Date:   Tue, 4 Jan 2022 21:11:56 +0800
From:   zhuyinbo <zhuyinbo@...ngson.cn>
To:     Andrew Lunn <andrew@...n.ch>,
        "Russell King (Oracle)" <linux@...linux.org.uk>, andrew@...n.ch,
        hkallweit1@...il.com, linux@...linux.org.uk, davem@...emloft.net,
        kuba@...nel.org, masahiroy@...nel.org, michal.lkml@...kovi.net,
        ndesaulniers@...gle.com
Cc:     Heiner Kallweit <hkallweit1@...il.com>, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org, linux-kbuild@...r.kernel.org,
        zhuyinbo@...ngson.cn, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org, linux-kbuild@...r.kernel.org,
        "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Masahiro Yamada <masahiroy@...nel.org>,
        Michal Marek <michal.lkml@...kovi.net>,
        Nick Desaulniers <ndesaulniers@...gle.com>
Subject: Re: [PATCH v2 1/2] modpost: file2alias: fixup mdio alias garbled code
 in modules.alias



在 2021/12/7 下午5:41, zhuyinbo 写道:
> 
> 
> 在 2021/12/1 上午8:38, Andrew Lunn 写道:
>>> However, this won't work for PHY devices created _before_ the kernel
>>> has mounted the rootfs, whether or not they end up being used. So,
>>> every PHY mentioned in DT will be created before the rootfs is mounted,
>>> and none of these PHYs will have their modules loaded.
>>
>> Hi Russell
>>
>> I think what you are saying here is, if the MAC or MDIO bus driver is
>> built in, the PHY driver also needs to be built in?
>>
>> If the MAC or MDIO bus driver is a module, it means the rootfs has
>> already been mounted in order to get these modules. And so the PHY
>> driver as a module will also work.
>>
>>> I believe this is the root cause of Yinbo Zhu's issue.
> 
> I think you should be right and I had did lots of test but use 
> rquest_module it doesn't load marvell module, and dts does't include any 
> phy node. even though I was use "marvell" as input's args of 
> request_module.
>>
>> You are speculating that in Yinbo Zhu case, the MAC driver is built
>> in, the PHY is a module. The initial request for the firmware fails.
>> Yinbo Zhu would like udev to try again later when the modules are
>> available.
>>
>>> What we _could_ do is review all device trees and PHY drivers to see
>>> whether DT modaliases are ever used for module loading. If they aren't,
>>> then we _could_ make the modalias published by the kernel conditional
>>> on the type of mdio device - continue with the DT approach for non-PHY
>>> devices, and switch to the mdio: scheme for PHY devices. I repeat, this
>>> can only happen if no PHY drivers match using the DT scheme, otherwise
>>> making this change _will_ cause a regression.
>>
> 
>> Take a look at
>> drivers/net/mdio/of_mdio.c:whitelist_phys[] and the comment above it.
>>
>> So there are some DT blobs out there with compatible strings for
>> PHYs. I've no idea if they actually load that way, or the standard PHY
>> mechanism is used.
>>
>>     Andrew
>>
> 
> 
>  > That is not true universally for all MDIO though - as
>  > xilinx_gmii2rgmii.c clearly shows. That is a MDIO driver which uses DT
>  > the compatible string to do the module load. So, we have proof there
>  > that Yinbo Zhu's change will definitely cause a regression which we
>  > can not allow.
> 
> I don't understand that what you said about regression.  My patch 
> doesn't cause  xilinx_gmii2rgmii.c driver load fail, in this time that 
> do_of_table and platform_uevent will be responsible "of" type driver 
> auto load and my patch was responsible for "mdio" type driver auto load,
> In default code. There are request_module to load phy driver, but as 
> Russell King said that request_module doesn't garantee auto load will 
> always work well, but udev mechanism can garantee it. and udev mechaism 
> is more mainstream, otherwise mdio_uevent is useless. if use udev 
> mechanism that my patch was needed. and if apply my patch it doesn't 
> cause request_module mechaism work bad because I will add following change:
> 
> 
> 
> -       ret = request_module(MDIO_MODULE_PREFIX MDIO_ID_FMT,
> -                            MDIO_ID_ARGS(phy_id));
> +       ret = request_module(MDIO_MODULE_PREFIX MDIO_ID_FMT, phy_id);
>          /* We only check for failures in executing the usermode binary,
>           * not whether a PHY driver module exists for the PHY ID.
>           * Accept -ENOENT because this may occur in case no initramfs 
> exists,
> diff --git a/include/linux/mod_devicetable.h 
> b/include/linux/mod_devicetable.h
> index 7bd23bf..bc6ea0d 100644
> --- a/include/linux/mod_devicetable.h
> +++ b/include/linux/mod_devicetable.h
> @@ -600,16 +600,7 @@ struct platform_device_id {
>   #define MDIO_NAME_SIZE         32
>   #define MDIO_MODULE_PREFIX     "mdio:"
> 
> -#define MDIO_ID_FMT 
> "%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u%u"
> -#define MDIO_ID_ARGS(_id) \
> -       ((_id)>>31) & 1, ((_id)>>30) & 1, ((_id)>>29) & 1, ((_id)>>28) & 
> 1, \
> -       ((_id)>>27) & 1, ((_id)>>26) & 1, ((_id)>>25) & 1, ((_id)>>24) & 
> 1, \
> -       ((_id)>>23) & 1, ((_id)>>22) & 1, ((_id)>>21) & 1, ((_id)>>20) & 
> 1, \
> -       ((_id)>>19) & 1, ((_id)>>18) & 1, ((_id)>>17) & 1, ((_id)>>16) & 
> 1, \
> -       ((_id)>>15) & 1, ((_id)>>14) & 1, ((_id)>>13) & 1, ((_id)>>12) & 
> 1, \
> -       ((_id)>>11) & 1, ((_id)>>10) & 1, ((_id)>>9) & 1, ((_id)>>8) & 1, \
> -       ((_id)>>7) & 1, ((_id)>>6) & 1, ((_id)>>5) & 1, ((_id)>>4) & 1, \
> -       ((_id)>>3) & 1, ((_id)>>2) & 1, ((_id)>>1) & 1, (_id) & 1
> +#define MDIO_ID_FMT "p%08x"
> 
> 
> 

 > > > > However, this won't work for PHY devices created _before_ the 
kernel
 > > > > has mounted the rootfs, whether or not they end up being used. So,
 > > > > every PHY mentioned in DT will be created before the rootfs is 
mounted,
 > > > > and none of these PHYs will have their modules loaded.
 > > >
 > > > Hi Russell
 > > >
 > > > I think what you are saying here is, if the MAC or MDIO bus driver is
 > > > built in, the PHY driver also needs to be built in?
 > > >
 > > > If the MAC or MDIO bus driver is a module, it means the rootfs has
 > > > already been mounted in order to get these modules. And so the PHY
 > > > driver as a module will also work.
 > > >
 > > > > I believe this is the root cause of Yinbo Zhu's issue.
 > >
 > > I think you should be right and I had did lots of test but use 
rquest_module
 > > it doesn't load marvell module, and dts does't include any phy 
node. even
 > > though I was use "marvell" as input's args of request_module.

 > Please can you report the contents of /proc/sys/kernel/modprobe, and
 > the kernel configuration of CONFIG_MODPROBE_PATH. I wonder if your
 > userspace has that module loading mechanism disabled, or your kernel
 > has CONFIG_MODPROBE_PATH as an empty string.

 > If the module is not present by the time this call is made, then
 > even if you load the appropriate driver module later, that module
 > will not be used - the PHY will end up being driven by the generic
 > clause 22 driver.

 > > > That is not true universally for all MDIO though - as
 > > > xilinx_gmii2rgmii.c clearly shows. That is a MDIO driver which 
uses DT
 > > > the compatible string to do the module load. So, we have proof there
 > > > that Yinbo Zhu's change will definitely cause a regression which we
 > > > can not allow.
 > >
 > > I don't understand that what you said about regression.  My patch 
doesn't
 > > cause  xilinx_gmii2rgmii.c driver load fail, in this time that 
do_of_table
 > >and platform_uevent will be responsible "of" type driver auto load 
and my
 > > patch was responsible for "mdio" type driver auto load,

 > xilinx_gmii2rgmii is not a platform driver. It is a mdio driver:

 > static struct mdio_driver xgmiitorgmii_driver = {
               ^^^^^^^^^^^

 > Therefore, platform_uevent() is irrelevant since this will never match
 > a platform device. It will only match mdio devices, and the uevent
 > generation for that is via mdio_uevent() which is the function you
 > are changing.


static const struct of_device_id xgmiitorgmii_of_match[] = {
         { .compatible = "xlnx,gmii-to-rgmii-1.0" },
         {},
};
MODULE_DEVICE_TABLE(of, xgmiitorgmii_of_match);

static struct mdio_driver xgmiitorgmii_driver = {
         .probe  = xgmiitorgmii_probe,
         .mdiodrv.driver = {
                 .name = "xgmiitorgmii",
                 .of_match_table = xgmiitorgmii_of_match,
         },
};
 From the present point of view, no matter what the situation, my 
supplement can cover udev or request_module for auto load module.

if that phy driver isn't platform driver my patch cover it I think there 
is no doubt, if phy driver is platform driver and platform driver udev 
will cover it. My only requestion is the request_module not work well.

about xgmiitorgmii_of_match that it belongs to platform driver load, 
please you note. and about your doubt usepace whether disable module 
load that module load function is okay becuase other device driver auto 
load is okay.

 > > In default code. There are request_module to load phy driver, but 
as > Russell
 > > King said that request_module doesn't garantee auto load will 
always work
 > > well, but udev mechanism can garantee it. and udev mechaism is more
 > > mainstream, otherwise mdio_uevent is useless. if use udev mechanism 
that my
 > > patch was needed. and if apply my patch it doesn't cause request_module
 > > mechaism work bad because I will add following change:

 > Please report back what the following command produces on your
 > problem system:

 > /sbin/modprobe -vn mdio:00000001010000010000110111010001

 > Thanks.

[root@...alhost ~]# lsmod | grep marvell
[root@...alhost ~]# ls 
/lib/modules/4.19.190+/kernel/drivers/net/phy/marvell.ko
/lib/modules/4.19.190+/kernel/drivers/net/phy/marvell.ko
[root@...alhost ~]# /sbin/modprobe -vn mdio:00000001010000010000110111010001
insmod /lib/modules/4.19.190+/kernel/drivers/net/phy/marvell.ko
insmod /lib/modules/4.19.190+/kernel/drivers/net/phy/marvell.ko
[root@...alhost ~]#
[root@...alhost ~]# cat /proc/sys/kernel/modprobe
/sbin/modprobe

BRs,
Yinbo

Powered by blists - more mailing lists