summaryrefslogtreecommitdiff
path: root/cpu/ixp/npe/IxEthAccDataPlane.c
diff options
context:
space:
mode:
authorWolfgang Denk <wd@castor.denx.de>2006-05-30 15:56:48 +0200
committerWolfgang Denk <wd@castor.denx.de>2006-05-30 15:56:48 +0200
commitba94a1bba3600d387edba7eb451990d9891e1c2f (patch)
treee84f737ac88e15342b4cab23c9e631987e8ee75e /cpu/ixp/npe/IxEthAccDataPlane.c
parent5770a1e488621a9e7e344afed7c921ff4e715a63 (diff)
downloadu-boot-imx-ba94a1bba3600d387edba7eb451990d9891e1c2f.zip
u-boot-imx-ba94a1bba3600d387edba7eb451990d9891e1c2f.tar.gz
u-boot-imx-ba94a1bba3600d387edba7eb451990d9891e1c2f.tar.bz2
* Update Intel IXP4xx support
- Add IXP4xx NPE ethernet MAC support - Add support for Intel IXDPG425 board - Add support for Prodrive PDNB3 board - Add IRQ support Patch by Stefan Roese, 23 May 2006 [This patch does not include cpu/ixp/npe/IxNpeMicrocode.c which still sufferes from licensing issues. Blame Intel.]
Diffstat (limited to 'cpu/ixp/npe/IxEthAccDataPlane.c')
-rw-r--r--cpu/ixp/npe/IxEthAccDataPlane.c2483
1 files changed, 2483 insertions, 0 deletions
diff --git a/cpu/ixp/npe/IxEthAccDataPlane.c b/cpu/ixp/npe/IxEthAccDataPlane.c
new file mode 100644
index 0000000..e46fc9b
--- /dev/null
+++ b/cpu/ixp/npe/IxEthAccDataPlane.c
@@ -0,0 +1,2483 @@
+/**
+ * @file IxEthDataPlane.c
+ *
+ * @author Intel Corporation
+ * @date 12-Feb-2002
+ *
+ * @brief This file contains the implementation of the IXPxxx
+ * Ethernet Access Data plane component
+ *
+ * Design Notes:
+ *
+ * @par
+ * IXP400 SW Release version 2.0
+ *
+ * -- Copyright Notice --
+ *
+ * @par
+ * Copyright 2001-2005, Intel Corporation.
+ * All rights reserved.
+ *
+ * @par
+ * 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. Neither the name of the Intel Corporation nor the names of its contributors
+ * may be used to endorse or promote products derived from this software
+ * without specific prior written permission.
+ *
+ * @par
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS ``AS IS''
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS 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.
+ *
+ * @par
+ * -- End of Copyright Notice --
+ */
+
+#include "IxNpeMh.h"
+#include "IxEthAcc.h"
+#include "IxEthDB.h"
+#include "IxOsal.h"
+#include "IxEthDBPortDefs.h"
+#include "IxFeatureCtrl.h"
+#include "IxEthAcc_p.h"
+#include "IxEthAccQueueAssign_p.h"
+
+extern PUBLIC IxEthAccMacState ixEthAccMacState[];
+extern PUBLIC UINT32 ixEthAccNewSrcMask;
+
+/**
+ * private functions prototype
+ */
+PRIVATE IX_OSAL_MBUF *
+ixEthAccEntryFromQConvert(UINT32 qEntry, UINT32 mask);
+
+PRIVATE UINT32
+ixEthAccMbufRxQPrepare(IX_OSAL_MBUF *mbuf);
+
+PRIVATE UINT32
+ixEthAccMbufTxQPrepare(IX_OSAL_MBUF *mbuf);
+
+PRIVATE IxEthAccStatus
+ixEthAccTxSwQHighestPriorityGet(IxEthAccPortId portId,
+ IxEthAccTxPriority *priorityPtr);
+
+PRIVATE IxEthAccStatus
+ixEthAccTxFromSwQ(IxEthAccPortId portId,
+ IxEthAccTxPriority priority);
+
+PRIVATE IxEthAccStatus
+ixEthAccRxFreeFromSwQ(IxEthAccPortId portId);
+
+PRIVATE void
+ixEthAccMbufFromTxQ(IX_OSAL_MBUF *mbuf);
+
+PRIVATE void
+ixEthAccMbufFromRxQ(IX_OSAL_MBUF *mbuf);
+
+PRIVATE IX_STATUS
+ixEthAccQmgrLockTxWrite(IxEthAccPortId portId,
+ UINT32 qBuffer);
+
+PRIVATE IX_STATUS
+ixEthAccQmgrLockRxWrite(IxEthAccPortId portId,
+ UINT32 qBuffer);
+
+PRIVATE IX_STATUS
+ixEthAccQmgrTxWrite(IxEthAccPortId portId,
+ UINT32 qBuffer,
+ UINT32 priority);
+
+/**
+ * @addtogroup IxEthAccPri
+ *@{
+ */
+
+/* increment a counter only when stats are enabled */
+#define TX_STATS_INC(port,field) \
+ IX_ETH_ACC_STATS_INC(ixEthAccPortData[port].ixEthAccTxData.stats.field)
+#define RX_STATS_INC(port,field) \
+ IX_ETH_ACC_STATS_INC(ixEthAccPortData[port].ixEthAccRxData.stats.field)
+
+/* always increment the counter (mainly used for unexpected errors) */
+#define TX_INC(port,field) \
+ ixEthAccPortData[port].ixEthAccTxData.stats.field++
+#define RX_INC(port,field) \
+ ixEthAccPortData[port].ixEthAccRxData.stats.field++
+
+PRIVATE IxEthAccDataPlaneStats ixEthAccDataStats;
+
+extern IxEthAccPortDataInfo ixEthAccPortData[];
+extern IxEthAccInfo ixEthAccDataInfo;
+
+PRIVATE IxOsalFastMutex txWriteMutex[IX_ETH_ACC_NUMBER_OF_PORTS];
+PRIVATE IxOsalFastMutex rxWriteMutex[IX_ETH_ACC_NUMBER_OF_PORTS];
+
+/**
+ *
+ * @brief Mbuf header conversion macros : they implement the
+ * different conversions using a temporary value. They also double-check
+ * that the parameters can be converted to/from NPE format.
+ *
+ */
+#if defined(__wince) && !defined(IN_KERNEL)
+#define PTR_VIRT2NPE(ptrSrc,dst) \
+ do { UINT32 temp; \
+ IX_OSAL_ENSURE(sizeof(ptrSrc) == sizeof(UINT32), "Wrong parameter type"); \
+ IX_OSAL_ENSURE(sizeof(dst) == sizeof(UINT32), "Wrong parameter type"); \
+ temp = (UINT32)IX_OSAL_MBUF_MBUF_VIRTUAL_TO_PHYSICAL_TRANSLATION((IX_OSAL_MBUF*)ptrSrc); \
+ (dst) = IX_OSAL_SWAP_BE_SHARED_LONG(temp); } \
+ while(0)
+
+#define PTR_NPE2VIRT(type,src,ptrDst) \
+ do { void *temp; \
+ IX_OSAL_ENSURE(sizeof(type) == sizeof(UINT32), "Wrong parameter type"); \
+ IX_OSAL_ENSURE(sizeof(src) == sizeof(UINT32), "Wrong parameter type"); \
+ IX_OSAL_ENSURE(sizeof(ptrDst) == sizeof(UINT32), "Wrong parameter type"); \
+ temp = (void *)IX_OSAL_SWAP_BE_SHARED_LONG(src); \
+ (ptrDst) = (type)IX_OSAL_MBUF_MBUF_PHYSICAL_TO_VIRTUAL_TRANSLATION(temp); } \
+ while(0)
+#else
+#define PTR_VIRT2NPE(ptrSrc,dst) \
+ do { UINT32 temp; \
+ IX_OSAL_ENSURE(sizeof(ptrSrc) == sizeof(UINT32), "Wrong parameter type"); \
+ IX_OSAL_ENSURE(sizeof(dst) == sizeof(UINT32), "Wrong parameter type"); \
+ temp = (UINT32)IX_OSAL_MMU_VIRT_TO_PHYS(ptrSrc); \
+ (dst) = IX_OSAL_SWAP_BE_SHARED_LONG(temp); } \
+ while(0)
+
+#define PTR_NPE2VIRT(type,src,ptrDst) \
+ do { void *temp; \
+ IX_OSAL_ENSURE(sizeof(type) == sizeof(UINT32), "Wrong parameter type"); \
+ IX_OSAL_ENSURE(sizeof(src) == sizeof(UINT32), "Wrong parameter type"); \
+ IX_OSAL_ENSURE(sizeof(ptrDst) == sizeof(UINT32), "Wrong parameter type"); \
+ temp = (void *)IX_OSAL_SWAP_BE_SHARED_LONG(src); \
+ (ptrDst) = (type)IX_OSAL_MMU_PHYS_TO_VIRT(temp); } \
+ while(0)
+#endif
+
+/**
+ *
+ * @brief Mbuf payload pointer conversion macros : Wince has its own
+ * method to convert the buffer pointers
+ */
+#if defined(__wince) && !defined(IN_KERNEL)
+#define DATAPTR_VIRT2NPE(ptrSrc,dst) \
+ do { UINT32 temp; \
+ temp = (UINT32)IX_OSAL_MBUF_DATA_VIRTUAL_TO_PHYSICAL_TRANSLATION(ptrSrc); \
+ (dst) = IX_OSAL_SWAP_BE_SHARED_LONG(temp); } \
+ while(0)
+
+#else
+#define DATAPTR_VIRT2NPE(ptrSrc,dst) PTR_VIRT2NPE(IX_OSAL_MBUF_MDATA(ptrSrc),dst)
+#endif
+
+
+/* Flush the shared part of the mbuf header */
+#define IX_ETHACC_NE_CACHE_FLUSH(mbufPtr) \
+ do { \
+ IX_OSAL_CACHE_FLUSH(IX_ETHACC_NE_SHARED(mbufPtr), \
+ sizeof(IxEthAccNe)); \
+ } \
+ while(0)
+
+/* Invalidate the shared part of the mbuf header */
+#define IX_ETHACC_NE_CACHE_INVALIDATE(mbufPtr) \
+ do { \
+ IX_OSAL_CACHE_INVALIDATE(IX_ETHACC_NE_SHARED(mbufPtr), \
+ sizeof(IxEthAccNe)); \
+ } \
+ while(0)
+
+/* Preload one cache line (shared mbuf headers are aligned
+ * and their size is 1 cache line)
+ *
+ * IX_OSAL_CACHED is defined when the mbuf headers are
+ * allocated from cached memory.
+ *
+ * Other processor on emulation environment may not implement
+ * preload function
+ */
+#ifdef IX_OSAL_CACHED
+ #if (CPU!=SIMSPARCSOLARIS) && !defined (__wince)
+ #define IX_ACC_DATA_CACHE_PRELOAD(ptr) \
+ do { /* preload a cache line (Xscale Processor) */ \
+ __asm__ (" pld [%0]\n": : "r" (ptr)); \
+ } \
+ while(0)
+ #else
+ /* preload not implemented on different processor */
+ #define IX_ACC_DATA_CACHE_PRELOAD(mbufPtr) \
+ do { /* nothing */ } while (0)
+ #endif
+#else
+ /* preload not needed if cache is not enabled */
+ #define IX_ACC_DATA_CACHE_PRELOAD(mbufPtr) \
+ do { /* nothing */ } while (0)
+#endif
+
+/**
+ *
+ * @brief function to retrieve the correct pointer from
+ * a queue entry posted by the NPE
+ *
+ * @param qEntry : entry from qmgr queue
+ * mask : applicable mask for this queue
+ * (4 most significant bits are used for additional informations)
+ *
+ * @return IX_OSAL_MBUF * pointer to mbuf header
+ *
+ * @internal
+ */
+PRIVATE IX_OSAL_MBUF *
+ixEthAccEntryFromQConvert(UINT32 qEntry, UINT32 mask)
+{
+ IX_OSAL_MBUF *mbufPtr;
+
+ if (qEntry != 0)
+ {
+ /* mask NPE bits (e.g. priority, port ...) */
+ qEntry &= mask;
+
+#if IX_ACC_DRAM_PHYS_OFFSET != 0
+ /* restore the original address pointer (if PHYS_OFFSET is not 0) */
+ qEntry |= (IX_ACC_DRAM_PHYS_OFFSET & ~IX_ETHNPE_QM_Q_RXENET_ADDR_MASK);
+#endif
+ /* get the mbuf pointer address from the npe-shared address */
+ qEntry -= offsetof(IX_OSAL_MBUF,ix_ne);
+
+ /* phys2virt mbuf */
+ mbufPtr = (IX_OSAL_MBUF *)IX_OSAL_MMU_PHYS_TO_VIRT(qEntry);
+
+ /* preload the cacheline shared with NPE */
+ IX_ACC_DATA_CACHE_PRELOAD(IX_ETHACC_NE_SHARED(mbufPtr));
+
+ /* preload the cacheline used by xscale */
+ IX_ACC_DATA_CACHE_PRELOAD(mbufPtr);
+ }
+ else
+ {
+ mbufPtr = NULL;
+ }
+
+ return mbufPtr;
+}
+
+/* Convert the mbuf header for NPE transmission */
+PRIVATE UINT32
+ixEthAccMbufTxQPrepare(IX_OSAL_MBUF *mbuf)
+{
+ UINT32 qbuf;
+ UINT32 len;
+
+ /* endianess swap for tci and flags
+ note: this is done only once, even for chained buffers */
+ IX_ETHACC_NE_FLAGS(mbuf) = IX_OSAL_SWAP_BE_SHARED_SHORT(IX_ETHACC_NE_FLAGS(mbuf));
+ IX_ETHACC_NE_VLANTCI(mbuf) = IX_OSAL_SWAP_BE_SHARED_SHORT(IX_ETHACC_NE_VLANTCI(mbuf));
+
+ /* test for unchained mbufs */
+ if (IX_OSAL_MBUF_NEXT_BUFFER_IN_PKT_PTR(mbuf) == NULL)
+ {
+ /* "best case" scenario : unchained mbufs */
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.unchainedTxMBufs);
+
+ /* payload pointer conversion */
+ DATAPTR_VIRT2NPE(mbuf, IX_ETHACC_NE_DATA(mbuf));
+
+ /* unchained mbufs : the frame length is the mbuf length
+ * and the 2 identical lengths are stored in the same
+ * word.
+ */
+ len = IX_OSAL_MBUF_MLEN(mbuf);
+
+ /* set the length in both length and pktLen 16-bits fields */
+ len |= (len << IX_ETHNPE_ACC_LENGTH_OFFSET);
+ IX_ETHACC_NE_LEN(mbuf) = IX_OSAL_SWAP_BE_SHARED_LONG(len);
+
+ /* unchained mbufs : next contains 0 */
+ IX_ETHACC_NE_NEXT(mbuf) = 0;
+
+ /* flush shared header after all address conversions */
+ IX_ETHACC_NE_CACHE_FLUSH(mbuf);
+ }
+ else
+ {
+ /* chained mbufs */
+ IX_OSAL_MBUF *ptr = mbuf;
+ IX_OSAL_MBUF *nextPtr;
+ UINT32 frmLen;
+
+ /* get the frame length from the header of the first buffer */
+ frmLen = IX_OSAL_MBUF_PKT_LEN(mbuf);
+
+ do
+ {
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.chainedTxMBufs);
+
+ /* payload pointer */
+ DATAPTR_VIRT2NPE(ptr,IX_ETHACC_NE_DATA(ptr));
+ /* Buffer length and frame length are stored in the same word */
+ len = IX_OSAL_MBUF_MLEN(ptr);
+ len = frmLen | (len << IX_ETHNPE_ACC_LENGTH_OFFSET);
+ IX_ETHACC_NE_LEN(ptr) = IX_OSAL_SWAP_BE_SHARED_LONG(len);
+
+ /* get the virtual next chain pointer */
+ nextPtr = IX_OSAL_MBUF_NEXT_BUFFER_IN_PKT_PTR(ptr);
+ if (nextPtr != NULL)
+ {
+ /* shared pointer of the next buffer is chained */
+ PTR_VIRT2NPE(IX_ETHACC_NE_SHARED(nextPtr),
+ IX_ETHACC_NE_NEXT(ptr));
+ }
+ else
+ {
+ IX_ETHACC_NE_NEXT(ptr) = 0;
+ }
+
+ /* flush shared header after all address conversions */
+ IX_ETHACC_NE_CACHE_FLUSH(ptr);
+
+ /* move to next buffer */
+ ptr = nextPtr;
+
+ /* the frame length field is set only in the first buffer
+ * and is zeroed in the next buffers
+ */
+ frmLen = 0;
+ }
+ while(ptr != NULL);
+
+ }
+
+ /* virt2phys mbuf itself */
+ qbuf = (UINT32)IX_OSAL_MMU_VIRT_TO_PHYS(
+ IX_ETHACC_NE_SHARED(mbuf));
+
+ /* Ensure the bits which are reserved to exchange information with
+ * the NPE are cleared
+ *
+ * If the mbuf address is not correctly aligned, or from an
+ * incompatible memory range, there is no point to continue
+ */
+ IX_OSAL_ENSURE(((qbuf & ~IX_ETHNPE_QM_Q_TXENET_ADDR_MASK) == 0),
+ "Invalid address range");
+
+ return qbuf;
+}
+
+/* Convert the mbuf header for NPE reception */
+PRIVATE UINT32
+ixEthAccMbufRxQPrepare(IX_OSAL_MBUF *mbuf)
+{
+ UINT32 len;
+ UINT32 qbuf;
+
+ if (IX_OSAL_MBUF_NEXT_BUFFER_IN_PKT_PTR(mbuf) == NULL)
+ {
+ /* "best case" scenario : unchained mbufs */
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.unchainedRxFreeMBufs);
+
+ /* unchained mbufs : payload pointer */
+ DATAPTR_VIRT2NPE(mbuf, IX_ETHACC_NE_DATA(mbuf));
+
+ /* unchained mbufs : set the buffer length
+ * and the frame length field is zeroed
+ */
+ len = (IX_OSAL_MBUF_MLEN(mbuf) << IX_ETHNPE_ACC_LENGTH_OFFSET);
+ IX_ETHACC_NE_LEN(mbuf) = IX_OSAL_SWAP_BE_SHARED_LONG(len);
+
+ /* unchained mbufs : next pointer is null */
+ IX_ETHACC_NE_NEXT(mbuf) = 0;
+
+ /* flush shared header after all address conversions */
+ IX_ETHACC_NE_CACHE_FLUSH(mbuf);
+
+ /* remove shared header cache line */
+ IX_ETHACC_NE_CACHE_INVALIDATE(mbuf);
+ }
+ else
+ {
+ /* chained mbufs */
+ IX_OSAL_MBUF *ptr = mbuf;
+ IX_OSAL_MBUF *nextPtr;
+
+ do
+ {
+ /* chained mbufs */
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.chainedRxFreeMBufs);
+
+ /* we must save virtual next chain pointer */
+ nextPtr = IX_OSAL_MBUF_NEXT_BUFFER_IN_PKT_PTR(ptr);
+
+ if (nextPtr != NULL)
+ {
+ /* chaining pointer for NPE */
+ PTR_VIRT2NPE(IX_ETHACC_NE_SHARED(nextPtr),
+ IX_ETHACC_NE_NEXT(ptr));
+ }
+ else
+ {
+ IX_ETHACC_NE_NEXT(ptr) = 0;
+ }
+
+ /* payload pointer */
+ DATAPTR_VIRT2NPE(ptr,IX_ETHACC_NE_DATA(ptr));
+
+ /* buffer length */
+ len = (IX_OSAL_MBUF_MLEN(ptr) << IX_ETHNPE_ACC_LENGTH_OFFSET);
+ IX_ETHACC_NE_LEN(ptr) = IX_OSAL_SWAP_BE_SHARED_LONG(len);
+
+ /* flush shared header after all address conversions */
+ IX_ETHACC_NE_CACHE_FLUSH(ptr);
+
+ /* remove shared header cache line */
+ IX_ETHACC_NE_CACHE_INVALIDATE(ptr);
+
+ /* next mbuf in the chain */
+ ptr = nextPtr;
+ }
+ while(ptr != NULL);
+ }
+
+ /* virt2phys mbuf itself */
+ qbuf = (UINT32)IX_OSAL_MMU_VIRT_TO_PHYS(
+ IX_ETHACC_NE_SHARED(mbuf));
+
+ /* Ensure the bits which are reserved to exchange information with
+ * the NPE are cleared
+ *
+ * If the mbuf address is not correctly aligned, or from an
+ * incompatible memory range, there is no point to continue
+ */
+ IX_OSAL_ENSURE(((qbuf & ~IX_ETHNPE_QM_Q_RXENET_ADDR_MASK) == 0),
+ "Invalid address range");
+
+ return qbuf;
+}
+
+/* Convert the mbuf header after NPE transmission
+ * Since there is nothing changed by the NPE, there is no need
+ * to process anything but the update of internal stats
+ * when they are enabled
+*/
+PRIVATE void
+ixEthAccMbufFromTxQ(IX_OSAL_MBUF *mbuf)
+{
+#ifndef NDEBUG
+ /* test for unchained mbufs */
+ if (IX_ETHACC_NE_NEXT(mbuf) == 0)
+ {
+ /* unchained mbufs : update the stats */
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.unchainedTxDoneMBufs);
+ }
+ else
+ {
+ /* chained mbufs : walk the chain and update the stats */
+ IX_OSAL_MBUF *ptr = mbuf;
+
+ do
+ {
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.chainedTxDoneMBufs);
+ ptr = IX_OSAL_MBUF_NEXT_BUFFER_IN_PKT_PTR(ptr);
+ }
+ while (ptr != NULL);
+ }
+#endif
+}
+
+/* Convert the mbuf header after NPE reception */
+PRIVATE void
+ixEthAccMbufFromRxQ(IX_OSAL_MBUF *mbuf)
+{
+ UINT32 len;
+
+ /* endianess swap for tci and flags
+ note: this is done only once, even for chained buffers */
+ IX_ETHACC_NE_FLAGS(mbuf) = IX_OSAL_SWAP_BE_SHARED_SHORT(IX_ETHACC_NE_FLAGS(mbuf));
+ IX_ETHACC_NE_VLANTCI(mbuf) = IX_OSAL_SWAP_BE_SHARED_SHORT(IX_ETHACC_NE_VLANTCI(mbuf));
+
+ /* test for unchained mbufs */
+ if (IX_ETHACC_NE_NEXT(mbuf) == 0)
+ {
+ /* unchained mbufs */
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.unchainedRxMBufs);
+
+ /* get the frame length. it is the same than the buffer length */
+ len = IX_OSAL_SWAP_BE_SHARED_LONG(IX_ETHACC_NE_LEN(mbuf));
+ len &= IX_ETHNPE_ACC_PKTLENGTH_MASK;
+ IX_OSAL_MBUF_PKT_LEN(mbuf) = IX_OSAL_MBUF_MLEN(mbuf) = len;
+
+ /* clears the next packet field */
+ IX_OSAL_MBUF_NEXT_BUFFER_IN_PKT_PTR(mbuf) = NULL;
+ }
+ else
+ {
+ IX_OSAL_MBUF *ptr = mbuf;
+ IX_OSAL_MBUF *nextPtr;
+ UINT32 frmLen;
+
+ /* convert the frame length */
+ frmLen = IX_OSAL_SWAP_BE_SHARED_LONG(IX_ETHACC_NE_LEN(mbuf));
+ IX_OSAL_MBUF_PKT_LEN(mbuf) = (frmLen & IX_ETHNPE_ACC_PKTLENGTH_MASK);
+
+ /* chained mbufs */
+ do
+ {
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.chainedRxMBufs);
+
+ /* convert the length */
+ len = IX_OSAL_SWAP_BE_SHARED_LONG(IX_ETHACC_NE_LEN(ptr));
+ IX_OSAL_MBUF_MLEN(ptr) = (len >> IX_ETHNPE_ACC_LENGTH_OFFSET);
+
+ /* get the next pointer */
+ PTR_NPE2VIRT(IX_OSAL_MBUF *,IX_ETHACC_NE_NEXT(ptr), nextPtr);
+ if (nextPtr != NULL)
+ {
+ nextPtr = (IX_OSAL_MBUF *)((UINT8 *)nextPtr - offsetof(IX_OSAL_MBUF,ix_ne));
+ }
+ /* set the next pointer */
+ IX_OSAL_MBUF_NEXT_BUFFER_IN_PKT_PTR(ptr) = nextPtr;
+
+ /* move to the next buffer */
+ ptr = nextPtr;
+ }
+ while (ptr != NULL);
+ }
+}
+
+/* write to qmgr if possible and report an overflow if not possible
+ * Use a fast lock to protect the queue write.
+ * This way, the tx feature is reentrant.
+ */
+PRIVATE IX_STATUS
+ixEthAccQmgrLockTxWrite(IxEthAccPortId portId, UINT32 qBuffer)
+{
+ IX_STATUS qStatus;
+ if (ixOsalFastMutexTryLock(&txWriteMutex[portId]) == IX_SUCCESS)
+ {
+ qStatus = ixQMgrQWrite(
+ IX_ETH_ACC_PORT_TO_TX_Q_ID(portId),
+ &qBuffer);
+#ifndef NDEBUG
+ if (qStatus != IX_SUCCESS)
+ {
+ TX_STATS_INC(portId, txOverflow);
+ }
+#endif
+ ixOsalFastMutexUnlock(&txWriteMutex[portId]);
+ }
+ else
+ {
+ TX_STATS_INC(portId, txLock);
+ qStatus = IX_QMGR_Q_OVERFLOW;
+ }
+ return qStatus;
+}
+
+/* write to qmgr if possible and report an overflow if not possible
+ * Use a fast lock to protect the queue write.
+ * This way, the Rx feature is reentrant.
+ */
+PRIVATE IX_STATUS
+ixEthAccQmgrLockRxWrite(IxEthAccPortId portId, UINT32 qBuffer)
+{
+ IX_STATUS qStatus;
+ if (ixOsalFastMutexTryLock(&rxWriteMutex[portId]) == IX_SUCCESS)
+ {
+ qStatus = ixQMgrQWrite(
+ IX_ETH_ACC_PORT_TO_RX_FREE_Q_ID(portId),
+ &qBuffer);
+#ifndef NDEBUG
+ if (qStatus != IX_SUCCESS)
+ {
+ RX_STATS_INC(portId, rxFreeOverflow);
+ }
+#endif
+ ixOsalFastMutexUnlock(&rxWriteMutex[portId]);
+ }
+ else
+ {
+ RX_STATS_INC(portId, rxFreeLock);
+ qStatus = IX_QMGR_Q_OVERFLOW;
+ }
+ return qStatus;
+}
+
+/*
+ * Set the priority and write to a qmgr queue.
+ */
+PRIVATE IX_STATUS
+ixEthAccQmgrTxWrite(IxEthAccPortId portId, UINT32 qBuffer, UINT32 priority)
+{
+ /* fill the priority field */
+ qBuffer |= (priority << IX_ETHNPE_QM_Q_FIELD_PRIOR_R);
+
+ return ixEthAccQmgrLockTxWrite(portId, qBuffer);
+}
+
+/**
+ *
+ * @brief This function will discover the highest priority S/W Tx Q that
+ * has entries in it
+ *
+ * @param portId - (in) the id of the port whose S/W Tx queues are to be searched
+ * priorityPtr - (out) the priority of the highest priority occupied q will be written
+ * here
+ *
+ * @return IX_ETH_ACC_SUCCESS if an occupied Q is found
+ * IX_ETH_ACC_FAIL if no Q has entries
+ *
+ * @internal
+ */
+PRIVATE IxEthAccStatus
+ixEthAccTxSwQHighestPriorityGet(IxEthAccPortId portId,
+ IxEthAccTxPriority *priorityPtr)
+{
+ if (ixEthAccPortData[portId].ixEthAccTxData.schDiscipline
+ == FIFO_NO_PRIORITY)
+ {
+ if(IX_ETH_ACC_DATAPLANE_IS_Q_EMPTY(ixEthAccPortData[portId].
+ ixEthAccTxData.txQ[IX_ETH_ACC_TX_DEFAULT_PRIORITY]))
+ {
+ return IX_ETH_ACC_FAIL;
+ }
+ else
+ {
+ *priorityPtr = IX_ETH_ACC_TX_DEFAULT_PRIORITY;
+ TX_STATS_INC(portId,txPriority[*priorityPtr]);
+ return IX_ETH_ACC_SUCCESS;
+ }
+ }
+ else
+ {
+ IxEthAccTxPriority highestPriority = IX_ETH_ACC_TX_PRIORITY_7;
+ while(1)
+ {
+ if(!IX_ETH_ACC_DATAPLANE_IS_Q_EMPTY(ixEthAccPortData[portId].
+ ixEthAccTxData.txQ[highestPriority]))
+ {
+
+ *priorityPtr = highestPriority;
+ TX_STATS_INC(portId,txPriority[highestPriority]);
+ return IX_ETH_ACC_SUCCESS;
+
+ }
+ if (highestPriority == IX_ETH_ACC_TX_PRIORITY_0)
+ {
+ return IX_ETH_ACC_FAIL;
+ }
+ highestPriority--;
+ }
+ }
+}
+
+/**
+ *
+ * @brief This function will take a buffer from a TX S/W Q and attempt
+ * to add it to the relevant TX H/W Q
+ *
+ * @param portId - the port whose TX queue is to be written to
+ * priority - identifies the queue from which the entry is to be read
+ *
+ * @internal
+ */
+PRIVATE IxEthAccStatus
+ixEthAccTxFromSwQ(IxEthAccPortId portId,
+ IxEthAccTxPriority priority)
+{
+ IX_OSAL_MBUF *mbuf;
+ IX_STATUS qStatus;
+
+ IX_OSAL_ENSURE((UINT32)priority <= (UINT32)7, "Invalid priority");
+
+ IX_ETH_ACC_DATAPLANE_REMOVE_MBUF_FROM_Q_HEAD(
+ ixEthAccPortData[portId].ixEthAccTxData.txQ[priority],
+ mbuf);
+
+ if (mbuf != NULL)
+ {
+ /*
+ * Add the Tx buffer to the H/W Tx Q
+ * We do not need to flush here as it is already done
+ * in TxFrameSubmit().
+ */
+ qStatus = ixEthAccQmgrTxWrite(
+ portId,
+ IX_OSAL_MMU_VIRT_TO_PHYS((UINT32)IX_ETHACC_NE_SHARED(mbuf)),
+ priority);
+
+ if (qStatus == IX_SUCCESS)
+ {
+ TX_STATS_INC(portId,txFromSwQOK);
+ return IX_SUCCESS;
+ }
+ else if (qStatus == IX_QMGR_Q_OVERFLOW)
+ {
+ /*
+ * H/W Q overflow, need to save the buffer
+ * back on the s/w Q.
+ * we must put it back on the head of the q to avoid
+ * reordering packet tx
+ */
+ TX_STATS_INC(portId,txFromSwQDelayed);
+ IX_ETH_ACC_DATAPLANE_ADD_MBUF_TO_Q_HEAD(
+ ixEthAccPortData[portId].ixEthAccTxData.txQ[priority],
+ mbuf);
+
+ /*enable Q notification*/
+ qStatus = ixQMgrNotificationEnable(
+ IX_ETH_ACC_PORT_TO_TX_Q_ID(portId),
+ IX_ETH_ACC_PORT_TO_TX_Q_SOURCE(portId));
+
+ if (qStatus != IX_SUCCESS && qStatus != IX_QMGR_WARNING)
+ {
+ TX_INC(portId,txUnexpectedError);
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthAccTxFromSwQ:Unexpected Error: %u\n",
+ qStatus, 0, 0, 0, 0, 0);
+ }
+ }
+ else
+ {
+ TX_INC(portId,txUnexpectedError);
+
+ /* recovery attempt */
+ IX_ETH_ACC_DATAPLANE_ADD_MBUF_TO_Q_HEAD(
+ ixEthAccPortData[portId].ixEthAccTxData.txQ[priority],
+ mbuf);
+
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthAccTxFromSwQ:Error: unexpected QM status 0x%08X\n",
+ qStatus, 0, 0, 0, 0, 0);
+ }
+ }
+ else
+ {
+ /* sw queue is empty */
+ }
+ return IX_ETH_ACC_FAIL;
+}
+
+/**
+ *
+ * @brief This function will take a buffer from a RXfree S/W Q and attempt
+ * to add it to the relevant RxFree H/W Q
+ *
+ * @param portId - the port whose RXFree queue is to be written to
+ *
+ * @internal
+ */
+PRIVATE IxEthAccStatus
+ixEthAccRxFreeFromSwQ(IxEthAccPortId portId)
+{
+ IX_OSAL_MBUF *mbuf;
+ IX_STATUS qStatus = IX_SUCCESS;
+
+ IX_ETH_ACC_DATAPLANE_REMOVE_MBUF_FROM_Q_HEAD(
+ ixEthAccPortData[portId].ixEthAccRxData.freeBufferList,
+ mbuf);
+ if (mbuf != NULL)
+ {
+ /*
+ * Add The Rx Buffer to the H/W Free buffer Q if possible
+ */
+ qStatus = ixEthAccQmgrLockRxWrite(portId,
+ IX_OSAL_MMU_VIRT_TO_PHYS(
+ (UINT32)IX_ETHACC_NE_SHARED(mbuf)));
+
+ if (qStatus == IX_SUCCESS)
+ {
+ RX_STATS_INC(portId,rxFreeRepFromSwQOK);
+ /*
+ * Buffer added to h/w Q.
+ */
+ return IX_SUCCESS;
+ }
+ else if (qStatus == IX_QMGR_Q_OVERFLOW)
+ {
+ /*
+ * H/W Q overflow, need to save the buffer back on the s/w Q.
+ */
+ RX_STATS_INC(portId,rxFreeRepFromSwQDelayed);
+
+ IX_ETH_ACC_DATAPLANE_ADD_MBUF_TO_Q_HEAD(
+ ixEthAccPortData[portId].ixEthAccRxData.freeBufferList,
+ mbuf);
+ }
+ else
+ {
+ /* unexpected qmgr error */
+ RX_INC(portId,rxUnexpectedError);
+
+ IX_ETH_ACC_DATAPLANE_ADD_MBUF_TO_Q_HEAD(
+ ixEthAccPortData[portId].ixEthAccRxData.freeBufferList,
+ mbuf);
+
+ IX_ETH_ACC_FATAL_LOG("IxEthAccRxFreeFromSwQ:Error: unexpected QM status 0x%08X\n",
+ qStatus, 0, 0, 0, 0, 0);
+ }
+ }
+ else
+ {
+ /* sw queue is empty */
+ }
+ return IX_ETH_ACC_FAIL;
+}
+
+
+IX_ETH_ACC_PUBLIC
+IxEthAccStatus ixEthAccInitDataPlane()
+{
+ UINT32 portId;
+
+ /*
+ * Initialize the service and register callback to other services.
+ */
+
+ IX_ETH_ACC_MEMSET(&ixEthAccDataStats,
+ 0,
+ sizeof(ixEthAccDataStats));
+
+ for(portId=0; portId < IX_ETH_ACC_NUMBER_OF_PORTS; portId++)
+ {
+ ixOsalFastMutexInit(&txWriteMutex[portId]);
+ ixOsalFastMutexInit(&rxWriteMutex[portId]);
+
+ IX_ETH_ACC_MEMSET(&ixEthAccPortData[portId],
+ 0,
+ sizeof(ixEthAccPortData[portId]));
+
+ ixEthAccPortData[portId].ixEthAccTxData.schDiscipline = FIFO_NO_PRIORITY;
+ }
+
+ return (IX_ETH_ACC_SUCCESS);
+}
+
+
+IX_ETH_ACC_PUBLIC
+IxEthAccStatus ixEthAccPortTxDoneCallbackRegister(IxEthAccPortId portId,
+ IxEthAccPortTxDoneCallback
+ txCallbackFn,
+ UINT32 callbackTag)
+{
+ if (!IX_ETH_ACC_IS_SERVICE_INITIALIZED())
+ {
+ return (IX_ETH_ACC_FAIL);
+ }
+ if (!IX_ETH_ACC_IS_PORT_VALID(portId))
+ {
+ return (IX_ETH_ACC_INVALID_PORT);
+ }
+
+/* HACK: removing this code to enable NPE-A preliminary testing
+ * if (IX_ETH_ACC_SUCCESS != ixEthAccSingleEthNpeCheck(portId))
+ * {
+ * IX_ETH_ACC_WARNING_LOG("ixEthAccPortTxDoneCallbackRegister: Unavailable Eth %d: Cannot register TxDone Callback.\n",(INT32)portId,0,0,0,0,0);
+ * return IX_ETH_ACC_SUCCESS ;
+ * }
+ */
+
+ if (!IX_ETH_IS_PORT_INITIALIZED(portId))
+ {
+ return (IX_ETH_ACC_PORT_UNINITIALIZED);
+ }
+ if (txCallbackFn == 0)
+ /* Check for null function pointer here. */
+ {
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+ ixEthAccPortData[portId].ixEthAccTxData.txBufferDoneCallbackFn = txCallbackFn;
+ ixEthAccPortData[portId].ixEthAccTxData.txCallbackTag = callbackTag;
+ return (IX_ETH_ACC_SUCCESS);
+}
+
+
+IX_ETH_ACC_PUBLIC
+IxEthAccStatus ixEthAccPortRxCallbackRegister(IxEthAccPortId portId,
+ IxEthAccPortRxCallback
+ rxCallbackFn,
+ UINT32 callbackTag)
+{
+ IxEthAccPortId port;
+
+ if (!IX_ETH_ACC_IS_SERVICE_INITIALIZED())
+ {
+ return (IX_ETH_ACC_FAIL);
+ }
+ if (!IX_ETH_ACC_IS_PORT_VALID(portId))
+ {
+ return (IX_ETH_ACC_INVALID_PORT);
+ }
+
+ if (IX_ETH_ACC_SUCCESS != ixEthAccSingleEthNpeCheck(portId))
+ {
+ IX_ETH_ACC_WARNING_LOG("ixEthAccPortRxCallbackRegister: Unavailable Eth %d: Cannot register Rx Callback.\n",(INT32)portId,0,0,0,0,0);
+ return IX_ETH_ACC_SUCCESS ;
+ }
+
+ if (!IX_ETH_IS_PORT_INITIALIZED(portId))
+ {
+ return (IX_ETH_ACC_PORT_UNINITIALIZED);
+ }
+
+ /* Check for null function pointer here. */
+ if (rxCallbackFn == NULL)
+ {
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+
+ /* Check the user is not changing the callback type
+ * when the port is enabled.
+ */
+ if (ixEthAccMacState[portId].portDisableState == ACTIVE)
+ {
+ for (port = 0; port < IX_ETH_ACC_NUMBER_OF_PORTS; port++)
+ {
+ if ((ixEthAccMacState[port].portDisableState == ACTIVE)
+ && (ixEthAccPortData[port].ixEthAccRxData.rxMultiBufferCallbackInUse == TRUE))
+ {
+ /* one of the active ports has a different rx callback type.
+ * Changing the callback type when the port is enabled
+ * is not safe
+ */
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+ }
+ }
+
+ /* update the callback pointer : this is done before
+ * registering the new qmgr callback
+ */
+ ixEthAccPortData[portId].ixEthAccRxData.rxCallbackFn = rxCallbackFn;
+ ixEthAccPortData[portId].ixEthAccRxData.rxCallbackTag = callbackTag;
+
+ /* update the qmgr callback for rx queues */
+ if (ixEthAccQMgrRxCallbacksRegister(ixEthRxFrameQMCallback)
+ != IX_ETH_ACC_SUCCESS)
+ {
+ /* unexpected qmgr error */
+ IX_ETH_ACC_FATAL_LOG("ixEthAccPortRxCallbackRegister: unexpected QMgr error, " \
+ "could not register Rx single-buffer callback\n", 0, 0, 0, 0, 0, 0);
+
+ RX_INC(portId,rxUnexpectedError);
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+
+ ixEthAccPortData[portId].ixEthAccRxData.rxMultiBufferCallbackInUse = FALSE;
+
+ return (IX_ETH_ACC_SUCCESS);
+}
+
+IX_ETH_ACC_PUBLIC
+IxEthAccStatus ixEthAccPortMultiBufferRxCallbackRegister(
+ IxEthAccPortId portId,
+ IxEthAccPortMultiBufferRxCallback
+ rxCallbackFn,
+ UINT32 callbackTag)
+{
+ IxEthAccPortId port;
+
+ if (!IX_ETH_ACC_IS_SERVICE_INITIALIZED())
+ {
+ return (IX_ETH_ACC_FAIL);
+ }
+ if (!IX_ETH_ACC_IS_PORT_VALID(portId))
+ {
+ return (IX_ETH_ACC_INVALID_PORT);
+ }
+
+ if (IX_ETH_ACC_SUCCESS != ixEthAccSingleEthNpeCheck(portId))
+ {
+ IX_ETH_ACC_WARNING_LOG("ixEthAccPortMultiBufferRxCallbackRegister: Unavailable Eth %d: Cannot register Rx Callback.\n",(INT32)portId,0,0,0,0,0);
+ return IX_ETH_ACC_SUCCESS ;
+ }
+
+ if (!IX_ETH_IS_PORT_INITIALIZED(portId))
+ {
+ return (IX_ETH_ACC_PORT_UNINITIALIZED);
+ }
+
+ /* Check for null function pointer here. */
+ if (rxCallbackFn == NULL)
+ {
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+
+ /* Check the user is not changing the callback type
+ * when the port is enabled.
+ */
+ if (ixEthAccMacState[portId].portDisableState == ACTIVE)
+ {
+ for (port = 0; port < IX_ETH_ACC_NUMBER_OF_PORTS; port++)
+ {
+ if ((ixEthAccMacState[port].portDisableState == ACTIVE)
+ && (ixEthAccPortData[port].ixEthAccRxData.rxMultiBufferCallbackInUse == FALSE))
+ {
+ /* one of the active ports has a different rx callback type.
+ * Changing the callback type when the port is enabled
+ * is not safe
+ */
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+ }
+ }
+
+ /* update the callback pointer : this is done before
+ * registering the new qmgr callback
+ */
+ ixEthAccPortData[portId].ixEthAccRxData.rxMultiBufferCallbackFn = rxCallbackFn;
+ ixEthAccPortData[portId].ixEthAccRxData.rxMultiBufferCallbackTag = callbackTag;
+
+ /* update the qmgr callback for rx queues */
+ if (ixEthAccQMgrRxCallbacksRegister(ixEthRxMultiBufferQMCallback)
+ != IX_ETH_ACC_SUCCESS)
+ {
+ /* unexpected qmgr error */
+ RX_INC(portId,rxUnexpectedError);
+
+ IX_ETH_ACC_FATAL_LOG("ixEthAccPortMultiBufferRxCallbackRegister: unexpected QMgr error, " \
+ "could not register Rx multi-buffer callback\n", 0, 0, 0, 0, 0, 0);
+
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+
+ ixEthAccPortData[portId].ixEthAccRxData.rxMultiBufferCallbackInUse = TRUE;
+
+ return (IX_ETH_ACC_SUCCESS);
+}
+
+IX_ETH_ACC_PUBLIC
+IxEthAccStatus ixEthAccPortTxFrameSubmit(IxEthAccPortId portId,
+ IX_OSAL_MBUF *buffer,
+ IxEthAccTxPriority priority)
+{
+ IX_STATUS qStatus = IX_SUCCESS;
+ UINT32 qBuffer;
+ IxEthAccTxPriority highestPriority;
+ IxQMgrQStatus txQStatus;
+
+#ifndef NDEBUG
+ if (buffer == NULL)
+ {
+ return (IX_ETH_ACC_FAIL);
+ }
+ if (!IX_ETH_ACC_IS_SERVICE_INITIALIZED())
+ {
+ return (IX_ETH_ACC_FAIL);
+ }
+ if (!IX_ETH_ACC_IS_PORT_VALID(portId))
+ {
+ return (IX_ETH_ACC_INVALID_PORT);
+ }
+
+ if (IX_ETH_ACC_SUCCESS != ixEthAccSingleEthNpeCheck(portId))
+ {
+ IX_ETH_ACC_FATAL_LOG("ixEthAccPortTxFrameSubmit: Unavailable Eth %d: Cannot submit Tx Frame.\n",
+ (INT32)portId,0,0,0,0,0);
+ return IX_ETH_ACC_PORT_UNINITIALIZED ;
+ }
+
+ if (!IX_ETH_IS_PORT_INITIALIZED(portId))
+ {
+ return (IX_ETH_ACC_PORT_UNINITIALIZED);
+ }
+ if ((UINT32)priority > (UINT32)IX_ETH_ACC_TX_PRIORITY_7)
+ {
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+#endif
+
+ /*
+ * Need to Flush the MBUF and its contents (data) as it may be
+ * read from the NPE. Convert virtual addresses to physical addresses also.
+ */
+ qBuffer = ixEthAccMbufTxQPrepare(buffer);
+
+ /*
+ * If no fifo priority set on Xscale ...
+ */
+ if (ixEthAccPortData[portId].ixEthAccTxData.schDiscipline ==
+ FIFO_NO_PRIORITY)
+ {
+ /*
+ * Add The Tx Buffer to the H/W Tx Q if possible
+ * (the priority is passed to the NPE, because
+ * the NPE is able to reorder the frames
+ * before transmission to the underlying hardware)
+ */
+ qStatus = ixEthAccQmgrTxWrite(portId,
+ qBuffer,
+ IX_ETH_ACC_TX_DEFAULT_PRIORITY);
+
+ if (qStatus == IX_SUCCESS)
+ {
+ TX_STATS_INC(portId,txQOK);
+
+ /*
+ * "best case" scenario : Buffer added to h/w Q.
+ */
+ return (IX_SUCCESS);
+ }
+ else if (qStatus == IX_QMGR_Q_OVERFLOW)
+ {
+ /*
+ * We were unable to write the buffer to the
+ * appropriate H/W Q, Save it in the sw Q.
+ * (use the default priority queue regardless of
+ * input parameter)
+ */
+ priority = IX_ETH_ACC_TX_DEFAULT_PRIORITY;
+ }
+ else
+ {
+ /* unexpected qmgr error */
+ TX_INC(portId,txUnexpectedError);
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthAccPortTxFrameSubmit:Error: qStatus = %u\n",
+ (UINT32)qStatus, 0, 0, 0, 0, 0);
+ return (IX_ETH_ACC_FAIL);
+ }
+ }
+ else if (ixEthAccPortData[portId].ixEthAccTxData.schDiscipline ==
+ FIFO_PRIORITY)
+ {
+
+ /*
+ * For priority transmission, put the frame directly on the H/W queue
+ * if the H/W queue is empty, otherwise, put it in a S/W Q
+ */
+ ixQMgrQStatusGet(IX_ETH_ACC_PORT_TO_TX_Q_ID(portId), &txQStatus);
+ if((txQStatus & IX_QMGR_Q_STATUS_E_BIT_MASK) != 0)
+ {
+ /*The tx queue is empty, check whether there are buffers on the s/w queues*/
+ if(ixEthAccTxSwQHighestPriorityGet(portId, &highestPriority)
+ !=IX_ETH_ACC_FAIL)
+ {
+ /*there are buffers on the s/w queues, submit them*/
+ ixEthAccTxFromSwQ(portId, highestPriority);
+
+ /* the queue was empty, 1 buffer is already supplied
+ * but is likely to be immediately transmitted and the
+ * hw queue is likely to be empty again, so submit
+ * more from the sw queues
+ */
+ if(ixEthAccTxSwQHighestPriorityGet(portId, &highestPriority)
+ !=IX_ETH_ACC_FAIL)
+ {
+ ixEthAccTxFromSwQ(portId, highestPriority);
+ /*
+ * and force the buffer supplied to be placed
+ * on a priority queue
+ */
+ qStatus = IX_QMGR_Q_OVERFLOW;
+ }
+ else
+ {
+ /*there are no buffers in the s/w queues, submit directly*/
+ qStatus = ixEthAccQmgrTxWrite(portId, qBuffer, priority);
+ }
+ }
+ else
+ {
+ /*there are no buffers in the s/w queues, submit directly*/
+ qStatus = ixEthAccQmgrTxWrite(portId, qBuffer, priority);
+ }
+ }
+ else
+ {
+ qStatus = IX_QMGR_Q_OVERFLOW;
+ }
+ }
+ else
+ {
+ TX_INC(portId,txUnexpectedError);
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthAccPortTxFrameSubmit:Error: wrong schedule discipline setup\n",
+ 0, 0, 0, 0, 0, 0);
+ return (IX_ETH_ACC_FAIL);
+ }
+
+ if(qStatus == IX_SUCCESS )
+ {
+ TX_STATS_INC(portId,txQOK);
+ return IX_ETH_ACC_SUCCESS;
+ }
+ else if(qStatus == IX_QMGR_Q_OVERFLOW)
+ {
+ TX_STATS_INC(portId,txQDelayed);
+ /*
+ * We were unable to write the buffer to the
+ * appropriate H/W Q, Save it in a s/w Q.
+ */
+ IX_ETH_ACC_DATAPLANE_ADD_MBUF_TO_Q_TAIL(
+ ixEthAccPortData[portId].
+ ixEthAccTxData.txQ[priority],
+ buffer);
+
+ qStatus = ixQMgrNotificationEnable(
+ IX_ETH_ACC_PORT_TO_TX_Q_ID(portId),
+ IX_ETH_ACC_PORT_TO_TX_Q_SOURCE(portId));
+
+ if (qStatus != IX_SUCCESS)
+ {
+ if (qStatus == IX_QMGR_WARNING)
+ {
+ /* notification is enabled for a queue
+ * which is already empty (the condition is already met)
+ * and there will be no more queue event to drain the sw queue
+ */
+ TX_STATS_INC(portId,txLateNotificationEnabled);
+
+ /* pull a buffer from the sw queue */
+ if(ixEthAccTxSwQHighestPriorityGet(portId, &highestPriority)
+ !=IX_ETH_ACC_FAIL)
+ {
+ /*there are buffers on the s/w queues, submit from them*/
+ ixEthAccTxFromSwQ(portId, highestPriority);
+ }
+ }
+ else
+ {
+ TX_INC(portId,txUnexpectedError);
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthAccPortTxFrameSubmit: unexpected Error: %u\n",
+ qStatus, 0, 0, 0, 0, 0);
+ }
+ }
+ }
+ else
+ {
+ TX_INC(portId,txUnexpectedError);
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthAccPortTxFrameSubmit: unexpected Error: %u\n",
+ qStatus, 0, 0, 0, 0, 0);
+ return (IX_ETH_ACC_FAIL);
+ }
+
+ return (IX_ETH_ACC_SUCCESS);
+}
+
+
+/**
+ *
+ * @brief replenish: convert a chain of mbufs to the format
+ * expected by the NPE
+ *
+ */
+
+IX_ETH_ACC_PUBLIC
+IxEthAccStatus ixEthAccPortRxFreeReplenish(IxEthAccPortId portId,
+ IX_OSAL_MBUF *buffer)
+{
+ IX_STATUS qStatus = IX_SUCCESS;
+ UINT32 qBuffer;
+
+ /*
+ * Check buffer is valid.
+ */
+
+#ifndef NDEBUG
+ /* check parameter value */
+ if (buffer == 0)
+ {
+ return (IX_ETH_ACC_FAIL);
+ }
+ if (!IX_ETH_ACC_IS_SERVICE_INITIALIZED())
+ {
+ return (IX_ETH_ACC_FAIL);
+ }
+ if (!IX_ETH_ACC_IS_PORT_VALID(portId))
+ {
+ return (IX_ETH_ACC_INVALID_PORT);
+ }
+
+ /* check initialisation is done */
+ if (IX_ETH_ACC_SUCCESS != ixEthAccSingleEthNpeCheck(portId))
+ {
+ IX_ETH_ACC_FATAL_LOG(" ixEthAccPortRxFreeReplenish: Unavailable Eth %d: Cannot replenish Rx Free Q.\n",(INT32)portId,0,0,0,0,0);
+ return IX_ETH_ACC_PORT_UNINITIALIZED ;
+ }
+
+ if (!IX_ETH_IS_PORT_INITIALIZED(portId))
+ {
+ return (IX_ETH_ACC_PORT_UNINITIALIZED);
+ }
+ /* check boundaries and constraints */
+ if (IX_OSAL_MBUF_MLEN(buffer) < IX_ETHNPE_ACC_RXFREE_BUFFER_LENGTH_MIN)
+ {
+ return (IX_ETH_ACC_FAIL);
+ }
+#endif
+
+ qBuffer = ixEthAccMbufRxQPrepare(buffer);
+
+ /*
+ * Add The Rx Buffer to the H/W Free buffer Q if possible
+ */
+ qStatus = ixEthAccQmgrLockRxWrite(portId, qBuffer);
+
+ if (qStatus == IX_SUCCESS)
+ {
+ RX_STATS_INC(portId,rxFreeRepOK);
+ /*
+ * Buffer added to h/w Q.
+ */
+ return (IX_SUCCESS);
+ }
+ else if (qStatus == IX_QMGR_Q_OVERFLOW)
+ {
+ RX_STATS_INC(portId,rxFreeRepDelayed);
+ /*
+ * We were unable to write the buffer to the approprate H/W Q,
+ * Save it in a s/w Q.
+ */
+ IX_ETH_ACC_DATAPLANE_ADD_MBUF_TO_Q_TAIL(
+ ixEthAccPortData[portId].ixEthAccRxData.freeBufferList,
+ buffer);
+
+ qStatus = ixQMgrNotificationEnable(
+ IX_ETH_ACC_PORT_TO_RX_FREE_Q_ID(portId),
+ IX_ETH_ACC_PORT_TO_RX_FREE_Q_SOURCE(portId));
+
+ if (qStatus != IX_SUCCESS)
+ {
+ if (qStatus == IX_QMGR_WARNING)
+ {
+ /* notification is enabled for a queue
+ * which is already empty (the condition is already met)
+ * and there will be no more queue event to drain the sw queue
+ * move an entry from the sw queue to the hw queue */
+ RX_STATS_INC(portId,rxFreeLateNotificationEnabled);
+ ixEthAccRxFreeFromSwQ(portId);
+ }
+ else
+ {
+ RX_INC(portId,rxUnexpectedError);
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthAccRxPortFreeReplenish:Error: %u\n",
+ qStatus, 0, 0, 0, 0, 0);
+ }
+ }
+ }
+ else
+ {
+ RX_INC(portId,rxUnexpectedError);
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthAccRxPortFreeReplenish:Error: qStatus = %u\n",
+ (UINT32)qStatus, 0, 0, 0, 0, 0);
+ return(IX_ETH_ACC_FAIL);
+ }
+ return (IX_ETH_ACC_SUCCESS);
+}
+
+
+IX_ETH_ACC_PUBLIC
+IxEthAccStatus ixEthAccTxSchedulingDisciplineSetPriv(IxEthAccPortId portId,
+ IxEthAccSchedulerDiscipline
+ sched)
+{
+ if (!IX_ETH_ACC_IS_PORT_VALID(portId))
+ {
+ return (IX_ETH_ACC_INVALID_PORT);
+ }
+
+ if (IX_ETH_ACC_SUCCESS != ixEthAccSingleEthNpeCheck(portId))
+ {
+ IX_ETH_ACC_WARNING_LOG("ixEthAccTxSchedulingDisciplineSet: Unavailable Eth %d: Cannot set Tx Scheduling Discipline.\n",(INT32)portId,0,0,0,0,0);
+ return IX_ETH_ACC_SUCCESS ;
+ }
+
+ if (!IX_ETH_IS_PORT_INITIALIZED(portId))
+ {
+ return (IX_ETH_ACC_PORT_UNINITIALIZED);
+ }
+
+ if (sched != FIFO_PRIORITY && sched != FIFO_NO_PRIORITY)
+ {
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+
+ ixEthAccPortData[portId].ixEthAccTxData.schDiscipline = sched;
+ return (IX_ETH_ACC_SUCCESS);
+}
+
+IX_ETH_ACC_PUBLIC
+IxEthAccStatus ixEthAccRxSchedulingDisciplineSetPriv(IxEthAccSchedulerDiscipline
+ sched)
+{
+ if (sched != FIFO_PRIORITY && sched != FIFO_NO_PRIORITY)
+ {
+ return (IX_ETH_ACC_INVALID_ARG);
+ }
+
+ ixEthAccDataInfo.schDiscipline = sched;
+
+ return (IX_ETH_ACC_SUCCESS);
+}
+
+
+/**
+ * @fn ixEthRxFrameProcess(IxEthAccPortId portId, IX_OSAL_MBUF *mbufPtr)
+ *
+ * @brief process incoming frame :
+ *
+ * @param @ref IxQMgrCallback IxQMgrMultiBufferCallback
+ *
+ * @return none
+ *
+ * @internal
+ *
+ */
+IX_ETH_ACC_PRIVATE BOOL
+ixEthRxFrameProcess(IxEthAccPortId portId, IX_OSAL_MBUF *mbufPtr)
+{
+ UINT32 flags;
+ IxEthDBStatus result;
+
+#ifndef NDEBUG
+ /* Prudent to at least check the port is within range */
+ if (portId >= IX_ETH_ACC_NUMBER_OF_PORTS)
+ {
+ ixEthAccDataStats.unexpectedError++;
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthRxFrameProcess: Illegal port: %u\n",
+ (UINT32)portId, 0, 0, 0, 0, 0);
+ return FALSE;
+ }
+#endif
+
+ /* convert fields from mbuf header */
+ ixEthAccMbufFromRxQ(mbufPtr);
+
+ /* check about any special processing for this frame */
+ flags = IX_ETHACC_NE_FLAGS(mbufPtr);
+ if ((flags & (IX_ETHACC_NE_FILTERMASK | IX_ETHACC_NE_NEWSRCMASK)) == 0)
+ {
+ /* "best case" scenario : nothing special to do for this frame */
+ return TRUE;
+ }
+
+#ifdef CONFIG_IXP425_COMPONENT_ETHDB
+ /* if a new source MAC address is detected by the NPE,
+ * update IxEthDB with the portId and the MAC address.
+ */
+ if ((flags & IX_ETHACC_NE_NEWSRCMASK & ixEthAccNewSrcMask) != 0)
+ {
+ result = ixEthDBFilteringDynamicEntryProvision(portId,
+ (IxEthDBMacAddr *) IX_ETHACC_NE_SOURCEMAC(mbufPtr));
+
+ if (result != IX_ETH_DB_SUCCESS && result != IX_ETH_DB_FEATURE_UNAVAILABLE)
+ {
+ if ((ixEthAccMacState[portId].portDisableState == ACTIVE) && (result != IX_ETH_DB_BUSY))
+ {
+ RX_STATS_INC(portId, rxUnexpectedError);
+ IX_ETH_ACC_FATAL_LOG("ixEthRxFrameProcess: Failed to add source MAC \
+ to the Learning/Filtering database\n", 0, 0, 0, 0, 0, 0);
+ }
+ else
+ {
+ /* we expect this to fail during PortDisable, as EthDB is disabled for
+ * that port and will refuse to learn new addresses
+ */
+ }
+ }
+ else
+ {
+ RX_STATS_INC(portId, rxUnlearnedMacAddress);
+ }
+ }
+#endif
+
+ /* check if this frame should have been filtered
+ * by the NPE and take the appropriate action
+ */
+ if (((flags & IX_ETHACC_NE_FILTERMASK) != 0)
+ && (ixEthAccMacState[portId].portDisableState == ACTIVE))
+ {
+ /* If the mbuf was allocated with a small data size, or the current data pointer is not
+ * within the allocated data area, then the buffer is non-standard and has to be
+ * replenished with the minimum size only
+ */
+ if( (IX_OSAL_MBUF_ALLOCATED_BUFF_LEN(mbufPtr) < IX_ETHNPE_ACC_RXFREE_BUFFER_LENGTH_MIN)
+ || ((UINT8 *)IX_OSAL_MBUF_ALLOCATED_BUFF_DATA(mbufPtr) > IX_OSAL_MBUF_MDATA(mbufPtr))
+ || ((UINT8 *)(IX_OSAL_MBUF_ALLOCATED_BUFF_DATA(mbufPtr) +
+ IX_OSAL_MBUF_ALLOCATED_BUFF_LEN(mbufPtr))
+ < IX_OSAL_MBUF_MDATA(mbufPtr)) )
+ {
+ /* set to minimum length */
+ IX_OSAL_MBUF_MLEN(mbufPtr) = IX_OSAL_MBUF_PKT_LEN(mbufPtr) =
+ IX_ETHNPE_ACC_RXFREE_BUFFER_LENGTH_MIN;
+ }
+ else
+ {
+ /* restore original length */
+ IX_OSAL_MBUF_MLEN(mbufPtr) = IX_OSAL_MBUF_PKT_LEN(mbufPtr) =
+ ( IX_OSAL_MBUF_ALLOCATED_BUFF_LEN(mbufPtr) -
+ (IX_OSAL_MBUF_MDATA(mbufPtr) - (UINT8 *)IX_OSAL_MBUF_ALLOCATED_BUFF_DATA(mbufPtr)) );
+ }
+
+ /* replenish from here */
+ if (ixEthAccPortRxFreeReplenish(portId, mbufPtr) != IX_ETH_ACC_SUCCESS)
+ {
+ IX_ETH_ACC_FATAL_LOG("ixEthRxFrameProcess: Failed to replenish with filtered frame\
+ on port %d\n", portId, 0, 0, 0, 0, 0);
+ }
+
+ RX_STATS_INC(portId, rxFiltered);
+
+ /* indicate that frame should not be subjected to further processing */
+ return FALSE;
+ }
+
+ return TRUE;
+}
+
+
+/**
+ * @fn ixEthRxFrameQMCallback
+ *
+ * @brief receive callback for Frame receive Q from NPE
+ *
+ * Frames are passed one-at-a-time to the user
+ *
+ * @param @ref IxQMgrCallback
+ *
+ * @return none
+ *
+ * @internal
+ *
+ * Design note : while processing the entry X, entry X+1 is preloaded
+ * into memory to reduce the number of stall cycles
+ *
+ */
+void ixEthRxFrameQMCallback(IxQMgrQId qId, IxQMgrCallbackId callbackId)
+{
+ IX_OSAL_MBUF *mbufPtr;
+ IX_OSAL_MBUF *nextMbufPtr;
+ UINT32 qEntry;
+ UINT32 nextQEntry;
+ UINT32 *qEntryPtr;
+ UINT32 portId;
+ UINT32 destPortId;
+ UINT32 npeId;
+ UINT32 rxQReadStatus;
+
+ /*
+ * Design note : entries are read in a buffer, This buffer contains
+ * an extra zeroed entry so the loop will
+ * always terminate on a null entry, whatever the result of Burst read is.
+ */
+ UINT32 rxQEntry[IX_ETH_ACC_MAX_RX_FRAME_CONSUME_PER_CALLBACK + 1];
+
+ /*
+ * Indication of the number of times the callback is used.
+ */
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.rxCallbackCounter);
+
+ do
+ {
+ /*
+ * Indication of the number of times the queue is drained
+ */
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.rxCallbackBurstRead);
+
+ /* ensure the last entry of the array contains a zeroed value */
+ qEntryPtr = rxQEntry;
+ qEntryPtr[IX_ETH_ACC_MAX_RX_FRAME_CONSUME_PER_CALLBACK] = 0;
+
+ rxQReadStatus = ixQMgrQBurstRead(qId,
+ IX_ETH_ACC_MAX_RX_FRAME_CONSUME_PER_CALLBACK,
+ qEntryPtr);
+
+#ifndef NDEBUG
+ if ((rxQReadStatus != IX_QMGR_Q_UNDERFLOW)
+ && (rxQReadStatus != IX_SUCCESS))
+ {
+ ixEthAccDataStats.unexpectedError++;
+ /*major error*/
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthRxFrameQMCallback:Error: %u\n",
+ (UINT32)rxQReadStatus, 0, 0, 0, 0, 0);
+ return;
+ }
+#endif
+
+ /* convert and preload the next entry
+ * (the conversion function takes care about null pointers which
+ * are used to mark the end of the loop)
+ */
+ nextQEntry = *qEntryPtr;
+ nextMbufPtr = ixEthAccEntryFromQConvert(nextQEntry,
+ IX_ETHNPE_QM_Q_RXENET_ADDR_MASK);
+
+ while(nextQEntry != 0)
+ {
+ /* get the next entry */
+ qEntry = nextQEntry;
+ mbufPtr = nextMbufPtr;
+
+#ifndef NDEBUG
+ if (mbufPtr == NULL)
+ {
+ ixEthAccDataStats.unexpectedError++;
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthRxFrameQMCallback: Null Mbuf Ptr\n",
+ 0, 0, 0, 0, 0, 0);
+ return;
+ }
+#endif
+
+ /* convert the next entry
+ * (the conversion function takes care about null pointers which
+ * are used to mark the end of the loop)
+ */
+ nextQEntry = *(++qEntryPtr);
+ nextMbufPtr = ixEthAccEntryFromQConvert(nextQEntry,
+ IX_ETHNPE_QM_Q_RXENET_ADDR_MASK);
+
+ /*
+ * Get Port and Npe ID from message.
+ */
+ npeId = ((IX_ETHNPE_QM_Q_RXENET_NPEID_MASK &
+ qEntry) >> IX_ETHNPE_QM_Q_FIELD_NPEID_R);
+ portId = IX_ETH_ACC_NPE_TO_PORT_ID(npeId);
+
+ /* process frame, check the return code and skip the remaining of
+ * the loop if the frame is to be filtered out
+ */
+ if (ixEthRxFrameProcess(portId, mbufPtr))
+ {
+ /* destination portId for this packet */
+ destPortId = IX_ETHACC_NE_DESTPORTID(mbufPtr);
+
+ if (destPortId != IX_ETH_DB_UNKNOWN_PORT)
+ {
+ destPortId = IX_ETH_DB_NPE_LOGICAL_ID_TO_PORT_ID(destPortId);
+ }
+
+ /* test if QoS is enabled in ethAcc
+ */
+ if (ixEthAccDataInfo.schDiscipline == FIFO_PRIORITY)
+ {
+ /* check if there is a higher priority queue
+ * which may require processing and then process it.
+ */
+ if (ixEthAccDataInfo.higherPriorityQueue[qId] < IX_QMGR_MAX_NUM_QUEUES)
+ {
+ ixEthRxFrameQMCallback(ixEthAccDataInfo.higherPriorityQueue[qId],
+ callbackId);
+ }
+ }
+
+ /*
+ * increment priority stats
+ */
+ RX_STATS_INC(portId,rxPriority[IX_ETHACC_NE_QOS(mbufPtr)]);
+
+ /*
+ * increment callback count stats
+ */
+ RX_STATS_INC(portId,rxFrameClientCallback);
+
+ /*
+ * Call user level callback.
+ */
+ ixEthAccPortData[portId].ixEthAccRxData.rxCallbackFn(
+ ixEthAccPortData[portId].ixEthAccRxData.rxCallbackTag,
+ mbufPtr,
+ destPortId);
+ }
+ }
+ } while (rxQReadStatus == IX_SUCCESS);
+}
+
+/**
+ * @fn ixEthRxMultiBufferQMCallback
+ *
+ * @brief receive callback for Frame receive Q from NPE
+ *
+ * Frames are passed as an array to the user
+ *
+ * @param @ref IxQMgrCallback
+ *
+ * @return none
+ *
+ * @internal
+ *
+ * Design note : while processing the entry X, entry X+1 is preloaded
+ * into memory to reduce the number of stall cycles
+ *
+ */
+void ixEthRxMultiBufferQMCallback(IxQMgrQId qId, IxQMgrCallbackId callbackId)
+{
+ IX_OSAL_MBUF *mbufPtr;
+ IX_OSAL_MBUF *nextMbufPtr;
+ UINT32 qEntry;
+ UINT32 nextQEntry;
+ UINT32 *qEntryPtr;
+ UINT32 portId;
+ UINT32 npeId;
+ UINT32 rxQReadStatus;
+ /*
+ * Design note : entries are read in a static buffer, This buffer contains
+ * an extra zeroed entry so the loop will
+ * always terminate on a null entry, whatever the result of Burst read is.
+ */
+ static UINT32 rxQEntry[IX_ETH_ACC_MAX_RX_FRAME_CONSUME_PER_CALLBACK + 1];
+ static IX_OSAL_MBUF *rxMbufPortArray[IX_ETH_ACC_NUMBER_OF_PORTS][IX_ETH_ACC_MAX_RX_FRAME_CONSUME_PER_CALLBACK + 1];
+ IX_OSAL_MBUF **rxMbufPtr[IX_ETH_ACC_NUMBER_OF_PORTS];
+
+ for (portId = 0; portId < IX_ETH_ACC_NUMBER_OF_PORTS; portId++)
+ {
+ rxMbufPtr[portId] = rxMbufPortArray[portId];
+ }
+
+ /*
+ * Indication of the number of times the callback is used.
+ */
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.rxCallbackCounter);
+
+ do
+ {
+ /*
+ * Indication of the number of times the queue is drained
+ */
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.rxCallbackBurstRead);
+
+ /* ensure the last entry of the array contains a zeroed value */
+ qEntryPtr = rxQEntry;
+ qEntryPtr[IX_ETH_ACC_MAX_RX_FRAME_CONSUME_PER_CALLBACK] = 0;
+
+ rxQReadStatus = ixQMgrQBurstRead(qId,
+ IX_ETH_ACC_MAX_RX_FRAME_CONSUME_PER_CALLBACK,
+ qEntryPtr);
+
+#ifndef NDEBUG
+ if ((rxQReadStatus != IX_QMGR_Q_UNDERFLOW)
+ && (rxQReadStatus != IX_SUCCESS))
+ {
+ ixEthAccDataStats.unexpectedError++;
+ /*major error*/
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthRxFrameMultiBufferQMCallback:Error: %u\n",
+ (UINT32)rxQReadStatus, 0, 0, 0, 0, 0);
+ return;
+ }
+#endif
+
+ /* convert and preload the next entry
+ * (the conversion function takes care about null pointers which
+ * are used to mark the end of the loop)
+ */
+ nextQEntry = *qEntryPtr;
+ nextMbufPtr = ixEthAccEntryFromQConvert(nextQEntry,
+ IX_ETHNPE_QM_Q_RXENET_ADDR_MASK);
+
+ while(nextQEntry != 0)
+ {
+ /* get the next entry */
+ qEntry = nextQEntry;
+ mbufPtr = nextMbufPtr;
+
+#ifndef NDEBUG
+ if (mbufPtr == NULL)
+ {
+ ixEthAccDataStats.unexpectedError++;
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthRxFrameMultiBufferQMCallback:Error: Null Mbuf Ptr\n",
+ 0, 0, 0, 0, 0, 0);
+ return;
+ }
+#endif
+
+ /* convert the next entry
+ * (the conversion function takes care about null pointers which
+ * are used to mark the end of the loop)
+ */
+ nextQEntry = *(++qEntryPtr);
+ nextMbufPtr = ixEthAccEntryFromQConvert(nextQEntry,
+ IX_ETHNPE_QM_Q_RXENET_ADDR_MASK);
+
+ /*
+ * Get Port and Npe ID from message.
+ */
+ npeId = ((IX_ETHNPE_QM_Q_RXENET_NPEID_MASK &
+ qEntry) >>
+ IX_ETHNPE_QM_Q_FIELD_NPEID_R);
+ portId = IX_ETH_ACC_NPE_TO_PORT_ID(npeId);
+
+ /* skip the remaining of the loop if the frame is
+ * to be filtered out
+ */
+ if (ixEthRxFrameProcess(portId, mbufPtr))
+ {
+ /* store a mbuf pointer in an array */
+ *rxMbufPtr[portId]++ = mbufPtr;
+
+ /*
+ * increment priority stats
+ */
+ RX_STATS_INC(portId,rxPriority[IX_ETHACC_NE_QOS(mbufPtr)]);
+ }
+
+ /* test for QoS enabled in ethAcc */
+ if (ixEthAccDataInfo.schDiscipline == FIFO_PRIORITY)
+ {
+ /* check if there is a higher priority queue
+ * which may require processing and then process it.
+ */
+ if (ixEthAccDataInfo.higherPriorityQueue[qId] < IX_QMGR_MAX_NUM_QUEUES)
+ {
+ ixEthRxMultiBufferQMCallback(ixEthAccDataInfo.higherPriorityQueue[qId],
+ callbackId);
+ }
+ }
+ }
+
+ /* check if any of the the arrays contains any entry */
+ for (portId = 0; portId < IX_ETH_ACC_NUMBER_OF_PORTS; portId++)
+ {
+ if (rxMbufPtr[portId] != rxMbufPortArray[portId])
+ {
+ /* add a last NULL pointer at the end of the
+ * array of mbuf pointers
+ */
+ *rxMbufPtr[portId] = NULL;
+
+ /*
+ * increment callback count stats
+ */
+ RX_STATS_INC(portId,rxFrameClientCallback);
+
+ /*
+ * Call user level callback with an array of
+ * buffers (NULL terminated)
+ */
+ ixEthAccPortData[portId].ixEthAccRxData.
+ rxMultiBufferCallbackFn(
+ ixEthAccPortData[portId].ixEthAccRxData.
+ rxMultiBufferCallbackTag,
+ rxMbufPortArray[portId]);
+
+ /* reset the buffer pointer to the beginning of
+ * the array
+ */
+ rxMbufPtr[portId] = rxMbufPortArray[portId];
+ }
+ }
+
+ } while (rxQReadStatus == IX_SUCCESS);
+}
+
+
+/**
+ * @brief rxFree low event handler
+ *
+ */
+void ixEthRxFreeQMCallback(IxQMgrQId qId, IxQMgrCallbackId callbackId)
+{
+ IxEthAccPortId portId = (IxEthAccPortId) callbackId;
+ int lockVal;
+ UINT32 maxQWritesToPerform = IX_ETH_ACC_MAX_RX_FREE_BUFFERS_LOAD;
+ IX_STATUS qStatus = IX_SUCCESS;
+
+ /*
+ * We have reached a low threshold on one of the Rx Free Qs
+ */
+
+ /*note that due to the fact that we are working off an Empty threshold, this callback
+ need only write a single entry to the Rx Free queue in order to re-arm the notification
+ */
+
+ RX_STATS_INC(portId,rxFreeLowCallback);
+
+ /*
+ * Get buffers from approprite S/W Rx freeBufferList Q.
+ */
+
+#ifndef NDEBUG
+ if (!IX_ETH_ACC_IS_PORT_VALID(portId))
+ {
+ ixEthAccDataStats.unexpectedError++;
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthRxFreeQMCallback:Error: Invalid Port 0x%08X\n",
+ portId, 0, 0, 0, 0, 0);
+ return;
+ }
+#endif
+ IX_ETH_ACC_DATA_PLANE_LOCK(lockVal);
+ if (IX_ETH_ACC_DATAPLANE_IS_Q_EMPTY(ixEthAccPortData[portId].
+ ixEthAccRxData.freeBufferList))
+ {
+ /*
+ * Turn off Q callback notification for Q in Question.
+ */
+ qStatus = ixQMgrNotificationDisable(
+ IX_ETH_ACC_PORT_TO_RX_FREE_Q_ID(portId));
+
+
+ IX_ETH_ACC_DATA_PLANE_UNLOCK(lockVal);
+
+ if (qStatus != IX_SUCCESS)
+ {
+ RX_INC(portId,rxUnexpectedError);
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthRxFreeQMCallback:Error: unexpected QM status 0x%08X\n",
+ qStatus, 0, 0, 0, 0, 0);
+ return;
+ }
+ }
+ else
+ {
+ IX_ETH_ACC_DATA_PLANE_UNLOCK(lockVal);
+ /*
+ * Load the H/W Q with buffers from the s/w Q.
+ */
+
+ do
+ {
+ /*
+ * Consume Q entries. - Note Q contains Physical addresss,
+ * and have already been flushed to memory,
+ * And endianess converted if required.
+ */
+ if (ixEthAccRxFreeFromSwQ(portId) != IX_SUCCESS)
+ {
+ /*
+ * No more entries in s/w Q.
+ * Turn off Q callback indication
+ */
+
+ IX_ETH_ACC_DATA_PLANE_LOCK(lockVal);
+ if (IX_ETH_ACC_DATAPLANE_IS_Q_EMPTY(ixEthAccPortData[portId].
+ ixEthAccRxData.freeBufferList))
+ {
+ qStatus = ixQMgrNotificationDisable(
+ IX_ETH_ACC_PORT_TO_RX_FREE_Q_ID(portId));
+ }
+ IX_ETH_ACC_DATA_PLANE_UNLOCK(lockVal);
+ break;
+ }
+ }
+ while (--maxQWritesToPerform);
+ }
+}
+/**
+ * @fn Tx queue low event handler
+ *
+ */
+void
+ixEthTxFrameQMCallback(IxQMgrQId qId, IxQMgrCallbackId callbackId)
+{
+ IxEthAccPortId portId = (IxEthAccPortId) callbackId;
+ int lockVal;
+ UINT32 maxQWritesToPerform = IX_ETH_ACC_MAX_TX_FRAME_TX_CONSUME_PER_CALLBACK;
+ IX_STATUS qStatus = IX_SUCCESS;
+ IxEthAccTxPriority highestPriority;
+
+
+ /*
+ * We have reached a low threshold on the Tx Q, and are being asked to
+ * supply a buffer for transmission from our S/W TX queues
+ */
+ TX_STATS_INC(portId,txLowThreshCallback);
+
+ /*
+ * Get buffers from approprite Q.
+ */
+
+#ifndef NDEBUG
+ if (!IX_ETH_ACC_IS_PORT_VALID(portId))
+ {
+ ixEthAccDataStats.unexpectedError++;
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthTxFrameQMCallback:Error: Invalid Port 0x%08X\n",
+ portId, 0, 0, 0, 0, 0);
+ return;
+ }
+#endif
+
+ do
+ {
+ /*
+ * Consume Q entries. - Note Q contains Physical addresss,
+ * and have already been flushed to memory,
+ * and endianess already sone if required.
+ */
+
+ IX_ETH_ACC_DATA_PLANE_LOCK(lockVal);
+
+ if(ixEthAccTxSwQHighestPriorityGet(portId, &highestPriority) ==
+ IX_ETH_ACC_FAIL)
+ {
+ /*
+ * No more entries in s/w Q.
+ * Turn off Q callback indication
+ */
+ qStatus = ixQMgrNotificationDisable(
+ IX_ETH_ACC_PORT_TO_TX_Q_ID(portId));
+
+ IX_ETH_ACC_DATA_PLANE_UNLOCK(lockVal);
+
+ if (qStatus != IX_SUCCESS)
+ {
+ ixEthAccDataStats.unexpectedError++;
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthTxFrameQMCallback:Error: unexpected QM status 0x%08X\n",
+ qStatus, 0, 0, 0, 0, 0);
+ }
+
+ return;
+ }
+ else
+ {
+ IX_ETH_ACC_DATA_PLANE_UNLOCK(lockVal);
+ if (ixEthAccTxFromSwQ(portId,highestPriority)!=IX_SUCCESS)
+ {
+ /* nothing left in the sw queue or the hw queues are
+ * full. There is no point to continue to drain the
+ * sw queues
+ */
+ return;
+ }
+ }
+ }
+ while (--maxQWritesToPerform);
+}
+
+/**
+ * @brief TxDone event handler
+ *
+ * Design note : while processing the entry X, entry X+1 is preloaded
+ * into memory to reduce the number of stall cycles
+ *
+ */
+
+void
+ixEthTxFrameDoneQMCallback(IxQMgrQId qId, IxQMgrCallbackId callbackId)
+{
+ IX_OSAL_MBUF *mbufPtr;
+ UINT32 qEntry;
+ UINT32 *qEntryPtr;
+ UINT32 txDoneQReadStatus;
+ UINT32 portId;
+ UINT32 npeId;
+
+ /*
+ * Design note : entries are read in a static buffer, This buffer contains
+ * an extra entyry (which is zeroed by the compiler), so the loop will
+ * always terminate on a null entry, whatever the result of Burst read is.
+ */
+ static UINT32 txDoneQEntry[IX_ETH_ACC_MAX_TX_FRAME_DONE_CONSUME_PER_CALLBACK + 1];
+
+ /*
+ * Indication that Tx frames have been transmitted from the NPE.
+ */
+
+ IX_ETH_ACC_STATS_INC(ixEthAccDataStats.txDoneCallbackCounter);
+
+ do{
+ qEntryPtr = txDoneQEntry;
+ txDoneQReadStatus = ixQMgrQBurstRead(IX_ETH_ACC_TX_FRAME_DONE_ETH_Q,
+ IX_ETH_ACC_MAX_TX_FRAME_DONE_CONSUME_PER_CALLBACK,
+ qEntryPtr);
+
+#ifndef NDEBUG
+ if (txDoneQReadStatus != IX_QMGR_Q_UNDERFLOW
+ && (txDoneQReadStatus != IX_SUCCESS))
+ {
+ /*major error*/
+ ixEthAccDataStats.unexpectedError++;
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthTxFrameDoneQMCallback:Error: %u\n",
+ (UINT32)txDoneQReadStatus, 0, 0, 0, 0, 0);
+ return;
+ }
+#endif
+
+ qEntry = *qEntryPtr;
+
+ while(qEntry != 0)
+ {
+ mbufPtr = ixEthAccEntryFromQConvert(qEntry,
+ IX_ETHNPE_QM_Q_TXENET_ADDR_MASK);
+
+#ifndef NDEBUG
+ if (mbufPtr == NULL)
+ {
+ ixEthAccDataStats.unexpectedError++;
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthTxFrameDoneQMCallback:Error: Null Mbuf Ptr\n",
+ 0, 0, 0, 0, 0, 0);
+ return;
+ }
+#endif
+
+ /* endianness conversions and stats updates */
+ ixEthAccMbufFromTxQ(mbufPtr);
+
+ /*
+ * Get NPE id from message, then convert to portId.
+ */
+ npeId = ((IX_ETHNPE_QM_Q_TXENETDONE_NPEID_MASK &
+ qEntry) >>
+ IX_ETHNPE_QM_Q_FIELD_NPEID_R);
+ portId = IX_ETH_ACC_NPE_TO_PORT_ID(npeId);
+
+#ifndef NDEBUG
+ /* Prudent to at least check the port is within range */
+ if (portId >= IX_ETH_ACC_NUMBER_OF_PORTS)
+ {
+ ixEthAccDataStats.unexpectedError++;
+ IX_ETH_ACC_FATAL_LOG(
+ "ixEthTxFrameDoneQMCallback: Illegal port: %u\n",
+ (UINT32)portId, 0, 0, 0, 0, 0);
+ return;
+ }
+#endif
+
+ TX_STATS_INC(portId,txDoneClientCallback);
+
+ /*
+ * Call user level callback.
+ */
+ ixEthAccPortData[portId].ixEthAccTxData.txBufferDoneCallbackFn(
+ ixEthAccPortData[portId].ixEthAccTxData.txCallbackTag,
+ mbufPtr);
+
+ /* move to next queue entry */
+ qEntry = *(++qEntryPtr);
+
+ }
+ } while( txDoneQReadStatus == IX_SUCCESS );
+}
+
+IX_ETH_ACC_PUBLIC
+void ixEthAccDataPlaneShow(void)
+{
+ UINT32 numTx0Entries;
+ UINT32 numTx1Entries;
+ UINT32 numTxDoneEntries;
+ UINT32 numRxEntries;
+ UINT32 numRxFree0Entries;
+ UINT32 numRxFree1Entries;
+ UINT32 portId;
+#ifdef __ixp46X
+ UINT32 numTx2Entries;
+ UINT32 numRxFree2Entries;
+#endif
+#ifndef NDEBUG
+ UINT32 priority;
+ UINT32 numBuffersInRx=0;
+ UINT32 numBuffersInTx=0;
+ UINT32 numBuffersInSwQ=0;
+ UINT32 totalBuffers=0;
+ UINT32 rxFreeCallbackCounter = 0;
+ UINT32 txCallbackCounter = 0;
+#endif
+ UINT32 key;
+
+ /* snapshot of stats */
+ IxEthAccTxDataStats tx[IX_ETH_ACC_NUMBER_OF_PORTS];
+ IxEthAccRxDataStats rx[IX_ETH_ACC_NUMBER_OF_PORTS];
+ IxEthAccDataPlaneStats stats;
+
+ if (!IX_ETH_ACC_IS_SERVICE_INITIALIZED())
+ {
+ return;
+ }
+
+ /* get a reliable snapshot */
+ key = ixOsalIrqLock();
+
+ numTx0Entries = 0;
+ ixQMgrQNumEntriesGet(IX_ETH_ACC_TX_FRAME_ENET0_Q, &numTx0Entries);
+ numTx1Entries = 0;
+ ixQMgrQNumEntriesGet(IX_ETH_ACC_TX_FRAME_ENET1_Q, &numTx1Entries);
+ numTxDoneEntries = 0;
+ ixQMgrQNumEntriesGet( IX_ETH_ACC_TX_FRAME_DONE_ETH_Q, &numTxDoneEntries);
+ numRxEntries = 0;
+ ixEthAccQMgrRxQEntryGet(&numRxEntries);
+ numRxFree0Entries = 0;
+ ixQMgrQNumEntriesGet(IX_ETH_ACC_RX_FREE_BUFF_ENET0_Q, &numRxFree0Entries);
+ numRxFree1Entries = 0;
+ ixQMgrQNumEntriesGet(IX_ETH_ACC_RX_FREE_BUFF_ENET1_Q, &numRxFree1Entries);
+
+#ifdef __ixp46X
+ numTx2Entries = 0;
+ ixQMgrQNumEntriesGet(IX_ETH_ACC_TX_FRAME_ENET2_Q, &numTx2Entries);
+ numRxFree2Entries = 0;
+ ixQMgrQNumEntriesGet(IX_ETH_ACC_RX_FREE_BUFF_ENET2_Q, &numRxFree2Entries);
+#endif
+
+ for(portId=IX_ETH_PORT_1; portId < IX_ETH_ACC_NUMBER_OF_PORTS; portId++)
+ {
+ memcpy(&tx[portId],
+ &ixEthAccPortData[portId].ixEthAccTxData.stats,
+ sizeof(tx[portId]));
+ memcpy(&rx[portId],
+ &ixEthAccPortData[portId].ixEthAccRxData.stats,
+ sizeof(rx[portId]));
+ }
+ memcpy(&stats, &ixEthAccDataStats, sizeof(stats));
+
+ ixOsalIrqUnlock(key);
+
+#ifdef NDEBUG
+ printf("Detailed statistics collection not supported in this load\n");
+#endif
+
+ /* print snapshot */
+ for(portId=0; portId < IX_ETH_ACC_NUMBER_OF_PORTS; portId++)
+ {
+ /* If not IXP42X A0 stepping, proceed to check for existence of coprocessors */
+ if ((IX_FEATURE_CTRL_SILICON_TYPE_A0 !=
+ (ixFeatureCtrlProductIdRead() & IX_FEATURE_CTRL_SILICON_STEPPING_MASK))
+ || (IX_FEATURE_CTRL_DEVICE_TYPE_IXP42X != ixFeatureCtrlDeviceRead ()))
+ {
+ if ((IX_ETH_PORT_1 == portId) &&
+ (ixFeatureCtrlComponentCheck(IX_FEATURECTRL_ETH0) ==
+ IX_FEATURE_CTRL_COMPONENT_DISABLED))
+ {
+ continue ;
+ }
+ if ((IX_ETH_PORT_2 == portId) &&
+ (ixFeatureCtrlComponentCheck(IX_FEATURECTRL_ETH1) ==
+ IX_FEATURE_CTRL_COMPONENT_DISABLED))
+ {
+ continue ;
+ }
+ if ((IX_ETH_PORT_3 == portId) &&
+ (ixFeatureCtrlComponentCheck(IX_FEATURECTRL_NPEA_ETH) ==
+ IX_FEATURE_CTRL_COMPONENT_DISABLED))
+ {
+ continue ;
+ }
+ }
+
+ printf("PORT %u --------------------------------\n",
+ portId);
+#ifndef NDEBUG
+ printf("Tx Done Frames : %u\n",
+ tx[portId].txDoneClientCallback +
+ tx[portId].txDoneSwQDuringDisable +
+ tx[portId].txDoneDuringDisable);
+ printf("Tx Frames : %u\n",
+ tx[portId].txQOK + tx[portId].txQDelayed);
+ printf("Tx H/W Q Added OK : %u\n",
+ tx[portId].txQOK);
+ printf("Tx H/W Q Delayed : %u\n",
+ tx[portId].txQDelayed);
+ printf("Tx From S/W Q Added OK : %u\n",
+ tx[portId].txFromSwQOK);
+ printf("Tx From S/W Q Delayed : %u\n",
+ tx[portId].txFromSwQDelayed);
+ printf("Tx Overflow : %u\n",
+ tx[portId].txOverflow);
+ printf("Tx Mutual Lock : %u\n",
+ tx[portId].txLock);
+ printf("Tx Late Ntf Enabled : %u\n",
+ tx[portId].txLateNotificationEnabled);
+ printf("Tx Low Thresh CB : %u\n",
+ tx[portId].txLowThreshCallback);
+ printf("Tx Done from H/W Q (Disable) : %u\n",
+ tx[portId].txDoneDuringDisable);
+ printf("Tx Done from S/W Q (Disable) : %u\n",
+ tx[portId].txDoneSwQDuringDisable);
+ for (priority = IX_ETH_ACC_TX_PRIORITY_0;
+ priority <= IX_ETH_ACC_TX_PRIORITY_7;
+ priority++)
+ {
+ if (tx[portId].txPriority[priority])
+ {
+ printf("Tx Priority %u : %u\n",
+ priority,
+ tx[portId].txPriority[priority]);
+ }
+ }
+#endif
+ printf("Tx unexpected errors : %u (should be 0)\n",
+ tx[portId].txUnexpectedError);
+
+#ifndef NDEBUG
+ printf("Rx Frames : %u\n",
+ rx[portId].rxFrameClientCallback +
+ rx[portId].rxSwQDuringDisable+
+ rx[portId].rxDuringDisable);
+ printf("Rx Free Replenish : %u\n",
+ rx[portId].rxFreeRepOK + rx[portId].rxFreeRepDelayed);
+ printf("Rx Free H/W Q Added OK : %u\n",
+ rx[portId].rxFreeRepOK);
+ printf("Rx Free H/W Q Delayed : %u\n",
+ rx[portId].rxFreeRepDelayed);
+ printf("Rx Free From S/W Q Added OK : %u\n",
+ rx[portId].rxFreeRepFromSwQOK);
+ printf("Rx Free From S/W Q Delayed : %u\n",
+ rx[portId].rxFreeRepFromSwQDelayed);
+ printf("Rx Free Overflow : %u\n",
+ rx[portId].rxFreeOverflow);
+ printf("Rx Free Mutual Lock : %u\n",
+ rx[portId].rxFreeLock);
+ printf("Rx Free Late Ntf Enabled : %u\n",
+ rx[portId].rxFreeLateNotificationEnabled);
+ printf("Rx Free Low CB : %u\n",
+ rx[portId].rxFreeLowCallback);
+ printf("Rx From H/W Q (Disable) : %u\n",
+ rx[portId].rxDuringDisable);
+ printf("Rx From S/W Q (Disable) : %u\n",
+ rx[portId].rxSwQDuringDisable);
+ printf("Rx unlearned Mac Address : %u\n",
+ rx[portId].rxUnlearnedMacAddress);
+ printf("Rx Filtered (Rx => RxFree) : %u\n",
+ rx[portId].rxFiltered);
+
+ for (priority = IX_ETH_ACC_TX_PRIORITY_0;
+ priority <= IX_ETH_ACC_TX_PRIORITY_7;
+ priority++)
+ {
+ if (rx[portId].rxPriority[priority])
+ {
+ printf("Rx Priority %u : %u\n",
+ priority,
+ rx[portId].rxPriority[priority]);
+ }
+ }
+#endif
+ printf("Rx unexpected errors : %u (should be 0)\n",
+ rx[portId].rxUnexpectedError);
+
+#ifndef NDEBUG
+ numBuffersInTx = tx[portId].txQOK +
+ tx[portId].txQDelayed -
+ tx[portId].txDoneClientCallback -
+ tx[portId].txDoneSwQDuringDisable -
+ tx[portId].txDoneDuringDisable;
+
+ printf("# Tx Buffers currently for transmission : %u\n",
+ numBuffersInTx);
+
+ numBuffersInRx = rx[portId].rxFreeRepOK +
+ rx[portId].rxFreeRepDelayed -
+ rx[portId].rxFrameClientCallback -
+ rx[portId].rxSwQDuringDisable -
+ rx[portId].rxDuringDisable;
+
+ printf("# Rx Buffers currently for reception : %u\n",
+ numBuffersInRx);
+
+ totalBuffers += numBuffersInRx + numBuffersInTx;
+#endif
+ }
+
+ printf("---------------------------------------\n");
+
+#ifndef NDEBUG
+ printf("\n");
+ printf("Mbufs :\n");
+ printf("Tx Unchained mbufs : %u\n",
+ stats.unchainedTxMBufs);
+ printf("Tx Chained bufs : %u\n",
+ stats.chainedTxMBufs);
+ printf("TxDone Unchained mbufs : %u\n",
+ stats.unchainedTxDoneMBufs);
+ printf("TxDone Chained bufs : %u\n",
+ stats.chainedTxDoneMBufs);
+ printf("RxFree Unchained mbufs : %u\n",
+ stats.unchainedRxFreeMBufs);
+ printf("RxFree Chained bufs : %u\n",
+ stats.chainedRxFreeMBufs);
+ printf("Rx Unchained mbufs : %u\n",
+ stats.unchainedRxMBufs);
+ printf("Rx Chained bufs : %u\n",
+ stats.chainedRxMBufs);
+
+ printf("\n");
+ printf("Software queue usage :\n");
+ printf("Buffers added to S/W Q : %u\n",
+ stats.addToSwQ);
+ printf("Buffers removed from S/W Q : %u\n",
+ stats.removeFromSwQ);
+
+ printf("\n");
+ printf("Hardware queues callbacks :\n");
+
+ for(portId=0; portId < IX_ETH_ACC_NUMBER_OF_PORTS; portId++)
+ {
+ rxFreeCallbackCounter += rx[portId].rxFreeLowCallback;
+ txCallbackCounter += tx[portId].txLowThreshCallback;
+ }
+ printf("Tx Done QM Callback invoked : %u\n",
+ stats.txDoneCallbackCounter);
+ printf("Tx QM Callback invoked : %u\n",
+ txCallbackCounter);
+ printf("Rx QM Callback invoked : %u\n",
+ stats.rxCallbackCounter);
+ printf("Rx QM Callback burst read : %u\n",
+ stats.rxCallbackBurstRead);
+ printf("Rx Free QM Callback invoked : %u\n",
+ rxFreeCallbackCounter);
+#endif
+ printf("Unexpected errors in CB : %u (should be 0)\n",
+ stats.unexpectedError);
+ printf("\n");
+
+ printf("Hardware queues levels :\n");
+ printf("Transmit Port 1 Q : %u \n",numTx0Entries);
+ printf("Transmit Port 2 Q : %u \n",numTx1Entries);
+#ifdef __ixp46X
+ printf("Transmit Port 3 Q : %u \n",numTx2Entries);
+#endif
+ printf("Transmit Done Q : %u \n",numTxDoneEntries);
+ printf("Receive Q : %u \n",numRxEntries);
+ printf("Receive Free Port 1 Q : %u \n",numRxFree0Entries);
+ printf("Receive Free Port 2 Q : %u \n",numRxFree1Entries);
+#ifdef __ixp46X
+ printf("Receive Free Port 3 Q : %u \n",numRxFree2Entries);
+#endif
+
+#ifndef NDEBUG
+ printf("\n");
+ printf("# Total Buffers accounted for : %u\n",
+ totalBuffers);
+
+ numBuffersInSwQ = ixEthAccDataStats.addToSwQ -
+ ixEthAccDataStats.removeFromSwQ;
+
+ printf(" Buffers in S/W Qs : %u\n",
+ numBuffersInSwQ);
+ printf(" Buffers in H/W Qs or NPEs : %u\n",
+ totalBuffers - numBuffersInSwQ);
+#endif
+
+ printf("Rx QoS Discipline : %s\n",
+ (ixEthAccDataInfo.schDiscipline ==
+ FIFO_PRIORITY ) ? "Enabled" : "Disabled");
+
+ for(portId=0; portId < IX_ETH_ACC_NUMBER_OF_PORTS; portId++)
+ {
+ printf("Tx QoS Discipline port %u : %s\n",
+ portId,
+ (ixEthAccPortData[portId].ixEthAccTxData.schDiscipline ==
+ FIFO_PRIORITY ) ? "Enabled" : "Disabled");
+ }
+ printf("\n");
+}
+
+
+
+
+