net: dsa: Add support for mv88e6020 and mv88e6071
Subsystems (4)
cover net: dsa: Add support for mv88e6020 and mv88e6071
This patch series provides support for Marvell's mv88e6020 and mv 88e6071 switches and can be divided to two parts;
1. Code necessary to support aforementioned switches
2. Driver modification to support proper size of frames for mv88e6071
Lukasz Majewski (5):
net: dsa: mv88e6xxx: Add support for MV88E6071 switch
net: dsa: marvell: Provide per device information about max frame size
net: dsa: mv88e6071: Define max frame size (2048 bytes)
net: dsa: mv88e6071: Provide struct mv88e6xxx_ops
net: dsa: mv88e6071: Set .set_max_frame_size callback
Matthias Schiffer (4):
net: dsa: allow switch drivers to override default slave PHY addresses
net: dsa: mv88e6xxx: account for PHY base address offset in dual chip
mode
net: dsa: mv88e6xxx: implement get_phy_address
net: dsa: mv88e6xxx: add support for MV88E6020 switch
drivers/net/dsa/mv88e6xxx/chip.c | 102 +++++++++++++++++++++++++++- drivers/net/dsa/mv88e6xxx/chip.h | 4 ++ drivers/net/dsa/mv88e6xxx/global2.c | 2 +- drivers/net/dsa/mv88e6xxx/port.h | 2 + drivers/net/dsa/mv88e6xxx/smi.c | 4 ++ include/net/dsa.h | 1 + net/dsa/slave.c | 9 ++- 7 files changed, 119 insertions(+), 5 deletions(-)
Signature · 1 line
2.37.2
PATCH 1/9 net: dsa: allow switch drivers to override default slave PHY addresses
From: Matthias Schiffer <[email protected]>
Avoid having to define a PHY for every physical port when PHY addresses are fixed, but port index != PHY address.
--- include/net/dsa.h | 1 + net/dsa/slave.c | 9 +++++++-- 2 files changed, 8 insertions(+), 2 deletions(-)
include/net/dsa.h +1 -0
@@ -858,6 +858,7 @@ struct dsa_switch_ops {int (*port_setup)(struct dsa_switch *ds,int port);void (*port_teardown)(struct dsa_switch *ds,int port); +int (*get_phy_address)(struct dsa_switch *ds,int port);u32 (*get_phy_flags)(struct dsa_switch *ds,int port);/*
net/dsa/slave.c +7 -2
@@ -2273,7 +2273,7 @@ static int dsa_slave_phy_setup(struct net_device *slave_dev)struct device_node *port_dn = dp->dn;struct dsa_switch *ds = dp->ds;u32 phy_flags =0 ; -int ret; +int ret, addr; dp->pl_config.dev = &slave_dev->dev; dp->pl_config.type =PHYLINK_NETDEV ;
@@ -2299,7 +2299,12 @@ static int dsa_slave_phy_setup(struct net_device *slave_dev)/* We could not connect to a designated PHY or SFP, so try to * use the switch internal MDIO bus instead */ - ret =dsa_slave_phy_connect (slave_dev, dp->index, phy_flags); +if (ds->ops->get_phy_address) + addr = ds->ops->get_phy_address (ds, dp->index); +else + addr = dp->index; + + ret =dsa_slave_phy_connect (slave_dev, addr, phy_flags); }if (ret) { netdev_err(slave_dev, "failed to connect to PHY: %pe\n",
Signature · 1 line
2.37.2
Vladimir Oltean
maintainerRe: [PATCH 1/9] net: dsa: allow switch drivers to override default slave PHY addresses
Hi Lukasz,
Quoted 8 lines · Lukasz Majewski
From: Matthias Schiffer <[email protected]>
Avoid having to define a PHY for every physical port when PHY addresses are fixed, but port index != PHY address.
Signed-off-by Matthias Schiffer <[email protected]>Signed-off-by Lukasz Majewski <[email protected]>[Adjustments for newest kernel upstreaming]
(I got reminded by the message)
How are things going with the imx28 MTIP L2 switch? Upstreaming went silent on that.
Re: [PATCH 1/9] net: dsa: allow switch drivers to override default slave PHY addresses
Hi Vladimir,
Quoted 8 lines
Hi Lukasz,
Quoted 5 lines · Lukasz Majewski
From: Matthias Schiffer <[email protected]>
Avoid having to define a PHY for every physical port when PHY addresses are fixed, but port index != PHY address.
Quoted 7 lines
Quoted 3 lines
Signed-off-by Matthias Schiffer <[email protected]>Signed-off-by Lukasz Majewski <[email protected]>[Adjustments for newest kernel upstreaming](I got reminded by the message)
How are things going with the imx28 MTIP L2 switch? Upstreaming went silent on that.
Yes, this task has been postponed as:
- Customer decided to use Linux 4.19 CIP for which I've forward ported the original NXP patches [1][2]
- Considering the above - I would need to change the structure of the switch driver (which up till now is based on old - i.e. 2.6.3x - FEC driver) and modify the current FEC to add acceleration in similar way to TI's driver.
This is IMHO quite time consuming task ...
(The attempt to add it as DSA [2] was a bit less intrusive, but is conceptually wrong).
Links:
[1] - https://github.com/lmajewski/linux-imx28-l2switch/tree/master
[2] -
https://github.com/lmajewski/linux-imx28-l2switch/tree/imx28-v5.12-L2-upstream-RFC_v1
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
Andrew Lunn
maintainerRe: [PATCH 1/9] net: dsa: allow switch drivers to override default slave PHY addresses
In reply to Lukasz MajewskiQuoted 4 lines · Lukasz Majewski
From: Matthias Schiffer <[email protected]>
Avoid having to define a PHY for every physical port when PHY addresses are fixed, but port index != PHY address.
Please could you expand the commit message. What i think is going on is that for the lower device, port 0 has phy address 0, port 1 phy address 1. But the upper switch has port 0 phy address 16, port 1 phy addr 17?
6141 and 6341 set phy_base_addr to 0x10. Oddly, this is only used for the interrupt. So i assume these two devices also need device tree phy-handle descriptions?
It would be nice to fix the 6141 and the 6341 as well.
What might help with understanding is have the patch for the mv88e6xxx op first, and then wire it up in the core afterwards. Reviews tend to happen first to last, so either your commit message needs to explain what is coming, or you do things in the order which helps the reviewer the most.
Andrew
Re: [PATCH 1/9] net: dsa: allow switch drivers to override default slave PHY addresses
Hi Andrew,
Quoted 7 lines · Lukasz Majewski
Quoted 4 lines
From: Matthias Schiffer <[email protected]>
Avoid having to define a PHY for every physical port when PHY addresses are fixed, but port index != PHY address.
Please could you expand the commit message.
I've left the comment untouched from Matthias...
Quoted 4 lines
What i think is going on
is that for the lower device, port 0 has phy address 0, port 1 phy address 1. But the upper switch has port 0 phy address 16, port 1 phy addr 17?
To be more specific -> for mv88e6071 and mv88e6020:
PHY ports have SMI addresses from 0 to 0x6 and
Switch PORTS have addresses from 0x8 to 0xE
Global 2 -> 0x7
Global 1 -> 0xF
Access to PHYs is ONLY possible via indirect access from Global 2 (0x7 SMI addr) - offsets 0x18 and 0x19.
but :-)
there is also RO_LED/ADDR4 pin (bootstrap), which when set changes the SMI address of PHYs (from 0x00 - 0x04 to 0x10 to 0x14). This allows easy expansion of the number of ports for switch...
Quoted 2 lines
6141 and 6341 set
phy_base_addrto0x10.
It depends if the HW designer set this bootstrap pin low or high :-) (often this pin is not concerned until mainline/BSP driver is not working :-) )
As it costs $ to fix this - it is easier to add "quirk" to the code.
Quoted 3 lines
Oddly, this is only used for
the interrupt. So i assume these two devices also need device tree phy-handle descriptions?
I only have access to 6071 and 6020 switches.
Quoted 2 lines
It would be nice to fix the 6141 and the 6341 as well.
As Vladimir pointed out - many Marvell switches use the "old" DTS description ...
Quoted 6 lines
What might help with understanding is have the patch for the mv88e6xxx op first, and then wire it up in the core afterwards. Reviews tend to happen first to last, so either your commit message needs to explain what is coming, or you do things in the order which helps the reviewer the most.
I must admit, that I've just used code (his patch sert) from Matthias as a starting point (to keep his credits).
Quoted 2 lines
Andrew
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
Andrew Lunn
maintainerRe: [PATCH 1/9] net: dsa: allow switch drivers to override default slave PHY addresses
Quoted 5 lines
It depends if the HW designer set this bootstrap pin low or high :-) (often this pin is not concerned until mainline/BSP driver is not working :-) )
As it costs $ to fix this - it is easier to add "quirk" to the code.
Can you read the strapping configuration via the scratch registers? Then you can detect it at probe time and do the right thing.
Andrew
Re: [PATCH 1/9] net: dsa: allow switch drivers to override default slave PHY addresses
Hi Andrew,
Quoted 10 lines
Quoted 6 lines
It depends if the HW designer set this bootstrap pin low or high :-) (often this pin is not concerned until mainline/BSP driver is not working :-) )
As it costs $ to fix this - it is easier to add "quirk" to the code.
Can you read the strapping configuration via the scratch registers? Then you can detect it at probe time and do the right thing.
This is how I do it, yes. My impression is that some DTS descriptions have this hardcoded instead.
Quoted 1 line
Andrew
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
Re: [PATCH 1/9] net: dsa: allow switch drivers to override default slave PHY addresses
In reply to Lukasz MajewskiQuoted 8 lines · Lukasz Majewski
From: Matthias Schiffer <[email protected]>
Avoid having to define a PHY for every physical port when PHY addresses are fixed, but port index != PHY address.
Signed-off-by Matthias Schiffer <[email protected]>Signed-off-by Lukasz Majewski <[email protected]>[Adjustments for newest kernel upstreaming]
Seems unnecessary when this can be specified through Device Tree entirely.
It is convenient to be able not to declare the switch internal MDIO bus and how the PHY address to port mapping is established, but Device Tree remains the most flexible way of mapping all possible types of configurations.
Signature · 1 line
Florian
PATCH 2/9 net: dsa: mv88e6xxx: account for PHY base address offset in dual chip mode
In reply to Lukasz MajewskiFrom: Matthias Schiffer <[email protected]>
In dual chip mode (6250 family), not only global and port registers are shifted by sw_addr, but also the PHY addresses. Account for this in the IRQ mapping.
--- drivers/net/dsa/mv88e6xxx/chip.h | 1 + drivers/net/dsa/mv88e6xxx/global2.c | 2 +- drivers/net/dsa/mv88e6xxx/smi.c | 4 ++++ 3 files changed, 6 insertions(+), 1 deletion(-)
drivers/net/dsa/mv88e6xxx/chip.h +1 -0
@@ -328,6 +328,7 @@ struct mv88e6xxx_chip {const struct mv88e6xxx_bus_ops *smi_ops;struct mii_bus *bus;int sw_addr; +unsigned int phy_base_addr;/* Handles automatic disabling and re-enabling of the PHY * polling unit.
drivers/net/dsa/mv88e6xxx/global2.c +1 -1
@@ -1172,7 +1172,7 @@ int mv88e6xxx_g2_irq_mdio_setup(struct mv88e6xxx_chip *chip, err = irq; goto out; } - bus->irq[chip->info->phy_base_addr + phy] = irq; + bus->irq[chip->phy_base_addr + phy] = irq; }return 0 ; out:
drivers/net/dsa/mv88e6xxx/smi.c +4 -0
@@ -186,5 +186,9 @@ int mv88e6xxx_smi_init(struct mv88e6xxx_chip *chip,if (chip->smi_ops->init)return chip->smi_ops->init (chip); + chip->phy_base_addr = chip->info->phy_base_addr; +if (chip->info->dual_chip) + chip->phy_base_addr += sw_addr; +return 0 ; }
Signature · 1 line
2.37.2
Andrew Lunn
maintainerRe: [PATCH 2/9] net: dsa: mv88e6xxx: account for PHY base address offset in dual chip mode
Quoted 8 lines · Lukasz Majewski
From: Matthias Schiffer <[email protected]>
In dual chip mode (6250 family), not only global and port registers are shifted by
sw_addr, but also the PHY addresses. Account for this in the IRQ mapping.+++
b/drivers/net/dsa/mv88e6xxx/smi.c
drivers/net/dsa/mv88e6xxx/smi.c int mv88e6xxx_smi_init(struct mv88e6xxx_chip *chip, · line 193
Quoted 8 lines
drivers/net/dsa/mv88e6xxx/smi.c +4
-0@@ -186,5 +186,9 @@ int mv88e6xxx_smi_init(struct mv88e6xxx_chip *chip,if (chip->smi_ops->init)return chip->smi_ops->init (chip); + chip->phy_base_addr = chip->info->phy_base_addr; +if (chip->info->dual_chip) + chip->phy_base_addr += sw_addr; +return 0 ;
Again, reviewing first to last, i assume the will be a patch soon implementing get_phy_address(), and it will have the same logic. Why not call it here, a default implementation which returns info->phy_base_addr, and a version for 6250 which returns sw_addr.
Andrew
Re: [PATCH 2/9] net: dsa: mv88e6xxx: account for PHY base address offset in dual chip mode
Hi Andrew,
Quoted 12 lines · Lukasz Majewski
Quoted 11 lines
From: Matthias Schiffer <[email protected]>
In dual chip mode (6250 family), not only global and port registers are shifted by
sw_addr, but also the PHY addresses. Account for this in the IRQ mapping.+++
b/drivers/net/dsa/mv88e6xxx/smi.cdrivers/net/dsa/mv88e6xxx/smi.c +0
-0@@ -186,5 +186,9 @@ int mv88e6xxx_smi_init(struct mv88e6xxx_chip *chip, if (chip->smi_ops->init)return chip->smi_ops->init (chip);
Quoted 12 lines
drivers/net/dsa/mv88e6xxx/smi.c int mv88e6xxx_smi_init(struct mv88e6xxx_chip *chip, · line 193
Quoted 5 lines
drivers/net/dsa/mv88e6xxx/smi.c +4
-0+ chip->phy_base_addr = chip->info->phy_base_addr; +if (chip->info->dual_chip) + chip->phy_base_addr += sw_addr; +return 0 ;Again, reviewing first to last, i assume the will be a patch soon implementing
get_phy_address(), and it will have the same logic. Why not call it here, a default implementation which returnsinfo->phy_base_addr, and a version for 6250 which returnssw_addr.
I will squash Matthias patches and prepare new set of them with proper operation's ordering.
Quoted 1 line
Andrew
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
PATCH 3/9 net: dsa: mv88e6xxx: implement get_phy_address
In reply to Lukasz MajewskiFrom: Matthias Schiffer <[email protected]>
Avoid the need to specify a PHY for each physical port in the device tree when phy_base_addr is not 0 (6250 and 6341 families).
This change should be backwards-compatible with existing device trees, as it only adds sensible defaults where explicit definitions were required before.
--- drivers/net/dsa/mv88e6xxx/chip.c | 8 ++++++++ 1 file changed, 8 insertions(+)
drivers/net/dsa/mv88e6xxx/chip.c +8 -0
@@ -6362,6 +6362,13 @@ static enum dsa_tag_protocol mv88e6xxx_get_tag_protocol(struct dsa_switch *ds,return chip->tag_protocol; } +static int mv88e6xxx_get_phy_address (struct dsa_switch *ds,int port) +{ +struct mv88e6xxx_chip *chip = ds->priv; + +return chip->phy_base_addr + port; +} +static int mv88e6xxx_change_tag_protocol (struct dsa_switch *ds,enum dsa_tag_protocol proto) {
@@ -6887,6 +6894,7 @@ static const struct dsa_switch_ops mv88e6xxx_switch_ops = { .phylink_mac_link_up = mv88e6xxx_mac_link_up, .get_strings = mv88e6xxx_get_strings, .get_ethtool_stats = mv88e6xxx_get_ethtool_stats, + .get_phy_address = mv88e6xxx_get_phy_address, .get_sset_count = mv88e6xxx_get_sset_count, .port_enable = mv88e6xxx_port_enable, .port_disable = mv88e6xxx_port_disable,
Signature · 1 line
2.37.2
Vladimir Oltean
maintainerRe: [PATCH 3/9] net: dsa: mv88e6xxx: implement get_phy_address
Quoted 10 lines · Lukasz Majewski
From: Matthias Schiffer <[email protected]>
Avoid the need to specify a PHY for each physical port in the device tree when
phy_base_addris not 0 (6250 and 6341 families).This change should be backwards-compatible with existing device trees, as it only adds sensible defaults where explicit definitions were required before.
Signed-off-by Matthias Schiffer <[email protected]>
Needs your Signed-off-by tag as well.
Quoted 1 line
---
Would it be possible to do like armada-3720-turris-mox.dts does, and put the phy-handle in the device tree, avoiding the need for so many PHY address translation quirks?
If you're going to have U-Boot support for this switch as well, the phy-handle mechanism is the only thing that U-Boot supports, so device trees written in this way will work for both (and can be passed by U-Boot to Linux):
switch1@11 {
compatible = "marvell,mv88e6190";
reg = <0x11>;
dsa,member = <0 1>;
interrupt-parent = <&moxtet>;
interrupts = <MOXTET_IRQ_PERIDOT(1)>;
status = "disabled";
mdio {
#address-cells = <1>;
#size-cells = <0>;
switch1phy1: switch1phy1@1 {
reg = <0x1>;
};
switch1phy2: switch1phy2@2 {
reg = <0x2>;
};
switch1phy3: switch1phy3@3 {
reg = <0x3>;
};
switch1phy4: switch1phy4@4 {
reg = <0x4>;
};
switch1phy5: switch1phy5@5 {
reg = <0x5>;
};
switch1phy6: switch1phy6@6 {
reg = <0x6>;
};
switch1phy7: switch1phy7@7 {
reg = <0x7>;
};
switch1phy8: switch1phy8@8 {
reg = <0x8>;
};
};
ports {
#address-cells = <1>;
#size-cells = <0>;
port@1 {
reg = <0x1>;
label = "lan9";
phy-handle = <&switch1phy1>;
};
port@2 {
reg = <0x2>;
label = "lan10";
phy-handle = <&switch1phy2>;
};
port@3 {
reg = <0x3>;
label = "lan11";
phy-handle = <&switch1phy3>;
};
port@4 {
reg = <0x4>;
label = "lan12";
phy-handle = <&switch1phy4>;
};
port@5 {
reg = <0x5>;
label = "lan13";
phy-handle = <&switch1phy5>;
};
port@6 {
reg = <0x6>;
label = "lan14";
phy-handle = <&switch1phy6>;
};
port@7 {
reg = <0x7>;
label = "lan15";
phy-handle = <&switch1phy7>;
};
port@8 {
reg = <0x8>;
label = "lan16";
phy-handle = <&switch1phy8>;
};
switch1port9: port@9 {
reg = <0x9>;
label = "dsa";
phy-mode = "2500base-x";
managed = "in-band-status";
link = <&switch0port10>;
};
switch1port10: port@a {
reg = <0xa>;
label = "dsa";
phy-mode = "2500base-x";
managed = "in-band-status";
link = <&switch2port9>;
status = "disabled";
};
port-sfp@a {
reg = <0xa>;
label = "sfp";
sfp = <&sfp>;
phy-mode = "sgmii";
managed = "in-band-status";
status = "disabled";
};
};
};Andrew Lunn
maintainerRe: [PATCH 3/9] net: dsa: mv88e6xxx: implement get_phy_address
Quoted 8 lines
Would it be possible to do like
armada-3720-turris-mox.dtsdoes, and put the phy-handle in the device tree, avoiding the need for so many PHY address translation quirks?If you're going to have U-Boot support for this switch as well, the phy-handle mechanism is the only thing that U-Boot supports, so device trees written in this way will work for both (and can be passed by U-Boot to Linux):
This is how i expect any board using the MV88E6141 and MV88E6341 work. It has the same issue that it is not a 1:1 mapping.
Portability with U-boot is an interesting argument. Maybe there are patches to u-boot to add the same sort of quirks?
Andrew
Re: [PATCH 3/9] net: dsa: mv88e6xxx: implement get_phy_address
Hi Andrew,
Quoted 6 lines
Quoted 6 lines
Would it be possible to do like
armada-3720-turris-mox.dtsdoes, and put the phy-handle in the device tree, avoiding the need for so many PHY address translation quirks?If you're going to have U-Boot support for this switch as well, the phy-handle mechanism is the only thing that U-Boot supports,
Maybe in the generic case of PHY, yes (via Driver Model).
However, when you delve into the mv88e6xxx driver [1] - you would find that this is not supporting it yet ...
Quoted 6 lines
Quoted 3 lines
so
device trees written in this way will work for both (and can be passed by U-Boot to Linux):This is how i expect any board using the MV88E6141 and MV88E6341 work. It has the same issue that it is not a 1:1 mapping.
Please be aware that there is also majority of DTS entries, which use the "old" switch description ...
Quoted 3 lines
Portability with U-boot is an interesting argument. Maybe there are patches to u-boot to add the same sort of quirks?
As fair as I know - for the driver [1] - there was no ongoing effort recently.
Quoted 2 lines
Andrew
Links:
[1] -
https://source.denx.de/u-boot/u-boot/-/blob/master/drivers/net/phy/mv88e61xx.c
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
Vladimir Oltean
maintainerRe: [PATCH 3/9] net: dsa: mv88e6xxx: implement get_phy_address
Quoted 11 lines · Lukasz Majewski
Maybe in the generic case of PHY, yes (via Driver Model).
However, when you delve into the mv88e6xxx driver [1] - you would find that this is not supporting it yet ...
As fair as I know - for the driver [1] - there was no ongoing effort recently.
Links:
[1] -
https://source.denx.de/u-boot/u-boot/-/blob/master/drivers/net/phy/mv88e61xx.c
U-Boot mailing list is moving a bit slower than netdev (and the patch set is not yet applied despite being ready), but I was talking about this:
https://patchwork.ozlabs.org/project/uboot/list/?series=324983
If you study DM_DSA, you'll see that only supporting PHY connection via phy-handle (even if the PHY is internal) (or fixed-link, in lack of a PHY) was an absolutely deliberate decision.
Re: [PATCH 3/9] net: dsa: mv88e6xxx: implement get_phy_address
Hi Vladimir,
Quoted 14 lines · Lukasz Majewski
Quoted 5 lines
Maybe in the generic case of PHY, yes (via Driver Model).
However, when you delve into the mv88e6xxx driver [1] - you would find that this is not supporting it yet ...
Quoted 7 lines
As fair as I know - for the driver [1] - there was no ongoing effort recently.
Links:
[1] -
https://source.denx.de/u-boot/u-boot/-/blob/master/drivers/net/phy/mv88e61xx.c
Quoted 5 lines
U-Boot mailing list is moving a bit slower than netdev (and the patch set is not yet applied despite being ready), but I was talking about this:
https://patchwork.ozlabs.org/project/uboot/list/?series=324983
Thanks for sharing this link. It looks like this driver also supports switchnig addresses for Marvell ICs.
(This was the goal for mine previous patches).
Quoted 3 lines
If you study
DM_DSA, you'll see that only supporting PHY connection via phy-handle (even if the PHY is internal) (or fixed-link, in lack of a PHY) was an absolutely deliberate decision.
Ok. I do agree now - will adjust the code accordingly.
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
Re: [PATCH 3/9] net: dsa: mv88e6xxx: implement get_phy_address
In reply to Vladimir OlteanHi Vladimir,
Quoted 6 lines · Lukasz Majewski
Quoted 5 lines
From: Matthias Schiffer <[email protected]>
Avoid the need to specify a PHY for each physical port in the device tree when
phy_base_addris not 0 (6250 and 6341 families).
Quoted 8 lines
Quoted 6 lines
This change should be backwards-compatible with existing device trees, as it only adds sensible defaults where explicit definitions were required before.
Signed-off-by: Matthias Schiffer
<[email protected]>Needs your Signed-off-by tag as well.
Ok.
Quoted 6 lines
Quoted 2 lines
---
Would it be possible to do like
armada-3720-turris-mox.dtsdoes, and put the phy-handle in the device tree, avoiding the need for so many PHY address translation quirks?
As far as I can tell - the mv88e6xxx driver (./drivers/net/dsa/mv88e6xxx) now uses hardcoded values for each member of mv88e6xxx_info struct.
Those values are "port_base_addr" and "phy_base_addr". Those values could be read from DTS description as pasted below.
Quoted 122 lines
If you're going to have U-Boot support for this switch as well, the phy-handle mechanism is the only thing that U-Boot supports, so device trees written in this way will work for both (and can be passed by U-Boot to Linux):
switch1@11 { compatible = "marvell,mv88e6190"; reg = <0x11>; dsa,member = <0 1>; interrupt-parent = <&moxtet>; interrupts = <MOXTET_IRQ_PERIDOT(1)>; status = "disabled"; mdio { #address-cells = <1>; #size-cells = <0>; switch1phy1: switch1phy1@1 { reg = <0x1>; }; switch1phy2: switch1phy2@2 { reg = <0x2>; }; switch1phy3: switch1phy3@3 { reg = <0x3>; }; switch1phy4: switch1phy4@4 { reg = <0x4>; }; switch1phy5: switch1phy5@5 { reg = <0x5>; }; switch1phy6: switch1phy6@6 { reg = <0x6>; }; switch1phy7: switch1phy7@7 { reg = <0x7>; }; switch1phy8: switch1phy8@8 { reg = <0x8>; }; }; ports { #address-cells = <1>; #size-cells = <0>; port@1 { reg = <0x1>; label = "lan9"; phy-handle = <&switch1phy1>; }; port@2 { reg = <0x2>; label = "lan10"; phy-handle = <&switch1phy2>; }; port@3 { reg = <0x3>; label = "lan11"; phy-handle = <&switch1phy3>; }; port@4 { reg = <0x4>; label = "lan12"; phy-handle = <&switch1phy4>; }; port@5 { reg = <0x5>; label = "lan13"; phy-handle = <&switch1phy5>; }; port@6 { reg = <0x6>; label = "lan14"; phy-handle = <&switch1phy6>; }; port@7 { reg = <0x7>; label = "lan15"; phy-handle = <&switch1phy7>; }; port@8 { reg = <0x8>; label = "lan16"; phy-handle = <&switch1phy8>; }; switch1port9: port@9 { reg = <0x9>; label = "dsa"; phy-mode = "2500base-x"; managed = "in-band-status"; link = <&switch0port10>; }; switch1port10: port@a { reg = <0xa>; label = "dsa"; phy-mode = "2500base-x"; managed = "in-band-status"; link = <&switch2port9>; status = "disabled"; }; port-sfp@a { reg = <0xa>;
Quoted 8 lines
label = "sfp"; sfp = <&sfp>; phy-mode = "sgmii"; managed = "in-band-status"; status = "disabled"; }; }; };
The u-boot mailine has basic support for mv88e6071 and mv88e6020 (and also some 'extension' patches which are floating around [1]).
For the current code - I'm using:
mdio {
#address-cells = <1>;
#size-cells = <0>;
switch@0 {
compatible = "marvell,mv88e6250";
reg = <0x00>;
interrupt-parent = <&gpio2>;
interrupts = <20 IRQ_TYPE_LEVEL_LOW>;
interrupt-controller;
#interrupt-cells = <2>;
ports {
#address-cells = <1>;
#size-cells = <0>;
port@0 {
reg = <0>;
label = "lan1";
};
port@1 {
reg = <1>;
label = "lan2";
};
port@2 {
reg = <2>;
label = "lan3";
};
port@5 {
reg = <5>;
label = "cpu";
phy-mode = "rgmii-id";
ethernet = <&fec1>;
fixed-link {
speed = <100>;
full-duplex;
};
};
};The only "hack" which I see from time to time is the replacement of 'switch@0' with 'switch@8' to take into account the R0_LED/ADDRES4 bootstrap pin value (to shift up ports addresses).
Links:
[1] - https://lists.denx.de/pipermail/u-boot/2021-March/444827.html
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
PATCH 4/9 net: dsa: mv88e6xxx: add support for MV88E6020 switch
In reply to Lukasz MajewskiFrom: Matthias Schiffer <[email protected]>
A 6250 family switch with 5 internal PHYs and no PTP support.
--- drivers/net/dsa/mv88e6xxx/chip.c | 20 ++++++++++++++++++++ drivers/net/dsa/mv88e6xxx/chip.h | 1 + drivers/net/dsa/mv88e6xxx/port.h | 1 + 3 files changed, 22 insertions(+)
drivers/net/dsa/mv88e6xxx/chip.c +20 -0
@@ -5527,6 +5527,26 @@ static const struct mv88e6xxx_ops mv88e6393x_ops = { };static const struct mv88e6xxx_info mv88e6xxx_table[] = { + [MV88E6020 ] = { + .prod_num =MV88E6XXX_PORT_SWITCH_ID_PROD_6020 , + .family =MV88E6XXX_FAMILY_6250 , + .name ="Marvell 88E6020" , + .num_databases =64 , + .num_ports =7 , + .num_internal_phys =5 , + .max_vid =4095 , + .port_base_addr =0x8 , + .phy_base_addr =0x0 , + .global1_addr =0xf , + .global2_addr =0x7 , + .age_time_coeff =15000 , + .g1_irqs =9 , + .g2_irqs =5 , + .atu_move_port_mask =0xf , + .dual_chip = true, + .ops = &mv88e6250_ops, + }, + [MV88E6085 ] = { .prod_num =MV88E6XXX_PORT_SWITCH_ID_PROD_6085 , .family =MV88E6XXX_FAMILY_6097 ,
drivers/net/dsa/mv88e6xxx/chip.h +1 -0
@@ -54,6 +54,7 @@ enum mv88e6xxx_frame_mode {/* List of supported models */ enum mv88e6xxx_model { +MV88E6020 ,MV88E6085 ,MV88E6095 ,MV88E6097 ,
drivers/net/dsa/mv88e6xxx/port.h +1 -0
@@ -111,6 +111,7 @@/* Offset 0x03: Switch Identifier Register */ #define MV88E6XXX_PORT_SWITCH_ID 0x03#define MV88E6XXX_PORT_SWITCH_ID_PROD_MASK 0xfff0 +#define MV88E6XXX_PORT_SWITCH_ID_PROD_6020 0x0200#define MV88E6XXX_PORT_SWITCH_ID_PROD_6085 0x04a0#define MV88E6XXX_PORT_SWITCH_ID_PROD_6095 0x0950#define MV88E6XXX_PORT_SWITCH_ID_PROD_6097 0x0990
Signature · 1 line
2.37.2
PATCH 5/9 net: dsa: mv88e6xxx: Add support for MV88E6071 switch
In reply to Lukasz MajewskiThis patch provides support for MV88E6071 Marvell switch.
--- drivers/net/dsa/mv88e6xxx/chip.c | 21 +++++++++++++++++++++ drivers/net/dsa/mv88e6xxx/chip.h | 1 + drivers/net/dsa/mv88e6xxx/port.h | 1 + 3 files changed, 23 insertions(+)
drivers/net/dsa/mv88e6xxx/chip.c +21 -0
@@ -5547,6 +5547,27 @@ static const struct mv88e6xxx_info mv88e6xxx_table[] = { .ops = &mv88e6250_ops, }, + [MV88E6071] = { + .prod_num = MV88E6XXX_PORT_SWITCH_ID_PROD_6071, + .family = MV88E6XXX_FAMILY_6250, + .name = "Marvell 88E6071", + .num_databases = 64, + .num_ports = 7, + .num_internal_phys = 5, + .max_vid = 4095, + .port_base_addr = 0x08, + .phy_base_addr = 0x00, + .global1_addr = 0x0f, + .global2_addr = 0x07, + .age_time_coeff = 15000, + .g1_irqs = 9, + .g2_irqs = 5, + .atu_move_port_mask = 0xf, + .dual_chip = true, + .ptp_support = true, + .ops = &mv88e6250_ops, + }, + [MV88E6085] = { .prod_num = MV88E6XXX_PORT_SWITCH_ID_PROD_6085, .family = MV88E6XXX_FAMILY_6097,
drivers/net/dsa/mv88e6xxx/chip.h +1 -0
@@ -55,6 +55,7 @@ enum mv88e6xxx_frame_mode {/* List of supported models */ enum mv88e6xxx_model {MV88E6020 , +MV88E6071 ,MV88E6085 ,MV88E6095 ,MV88E6097 ,
drivers/net/dsa/mv88e6xxx/port.h +1 -0
@@ -112,6 +112,7 @@#define MV88E6XXX_PORT_SWITCH_ID 0x03#define MV88E6XXX_PORT_SWITCH_ID_PROD_MASK 0xfff0#define MV88E6XXX_PORT_SWITCH_ID_PROD_6020 0x0200 +#define MV88E6XXX_PORT_SWITCH_ID_PROD_6071 0x0710#define MV88E6XXX_PORT_SWITCH_ID_PROD_6085 0x04a0#define MV88E6XXX_PORT_SWITCH_ID_PROD_6095 0x0950#define MV88E6XXX_PORT_SWITCH_ID_PROD_6097 0x0990
Signature · 1 line
2.37.2
PATCH 6/9 net: dsa: marvell: Provide per device information about max frame size
In reply to Lukasz MajewskiDifferent Marvell DSA switches support different size of max frame bytes to be sent.
For example mv88e6185 supports max 1632B, which is now in-driver standard value. On the other hand - mv88e6071 supports 2048 bytes.
As this value is internal and may be different for each switch IC new entry in struct mv88e6xxx_info has been added to store it.
When the 'max_frame_size' is not defined (and hence zeroed by the kvzalloc()) the default of 1632 bytes is used.
--- drivers/net/dsa/mv88e6xxx/chip.c | 12 ++++++++++-- drivers/net/dsa/mv88e6xxx/chip.h | 1 + 2 files changed, 11 insertions(+), 2 deletions(-)
drivers/net/dsa/mv88e6xxx/chip.c +10 -2
@@ -3542,11 +3542,19 @@ static int mv88e6xxx_setup_port(struct mv88e6xxx_chip *chip, int port)static int mv88e6xxx_get_max_mtu (struct dsa_switch *ds,int port) {struct mv88e6xxx_chip *chip = ds->priv; +int max_frame_size;if (chip->info->ops->port_set_jumbo_size)return 10240 -VLAN_ETH_HLEN -EDSA_HLEN -ETH_FCS_LEN ; -else if (chip->info->ops->set_max_frame_size) -return 1632 -VLAN_ETH_HLEN -EDSA_HLEN -ETH_FCS_LEN ; +else if (chip->info->ops->set_max_frame_size) { +if (chip->info->max_frame_size) + max_frame_size = chip->info->max_frame_size; +else + max_frame_size =1632 ; + +return max_frame_size -VLAN_ETH_HLEN -EDSA_HLEN -ETH_FCS_LEN ; + } +return 1522 -VLAN_ETH_HLEN -EDSA_HLEN -ETH_FCS_LEN ; }
drivers/net/dsa/mv88e6xxx/chip.h +1 -0
@@ -132,6 +132,7 @@ struct mv88e6xxx_info {unsigned int num_ports;unsigned int num_internal_phys;unsigned int num_gpio; +unsigned int max_frame_size;unsigned int max_vid;unsigned int max_sid;unsigned int port_base_addr;
Signature · 1 line
2.37.2
Andrew Lunn
maintainerRe: [PATCH 6/9] net: dsa: marvell: Provide per device information about max frame size
Quoted 11 lines · Lukasz Majewski
Different Marvell DSA switches support different size of max frame bytes to be sent.
For example mv88e6185 supports max 1632B, which is now in-driver standard value. On the other hand - mv88e6071 supports 2048 bytes.
As this value is internal and may be different for each switch IC new entry in struct
mv88e6xxx_infohas been added to store it.When the '
max_frame_size' is not defined (and hence zeroed by thekvzalloc()) the default of 1632 bytes is used.
I would prefer every entry states the value. That both simplifies the code, and probably reduces the likelihood of somebody forgetting to set it when adding a new chip.
Andrew
Re: [PATCH 6/9] net: dsa: marvell: Provide per device information about max frame size
Hi Andrew,
Quoted 7 lines · Lukasz Majewski
Quoted 6 lines
Different Marvell DSA switches support different size of max frame bytes to be sent.
For example mv88e6185 supports max 1632B, which is now in-driver standard value. On the other hand - mv88e6071 supports 2048 bytes.
Quoted 7 lines
Quoted 5 lines
As this value is internal and may be different for each switch IC new entry in struct
mv88e6xxx_infohas been added to store it.When the '
max_frame_size' is not defined (and hence zeroed by thekvzalloc()) the default of 1632 bytes is used.I would prefer every entry states the value.
You mean to add it explicitly (i.e. for each supported switch version) to the struct mv88e6xxx_ops ?
Quoted 3 lines
That both simplifies the
code, and probably reduces the likelihood of somebody forgetting to set it when adding a new chip.
Ok.
Quoted 2 lines
Andrew
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
Andrew Lunn
maintainerRe: [PATCH 6/9] net: dsa: marvell: Provide per device information about max frame size
Quoted 12 lines · Lukasz Majewski
Hi Andrew,
Quoted 10 lines · Lukasz Majewski
Quoted 9 lines
Different Marvell DSA switches support different size of max frame bytes to be sent.
For example mv88e6185 supports max 1632B, which is now in-driver standard value. On the other hand - mv88e6071 supports 2048 bytes.
As this value is internal and may be different for each switch IC new entry in struct
mv88e6xxx_infohas been added to store it.
Quoted 7 lines
Quoted 4 lines
Quoted 2 lines
When the '
max_frame_size' is not defined (and hence zeroed by thekvzalloc()) the default of 1632 bytes is used.I would prefer every entry states the value.
You mean to add it explicitly (i.e. for each supported switch version) to the struct
mv88e6xxx_ops?
To the info structure. You added it for just your devices and left it to 0 for the rest. Rather than special case 0, always set the value and remove the special case. Maybe even -EINVAL if you find a 0 there.
Andrew
Re: [PATCH 6/9] net: dsa: marvell: Provide per device information about max frame size
Hi Andrew,
Quoted 18 lines · Lukasz Majewski
Quoted 17 lines
Hi Andrew,
Quoted 15 lines · Lukasz Majewski
Quoted 13 lines
Different Marvell DSA switches support different size of max frame bytes to be sent.
For example mv88e6185 supports max 1632B, which is now in-driver standard value. On the other hand - mv88e6071 supports 2048 bytes.
As this value is internal and may be different for each switch IC new entry in struct
mv88e6xxx_infohas been added to store it.When the '
max_frame_size' is not defined (and hence zeroed by thekvzalloc()) the default of 1632 bytes is used.
Quoted 9 lines
Quoted 4 lines
Quoted 1 line
I would prefer every entry states the value.
You mean to add it explicitly (i.e. for each supported switch version) to the struct
mv88e6xxx_ops?To the info structure. You added it for just your devices and left it to 0 for the rest. Rather than special case 0, always set the value and remove the special case. Maybe even
-EINVALif you find a 0 there.
Ok.
Quoted 1 line
Andrew
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
PATCH 7/9 net: dsa: mv88e6071: Define max frame size (2048 bytes)
In reply to Lukasz MajewskiAccroding to the documentation - the mv88e6071 can support frame size up to 2048 bytes.
--- drivers/net/dsa/mv88e6xxx/chip.c | 1 + 1 file changed, 1 insertion(+)
drivers/net/dsa/mv88e6xxx/chip.c +1 -0
@@ -5563,6 +5563,7 @@ static const struct mv88e6xxx_info mv88e6xxx_table[] = { .num_ports = 7, .num_internal_phys = 5, .max_vid = 4095, + .max_frame_size = 2048, .port_base_addr = 0x08, .phy_base_addr = 0x00, .global1_addr = 0x0f,
Signature · 1 line
2.37.2
Andrew Lunn
maintainerRe: [PATCH 7/9] net: dsa: mv88e6071: Define max frame size (2048 bytes)
Quoted 2 lines · Lukasz Majewski
Accroding to the documentation - the mv88e6071 can support frame size up to 2048 bytes.
Since the mv88e6020 is in the same family, it probably is the same? And what about the mv88e66220?
Andrew
Re: [PATCH 7/9] net: dsa: mv88e6071: Define max frame size (2048 bytes)
Hi Andrew,
Quoted 5 lines · Lukasz Majewski
Quoted 2 lines
Accroding to the documentation - the mv88e6071 can support frame size up to 2048 bytes.
Since the mv88e6020 is in the same family, it probably is the same?
Yes it is 2048 B
Quoted 1 line
And what about the mv88e66220?
You mean mv88e6220 ?
IIRC they are from the same family of ICs, so my guess :-) is that they also have the same value.
Quoted 2 lines
Andrew
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
Andrew Lunn
maintainerRe: [PATCH 7/9] net: dsa: mv88e6071: Define max frame size (2048 bytes)
Quoted 6 lines · Lukasz Majewski
Hi Andrew,
Quoted 4 lines · Lukasz Majewski
Quoted 2 lines
Accroding to the documentation - the mv88e6071 can support frame size up to 2048 bytes.
Quoted 7 lines
Quoted 1 line
Since the mv88e6020 is in the same family, it probably is the same?
Yes it is 2048 B
Quoted 1 line
And what about the mv88e66220?
You mean mv88e6220 ?
Upps, sorry, yes.
Quoted 3 lines
IIRC they are from the same family of ICs, so my guess :-) is that they also have the same value.
My point being, you created a new _ops structure when i don't think it is needed.
Andrew
Re: [PATCH 7/9] net: dsa: mv88e6071: Define max frame size (2048 bytes)
Hi Andrew,
Quoted 16 lines · Lukasz Majewski
Quoted 6 lines
Hi Andrew,
Quoted 4 lines · Lukasz Majewski
Quoted 2 lines
Accroding to the documentation - the mv88e6071 can support frame size up to 2048 bytes.
Quoted 8 lines
Quoted 2 lines
Since the mv88e6020 is in the same family, it probably is the same?
Yes it is 2048 B
Quoted 1 line
And what about the mv88e66220?
You mean mv88e6220 ?
Quoted 8 lines
Upps, sorry, yes.
Quoted 3 lines
IIRC they are from the same family of ICs, so my guess :-) is that they also have the same value.
My point being, you created a new
_opsstructure when i don't think it is needed.
This was mostly my precaution to not introduce regression for other supported ICs.
However, this shall not be the problem as long as the max supported frame size is set for each of it. Then calling set_max_frame_size() callback will always provide correct value.
Quoted 2 lines
Andrew
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]
PATCH 8/9 net: dsa: mv88e6071: Provide struct mv88e6xxx_ops
In reply to Lukasz MajewskiThis structure is the same as the one for mv88e6250 and is supposed to ease further development of mv88e6071 (to put functions not present on the mv88e6250 device).
--- drivers/net/dsa/mv88e6xxx/chip.c | 43 ++++++++++++++++++++++++++++++-- 1 file changed, 41 insertions(+), 2 deletions(-)
drivers/net/dsa/mv88e6xxx/chip.c +41 -2
@@ -4961,6 +4961,45 @@ static const struct mv88e6xxx_ops mv88e6250_ops = { .phylink_get_caps = mv88e6250_phylink_get_caps, }; +static const struct mv88e6xxx_ops mv88e6071_ops = { +/* MV88E6XXX_FAMILY_6071 */ + .ieee_pri_map = mv88e6250_g1_ieee_pri_map, + .ip_pri_map = mv88e6085_g1_ip_pri_map, + .irl_init_all = mv88e6352_g2_irl_init_all, + .get_eeprom = mv88e6xxx_g2_get_eeprom16, + .set_eeprom = mv88e6xxx_g2_set_eeprom16, + .set_switch_mac = mv88e6xxx_g2_set_switch_mac, + .phy_read = mv88e6xxx_g2_smi_phy_read, + .phy_write = mv88e6xxx_g2_smi_phy_write, + .port_set_link = mv88e6xxx_port_set_link, + .port_sync_link = mv88e6xxx_port_sync_link, + .port_set_rgmii_delay = mv88e6352_port_set_rgmii_delay, + .port_set_speed_duplex = mv88e6250_port_set_speed_duplex, + .port_tag_remap = mv88e6095_port_tag_remap, + .port_set_frame_mode = mv88e6351_port_set_frame_mode, + .port_set_ucast_flood = mv88e6352_port_set_ucast_flood, + .port_set_mcast_flood = mv88e6352_port_set_mcast_flood, + .port_set_ether_type = mv88e6351_port_set_ether_type, + .port_egress_rate_limiting = mv88e6097_port_egress_rate_limiting, + .port_pause_limit = mv88e6097_port_pause_limit, + .port_disable_pri_override = mv88e6xxx_port_disable_pri_override, + .stats_snapshot = mv88e6320_g1_stats_snapshot, + .stats_set_histogram = mv88e6095_g1_stats_set_histogram, + .stats_get_sset_count = mv88e6250_stats_get_sset_count, + .stats_get_strings = mv88e6250_stats_get_strings, + .stats_get_stats = mv88e6250_stats_get_stats, + .set_cpu_port = mv88e6095_g1_set_cpu_port, + .set_egress_port = mv88e6095_g1_set_egress_port, + .watchdog_ops = &mv88e6250_watchdog_ops, + .mgmt_rsvd2cpu = mv88e6352_g2_mgmt_rsvd2cpu, + .pot_clear = mv88e6xxx_g2_pot_clear, + .reset = mv88e6250_g1_reset, + .vtu_getnext = mv88e6185_g1_vtu_getnext, + .vtu_loadpurge = mv88e6185_g1_vtu_loadpurge, + .avb_ops = &mv88e6352_avb_ops, + .ptp_ops = &mv88e6250_ptp_ops, +}; +static const struct mv88e6xxx_ops mv88e6290_ops = {/* MV88E6XXX_FAMILY_6390 */ .setup_errata = mv88e6390_setup_errata,
@@ -5552,7 +5591,7 @@ static const struct mv88e6xxx_info mv88e6xxx_table[] = { .g2_irqs = 5, .atu_move_port_mask = 0xf, .dual_chip = true, - .ops = &mv88e6250_ops, + .ops = &mv88e6071_ops, }, [MV88E6071] = {
@@ -5574,7 +5613,7 @@ static const struct mv88e6xxx_info mv88e6xxx_table[] = { .atu_move_port_mask = 0xf, .dual_chip = true, .ptp_support = true, - .ops = &mv88e6250_ops, + .ops = &mv88e6071_ops, }, [MV88E6085] = {
Signature · 1 line
2.37.2
PATCH 9/9 net: dsa: mv88e6071: Set .set_max_frame_size callback
In reply to Lukasz MajewskiThe .set_max_frame_size is now set to the mv88e6185_g1_set_max_frame_size() function.
The global switch control register (0x4 offset) used as well as the bit (10) are the same.
The only difference is the misleading suffix (1632) as the mv88e6071/mv88e6020 supports 2048 bytes as a maximal size of the frame.
--- drivers/net/dsa/mv88e6xxx/chip.c | 1 + 1 file changed, 1 insertion(+)
drivers/net/dsa/mv88e6xxx/chip.c +1 -0
@@ -4998,6 +4998,7 @@ static const struct mv88e6xxx_ops mv88e6071_ops = { .vtu_loadpurge = mv88e6185_g1_vtu_loadpurge, .avb_ops = &mv88e6352_avb_ops, .ptp_ops = &mv88e6250_ptp_ops, + .set_max_frame_size = mv88e6185_g1_set_max_frame_size, };static const struct mv88e6xxx_ops mv88e6290_ops = {
Signature · 1 line
2.37.2
Andrew Lunn
maintainerRe: [PATCH 9/9] net: dsa: mv88e6071: Set .set_max_frame_size callback
Quoted 9 lines · Lukasz Majewski
The .set_max_frame_size is now set to the
mv88e6185_g1_set_max_frame_size()function.The global switch control register (
0x4offset) used as well as the bit (10) are the same.The only difference is the misleading suffix (1632) as the mv88e6071/mv88e6020 supports 2048 bytes as a maximal size of the frame.
Are you really sure that different members of the 6250 family have different maximum frame sizes?
Marvells GPL DSDT SDK has:
#define G1_DEV_88ESPANNAK_FAMILY (DEV_88E3020 | DEV_88E6020 | DEV_88E6070 | DEV_88E6071 | DEV_88E6220 | DEV_88E6250 )
The differences within a family tend to be the number of ports, if PTP is provided, if AVB is provided etc.
Andrew
Re: [PATCH 9/9] net: dsa: mv88e6071: Set .set_max_frame_size callback
Hi Andrew,
Quoted 11 lines · Lukasz Majewski
Quoted 9 lines
The .set_max_frame_size is now set to the
mv88e6185_g1_set_max_frame_size()function.The global switch control register (
0x4offset) used as well as the bit (10) are the same.The only difference is the misleading suffix (1632) as the mv88e6071/mv88e6020 supports 2048 bytes as a maximal size of the frame.
Quoted 7 lines
Are you really sure that different members of the 6250 family have different maximum frame sizes?
Marvells GPL DSDT SDK has:
#define
G1_DEV_88ESPANNAK_FAMILY(DEV_88E3020|DEV_88E6020|DEV_88E6070|DEV_88E6071|DEV_88E6220|DEV_88E6250)
From the above list - all but 88E3020 are from the same IC "family" (they are listed in the same documentation pdf). They all have the 2048 B max packet size.
I've re-used the mv88e6185_g1_set_max_frame_size, as it does the job
Quoted 5 lines
The differences within a family tend to be the number of ports, if PTP is provided, if AVB is provided etc.
Andrew
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: [email protected]