rt-thread-official/bsp/tm4c129x/drivers/drv_eth.c

1520 lines
46 KiB
C
Raw Permalink Normal View History

/*
2021-03-21 22:30:55 +08:00
* Copyright (c) 2006-2021, RT-Thread Development Team
*
2021-03-21 22:30:55 +08:00
* SPDX-License-Identifier: Apache-2.0
*
* Change Logs:
* Date Author Notes
* 2014-07-25 ArdaFu Port to TM4C129X
*/
/**
* @file - tivaif.c
* lwIP Ethernet interface for Stellaris LM4F Devices
*
*/
/**
* Copyright (c) 2001-2004 Swedish Institute of Computer Science.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without modification,
* are permitted provided that the following conditions are met:
*
* 1. Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
* 3. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS FOR A PARTICui32AR PURPOSE ARE DISCLAIMED. IN NO EVENT
* SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT
* OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
* IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY
* OF SUCH DAMAGE.
*
* This file is part of the lwIP TCP/IP stack.
*
* Author: Adam Dunkels <adam@sics.se>
*
*/
/**
* Copyright (c) 2008-2012 Texas Instruments Incorporated
*
* This file is derived from the ``ethernetif.c'' skeleton Ethernet network
* interface driver for lwIP.
*
*/
#include "lwip/opt.h"
#include "lwip/def.h"
#include "lwip/mem.h"
#include "lwip/pbuf.h"
#include "lwip/sys.h"
#include <lwip/stats.h>
#include <lwip/snmp.h>
#include "lwip/tcpip.h"
#include "netif/etharp.h"
#include "netif/ppp_oe.h"
/**
* Sanity Check: This interface driver will NOT work if the following defines
* are incorrect.
*
*/
#if (PBUF_LINK_HLEN != 16)
#error "PBUF_LINK_HLEN must be 16 for this interface driver!"
#endif
#if (ETH_PAD_SIZE != 0)
#error "ETH_PAD_SIZE must be 0 for this interface driver!"
#endif
#if (!SYS_LIGHTWEIGHT_PROT)
#error "SYS_LIGHTWEIGHT_PROT must be enabled for this interface driver!"
#endif
/**
* Set the physical address of the PHY we will be using if this is not
* specified in lwipopts.h. We assume 0 for the internal PHY.
*/
#ifndef PHY_PHYS_ADDR
#define PHY_PHYS_ADDR 0
#endif
#if 1
#ifndef EMAC_PHY_CONFIG
#define EMAC_PHY_CONFIG (EMAC_PHY_TYPE_INTERNAL | EMAC_PHY_INT_MDIX_EN | \
EMAC_PHY_AN_100B_T_FULL_DUPLEX)
#endif
#endif
/**
* If necessary, set the defaui32t number of transmit and receive DMA descriptors
* used by the Ethernet MAC.
*
*/
#ifndef NUM_RX_DESCRIPTORS
#define NUM_RX_DESCRIPTORS 4
#endif
#ifndef NUM_TX_DESCRIPTORS
#define NUM_TX_DESCRIPTORS 8
#endif
/**
* Setup processing for PTP (IEEE-1588).
*
*/
#if LWIP_PTPD
extern uint32_t g_ui32SysClk;
extern uint32_t g_ui32PTPTickRate;
extern void lwIPHostGetTime(u32_t *time_s, u32_t *time_ns);
#endif
/**
* Stellaris DriverLib Header Files required for this interface driver.
*
*/
#include <stdint.h>
#include <stdbool.h>
#include "inc/hw_emac.h"
#include "inc/hw_ints.h"
#include "inc/hw_memmap.h"
#include "inc/hw_types.h"
#include "driverlib/emac.h"
#include "driverlib/interrupt.h"
#include "driverlib/sysctl.h"
#include "driverlib/flash.h"
#include "driverlib/interrupt.h"
#include "driverlib/pin_map.h"
#include "driverlib/rom_map.h"
#include "driverlib/gpio.h"
#include <netif/ethernetif.h>
#include "lwipopts.h"
#include "drv_eth.h"
/* Define those to better describe your network interface. */
#define IFNAME0 't'
#define IFNAME1 'i'
/**
* A structure used to keep track of driver state and error counts.
*/
typedef struct {
uint32_t ui32TXCount;
uint32_t ui32TXCopyCount;
uint32_t ui32TXCopyFailCount;
uint32_t ui32TXNoDescCount;
uint32_t ui32TXBufQueuedCount;
uint32_t ui32TXBufFreedCount;
uint32_t ui32RXBufReadCount;
uint32_t ui32RXPacketReadCount;
uint32_t ui32RXPacketErrCount;
uint32_t ui32RXPacketCBErrCount;
uint32_t ui32RXNoBufCount;
}
tDriverStats;
tDriverStats g_sDriverStats;
#ifdef DEBUG
/**
* Note: This rather weird construction where we invoke the macro with the
* name of the field minus its Hungarian prefix is a workaround for a problem
* experienced with GCC which does not like concatenating tokens after an
* operator, specifically '.' or '->', in a macro.
*/
#define DRIVER_STATS_INC(x) do{ g_sDriverStats.ui32##x++; } while(0)
#define DRIVER_STATS_DEC(x) do{ g_sDriverStats.ui32##x--; } while(0)
#define DRIVER_STATS_ADD(x, inc) do{ g_sDriverStats.ui32##x += (inc); } while(0)
#define DRIVER_STATS_SUB(x, dec) do{ g_sDriverStats.ui32##x -= (dec); } while(0)
#else
#define DRIVER_STATS_INC(x)
#define DRIVER_STATS_DEC(x)
#define DRIVER_STATS_ADD(x, inc)
#define DRIVER_STATS_SUB(x, dec)
#endif
/**
* Helper struct holding a DMA descriptor and the pbuf it currently refers
* to.
*/
typedef struct {
tEMACDMADescriptor Desc;
struct pbuf *pBuf;
} tDescriptor;
typedef struct {
tDescriptor *pDescriptors;
uint32_t ui32NumDescs;
uint32_t ui32Read;
uint32_t ui32Write;
} tDescriptorList;
/**
* Helper struct to hold private data used to operate your ethernet interface.
* Keeping the ethernet address of the MAC in this struct is not necessary
* as it is already kept in the struct netif.
* But this is only an example, anyway...
*/
typedef struct {
struct eth_addr *ethaddr;
/* Add whatever per-interface state that is needed here. */
tDescriptorList *pTxDescList;
tDescriptorList *pRxDescList;
} tStellarisIF;
/**
* Global variable for this interface's private data. Needed to allow
* the interrupt handlers access to this information outside of the
* context of the lwIP netif.
*
*/
tDescriptor g_pTxDescriptors[NUM_TX_DESCRIPTORS];
tDescriptor g_pRxDescriptors[NUM_RX_DESCRIPTORS];
tDescriptorList g_TxDescList = {
g_pTxDescriptors, NUM_TX_DESCRIPTORS, 0, 0
};
tDescriptorList g_RxDescList = {
g_pRxDescriptors, NUM_RX_DESCRIPTORS, 0, 0
};
static tStellarisIF g_StellarisIFData = {
0, &g_TxDescList, &g_RxDescList
};
/**
* Interrupt counters (for debug purposes).
*/
volatile uint32_t g_ui32NormalInts;
volatile uint32_t g_ui32AbnormalInts;
/**
* Status flag for EEE link established
*/
#if EEE_SUPPORT
volatile bool g_bEEELinkActive;
#endif
/**
* A macro which determines whether a pointer is within the SRAM address
* space and, hence, points to a buffer that the Ethernet MAC can directly
* DMA from.
*/
#define PTR_SAFE_FOR_EMAC_DMA(ptr) (((uint32_t)(ptr) >= 0x2000000) && \
((uint32_t)(ptr) < 0x20070000))
2021-03-21 22:30:55 +08:00
typedef struct
{
/* inherit from ethernet device */
struct eth_device parent;
tStellarisIF* dma_if;
/* for rx_thread async get pbuf */
rt_mailbox_t rx_pbuf_mb;
} net_device;
2021-03-21 22:30:55 +08:00
typedef net_device* net_device_t;
static char rx_pbuf_mb_pool[8*4];
static struct rt_mailbox eth_rx_pbuf_mb;
static net_device eth_dev_entry;
static net_device_t eth_dev = &eth_dev_entry;
/**
* Initialize the transmit and receive DMA descriptor lists.
*/
void
InitDMADescriptors(void)
{
uint32_t ui32Loop;
/* Transmit list - mark all descriptors as not owned by the hardware */
for(ui32Loop = 0; ui32Loop < NUM_TX_DESCRIPTORS; ui32Loop++)
{
g_pTxDescriptors[ui32Loop].pBuf = (struct pbuf *)0;
g_pTxDescriptors[ui32Loop].Desc.ui32Count = 0;
g_pTxDescriptors[ui32Loop].Desc.pvBuffer1 = 0;
g_pTxDescriptors[ui32Loop].Desc.DES3.pLink =
((ui32Loop == (NUM_TX_DESCRIPTORS - 1)) ?
&g_pTxDescriptors[0].Desc : &g_pTxDescriptors[ui32Loop + 1].Desc);
g_pTxDescriptors[ui32Loop].Desc.ui32CtrlStatus = DES0_TX_CTRL_INTERRUPT |
DES0_TX_CTRL_CHAINED | DES0_TX_CTRL_IP_ALL_CKHSUMS;
}
g_TxDescList.ui32Read = 0;
g_TxDescList.ui32Write = 0;
/* Receive list - tag each descriptor with a pbuf and set all fields to
* allow packets to be received.
*/
for(ui32Loop = 0; ui32Loop < NUM_RX_DESCRIPTORS; ui32Loop++)
{
g_pRxDescriptors[ui32Loop].pBuf = pbuf_alloc(PBUF_RAW, PBUF_POOL_BUFSIZE,
PBUF_POOL);
g_pRxDescriptors[ui32Loop].Desc.ui32Count = DES1_RX_CTRL_CHAINED;
if(g_pRxDescriptors[ui32Loop].pBuf)
{
/* Set the DMA to write directly into the pbuf payload. */
g_pRxDescriptors[ui32Loop].Desc.pvBuffer1 =
g_pRxDescriptors[ui32Loop].pBuf->payload;
g_pRxDescriptors[ui32Loop].Desc.ui32Count |=
(g_pRxDescriptors[ui32Loop].pBuf->len << DES1_RX_CTRL_BUFF1_SIZE_S);
g_pRxDescriptors[ui32Loop].Desc.ui32CtrlStatus = DES0_RX_CTRL_OWN;
}
else
{
LWIP_DEBUGF(NETIF_DEBUG, ("tivaif_init: pbuf_alloc error\n"));
/* No pbuf available so leave the buffer pointer empty. */
g_pRxDescriptors[ui32Loop].Desc.pvBuffer1 = 0;
g_pRxDescriptors[ui32Loop].Desc.ui32CtrlStatus = 0;
}
g_pRxDescriptors[ui32Loop].Desc.DES3.pLink =
((ui32Loop == (NUM_RX_DESCRIPTORS - 1)) ?
&g_pRxDescriptors[0].Desc : &g_pRxDescriptors[ui32Loop + 1].Desc);
}
g_TxDescList.ui32Read = 0;
g_TxDescList.ui32Write = 0;
//
// Set the descriptor pointers in the hardware.
//
EMACRxDMADescriptorListSet(EMAC0_BASE, &g_pRxDescriptors[0].Desc);
EMACTxDMADescriptorListSet(EMAC0_BASE, &g_pTxDescriptors[0].Desc);
}
/**
* In this function, the hardware should be initialized.
* Called from tivaif_init().
*
* @param netif the already initialized lwip network interface structure
* for this ethernetif
*/
static void
tivaif_hwinit(struct netif *psNetif)
{
uint16_t ui16Val;
/* clear the EEE Link Active flag */
#if EEE_SUPPORT
g_bEEELinkActive = false;
#endif
/* Set MAC hardware address length */
psNetif->hwaddr_len = ETHARP_HWADDR_LEN;
/* Set MAC hardware address */
EMACAddrGet(EMAC0_BASE, 0, &(psNetif->hwaddr[0]));
/* Maximum transfer unit */
psNetif->mtu = 1500;
/* Device capabilities */
psNetif->flags = NETIF_FLAG_BROADCAST | NETIF_FLAG_ETHARP | NETIF_FLAG_LINK_UP;
/* Initialize the DMA descriptors. */
InitDMADescriptors();
#if defined(EMAC_PHY_IS_EXT_MII) || defined(EMAC_PHY_IS_EXT_RMII)
/* If PHY is external then reset the PHY before configuring it */
EMACPHYWrite(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_BMCR,
EPHY_BMCR_MIIRESET);
while((EMACPHYRead(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_BMCR) &
EPHY_BMCR_MIIRESET) == EPHY_BMCR_MIIRESET);
#endif
/* Clear any stray PHY interrupts that may be set. */
ui16Val = EMACPHYRead(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_MISR1);
ui16Val = EMACPHYRead(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_MISR2);
/* Configure and enable the link status change interrupt in the PHY. */
ui16Val = EMACPHYRead(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_SCR);
ui16Val |= (EPHY_SCR_INTEN_EXT | EPHY_SCR_INTOE_EXT);
EMACPHYWrite(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_SCR, ui16Val);
EMACPHYWrite(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_MISR1, (EPHY_MISR1_LINKSTATEN |
EPHY_MISR1_SPEEDEN | EPHY_MISR1_DUPLEXMEN | EPHY_MISR1_ANCEN));
/* Read the PHY interrupt status to clear any stray events. */
ui16Val = EMACPHYRead(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_MISR1);
/**
* Set MAC filtering options. We receive all broadcast and mui32ticast
* packets along with those addressed specifically for us.
*/
EMACFrameFilterSet(EMAC0_BASE, (EMAC_FRMFILTER_HASH_AND_PERFECT |
EMAC_FRMFILTER_PASS_MULTICAST));
#if LWIP_PTPD
//
// Enable timestamping on all received packets.
//
// We set the fine clock adjustment mode and configure the subsecond
// increment to half the 25MHz PTPD clock. This will give us maximum control
// over the clock rate adjustment and keep the arithmetic easy later. It
// should be possible to synchronize with higher accuracy than this with
// appropriate juggling of the subsecond increment count and the addend
// register value, though.
//
EMACTimestampConfigSet(EMAC0_BASE, (EMAC_TS_ALL_RX_FRAMES |
EMAC_TS_DIGITAL_ROLLOVER |
EMAC_TS_PROCESS_IPV4_UDP | EMAC_TS_ALL |
EMAC_TS_PTP_VERSION_1 | EMAC_TS_UPDATE_FINE),
(1000000000 / (25000000 / 2)));
EMACTimestampAddendSet(EMAC0_BASE, 0x80000000);
EMACTimestampEnable(EMAC0_BASE);
#endif
/* Clear any pending MAC interrupts. */
EMACIntClear(EMAC0_BASE, EMACIntStatus(EMAC0_BASE, false));
/* Enable the Ethernet MAC transmitter and receiver. */
EMACTxEnable(EMAC0_BASE);
EMACRxEnable(EMAC0_BASE);
/* Enable the Ethernet RX and TX interrupt source. */
EMACIntEnable(EMAC0_BASE, (EMAC_INT_RECEIVE | EMAC_INT_TRANSMIT |
EMAC_INT_TX_STOPPED | EMAC_INT_RX_NO_BUFFER |
EMAC_INT_RX_STOPPED | EMAC_INT_PHY));
/* Enable the Ethernet interrupt. */
IntEnable(INT_EMAC0);
/* Enable all processor interrupts. */
IntMasterEnable();
/* Tell the PHY to start an auto-negotiation cycle. */
#if defined(EMAC_PHY_IS_EXT_MII) || defined(EMAC_PHY_IS_EXT_RMII)
EMACPHYWrite(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_BMCR, (EPHY_BMCR_SPEED |
EPHY_BMCR_DUPLEXM | EPHY_BMCR_ANEN | EPHY_BMCR_RESTARTAN));
#else
EMACPHYWrite(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_BMCR, (EPHY_BMCR_ANEN |
EPHY_BMCR_RESTARTAN));
#endif
}
#ifdef DEBUG
/**
* Dump the chain of pbuf pointers to the debug output.
*/
void
tivaif_trace_pbuf(const char *pcTitle, struct pbuf *p)
{
LWIP_DEBUGF(NETIF_DEBUG, ("%s %08x (%d, %d)", pcTitle, p, p->tot_len,
p->len));
do
{
p = p->next;
if(p)
{
LWIP_DEBUGF(NETIF_DEBUG, ("->%08x(%d)", p, p->len));
}
else
{
LWIP_DEBUGF(NETIF_DEBUG, ("->%08x", p));
}
} while((p != NULL) && (p->tot_len != p->len));
LWIP_DEBUGF(NETIF_DEBUG, ("\n"));
}
#endif
/**
* This function is used to check whether a passed pbuf contains only buffers
* resident in regions of memory that the Ethernet MAC can access. If any
* buffers in the chain are outside a directly-DMAable section of memory,
* the pbuf is copied to SRAM and a different pointer returned. If all
* buffers are safe, the pbuf reference count is incremented and the original
* pointer returned.
*/
static struct pbuf *
tivaif_check_pbuf(struct pbuf *p)
{
struct pbuf *pBuf;
rt_err_t Err;
pBuf = p;
#ifdef DEBUG
tivaif_trace_pbuf("Original:", p);
#endif
/* Walk the list of buffers in the pbuf checking each. */
do
{
/* Does this pbuf's payload reside in memory that the Ethernet DMA
* can access?
*/
if(!PTR_SAFE_FOR_EMAC_DMA(pBuf->payload))
{
/* This buffer is outside the DMA-able memory space so we need
* to copy the pbuf.
*/
pBuf = pbuf_alloc(PBUF_RAW, p->tot_len, PBUF_POOL);
/* If we got a new pbuf... */
if(pBuf)
{
/* ...copy the old pbuf into the new one. */
Err = pbuf_copy(pBuf, p);
/* If we failed to copy the pbuf, free the newly allocated one
* and make sure we return a NULL to show a problem.
*/
if(Err != RT_EOK)
{
DRIVER_STATS_INC(TXCopyFailCount);
pbuf_free(pBuf);
pBuf = NULL;
}
else
{
#ifdef DEBUG
tivaif_trace_pbuf("Copied:", pBuf);
#endif
DRIVER_STATS_INC(TXCopyCount);
}
}
/* Reduce the reference count on the original pbuf since we're not
* going to hold on to it after returning from tivaif_transmit.
* Note that we already bumped the reference count at the top of
* tivaif_transmit.
*/
pbuf_free(p);
/* Send back the new pbuf pointer or NULL if an error occurred. */
return(pBuf);
}
/* Move on to the next buffer in the queue */
pBuf = pBuf->next;
}
while(pBuf);
/**
* If we get here, the passed pbuf can be safely used without needing to
* be copied.
*/
return(p);
}
/**
* This function should do the actual transmission of the packet. The packet is
* contained in the pbuf that is passed to the function. This pbuf might be
* chained.
*
* @param psNetif the lwip network interface structure for this ethernetif
* @param p the MAC packet to send (e.g. IP packet including MAC addresses and type)
* @return RT_EOK if the packet coui32d be sent
* an err_t value if the packet coui32dn't be sent
*/
static rt_err_t
tivaif_transmit(net_device_t dev, struct pbuf *p)
{
tStellarisIF *pIF;
tDescriptor *pDesc;
struct pbuf *pBuf;
uint32_t ui32NumChained, ui32NumDescs;
bool bFirst;
SYS_ARCH_DECL_PROTECT(lev);
LWIP_DEBUGF(NETIF_DEBUG, ("tivaif_transmit 0x%08x, len %d\n", p,
p->tot_len));
/**
* This entire function must run within a "critical section" to preserve
* the integrity of the transmit pbuf queue.
*/
SYS_ARCH_PROTECT(lev);
/* Update our transmit attempt counter. */
DRIVER_STATS_INC(TXCount);
/**
* Increase the reference count on the packet provided so that we can
* hold on to it until we are finished transmitting its content.
*/
pbuf_ref(p);
/**
* Determine whether all buffers passed are within SRAM and, if not, copy
* the pbuf into SRAM-resident buffers so that the Ethernet DMA can access
* the data.
*/
p = tivaif_check_pbuf(p);
/* Make sure we still have a valid buffer (it may have been copied) */
if(!p)
{
LINK_STATS_INC(link.memerr);
SYS_ARCH_UNPROTECT(lev);
return(-RT_ENOMEM);
}
/* Get our state data from the netif structure we were passed. */
//pIF = (tStellarisIF *)psNetif->state;
pIF = dev->dma_if;
2021-03-21 22:30:55 +08:00
/* Make sure that the transmit descriptors are not all in use */
pDesc = &(pIF->pTxDescList->pDescriptors[pIF->pTxDescList->ui32Write]);
if(pDesc->pBuf)
{
/**
* The current write descriptor has a pbuf attached to it so this
* implies that the ring is full. Reject this transmit request with a
* memory error since we can't satisfy it just now.
*/
pbuf_free(p);
LINK_STATS_INC(link.memerr);
DRIVER_STATS_INC(TXNoDescCount);
SYS_ARCH_UNPROTECT(lev);
return (-RT_ENOMEM);
}
/* How many pbufs are in the chain passed? */
ui32NumChained = (uint32_t)pbuf_clen(p);
/* How many free transmit descriptors do we have? */
ui32NumDescs = (pIF->pTxDescList->ui32Read > pIF->pTxDescList->ui32Write) ?
(pIF->pTxDescList->ui32Read - pIF->pTxDescList->ui32Write) :
((NUM_TX_DESCRIPTORS - pIF->pTxDescList->ui32Write) +
pIF->pTxDescList->ui32Read);
/* Do we have enough free descriptors to send the whole packet? */
if(ui32NumDescs < ui32NumChained)
{
/* No - we can't transmit this whole packet so return an error. */
pbuf_free(p);
LINK_STATS_INC(link.memerr);
DRIVER_STATS_INC(TXNoDescCount);
SYS_ARCH_UNPROTECT(lev);
return (-RT_ENOMEM);
}
/* Tag the first descriptor as the start of the packet. */
bFirst = true;
pDesc->Desc.ui32CtrlStatus = DES0_TX_CTRL_FIRST_SEG;
/* Here, we know we can send the packet so write it to the descriptors */
pBuf = p;
while(ui32NumChained)
{
/* Get a pointer to the descriptor we will write next. */
pDesc = &(pIF->pTxDescList->pDescriptors[pIF->pTxDescList->ui32Write]);
/* Fill in the buffer pointer and length */
pDesc->Desc.ui32Count = (uint32_t)pBuf->len;
pDesc->Desc.pvBuffer1 = pBuf->payload;
/* Tag the first descriptor as the start of the packet. */
if(bFirst)
{
bFirst = false;
pDesc->Desc.ui32CtrlStatus = DES0_TX_CTRL_FIRST_SEG;
}
else
{
pDesc->Desc.ui32CtrlStatus = 0;
}
pDesc->Desc.ui32CtrlStatus |= (DES0_TX_CTRL_IP_ALL_CKHSUMS |
DES0_TX_CTRL_CHAINED);
/* Decrement our descriptor counter, move on to the next buffer in the
* pbuf chain. */
ui32NumChained--;
pBuf = pBuf->next;
/* Update the descriptor list write index. */
pIF->pTxDescList->ui32Write++;
if(pIF->pTxDescList->ui32Write == NUM_TX_DESCRIPTORS)
{
pIF->pTxDescList->ui32Write = 0;
}
/* If this is the last descriptor, mark it as the end of the packet. */
if(!ui32NumChained)
{
pDesc->Desc.ui32CtrlStatus |= (DES0_TX_CTRL_LAST_SEG |
DES0_TX_CTRL_INTERRUPT);
/* Tag the descriptor with the original pbuf pointer. */
pDesc->pBuf = p;
}
else
{
/* Set the lsb of the pbuf pointer. We use this as a signal that
* we should not free the pbuf when we are walking the descriptor
* list while processing the transmit interrupt. We only free the
* pbuf when processing the last descriptor used to transmit its
* chain.
*/
pDesc->pBuf = (struct pbuf *)((uint32_t)p + 1);
}
DRIVER_STATS_INC(TXBufQueuedCount);
/* Hand the descriptor over to the hardware. */
pDesc->Desc.ui32CtrlStatus |= DES0_TX_CTRL_OWN;
}
/* Tell the transmitter to start (in case it had stopped). */
EMACTxDMAPollDemand(EMAC0_BASE);
/* Update lwIP statistics */
LINK_STATS_INC(link.xmit);
SYS_ARCH_UNPROTECT(lev);
return(RT_EOK);
}
/**
* This function will process all transmit descriptors and free pbufs attached
* to any that have been transmitted since we last checked.
*
* This function is called only from the Ethernet interrupt handler.
*
* @param netif the lwip network interface structure for this ethernetif
* @return None.
*/
static void
tivaif_process_transmit(tStellarisIF *pIF)
{
tDescriptorList *pDescList;
uint32_t ui32NumDescs;
/* Get a pointer to the transmit descriptor list. */
pDescList = pIF->pTxDescList;
/* Walk the list until we have checked all descriptors or we reach the
* write pointer or find a descriptor that the hardware is still working
* on.
*/
for(ui32NumDescs = 0; ui32NumDescs < pDescList->ui32NumDescs; ui32NumDescs++)
{
/* Has the buffer attached to this descriptor been transmitted? */
if(pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32CtrlStatus &
DES0_TX_CTRL_OWN)
{
/* No - we're finished. */
break;
}
/* Does this descriptor have a buffer attached to it? */
if(pDescList->pDescriptors[pDescList->ui32Read].pBuf)
{
/* Yes - free it if it's not marked as an intermediate pbuf */
if(!((uint32_t)(pDescList->pDescriptors[pDescList->ui32Read].pBuf) & 1))
{
pbuf_free(pDescList->pDescriptors[pDescList->ui32Read].pBuf);
DRIVER_STATS_INC(TXBufFreedCount);
}
pDescList->pDescriptors[pDescList->ui32Read].pBuf = NULL;
}
else
{
/* If the descriptor has no buffer, we are finished. */
break;
}
/* Move on to the next descriptor. */
pDescList->ui32Read++;
if(pDescList->ui32Read == pDescList->ui32NumDescs)
{
pDescList->ui32Read = 0;
}
}
}
/**
* This function will process all receive descriptors that contain newly read
* data and pass complete frames up the lwIP stack as they are found. The
* timestamp of the packet will be placed into the pbuf structure if PTPD is
* enabled.
*
* This function is called only from the Ethernet interrupt handler.
*
* @param psNetif the lwip network interface structure for this ethernetif
* @return None.
*/
static void
tivaif_receive(net_device_t dev)
{
tDescriptorList *pDescList;
tStellarisIF *pIF;
static struct pbuf *pBuf = NULL;
uint32_t ui32DescEnd;
/* Get a pointer to our state data */
pIF = dev->dma_if;
/* Get a pointer to the receive descriptor list. */
pDescList = pIF->pRxDescList;
/* Start with a NULL pbuf so that we don't try to link chain the first
* time round.
*/
//pBuf = NULL;
/* Determine where we start and end our walk of the descriptor list */
ui32DescEnd = pDescList->ui32Read ? (pDescList->ui32Read - 1) : (pDescList->ui32NumDescs - 1);
/* Step through the descriptors that are marked for CPU attention. */
while(pDescList->ui32Read != ui32DescEnd)
{
/* Does the current descriptor have a buffer attached to it? */
if(pDescList->pDescriptors[pDescList->ui32Read].pBuf)
{
/* Yes - determine if the host has filled it yet. */
if(pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32CtrlStatus &
DES0_RX_CTRL_OWN)
{
/* The DMA engine still owns the descriptor so we are finished */
break;
}
DRIVER_STATS_INC(RXBufReadCount);
/* If this descriptor contains the end of the packet, fix up the
* buffer size accordingly.
*/
if(pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32CtrlStatus &
DES0_RX_STAT_LAST_DESC)
{
/* This is the last descriptor for the frame so fix up the
* length. It is safe for us to modify the internal fields
* directly here (rather than calling pbuf_realloc) since we
* know each of these pbufs is never chained.
*/
pDescList->pDescriptors[pDescList->ui32Read].pBuf->len =
(pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32CtrlStatus &
DES0_RX_STAT_FRAME_LENGTH_M) >>
DES0_RX_STAT_FRAME_LENGTH_S;
pDescList->pDescriptors[pDescList->ui32Read].pBuf->tot_len =
pDescList->pDescriptors[pDescList->ui32Read].pBuf->len;
}
if(pBuf)
{
/* Link this pbuf to the last one we looked at since this buffer
* is a continuation of an existing frame (split across multiple
* pbufs). Note that we use pbuf_cat() here rather than
* pbuf_chain() since we don't want to increase the reference
* count of either pbuf - we only want to link them together.
*/
pbuf_cat(pBuf, pDescList->pDescriptors[pDescList->ui32Read].pBuf);
pDescList->pDescriptors[pDescList->ui32Read].pBuf = pBuf;
}
/* Remember the buffer associated with this descriptor. */
pBuf = pDescList->pDescriptors[pDescList->ui32Read].pBuf;
/* Is this the last descriptor for the current frame? */
if(pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32CtrlStatus &
DES0_RX_STAT_LAST_DESC)
{
/* Yes - does the frame contain errors? */
if(pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32CtrlStatus &
DES0_RX_STAT_ERR)
{
/* This is a bad frame so discard it and update the relevant
* statistics.
*/
LWIP_DEBUGF(NETIF_DEBUG, ("tivaif_receive: packet error\n"));
pbuf_free(pBuf);
LINK_STATS_INC(link.drop);
DRIVER_STATS_INC(RXPacketErrCount);
pBuf = NULL;
}
else
{
/* This is a good frame so pass it up the stack. */
LINK_STATS_INC(link.recv);
DRIVER_STATS_INC(RXPacketReadCount);
#if LWIP_PTPD
/* Place the timestamp in the PBUF if PTPD is enabled */
pBuf->time_s =
pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32IEEE1588TimeHi;
pBuf->time_ns =
pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32IEEE1588TimeLo;
#endif
#if NO_SYS
if(ethernet_input(pBuf, psNetif) != RT_EOK)
#else
//if(tcpip_input(pBuf, psNetif) != ERR_OK)
if(rt_mb_send(dev->rx_pbuf_mb, (rt_uint32_t)pBuf) != RT_EOK)
#endif
{
/* drop the packet */
LWIP_DEBUGF(NETIF_DEBUG, ("tivaif_input: input error\n"));
pbuf_free(pBuf);
/* Adjust the link statistics */
LINK_STATS_INC(link.memerr);
LINK_STATS_INC(link.drop);
DRIVER_STATS_INC(RXPacketCBErrCount);
}
/* We're finished with this packet so make sure we don't try
* to link the next buffer to it.
*/
pBuf = NULL;
}
}
}
/* Allocate a new buffer for this descriptor */
pDescList->pDescriptors[pDescList->ui32Read].pBuf = pbuf_alloc(PBUF_RAW,
PBUF_POOL_BUFSIZE,
PBUF_POOL);
pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32Count =
DES1_RX_CTRL_CHAINED;
if(pDescList->pDescriptors[pDescList->ui32Read].pBuf)
{
/* We got a buffer so fill in the payload pointer and size. */
pDescList->pDescriptors[pDescList->ui32Read].Desc.pvBuffer1 =
pDescList->pDescriptors[pDescList->ui32Read].pBuf->payload;
pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32Count |=
(pDescList->pDescriptors[pDescList->ui32Read].pBuf->len <<
DES1_RX_CTRL_BUFF1_SIZE_S);
/* Give this descriptor back to the hardware */
pDescList->pDescriptors[pDescList->ui32Read].Desc.ui32CtrlStatus =
DES0_RX_CTRL_OWN;
}
else
{
LWIP_DEBUGF(NETIF_DEBUG, ("tivaif_receive: pbuf_alloc error\n"));
pDescList->pDescriptors[pDescList->ui32Read].Desc.pvBuffer1 = 0;
/* Update the stats to show we coui32dn't allocate a pbuf. */
DRIVER_STATS_INC(RXNoBufCount);
LINK_STATS_INC(link.memerr);
/* Stop parsing here since we can't leave a broken descriptor in
* the chain.
*/
break;
}
/* Move on to the next descriptor in the chain, taking care to wrap. */
pDescList->ui32Read++;
if(pDescList->ui32Read == pDescList->ui32NumDescs)
{
pDescList->ui32Read = 0;
}
}
}
/**
* Process interrupts from the PHY.
*
* should be called from the Stellaris Ethernet Interrupt Handler. This
* function will read packets from the Stellaris Ethernet fifo and place them
* into a pbuf queue. If the transmitter is idle and there is at least one packet
* on the transmit queue, it will place it in the transmit fifo and start the
* transmitter.
*
*/
void
tivaif_process_phy_interrupt(net_device_t dev)
{
uint16_t ui16Val, ui16Status;
#if EEE_SUPPORT
uint16_t ui16EEEStatus;
#endif
uint32_t ui32Config, ui32Mode, ui32RxMaxFrameSize;
/* Read the PHY interrupt status. This clears all interrupt sources.
* Note that we are only enabling sources in EPHY_MISR1 so we don't
* read EPHY_MISR2.
*/
ui16Val = EMACPHYRead(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_MISR1);
/* Read the current PHY status. */
ui16Status = EMACPHYRead(EMAC0_BASE, PHY_PHYS_ADDR, EPHY_STS);
/* If EEE mode support is requested then read the value of the Link
* partners status
*/
#if EEE_SUPPORT
ui16EEEStatus = EMACPHYMMDRead(EMAC0_BASE, PHY_PHYS_ADDR, 0x703D);
#endif
/* Has the link status changed? */
if(ui16Val & EPHY_MISR1_LINKSTAT)
{
/* Is link up or down now? */
if(ui16Status & EPHY_STS_LINK)
{
/* Tell lwIP the link is up. */
#if NO_SYS
netif_set_link_up(psNetif);
#else
//tcpip_callback((tcpip_callback_fn)netif_set_link_up, psNetif);
eth_device_linkchange(&(dev->parent), RT_TRUE);
#endif
/* if the link has been advertised as EEE capable then configure
* the MAC register for LPI timers and manually set the PHY link
* status bit
*/
#if EEE_SUPPORT
if(ui16EEEStatus & 0x2)
{
EMACLPIConfig(EMAC0_BASE, true, 1000, 36);
EMACLPILinkSet(EMAC0_BASE);
g_bEEELinkActive = true;
}
#endif
/* In this case we drop through since we may need to reconfigure
* the MAC depending upon the speed and half/fui32l-duplex settings.
*/
}
else
{
/* Tell lwIP the link is down */
#if NO_SYS
netif_set_link_down(psNetif);
#else
//tcpip_callback((tcpip_callback_fn)netif_set_link_down, psNetif);
eth_device_linkchange(&(dev->parent), RT_FALSE);
#endif
/* if the link has been advertised as EEE capable then clear the
* MAC register LPI timers and manually clear the PHY link status
* bit
*/
#if EEE_SUPPORT
g_bEEELinkActive = false;
EMACLPILinkClear(EMAC0_BASE);
EMACLPIConfig(EMAC0_BASE, false, 1000, 0);
#endif
}
}
/* Has the speed or duplex status changed? */
if(ui16Val & (EPHY_MISR1_SPEED | EPHY_MISR1_SPEED | EPHY_MISR1_ANC))
{
/* Get the current MAC configuration. */
EMACConfigGet(EMAC0_BASE, &ui32Config, &ui32Mode,
&ui32RxMaxFrameSize);
/* What speed is the interface running at now?
*/
if(ui16Status & EPHY_STS_SPEED)
{
/* 10Mbps is selected */
ui32Config &= ~EMAC_CONFIG_100MBPS;
}
else
{
/* 100Mbps is selected */
ui32Config |= EMAC_CONFIG_100MBPS;
}
/* Are we in fui32l- or half-duplex mode? */
if(ui16Status & EPHY_STS_DUPLEX)
{
/* Fui32l duplex. */
ui32Config |= EMAC_CONFIG_FULL_DUPLEX;
}
else
{
/* Half duplex. */
ui32Config &= ~EMAC_CONFIG_FULL_DUPLEX;
}
/* Reconfigure the MAC */
EMACConfigSet(EMAC0_BASE, ui32Config, ui32Mode, ui32RxMaxFrameSize);
}
}
/**
* Process tx and rx packets at the low-level interrupt.
*
* should be called from the Stellaris Ethernet Interrupt Handler. This
* function will read packets from the Stellaris Ethernet fifo and place them
* into a pbuf queue. If the transmitter is idle and there is at least one packet
* on the transmit queue, it will place it in the transmit fifo and start the
* transmitter.
*
*/
void
tivaif_interrupt(net_device_t dev, uint32_t ui32Status)
{
/* Update our debug interrupt counters. */
if(ui32Status & EMAC_INT_NORMAL_INT)
{
g_ui32NormalInts++;
}
if(ui32Status & EMAC_INT_ABNORMAL_INT)
{
g_ui32AbnormalInts++;
}
/* Is this an interrupt from the PHY? */
if(ui32Status & EMAC_INT_PHY)
{
tivaif_process_phy_interrupt(dev);
}
/* Process the transmit DMA list, freeing any buffers that have been
* transmitted since our last interrupt.
*/
if(ui32Status & EMAC_INT_TRANSMIT)
{
#if EEE_SUPPORT
if(g_bEEELinkActive)
{
EMACLPIEnter(EMAC0_BASE);
}
#endif
tivaif_process_transmit(dev->dma_if);
}
/**
* Process the receive DMA list and pass all successfully received packets
* up the stack. We also call this function in cases where the receiver has
* stalled due to missing buffers since the receive function will attempt to
* allocate new pbufs for descriptor entries which have none.
*/
if(ui32Status & (EMAC_INT_RECEIVE | EMAC_INT_RX_NO_BUFFER |
EMAC_INT_RX_STOPPED))
{
tivaif_receive(dev);
}
}
#if NETIF_DEBUG
/* Print an IP header by using LWIP_DEBUGF
* @param p an IP packet, p->payload pointing to the IP header
*/
void
tivaif_debug_print(struct pbuf *p)
{
struct eth_hdr *ethhdr = (struct eth_hdr *)p->payload;
u16_t *plen = (u16_t *)p->payload;
LWIP_DEBUGF(NETIF_DEBUG, ("ETH header:\n"));
LWIP_DEBUGF(NETIF_DEBUG, ("Packet Length:%5"U16_F" \n",*plen));
LWIP_DEBUGF(NETIF_DEBUG, ("Destination: %02"X8_F"-%02"X8_F"-%02"X8_F"-%02"X8_F"-%02"X8_F"-%02"X8_F"\n",
ethhdr->dest.addr[0],
ethhdr->dest.addr[1],
ethhdr->dest.addr[2],
ethhdr->dest.addr[3],
ethhdr->dest.addr[4],
ethhdr->dest.addr[5]));
LWIP_DEBUGF(NETIF_DEBUG, ("Source: %02"X8_F"-%02"X8_F"-%02"X8_F"-%02"X8_F"-%02"X8_F"-%02"X8_F"\n",
ethhdr->src.addr[0],
ethhdr->src.addr[1],
ethhdr->src.addr[2],
ethhdr->src.addr[3],
ethhdr->src.addr[4],
ethhdr->src.addr[5]));
LWIP_DEBUGF(NETIF_DEBUG, ("Packet Type:0x%04"U16_F" \n", ethhdr->type));
}
#endif /* NETIF_DEBUG */
void lwIPEthernetIntHandler(void)
{
uint32_t ui32Status;
#ifdef DEF_INT_TEMPSTAMP
uint32_t ui32TimerStatus;
#endif
//
// Read and Clear the interrupt.
//
ui32Status = MAP_EMACIntStatus(EMAC0_BASE, true);
//
// If the interrupt really came from the Ethernet and not our
// timer, clear it.
//
if(ui32Status)
{
MAP_EMACIntClear(EMAC0_BASE, ui32Status);
}
#ifdef DEF_INT_TEMPSTAMP
//
// Check to see whether a hardware timer interrupt has been reported.
//
if(ui32Status & EMAC_INT_TIMESTAMP)
{
//
// Yes - read and clear the timestamp interrupt status.
//
ui32TimerStatus = EMACTimestampIntStatus(EMAC0_BASE);
//
// If a timer interrupt handler has been registered, call it.
//
if(g_pfnTimerHandler)
{
g_pfnTimerHandler(EMAC0_BASE, ui32TimerStatus);
}
}
#endif
//
// The handling of the interrupt is different based on the use of a RTOS.
//
//
// No RTOS is being used. If a transmit/receive interrupt was active,
// run the low-level interrupt handler.
//
if(ui32Status)
{
tivaif_interrupt(eth_dev, ui32Status);
}
//
// Service the lwIP timers.
//
//lwIPServiceTimers();
}
2021-03-21 22:30:55 +08:00
// OUI:00-12-37 (hex) Texas Instruments, only for test
static int tiva_eth_mac_addr_init(void)
{
2021-03-21 22:30:55 +08:00
int retVal =0;
uint32_t ulUser[2];
uint8_t mac_addr[6];
2021-03-21 22:30:55 +08:00
MAP_FlashUserGet(&ulUser[0], &ulUser[1]);
if((ulUser[0] == 0xffffffff) || (ulUser[1] == 0xffffffff))
{
rt_kprintf("Fail to get mac address from eeprom.\n");
rt_kprintf("Using default mac address\n");
2021-03-21 22:30:55 +08:00
// OUI:00-12-37 (hex) Texas Instruments, only for test
// Configure the hardware MAC address
ulUser[0] = 0x00371200;
ulUser[1] = 0x00563412;
2021-03-21 22:30:55 +08:00
//FlashUserSet(ulUser0, ulUser1);
retVal =-1;
}
2021-03-21 22:30:55 +08:00
//Convert the 24/24 split MAC address from NV ram into a 32/16 split MAC
//address needed to program the hardware registers, then program the MAC
//address into the Ethernet Controller registers.
2021-03-21 22:30:55 +08:00
mac_addr[0] = ((ulUser[0] >> 0) & 0xff);
mac_addr[1] = ((ulUser[0] >> 8) & 0xff);
mac_addr[2] = ((ulUser[0] >> 16) & 0xff);
mac_addr[3] = ((ulUser[1] >> 0) & 0xff);
mac_addr[4] = ((ulUser[1] >> 8) & 0xff);
mac_addr[5] = ((ulUser[1] >> 16) & 0xff);
2021-03-21 22:30:55 +08:00
//
// Program the hardware with its MAC address (for filtering).
//
MAP_EMACAddrSet(EMAC0_BASE, 0, mac_addr);
return retVal;
}
void tiva_eth_lowlevel_init(void)
{
MAP_SysCtlPeripheralEnable(SYSCTL_PERIPH_GPIOF);
//
// PF1/PK4/PK6 are used for Ethernet LEDs.
//
MAP_GPIOPinConfigure(GPIO_PF0_EN0LED0);
MAP_GPIOPinConfigure(GPIO_PF4_EN0LED1);
GPIOPinTypeEthernetLED(GPIO_PORTF_BASE, GPIO_PIN_0);
GPIOPinTypeEthernetLED(GPIO_PORTF_BASE, GPIO_PIN_4);
2021-03-21 22:30:55 +08:00
//
// Enable the ethernet peripheral.
//
MAP_SysCtlPeripheralEnable(SYSCTL_PERIPH_EMAC0);
MAP_SysCtlPeripheralReset(SYSCTL_PERIPH_EMAC0);
//
// Enable the internal PHY if it's present and we're being
// asked to use it.
//
if((EMAC_PHY_CONFIG & EMAC_PHY_TYPE_MASK) == EMAC_PHY_TYPE_INTERNAL)
{
//
// We've been asked to configure for use with the internal
// PHY. Is it present?
//
if(SysCtlPeripheralPresent(SYSCTL_PERIPH_EPHY0))
{
//
// Yes - enable and reset it.
//
MAP_SysCtlPeripheralEnable(SYSCTL_PERIPH_EPHY0);
MAP_SysCtlPeripheralReset(SYSCTL_PERIPH_EPHY0);
}
else
{
//
// Internal PHY is not present on this part so hang here.
//
rt_kprintf("Internal PHY is not present on this part.\n");
while(1)
{
}
}
}
//
// Wait for the MAC to come out of reset.
//
while(!MAP_SysCtlPeripheralReady(SYSCTL_PERIPH_EMAC0))
{
}
//
// Configure for use with whichever PHY the user requires.
//
MAP_EMACPHYConfigSet(EMAC0_BASE, EMAC_PHY_CONFIG);
//
// Initialize the MAC and set the DMA mode.
//
MAP_EMACInit(EMAC0_BASE, 120000000, //system clock = 120MHz
EMAC_BCONFIG_MIXED_BURST | EMAC_BCONFIG_PRIORITY_FIXED,
4, 4, 0);
//
// Set MAC configuration options.
//
MAP_EMACConfigSet(EMAC0_BASE, (EMAC_CONFIG_FULL_DUPLEX |
EMAC_CONFIG_CHECKSUM_OFFLOAD |
EMAC_CONFIG_7BYTE_PREAMBLE |
EMAC_CONFIG_IF_GAP_96BITS |
EMAC_CONFIG_USE_MACADDR0 |
EMAC_CONFIG_SA_FROM_DESCRIPTOR |
EMAC_CONFIG_BO_LIMIT_1024),
(EMAC_MODE_RX_STORE_FORWARD |
EMAC_MODE_TX_STORE_FORWARD |
EMAC_MODE_TX_THRESHOLD_64_BYTES |
EMAC_MODE_RX_THRESHOLD_64_BYTES), 0);
2021-03-21 22:30:55 +08:00
EMACIntRegister(EMAC0_BASE, lwIPEthernetIntHandler);
}
static rt_err_t eth_dev_init(rt_device_t device)
{
net_device_t net_dev = (net_device_t)device;
struct netif *psNetif = (net_dev->parent.netif);
2021-03-21 22:30:55 +08:00
LWIP_ASSERT("psNetif != NULL", (psNetif != NULL));
#if LWIP_NETIF_HOSTNAME
/* Initialize interface hostname */
psNetif->hostname = "t4mc";
#endif /* LWIP_NETIF_HOSTNAME */
/*
* Initialize the snmp variables and counters inside the struct netif.
* The last argument should be replaced with your link speed, in units
* of bits per second.
*/
//NETIF_INIT_SNMP(psNetif, snmp_ifType_ethernet_csmacd, 1000000);
2021-03-21 22:30:55 +08:00
net_dev->dma_if = &g_StellarisIFData;
/* Remember our MAC address. */
g_StellarisIFData.ethaddr = (struct eth_addr *)&(psNetif->hwaddr[0]);
/* Initialize the hardware */
tivaif_hwinit(psNetif);
return RT_EOK;
}
/* control the interface */
static rt_err_t eth_dev_control(rt_device_t dev, int cmd, void *args)
{
switch(cmd)
{
case NIOCTL_GADDR:
/* get mac address */
2021-03-21 22:30:55 +08:00
if(args)
MAP_EMACAddrGet(EMAC0_BASE, 0, (uint8_t*)args);
2021-03-21 22:30:55 +08:00
else
return -RT_ERROR;
break;
default :
break;
}
return RT_EOK;
}
/* Open the interface */
static rt_err_t eth_dev_open(rt_device_t dev, rt_uint16_t oflag)
{
return RT_EOK;
}
/* Close the interface */
static rt_err_t eth_dev_close(rt_device_t dev)
{
return RT_EOK;
}
/* Read */
static rt_ssize_t eth_dev_read(rt_device_t dev, rt_off_t pos, void* buffer, rt_size_t size)
{
rt_set_errno(-RT_ENOSYS);
return 0;
}
/* Write */
static rt_ssize_t eth_dev_write(rt_device_t dev, rt_off_t pos, const void* buffer, rt_size_t size)
{
rt_set_errno(-RT_ENOSYS);
return 0;
}
static rt_err_t eth_dev_tx(rt_device_t dev, struct pbuf *p)
{
return tivaif_transmit((net_device_t)dev, p);
}
static struct pbuf* eth_dev_rx(rt_device_t dev)
{
rt_err_t result;
rt_uint32_t temp =0;
net_device_t net_dev = (net_device_t)dev;
result = rt_mb_recv(net_dev->rx_pbuf_mb, (rt_ubase_t *)&temp, RT_WAITING_NO);
2021-03-21 22:30:55 +08:00
return (result == RT_EOK)? (struct pbuf*)temp : RT_NULL;
}
int rt_hw_tiva_eth_init(void)
{
rt_err_t result;
/* Clock GPIO and etc */
2021-03-21 22:30:55 +08:00
tiva_eth_lowlevel_init();
tiva_eth_mac_addr_init();
/* init rt-thread device interface */
eth_dev->parent.parent.init = eth_dev_init;
eth_dev->parent.parent.open = eth_dev_open;
eth_dev->parent.parent.close = eth_dev_close;
eth_dev->parent.parent.read = eth_dev_read;
eth_dev->parent.parent.write = eth_dev_write;
eth_dev->parent.parent.control = eth_dev_control;
eth_dev->parent.eth_rx = eth_dev_rx;
eth_dev->parent.eth_tx = eth_dev_tx;
2021-03-21 22:30:55 +08:00
result = rt_mb_init(&eth_rx_pbuf_mb, "epbuf",
&rx_pbuf_mb_pool[0], sizeof(rx_pbuf_mb_pool)/4,
RT_IPC_FLAG_FIFO);
RT_ASSERT(result == RT_EOK);
eth_dev->rx_pbuf_mb = &eth_rx_pbuf_mb;
2021-03-21 22:30:55 +08:00
result = eth_device_init(&(eth_dev->parent), "e0");
return result;
}
// eth_device_init using malloc
// We use INIT_COMPONENT_EXPORT insted of INIT_BOARD_EXPORT
INIT_COMPONENT_EXPORT(rt_hw_tiva_eth_init);
#if 0
#ifdef RT_USING_FINSH
#include "finsh.h"
void PHY_Read(uint8_t addr)
{
uint16_t data = EMACPHYRead(EMAC0_BASE, PHY_PHYS_ADDR, addr);
rt_kprintf("R PHY_REG[0x%02X] = 0x%04X\n", addr, data);
}
FINSH_FUNCTION_EXPORT(PHY_Read, (add));
void PHY_Write(uint8_t addr , uint16_t data)
{
EMACPHYWrite(EMAC0_BASE, PHY_PHYS_ADDR, addr, data);
rt_kprintf("W PHY_REG[0x%02X] = 0x%04X\n", addr, data);
}
FINSH_FUNCTION_EXPORT(PHY_Write, (add, data));
2021-03-21 22:30:55 +08:00
void PHY_SetAdd(uint8_t addr0, uint8_t addr1, uint8_t addr2,
uint8_t addr3, uint8_t addr4, uint8_t addr5)
{
uint32_t ulUser[2];
ulUser[0] = (((addr2<<8)|addr1)<<8)|addr0;
ulUser[1] = (((addr5<<8)|addr4)<<8)|addr3;
2021-03-21 22:30:55 +08:00
MAP_FlashUserSet(ulUser[0], ulUser[1]);
MAP_FlashUserSave();
rt_kprintf("Save to EEPROM. please reboot.");
}
FINSH_FUNCTION_EXPORT(PHY_SetAdd, (add0-add5));
#endif //RT_USING_FINSH
#endif