[PATCH net-next] net: fec: Use a proper ID allocation scheme

[PATCH net-next] net: fec: Use a proper ID allocation scheme

[Fabio Estevam]

Currently when performing an unbind/bind operation network is no
longer functional:

# echo 2188000.ethernet > /sys/bus/platform/drivers/fec/unbind
# echo 2188000.ethernet > /sys/bus/platform/drivers/fec/bind
[   10.756519] pps pps0: new PPS source ptp0
[   10.792626] libphy: fec_enet_mii_bus: probed
[   10.799330] fec 2188000.ethernet eth0: registered PHC device 1
# udhcpc -i eth0
udhcpc: started, v1.31.1
[   14.985211] fec 2188000.ethernet eth0: no PHY, assuming direct connection to switch
[   14.993140] libphy: PHY fixed-0:00 not found
[   14.997643] fec 2188000.ethernet eth0: could not attach to PHY
ifconfig: SIOCSIFFLAGS: No such device

The reason for the failure is that fec_drv_remove() does not
decrement the dev_id variable.

Instead of using such poor mechanism for counting the network interfaces
IDs, use a proper allocation scheme, such as IDR.

This fixes the network behavior after unbind/bind.

Tested on a imx6qp-sabresd board.

Suggested-by: David Miller <davem@davemloft.net>
Signed-off-by: Fabio Estevam <festevam@gmail.com>
---
Hi,

There was a prior attempt to fix this problem:
https://www.spinics.net/lists/netdev/msg616487.html

, but it was rejected and David suggested the usage of IDR.

 drivers/net/ethernet/freescale/fec.h      |  1 +
 drivers/net/ethernet/freescale/fec_main.c | 15 +++++++++------
 2 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec.h b/drivers/net/ethernet/freescale/fec.h
index f79e57f735b3..0d718545b9a2 100644
--- a/drivers/net/ethernet/freescale/fec.h
+++ b/drivers/net/ethernet/freescale/fec.h
@@ -528,6 +528,7 @@ struct fec_enet_private {
 	struct	platform_device *pdev;
 
 	int	dev_id;
+	struct  idr idr;
 
 	/* Phylib and MDIO interface */
 	struct	mii_bus *mii_bus;
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 4432a59904c7..77b63ecf96b4 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -62,6 +62,7 @@
 #include <linux/if_vlan.h>
 #include <linux/pinctrl/consumer.h>
 #include <linux/prefetch.h>
+#include <linux/idr.h>
 #include <soc/imx/cpuidle.h>
 
 #include <asm/cacheflush.h>
@@ -1949,7 +1950,6 @@ static int fec_enet_mii_probe(struct net_device *ndev)
 	char mdio_bus_id[MII_BUS_ID_SIZE];
 	char phy_name[MII_BUS_ID_SIZE + 3];
 	int phy_id;
-	int dev_id = fep->dev_id;
 
 	if (fep->phy_node) {
 		phy_dev = of_phy_connect(ndev, fep->phy_node,
@@ -1964,8 +1964,6 @@ static int fec_enet_mii_probe(struct net_device *ndev)
 		for (phy_id = 0; (phy_id < PHY_MAX_ADDR); phy_id++) {
 			if (!mdiobus_is_registered_device(fep->mii_bus, phy_id))
 				continue;
-			if (dev_id--)
-				continue;
 			strlcpy(mdio_bus_id, fep->mii_bus->id, MII_BUS_ID_SIZE);
 			break;
 		}
@@ -3406,7 +3404,6 @@ fec_probe(struct platform_device *pdev)
 	struct net_device *ndev;
 	int i, irq, ret = 0;
 	const struct of_device_id *of_id;
-	static int dev_id;
 	struct device_node *np = pdev->dev.of_node, *phy_node;
 	int num_tx_qs;
 	int num_rx_qs;
@@ -3451,7 +3448,13 @@ fec_probe(struct platform_device *pdev)
 	}
 
 	fep->pdev = pdev;
-	fep->dev_id = dev_id++;
+	idr_init(&fep->idr);
+
+	ret = idr_alloc_cyclic(&fep->idr, fep, 0, INT_MAX, GFP_KERNEL);
+	if (ret < 0)
+		return ret;
+
+	fep->dev_id = ret;
 
 	platform_set_drvdata(pdev, ndev);
 
@@ -3632,7 +3635,6 @@ fec_probe(struct platform_device *pdev)
 		of_phy_deregister_fixed_link(np);
 	of_node_put(phy_node);
 failed_phy:
-	dev_id--;
 failed_ioremap:
 	free_netdev(ndev);
 
@@ -3661,6 +3663,7 @@ fec_drv_remove(struct platform_device *pdev)
 	if (of_phy_is_fixed_link(np))
 		of_phy_deregister_fixed_link(np);
 	of_node_put(fep->phy_node);
+	idr_destroy(&fep->idr);
 	free_netdev(ndev);
 
 	clk_disable_unprepare(fep->clk_ahb);
-- 
2.17.1

Re: [PATCH net-next] net: fec: Use a proper ID allocation scheme

[David Miller]

From: Fabio Estevam <festevam@gmail.com>
Date: Mon, 17 Feb 2020 19:36:51 -0300

> Instead of using such poor mechanism for counting the network interfaces
> IDs, use a proper allocation scheme, such as IDR.
> 
> This fixes the network behavior after unbind/bind.

What about:

1) unbind fec0
2) unbind fec1
3) bind fec0

It doesn't work even with the IDR scheme.

RE: [EXT] Re: [PATCH net-next] net: fec: Use a proper ID allocation scheme

[Andy Duan]

From: David Miller <davem@davemloft.net> Sent: Tuesday, February 18, 2020 1:49 PM
> From: Fabio Estevam <festevam@gmail.com>
> Date: Mon, 17 Feb 2020 19:36:51 -0300
> 
> > Instead of using such poor mechanism for counting the network
> > interfaces IDs, use a proper allocation scheme, such as IDR.
> >
> > This fixes the network behavior after unbind/bind.
> 
> What about:
> 
> 1) unbind fec0
> 2) unbind fec1
> 3) bind fec0
> 
> It doesn't work even with the IDR scheme.

Not only such case, instance#A (maybe fec0 or fec1) depends on instance#B (maybe fec1 or fec0),
Unbind instance#B firstly has problem.
Bind instance#A firstly also has problem.

Re: [EXT] Re: [PATCH net-next] net: fec: Use a proper ID allocation scheme

[Fabio Estevam]

Hi Andy,

On Tue, Feb 18, 2020 at 3:51 AM Andy Duan <fugang.duan@nxp.com> wrote:

> > What about:
> >
> > 1) unbind fec0
> > 2) unbind fec1
> > 3) bind fec0
> >
> > It doesn't work even with the IDR scheme.
>
> Not only such case, instance#A (maybe fec0 or fec1) depends on instance#B (maybe fec1 or fec0),
> Unbind instance#B firstly has problem.
> Bind instance#A firstly also has problem.

Yes, I do see the error now with the sequence suggested by David.

I have also noticed in the fec_main.c comments:

/*
* The i.MX28 dual fec interfaces are not equal.
* Here are the differences:
*
*  - fec0 supports MII & RMII modes while fec1 only supports RMII
*  - fec0 acts as the 1588 time master while fec1 is slave
*  - external phys can only be configured by fec0
*
* That is to say fec1 can not work independently. It only works
* when fec0 is working. The reason behind this design is that the
* second interface is added primarily for Switch mode.
*
* Because of the last point above, both phys are attached on fec0
* mdio interface in board design, and need to be configured by
* fec0 mii_bus.
*/

Should we prevent unbind operation from this driver like this?

diff --git a/drivers/net/ethernet/freescale/fec_main.c
b/drivers/net/ethernet/freescale/fec_main.c
index 4432a59904c7..1d348c5d0794 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -3793,6 +3793,7 @@ static struct platform_driver fec_driver = {
                .name   = DRIVER_NAME,
                .pm     = &fec_pm_ops,
                .of_match_table = fec_dt_ids,
+               .suppress_bind_attrs = true
        },
        .id_table = fec_devtype,
        .probe  = fec_probe,

Please advise.

Thanks

RE: [EXT] Re: [PATCH net-next] net: fec: Use a proper ID allocation scheme

[Andy Duan]

From: Fabio Estevam <festevam@gmail.com> Sent: Tuesday, February 18, 2020 9:25 PM
> Hi Andy,
> 
> On Tue, Feb 18, 2020 at 3:51 AM Andy Duan <fugang.duan@nxp.com> wrote:
> 
> > > What about:
> > >
> > > 1) unbind fec0
> > > 2) unbind fec1
> > > 3) bind fec0
> > >
> > > It doesn't work even with the IDR scheme.
> >
> > Not only such case, instance#A (maybe fec0 or fec1) depends on
> > instance#B (maybe fec1 or fec0), Unbind instance#B firstly has problem.
> > Bind instance#A firstly also has problem.
> 
> Yes, I do see the error now with the sequence suggested by David.
> 
> I have also noticed in the fec_main.c comments:
> 
> /*
> * The i.MX28 dual fec interfaces are not equal.
> * Here are the differences:
> *
> *  - fec0 supports MII & RMII modes while fec1 only supports RMII
> *  - fec0 acts as the 1588 time master while fec1 is slave
> *  - external phys can only be configured by fec0
> *
> * That is to say fec1 can not work independently. It only works
> * when fec0 is working. The reason behind this design is that the
> * second interface is added primarily for Switch mode.
> *
> * Because of the last point above, both phys are attached on fec0
> * mdio interface in board design, and need to be configured by
> * fec0 mii_bus.
> */
> 
> Should we prevent unbind operation from this driver like this?
> 
> diff --git a/drivers/net/ethernet/freescale/fec_main.c
> b/drivers/net/ethernet/freescale/fec_main.c
> index 4432a59904c7..1d348c5d0794 100644
> --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -3793,6 +3793,7 @@ static struct platform_driver fec_driver = {
>                 .name   = DRIVER_NAME,
>                 .pm     = &fec_pm_ops,
>                 .of_match_table = fec_dt_ids,
> +               .suppress_bind_attrs = true
>         },
>         .id_table = fec_devtype,
>         .probe  = fec_probe,
> 
> Please advise.
> 
> Thanks

For imx6sl/imx8mp/imx8mm/imx8mn, soc only has one instance, bind operation
is supported and has no problem.

Regards,
Andy

Re: [EXT] Re: [PATCH net-next] net: fec: Use a proper ID allocation scheme

[Fabio Estevam]

Hi Andy,

On Tue, Feb 18, 2020 at 10:54 AM Andy Duan <fugang.duan@nxp.com> wrote:

> For imx6sl/imx8mp/imx8mm/imx8mn, soc only has one instance, bind operation
> is supported and has no problem.

This is not true.

As per the commit log, here is the result of unbind/bind on a i.mx6qp,
which only has a single FEC instance:

# echo 2188000.ethernet > /sys/bus/platform/drivers/fec/unbind
# echo 2188000.ethernet > /sys/bus/platform/drivers/fec/bind
[   10.756519] pps pps0: new PPS source ptp0
[   10.792626] libphy: fec_enet_mii_bus: probed
[   10.799330] fec 2188000.ethernet eth0: registered PHC device 1
# udhcpc -i eth0
udhcpc: started, v1.31.1
[   14.985211] fec 2188000.ethernet eth0: no PHY, assuming direct
connection to switch
[   14.993140] libphy: PHY fixed-0:00 not found
[   14.997643] fec 2188000.ethernet eth0: could not attach to PHY

After performing unbind/bind operation the network is not functional at all.

Don't you agree that unbind/bind is currently broken here even for
SoCs with a single FEC?

Should we prevent unbind? Or any other suggestion?

RE: [EXT] Re: [PATCH net-next] net: fec: Use a proper ID allocation scheme

[Andy Duan]

From: Fabio Estevam <festevam@gmail.com> Sent: Tuesday, February 18, 2020 10:05 PM
> Hi Andy,
> 
> On Tue, Feb 18, 2020 at 10:54 AM Andy Duan <fugang.duan@nxp.com>
> wrote:
> 
> > For imx6sl/imx8mp/imx8mm/imx8mn, soc only has one instance, bind
> > operation is supported and has no problem.
> 
> This is not true.
> 
> As per the commit log, here is the result of unbind/bind on a i.mx6qp, which
> only has a single FEC instance:
I mean if apply the patch, it should work for one instance.

> 
> # echo 2188000.ethernet > /sys/bus/platform/drivers/fec/unbind
> # echo 2188000.ethernet > /sys/bus/platform/drivers/fec/bind
> [   10.756519] pps pps0: new PPS source ptp0
> [   10.792626] libphy: fec_enet_mii_bus: probed
> [   10.799330] fec 2188000.ethernet eth0: registered PHC device 1
> # udhcpc -i eth0
> udhcpc: started, v1.31.1
> [   14.985211] fec 2188000.ethernet eth0: no PHY, assuming direct
> connection to switch
> [   14.993140] libphy: PHY fixed-0:00 not found
> [   14.997643] fec 2188000.ethernet eth0: could not attach to PHY
> 
> After performing unbind/bind operation the network is not functional at all.
> 
> Don't you agree that unbind/bind is currently broken here even for SoCs with
> a single FEC?
> 
> Should we prevent unbind? Or any other suggestion?
Suppose apply the patch, it can work for one instance, but not for two instances.
Currently, I agree to prevent unbind operation.

Re: [EXT] Re: [PATCH net-next] net: fec: Use a proper ID allocation scheme

[Fabio Estevam]

On Tue, Feb 18, 2020 at 11:30 AM Andy Duan <fugang.duan@nxp.com> wrote:

> Currently, I agree to prevent unbind operation.

Ok, thanks for the feedback.

I will submit a patch doing this.