How to disable network support in U-Boot?

Hi, Prabhakar

I disabled 'CONFIG_DRIVER_TI_EMAC' in 'include/configs/davinci_dm365evm.h'

But I can see now linker errors:

drivers/net/libnet.o: In function `davinci_emac_initialize':
drivers/net/davinci_emac.c:691: undefined reference to `lxt972_init_phy'
drivers/net/davinci_emac.c:691: undefined reference to `lxt972_is_phy_connected'
drivers/net/davinci_emac.c:691: undefined reference to `lxt972_get_link_speed'
drivers/net/davinci_emac.c:691: undefined reference to `lxt972_auto_negotiate'
drivers/net/davinci_emac.c:691: undefined reference to `dp83848_init_phy'
drivers/net/davinci_emac.c:691: undefined reference to `dp83848_is_phy_connected'
drivers/net/davinci_emac.c:691: undefined reference to `dp83848_get_link_speed'
drivers/net/davinci_emac.c:691: undefined reference to `dp83848_auto_negotiate'
drivers/net/davinci_emac.c:691: undefined reference to `et1011c_get_link_speed'
make[1]: *** [u-boot] Error 1

Hi Senchuss,

disabled in the sense you completely removed it right? I find no issue when i build(no linker errors).

what is the source of your u-boot is denx git?

Regards

Prabhakar Lad

U-Boot sources used: http://arago-project.org/git/projects/?p=u-boot-davinci.git;a=commit;h=1254a0b58d900b7035cfb36bf4e1b05dfbf09997

Compiler: Mentor CodeSourcery 2011.03-42 (arm-none-eabi), i.e. "bare metal" arm compiler (not for linux).

Opppsss... i make rebuild now (i.e. make clean  before make build). It's built without errors!!!

Yes! You are right!

I can see now in U-Boot

Net:   No ethernet found.

Thank you. You've helped me!

But output 'u-boot.bin' still looks like big ~288Kb.

I'll try to disable something else (cmd_ping and etc)

THANK YOU!

Here a small addition.

in 'include/config_cmd_default.h'

// #define CONFIG_CMD_NET        /* bootp, tftpboot, rarpboot    */
// #define CONFIG_CMD_NFS        /* NFS support            */

The size of 'u-boot.bin' decreased by 2 Kb.

A string with the "Net: "-message is no longer printed on the U-Boot screen. Good!!!

A small note about U-Boot programming style. U-Boot tends to use #Undef instead of commenting out #define. This is required usually to undo a previous define, eg config_cmd_default.h. To remove most network stuff:

#undef CONFIG_DRIVER_TI_EMAC
#undef CONFIG_CMD_DHCP
#undef CONFIG_CMD_MII
#undef CONFIG_CMD_NET
#undef CONFIG_CMD_PING
#undef CONFIG_CMD_NFS

If your platform has MAC in EEPROM. I believe the MAC address and EEPROM code still get included and executed. Removing thats takes a bit more hacking and slashing. My platform included USB host code. That takes up a lot of room as well.

Hi, Prabhakar !

Hi, Norman !

Excuse me please!
The problem is again relevant.
I just now discovered that if I disable the support of the network, it will crash i2c-subsystem.
Unbelievable!

I've restored all the definitions back as it was originally.

We will now discuss only about the "CONFIG_DRIVER_TI_EMAC",
all other 'defines' are not touched (all of them are restored to initial 'davinci_dm365evm.h' configuration).

We are touch this line only.

in "include/configs/davinci_dm365evm.h"
#define CONFIG_DRIVER_TI_EMAC
Build, Flash and Test.
u-boot# i2c probe
It's works. U-boot discovers my i2c-chips and shows their addresses (I can see three addresses, it's right). All is good.

in "include/configs/davinci_dm365evm.h"
#undef CONFIG_DRIVER_TI_EMAC
Build, Flash and Test.
u-boot# i2c probe
Displays a sequence "01 02 03 04 ... 7F"
This means that the u-boot can't finds i2c-devices, i.e. the i2c subsystem crashed.

Unbelievable!
The network subsystem has influence to i2c-subsytem?

Hi senchuss ,

What Norman suggested should work out.

hi Norman ,the piece of code you suggested might go as patch or is it a rare requirement of disabling the ethernet ?

Regards

--Prabhakar Lad

Hi Prabhakar,

The modification I suggested is just a quick hack. A proper patch would move the I2C pinmux code out of board_eth_init() to avoid configuring the pinmux twice. Although there is probably no harm. The I2C pinmux code would need to test both I2C and EMAC flags. Ideally, a board_i2c_init() function is needed. Hooking that in would take quite a bit more work.

Norman

Hi !

That's right exactly.

U-Boot: v2010.12_DAVINCIPSP_03.21.00.04 (from arago-project.org)

File: board/davinci/dm365evm/dm365evm.c

Lines: 43-81

---

  43 #ifdef CONFIG_DRIVER_TI_EMAC
  44 int board_eth_init(bd_t *bis)
  45 {
  46         uint8_t eeprom_enetaddr[6];
...
  70         /* Configure I2C pins so that EEPROM can be read */
  71         writel((readl(PINMUX3) | 0x01400000), PINMUX3);
...
  80 }
  81 #endif

---

If we disable "CONFIG_DRIVER_TI_EMAC" then no PINMUX3 settings for I2C (GIO21=2=I2C_SDA, GIO20=2=I2C_SCL).

Also, I confess that I2C-support in register 'pinmux3' is not enabled in my 'ubl'-code.

There are two ways.

1: Enable I2C in U-Boot ( "CONFIG_DRIVER_TI_EMAC" independent)

in "board/davinci/dm365evm/dm365evm.c"

int board_init(void)
{
    gd->bd->bi_arch_number = MACH_TYPE_DAVINCI_DM365_EVM;
    gd->bd->bi_boot_params = PHYS_SDRAM_1 + 0x100;
#ifdef CONFIG_DRIVER_DAVINCI_I2C
// Configure I2C pins
    writel((readl(PINMUX3) | 0x01400000), PINMUX3);
#endif
    return 0;
}

2. Enable I2C in your UBL-code

(for example carefully check PINMUX3 in your 'flash-utils/DM36x/Common/src/device.c' file.

---

I do not know what is more correct ideology.
At first, 'U-boot' does everything right. I2C is enabled because it needs access to the EEPROM to read the MAC if "CONFIG_DRIVER_TI_EMAC" is defined.
If not, there is no reason to read MAC from EEPROM.
At second, UBL has main goal that is setup DDR and NAND and load U-Boot. I.e. there is no reason to enable I2C in UBL.

But I think that if U-boot has i2c-commands (like 'i2c probe', 'i2c read' etc) then i2c must be enabled in U-boot.

Thank you Prabhakar! Thank you Norman!

Now it's works!

You've helped me!