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@localhost ~]# lsmod | grep marvell
[root@localhost ~]# ls
/lib/modules/4.19.190+/kernel/drivers/net/phy/marvell.ko
/lib/modules/4.19.190+/kernel/drivers/net/phy/marvell.ko
[root@localhost ~]# /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@localhost ~]#
[root@localhost ~]# cat /proc/sys/kernel/modprobe
/sbin/modprobe
BRs,
Yinbo
