linuxdebug/drivers/net/fddi/skfp/hwmtm.c

2171 lines
55 KiB
C

// SPDX-License-Identifier: GPL-2.0-or-later
/******************************************************************************
*
* (C)Copyright 1998,1999 SysKonnect,
* a business unit of Schneider & Koch & Co. Datensysteme GmbH.
*
* See the file "skfddi.c" for further information.
*
* The information in this file is provided "AS IS" without warranty.
*
******************************************************************************/
#define HWMTM
#ifndef FDDI
#define FDDI
#endif
#include "h/types.h"
#include "h/fddi.h"
#include "h/smc.h"
#include "h/supern_2.h"
#include "h/skfbiinc.h"
/*
-------------------------------------------------------------
DOCUMENTATION
-------------------------------------------------------------
BEGIN_MANUAL_ENTRY(DOCUMENTATION)
T B D
END_MANUAL_ENTRY
*/
/*
-------------------------------------------------------------
LOCAL VARIABLES:
-------------------------------------------------------------
*/
#ifdef COMMON_MB_POOL
static SMbuf *mb_start;
static SMbuf *mb_free;
static int mb_init = FALSE ;
static int call_count;
#endif
/*
-------------------------------------------------------------
EXTERNE VARIABLES:
-------------------------------------------------------------
*/
#ifdef DEBUG
#ifndef DEBUG_BRD
extern struct smt_debug debug ;
#endif
#endif
#ifdef NDIS_OS2
extern u_char offDepth ;
extern u_char force_irq_pending ;
#endif
/*
-------------------------------------------------------------
LOCAL FUNCTIONS:
-------------------------------------------------------------
*/
static void queue_llc_rx(struct s_smc *smc, SMbuf *mb);
static void smt_to_llc(struct s_smc *smc, SMbuf *mb);
static void init_txd_ring(struct s_smc *smc);
static void init_rxd_ring(struct s_smc *smc);
static void queue_txd_mb(struct s_smc *smc, SMbuf *mb);
static u_long init_descr_ring(struct s_smc *smc, union s_fp_descr volatile *start,
int count);
static u_long repair_txd_ring(struct s_smc *smc, struct s_smt_tx_queue *queue);
static u_long repair_rxd_ring(struct s_smc *smc, struct s_smt_rx_queue *queue);
static SMbuf* get_llc_rx(struct s_smc *smc);
static SMbuf* get_txd_mb(struct s_smc *smc);
static void mac_drv_clear_txd(struct s_smc *smc);
/*
-------------------------------------------------------------
EXTERNAL FUNCTIONS:
-------------------------------------------------------------
*/
/* The external SMT functions are listed in cmtdef.h */
extern void* mac_drv_get_space(struct s_smc *smc, unsigned int size);
extern void* mac_drv_get_desc_mem(struct s_smc *smc, unsigned int size);
extern void mac_drv_fill_rxd(struct s_smc *smc);
extern void mac_drv_tx_complete(struct s_smc *smc,
volatile struct s_smt_fp_txd *txd);
extern void mac_drv_rx_complete(struct s_smc *smc,
volatile struct s_smt_fp_rxd *rxd,
int frag_count, int len);
extern void mac_drv_requeue_rxd(struct s_smc *smc,
volatile struct s_smt_fp_rxd *rxd,
int frag_count);
extern void mac_drv_clear_rxd(struct s_smc *smc,
volatile struct s_smt_fp_rxd *rxd, int frag_count);
#ifdef USE_OS_CPY
extern void hwm_cpy_rxd2mb(void);
extern void hwm_cpy_txd2mb(void);
#endif
#ifdef ALL_RX_COMPLETE
extern void mac_drv_all_receives_complete(void);
#endif
extern u_long mac_drv_virt2phys(struct s_smc *smc, void *virt);
extern u_long dma_master(struct s_smc *smc, void *virt, int len, int flag);
#ifdef NDIS_OS2
extern void post_proc(void);
#else
extern void dma_complete(struct s_smc *smc, volatile union s_fp_descr *descr,
int flag);
#endif
extern int mac_drv_rx_init(struct s_smc *smc, int len, int fc, char *look_ahead,
int la_len);
/*
-------------------------------------------------------------
PUBLIC FUNCTIONS:
-------------------------------------------------------------
*/
void process_receive(struct s_smc *smc);
void fddi_isr(struct s_smc *smc);
void smt_free_mbuf(struct s_smc *smc, SMbuf *mb);
void init_driver_fplus(struct s_smc *smc);
void mac_drv_rx_mode(struct s_smc *smc, int mode);
void init_fddi_driver(struct s_smc *smc, u_char *mac_addr);
void mac_drv_clear_tx_queue(struct s_smc *smc);
void mac_drv_clear_rx_queue(struct s_smc *smc);
void hwm_tx_frag(struct s_smc *smc, char far *virt, u_long phys, int len,
int frame_status);
void hwm_rx_frag(struct s_smc *smc, char far *virt, u_long phys, int len,
int frame_status);
int mac_drv_init(struct s_smc *smc);
int hwm_tx_init(struct s_smc *smc, u_char fc, int frag_count, int frame_len,
int frame_status);
u_int mac_drv_check_space(void);
SMbuf* smt_get_mbuf(struct s_smc *smc);
#ifdef DEBUG
void mac_drv_debug_lev(struct s_smc *smc, int flag, int lev);
#endif
/*
-------------------------------------------------------------
MACROS:
-------------------------------------------------------------
*/
#ifndef UNUSED
#ifdef lint
#define UNUSED(x) (x) = (x)
#else
#define UNUSED(x)
#endif
#endif
#ifdef USE_CAN_ADDR
#define MA smc->hw.fddi_canon_addr.a
#define GROUP_ADDR_BIT 0x01
#else
#define MA smc->hw.fddi_home_addr.a
#define GROUP_ADDR_BIT 0x80
#endif
#define RXD_TXD_COUNT (HWM_ASYNC_TXD_COUNT+HWM_SYNC_TXD_COUNT+\
SMT_R1_RXD_COUNT+SMT_R2_RXD_COUNT)
#ifdef MB_OUTSIDE_SMC
#define EXT_VIRT_MEM ((RXD_TXD_COUNT+1)*sizeof(struct s_smt_fp_txd) +\
MAX_MBUF*sizeof(SMbuf))
#define EXT_VIRT_MEM_2 ((RXD_TXD_COUNT+1)*sizeof(struct s_smt_fp_txd))
#else
#define EXT_VIRT_MEM ((RXD_TXD_COUNT+1)*sizeof(struct s_smt_fp_txd))
#endif
/*
* define critical read for 16 Bit drivers
*/
#if defined(NDIS_OS2) || defined(ODI2)
#define CR_READ(var) ((var) & 0xffff0000 | ((var) & 0xffff))
#else
#define CR_READ(var) (__le32)(var)
#endif
#define IMASK_SLOW (IS_PLINT1 | IS_PLINT2 | IS_TIMINT | IS_TOKEN | \
IS_MINTR1 | IS_MINTR2 | IS_MINTR3 | IS_R1_P | \
IS_R1_C | IS_XA_C | IS_XS_C)
/*
-------------------------------------------------------------
INIT- AND SMT FUNCTIONS:
-------------------------------------------------------------
*/
/*
* BEGIN_MANUAL_ENTRY(mac_drv_check_space)
* u_int mac_drv_check_space()
*
* function DOWNCALL (drvsr.c)
* This function calculates the needed non virtual
* memory for MBufs, RxD and TxD descriptors etc.
* needed by the driver.
*
* return u_int memory in bytes
*
* END_MANUAL_ENTRY
*/
u_int mac_drv_check_space(void)
{
#ifdef MB_OUTSIDE_SMC
#ifdef COMMON_MB_POOL
call_count++ ;
if (call_count == 1) {
return EXT_VIRT_MEM;
}
else {
return EXT_VIRT_MEM_2;
}
#else
return EXT_VIRT_MEM;
#endif
#else
return 0;
#endif
}
/*
* BEGIN_MANUAL_ENTRY(mac_drv_init)
* void mac_drv_init(smc)
*
* function DOWNCALL (drvsr.c)
* In this function the hardware module allocates it's
* memory.
* The operating system dependent module should call
* mac_drv_init once, after the adatper is detected.
* END_MANUAL_ENTRY
*/
int mac_drv_init(struct s_smc *smc)
{
if (sizeof(struct s_smt_fp_rxd) % 16) {
SMT_PANIC(smc,HWM_E0001,HWM_E0001_MSG) ;
}
if (sizeof(struct s_smt_fp_txd) % 16) {
SMT_PANIC(smc,HWM_E0002,HWM_E0002_MSG) ;
}
/*
* get the required memory for the RxDs and TxDs
*/
if (!(smc->os.hwm.descr_p = (union s_fp_descr volatile *)
mac_drv_get_desc_mem(smc,(u_int)
(RXD_TXD_COUNT+1)*sizeof(struct s_smt_fp_txd)))) {
return 1; /* no space the hwm modul can't work */
}
/*
* get the memory for the SMT MBufs
*/
#ifndef MB_OUTSIDE_SMC
smc->os.hwm.mbuf_pool.mb_start=(SMbuf *)(&smc->os.hwm.mbuf_pool.mb[0]) ;
#else
#ifndef COMMON_MB_POOL
if (!(smc->os.hwm.mbuf_pool.mb_start = (SMbuf *) mac_drv_get_space(smc,
MAX_MBUF*sizeof(SMbuf)))) {
return 1; /* no space the hwm modul can't work */
}
#else
if (!mb_start) {
if (!(mb_start = (SMbuf *) mac_drv_get_space(smc,
MAX_MBUF*sizeof(SMbuf)))) {
return 1; /* no space the hwm modul can't work */
}
}
#endif
#endif
return 0;
}
/*
* BEGIN_MANUAL_ENTRY(init_driver_fplus)
* init_driver_fplus(smc)
*
* Sets hardware modul specific values for the mode register 2
* (e.g. the byte alignment for the received frames, the position of the
* least significant byte etc.)
* END_MANUAL_ENTRY
*/
void init_driver_fplus(struct s_smc *smc)
{
smc->hw.fp.mdr2init = FM_LSB | FM_BMMODE | FM_ENNPRQ | FM_ENHSRQ | 3 ;
#ifdef PCI
smc->hw.fp.mdr2init |= FM_CHKPAR | FM_PARITY ;
#endif
smc->hw.fp.mdr3init = FM_MENRQAUNLCK | FM_MENRS ;
#ifdef USE_CAN_ADDR
/* enable address bit swapping */
smc->hw.fp.frselreg_init = FM_ENXMTADSWAP | FM_ENRCVADSWAP ;
#endif
}
static u_long init_descr_ring(struct s_smc *smc,
union s_fp_descr volatile *start,
int count)
{
int i ;
union s_fp_descr volatile *d1 ;
union s_fp_descr volatile *d2 ;
u_long phys ;
DB_GEN(3, "descr ring starts at = %p", start);
for (i=count-1, d1=start; i ; i--) {
d2 = d1 ;
d1++ ; /* descr is owned by the host */
d2->r.rxd_rbctrl = cpu_to_le32(BMU_CHECK) ;
d2->r.rxd_next = &d1->r ;
phys = mac_drv_virt2phys(smc,(void *)d1) ;
d2->r.rxd_nrdadr = cpu_to_le32(phys) ;
}
DB_GEN(3, "descr ring ends at = %p", d1);
d1->r.rxd_rbctrl = cpu_to_le32(BMU_CHECK) ;
d1->r.rxd_next = &start->r ;
phys = mac_drv_virt2phys(smc,(void *)start) ;
d1->r.rxd_nrdadr = cpu_to_le32(phys) ;
for (i=count, d1=start; i ; i--) {
DRV_BUF_FLUSH(&d1->r,DDI_DMA_SYNC_FORDEV) ;
d1++;
}
return phys;
}
static void init_txd_ring(struct s_smc *smc)
{
struct s_smt_fp_txd volatile *ds ;
struct s_smt_tx_queue *queue ;
u_long phys ;
/*
* initialize the transmit descriptors
*/
ds = (struct s_smt_fp_txd volatile *) ((char *)smc->os.hwm.descr_p +
SMT_R1_RXD_COUNT*sizeof(struct s_smt_fp_rxd)) ;
queue = smc->hw.fp.tx[QUEUE_A0] ;
DB_GEN(3, "Init async TxD ring, %d TxDs", HWM_ASYNC_TXD_COUNT);
(void)init_descr_ring(smc,(union s_fp_descr volatile *)ds,
HWM_ASYNC_TXD_COUNT) ;
phys = le32_to_cpu(ds->txd_ntdadr) ;
ds++ ;
queue->tx_curr_put = queue->tx_curr_get = ds ;
ds-- ;
queue->tx_free = HWM_ASYNC_TXD_COUNT ;
queue->tx_used = 0 ;
outpd(ADDR(B5_XA_DA),phys) ;
ds = (struct s_smt_fp_txd volatile *) ((char *)ds +
HWM_ASYNC_TXD_COUNT*sizeof(struct s_smt_fp_txd)) ;
queue = smc->hw.fp.tx[QUEUE_S] ;
DB_GEN(3, "Init sync TxD ring, %d TxDs", HWM_SYNC_TXD_COUNT);
(void)init_descr_ring(smc,(union s_fp_descr volatile *)ds,
HWM_SYNC_TXD_COUNT) ;
phys = le32_to_cpu(ds->txd_ntdadr) ;
ds++ ;
queue->tx_curr_put = queue->tx_curr_get = ds ;
queue->tx_free = HWM_SYNC_TXD_COUNT ;
queue->tx_used = 0 ;
outpd(ADDR(B5_XS_DA),phys) ;
}
static void init_rxd_ring(struct s_smc *smc)
{
struct s_smt_fp_rxd volatile *ds ;
struct s_smt_rx_queue *queue ;
u_long phys ;
/*
* initialize the receive descriptors
*/
ds = (struct s_smt_fp_rxd volatile *) smc->os.hwm.descr_p ;
queue = smc->hw.fp.rx[QUEUE_R1] ;
DB_GEN(3, "Init RxD ring, %d RxDs", SMT_R1_RXD_COUNT);
(void)init_descr_ring(smc,(union s_fp_descr volatile *)ds,
SMT_R1_RXD_COUNT) ;
phys = le32_to_cpu(ds->rxd_nrdadr) ;
ds++ ;
queue->rx_curr_put = queue->rx_curr_get = ds ;
queue->rx_free = SMT_R1_RXD_COUNT ;
queue->rx_used = 0 ;
outpd(ADDR(B4_R1_DA),phys) ;
}
/*
* BEGIN_MANUAL_ENTRY(init_fddi_driver)
* void init_fddi_driver(smc,mac_addr)
*
* initializes the driver and it's variables
*
* END_MANUAL_ENTRY
*/
void init_fddi_driver(struct s_smc *smc, u_char *mac_addr)
{
SMbuf *mb ;
int i ;
init_board(smc,mac_addr) ;
(void)init_fplus(smc) ;
/*
* initialize the SMbufs for the SMT
*/
#ifndef COMMON_MB_POOL
mb = smc->os.hwm.mbuf_pool.mb_start ;
smc->os.hwm.mbuf_pool.mb_free = (SMbuf *)NULL ;
for (i = 0; i < MAX_MBUF; i++) {
mb->sm_use_count = 1 ;
smt_free_mbuf(smc,mb) ;
mb++ ;
}
#else
mb = mb_start ;
if (!mb_init) {
mb_free = 0 ;
for (i = 0; i < MAX_MBUF; i++) {
mb->sm_use_count = 1 ;
smt_free_mbuf(smc,mb) ;
mb++ ;
}
mb_init = TRUE ;
}
#endif
/*
* initialize the other variables
*/
smc->os.hwm.llc_rx_pipe = smc->os.hwm.llc_rx_tail = (SMbuf *)NULL ;
smc->os.hwm.txd_tx_pipe = smc->os.hwm.txd_tx_tail = NULL ;
smc->os.hwm.pass_SMT = smc->os.hwm.pass_NSA = smc->os.hwm.pass_DB = 0 ;
smc->os.hwm.pass_llc_promisc = TRUE ;
smc->os.hwm.queued_rx_frames = smc->os.hwm.queued_txd_mb = 0 ;
smc->os.hwm.detec_count = 0 ;
smc->os.hwm.rx_break = 0 ;
smc->os.hwm.rx_len_error = 0 ;
smc->os.hwm.isr_flag = FALSE ;
/*
* make sure that the start pointer is 16 byte aligned
*/
i = 16 - ((long)smc->os.hwm.descr_p & 0xf) ;
if (i != 16) {
DB_GEN(3, "i = %d", i);
smc->os.hwm.descr_p = (union s_fp_descr volatile *)
((char *)smc->os.hwm.descr_p+i) ;
}
DB_GEN(3, "pt to descr area = %p", smc->os.hwm.descr_p);
init_txd_ring(smc) ;
init_rxd_ring(smc) ;
mac_drv_fill_rxd(smc) ;
init_plc(smc) ;
}
SMbuf *smt_get_mbuf(struct s_smc *smc)
{
register SMbuf *mb ;
#ifndef COMMON_MB_POOL
mb = smc->os.hwm.mbuf_pool.mb_free ;
#else
mb = mb_free ;
#endif
if (mb) {
#ifndef COMMON_MB_POOL
smc->os.hwm.mbuf_pool.mb_free = mb->sm_next ;
#else
mb_free = mb->sm_next ;
#endif
mb->sm_off = 8 ;
mb->sm_use_count = 1 ;
}
DB_GEN(3, "get SMbuf: mb = %p", mb);
return mb; /* May be NULL */
}
void smt_free_mbuf(struct s_smc *smc, SMbuf *mb)
{
if (mb) {
mb->sm_use_count-- ;
DB_GEN(3, "free_mbuf: sm_use_count = %d", mb->sm_use_count);
/*
* If the use_count is != zero the MBuf is queued
* more than once and must not queued into the
* free MBuf queue
*/
if (!mb->sm_use_count) {
DB_GEN(3, "free SMbuf: mb = %p", mb);
#ifndef COMMON_MB_POOL
mb->sm_next = smc->os.hwm.mbuf_pool.mb_free ;
smc->os.hwm.mbuf_pool.mb_free = mb ;
#else
mb->sm_next = mb_free ;
mb_free = mb ;
#endif
}
}
else
SMT_PANIC(smc,HWM_E0003,HWM_E0003_MSG) ;
}
/*
* BEGIN_MANUAL_ENTRY(mac_drv_repair_descr)
* void mac_drv_repair_descr(smc)
*
* function called from SMT (HWM / hwmtm.c)
* The BMU is idle when this function is called.
* Mac_drv_repair_descr sets up the physical address
* for all receive and transmit queues where the BMU
* should continue.
* It may be that the BMU was reseted during a fragmented
* transfer. In this case there are some fragments which will
* never completed by the BMU. The OWN bit of this fragments
* must be switched to be owned by the host.
*
* Give a start command to the receive BMU.
* Start the transmit BMUs if transmit frames pending.
*
* END_MANUAL_ENTRY
*/
void mac_drv_repair_descr(struct s_smc *smc)
{
u_long phys ;
if (smc->hw.hw_state != STOPPED) {
SK_BREAK() ;
SMT_PANIC(smc,HWM_E0013,HWM_E0013_MSG) ;
return ;
}
/*
* repair tx queues: don't start
*/
phys = repair_txd_ring(smc,smc->hw.fp.tx[QUEUE_A0]) ;
outpd(ADDR(B5_XA_DA),phys) ;
if (smc->hw.fp.tx_q[QUEUE_A0].tx_used) {
outpd(ADDR(B0_XA_CSR),CSR_START) ;
}
phys = repair_txd_ring(smc,smc->hw.fp.tx[QUEUE_S]) ;
outpd(ADDR(B5_XS_DA),phys) ;
if (smc->hw.fp.tx_q[QUEUE_S].tx_used) {
outpd(ADDR(B0_XS_CSR),CSR_START) ;
}
/*
* repair rx queues
*/
phys = repair_rxd_ring(smc,smc->hw.fp.rx[QUEUE_R1]) ;
outpd(ADDR(B4_R1_DA),phys) ;
outpd(ADDR(B0_R1_CSR),CSR_START) ;
}
static u_long repair_txd_ring(struct s_smc *smc, struct s_smt_tx_queue *queue)
{
int i ;
int tx_used ;
u_long phys ;
u_long tbctrl ;
struct s_smt_fp_txd volatile *t ;
SK_UNUSED(smc) ;
t = queue->tx_curr_get ;
tx_used = queue->tx_used ;
for (i = tx_used+queue->tx_free-1 ; i ; i-- ) {
t = t->txd_next ;
}
phys = le32_to_cpu(t->txd_ntdadr) ;
t = queue->tx_curr_get ;
while (tx_used) {
DRV_BUF_FLUSH(t,DDI_DMA_SYNC_FORCPU) ;
tbctrl = le32_to_cpu(t->txd_tbctrl) ;
if (tbctrl & BMU_OWN) {
if (tbctrl & BMU_STF) {
break ; /* exit the loop */
}
else {
/*
* repair the descriptor
*/
t->txd_tbctrl &= ~cpu_to_le32(BMU_OWN) ;
}
}
phys = le32_to_cpu(t->txd_ntdadr) ;
DRV_BUF_FLUSH(t,DDI_DMA_SYNC_FORDEV) ;
t = t->txd_next ;
tx_used-- ;
}
return phys;
}
/*
* Repairs the receive descriptor ring and returns the physical address
* where the BMU should continue working.
*
* o The physical address where the BMU was stopped has to be
* determined. This is the next RxD after rx_curr_get with an OWN
* bit set.
* o The BMU should start working at beginning of the next frame.
* RxDs with an OWN bit set but with a reset STF bit should be
* skipped and owned by the driver (OWN = 0).
*/
static u_long repair_rxd_ring(struct s_smc *smc, struct s_smt_rx_queue *queue)
{
int i ;
int rx_used ;
u_long phys ;
u_long rbctrl ;
struct s_smt_fp_rxd volatile *r ;
SK_UNUSED(smc) ;
r = queue->rx_curr_get ;
rx_used = queue->rx_used ;
for (i = SMT_R1_RXD_COUNT-1 ; i ; i-- ) {
r = r->rxd_next ;
}
phys = le32_to_cpu(r->rxd_nrdadr) ;
r = queue->rx_curr_get ;
while (rx_used) {
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORCPU) ;
rbctrl = le32_to_cpu(r->rxd_rbctrl) ;
if (rbctrl & BMU_OWN) {
if (rbctrl & BMU_STF) {
break ; /* exit the loop */
}
else {
/*
* repair the descriptor
*/
r->rxd_rbctrl &= ~cpu_to_le32(BMU_OWN) ;
}
}
phys = le32_to_cpu(r->rxd_nrdadr) ;
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORDEV) ;
r = r->rxd_next ;
rx_used-- ;
}
return phys;
}
/*
-------------------------------------------------------------
INTERRUPT SERVICE ROUTINE:
-------------------------------------------------------------
*/
/*
* BEGIN_MANUAL_ENTRY(fddi_isr)
* void fddi_isr(smc)
*
* function DOWNCALL (drvsr.c)
* interrupt service routine, handles the interrupt requests
* generated by the FDDI adapter.
*
* NOTE: The operating system dependent module must guarantee that the
* interrupts of the adapter are disabled when it calls fddi_isr.
*
* About the USE_BREAK_ISR mechanismn:
*
* The main requirement of this mechanismn is to force an timer IRQ when
* leaving process_receive() with leave_isr set. process_receive() may
* be called at any time from anywhere!
* To be sure we don't miss such event we set 'force_irq' per default.
* We have to force and Timer IRQ if 'smc->os.hwm.leave_isr' AND
* 'force_irq' are set. 'force_irq' may be reset if a receive complete
* IRQ is pending.
*
* END_MANUAL_ENTRY
*/
void fddi_isr(struct s_smc *smc)
{
u_long is ; /* ISR source */
u_short stu, stl ;
SMbuf *mb ;
#ifdef USE_BREAK_ISR
int force_irq ;
#endif
#ifdef ODI2
if (smc->os.hwm.rx_break) {
mac_drv_fill_rxd(smc) ;
if (smc->hw.fp.rx_q[QUEUE_R1].rx_used > 0) {
smc->os.hwm.rx_break = 0 ;
process_receive(smc) ;
}
else {
smc->os.hwm.detec_count = 0 ;
smt_force_irq(smc) ;
}
}
#endif
smc->os.hwm.isr_flag = TRUE ;
#ifdef USE_BREAK_ISR
force_irq = TRUE ;
if (smc->os.hwm.leave_isr) {
smc->os.hwm.leave_isr = FALSE ;
process_receive(smc) ;
}
#endif
while ((is = GET_ISR() & ISR_MASK)) {
NDD_TRACE("CH0B",is,0,0) ;
DB_GEN(7, "ISA = 0x%lx", is);
if (is & IMASK_SLOW) {
NDD_TRACE("CH1b",is,0,0) ;
if (is & IS_PLINT1) { /* PLC1 */
plc1_irq(smc) ;
}
if (is & IS_PLINT2) { /* PLC2 */
plc2_irq(smc) ;
}
if (is & IS_MINTR1) { /* FORMAC+ STU1(U/L) */
stu = inpw(FM_A(FM_ST1U)) ;
stl = inpw(FM_A(FM_ST1L)) ;
DB_GEN(6, "Slow transmit complete");
mac1_irq(smc,stu,stl) ;
}
if (is & IS_MINTR2) { /* FORMAC+ STU2(U/L) */
stu= inpw(FM_A(FM_ST2U)) ;
stl= inpw(FM_A(FM_ST2L)) ;
DB_GEN(6, "Slow receive complete");
DB_GEN(7, "stl = %x : stu = %x", stl, stu);
mac2_irq(smc,stu,stl) ;
}
if (is & IS_MINTR3) { /* FORMAC+ STU3(U/L) */
stu= inpw(FM_A(FM_ST3U)) ;
stl= inpw(FM_A(FM_ST3L)) ;
DB_GEN(6, "FORMAC Mode Register 3");
mac3_irq(smc,stu,stl) ;
}
if (is & IS_TIMINT) { /* Timer 82C54-2 */
timer_irq(smc) ;
#ifdef NDIS_OS2
force_irq_pending = 0 ;
#endif
/*
* out of RxD detection
*/
if (++smc->os.hwm.detec_count > 4) {
/*
* check out of RxD condition
*/
process_receive(smc) ;
}
}
if (is & IS_TOKEN) { /* Restricted Token Monitor */
rtm_irq(smc) ;
}
if (is & IS_R1_P) { /* Parity error rx queue 1 */
/* clear IRQ */
outpd(ADDR(B4_R1_CSR),CSR_IRQ_CL_P) ;
SMT_PANIC(smc,HWM_E0004,HWM_E0004_MSG) ;
}
if (is & IS_R1_C) { /* Encoding error rx queue 1 */
/* clear IRQ */
outpd(ADDR(B4_R1_CSR),CSR_IRQ_CL_C) ;
SMT_PANIC(smc,HWM_E0005,HWM_E0005_MSG) ;
}
if (is & IS_XA_C) { /* Encoding error async tx q */
/* clear IRQ */
outpd(ADDR(B5_XA_CSR),CSR_IRQ_CL_C) ;
SMT_PANIC(smc,HWM_E0006,HWM_E0006_MSG) ;
}
if (is & IS_XS_C) { /* Encoding error sync tx q */
/* clear IRQ */
outpd(ADDR(B5_XS_CSR),CSR_IRQ_CL_C) ;
SMT_PANIC(smc,HWM_E0007,HWM_E0007_MSG) ;
}
}
/*
* Fast Tx complete Async/Sync Queue (BMU service)
*/
if (is & (IS_XS_F|IS_XA_F)) {
DB_GEN(6, "Fast tx complete queue");
/*
* clear IRQ, Note: no IRQ is lost, because
* we always service both queues
*/
outpd(ADDR(B5_XS_CSR),CSR_IRQ_CL_F) ;
outpd(ADDR(B5_XA_CSR),CSR_IRQ_CL_F) ;
mac_drv_clear_txd(smc) ;
llc_restart_tx(smc) ;
}
/*
* Fast Rx Complete (BMU service)
*/
if (is & IS_R1_F) {
DB_GEN(6, "Fast receive complete");
/* clear IRQ */
#ifndef USE_BREAK_ISR
outpd(ADDR(B4_R1_CSR),CSR_IRQ_CL_F) ;
process_receive(smc) ;
#else
process_receive(smc) ;
if (smc->os.hwm.leave_isr) {
force_irq = FALSE ;
} else {
outpd(ADDR(B4_R1_CSR),CSR_IRQ_CL_F) ;
process_receive(smc) ;
}
#endif
}
#ifndef NDIS_OS2
while ((mb = get_llc_rx(smc))) {
smt_to_llc(smc,mb) ;
}
#else
if (offDepth)
post_proc() ;
while (!offDepth && (mb = get_llc_rx(smc))) {
smt_to_llc(smc,mb) ;
}
if (!offDepth && smc->os.hwm.rx_break) {
process_receive(smc) ;
}
#endif
if (smc->q.ev_get != smc->q.ev_put) {
NDD_TRACE("CH2a",0,0,0) ;
ev_dispatcher(smc) ;
}
#ifdef NDIS_OS2
post_proc() ;
if (offDepth) { /* leave fddi_isr because */
break ; /* indications not allowed */
}
#endif
#ifdef USE_BREAK_ISR
if (smc->os.hwm.leave_isr) {
break ; /* leave fddi_isr */
}
#endif
/* NOTE: when the isr is left, no rx is pending */
} /* end of interrupt source polling loop */
#ifdef USE_BREAK_ISR
if (smc->os.hwm.leave_isr && force_irq) {
smt_force_irq(smc) ;
}
#endif
smc->os.hwm.isr_flag = FALSE ;
NDD_TRACE("CH0E",0,0,0) ;
}
/*
-------------------------------------------------------------
RECEIVE FUNCTIONS:
-------------------------------------------------------------
*/
#ifndef NDIS_OS2
/*
* BEGIN_MANUAL_ENTRY(mac_drv_rx_mode)
* void mac_drv_rx_mode(smc,mode)
*
* function DOWNCALL (fplus.c)
* Corresponding to the parameter mode, the operating system
* dependent module can activate several receive modes.
*
* para mode = 1: RX_ENABLE_ALLMULTI enable all multicasts
* = 2: RX_DISABLE_ALLMULTI disable "enable all multicasts"
* = 3: RX_ENABLE_PROMISC enable promiscuous
* = 4: RX_DISABLE_PROMISC disable promiscuous
* = 5: RX_ENABLE_NSA enable rec. of all NSA frames
* (disabled after 'driver reset' & 'set station address')
* = 6: RX_DISABLE_NSA disable rec. of all NSA frames
*
* = 21: RX_ENABLE_PASS_SMT ( see description )
* = 22: RX_DISABLE_PASS_SMT ( " " )
* = 23: RX_ENABLE_PASS_NSA ( " " )
* = 24: RX_DISABLE_PASS_NSA ( " " )
* = 25: RX_ENABLE_PASS_DB ( " " )
* = 26: RX_DISABLE_PASS_DB ( " " )
* = 27: RX_DISABLE_PASS_ALL ( " " )
* = 28: RX_DISABLE_LLC_PROMISC ( " " )
* = 29: RX_ENABLE_LLC_PROMISC ( " " )
*
*
* RX_ENABLE_PASS_SMT / RX_DISABLE_PASS_SMT
*
* If the operating system dependent module activates the
* mode RX_ENABLE_PASS_SMT, the hardware module
* duplicates all SMT frames with the frame control
* FC_SMT_INFO and passes them to the LLC receive channel
* by calling mac_drv_rx_init.
* The SMT Frames which are sent by the local SMT and the NSA
* frames whose A- and C-Indicator is not set are also duplicated
* and passed.
* The receive mode RX_DISABLE_PASS_SMT disables the passing
* of SMT frames.
*
* RX_ENABLE_PASS_NSA / RX_DISABLE_PASS_NSA
*
* If the operating system dependent module activates the
* mode RX_ENABLE_PASS_NSA, the hardware module
* duplicates all NSA frames with frame control FC_SMT_NSA
* and a set A-Indicator and passed them to the LLC
* receive channel by calling mac_drv_rx_init.
* All NSA Frames which are sent by the local SMT
* are also duplicated and passed.
* The receive mode RX_DISABLE_PASS_NSA disables the passing
* of NSA frames with the A- or C-Indicator set.
*
* NOTE: For fear that the hardware module receives NSA frames with
* a reset A-Indicator, the operating system dependent module
* has to call mac_drv_rx_mode with the mode RX_ENABLE_NSA
* before activate the RX_ENABLE_PASS_NSA mode and after every
* 'driver reset' and 'set station address'.
*
* RX_ENABLE_PASS_DB / RX_DISABLE_PASS_DB
*
* If the operating system dependent module activates the
* mode RX_ENABLE_PASS_DB, direct BEACON frames
* (FC_BEACON frame control) are passed to the LLC receive
* channel by mac_drv_rx_init.
* The receive mode RX_DISABLE_PASS_DB disables the passing
* of direct BEACON frames.
*
* RX_DISABLE_PASS_ALL
*
* Disables all special receives modes. It is equal to
* call mac_drv_set_rx_mode successively with the
* parameters RX_DISABLE_NSA, RX_DISABLE_PASS_SMT,
* RX_DISABLE_PASS_NSA and RX_DISABLE_PASS_DB.
*
* RX_ENABLE_LLC_PROMISC
*
* (default) all received LLC frames and all SMT/NSA/DBEACON
* frames depending on the attitude of the flags
* PASS_SMT/PASS_NSA/PASS_DBEACON will be delivered to the
* LLC layer
*
* RX_DISABLE_LLC_PROMISC
*
* all received SMT/NSA/DBEACON frames depending on the
* attitude of the flags PASS_SMT/PASS_NSA/PASS_DBEACON
* will be delivered to the LLC layer.
* all received LLC frames with a directed address, Multicast
* or Broadcast address will be delivered to the LLC
* layer too.
*
* END_MANUAL_ENTRY
*/
void mac_drv_rx_mode(struct s_smc *smc, int mode)
{
switch(mode) {
case RX_ENABLE_PASS_SMT:
smc->os.hwm.pass_SMT = TRUE ;
break ;
case RX_DISABLE_PASS_SMT:
smc->os.hwm.pass_SMT = FALSE ;
break ;
case RX_ENABLE_PASS_NSA:
smc->os.hwm.pass_NSA = TRUE ;
break ;
case RX_DISABLE_PASS_NSA:
smc->os.hwm.pass_NSA = FALSE ;
break ;
case RX_ENABLE_PASS_DB:
smc->os.hwm.pass_DB = TRUE ;
break ;
case RX_DISABLE_PASS_DB:
smc->os.hwm.pass_DB = FALSE ;
break ;
case RX_DISABLE_PASS_ALL:
smc->os.hwm.pass_SMT = smc->os.hwm.pass_NSA = FALSE ;
smc->os.hwm.pass_DB = FALSE ;
smc->os.hwm.pass_llc_promisc = TRUE ;
mac_set_rx_mode(smc,RX_DISABLE_NSA) ;
break ;
case RX_DISABLE_LLC_PROMISC:
smc->os.hwm.pass_llc_promisc = FALSE ;
break ;
case RX_ENABLE_LLC_PROMISC:
smc->os.hwm.pass_llc_promisc = TRUE ;
break ;
case RX_ENABLE_ALLMULTI:
case RX_DISABLE_ALLMULTI:
case RX_ENABLE_PROMISC:
case RX_DISABLE_PROMISC:
case RX_ENABLE_NSA:
case RX_DISABLE_NSA:
default:
mac_set_rx_mode(smc,mode) ;
break ;
}
}
#endif /* ifndef NDIS_OS2 */
/*
* process receive queue
*/
void process_receive(struct s_smc *smc)
{
int i ;
int n ;
int frag_count ; /* number of RxDs of the curr rx buf */
int used_frags ; /* number of RxDs of the curr frame */
struct s_smt_rx_queue *queue ; /* points to the queue ctl struct */
struct s_smt_fp_rxd volatile *r ; /* rxd pointer */
struct s_smt_fp_rxd volatile *rxd ; /* first rxd of rx frame */
u_long rbctrl ; /* receive buffer control word */
u_long rfsw ; /* receive frame status word */
u_short rx_used ;
u_char far *virt ;
char far *data ;
SMbuf *mb ;
u_char fc ; /* Frame control */
int len ; /* Frame length */
smc->os.hwm.detec_count = 0 ;
queue = smc->hw.fp.rx[QUEUE_R1] ;
NDD_TRACE("RHxB",0,0,0) ;
for ( ; ; ) {
r = queue->rx_curr_get ;
rx_used = queue->rx_used ;
frag_count = 0 ;
#ifdef USE_BREAK_ISR
if (smc->os.hwm.leave_isr) {
goto rx_end ;
}
#endif
#ifdef NDIS_OS2
if (offDepth) {
smc->os.hwm.rx_break = 1 ;
goto rx_end ;
}
smc->os.hwm.rx_break = 0 ;
#endif
#ifdef ODI2
if (smc->os.hwm.rx_break) {
goto rx_end ;
}
#endif
n = 0 ;
do {
DB_RX(5, "Check RxD %p for OWN and EOF", r);
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORCPU) ;
rbctrl = le32_to_cpu(CR_READ(r->rxd_rbctrl));
if (rbctrl & BMU_OWN) {
NDD_TRACE("RHxE",r,rfsw,rbctrl) ;
DB_RX(4, "End of RxDs");
goto rx_end ;
}
/*
* out of RxD detection
*/
if (!rx_used) {
SK_BREAK() ;
SMT_PANIC(smc,HWM_E0009,HWM_E0009_MSG) ;
/* Either we don't have an RxD or all
* RxDs are filled. Therefore it's allowed
* for to set the STOPPED flag */
smc->hw.hw_state = STOPPED ;
mac_drv_clear_rx_queue(smc) ;
smc->hw.hw_state = STARTED ;
mac_drv_fill_rxd(smc) ;
smc->os.hwm.detec_count = 0 ;
goto rx_end ;
}
rfsw = le32_to_cpu(r->rxd_rfsw) ;
if ((rbctrl & BMU_STF) != ((rbctrl & BMU_ST_BUF) <<5)) {
/*
* The BMU_STF bit is deleted, 1 frame is
* placed into more than 1 rx buffer
*
* skip frame by setting the rx len to 0
*
* if fragment count == 0
* The missing STF bit belongs to the
* current frame, search for the
* EOF bit to complete the frame
* else
* the fragment belongs to the next frame,
* exit the loop and process the frame
*/
SK_BREAK() ;
rfsw = 0 ;
if (frag_count) {
break ;
}
}
n += rbctrl & 0xffff ;
r = r->rxd_next ;
frag_count++ ;
rx_used-- ;
} while (!(rbctrl & BMU_EOF)) ;
used_frags = frag_count ;
DB_RX(5, "EOF set in RxD, used_frags = %d", used_frags);
/* may be next 2 DRV_BUF_FLUSH() can be skipped, because */
/* BMU_ST_BUF will not be changed by the ASIC */
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORCPU) ;
while (rx_used && !(r->rxd_rbctrl & cpu_to_le32(BMU_ST_BUF))) {
DB_RX(5, "Check STF bit in %p", r);
r = r->rxd_next ;
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORCPU) ;
frag_count++ ;
rx_used-- ;
}
DB_RX(5, "STF bit found");
/*
* The received frame is finished for the process receive
*/
rxd = queue->rx_curr_get ;
queue->rx_curr_get = r ;
queue->rx_free += frag_count ;
queue->rx_used = rx_used ;
/*
* ASIC Errata no. 7 (STF - Bit Bug)
*/
rxd->rxd_rbctrl &= cpu_to_le32(~BMU_STF) ;
for (r=rxd, i=frag_count ; i ; r=r->rxd_next, i--){
DB_RX(5, "dma_complete for RxD %p", r);
dma_complete(smc,(union s_fp_descr volatile *)r,DMA_WR);
}
smc->hw.fp.err_stats.err_valid++ ;
smc->mib.m[MAC0].fddiMACCopied_Ct++ ;
/* the length of the data including the FC */
len = (rfsw & RD_LENGTH) - 4 ;
DB_RX(4, "frame length = %d", len);
/*
* check the frame_length and all error flags
*/
if (rfsw & (RX_MSRABT|RX_FS_E|RX_FS_CRC|RX_FS_IMPL)){
if (rfsw & RD_S_MSRABT) {
DB_RX(2, "Frame aborted by the FORMAC");
smc->hw.fp.err_stats.err_abort++ ;
}
/*
* check frame status
*/
if (rfsw & RD_S_SEAC2) {
DB_RX(2, "E-Indicator set");
smc->hw.fp.err_stats.err_e_indicator++ ;
}
if (rfsw & RD_S_SFRMERR) {
DB_RX(2, "CRC error");
smc->hw.fp.err_stats.err_crc++ ;
}
if (rfsw & RX_FS_IMPL) {
DB_RX(2, "Implementer frame");
smc->hw.fp.err_stats.err_imp_frame++ ;
}
goto abort_frame ;
}
if (len > FDDI_RAW_MTU-4) {
DB_RX(2, "Frame too long error");
smc->hw.fp.err_stats.err_too_long++ ;
goto abort_frame ;
}
/*
* SUPERNET 3 Bug: FORMAC delivers status words
* of aborted frames to the BMU
*/
if (len <= 4) {
DB_RX(2, "Frame length = 0");
goto abort_frame ;
}
if (len != (n-4)) {
DB_RX(4, "BMU: rx len differs: [%d:%d]", len, n);
smc->os.hwm.rx_len_error++ ;
goto abort_frame ;
}
/*
* Check SA == MA
*/
virt = (u_char far *) rxd->rxd_virt ;
DB_RX(2, "FC = %x", *virt);
if (virt[12] == MA[5] &&
virt[11] == MA[4] &&
virt[10] == MA[3] &&
virt[9] == MA[2] &&
virt[8] == MA[1] &&
(virt[7] & ~GROUP_ADDR_BIT) == MA[0]) {
goto abort_frame ;
}
/*
* test if LLC frame
*/
if (rfsw & RX_FS_LLC) {
/*
* if pass_llc_promisc is disable
* if DA != Multicast or Broadcast or DA!=MA
* abort the frame
*/
if (!smc->os.hwm.pass_llc_promisc) {
if(!(virt[1] & GROUP_ADDR_BIT)) {
if (virt[6] != MA[5] ||
virt[5] != MA[4] ||
virt[4] != MA[3] ||
virt[3] != MA[2] ||
virt[2] != MA[1] ||
virt[1] != MA[0]) {
DB_RX(2, "DA != MA and not multi- or broadcast");
goto abort_frame ;
}
}
}
/*
* LLC frame received
*/
DB_RX(4, "LLC - receive");
mac_drv_rx_complete(smc,rxd,frag_count,len) ;
}
else {
if (!(mb = smt_get_mbuf(smc))) {
smc->hw.fp.err_stats.err_no_buf++ ;
DB_RX(4, "No SMbuf; receive terminated");
goto abort_frame ;
}
data = smtod(mb,char *) - 1 ;
/*
* copy the frame into a SMT_MBuf
*/
#ifdef USE_OS_CPY
hwm_cpy_rxd2mb(rxd,data,len) ;
#else
for (r=rxd, i=used_frags ; i ; r=r->rxd_next, i--){
n = le32_to_cpu(r->rxd_rbctrl) & RD_LENGTH ;
DB_RX(6, "cp SMT frame to mb: len = %d", n);
memcpy(data,r->rxd_virt,n) ;
data += n ;
}
data = smtod(mb,char *) - 1 ;
#endif
fc = *(char *)mb->sm_data = *data ;
mb->sm_len = len - 1 ; /* len - fc */
data++ ;
/*
* SMT frame received
*/
switch(fc) {
case FC_SMT_INFO :
smc->hw.fp.err_stats.err_smt_frame++ ;
DB_RX(5, "SMT frame received");
if (smc->os.hwm.pass_SMT) {
DB_RX(5, "pass SMT frame");
mac_drv_rx_complete(smc, rxd,
frag_count,len) ;
}
else {
DB_RX(5, "requeue RxD");
mac_drv_requeue_rxd(smc,rxd,frag_count);
}
smt_received_pack(smc,mb,(int)(rfsw>>25)) ;
break ;
case FC_SMT_NSA :
smc->hw.fp.err_stats.err_smt_frame++ ;
DB_RX(5, "SMT frame received");
/* if pass_NSA set pass the NSA frame or */
/* pass_SMT set and the A-Indicator */
/* is not set, pass the NSA frame */
if (smc->os.hwm.pass_NSA ||
(smc->os.hwm.pass_SMT &&
!(rfsw & A_INDIC))) {
DB_RX(5, "pass SMT frame");
mac_drv_rx_complete(smc, rxd,
frag_count,len) ;
}
else {
DB_RX(5, "requeue RxD");
mac_drv_requeue_rxd(smc,rxd,frag_count);
}
smt_received_pack(smc,mb,(int)(rfsw>>25)) ;
break ;
case FC_BEACON :
if (smc->os.hwm.pass_DB) {
DB_RX(5, "pass DB frame");
mac_drv_rx_complete(smc, rxd,
frag_count,len) ;
}
else {
DB_RX(5, "requeue RxD");
mac_drv_requeue_rxd(smc,rxd,frag_count);
}
smt_free_mbuf(smc,mb) ;
break ;
default :
/*
* unknown FC abort the frame
*/
DB_RX(2, "unknown FC error");
smt_free_mbuf(smc,mb) ;
DB_RX(5, "requeue RxD");
mac_drv_requeue_rxd(smc,rxd,frag_count) ;
if ((fc & 0xf0) == FC_MAC)
smc->hw.fp.err_stats.err_mac_frame++ ;
else
smc->hw.fp.err_stats.err_imp_frame++ ;
break ;
}
}
DB_RX(3, "next RxD is %p", queue->rx_curr_get);
NDD_TRACE("RHx1",queue->rx_curr_get,0,0) ;
continue ;
/*--------------------------------------------------------------------*/
abort_frame:
DB_RX(5, "requeue RxD");
mac_drv_requeue_rxd(smc,rxd,frag_count) ;
DB_RX(3, "next RxD is %p", queue->rx_curr_get);
NDD_TRACE("RHx2",queue->rx_curr_get,0,0) ;
}
rx_end:
#ifdef ALL_RX_COMPLETE
mac_drv_all_receives_complete(smc) ;
#endif
return ; /* lint bug: needs return detect end of function */
}
static void smt_to_llc(struct s_smc *smc, SMbuf *mb)
{
u_char fc ;
DB_RX(4, "send a queued frame to the llc layer");
smc->os.hwm.r.len = mb->sm_len ;
smc->os.hwm.r.mb_pos = smtod(mb,char *) ;
fc = *smc->os.hwm.r.mb_pos ;
(void)mac_drv_rx_init(smc,(int)mb->sm_len,(int)fc,
smc->os.hwm.r.mb_pos,(int)mb->sm_len) ;
smt_free_mbuf(smc,mb) ;
}
/*
* BEGIN_MANUAL_ENTRY(hwm_rx_frag)
* void hwm_rx_frag(smc,virt,phys,len,frame_status)
*
* function MACRO (hardware module, hwmtm.h)
* This function calls dma_master for preparing the
* system hardware for the DMA transfer and initializes
* the current RxD with the length and the physical and
* virtual address of the fragment. Furthermore, it sets the
* STF and EOF bits depending on the frame status byte,
* switches the OWN flag of the RxD, so that it is owned by the
* adapter and issues an rx_start.
*
* para virt virtual pointer to the fragment
* len the length of the fragment
* frame_status status of the frame, see design description
*
* NOTE: It is possible to call this function with a fragment length
* of zero.
*
* END_MANUAL_ENTRY
*/
void hwm_rx_frag(struct s_smc *smc, char far *virt, u_long phys, int len,
int frame_status)
{
struct s_smt_fp_rxd volatile *r ;
__le32 rbctrl;
NDD_TRACE("RHfB",virt,len,frame_status) ;
DB_RX(2, "hwm_rx_frag: len = %d, frame_status = %x", len, frame_status);
r = smc->hw.fp.rx_q[QUEUE_R1].rx_curr_put ;
r->rxd_virt = virt ;
r->rxd_rbadr = cpu_to_le32(phys) ;
rbctrl = cpu_to_le32( (((__u32)frame_status &
(FIRST_FRAG|LAST_FRAG))<<26) |
(((u_long) frame_status & FIRST_FRAG) << 21) |
BMU_OWN | BMU_CHECK | BMU_EN_IRQ_EOF | len) ;
r->rxd_rbctrl = rbctrl ;
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORDEV) ;
outpd(ADDR(B0_R1_CSR),CSR_START) ;
smc->hw.fp.rx_q[QUEUE_R1].rx_free-- ;
smc->hw.fp.rx_q[QUEUE_R1].rx_used++ ;
smc->hw.fp.rx_q[QUEUE_R1].rx_curr_put = r->rxd_next ;
NDD_TRACE("RHfE",r,le32_to_cpu(r->rxd_rbadr),0) ;
}
/*
* BEGINN_MANUAL_ENTRY(mac_drv_clear_rx_queue)
*
* void mac_drv_clear_rx_queue(smc)
* struct s_smc *smc ;
*
* function DOWNCALL (hardware module, hwmtm.c)
* mac_drv_clear_rx_queue is called by the OS-specific module
* after it has issued a card_stop.
* In this case, the frames in the receive queue are obsolete and
* should be removed. For removing mac_drv_clear_rx_queue
* calls dma_master for each RxD and mac_drv_clear_rxd for each
* receive buffer.
*
* NOTE: calling sequence card_stop:
* CLI_FBI(), card_stop(),
* mac_drv_clear_tx_queue(), mac_drv_clear_rx_queue(),
*
* NOTE: The caller is responsible that the BMUs are idle
* when this function is called.
*
* END_MANUAL_ENTRY
*/
void mac_drv_clear_rx_queue(struct s_smc *smc)
{
struct s_smt_fp_rxd volatile *r ;
struct s_smt_fp_rxd volatile *next_rxd ;
struct s_smt_rx_queue *queue ;
int frag_count ;
int i ;
if (smc->hw.hw_state != STOPPED) {
SK_BREAK() ;
SMT_PANIC(smc,HWM_E0012,HWM_E0012_MSG) ;
return ;
}
queue = smc->hw.fp.rx[QUEUE_R1] ;
DB_RX(5, "clear_rx_queue");
/*
* dma_complete and mac_drv_clear_rxd for all RxDs / receive buffers
*/
r = queue->rx_curr_get ;
while (queue->rx_used) {
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORCPU) ;
DB_RX(5, "switch OWN bit of RxD 0x%p", r);
r->rxd_rbctrl &= ~cpu_to_le32(BMU_OWN) ;
frag_count = 1 ;
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORDEV) ;
r = r->rxd_next ;
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORCPU) ;
while (r != queue->rx_curr_put &&
!(r->rxd_rbctrl & cpu_to_le32(BMU_ST_BUF))) {
DB_RX(5, "Check STF bit in %p", r);
r->rxd_rbctrl &= ~cpu_to_le32(BMU_OWN) ;
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORDEV) ;
r = r->rxd_next ;
DRV_BUF_FLUSH(r,DDI_DMA_SYNC_FORCPU) ;
frag_count++ ;
}
DB_RX(5, "STF bit found");
next_rxd = r ;
for (r=queue->rx_curr_get,i=frag_count; i ; r=r->rxd_next,i--){
DB_RX(5, "dma_complete for RxD %p", r);
dma_complete(smc,(union s_fp_descr volatile *)r,DMA_WR);
}
DB_RX(5, "mac_drv_clear_rxd: RxD %p frag_count %d",
queue->rx_curr_get, frag_count);
mac_drv_clear_rxd(smc,queue->rx_curr_get,frag_count) ;
queue->rx_curr_get = next_rxd ;
queue->rx_used -= frag_count ;
queue->rx_free += frag_count ;
}
}
/*
-------------------------------------------------------------
SEND FUNCTIONS:
-------------------------------------------------------------
*/
/*
* BEGIN_MANUAL_ENTRY(hwm_tx_init)
* int hwm_tx_init(smc,fc,frag_count,frame_len,frame_status)
*
* function DOWN_CALL (hardware module, hwmtm.c)
* hwm_tx_init checks if the frame can be sent through the
* corresponding send queue.
*
* para fc the frame control. To determine through which
* send queue the frame should be transmitted.
* 0x50 - 0x57: asynchronous LLC frame
* 0xD0 - 0xD7: synchronous LLC frame
* 0x41, 0x4F: SMT frame to the network
* 0x42: SMT frame to the network and to the local SMT
* 0x43: SMT frame to the local SMT
* frag_count count of the fragments for this frame
* frame_len length of the frame
* frame_status status of the frame, the send queue bit is already
* specified
*
* return frame_status
*
* END_MANUAL_ENTRY
*/
int hwm_tx_init(struct s_smc *smc, u_char fc, int frag_count, int frame_len,
int frame_status)
{
NDD_TRACE("THiB",fc,frag_count,frame_len) ;
smc->os.hwm.tx_p = smc->hw.fp.tx[frame_status & QUEUE_A0] ;
smc->os.hwm.tx_descr = TX_DESCRIPTOR | (((u_long)(frame_len-1)&3)<<27) ;
smc->os.hwm.tx_len = frame_len ;
DB_TX(3, "hwm_tx_init: fc = %x, len = %d", fc, frame_len);
if ((fc & ~(FC_SYNC_BIT|FC_LLC_PRIOR)) == FC_ASYNC_LLC) {
frame_status |= LAN_TX ;
}
else {
switch (fc) {
case FC_SMT_INFO :
case FC_SMT_NSA :
frame_status |= LAN_TX ;
break ;
case FC_SMT_LOC :
frame_status |= LOC_TX ;
break ;
case FC_SMT_LAN_LOC :
frame_status |= LAN_TX | LOC_TX ;
break ;
default :
SMT_PANIC(smc,HWM_E0010,HWM_E0010_MSG) ;
}
}
if (!smc->hw.mac_ring_is_up) {
frame_status &= ~LAN_TX ;
frame_status |= RING_DOWN ;
DB_TX(2, "Ring is down: terminate LAN_TX");
}
if (frag_count > smc->os.hwm.tx_p->tx_free) {
#ifndef NDIS_OS2
mac_drv_clear_txd(smc) ;
if (frag_count > smc->os.hwm.tx_p->tx_free) {
DB_TX(2, "Out of TxDs, terminate LAN_TX");
frame_status &= ~LAN_TX ;
frame_status |= OUT_OF_TXD ;
}
#else
DB_TX(2, "Out of TxDs, terminate LAN_TX");
frame_status &= ~LAN_TX ;
frame_status |= OUT_OF_TXD ;
#endif
}
DB_TX(3, "frame_status = %x", frame_status);
NDD_TRACE("THiE",frame_status,smc->os.hwm.tx_p->tx_free,0) ;
return frame_status;
}
/*
* BEGIN_MANUAL_ENTRY(hwm_tx_frag)
* void hwm_tx_frag(smc,virt,phys,len,frame_status)
*
* function DOWNCALL (hardware module, hwmtm.c)
* If the frame should be sent to the LAN, this function calls
* dma_master, fills the current TxD with the virtual and the
* physical address, sets the STF and EOF bits dependent on
* the frame status, and requests the BMU to start the
* transmit.
* If the frame should be sent to the local SMT, an SMT_MBuf
* is allocated if the FIRST_FRAG bit is set in the frame_status.
* The fragment of the frame is copied into the SMT MBuf.
* The function smt_received_pack is called if the LAST_FRAG
* bit is set in the frame_status word.
*
* para virt virtual pointer to the fragment
* len the length of the fragment
* frame_status status of the frame, see design description
*
* return nothing returned, no parameter is modified
*
* NOTE: It is possible to invoke this macro with a fragment length
* of zero.
*
* END_MANUAL_ENTRY
*/
void hwm_tx_frag(struct s_smc *smc, char far *virt, u_long phys, int len,
int frame_status)
{
struct s_smt_fp_txd volatile *t ;
struct s_smt_tx_queue *queue ;
__le32 tbctrl ;
queue = smc->os.hwm.tx_p ;
NDD_TRACE("THfB",virt,len,frame_status) ;
/* Bug fix: AF / May 31 1999 (#missing)
* snmpinfo problem reported by IBM is caused by invalid
* t-pointer (txd) if LAN_TX is not set but LOC_TX only.
* Set: t = queue->tx_curr_put here !
*/
t = queue->tx_curr_put ;
DB_TX(2, "hwm_tx_frag: len = %d, frame_status = %x", len, frame_status);
if (frame_status & LAN_TX) {
/* '*t' is already defined */
DB_TX(3, "LAN_TX: TxD = %p, virt = %p", t, virt);
t->txd_virt = virt ;
t->txd_txdscr = cpu_to_le32(smc->os.hwm.tx_descr) ;
t->txd_tbadr = cpu_to_le32(phys) ;
tbctrl = cpu_to_le32((((__u32)frame_status &
(FIRST_FRAG|LAST_FRAG|EN_IRQ_EOF))<< 26) |
BMU_OWN|BMU_CHECK |len) ;
t->txd_tbctrl = tbctrl ;
#ifndef AIX
DRV_BUF_FLUSH(t,DDI_DMA_SYNC_FORDEV) ;
outpd(queue->tx_bmu_ctl,CSR_START) ;
#else /* ifndef AIX */
DRV_BUF_FLUSH(t,DDI_DMA_SYNC_FORDEV) ;
if (frame_status & QUEUE_A0) {
outpd(ADDR(B0_XA_CSR),CSR_START) ;
}
else {
outpd(ADDR(B0_XS_CSR),CSR_START) ;
}
#endif
queue->tx_free-- ;
queue->tx_used++ ;
queue->tx_curr_put = t->txd_next ;
if (frame_status & LAST_FRAG) {
smc->mib.m[MAC0].fddiMACTransmit_Ct++ ;
}
}
if (frame_status & LOC_TX) {
DB_TX(3, "LOC_TX:");
if (frame_status & FIRST_FRAG) {
if(!(smc->os.hwm.tx_mb = smt_get_mbuf(smc))) {
smc->hw.fp.err_stats.err_no_buf++ ;
DB_TX(4, "No SMbuf; transmit terminated");
}
else {
smc->os.hwm.tx_data =
smtod(smc->os.hwm.tx_mb,char *) - 1 ;
#ifdef USE_OS_CPY
#ifdef PASS_1ST_TXD_2_TX_COMP
hwm_cpy_txd2mb(t,smc->os.hwm.tx_data,
smc->os.hwm.tx_len) ;
#endif
#endif
}
}
if (smc->os.hwm.tx_mb) {
#ifndef USE_OS_CPY
DB_TX(3, "copy fragment into MBuf");
memcpy(smc->os.hwm.tx_data,virt,len) ;
smc->os.hwm.tx_data += len ;
#endif
if (frame_status & LAST_FRAG) {
#ifdef USE_OS_CPY
#ifndef PASS_1ST_TXD_2_TX_COMP
/*
* hwm_cpy_txd2mb(txd,data,len) copies 'len'
* bytes from the virtual pointer in 'rxd'
* to 'data'. The virtual pointer of the
* os-specific tx-buffer should be written
* in the LAST txd.
*/
hwm_cpy_txd2mb(t,smc->os.hwm.tx_data,
smc->os.hwm.tx_len) ;
#endif /* nPASS_1ST_TXD_2_TX_COMP */
#endif /* USE_OS_CPY */
smc->os.hwm.tx_data =
smtod(smc->os.hwm.tx_mb,char *) - 1 ;
*(char *)smc->os.hwm.tx_mb->sm_data =
*smc->os.hwm.tx_data ;
smc->os.hwm.tx_data++ ;
smc->os.hwm.tx_mb->sm_len =
smc->os.hwm.tx_len - 1 ;
DB_TX(3, "pass LLC frame to SMT");
smt_received_pack(smc,smc->os.hwm.tx_mb,
RD_FS_LOCAL) ;
}
}
}
NDD_TRACE("THfE",t,queue->tx_free,0) ;
}
/*
* queues a receive for later send
*/
static void queue_llc_rx(struct s_smc *smc, SMbuf *mb)
{
DB_GEN(4, "queue_llc_rx: mb = %p", mb);
smc->os.hwm.queued_rx_frames++ ;
mb->sm_next = (SMbuf *)NULL ;
if (smc->os.hwm.llc_rx_pipe == NULL) {
smc->os.hwm.llc_rx_pipe = mb ;
}
else {
smc->os.hwm.llc_rx_tail->sm_next = mb ;
}
smc->os.hwm.llc_rx_tail = mb ;
/*
* force an timer IRQ to receive the data
*/
if (!smc->os.hwm.isr_flag) {
smt_force_irq(smc) ;
}
}
/*
* get a SMbuf from the llc_rx_queue
*/
static SMbuf *get_llc_rx(struct s_smc *smc)
{
SMbuf *mb ;
if ((mb = smc->os.hwm.llc_rx_pipe)) {
smc->os.hwm.queued_rx_frames-- ;
smc->os.hwm.llc_rx_pipe = mb->sm_next ;
}
DB_GEN(4, "get_llc_rx: mb = 0x%p", mb);
return mb;
}
/*
* queues a transmit SMT MBuf during the time were the MBuf is
* queued the TxD ring
*/
static void queue_txd_mb(struct s_smc *smc, SMbuf *mb)
{
DB_GEN(4, "_rx: queue_txd_mb = %p", mb);
smc->os.hwm.queued_txd_mb++ ;
mb->sm_next = (SMbuf *)NULL ;
if (smc->os.hwm.txd_tx_pipe == NULL) {
smc->os.hwm.txd_tx_pipe = mb ;
}
else {
smc->os.hwm.txd_tx_tail->sm_next = mb ;
}
smc->os.hwm.txd_tx_tail = mb ;
}
/*
* get a SMbuf from the txd_tx_queue
*/
static SMbuf *get_txd_mb(struct s_smc *smc)
{
SMbuf *mb ;
if ((mb = smc->os.hwm.txd_tx_pipe)) {
smc->os.hwm.queued_txd_mb-- ;
smc->os.hwm.txd_tx_pipe = mb->sm_next ;
}
DB_GEN(4, "get_txd_mb: mb = 0x%p", mb);
return mb;
}
/*
* SMT Send function
*/
void smt_send_mbuf(struct s_smc *smc, SMbuf *mb, int fc)
{
char far *data ;
int len ;
int n ;
int i ;
int frag_count ;
int frame_status ;
SK_LOC_DECL(char far,*virt[3]) ;
int frag_len[3] ;
struct s_smt_tx_queue *queue ;
struct s_smt_fp_txd volatile *t ;
u_long phys ;
__le32 tbctrl;
NDD_TRACE("THSB",mb,fc,0) ;
DB_TX(4, "smt_send_mbuf: mb = 0x%p, fc = 0x%x", mb, fc);
mb->sm_off-- ; /* set to fc */
mb->sm_len++ ; /* + fc */
data = smtod(mb,char *) ;
*data = fc ;
if (fc == FC_SMT_LOC)
*data = FC_SMT_INFO ;
/*
* determine the frag count and the virt addresses of the frags
*/
frag_count = 0 ;
len = mb->sm_len ;
while (len) {
n = SMT_PAGESIZE - ((long)data & (SMT_PAGESIZE-1)) ;
if (n >= len) {
n = len ;
}
DB_TX(5, "frag: virt/len = 0x%p/%d", data, n);
virt[frag_count] = data ;
frag_len[frag_count] = n ;
frag_count++ ;
len -= n ;
data += n ;
}
/*
* determine the frame status
*/
queue = smc->hw.fp.tx[QUEUE_A0] ;
if (fc == FC_BEACON || fc == FC_SMT_LOC) {
frame_status = LOC_TX ;
}
else {
frame_status = LAN_TX ;
if ((smc->os.hwm.pass_NSA &&(fc == FC_SMT_NSA)) ||
(smc->os.hwm.pass_SMT &&(fc == FC_SMT_INFO)))
frame_status |= LOC_TX ;
}
if (!smc->hw.mac_ring_is_up || frag_count > queue->tx_free) {
frame_status &= ~LAN_TX;
if (frame_status) {
DB_TX(2, "Ring is down: terminate LAN_TX");
}
else {
DB_TX(2, "Ring is down: terminate transmission");
smt_free_mbuf(smc,mb) ;
return ;
}
}
DB_TX(5, "frame_status = 0x%x", frame_status);
if ((frame_status & LAN_TX) && (frame_status & LOC_TX)) {
mb->sm_use_count = 2 ;
}
if (frame_status & LAN_TX) {
t = queue->tx_curr_put ;
frame_status |= FIRST_FRAG ;
for (i = 0; i < frag_count; i++) {
DB_TX(5, "init TxD = 0x%p", t);
if (i == frag_count-1) {
frame_status |= LAST_FRAG ;
t->txd_txdscr = cpu_to_le32(TX_DESCRIPTOR |
(((__u32)(mb->sm_len-1)&3) << 27)) ;
}
t->txd_virt = virt[i] ;
phys = dma_master(smc, (void far *)virt[i],
frag_len[i], DMA_RD|SMT_BUF) ;
t->txd_tbadr = cpu_to_le32(phys) ;
tbctrl = cpu_to_le32((((__u32)frame_status &
(FIRST_FRAG|LAST_FRAG)) << 26) |
BMU_OWN | BMU_CHECK | BMU_SMT_TX |frag_len[i]) ;
t->txd_tbctrl = tbctrl ;
#ifndef AIX
DRV_BUF_FLUSH(t,DDI_DMA_SYNC_FORDEV) ;
outpd(queue->tx_bmu_ctl,CSR_START) ;
#else
DRV_BUF_FLUSH(t,DDI_DMA_SYNC_FORDEV) ;
outpd(ADDR(B0_XA_CSR),CSR_START) ;
#endif
frame_status &= ~FIRST_FRAG ;
queue->tx_curr_put = t = t->txd_next ;
queue->tx_free-- ;
queue->tx_used++ ;
}
smc->mib.m[MAC0].fddiMACTransmit_Ct++ ;
queue_txd_mb(smc,mb) ;
}
if (frame_status & LOC_TX) {
DB_TX(5, "pass Mbuf to LLC queue");
queue_llc_rx(smc,mb) ;
}
/*
* We need to unqueue the free SMT_MBUFs here, because it may
* be that the SMT want's to send more than 1 frame for one down call
*/
mac_drv_clear_txd(smc) ;
NDD_TRACE("THSE",t,queue->tx_free,frag_count) ;
}
/* BEGIN_MANUAL_ENTRY(mac_drv_clear_txd)
* void mac_drv_clear_txd(smc)
*
* function DOWNCALL (hardware module, hwmtm.c)
* mac_drv_clear_txd searches in both send queues for TxD's
* which were finished by the adapter. It calls dma_complete
* for each TxD. If the last fragment of an LLC frame is
* reached, it calls mac_drv_tx_complete to release the
* send buffer.
*
* return nothing
*
* END_MANUAL_ENTRY
*/
static void mac_drv_clear_txd(struct s_smc *smc)
{
struct s_smt_tx_queue *queue ;
struct s_smt_fp_txd volatile *t1 ;
struct s_smt_fp_txd volatile *t2 = NULL ;
SMbuf *mb ;
u_long tbctrl ;
int i ;
int frag_count ;
int n ;
NDD_TRACE("THcB",0,0,0) ;
for (i = QUEUE_S; i <= QUEUE_A0; i++) {
queue = smc->hw.fp.tx[i] ;
t1 = queue->tx_curr_get ;
DB_TX(5, "clear_txd: QUEUE = %d (0=sync/1=async)", i);
for ( ; ; ) {
frag_count = 0 ;
do {
DRV_BUF_FLUSH(t1,DDI_DMA_SYNC_FORCPU) ;
DB_TX(5, "check OWN/EOF bit of TxD 0x%p", t1);
tbctrl = le32_to_cpu(CR_READ(t1->txd_tbctrl));
if (tbctrl & BMU_OWN || !queue->tx_used){
DB_TX(4, "End of TxDs queue %d", i);
goto free_next_queue ; /* next queue */
}
t1 = t1->txd_next ;
frag_count++ ;
} while (!(tbctrl & BMU_EOF)) ;
t1 = queue->tx_curr_get ;
for (n = frag_count; n; n--) {
tbctrl = le32_to_cpu(t1->txd_tbctrl) ;
dma_complete(smc,
(union s_fp_descr volatile *) t1,
(int) (DMA_RD |
((tbctrl & BMU_SMT_TX) >> 18))) ;
t2 = t1 ;
t1 = t1->txd_next ;
}
if (tbctrl & BMU_SMT_TX) {
mb = get_txd_mb(smc) ;
smt_free_mbuf(smc,mb) ;
}
else {
#ifndef PASS_1ST_TXD_2_TX_COMP
DB_TX(4, "mac_drv_tx_comp for TxD 0x%p", t2);
mac_drv_tx_complete(smc,t2) ;
#else
DB_TX(4, "mac_drv_tx_comp for TxD 0x%x",
queue->tx_curr_get);
mac_drv_tx_complete(smc,queue->tx_curr_get) ;
#endif
}
queue->tx_curr_get = t1 ;
queue->tx_free += frag_count ;
queue->tx_used -= frag_count ;
}
free_next_queue: ;
}
NDD_TRACE("THcE",0,0,0) ;
}
/*
* BEGINN_MANUAL_ENTRY(mac_drv_clear_tx_queue)
*
* void mac_drv_clear_tx_queue(smc)
* struct s_smc *smc ;
*
* function DOWNCALL (hardware module, hwmtm.c)
* mac_drv_clear_tx_queue is called from the SMT when
* the RMT state machine has entered the ISOLATE state.
* This function is also called by the os-specific module
* after it has called the function card_stop().
* In this case, the frames in the send queues are obsolete and
* should be removed.
*
* note calling sequence:
* CLI_FBI(), card_stop(),
* mac_drv_clear_tx_queue(), mac_drv_clear_rx_queue(),
*
* NOTE: The caller is responsible that the BMUs are idle
* when this function is called.
*
* END_MANUAL_ENTRY
*/
void mac_drv_clear_tx_queue(struct s_smc *smc)
{
struct s_smt_fp_txd volatile *t ;
struct s_smt_tx_queue *queue ;
int tx_used ;
int i ;
if (smc->hw.hw_state != STOPPED) {
SK_BREAK() ;
SMT_PANIC(smc,HWM_E0011,HWM_E0011_MSG) ;
return ;
}
for (i = QUEUE_S; i <= QUEUE_A0; i++) {
queue = smc->hw.fp.tx[i] ;
DB_TX(5, "clear_tx_queue: QUEUE = %d (0=sync/1=async)", i);
/*
* switch the OWN bit of all pending frames to the host
*/
t = queue->tx_curr_get ;
tx_used = queue->tx_used ;
while (tx_used) {
DRV_BUF_FLUSH(t,DDI_DMA_SYNC_FORCPU) ;
DB_TX(5, "switch OWN bit of TxD 0x%p", t);
t->txd_tbctrl &= ~cpu_to_le32(BMU_OWN) ;
DRV_BUF_FLUSH(t,DDI_DMA_SYNC_FORDEV) ;
t = t->txd_next ;
tx_used-- ;
}
}
/*
* release all TxD's for both send queues
*/
mac_drv_clear_txd(smc) ;
for (i = QUEUE_S; i <= QUEUE_A0; i++) {
queue = smc->hw.fp.tx[i] ;
t = queue->tx_curr_get ;
/*
* write the phys pointer of the NEXT descriptor into the
* BMU's current address descriptor pointer and set
* tx_curr_get and tx_curr_put to this position
*/
if (i == QUEUE_S) {
outpd(ADDR(B5_XS_DA),le32_to_cpu(t->txd_ntdadr)) ;
}
else {
outpd(ADDR(B5_XA_DA),le32_to_cpu(t->txd_ntdadr)) ;
}
queue->tx_curr_put = queue->tx_curr_get->txd_next ;
queue->tx_curr_get = queue->tx_curr_put ;
}
}
/*
-------------------------------------------------------------
TEST FUNCTIONS:
-------------------------------------------------------------
*/
#ifdef DEBUG
/*
* BEGIN_MANUAL_ENTRY(mac_drv_debug_lev)
* void mac_drv_debug_lev(smc,flag,lev)
*
* function DOWNCALL (drvsr.c)
* To get a special debug info the user can assign a debug level
* to any debug flag.
*
* para flag debug flag, possible values are:
* = 0: reset all debug flags (the defined level is
* ignored)
* = 1: debug.d_smtf
* = 2: debug.d_smt
* = 3: debug.d_ecm
* = 4: debug.d_rmt
* = 5: debug.d_cfm
* = 6: debug.d_pcm
*
* = 10: debug.d_os.hwm_rx (hardware module receive path)
* = 11: debug.d_os.hwm_tx(hardware module transmit path)
* = 12: debug.d_os.hwm_gen(hardware module general flag)
*
* lev debug level
*
* END_MANUAL_ENTRY
*/
void mac_drv_debug_lev(struct s_smc *smc, int flag, int lev)
{
switch(flag) {
case (int)NULL:
DB_P.d_smtf = DB_P.d_smt = DB_P.d_ecm = DB_P.d_rmt = 0 ;
DB_P.d_cfm = 0 ;
DB_P.d_os.hwm_rx = DB_P.d_os.hwm_tx = DB_P.d_os.hwm_gen = 0 ;
#ifdef SBA
DB_P.d_sba = 0 ;
#endif
#ifdef ESS
DB_P.d_ess = 0 ;
#endif
break ;
case DEBUG_SMTF:
DB_P.d_smtf = lev ;
break ;
case DEBUG_SMT:
DB_P.d_smt = lev ;
break ;
case DEBUG_ECM:
DB_P.d_ecm = lev ;
break ;
case DEBUG_RMT:
DB_P.d_rmt = lev ;
break ;
case DEBUG_CFM:
DB_P.d_cfm = lev ;
break ;
case DEBUG_PCM:
DB_P.d_pcm = lev ;
break ;
case DEBUG_SBA:
#ifdef SBA
DB_P.d_sba = lev ;
#endif
break ;
case DEBUG_ESS:
#ifdef ESS
DB_P.d_ess = lev ;
#endif
break ;
case DB_HWM_RX:
DB_P.d_os.hwm_rx = lev ;
break ;
case DB_HWM_TX:
DB_P.d_os.hwm_tx = lev ;
break ;
case DB_HWM_GEN:
DB_P.d_os.hwm_gen = lev ;
break ;
default:
break ;
}
}
#endif