[PATCH net v1 1/2] net: dsa: microchip: fix devicetree parsing of cpu node

[PATCH net v1 1/2] net: dsa: microchip: fix devicetree parsing of cpu node

[Sven Van Asbroeck]

From: Sven Van Asbroeck <thesven73@gmail.com>

On the ksz8795, if the devicetree contains a cpu node,
devicetree parsing fails and the whole driver errors out.

Fix the devicetree parsing code by making it use the
correct number of ports.

Fixes: 912aae27c6af ("net: dsa: microchip: really look for phy-mode in port nodes")
Tested-by: Sven Van Asbroeck <thesven73@gmail.com> # ksz8795
Signed-off-by: Sven Van Asbroeck <thesven73@gmail.com>
---

Tree: git://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git # 905b2032fa42

To: Woojung Huh <woojung.huh@microchip.com>
To: Microchip Linux Driver Support <UNGLinuxDriver@microchip.com>
To: Andrew Lunn <andrew@lunn.ch>
To: Vivien Didelot <vivien.didelot@gmail.com>
To: Florian Fainelli <f.fainelli@gmail.com>
To: Vladimir Oltean <olteanv@gmail.com>
To: "David S. Miller" <davem@davemloft.net>
To: Jakub Kicinski <kuba@kernel.org>
Cc: Helmut Grohne <helmut.grohne@intenta.de>
Cc: netdev@vger.kernel.org
Cc: linux-kernel@vger.kernel.org

 drivers/net/dsa/microchip/ksz_common.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/dsa/microchip/ksz_common.c b/drivers/net/dsa/microchip/ksz_common.c
index e5f047129b15..17b804c44c53 100644
--- a/drivers/net/dsa/microchip/ksz_common.c
+++ b/drivers/net/dsa/microchip/ksz_common.c
@@ -431,7 +431,7 @@ int ksz_switch_register(struct ksz_device *dev,
 				if (of_property_read_u32(port, "reg",
 							 &port_num))
 					continue;
-				if (port_num >= dev->port_cnt)
+				if (port_num >= dev->ds->num_ports)
 					return -EINVAL;
 				of_get_phy_mode(port,
 						&dev->ports[port_num].interface);
-- 
2.17.1

[PATCH net v1 2/2] net: dsa: microchip: improve port count comments

[Sven Van Asbroeck]

From: Sven Van Asbroeck <thesven73@gmail.com>

Port counts in microchip dsa drivers can be quite confusing:
on the ksz8795, ksz_chip_data->port_cnt excludes the cpu port,
yet on the ksz9477, it includes the cpu port.

Add comments to document this situation explicitly.

Fixes: 912aae27c6af ("net: dsa: microchip: really look for phy-mode in port nodes")
Signed-off-by: Sven Van Asbroeck <thesven73@gmail.com>
---

Tree: git://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git # 905b2032fa42

To: Woojung Huh <woojung.huh@microchip.com>
To: Microchip Linux Driver Support <UNGLinuxDriver@microchip.com>
To: Andrew Lunn <andrew@lunn.ch>
To: Vivien Didelot <vivien.didelot@gmail.com>
To: Florian Fainelli <f.fainelli@gmail.com>
To: Vladimir Oltean <olteanv@gmail.com>
To: "David S. Miller" <davem@davemloft.net>
To: Jakub Kicinski <kuba@kernel.org>
Cc: Helmut Grohne <helmut.grohne@intenta.de>
Cc: netdev@vger.kernel.org
Cc: linux-kernel@vger.kernel.org

 drivers/net/dsa/microchip/ksz8795.c | 12 +++++++++---
 drivers/net/dsa/microchip/ksz9477.c | 16 ++++++++++++----
 2 files changed, 21 insertions(+), 7 deletions(-)

diff --git a/drivers/net/dsa/microchip/ksz8795.c b/drivers/net/dsa/microchip/ksz8795.c
index f5779e152377..99183347053f 100644
--- a/drivers/net/dsa/microchip/ksz8795.c
+++ b/drivers/net/dsa/microchip/ksz8795.c
@@ -1190,7 +1190,9 @@ static const struct ksz_chip_data ksz8795_switch_chips[] = {
 		.num_alus = 0,
 		.num_statics = 8,
 		.cpu_ports = 0x10,	/* can be configured as cpu port */
-		.port_cnt = 4,		/* total physical port count */
+		.port_cnt = 4,		/* total physical port count, excluding
+					 * the cpu port
+					 */
 	},
 	{
 		.chip_id = 0x8794,
@@ -1199,7 +1201,9 @@ static const struct ksz_chip_data ksz8795_switch_chips[] = {
 		.num_alus = 0,
 		.num_statics = 8,
 		.cpu_ports = 0x10,	/* can be configured as cpu port */
-		.port_cnt = 3,		/* total physical port count */
+		.port_cnt = 3,		/* total physical port count, excluding
+					 * the cpu port
+					 */
 	},
 	{
 		.chip_id = 0x8765,
@@ -1208,7 +1212,9 @@ static const struct ksz_chip_data ksz8795_switch_chips[] = {
 		.num_alus = 0,
 		.num_statics = 8,
 		.cpu_ports = 0x10,	/* can be configured as cpu port */
-		.port_cnt = 4,		/* total physical port count */
+		.port_cnt = 4,		/* total physical port count, excluding
+					 * the cpu port
+					 */
 	},
 };
 
diff --git a/drivers/net/dsa/microchip/ksz9477.c b/drivers/net/dsa/microchip/ksz9477.c
index 2f5506ac7d19..e56122ffd495 100644
--- a/drivers/net/dsa/microchip/ksz9477.c
+++ b/drivers/net/dsa/microchip/ksz9477.c
@@ -1482,7 +1482,9 @@ static const struct ksz_chip_data ksz9477_switch_chips[] = {
 		.num_alus = 4096,
 		.num_statics = 16,
 		.cpu_ports = 0x7F,	/* can be configured as cpu port */
-		.port_cnt = 7,		/* total physical port count */
+		.port_cnt = 7,		/* total physical port count, including
+					 * the cpu port
+					 */
 		.phy_errata_9477 = true,
 	},
 	{
@@ -1492,7 +1494,9 @@ static const struct ksz_chip_data ksz9477_switch_chips[] = {
 		.num_alus = 4096,
 		.num_statics = 16,
 		.cpu_ports = 0x7F,	/* can be configured as cpu port */
-		.port_cnt = 7,		/* total physical port count */
+		.port_cnt = 7,		/* total physical port count, including
+					 * the cpu port
+					 */
 		.phy_errata_9477 = true,
 	},
 	{
@@ -1502,7 +1506,9 @@ static const struct ksz_chip_data ksz9477_switch_chips[] = {
 		.num_alus = 4096,
 		.num_statics = 16,
 		.cpu_ports = 0x07,	/* can be configured as cpu port */
-		.port_cnt = 3,		/* total port count */
+		.port_cnt = 3,		/* total physical port count, including
+					 * the cpu port
+					 */
 	},
 	{
 		.chip_id = 0x00956700,
@@ -1511,7 +1517,9 @@ static const struct ksz_chip_data ksz9477_switch_chips[] = {
 		.num_alus = 4096,
 		.num_statics = 16,
 		.cpu_ports = 0x7F,	/* can be configured as cpu port */
-		.port_cnt = 7,		/* total physical port count */
+		.port_cnt = 7,		/* total physical port count, including
+					 * the cpu port
+					 */
 	},
 };
 
-- 
2.17.1

Re: [PATCH net v1 2/2] net: dsa: microchip: improve port count comments

[Andrew Lunn]

On Sat, Dec 05, 2020 at 10:28:14AM -0500, Sven Van Asbroeck wrote:
> From: Sven Van Asbroeck <thesven73@gmail.com>
> 
> Port counts in microchip dsa drivers can be quite confusing:
> on the ksz8795, ksz_chip_data->port_cnt excludes the cpu port,
> yet on the ksz9477, it includes the cpu port.
> 
> Add comments to document this situation explicitly.

Rather than document it, we should make it uniform. Unless there is a
valid reason to require them to mean different things.

      Andrew

Re: [PATCH net v1 2/2] net: dsa: microchip: improve port count comments

[Jakub Kicinski]

On Tue, 8 Dec 2020 00:31:16 +0100 Andrew Lunn wrote:
> On Sat, Dec 05, 2020 at 10:28:14AM -0500, Sven Van Asbroeck wrote:
> > From: Sven Van Asbroeck <thesven73@gmail.com>
> > 
> > Port counts in microchip dsa drivers can be quite confusing:
> > on the ksz8795, ksz_chip_data->port_cnt excludes the cpu port,
> > yet on the ksz9477, it includes the cpu port.
> > 
> > Add comments to document this situation explicitly.  
> 
> Rather than document it, we should make it uniform. Unless there is a
> valid reason to require them to mean different things.

Agreed.

I wonder if we should make this effort target net-next.

My concern is that for the 3 port switch the cpu_ports variable is set
to 0x10, the same as for the 4 port one. Which makes me worried that 
if we just allow the "+ 1" - the CPU port will not actually hit the
register offsets its supposed to on 3 port platforms.

Since configuring the CPU port never worked here (AFAICT) we can view
this as a new feature / config option (even tho an important one).

So let's move to net-next, and we can "do this right".

Does that sound sane?

Re: [PATCH net v1 1/2] net: dsa: microchip: fix devicetree parsing of cpu node

[Sven Van Asbroeck]

Andrew, Jakub,

On Sat, Dec 5, 2020 at 10:28 AM Sven Van Asbroeck <thesven73@gmail.com> wrote:
>
> From: Sven Van Asbroeck <thesven73@gmail.com>
>
> On the ksz8795, if the devicetree contains a cpu node,
> devicetree parsing fails and the whole driver errors out.
>
> Fix the devicetree parsing code by making it use the
> correct number of ports.
>
> Fixes: 912aae27c6af ("net: dsa: microchip: really look for phy-mode in port nodes")
> Tested-by: Sven Van Asbroeck <thesven73@gmail.com> # ksz8795
> Signed-off-by: Sven Van Asbroeck <thesven73@gmail.com>
> ---

Any chance that this patch could still get merged?
I believe this will work fine on both ksz8795 and ksz9477, even though num_ports
is defined differently, because:

ksz8795:
/* set the real number of ports */
dev->ds->num_ports = dev->port_cnt + 1;
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/dsa/microchip/ksz8795.c?h=v5.10-rc7#n1266

ksz9477:
/* set the real number of ports */
dev->ds->num_ports = dev->port_cnt;
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/dsa/microchip/ksz9477.c?h=v5.10-rc7#n1585

Would it be possible to merge this into net, so users get working cpu nodes?
I don't think this will prevent you from harmonizing port_cnt in net-next.

Re: [PATCH net v1 1/2] net: dsa: microchip: fix devicetree parsing of cpu node

[Jakub Kicinski]

On Tue, 8 Dec 2020 10:33:28 -0500 Sven Van Asbroeck wrote:
> Andrew, Jakub,
> 
> On Sat, Dec 5, 2020 at 10:28 AM Sven Van Asbroeck <thesven73@gmail.com> wrote:
> >
> > From: Sven Van Asbroeck <thesven73@gmail.com>
> >
> > On the ksz8795, if the devicetree contains a cpu node,
> > devicetree parsing fails and the whole driver errors out.
> >
> > Fix the devicetree parsing code by making it use the
> > correct number of ports.
> >
> > Fixes: 912aae27c6af ("net: dsa: microchip: really look for phy-mode in port nodes")
> > Tested-by: Sven Van Asbroeck <thesven73@gmail.com> # ksz8795
> > Signed-off-by: Sven Van Asbroeck <thesven73@gmail.com>
> > ---  
> 
> Any chance that this patch could still get merged?
> I believe this will work fine on both ksz8795 and ksz9477, even though num_ports
> is defined differently, because:
> 
> ksz8795:
> /* set the real number of ports */
> dev->ds->num_ports = dev->port_cnt + 1;
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/dsa/microchip/ksz8795.c?h=v5.10-rc7#n1266
> 
> ksz9477:
> /* set the real number of ports */
> dev->ds->num_ports = dev->port_cnt;
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/dsa/microchip/ksz9477.c?h=v5.10-rc7#n1585
> 
> Would it be possible to merge this into net, so users get working cpu nodes?
> I don't think this will prevent you from harmonizing port_cnt in net-next.

What I was talking about in the email yesterday was 0x8794 support
in ksz8795.c. Is the cpu port configuration going to work there?
Isn't the CPU port always port 5 (index 4)?

It sure as hell looked like it until commit c9f4633b93ea ("net: dsa:
microchip: remove usage of mib_port_count") came along. I wonder if 
ksz8794 works on net-next :/

Re: [PATCH net v1 1/2] net: dsa: microchip: fix devicetree parsing of cpu node

[Sven Van Asbroeck]

On Tue, Dec 8, 2020 at 11:52 AM Jakub Kicinski <kuba@kernel.org> wrote:
>
> What I was talking about in the email yesterday was 0x8794 support
> in ksz8795.c. Is the cpu port configuration going to work there?
> Isn't the CPU port always port 5 (index 4)?

Understood. You expressed concern that .cpu_ports = 0x10 even on the 3-port
switch.

I noticed that .cpu_ports is never used in ksz8794.c, so I hoped the fix
would be fine. But digging a bit deeper, I see in ksz8795.c:

        dev->mib_port_cnt = TOTAL_PORT_NUM; /* = 5 */
        ...
        dev->cpu_port = dev->mib_port_cnt - 1; /* = 4 */

And that is unlikely to work on the 3-port switch.
So yeah, you are right, my patch won't fix the general issues here :(