform, at91_twi_devtypes);
> 
>  #if defined(CONFIG_OF)
>  static struct at91_twi_pdata at91sam9x5_config = {
> -- 

This device is always probed from DT, so a better fix would
be to remove the table and the #ifdef/of_match_ptr() around
the atmel_twi_dt_ids.

     Arnd

[PATCH -next 5/9] mpc85xx_edac: fix module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
---
 drivers/edac/mpc85xx_edac.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/edac/mpc85xx_edac.c b/drivers/edac/mpc85xx_edac.c
index d0266cbcbeda..c8b94aab9819 100644
--- a/drivers/edac/mpc85xx_edac.c
+++ b/drivers/edac/mpc85xx_edac.c
@@ -320,6 +320,7 @@ static const struct platform_device_id mpc85xx_pci_err_match[] = {
 	},
 	{}
 };
+MODULE_DEVICE_TABLE(platform, mpc85xx_pci_err_match);
 
 static struct platform_driver mpc85xx_pci_err_driver = {
 	.probe = mpc85xx_pci_err_probe,
-- 
2.34.1

Re: [PATCH -next 5/9] mpc85xx_edac: fix module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>

Acked-by: Arnd Bergmann <arnd@arndb.de>

>  	},
>  	{}
>  };
> +MODULE_DEVICE_TABLE(platform, mpc85xx_pci_err_match);
> 

I see that this device is created in arch/powerpc/sysdev/fsl_pci.c,
so your change makes sense here.

     Arnd

[PATCH -next 6/9] dmaengine: pxa: fix module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
---
 drivers/dma/pxa_dma.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/dma/pxa_dma.c b/drivers/dma/pxa_dma.c
index 31f8da810c05..56785805e7a7 100644
--- a/drivers/dma/pxa_dma.c
+++ b/drivers/dma/pxa_dma.c
@@ -1434,6 +1434,7 @@ static const struct platform_device_id pxad_id_table[] = {
 	{ "pxa-dma", },
 	{ },
 };
+MODULE_DEVICE_TABLE(platform, pxad_id_table);
 
 static struct platform_driver pxad_driver = {
 	.driver		= {
-- 
2.34.1

Re: [PATCH -next 6/9] dmaengine: pxa: fix module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
> ---
>  drivers/dma/pxa_dma.c | 1 +
>  1 file changed, 1 insertion(+)
>

The legacy probe will soon be gone for pxa, so I would skip this
one, like the other pxa patches.

      Arnd

[PATCH -next 7/9] dmaengine: mmp_pdma: fix module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
---
 drivers/dma/mmp_pdma.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/dma/mmp_pdma.c b/drivers/dma/mmp_pdma.c
index 136fcaeff8dd..05d051ecf833 100644
--- a/drivers/dma/mmp_pdma.c
+++ b/drivers/dma/mmp_pdma.c
@@ -1129,6 +1129,7 @@ static const struct platform_device_id mmp_pdma_id_table[] = {
 	{ "mmp-pdma", },
 	{ },
 };
+MODULE_DEVICE_TABLE(platform, mmp_pdma_id_table);
 
 static struct platform_driver mmp_pdma_driver = {
 	.driver		= {
-- 
2.34.1

Re: [PATCH -next 7/9] dmaengine: mmp_pdma: fix module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
> ---
>  drivers/dma/mmp_pdma.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/dma/mmp_pdma.c b/drivers/dma/mmp_pdma.c
> index 136fcaeff8dd..05d051ecf833 100644
> --- a/drivers/dma/mmp_pdma.c
> +++ b/drivers/dma/mmp_pdma.c
> @@ -1129,6 +1129,7 @@ static const struct platform_device_id 
> mmp_pdma_id_table[] = {
>  	{ "mmp-pdma", },
>  	{ },
>  };
> +MODULE_DEVICE_TABLE(platform, mmp_pdma_id_table);

It appears that this table was never used in upstream kernels,
as the driver already used DT when it was first added.

     Arnd

[PATCH -next 8/9] dmaengine: at_hdmac: fix module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
---
 drivers/dma/at_hdmac.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/dma/at_hdmac.c b/drivers/dma/at_hdmac.c
index 40052d1bd0b5..614c60b16ee5 100644
--- a/drivers/dma/at_hdmac.c
+++ b/drivers/dma/at_hdmac.c
@@ -1904,6 +1904,7 @@ static const struct platform_device_id atdma_devtypes[] = {
 		/* sentinel */
 	}
 };
+MODULE_DEVICE_TABLE(platform, atdma_devtypes);
 
 static inline const struct at_dma_platform_data * __init at_dma_get_driver_data(
 						struct platform_device *pdev)
-- 
2.34.1

Re: [PATCH -next 8/9] dmaengine: at_hdmac: fix module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>

This table is again unused because at91 uses DT based
probing. Please just remove the table and the #ifdef/of_match_ptr()
around the of_device_id table.

     Arnd

[PATCH -next 9/9] ipmi: ipmi_ssif: fix module autoloading

[Yuntao Liu]

Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
based on the alias from platform_device_id table.

Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
---
 drivers/char/ipmi/ipmi_ssif.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/char/ipmi/ipmi_ssif.c b/drivers/char/ipmi/ipmi_ssif.c
index 96ad571d041a..e8e7b832c060 100644
--- a/drivers/char/ipmi/ipmi_ssif.c
+++ b/drivers/char/ipmi/ipmi_ssif.c
@@ -2085,6 +2085,7 @@ static const struct platform_device_id ssif_plat_ids[] = {
     { "dmi-ipmi-ssif", 0 },
     { }
 };
+MODULE_DEVICE_TABLE(platform, ssif_plat_ids);
 
 static struct platform_driver ipmi_driver = {
 	.driver = {
-- 
2.34.1

Re: [PATCH -next 9/9] ipmi: ipmi_ssif: fix module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Signed-off-by: Yuntao Liu <liuyuntao12@huawei.com>
> ---

The driver already has a MODULE_ALIAS() with the same string.

I think the MODULE_DEVICE_TABLE() entry is slightly cleaner here,
but it should only have one of the two, not both.

     Arnd

Re: [PATCH -next 0/9] drivers: fix some module autoloading

[Arnd Bergmann]

On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> based on the alias from platform_device_id table.
>
> Yuntao Liu (9):
>   usb: ehci-mv: fix module autoloading
>   soc: pxa: ssp: fix module autoloading
>   misc: atmel-ssc: fix module autoloading
>   i2c: at91: fix module autoloading
>   mpc85xx_edac: fix module autoloading
>   dmaengine: pxa: fix module autoloading
>   dmaengine: mmp_pdma: fix module autoloading
>   dmaengine: at_hdmac: fix module autoloading
>   ipmi: ipmi_ssif: fix module autoloading

I looked at all the patches and found that most of them do not
use the table any more, or will stop using it in the near future.

I think your work to validate the correctness of the entries
is useful, but it may be more helpful to focus on removing
all the unused tables, including those that have a
MODULE_DEVICE_TABLE() tag.

If you are planning to do more such cleanups, maybe you can
go through them one subsystem at a time and look for drivers
that have both of_device_id and i2c_device_id/platform_device_id/
spi_device_id tables. If nothing in the kernel creates a device
with the legacy string, you can then send a patch that removes
the old device ID list and at the same time makes the DT support
unconditional in case there is an #ifdef CONFIG_OF check.

If the probe() function accesses platform_data, this would also
be unused, allowing an even nicer cleanup of removing the
platofrm_data path in favor of OF properties.

      Arnd

Re: [PATCH -next 0/9] drivers: fix some module autoloading

[liuyuntao (F)]

yes sir^^.

Regards!
Yuntao

On 2024/8/19 20:09, Arnd Bergmann wrote:
> On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
>> Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
>> based on the alias from platform_device_id table.
>>
>> Yuntao Liu (9):
>>    usb: ehci-mv: fix module autoloading
>>    soc: pxa: ssp: fix module autoloading
>>    misc: atmel-ssc: fix module autoloading
>>    i2c: at91: fix module autoloading
>>    mpc85xx_edac: fix module autoloading
>>    dmaengine: pxa: fix module autoloading
>>    dmaengine: mmp_pdma: fix module autoloading
>>    dmaengine: at_hdmac: fix module autoloading
>>    ipmi: ipmi_ssif: fix module autoloading
> 
> I looked at all the patches and found that most of them do not
> use the table any more, or will stop using it in the near future.
> 
> I think your work to validate the correctness of the entries
> is useful, but it may be more helpful to focus on removing
> all the unused tables, including those that have a
> MODULE_DEVICE_TABLE() tag.
> 
> If you are planning to do more such cleanups, maybe you can
> go through them one subsystem at a time and look for drivers
> that have both of_device_id and i2c_device_id/platform_device_id/
> spi_device_id tables. If nothing in the kernel creates a device
> with the legacy string, you can then send a patch that removes
> the old device ID list and at the same time makes the DT support
> unconditional in case there is an #ifdef CONFIG_OF check.
> 
> If the probe() function accesses platform_data, this would also
> be unused, allowing an even nicer cleanup of removing the
> platofrm_data path in favor of OF properties.
> 
>        Arnd

Re: [PATCH -next 0/9] drivers: fix some module autoloading

[Andi Shyti]

Hi,

On Mon, Aug 19, 2024 at 02:09:49PM GMT, Arnd Bergmann wrote:
> On Mon, Aug 19, 2024, at 13:38, Yuntao Liu wrote:
> > Add MODULE_DEVICE_TABLE(), so modules could be properly autoloaded
> > based on the alias from platform_device_id table.
> >
> > Yuntao Liu (9):
> >   usb: ehci-mv: fix module autoloading
> >   soc: pxa: ssp: fix module autoloading
> >   misc: atmel-ssc: fix module autoloading
> >   i2c: at91: fix module autoloading
> >   mpc85xx_edac: fix module autoloading
> >   dmaengine: pxa: fix module autoloading
> >   dmaengine: mmp_pdma: fix module autoloading
> >   dmaengine: at_hdmac: fix module autoloading
> >   ipmi: ipmi_ssif: fix module autoloading
> 
> I looked at all the patches and found that most of them do not
> use the table any more, or will stop using it in the near future.
> 
> I think your work to validate the correctness of the entries
> is useful, but it may be more helpful to focus on removing
> all the unused tables, including those that have a
> MODULE_DEVICE_TABLE() tag.
> 
> If you are planning to do more such cleanups, maybe you can
> go through them one subsystem at a time and look for drivers
> that have both of_device_id and i2c_device_id/platform_device_id/
> spi_device_id tables. If nothing in the kernel creates a device
> with the legacy string, you can then send a patch that removes
> the old device ID list and at the same time makes the DT support
> unconditional in case there is an #ifdef CONFIG_OF check.
> 
> If the probe() function accesses platform_data, this would also
> be unused, allowing an even nicer cleanup of removing the
> platofrm_data path in favor of OF properties.

Thanks for looking into these patches, Arnd!

Andi