gadget g_serial on mini2440SDK and kernel 2.6.32

Hi,
we want to use the Mini2440 with the gadget g_serial .
We configure it with the downloaded sources 2.6.32.2-mini2440_20100113.

When booting in NOR in supervivi, the Mini2440 is recognized by the host.
Then I press [b] to boot the image, and the Mini2440 is recognized as a
serial gadget. It's the correct way.

But, when booting directly in NAND, the gadget is mounted by the kernel
without any error (/dev/ttyGS0 exists) and the host PC never see anything
connected.

What is missing in my kernel that is done by supervivi ?

Thanks

Michel

I'm pulling my hair out over the same problem. Let me know if you have any
progress.

My analysis so far: It seems like the low-level UDC driver isn't setting up
the device right. For instance, when I boot through supervivi:

    [root@forehead0 /]# cat /sys/kernel/debug/s3c2410_udc/registers   
    FUNC_ADDR_REG  : 0x000D
    PWR_REG        : 0x0000
    EP_INT_REG     : 0x0000
    USB_INT_REG    : 0x0000
    EP_INT_EN_REG  : 0x001F
    USB_INT_EN_REG : 0x0004
    EP0_CSR        : 0x0000
    EP1_I_CSR1     : 0x0000
    EP1_I_CSR2     : 0x0030
    EP1_O_CSR1     : 0x0000
    EP1_O_CSR2     : 0x0030
    EP2_I_CSR1     : 0x0000
    EP2_I_CSR2     : 0x0010
    EP2_O_CSR1     : 0x0000
    EP2_O_CSR2     : 0x0010

but when I boot directly from nand:

    [root@forehead0 /]# cat /sys/kernel/debug/s3c2410_udc/registers   
    FUNC_ADDR_REG  : 0x0000
    PWR_REG        : 0x0008
    EP_INT_REG     : 0x0000
    USB_INT_REG    : 0x0000
    EP_INT_EN_REG  : 0x0001
    USB_INT_EN_REG : 0x0005
    EP0_CSR        : 0x0000
    EP1_I_CSR1     : 0x0000
    EP1_I_CSR2     : 0x0020
    EP1_O_CSR1     : 0x0000
    EP1_O_CSR2     : 0x0020
    EP2_I_CSR1     : 0x0000
    EP2_I_CSR2     : 0x0020
    EP2_O_CSR1     : 0x0000
    EP2_O_CSR2     : 0x0020

Note the bit PWR_REG&0x08, which indicates the USB device interface is
receiving reset.

Aha, the problem is that there is a pullup resistor on the USB line which
is supported for many boards, but not by the drivers for the mini2440. Here
is a patch that seems like the right way of fixing it:


Index: linux/arch/arm/mach-s3c2440/mach-mini2440.c
===================================================================
--- linux/arch/arm/mach-s3c2440/mach-mini2440.c  (revision 5)
+++ linux/arch/arm/mach-s3c2440/mach-mini2440.c  (working copy)
@@ -66,6 +66,7 @@
 #include <plat/nand.h>
 #include <plat/pm.h>
 #include <plat/mci.h>
+#include <plat/udc.h>
 
 
 #include <sound/s3c24xx_uda134x.h>
@@ -290,6 +291,34 @@
   .ignore_unset_ecc = 1,
 };
 
+/*
+  Enable the pullup resistor for the device-side ("gadget") USB port.
According to the
+  Microarmsystems.com schematic there is a 1.5k resistor (R29) between
GPC5 and the D+
+  line of the USB.
+ */
+static void mini2440_udc_pullup(enum s3c2410_udc_cmd_e cmd)
+{
+  switch (cmd) {
+  case S3C2410_UDC_P_ENABLE :
+                printk(KERN_INFO "mini2440_udc_pullup enable\n");
+    s3c2410_gpio_setpin(S3C2410_GPC(5), 1);
+    break;
+  case S3C2410_UDC_P_DISABLE :
+                printk(KERN_INFO "mini2440_udc_pullup disable\n");
+    s3c2410_gpio_setpin(S3C2410_GPC(5), 0);
+    break;
+  case S3C2410_UDC_P_RESET :
+                printk(KERN_INFO "mini2440_udc_pullup reset\n");
+    break;
+  default:
+    break;
+  }
+}
+
+static struct s3c2410_udc_mach_info mini2440_udc_cfg __initdata = {
+  .udc_command    = mini2440_udc_pullup,
+};
+
 /* DM9000AEP 10/100 ethernet controller */
 #define MACH_MINI2440_DM9K_BASE (S3C2410_CS4 + 0x300)
 
@@ -369,8 +398,13 @@
 #endif
   s3c_i2c0_set_platdata(NULL);
 
+  s3c24xx_udc_set_platdata(&mini2440_udc_cfg);
+
   s3c2410_gpio_cfgpin(S3C2410_GPC(0), S3C2410_GPC0_LEND);
 
+        /* pullup for usb device (gadget) port */
+  s3c2410_gpio_cfgpin(S3C2410_GPC(5), S3C2410_GPIO_OUTPUT);
+
   s3c_device_nand.dev.platform_data = &friendly_arm_nand_info;
   s3c_device_sdi.dev.platform_data = &mini2440_mmc_cfg;
   platform_add_devices(mini2440_devices, ARRAY_SIZE(mini2440_devices));

Hi Trevor,
I found there were a power problem but not enough knowledge to solve it.

I applied your patch, and now the serial port is mounted in a nand boot
process.

Thanks