Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 1 | ----------------------- |
| 2 | Ethernet Driver Guide |
| 3 | ----------------------- |
| 4 | |
| 5 | The networking stack in Das U-Boot is designed for multiple network devices |
| 6 | to be easily added and controlled at runtime. This guide is meant for people |
| 7 | who wish to review the net driver stack with an eye towards implementing your |
| 8 | own ethernet device driver. Here we will describe a new pseudo 'APE' driver. |
| 9 | |
| 10 | ------------------ |
| 11 | Driver Functions |
| 12 | ------------------ |
| 13 | |
| 14 | All functions you will be implementing in this document have the return value |
| 15 | meaning of 0 for success and non-zero for failure. |
| 16 | |
| 17 | ---------- |
| 18 | Register |
| 19 | ---------- |
| 20 | |
| 21 | When U-Boot initializes, it will call the common function eth_initialize(). |
| 22 | This will in turn call the board-specific board_eth_init() (or if that fails, |
| 23 | the cpu-specific cpu_eth_init()). These board-specific functions can do random |
| 24 | system handling, but ultimately they will call the driver-specific register |
| 25 | function which in turn takes care of initializing that particular instance. |
| 26 | |
| 27 | Keep in mind that you should code the driver to avoid storing state in global |
Ben Warren | 99dbd4e | 2009-10-05 00:02:51 -0700 | [diff] [blame] | 28 | data as someone might want to hook up two of the same devices to one board. |
| 29 | Any such information that is specific to an interface should be stored in a |
| 30 | private, driver-defined data structure and pointed to by eth->priv (see below). |
Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 31 | |
| 32 | So the call graph at this stage would look something like: |
| 33 | board_init() |
| 34 | eth_initialize() |
| 35 | board_eth_init() / cpu_eth_init() |
| 36 | driver_register() |
| 37 | initialize eth_device |
| 38 | eth_register() |
| 39 | |
| 40 | At this point in time, the only thing you need to worry about is the driver's |
| 41 | register function. The pseudo code would look something like: |
| 42 | int ape_register(bd_t *bis, int iobase) |
| 43 | { |
| 44 | struct ape_priv *priv; |
| 45 | struct eth_device *dev; |
| 46 | |
| 47 | priv = malloc(sizeof(*priv)); |
| 48 | if (priv == NULL) |
| 49 | return 1; |
| 50 | |
| 51 | dev = malloc(sizeof(*dev)); |
| 52 | if (dev == NULL) { |
| 53 | free(priv); |
| 54 | return 1; |
| 55 | } |
| 56 | |
| 57 | /* setup whatever private state you need */ |
| 58 | |
| 59 | memset(dev, 0, sizeof(*dev)); |
| 60 | sprintf(dev->name, "APE"); |
| 61 | |
| 62 | /* if your device has dedicated hardware storage for the |
| 63 | * MAC, read it and initialize dev->enetaddr with it |
| 64 | */ |
| 65 | ape_mac_read(dev->enetaddr); |
| 66 | |
| 67 | dev->iobase = iobase; |
| 68 | dev->priv = priv; |
| 69 | dev->init = ape_init; |
| 70 | dev->halt = ape_halt; |
| 71 | dev->send = ape_send; |
| 72 | dev->recv = ape_recv; |
Ben Warren | ecee932 | 2010-04-26 11:11:46 -0700 | [diff] [blame] | 73 | dev->write_hwaddr = ape_write_hwaddr; |
Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 74 | |
| 75 | eth_register(dev); |
| 76 | |
| 77 | #ifdef CONFIG_CMD_MII) |
| 78 | miiphy_register(dev->name, ape_mii_read, ape_mii_write); |
| 79 | #endif |
| 80 | |
Ben Warren | 99dbd4e | 2009-10-05 00:02:51 -0700 | [diff] [blame] | 81 | return 1; |
Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 82 | } |
| 83 | |
| 84 | The exact arguments needed to initialize your device are up to you. If you |
| 85 | need to pass more/less arguments, that's fine. You should also add the |
Ben Warren | 99dbd4e | 2009-10-05 00:02:51 -0700 | [diff] [blame] | 86 | prototype for your new register function to include/netdev.h. |
| 87 | |
| 88 | The return value for this function should be as follows: |
| 89 | < 0 - failure (hardware failure, not probe failure) |
| 90 | >=0 - number of interfaces detected |
| 91 | |
Wolfgang Denk | 4946775 | 2009-10-28 00:49:47 +0100 | [diff] [blame] | 92 | You might notice that many drivers seem to use xxx_initialize() rather than |
Ben Warren | 99dbd4e | 2009-10-05 00:02:51 -0700 | [diff] [blame] | 93 | xxx_register(). This is the old naming convention and should be avoided as it |
| 94 | causes confusion with the driver-specific init function. |
Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 95 | |
| 96 | Other than locating the MAC address in dedicated hardware storage, you should |
| 97 | not touch the hardware in anyway. That step is handled in the driver-specific |
| 98 | init function. Remember that we are only registering the device here, we are |
| 99 | not checking its state or doing random probing. |
| 100 | |
| 101 | ----------- |
| 102 | Callbacks |
| 103 | ----------- |
| 104 | |
| 105 | Now that we've registered with the ethernet layer, we can start getting some |
Ben Warren | ecee932 | 2010-04-26 11:11:46 -0700 | [diff] [blame] | 106 | real work done. You will need five functions: |
Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 107 | int ape_init(struct eth_device *dev, bd_t *bis); |
| 108 | int ape_send(struct eth_device *dev, volatile void *packet, int length); |
| 109 | int ape_recv(struct eth_device *dev); |
| 110 | int ape_halt(struct eth_device *dev); |
Ben Warren | ecee932 | 2010-04-26 11:11:46 -0700 | [diff] [blame] | 111 | int ape_write_hwaddr(struct eth_device *dev); |
Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 112 | |
| 113 | The init function checks the hardware (probing/identifying) and gets it ready |
| 114 | for send/recv operations. You often do things here such as resetting the MAC |
| 115 | and/or PHY, and waiting for the link to autonegotiate. You should also take |
| 116 | the opportunity to program the device's MAC address with the dev->enetaddr |
| 117 | member. This allows the rest of U-Boot to dynamically change the MAC address |
| 118 | and have the new settings be respected. |
| 119 | |
| 120 | The send function does what you think -- transmit the specified packet whose |
| 121 | size is specified by length (in bytes). You should not return until the |
| 122 | transmission is complete, and you should leave the state such that the send |
| 123 | function can be called multiple times in a row. |
| 124 | |
| 125 | The recv function should process packets as long as the hardware has them |
| 126 | readily available before returning. i.e. you should drain the hardware fifo. |
Mike Frysinger | e5c5d9e | 2009-11-01 22:39:56 -0500 | [diff] [blame] | 127 | For each packet you receive, you should call the NetReceive() function on it |
| 128 | along with the packet length. The common code sets up packet buffers for you |
| 129 | already in the .bss (NetRxPackets), so there should be no need to allocate your |
| 130 | own. This doesn't mean you must use the NetRxPackets array however; you're |
| 131 | free to call the NetReceive() function with any buffer you wish. So the pseudo |
| 132 | code here would look something like: |
Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 133 | int ape_recv(struct eth_device *dev) |
| 134 | { |
| 135 | int length, i = 0; |
| 136 | ... |
| 137 | while (packets_are_available()) { |
| 138 | ... |
| 139 | length = ape_get_packet(&NetRxPackets[i]); |
| 140 | ... |
| 141 | NetReceive(&NetRxPackets[i], length); |
| 142 | ... |
| 143 | if (++i >= PKTBUFSRX) |
| 144 | i = 0; |
| 145 | ... |
| 146 | } |
| 147 | ... |
| 148 | return 0; |
| 149 | } |
| 150 | |
| 151 | The halt function should turn off / disable the hardware and place it back in |
Mike Frysinger | e5c5d9e | 2009-11-01 22:39:56 -0500 | [diff] [blame] | 152 | its reset state. It can be called at any time (before any call to the related |
| 153 | init function), so make sure it can handle this sort of thing. |
Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 154 | |
Ben Warren | ecee932 | 2010-04-26 11:11:46 -0700 | [diff] [blame] | 155 | The write_hwaddr function should program the MAC address stored in dev->enetaddr |
| 156 | into the Ethernet controller. |
| 157 | |
Mike Frysinger | 1f1e774 | 2009-09-09 14:41:22 -0400 | [diff] [blame] | 158 | So the call graph at this stage would look something like: |
| 159 | some net operation (ping / tftp / whatever...) |
| 160 | eth_init() |
| 161 | dev->init() |
| 162 | eth_send() |
| 163 | dev->send() |
| 164 | eth_rx() |
| 165 | dev->recv() |
| 166 | eth_halt() |
| 167 | dev->halt() |
| 168 | |
| 169 | ----------------------------- |
| 170 | CONFIG_MII / CONFIG_CMD_MII |
| 171 | ----------------------------- |
| 172 | |
| 173 | If your device supports banging arbitrary values on the MII bus (pretty much |
| 174 | every device does), you should add support for the mii command. Doing so is |
| 175 | fairly trivial and makes debugging mii issues a lot easier at runtime. |
| 176 | |
| 177 | After you have called eth_register() in your driver's register function, add |
| 178 | a call to miiphy_register() like so: |
| 179 | #if defined(CONFIG_MII) || defined(CONFIG_CMD_MII) |
| 180 | miiphy_register(dev->name, mii_read, mii_write); |
| 181 | #endif |
| 182 | |
| 183 | And then define the mii_read and mii_write functions if you haven't already. |
| 184 | Their syntax is straightforward: |
| 185 | int mii_read(char *devname, uchar addr, uchar reg, ushort *val); |
| 186 | int mii_write(char *devname, uchar addr, uchar reg, ushort val); |
| 187 | |
| 188 | The read function should read the register 'reg' from the phy at address 'addr' |
| 189 | and store the result in the pointer 'val'. The implementation for the write |
| 190 | function should logically follow. |