/*
 * FILE NAME cpmodem_wrapper.c
 *
 * BRIEF MODULE DESCRIPTION
 *  Custom USB modem wrapper module
 *
 *  Author: CradlePoint Technology, Inc.  <source@cradlepoint.com>
 *          Ben Kendall <benk@cradlepoint.com>
 *          Cory Atkin <catkin@cradlepoint.com>
 *
 * Copyright 2012-2023, CradlePoint Technology, Inc.
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License version 2
 *  as published by the Free Software Foundation.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to:
 *  Free Software Foundation
 *  51 Franklin Street, Fifth Floor
 *  Boston, MA  02111-1301  USA
 */


// Necessary includes for this device driver
#include <linux/module.h>       // Needed by all modules 
#include <linux/kernel.h>       // Needed for KERN_xxxx 
#include <linux/init.h>         // Needed for the macros 
#include <linux/types.h> 
#include <linux/skbuff.h> 		
#include <linux/list.h>
#include <linux/netdevice.h>
#include <cpmodem_wrapper.h>


#define RUNTIME_DEBUG_TRACE (1 << 0)
#define RUNTIME_DEBUG_INFO (1 << 1)
#define RUNTIME_DEBUG_WARN (1 << 2)
#define RUNTIME_DEBUG_ERROR (1 << 3)

//#undef RUNTIME_DEBUG
//#define RUNTIME_DEBUG ( RUNTIME_DEBUG_TRACE | RUNTIME_DEBUG_INFO | RUNTIME_DEBUG_WARN | RUNTIME_DEBUG_ERROR )

static int cp_lkm_wrapper_log_level = 0;

#ifdef RUNTIME_DEBUG
static const char *cp_lkm_wrapper_runtime_debug_level_str[] = {
	"ASSERT",
	"TRACE",
	"INFO",
	"WARN",
	"ERROR",
};
#else
static const char *cp_lkm_wrapper_debug_log_level_str[] = {
	"ASSERT",
	"ERROR",
	"WARN",
	"INFO",
	"TRACE",
	"PRINTF"
};
#endif

static int cp_out_get_level_index(int level)
{
	int level_index = 0;
	while (level) {
		level = level >> 1;
		level_index++;
	}
	return level_index;
}

static void cp_out(int level, const char * file, int line, const char *fmt, ...)
{
	int file_str_len = 0;
	char *file_pos = (char *)file; 
    char *fmt1;
    va_list arg;
	int level_index = 0;
	const char *level_str = NULL;

    if (level) { // level of 0 is ASSERT and log - always output
		level_index = cp_out_get_level_index(level);

#ifdef RUNTIME_DEBUG
        if (!(RUNTIME_DEBUG & level)) {
            return;
        }
		level_str = cp_lkm_wrapper_runtime_debug_level_str[level_index];
#else
		if (!(cp_lkm_wrapper_log_level & level)) {
			return;
		}
		level_str = cp_lkm_wrapper_debug_log_level_str[level_index];
#endif
    }

	va_start(arg, fmt);

	if (file) {
		char *pos = (char *)file;
		while ((pos = strchr(pos, '/'))) {
			pos++;
			file_pos = pos;
		}

		file_str_len = strlen(file_pos);
	}

	fmt1 = kmalloc(strlen(fmt) + file_str_len + 12 + 6, GFP_ATOMIC); // +6 for debug type indication
	if (!fmt1) {
		return;
	}
	if (level_str) {
		if (file) {
			sprintf(fmt1, "%6s  %s(%4d):%s\n", level_str, file_pos, line, fmt);
		} else {
			sprintf(fmt1, "%6s  %s\n", level_str, fmt);
		}
	} else {
		if (file) {
			sprintf(fmt1, "%s(%4d):%s\n", file_pos, line, fmt);
		} else {
			sprintf(fmt1, "%s\n", fmt);
		}
	}
    vprintk(fmt1, arg);
    kfree(fmt1);
    va_end(arg);
}

#ifdef RUNTIME_DEBUG
// assert is always defined if RUNTIME_DEBUG is defined
#define DEBUG_ASSERT(a, args...) \
    if (!(a)) { \
        cp_out(0, __FILE__, __LINE__, args); \
		dump_stack(); \
        while(1) { }; \
    }
#define DEBUG_TRACE(args...) cp_out(RUNTIME_DEBUG_TRACE, __FILE__, __LINE__, args)
#define DEBUG_INFO(args...)  cp_out(RUNTIME_DEBUG_INFO, __FILE__, __LINE__, args)
#define DEBUG_WARN(args...)  cp_out(RUNTIME_DEBUG_WARN, __FILE__, __LINE__, args)
#define DEBUG_ERROR(args...) cp_out(RUNTIME_DEBUG_ERROR, __FILE__, __LINE__, args)

#else
#define DEBUG_ASSERT(a, args...) 
#define DEBUG_TRACE(args...)	cp_out(LOG_DEBUG_LEVEL_TRACE, __FILE__, __LINE__, args)
								
#define DEBUG_INFO(args...) 	cp_out(LOG_DEBUG_LEVEL_INFO, __FILE__, __LINE__, args)
								
#define DEBUG_WARN(args...) 	cp_out(LOG_DEBUG_LEVEL_WARN, __FILE__, __LINE__, args)
								
#define DEBUG_ERROR(args...)	cp_out(LOG_DEBUG_LEVEL_ERROR, __FILE__, __LINE__, args)
								
#define DEBUG_PRINTF(args...)   cp_out(LOG_DEBUG_LEVEL_PRINTF, __FILE__, __LINE__, args)
								
#endif

#define LOG(args...) cp_out(0, NULL, 0, args)


void cp_lkm_wrapper_set_log_level(int level)
{
	DEBUG_TRACE("%s(%d)", __FUNCTION__, level);

	cp_lkm_wrapper_log_level = level;
}

/*******************************  usb wrapper module functionality **********************************/

typedef int (*cp_lkm_wrapper_send_op)(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
typedef int (*cp_lkm_wrapper_recv_op)(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);

static int cp_lkm_generic_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_generic_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_asix_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_asix_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_asix88179_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_asix88179_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_dip_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_dip_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_ncm_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_ncm_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_msrndis_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_msrndis_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_pegasus_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_pegasus_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_qmap_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);
static int cp_lkm_qmap_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out);


static void* cp_lkm_msrndis_wrapper_alloc(cp_lkm_wrapper_type_t wrapper, void* wrapper_info, int len);
static void* cp_lkm_ncm_wrapper_alloc(cp_lkm_wrapper_type_t wrapper, void* wrapper_info, int len);
static void* cp_lkm_asix88179_wrapper_alloc(cp_lkm_wrapper_type_t wrapper, void* wrapper_info, int len);
static void* cp_lkm_qmap_wrapper_alloc(cp_lkm_wrapper_type_t wrapper, void* wrapper_info, int len);

#define CP_LKM_WRAPPER_STATE_INIT 0
#define CP_LKM_WRAPPER_STATE_SPLIT 1

#define WRAPPER_WRITE_U8(ptr,val) (*((u8*)(ptr)) = val)
#define WRAPPER_WRITE_U16(ptr,val) (*((u16*)(ptr)) = val)
#define WRAPPER_WRITE_U32(ptr,val) (*((u32*)(ptr)) = val)

#define WRAPPER_READ_U8(ptr) (*((u8*)(ptr)))
#define WRAPPER_READ_U16(ptr) (*((u16*)(ptr)))
#define WRAPPER_READ_U32(ptr) (*((u32*)(ptr)))

struct cp_lkm_wrapper_state_map{
	int id;
	cp_lkm_wrapper_state_t wrapper_state;
};
#define MAX_STATE_MAPS 16

struct cp_lkm_wrapper_context
{
    cp_lkm_wrapper_type_t wrapper;
    int send_state;                          //generic send state that can be used by all wrappers
    int recv_state;                          //generic recv state that can be used by all wrappers
    cp_lkm_wrapper_send_op send;
    cp_lkm_wrapper_recv_op recv;
    int hdr_size;
	spinlock_t lock; 	
	struct cp_lkm_wrapper_state_map state_maps[MAX_STATE_MAPS];
	int num_state_maps;	  
    struct sk_buff_head skb_ctrl_recv_list;
    struct sk_buff_head skb_data_recv_list;
    struct sk_buff_head skb_ctrl_send_list;
    struct sk_buff_head skb_data_send_list;
};

static void cp_lkm_wrapper_common_init(struct cp_lkm_wrapper_context* cpwc)
{
    cpwc->recv_state = CP_LKM_WRAPPER_STATE_INIT;
    cpwc->send_state = CP_LKM_WRAPPER_STATE_INIT;
	spin_lock_init(&cpwc->lock);
    skb_queue_head_init(&cpwc->skb_ctrl_recv_list);
    skb_queue_head_init(&cpwc->skb_ctrl_send_list);
    skb_queue_head_init(&cpwc->skb_data_recv_list);
    skb_queue_head_init(&cpwc->skb_data_send_list);
}

static void cp_lkm_wrapper_clean_list(struct sk_buff_head* list)
{
    struct sk_buff *skb;
    while((skb = skb_dequeue(list)) != NULL){
        DEBUG_INFO("%s() found a straggler", __FUNCTION__);

        dev_kfree_skb_any(skb);
    }
}

static void cp_lkm_wrapper_common_cleanup(struct cp_lkm_wrapper_context* cpwc)
{
    cp_lkm_wrapper_clean_list(&cpwc->skb_ctrl_recv_list);
    cp_lkm_wrapper_clean_list(&cpwc->skb_ctrl_send_list);
    cp_lkm_wrapper_clean_list(&cpwc->skb_data_recv_list);
    cp_lkm_wrapper_clean_list(&cpwc->skb_data_send_list);
}

static struct sk_buff* cp_lkm_wrapper_skb_make_space(struct sk_buff* skb_in, int headspace, int tailspace)
{
    int headroom = skb_headroom(skb_in);
    int tailroom = skb_tailroom(skb_in);
    int space = headspace + tailspace;
    if(skb_in == NULL) {
        DEBUG_ERROR("%s() NULL skb_in, shouldn't happen", __FUNCTION__);
        return NULL;
    }

    if ((!skb_cloned(skb_in)) && ((headroom + tailroom) >= space)) {
        if (headroom < headspace || tailroom < tailspace) {
			//printk("%s() move it\n", __FUNCTION__);
            skb_in->data = memmove(skb_in->head + headspace, skb_in->data, skb_in->len);
            skb_set_tail_pointer(skb_in, skb_in->len);
        }
    } else {
        struct sk_buff *skb2;
        //printk("%s() copy it\n", __FUNCTION__);
        skb2 = skb_copy_expand(skb_in, headspace, tailspace, GFP_ATOMIC);
        dev_kfree_skb_any(skb_in);
        skb_in = skb2;
    }
    return skb_in;
}

// generic helper function for getting the state for id from the ctxt
static cp_lkm_wrapper_state_t cp_lkm_generic_wrapper_get_state(void* ctxt, int id)
{
	struct cp_lkm_wrapper_context* cpwc = (struct cp_lkm_wrapper_context*)ctxt;
	int i;
	cp_lkm_wrapper_state_t wrapper_state = CP_LKM_WRAPPER_INVALID;
	for (i = 0; i < cpwc->num_state_maps; i++) {
		if (cpwc->state_maps[i].id == id) {
			wrapper_state = cpwc->state_maps[i].wrapper_state;
			break;
		}
	}
	//printk("%s() id: %d, state: %d\n",__FUNCTION__,id,wrapper_state);
	return wrapper_state;
}

static int cp_lkm_generic_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    DEBUG_TRACE("%s()", __FUNCTION__);
    *skb_out = skb_in;
    return CP_LKM_WRAPPER_RES_DONE;
}

static int cp_lkm_generic_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    int result = CP_LKM_WRAPPER_RES_DONE;
	cp_lkm_wrapper_state_t wrapper_state = cp_lkm_generic_wrapper_get_state(ctxt, CP_LKM_WRAPPER_DEFAULT_ID);

    //printk("%s() state: %d\n", __FUNCTION__, wrapper_state);
    *skb_out = skb_in;

    if(wrapper_state == CP_LKM_WRAPPER_CTRL) {
		//PPP modems will often use the data endpoints for AT while connecting and then PPP data once connected.
		//That's why we need to check the state here
        DEBUG_TRACE("%s() ctrl pkt", __FUNCTION__);
        *dst = CP_LKM_WRAPPER_DST_CTRL;
    }
    else{
        *dst = CP_LKM_WRAPPER_DST_DATA;
    }
    return result;
}

#define ASIX_ENABLE_PADDING 0xffff0000
#define ASIX_HDR_MASK 0x0000ffff
#define ASIX_16BIT_EVEN_MASK 0xfffe

//============================================== wrapper specific functions 
static int cp_lkm_asix_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    int pad_len;
    u32 pkt_len;
    u32 padding = ASIX_ENABLE_PADDING;

    pad_len = ((skb_in->len + sizeof(u32)) % 512) ? 0 : sizeof(u32);

    *skb_out = NULL;

    if(!skb_in) {
        DEBUG_ERROR("%s() NULL skb_in, shouldn't happen", __FUNCTION__);
        return CP_LKM_WRAPPER_RES_ERROR;
    }
    //DEBUG_INFO("%s() wrapping", __FUNCTION__);

    skb_in = cp_lkm_wrapper_skb_make_space(skb_in, sizeof(u32), pad_len);
    if (!skb_in){
        DEBUG_INFO("%s() couldn't expand", __FUNCTION__);
        *skb_out = NULL;
        return CP_LKM_WRAPPER_RES_ERROR;
    }

    //generate the len for the header
    pkt_len = ((skb_in->len ^ ASIX_HDR_MASK) << 16) + skb_in->len;
    skb_push(skb_in, sizeof(u32));
    cpu_to_le32s(&pkt_len);
    memcpy(skb_in->data, &pkt_len, sizeof(u32));

    if (pad_len) {
        cpu_to_le32s(&padding);
        memcpy(skb_tail_pointer(skb_in), &padding, sizeof(u32));
        skb_put(skb_in, sizeof(u32));
    }
    //DEBUG_INFO("%s() wrapped", __FUNCTION__);
    *skb_out = skb_in;
    return CP_LKM_WRAPPER_RES_DONE;
}

static int cp_lkm_asix_wrapper_recv(void *ctxt, int *dst, int *mux_id, struct sk_buff *skb_in, struct sk_buff **skb_out)
{
	u8  *head;
	u32  hdr;
	char *pkt;
	struct sk_buff *pkt_skb;
	u16 size;
    struct cp_lkm_wrapper_context *cpwc = (struct cp_lkm_wrapper_context *)ctxt;
    int result = CP_LKM_WRAPPER_RES_DONE;
	cp_lkm_wrapper_state_t wrapper_state = cp_lkm_generic_wrapper_get_state(ctxt, CP_LKM_WRAPPER_DEFAULT_ID);

    *skb_out = NULL;

    //skb_in is NULL when we returned 'again' previously and so the caller is recalling us. This means there should be
    //a queue'd skb for us to process.
    if(!skb_in) {
        DEBUG_TRACE("%s() had a pending", __FUNCTION__);
        skb_in = skb_dequeue(&cpwc->skb_data_recv_list);
    }
    if(!skb_in) {
        //nothing more to do
        DEBUG_TRACE("%s() done", __FUNCTION__);
        goto asix_recv_done;
    }
    if(skb_in->len < sizeof(u32)){
        DEBUG_ERROR("%s() not enough data", __FUNCTION__);
        result = CP_LKM_WRAPPER_RES_ERROR;
        goto asix_recv_done;
    }

    //read the hdr off the front
	head = (u8 *) skb_in->data;
	memcpy(&hdr, head, sizeof(u32));
	le32_to_cpus(&hdr);
	pkt = head + sizeof(u32);
	skb_pull(skb_in, sizeof(u32));

    //the complement sizes don't match, what to do? just keep going
    if ((short)(hdr & ASIX_HDR_MASK) !=
        ~((short)((hdr & ~(ASIX_HDR_MASK)) >> 16))) {
        DEBUG_INFO("%s(), bad length", __FUNCTION__);
    }
    // get the packet length 
    size = (u16) (hdr & ASIX_HDR_MASK);

    //if exact fit, send it
    if ((skb_in->len) - ((size + 1) & ASIX_16BIT_EVEN_MASK) == 0){
        DEBUG_TRACE("%s(), exact fit", __FUNCTION__);
        *skb_out = skb_in;
        skb_in = NULL;   //so we don't free it below
        goto asix_recv_done;
    }

    if (size > ETH_FRAME_LEN || size > skb_in->len) {
        //deverr(dev,"asix_rx_fixup() Bad RX Length %d", size);
        DEBUG_ERROR("%s() too big or buff too small", __FUNCTION__);

        result = CP_LKM_WRAPPER_RES_ERROR;
        goto asix_recv_done;
    }

    //multiple pkts in this one. Have to copy them
    pkt_skb = skb_clone(skb_in, GFP_ATOMIC);
    if (!pkt_skb) {
        result = CP_LKM_WRAPPER_RES_ERROR;
        goto asix_recv_done;
    }
    pkt_skb->len = size;
    pkt_skb->data = pkt;
    skb_set_tail_pointer(pkt_skb, size);
    *skb_out = pkt_skb;
        
    //This skb has multiple pkts. We just cloned the first pkt into pkt_skb above. Move past that data and if there
    //is any more data left, enqueue it and return 'again' so we can process it.
    skb_pull(skb_in, (size + 1) & ASIX_16BIT_EVEN_MASK);

    //if have more (at least hdr size worth), requeue and tell caller to come again sometime
    if (skb_in->len <= sizeof(u32)){
        DEBUG_ERROR("%s() overflowed", __FUNCTION__);
        result = CP_LKM_WRAPPER_RES_ERROR;
        goto asix_recv_done;
    }

    DEBUG_TRACE("%s() more to do", __FUNCTION__);
    skb_queue_tail(&cpwc->skb_data_recv_list, skb_in);
    skb_in = NULL;
    result = CP_LKM_WRAPPER_RES_AGAIN;

asix_recv_done:
    if(skb_in) {
        dev_kfree_skb_any(skb_in);
    }
    //if error, clear the out skb if any
    if(result == CP_LKM_WRAPPER_RES_ERROR) {
        if(*skb_out) {
            dev_kfree_skb_any(*skb_out);
            *skb_out = NULL;
        }
    }
    DEBUG_TRACE("%s() done result: 0x%x skb_out:%p", __FUNCTION__, result, *skb_out);
    if(wrapper_state == CP_LKM_WRAPPER_CTRL) {
        DEBUG_TRACE("%s() ctrl pkt", __FUNCTION__);
        *dst = CP_LKM_WRAPPER_DST_CTRL;
    }
    else{
        *dst = CP_LKM_WRAPPER_DST_DATA;
    }

	return result;
}

// asix88179 defines
#define RX_HDR_CRC_ERR (1 << 31) // should this be 29?
#define RX_HDR_DROP_ERR (1 << 30) // should this be 31?
#define RX_HDR_L3CSUM_ERR 2
#define RX_HDR_L4CSUM_ERR 1
#define RX_HDR_L4_TYPE_UDP 4
#define RX_HDR_L4_TYPE_TCP 16
#define RX_HDR_L4_TYPE_MASK 0x1c

struct cp_lkm_asix88179_wrapper_context {
    struct cp_lkm_wrapper_context common;
	u32 max_transfer_len;
    u32 *pkt_hdr;
    int pkt_cnt;
};

#define ASIX_88179_ENABLE_PADDING 0x80008000
#define ASIX_88179_13BIT_MASK 0x1fff
#define ASIX_88179_8BIT_BOUNDARY_MASK 0xFFF8

static int cp_lkm_asix88179_wrapper_send(void *ctxt, int src, int mux_id, struct sk_buff *skb_in, struct sk_buff **skb_out)
{
	struct sk_buff *skb2;
	struct cp_lkm_asix88179_wrapper_context *asix88179_wc = (struct cp_lkm_asix88179_wrapper_context *)ctxt;
    u32 hdr1;
    u32 hdr2;
    int frame_size = asix88179_wc->max_transfer_len;
    u32 mss;
    *skb_out = NULL;

	mss = skb_shinfo(skb_in)->gso_size;

	hdr1 = skb_in->len;
	hdr2 = mss;
	if (((skb_in->len + 8) % frame_size) == 0) {
		hdr2 |= ASIX_88179_ENABLE_PADDING;	// enable padding
    }

    // make space for both headers
	skb2 = cp_lkm_wrapper_skb_make_space(skb_in, sizeof(u32) * 2, 0);
	if (!skb2) {
		// skb_in is already freed in cp_lkm_wrapper_skb_make_space
		printk("%s() - could not make space\n", __FUNCTION__);
		return CP_LKM_WRAPPER_RES_ERROR;
	}
	skb_in = skb2;

	cpu_to_le32s(&hdr2);
	skb_push(skb_in, sizeof(u32));
	skb_copy_to_linear_data(skb_in, &hdr2, sizeof(u32));

	cpu_to_le32s(&hdr1);
	skb_push(skb_in, sizeof(u32));
	skb_copy_to_linear_data(skb_in, &hdr1, sizeof(u32));

    *skb_out = skb_in;
    return CP_LKM_WRAPPER_RES_DONE;
}

static void cp_lkm_asix88179_check_csum(struct sk_buff *skb, u32 *pkt_hdr)
{
    u32 err_ind = *pkt_hdr;
    bool hdr_err = (err_ind & RX_HDR_L3CSUM_ERR) || (err_ind & RX_HDR_L4CSUM_ERR);
    bool csum_valid = ((err_ind & RX_HDR_L4_TYPE_MASK) == RX_HDR_L4_TYPE_TCP) || ((err_ind & RX_HDR_L4_TYPE_MASK) == RX_HDR_L4_TYPE_UDP);
	
    skb->ip_summed = CHECKSUM_NONE;

    if (!hdr_err && csum_valid) {
		skb->ip_summed = CHECKSUM_UNNECESSARY;
    }
}

static unsigned long total_pkt_cnt = 0;
static unsigned long total_pkt_processed = 0;

static int cp_lkm_asix88179_wrapper_recv(void *ctxt, int *dst, int *mux_id, struct sk_buff *skb_in, struct sk_buff **skb_out)
{
	u32 hdr;
	u16 hdr_off;
	struct cp_lkm_asix88179_wrapper_context* cp_88179wc = (struct cp_lkm_asix88179_wrapper_context*)ctxt;
	struct cp_lkm_wrapper_context *cpwc = (struct cp_lkm_wrapper_context *)ctxt;
    int result = CP_LKM_WRAPPER_RES_DONE;
	struct sk_buff *pkt_skb;
    u16 pkt_len;
    bool crc_runt;
    unsigned int end_len;

	cp_lkm_wrapper_state_t wrapper_state = cp_lkm_generic_wrapper_get_state(ctxt, CP_LKM_WRAPPER_DEFAULT_ID);

	*skb_out = NULL;

    DEBUG_TRACE("%s()", __FUNCTION__);

    //skb_in is NULL when we returned 'again' previously and so the caller is recalling us. This means there should be
    //a queue'd skb for us to process.
    if(!skb_in) {
        DEBUG_TRACE("%s() had a pending", __FUNCTION__);
        skb_in = skb_dequeue(&cpwc->skb_data_recv_list);
    } else {
        DEBUG_TRACE("%s() 1st pkt of queue, skb_in->len=%x", __FUNCTION__, skb_in->len);
    	skb_trim(skb_in, skb_in->len - 4);
    	memcpy(&hdr, skb_tail_pointer(skb_in), sizeof(u32));
    	le32_to_cpus(&hdr);

    	cp_88179wc->pkt_cnt = (u16)hdr;
        total_pkt_cnt += cp_88179wc->pkt_cnt;
    	hdr_off = (u16)(hdr >> 16);
    	cp_88179wc->pkt_hdr = (u32 *)(skb_in->data + hdr_off);
        le32_to_cpus(cp_88179wc->pkt_hdr);
    }
    if(!skb_in) {
        //nothing more to do
        DEBUG_TRACE("%s() done", __FUNCTION__);
        goto asix_recv_done;
    }
    if(skb_in->len < sizeof(u32)){
        DEBUG_ERROR("%s() not enough data", __FUNCTION__);
        result = CP_LKM_WRAPPER_RES_ERROR;
        goto asix_recv_done;
    }

    while (cp_88179wc->pkt_cnt--) {

        pkt_len = (*cp_88179wc->pkt_hdr >> 16) & ASIX_88179_13BIT_MASK;
        end_len = (pkt_len + 7) & ASIX_88179_8BIT_BOUNDARY_MASK;

        DEBUG_TRACE("%s() rx_hdr = %x, pkt_cnt=%x, pkt_hdr=%x, pkt_len=%x", __FUNCTION__, hdr, cp_88179wc->pkt_cnt, cp_88179wc->pkt_hdr, pkt_len);
        // Check CRC or runt packet
        crc_runt = (*cp_88179wc->pkt_hdr & RX_HDR_CRC_ERR) || (*cp_88179wc->pkt_hdr & RX_HDR_DROP_ERR);
        if (crc_runt) {
            skb_pull(skb_in, end_len);
            cp_88179wc->pkt_hdr++;
            le32_to_cpus(cp_88179wc->pkt_hdr);

            DEBUG_TRACE("%s() crc error or runt", __FUNCTION__);
            continue;
        }

        total_pkt_processed++;

        //multiple packets in this one. Have to copy them
        pkt_skb = skb_clone(skb_in, GFP_ATOMIC);
        if (!pkt_skb) {
            result = CP_LKM_WRAPPER_RES_ERROR;
            goto asix_recv_done;
        }

        pkt_skb->data = skb_in->data + 2;
        pkt_skb->len = pkt_len;
        pkt_skb->truesize = pkt_len + sizeof(struct sk_buff);
        skb_set_tail_pointer(pkt_skb, pkt_len);
        cp_lkm_asix88179_check_csum(pkt_skb, cp_88179wc->pkt_hdr);
        *skb_out = pkt_skb;

        if (cp_88179wc->pkt_cnt != 0) {
            //This skb has multiple pkts. We just cloned the first pkt into pkt_skb above. Move past that data and if there
            //is any more data left, enqueue it and return 'again' so we can process it.
            skb_pull(skb_in, end_len);
            cp_88179wc->pkt_hdr++;
            le32_to_cpus(cp_88179wc->pkt_hdr);

            DEBUG_TRACE("%s() more to do", __FUNCTION__); 
            skb_queue_tail(&cpwc->skb_data_recv_list, skb_in);
            skb_in = NULL;
            result = CP_LKM_WRAPPER_RES_AGAIN;
        }
        break;
    }

asix_recv_done:
    if(skb_in) {
        dev_kfree_skb_any(skb_in);
    }
    //if error, clear the out skb if any
    if(result == CP_LKM_WRAPPER_RES_ERROR) {
        if(*skb_out) {
            dev_kfree_skb_any(*skb_out);
            *skb_out = NULL;
        }
    }
    DEBUG_TRACE("%s() done result: 0x%x skb_out:%p", __FUNCTION__, result, *skb_out);
    if(wrapper_state == CP_LKM_WRAPPER_CTRL) {
        DEBUG_TRACE("%s() ctrl pkt", __FUNCTION__);
        *dst =  CP_LKM_WRAPPER_DST_CTRL;
    } else{
        *dst = CP_LKM_WRAPPER_DST_DATA;
    }

	return result;
}

static void* cp_lkm_asix88179_wrapper_alloc(cp_lkm_wrapper_type_t wrapper, void* wrapper_info, int len)
{
    struct cp_lkm_asix88179_wrapper_context* asix88179_wc;
    struct cp_lkm_wrapper_context* wc;

    asix88179_wc = kzalloc(sizeof(struct cp_lkm_asix88179_wrapper_context), GFP_KERNEL);
    if(!asix88179_wc) {
        return NULL;
    }

	if(wrapper_info) {
		asix88179_wc->max_transfer_len = *((u32*)(wrapper_info));
		DEBUG_INFO("%s(), max transfer:%d", __FUNCTION__, asix88179_wc->max_transfer_len);
	} else {
		DEBUG_ERROR("%s(),no max transfer set", __FUNCTION__);
	}

	wc = (struct cp_lkm_wrapper_context*)asix88179_wc;
    cp_lkm_wrapper_common_init(wc);
	wc->wrapper = wrapper;
	wc->send = cp_lkm_asix88179_wrapper_send;
	wc->recv = cp_lkm_asix88179_wrapper_recv;

    return asix88179_wc;

}

// ===== pegasus wrapper
static int cp_lkm_pegasus_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    int padlen = 0;
    u32 packet_len;
    u32 hdrlen = 2;

    *skb_out = NULL;

    if(skb_in == NULL) {
        DEBUG_ERROR("%s() NULL skb_in, shouldn't happen", __FUNCTION__);
        return CP_LKM_WRAPPER_RES_ERROR;
    }
    //DEBUG_INFO("%s() wrapping", __FUNCTION__);

    skb_in = cp_lkm_wrapper_skb_make_space(skb_in, hdrlen, padlen);
    if (!skb_in){
        DEBUG_ERROR("%s() couldn't expand", __FUNCTION__);
        *skb_out = NULL;
        return CP_LKM_WRAPPER_RES_ERROR;
    }

    //generate the mirror'd len for the header
    packet_len = skb_in->len;
    skb_push(skb_in, sizeof(u16));
    WRAPPER_WRITE_U16(skb_in->data, cpu_to_le16(packet_len));

    //DEBUG_INFO("%s() wrapped", __FUNCTION__);
    *skb_out = skb_in;
    return CP_LKM_WRAPPER_RES_DONE;
}

static int cp_lkm_pegasus_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    u32 hdr_size;
    u32 pkt_size;

    //DEBUG_INFO("%s() unwrap it", __FUNCTION__);

    *skb_out = NULL;
	*dst = CP_LKM_WRAPPER_DST_DATA;
    hdr_size = 2;

    if(skb_in == NULL) {
        //nothing more to do
        DEBUG_TRACE("%s() done", __FUNCTION__);
        return CP_LKM_WRAPPER_RES_DONE;
    }

    // If don't have enough for the headers, it is an error
    if(skb_in->len < hdr_size) {
        dev_kfree_skb_any(skb_in);
        return CP_LKM_WRAPPER_RES_ERROR;
    }
    //read the pkt size and make sure have enough data. the pkt size
    //doesn't include the dip header so add it in for comparison
    pkt_size = le16_to_cpu(WRAPPER_READ_U16(skb_in->data));
    if(pkt_size > skb_in->len){
        DEBUG_ERROR("%s() bad data pkt pkt_size:%d, data size: %d", __FUNCTION__, pkt_size, skb_in->len);
        dev_kfree_skb_any(skb_in);
        return CP_LKM_WRAPPER_RES_ERROR;
    }
    //remove the dip and ethernet hdrs
    skb_pull(skb_in, hdr_size);
    *skb_out = skb_in;
    DEBUG_TRACE("%s() data pkt", __FUNCTION__);

    return CP_LKM_WRAPPER_RES_DONE; 
}

//===================direct ip wrapper
#define SIERRA_DIRECTIP_UPLINK_DATA_INDICATION_MSGID 0x3F
#define SIERRA_DIRECTIP_UPLINK_DATA_INDICATION_EXTENDED_MSGID 0x0002
#define SIERRA_DIRECTIP_UPLINK_DATA_INDICATION_MSG_SPECIFIC_ID 0x00
#define SIERRA_DIRECTIP_HDR_SIZE 6
#define SIERRA_DIRECTIP_ETHER_SIZE 14
static int cp_lkm_dip_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    u32 packet_len;
    u32 hdr_len;
	cp_lkm_wrapper_state_t wrapper_state = cp_lkm_generic_wrapper_get_state(ctxt, CP_LKM_WRAPPER_DEFAULT_ID);

    *skb_out = NULL;
    //DEBUG_INFO("%s() wrap it", __FUNCTION__);

    if(skb_in == NULL) {
        DEBUG_ERROR("%s() NULL skb_in, shouldn't happen", __FUNCTION__);
        return CP_LKM_WRAPPER_RES_ERROR;
    }

    //in ctrl mode, we don't put a wrapper on (only after data comes up)
    if(wrapper_state == CP_LKM_WRAPPER_CTRL) {
        *skb_out = skb_in;
        DEBUG_TRACE("%s() ctrl pkt", __FUNCTION__);
        return CP_LKM_WRAPPER_RES_DONE;
    }
    //DEBUG_INFO("%s() wrapping", __FUNCTION__);

    // Add header:
    //      HIP header: 6 bytes 
    //      Fake ethernet hdr: 14 bytes
    hdr_len = SIERRA_DIRECTIP_HDR_SIZE + SIERRA_DIRECTIP_ETHER_SIZE;
    skb_in = cp_lkm_wrapper_skb_make_space(skb_in, hdr_len, 0);
    if (!skb_in){
        DEBUG_ERROR("%s() couldn't expand", __FUNCTION__);
        return CP_LKM_WRAPPER_RES_ERROR;
    }

    packet_len = skb_in->len;
    packet_len += SIERRA_DIRECTIP_ETHER_SIZE;  //add bytes for the ethernet hdr (the dip hdr isn't counted in the len)

    //ethernet protocol
    skb_push(skb_in, sizeof(u16));
    WRAPPER_WRITE_U16(skb_in->data, cpu_to_be16(0x0800));

    //bogus ethernet addrs (modem side doesn't care)
    skb_push(skb_in, 12);
    memset(skb_in->data, 0, 12);

    //extended msg id
    skb_push(skb_in, sizeof(u16));
    WRAPPER_WRITE_U16(skb_in->data, cpu_to_be16(SIERRA_DIRECTIP_UPLINK_DATA_INDICATION_EXTENDED_MSGID));

    //msg specific id
    skb_push(skb_in, 1);
    WRAPPER_WRITE_U8(skb_in->data, SIERRA_DIRECTIP_UPLINK_DATA_INDICATION_MSG_SPECIFIC_ID);

    //msg indication id
    skb_push(skb_in, 1);
    WRAPPER_WRITE_U8(skb_in->data, SIERRA_DIRECTIP_UPLINK_DATA_INDICATION_MSGID);

    //len
    skb_push(skb_in, sizeof(u16));
    WRAPPER_WRITE_U16(skb_in->data, cpu_to_be16(packet_len));

    //DEBUG_INFO("%s() data pkt", __FUNCTION__);
    *skb_out = skb_in;
    return CP_LKM_WRAPPER_RES_DONE;
}

static int cp_lkm_dip_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    u32 hdr_size;
    u32 pkt_size;
	cp_lkm_wrapper_state_t wrapper_state = cp_lkm_generic_wrapper_get_state(ctxt, CP_LKM_WRAPPER_DEFAULT_ID);

    //DEBUG_INFO("%s() unwrap it", __FUNCTION__);

    *skb_out = NULL;
	*dst = CP_LKM_WRAPPER_DST_DATA;
    hdr_size = SIERRA_DIRECTIP_HDR_SIZE + SIERRA_DIRECTIP_ETHER_SIZE;

    if(skb_in == NULL) {
        //nothing more to do
        DEBUG_TRACE("%s() done", __FUNCTION__);
        return CP_LKM_WRAPPER_RES_DONE;
    }

    //There are no headers on the pkts when in ctrl mode. Only in data mode
    if(wrapper_state == CP_LKM_WRAPPER_CTRL) {
        DEBUG_TRACE("%s() ctrl pkt", __FUNCTION__);
        *skb_out = skb_in;
		*dst = CP_LKM_WRAPPER_DST_CTRL;
        return CP_LKM_WRAPPER_RES_DONE;
    }

    //from here down, they are data packets

    // If don't have enough for the headers, it is an error
    if(skb_in->len < hdr_size) {
        dev_kfree_skb_any(skb_in);
        return CP_LKM_WRAPPER_RES_ERROR;
    }
    //read the pkt size and make sure have enough data. the pkt size
    //doesn't include the dip header so add it in for comparison
    pkt_size = be16_to_cpu(WRAPPER_READ_U16(skb_in->data));
    if((pkt_size+SIERRA_DIRECTIP_HDR_SIZE) > skb_in->len){
        dev_kfree_skb_any(skb_in);
        return CP_LKM_WRAPPER_RES_ERROR;
    }
    //remove the dip and ethernet hdrs
	skb_pull(skb_in, hdr_size);
    *skb_out = skb_in;
    DEBUG_TRACE("%s() data pkt", __FUNCTION__);

    return CP_LKM_WRAPPER_RES_DONE; 
}

//===================== msrndis wrapper
#define MSRNDIS_REMOTE_NDIS_PACKET_MSG 0x00000001			// data packet

struct cp_lkm_msrndis_wrapper_context{
    struct cp_lkm_wrapper_context common;
	u32 max_transfer_len;
};

// data pkt header
struct msrndis_data_hdr { // data packet message header (msrndis_hdr preceeds this header) (payload immediately follows)
	u32 data_offset;
	u32 data_length;
	u32 OOB_data_offset;
	u32 OOB_data_length;
	u32 num_OOB_data_elements;
	u32 per_packet_info_offset;
	u32 per_packet_info_length;
	u32 reserved[2];
}__attribute__((packed));

struct msrndis_hdr { // general msrndis header at beginning of all messages
	u32 message_type;
	u32 message_length;
} __attribute__((packed));

static int cp_lkm_msrndis_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    u32 data_hdr_len;
    u32 msg_hdr_len;
    struct msrndis_data_hdr hdr;
    u32 packet_len;

    *skb_out = NULL;
    DEBUG_TRACE("%s() wrap it", __FUNCTION__);

    if(skb_in == NULL) {
        DEBUG_ERROR("%s() NULL skb_in, shouldn't happen", __FUNCTION__);
        return CP_LKM_WRAPPER_RES_ERROR;
    }

    // This bad boy has pkt data plus a data hdr plus a msg header (it was created by microsoft after all)
    packet_len = skb_in->len;
    data_hdr_len = sizeof(struct msrndis_data_hdr);
    msg_hdr_len = sizeof(struct msrndis_hdr);

    //need to add space for both headers  
    skb_in = cp_lkm_wrapper_skb_make_space(skb_in, data_hdr_len + msg_hdr_len, 0);
    if (!skb_in){
        DEBUG_ERROR("%s() couldn't expand", __FUNCTION__);
        *skb_out = NULL;
        return CP_LKM_WRAPPER_RES_ERROR;
    }

    //create the data hdr
    memset(&hdr, 0x00, data_hdr_len);
    hdr.data_offset = cpu_to_le32(data_hdr_len);  //data starts after the data hdr
    hdr.data_length = cpu_to_le32(packet_len);    //the data hdr doesn't include the hdr lenght in length, only the data
    skb_push(skb_in, data_hdr_len);
    memcpy(skb_in->data, &hdr, data_hdr_len);

    //Create the msg hdr, the length includes the msg header size as well
    packet_len = skb_in->len + msg_hdr_len;

    skb_push(skb_in, sizeof(u32));
    WRAPPER_WRITE_U32(skb_in->data, cpu_to_le32(packet_len));

    skb_push(skb_in, sizeof(u32));
    WRAPPER_WRITE_U32(skb_in->data, cpu_to_le32(MSRNDIS_REMOTE_NDIS_PACKET_MSG));

    DEBUG_TRACE("%s() data pkt", __FUNCTION__);
    *skb_out = skb_in;
    return CP_LKM_WRAPPER_RES_DONE;
}

static int cp_lkm_msrndis_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    u32 data_hdr_len = sizeof(struct msrndis_data_hdr);
    u32 msg_hdr_len = sizeof(struct msrndis_hdr);
    struct msrndis_data_hdr hdr;
	u32 adv = 0;
	u32 out_len;
    u32 pkt_len;
    u32 pkt_type;
	struct sk_buff *skb_working = NULL;

	struct cp_lkm_wrapper_context* cpwc = (struct cp_lkm_wrapper_context*)ctxt;
	struct cp_lkm_msrndis_wrapper_context* msrndis_wc = (struct cp_lkm_msrndis_wrapper_context*)ctxt;

//    DEBUG_INFO("%s() unwrap it", __FUNCTION__);
    *skb_out = NULL;
	*dst = CP_LKM_WRAPPER_DST_DATA;

	if (skb_in) {
		cpwc->recv_state = CP_LKM_WRAPPER_STATE_INIT;
        DEBUG_TRACE("%s() done", __FUNCTION__);
		if (0 == skb_in->len) {
			dev_kfree_skb_any(skb_in);
			skb_in = NULL;
		} else if (msrndis_wc->max_transfer_len == skb_in->len) {
			DEBUG_INFO("%s() - max transfer - setting split", __FUNCTION__);
			cpwc->recv_state = CP_LKM_WRAPPER_STATE_SPLIT;
		}
	}

	skb_working = skb_dequeue(&cpwc->skb_data_recv_list); 

	if (!skb_working) {
		skb_working = skb_in;
	} else if (skb_in) {
		// append data to skb_working
		skb_working = cp_lkm_wrapper_skb_make_space(skb_working, 0, skb_in->len);
		if(!skb_working) {
			DEBUG_WARN("%s() failed to make space", __FUNCTION__);
			return CP_LKM_WRAPPER_RES_ERROR;
		}
		memcpy(skb_tail_pointer(skb_working), skb_in->data, skb_in->len);
		skb_put(skb_working, skb_in->len);
		dev_kfree_skb_any(skb_in);
	}

	if (!skb_working) {
		return CP_LKM_WRAPPER_RES_DONE;
	}

	if(skb_working->len < msg_hdr_len) {
		if (CP_LKM_WRAPPER_STATE_SPLIT != cpwc->recv_state) {
			DEBUG_INFO("%s() - flushing remaining byte count:%d", __FUNCTION__, skb_working->len);
			dev_kfree_skb_any(skb_working);
			return CP_LKM_WRAPPER_RES_ERROR;
		}

		// expecting a split packet
		skb_queue_tail(&cpwc->skb_data_recv_list, skb_working);

		return CP_LKM_WRAPPER_RES_DONE;
	}

	pkt_type = le32_to_cpu(WRAPPER_READ_U32(skb_working->data));
	skb_pull(skb_working, 4);

	pkt_len = le32_to_cpu(WRAPPER_READ_U32(skb_working->data));
	skb_pull(skb_working, 4); 

	// try to determine if this packet len is reasonable
	if (pkt_len > (4 * 1024) || pkt_len < 0) {
		// probably bad packet length - drop the packets
		DEBUG_WARN("%s() - bad packet len:%x", __FUNCTION__, pkt_len);
		DEBUG_WARN("%s() - flushing remaining byte count:%d", __FUNCTION__, skb_working->len);

		dev_kfree_skb_any(skb_working);
//		DEBUG_ASSERT(0, "bad packet len:%d", pkt_len);

		return CP_LKM_WRAPPER_RES_ERROR;
	}

	if (skb_working->len < data_hdr_len) {
		if (CP_LKM_WRAPPER_STATE_SPLIT != cpwc->recv_state) {
			DEBUG_INFO("%s() - flushing remaining byte count:%d", __FUNCTION__, skb_working->len);
			dev_kfree_skb_any(skb_working);
			return CP_LKM_WRAPPER_RES_ERROR;
		}

		// expecting a split packet
		skb_push(skb_working, msg_hdr_len);
		skb_queue_tail(&cpwc->skb_data_recv_list, skb_working);

		return CP_LKM_WRAPPER_RES_DONE;
	}
    memcpy(&hdr, skb_working->data, data_hdr_len);
    hdr.data_offset = le32_to_cpu(hdr.data_offset);
    hdr.data_length = le32_to_cpu(hdr.data_length);
    skb_pull(skb_working, data_hdr_len);

    //account for any gaps between the end of the hdr and the start of data 
    if(hdr.data_offset > data_hdr_len) {
        adv = hdr.data_offset - data_hdr_len;
        if(skb_working->len < adv) {
			if (CP_LKM_WRAPPER_STATE_SPLIT != cpwc->recv_state) {
				DEBUG_INFO("%s() - flushing remaining byte count:%d", __FUNCTION__, skb_working->len);
				dev_kfree_skb_any(skb_working);
				return CP_LKM_WRAPPER_RES_ERROR;
			}

			// expecting a split packet
			skb_push(skb_working, msg_hdr_len + data_hdr_len);
			skb_queue_tail(&cpwc->skb_data_recv_list, skb_working);
			return CP_LKM_WRAPPER_RES_DONE;
        }

        skb_pull(skb_working, adv);
    }

	if(skb_working->len < hdr.data_length) {
		if (CP_LKM_WRAPPER_STATE_SPLIT != cpwc->recv_state) {
			DEBUG_INFO("%s() - flushing remaining byte count:%d", __FUNCTION__, skb_working->len);
			dev_kfree_skb_any(skb_working);
			return CP_LKM_WRAPPER_RES_ERROR;
		}
    DEBUG_TRACE("%s() data pkt", __FUNCTION__);

		// expecting a split packet
		skb_push(skb_working, msg_hdr_len + data_hdr_len + adv);
		skb_queue_tail(&cpwc->skb_data_recv_list, skb_working);
		return CP_LKM_WRAPPER_RES_DONE;
    }

	out_len = hdr.data_length;

	if (MSRNDIS_REMOTE_NDIS_PACKET_MSG != pkt_type) {
		out_len = msg_hdr_len + data_hdr_len + adv + hdr.data_length;
		skb_push(skb_working, msg_hdr_len + data_hdr_len + adv);

	}

	*skb_out = skb_clone(skb_working, GFP_ATOMIC);
	if (!(*skb_out)) {
		DEBUG_WARN("%s() - couldn't clone skb", __FUNCTION__);
		dev_kfree_skb_any(skb_working);
		return CP_LKM_WRAPPER_RES_ERROR;
	}
	skb_set_tail_pointer(*skb_out, out_len);
	(*skb_out)->len = out_len;

	skb_pull(skb_working, out_len);

	if (skb_working->len) {
		DEBUG_INFO("%s() complete pkt with remaining data: %d", __FUNCTION__, skb_working->len);
		skb_queue_tail(&cpwc->skb_data_recv_list, skb_working);
		*dst = (MSRNDIS_REMOTE_NDIS_PACKET_MSG == pkt_type) ? CP_LKM_WRAPPER_DST_DATA : CP_LKM_WRAPPER_DST_CTRL;
		return CP_LKM_WRAPPER_RES_AGAIN;
	}

	dev_kfree_skb_any(skb_working);
	*dst = (MSRNDIS_REMOTE_NDIS_PACKET_MSG == pkt_type) ? CP_LKM_WRAPPER_DST_DATA : CP_LKM_WRAPPER_DST_CTRL;
	return CP_LKM_WRAPPER_RES_DONE;

}

static void* cp_lkm_msrndis_wrapper_alloc(cp_lkm_wrapper_type_t wrapper, void* wrapper_info, int len)
{
    struct cp_lkm_msrndis_wrapper_context* msrndis_wc;
    struct cp_lkm_wrapper_context* wc;

    msrndis_wc = kzalloc(sizeof(struct cp_lkm_msrndis_wrapper_context), GFP_KERNEL);
    if(!msrndis_wc) {
        return NULL;
    }

	if(wrapper_info) {
		msrndis_wc->max_transfer_len = *((u32*)(wrapper_info));
		DEBUG_INFO("%s(), max transfer:%d", __FUNCTION__, msrndis_wc->max_transfer_len);
	}
	else{
		DEBUG_ERROR("%s(),no max transfer set", __FUNCTION__);
	}

	wc = (struct cp_lkm_wrapper_context*)msrndis_wc;
    cp_lkm_wrapper_common_init(wc);
	wc->wrapper = wrapper;
	wc->send = cp_lkm_msrndis_wrapper_send;
	wc->recv = cp_lkm_msrndis_wrapper_recv;

    return msrndis_wc;

}


//============== NCM wrapper
//There are 2 modes of operation for an NCM device, 16 bit and 32 bit. 16 bit block allows for transfer bkocks up to 64K in length,
//while 32 allows for 4G length blocks. We will be using the 16 bit, which is set in plug.

#define NTB_HEADER_SIGNATURE 0x484D434E //"NCMH" 16 bit transfer blocks signature.
#define NDP_SIGNATURE_NO_CRC 0x304D434E //"NCM0"

/// 
/// THIS STRUCTURE MUST BE THE SAME AS THE ONE IN ncm_modem.h
/// 
struct ncm_ntb_parameters{
    u16 wLength;
    u16 bmNtbFormatsSupported;
    u32 dwNtbInMaxSize;
    u16 wNdpInDivisor;
    u16 wNdpInPayloadRemainder;
    u16 wNdpInAlignment;
    u16 reserved;
    u32 dwNtbOutMaxSize;
    u16 wNdpOutDivisor;
    u16 wNdpOutPayloadRemainder;
    u16 wNdpOutAlignment;
    u16 wNtbOutMaxDatagrams;
};

//NCM Transfer Header (NTH)
struct ncm_transfer_header {
    u32 signature;
    u16 header_length;
    u16 sequence_number;
    u16 ntb_length;//length of entire block
    u16 ndp_index; //Offset in block in NDP  
}__attribute__ ((packed));

struct ncm_datagram_info {
    u16 index;
    u16 length;
}__attribute__ ((packed));

//NCM Datagram Pointers (NDP)
struct ncm_datagram_pointers {
    u32 signature;
    u16 length; // Size of this NDP. Must be multiple of 4, and at least 0x10
    u16 next_ndp_index; //offset of next ndp in NTB.
    struct ncm_datagram_info datagram_info[2]; //Setting to one datagream for now. It's 2 due to the NULL tail item required in list. 
}__attribute__ ((packed));

struct cp_lkm_ncm_wrapper_context{
    struct cp_lkm_wrapper_context common;
    u32 nth_seq_num;
    u32 datagram_offset;
    struct ncm_ntb_parameters ntb_parms; // usb max transfer size - used to detect runt HIM blocks
};

static int cp_lkm_ncm_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    u32 ndp_padding, datagram_padding, align_factor, total_size, header_size;
    u32 payload = skb_in->len;
    void *ptr = NULL;
    struct cp_lkm_ncm_wrapper_context* ncmwc = (struct cp_lkm_ncm_wrapper_context*)ctxt;

    *skb_out = NULL;

    //should never see this in here
    if(skb_in == NULL) {
        return CP_LKM_WRAPPER_RES_DONE; 
    }

    header_size = sizeof(struct ncm_transfer_header);

    //Need to align NDP by the align value. offset%align = 0.  Add align value -1 and mask off by it's inverse to get aligned offset.
    //Then subtract current header size to get the padding.
    align_factor = ncmwc->ntb_parms.wNdpInAlignment - 1;
    ndp_padding = ((header_size + align_factor) & ~align_factor) - header_size;
    header_size += ndp_padding;

    header_size += sizeof(struct ncm_datagram_pointers);

    //Alignment for 

    //Need to align NDP by the divisor value + the remainder value. offset%divisor = 0.  Add divisor value -1 and mask off by it's inverse to get aligned offset.
    //Then subtract current header size to get the alignment padding and add the remainder to get the total padding.
    align_factor = ncmwc->ntb_parms.wNdpInDivisor - 1;
    datagram_padding = (((header_size + align_factor) & ~align_factor) - header_size) + ncmwc->ntb_parms.wNdpInPayloadRemainder;

    header_size += datagram_padding;

    total_size = header_size + payload;

    //Need to account for our max transfer size allowed by modem. Current modem is 400K, should never hit this.
    if (ncmwc->ntb_parms.dwNtbInMaxSize < total_size) {
		dev_kfree_skb_any(skb_in);
        return CP_LKM_WRAPPER_RES_ERROR;
    }

    //add space for the header to skb_in
    skb_in = cp_lkm_wrapper_skb_make_space(skb_in, header_size, 0);
    if(!skb_in) {
        DEBUG_WARN("%s() couldn't make space", __FUNCTION__);
        return CP_LKM_WRAPPER_RES_ERROR;
    }

    //write NCM Pkt hdr
    ptr = (void *)skb_push(skb_in, header_size);
    memset(ptr, 0, header_size);

    WRAPPER_WRITE_U32(ptr, cpu_to_le32(NTB_HEADER_SIGNATURE));
    ptr +=4;

    //Write the header size
    WRAPPER_WRITE_U16(ptr, cpu_to_le16(sizeof(struct ncm_transfer_header)));
    ptr +=4; //Moving 2 to skip using optional sequence number

    //Total NTB size
    WRAPPER_WRITE_U16(ptr, cpu_to_le16(skb_in->len));
    ptr += 2;

    //Index of first ndp
    WRAPPER_WRITE_U16(ptr, cpu_to_le16(sizeof(struct ncm_transfer_header) + ndp_padding));
    ptr += (2 + ndp_padding);

    //Write the ndp
    WRAPPER_WRITE_U32(ptr, cpu_to_le32(NDP_SIGNATURE_NO_CRC));
    ptr +=4;

    //Write the ndp size
    WRAPPER_WRITE_U16(ptr, cpu_to_le16(sizeof(struct ncm_datagram_pointers)));
    ptr +=4; //Moving past 2 reserved as well

    //Write the datagram index. It's write after the ntb length
    WRAPPER_WRITE_U16(ptr, cpu_to_le16(header_size));
    ptr +=2; 

    //Write the datagram length.
    WRAPPER_WRITE_U16(ptr, cpu_to_le16(payload));

    //tail entry 0'd in memset.
   
    *skb_out = skb_in;
    return CP_LKM_WRAPPER_RES_DONE; 
}

/* 
 *  ------------------------------------- 
 *  |            Signature              | NCM Transfer Block
 *  -------------------------------------
 *  |          Header Length            |
 *  -------------------------------------
 *  |         Sequence Number           | 
 *  -------------------------------------
 *  |       Total Packet Length         |
 *  -------------------------------------
 *  |           NDP Index               |
 *  -------------------------------------
 *  
 *  
 *  -------------------------------------
 *  |            Signature              | NCM Datagram Pointers
 *  -------------------------------------
 *  |          Header Length            | 
 *  -------------------------------------
 *  |        Index to next NDP          |
 *  -------------------------------------
 *  |        Datagram[0] index          | 
 *  -------------------------------------
 *  |       Datagram[0] length          |
 *  -------------------------------------
 *  |        Datagram[1] index          | 
 *  -------------------------------------
 *  |       Datagram[1] length          |
 *  -------------------------------------
 *                   .
 *                   .
 *                   .
 *  -------------------------------------
 *  |        Datagram[n] index          | 
 *  -------------------------------------
 *  |       Datagram[n] length          |
 *  -------------------------------------
 *  |               0                   | Termination of header
 *  -------------------------------------
 *  |               0                   | Termination of header
 *  ------------------------------------
 *  
 *  Ethernet packets....
 *  
 *  
 *  This function processes the NCM Transfer Block. It can consist
 *  of multiple Ethernet pkts. We specified the max size we could
 *  handle during plug. Only a single SBK should ever be sent.
*/
static int cp_lkm_ncm_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    u32 tmp_val;
    u16 nth_len, ndp_len, datagram_index, datagram_len;
    struct sk_buff *ncm_skb_out;
    unsigned char *ptr = NULL, *tmp_ptr = NULL;   
    struct cp_lkm_wrapper_context* cpwc = (struct cp_lkm_wrapper_context*)ctxt;
    struct cp_lkm_ncm_wrapper_context* ncmwc = (struct cp_lkm_ncm_wrapper_context*)cpwc;
	cp_lkm_wrapper_state_t wrapper_state = cp_lkm_generic_wrapper_get_state(ctxt, CP_LKM_WRAPPER_DEFAULT_ID);

    *skb_out = NULL;
	*dst = CP_LKM_WRAPPER_DST_DATA;

    //skb_in is NULL when the caller is recalling us to finish processing the skb. 
    if(NULL != skb_in) {
        //print_hex_dump(KERN_INFO, "SKB_IN:", DUMP_PREFIX_ADDRESS, 16, 1, skb_in->data, 64, false);

        ptr = (void *)skb_in->data;
        //There are no headers on the pkts when in ctrl mode. Only in data mode. Shouldn't see control on data eps
        if(wrapper_state == CP_LKM_WRAPPER_CTRL) {
            *skb_out = skb_in;
			*dst = CP_LKM_WRAPPER_DST_CTRL;
            return CP_LKM_WRAPPER_RES_DONE;
        }    

        // Not enough data for the headers, it is an error.
        if(skb_in->len < sizeof(struct ncm_transfer_header) + sizeof(struct ncm_datagram_pointers)) {
            //DEBUG_ERROR("%s() NCM ERROR: NCM packet size error, len: %d", __FUNCTION__,skb_in->len);
            goto error;
        }

        //get the signature.
        tmp_val = le32_to_cpu(WRAPPER_READ_U32(ptr));
        ptr +=4;
        if (tmp_val != NTB_HEADER_SIGNATURE) {
            DEBUG_ERROR("%s() NCM ERROR: Invalid NCM Signature: 0x%lX", __FUNCTION__, tmp_val);
            goto error;
        }

        //Check the header length
        nth_len = le16_to_cpu(WRAPPER_READ_U16(ptr));
        ptr +=2; 
        if (nth_len != sizeof(struct ncm_transfer_header)) {
            DEBUG_ERROR("%s() NCM ERROR: Invalid NTH Size: %d", __FUNCTION__, nth_len);
            goto error;
        }

        ncmwc->nth_seq_num = le16_to_cpu(WRAPPER_READ_U16(ptr));
        ptr +=2; 

        //Get the total packet length
        tmp_val = le16_to_cpu(WRAPPER_READ_U16(ptr));
        ptr +=2;
        if (tmp_val != skb_in->len || tmp_val > ncmwc->ntb_parms.dwNtbOutMaxSize) {
            DEBUG_ERROR("%s() NCM ERROR: Invalid length: 0x%lX,  skb_in->len: 0x%lX,  dwNtbOutMaxSize: 0x%lX", __FUNCTION__, tmp_val, skb_in->len, ncmwc->ntb_parms.dwNtbOutMaxSize);
            goto error;
        }    

        //Get NDP index
        tmp_val = le16_to_cpu(WRAPPER_READ_U16(ptr));
        //Validate against spec. Table 3-2
        if (((tmp_val % 4) != 0) && (tmp_val < nth_len)) {
            DEBUG_ERROR("%s() NCM ERROR: Invalid NDP index: 0x%lX", __FUNCTION__, tmp_val);
            goto error;
        }    

        //Move pointer to ndp offset
        ptr = ((void *)skb_in->data) + tmp_val;

        //get the signature.
        tmp_val = le32_to_cpu(WRAPPER_READ_U32(ptr));
        ptr +=4;
        //We specified no CRC during plug
        if (tmp_val != NDP_SIGNATURE_NO_CRC) {
            DEBUG_ERROR("%s() NCM ERROR: Invalid NDP Signature: 0x%lX", __FUNCTION__, tmp_val);
            goto error;
        }

        //Check the header length
        ndp_len = le16_to_cpu(WRAPPER_READ_U16(ptr));
        ptr +=2; 
        //Need to subtract size of ncm_datagram_info from size of ncm_datagram_pointers to account form empty NTB.
        if ((ndp_len < sizeof(struct ncm_datagram_pointers)-sizeof(struct ncm_datagram_info))|| (ndp_len % 4 != 0)) {
            DEBUG_ERROR("%s() NCM ERROR: Invalid NDP Size: %ld", __FUNCTION__, ndp_len);
            goto error;
        }

        //Move past 2 bytes reserved.
        ptr += 2;

        //Validate datagram pointers. There must be a terminator entry or the 
        //entire packet is to be refused. Section 3.7
        tmp_ptr = ptr;
        ndp_len -= 8; //Subtrace header to get length of datagram pointers in bytes.
        while (0 < ndp_len) {
            datagram_index = le16_to_cpu(WRAPPER_READ_U16(tmp_ptr));
            tmp_ptr +=2;

            datagram_len = le16_to_cpu(WRAPPER_READ_U16(tmp_ptr));
            tmp_ptr +=2;

            //Need to check for early 0's. 
            if (0 == datagram_index &&  0 == datagram_len) {
                break;
            }

            ndp_len -= sizeof(struct ncm_datagram_info);
        } 

        //We should be at the terminator value.
        if (datagram_index != 0 &&  datagram_len != 0) {
            goto error;
        }

        datagram_index = le16_to_cpu(WRAPPER_READ_U16(ptr));
        ptr +=2;

        datagram_len = le16_to_cpu(WRAPPER_READ_U16(ptr));
        ptr +=2;

    } else {

        //We'd better have an offset
        if (0 == ncmwc->datagram_offset) {
            goto error;
        }

        skb_in = skb_dequeue(&cpwc->skb_data_recv_list);
        //We'd better have a queue'd skb for us to process. 
        if (NULL == skb_in) {
            goto error;
        }

        ptr = skb_in->data + ncmwc->datagram_offset;
        //print_hex_dump(KERN_INFO, "Data Gram PTRs:", DUMP_PREFIX_ADDRESS, 16, 1, ptr, 64, false);

        //read the next datagram info
        datagram_index = le16_to_cpu(WRAPPER_READ_U16(ptr));
        ptr +=2;
        datagram_len = le16_to_cpu(WRAPPER_READ_U16(ptr));
        ptr +=2;

        //DEBUG_TRACE("%s() dp_index: 0x%lX", __FUNCTION__, datagram_index);
        //DEBUG_TRACE("%s() datagram_len: 0x%lX", __FUNCTION__, datagram_len);
    }

    //Save offset to next datagram pointer
    ncmwc->datagram_offset = ptr - skb_in->data;

    //Handle NULL datagram pointer entries. Section 3.7. Terminator would be both having value of 0,
    //Spec says ignore anything after either of them is NULL
    if (0 == datagram_index || 0 == datagram_len) {
        if(skb_in) {
            dev_kfree_skb_any(skb_in); 
        }
        ncmwc->datagram_offset = 0;
        return CP_LKM_WRAPPER_RES_DONE;
    }

    //copy out the data packet
    ncm_skb_out = skb_clone(skb_in, GFP_ATOMIC);
    if (!ncm_skb_out) {
        DEBUG_ERROR("%s() Failed to clone skb_in", __FUNCTION__);
        goto error;
    }
    ncm_skb_out->len = datagram_len;
    ncm_skb_out->data += datagram_index;

    skb_set_tail_pointer(ncm_skb_out, ncm_skb_out->len);
    *skb_out = ncm_skb_out;

    //print_hex_dump(KERN_INFO, "skb_out:", DUMP_PREFIX_ADDRESS, 0, 1, ncm_skb_out->data, 64, false);

    //Check next datagram pointer for terminator
    datagram_index = le16_to_cpu(WRAPPER_READ_U16(ptr));
    ptr +=2;

    datagram_len = le16_to_cpu(WRAPPER_READ_U16(ptr));

    if (0 == datagram_index || 0 == datagram_len) {
        if(skb_in) {
            dev_kfree_skb_any(skb_in); 
        }
        ncmwc->datagram_offset = 0;
        return CP_LKM_WRAPPER_RES_DONE;
    }

    //Not done, so queue up for next call. We need to come back to process the terminator packet.
    skb_queue_tail(&cpwc->skb_data_recv_list, skb_in);
    return CP_LKM_WRAPPER_RES_AGAIN;

error:
    if(skb_in) {
        dev_kfree_skb_any(skb_in); 
    }

    ncmwc->datagram_offset = 0;
    return CP_LKM_WRAPPER_RES_ERROR;

}

static void* cp_lkm_ncm_wrapper_alloc(cp_lkm_wrapper_type_t wrapper, void* wrapper_info, int len)
{

    struct cp_lkm_ncm_wrapper_context* ncmwc;
    struct cp_lkm_wrapper_context* wc;

    DEBUG_TRACE("%s() ", __FUNCTION__);
    ncmwc = kzalloc(sizeof(struct cp_lkm_ncm_wrapper_context), GFP_KERNEL);
    if(!ncmwc) {
        return NULL;
    }
    if(wrapper_info) {
        memcpy(&ncmwc->ntb_parms,(struct ncm_ntb_parameters*)(wrapper_info), sizeof(struct ncm_ntb_parameters));
    }
    else{
        DEBUG_ERROR("%s(),no ncm ntb parameters", __FUNCTION__);
        return NULL;
    }
    wc = (struct cp_lkm_wrapper_context*)ncmwc;
    cp_lkm_wrapper_common_init(wc);
    wc->wrapper = wrapper;
    wc->send = cp_lkm_ncm_wrapper_send;
    wc->recv = cp_lkm_ncm_wrapper_recv;

    ncmwc->datagram_offset = 0;
    return ncmwc;

}


// ===== QMAP wrapper ================================================================

/* 
 * qmap mux header: 
 *  
 *         |-----------------------|-----------------------|-----------------------|-----------------------|                                    
 *  Octet: |           0           |           1           |           2           |           3           |                                      
 *         |-----------------------|-----------------------|-----------------------|-----------------------|                                       
 *  Bit  : |00|01|02|03|04|05|06|07|08|09|10|11|12|13|14|15|16|17|18|19|20|21|22|23|24|25|26|27|28|29|30|31|                                      
 *         |-----------------------|-----------------------|-----------------------|-----------------------|                                        
 *  Field: |C |R |    Pad Bytes    |        Mux ID         |            Payload Len With Padding           |
 *         |-----------------------|-----------------------|-----------------------|-----------------------|
 *  
 *         C          : QMAP control or data packet.
 *                           1 - QMAP control command
 *                           0 - Data packet
 *  	   R          : Reserved
 *  	   PAD        : Number of bytes padded to achieve 4 byte alignment. Padded bytes can be 0 or not.
 *  	                This is only needed if aggregating packets and need next packet to be 4 byte aligned
 *  	   Payload Len: Total payload length in bytes including padding (not including header)
 *  
 *  	   Notes: QMAP can aggregate IP packets, each with its own QMAP header in a single USB transfer.
 *  	          QMAP adds an empty header at the end, with mux id 0, and len 0. 
 *  
*/

struct qmap_hdr{
	u8 pad_bytes;
	u8 mux_id;
	u16 payload_len;
} __attribute__((packed));


#define CP_LKM_QMAP_DATA 0
#define CP_LKM_QMAP_CTRL 1

static void* cp_lkm_qmap_wrapper_alloc(cp_lkm_wrapper_type_t wrapper, void* wrapper_info, int len)
{
	struct cp_lkm_wrapper_context* cpwc;
	cpwc = kzalloc(sizeof(struct cp_lkm_wrapper_context), GFP_KERNEL);
	if(!cpwc) {
		return cpwc;
	}
	cp_lkm_wrapper_common_init(cpwc);
	cpwc->wrapper = wrapper;
	cpwc->hdr_size = sizeof(struct qmap_hdr); 
	cpwc->send = cp_lkm_qmap_wrapper_send;
	cpwc->recv = cp_lkm_qmap_wrapper_recv;
	return cpwc;
}

/*
   We only send one QMAP IP packet at a time.
   While the spec is not clear on this (at least to me), it appears we are always supposed to add
   a single empty QMAP header at the the end.
   For us this will look like this:
      QMAP Header
      IP pkt
      padding
      Empty QMAP Header (all values 0)
 
*/
static int cp_lkm_qmap_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
	struct qmap_hdr* qmh;
	int in_len;
	int hdr_size;
	int pad = 0;
	int result = CP_LKM_WRAPPER_RES_DONE;

	// don't currently care about the wrapper_state, but this is how we would get it if we did
	//cp_lkm_wrapper_state_t wrapper_state = cp_lkm_generic_wrapper_get_state(ctxt, mux_id);

	hdr_size = sizeof(struct qmap_hdr);
	in_len = skb_in->len;
	if(in_len & 3) {
		pad = 4 - (in_len & 3);
	}

	//printk("%s() src: %d, len: %d, mux_id: %d, pad: %d\n",__FUNCTION__,src,in_len,mux_id,pad);

	//add space for the initial header at the start, plus pad and ending header at the end 
	skb_in = cp_lkm_wrapper_skb_make_space(skb_in, hdr_size, pad);
	if(!skb_in) {
		DEBUG_WARN("%s() couldn't make space", __FUNCTION__);
		return CP_LKM_WRAPPER_RES_ERROR;
	}
	skb_push(skb_in, sizeof(struct qmap_hdr));

	//add the header at the front
	qmh = (struct qmap_hdr*)skb_in->data;
	qmh->pad_bytes = CP_LKM_QMAP_DATA + pad;
	qmh->mux_id = mux_id;
	qmh->payload_len = cpu_to_be16(in_len+pad);

	// CA: determined the empty header is not necessary, but not sure about pad so keeping it.
	// add pad (if needed) and empty header at the end.
	//memset(skb_tail_pointer(skb_in), 0, sizeof(struct qmap_hdr)+pad);
	//skb_put(skb_in, sizeof(struct qmap_hdr)+pad);
	memset(skb_tail_pointer(skb_in), 0, pad);
	skb_put(skb_in, pad);

	*skb_out = skb_in;
	return result;

}

static int cp_lkm_qmap_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
	int c, pad, len;
	struct cp_lkm_wrapper_context* cpwc = (struct cp_lkm_wrapper_context*)ctxt;
	struct qmap_hdr qmh;
	int hdr_size;
	struct sk_buff* tmp_skb;
	int result = CP_LKM_WRAPPER_RES_DONE;
	//cp_lkm_wrapper_state_t wrapper_state;

	hdr_size = sizeof(struct qmap_hdr);

	*skb_out = NULL;
	*dst = CP_LKM_WRAPPER_DST_DATA;

	//skb_in is NULL when we returned 'again' previously and so the caller is recalling us. This means there should be
	//a queue'd skb for us to process.
	if(skb_in == NULL) {
		//printk("%s() had a pending\n", __FUNCTION__);
		skb_in = skb_dequeue(&cpwc->skb_data_recv_list);
	}
	if(skb_in == NULL) {
		//nothing more to do
		//printk("%s() done\n", __FUNCTION__);
		goto qmap_recv_done;
	}
	if(skb_in->len < hdr_size){
		//printk("%s() not enough data, len: %d, hdr_size: %d\n", __FUNCTION__, skb_in->len, hdr_size);
		result = CP_LKM_WRAPPER_RES_ERROR;
		goto qmap_recv_done;
	}
	
	//read header
	memcpy(&qmh, skb_in->data, sizeof(struct qmap_hdr));
	qmh.payload_len = be16_to_cpu(qmh.payload_len);

	c = qmh.pad_bytes & 0x8;
	pad = qmh.pad_bytes & 0x7;
	*mux_id = qmh.mux_id;
	len = qmh.payload_len;  //payload plus pad (doesn't include hdr)
	skb_pull(skb_in, hdr_size);

	//printk("%s() c: 0x%x, pad: %d, mux_id: 0x%x, pkt len: %d, skb len: %d\n", __FUNCTION__, c,pad,qmh.mux_id,len,skb_in->len);

	// don't currently care about the usb state for processing, but if we did this is how we would get it
	//wrapper_state = cp_lkm_generic_wrapper_get_state(ctxt, *mux_id);

	if(skb_in->len < len){
		//printk("%s() not enough data, pkt len: %d, skb len: %d\n", __FUNCTION__, len, skb_in->len);
		result = CP_LKM_WRAPPER_RES_ERROR;
		goto qmap_recv_done;
	}

	//printk("%s() pkt len: %d, skb len: %d\n", __FUNCTION__, len, skb_in->len);

	if(skb_in->len == (len + sizeof(struct qmap_hdr))){
		//this is an exact fit plus an empty hdr at the end. 
		//Some modems add it, some don't. Dump the empty if present.
		skb_set_tail_pointer(skb_in, len); 
		skb_in->len -= sizeof(struct qmap_hdr);
	}

	//if exact fit, send it
	if (skb_in->len == len){
		//printk("%s(), exact fit\n", __FUNCTION__);
		skb_set_tail_pointer(skb_in, skb_in->len-pad); //dump the padding if any
		*skb_out = skb_in;
		skb_in = NULL;   //so we don't free it below
		if (c == CP_LKM_QMAP_CTRL) {
			//TODO: decode ctrl packets to find pauses and resumes if we decide to support that
			//      when not using flow control, what do I do here?
			*dst = CP_LKM_WRAPPER_DST_UNKNOWN;
		}
		else if (len == 0) {
			//this is the 0 len header at the end. Tell the outside world to dump it.
			*dst = CP_LKM_WRAPPER_DST_UNKNOWN;
		}
		goto qmap_recv_done;
	}

	//multiple packets in this one. Have to copy them
	tmp_skb = skb_clone(skb_in, GFP_ATOMIC);
	if (!tmp_skb) {
		//printk("%s() couldn't clone skb\n", __FUNCTION__);
		result = CP_LKM_WRAPPER_RES_ERROR;
		goto qmap_recv_done;
	}
	tmp_skb->len = len-pad;
	skb_set_tail_pointer(tmp_skb, len-pad);
	*skb_out = tmp_skb;

	//This skb has multiple pkts. We just cloned the first pkt into tmp_skb above. Move past that data and if there
	//is any more data left, enqueue it and return 'again' so we can process it.
	skb_pull(skb_in, len);

	//More data after this one, queue and tell caller to come again sometime
	//printk("%s() %d more to do\n", __FUNCTION__, skb_in->len);
	skb_queue_tail(&cpwc->skb_data_recv_list, skb_in);
	skb_in = NULL;
	result = CP_LKM_WRAPPER_RES_AGAIN;

	if (c == CP_LKM_QMAP_CTRL) {
		//TODO: decode ctrl packets to find pauses and resumes if we decide to support that
		//      when not using flow control, what do I do here?
		*dst = CP_LKM_WRAPPER_DST_UNKNOWN;
	}

qmap_recv_done:
	if(skb_in) {
		dev_kfree_skb_any(skb_in);
	}
	//if error, clear the out skb if any
	if(result == CP_LKM_WRAPPER_RES_ERROR) {
		if(*skb_out) {
			dev_kfree_skb_any(*skb_out);
			*skb_out = NULL;
		}
	}
	//printk("%s() done result: %d, dst: %d, mux_id: %d\n", __FUNCTION__, result, *dst, *mux_id);
	return result; 
}


//================================ API
//If any of the wrappers have wrapper_info passed in they need to save it in their structures since
//it is freed after this function returns
void *cp_lkm_wrapper_instance_alloc(cp_lkm_wrapper_type_t wrapper, void* wrapper_info, int len)
{
    struct cp_lkm_wrapper_context* cpwc = NULL;

    DEBUG_TRACE("%s() wrapper:%d", __FUNCTION__, wrapper);
    switch (wrapper) {
        case CP_LKM_WRAPPER_TYPE_ASIX:
            cpwc = kzalloc(sizeof(struct cp_lkm_wrapper_context), GFP_KERNEL);
            if(!cpwc) {
                goto wrap_alloc_done;
            }
            cp_lkm_wrapper_common_init(cpwc);
            cpwc->wrapper = wrapper;
			cpwc->hdr_size = 4; //4 byte asix hdr
            cpwc->send = cp_lkm_asix_wrapper_send;
            cpwc->recv = cp_lkm_asix_wrapper_recv;
            break;

		case CP_LKM_WRAPPER_TYPE_ASIX_88179:
			cpwc = cp_lkm_asix88179_wrapper_alloc(wrapper, wrapper_info, len);
            break;

        case CP_LKM_WRAPPER_TYPE_LG:
            // not supported
            break;

        case CP_LKM_WRAPPER_TYPE_DIRECT_IP:
            cpwc = kzalloc(sizeof(struct cp_lkm_wrapper_context), GFP_KERNEL);
            if(!cpwc) {
                goto wrap_alloc_done;
            }
            cp_lkm_wrapper_common_init(cpwc);
            cpwc->wrapper = wrapper;
            cpwc->send = cp_lkm_dip_wrapper_send;
            cpwc->recv = cp_lkm_dip_wrapper_recv;
			cpwc->hdr_size = 6; //6 byte dip hdr
            break;

	    case CP_LKM_WRAPPER_TYPE_MSRNDIS:
			cpwc = cp_lkm_msrndis_wrapper_alloc(wrapper, wrapper_info, len);
            break;

        case CP_LKM_WRAPPER_TYPE_PEGASUS:
            cpwc = kzalloc(sizeof(struct cp_lkm_wrapper_context), GFP_KERNEL);
            if(!cpwc) {
                goto wrap_alloc_done;
            }
            cp_lkm_wrapper_common_init(cpwc);
            cpwc->wrapper = wrapper;
            cpwc->send = cp_lkm_pegasus_wrapper_send;
            cpwc->recv = cp_lkm_pegasus_wrapper_recv;
			cpwc->hdr_size = 2; //2 byte pegasus hdr
            break;

        case CP_LKM_WRAPPER_TYPE_NCM:
            cpwc = cp_lkm_ncm_wrapper_alloc(wrapper, wrapper_info, len);
            break;

		case CP_LKM_WRAPPER_TYPE_QMAP:
			cpwc = cp_lkm_qmap_wrapper_alloc(wrapper, wrapper_info, len);
			break;

        default:
            cpwc = kzalloc(sizeof(struct cp_lkm_wrapper_context), GFP_KERNEL);
            if(!cpwc) {
                goto wrap_alloc_done;
            }
            cp_lkm_wrapper_common_init(cpwc);
            cpwc->wrapper = wrapper;
            cpwc->send = cp_lkm_generic_wrapper_send;
            cpwc->recv = cp_lkm_generic_wrapper_recv;
            break;
    }

wrap_alloc_done:
    return cpwc;
}

void cp_lkm_wrapper_instance_free(void* ctxt)
{
    struct cp_lkm_wrapper_context* cpwc = (struct cp_lkm_wrapper_context*)ctxt;

    DEBUG_TRACE("%s()", __FUNCTION__);

    switch (cpwc->wrapper) {
        case CP_LKM_WRAPPER_TYPE_LG:
            // not supported
            break;
        default:
            cp_lkm_wrapper_common_cleanup(cpwc);
            kfree(ctxt);
            break;
    }
}

int cp_lkm_wrapper_hdr_size(void* ctxt)
{
	struct cp_lkm_wrapper_context* cpwc = (struct cp_lkm_wrapper_context*)ctxt;
	return cpwc->hdr_size;
}

void cp_lkm_wrapper_set_state(void* ctxt, int id, cp_lkm_wrapper_state_t wrapper_state)
{
	struct cp_lkm_wrapper_context* cpwc = (struct cp_lkm_wrapper_context*)ctxt;
	int i;

	for (i = 0; i < cpwc->num_state_maps; i++) {
		if (cpwc->state_maps[i].id == id) {
			cpwc->state_maps[i].wrapper_state = wrapper_state;
			return;
		}
	}
	//if we get here, this is a new id
	if (cpwc->num_state_maps < MAX_STATE_MAPS) {
		cpwc->state_maps[cpwc->num_state_maps].wrapper_state = wrapper_state;
		cpwc->state_maps[cpwc->num_state_maps].id = id;
		cpwc->num_state_maps++;
	}
	else{
		//DEBUG_ASSERT(cpwc->num_state_maps < MAX_STATE_MAPS, "Too many wrapper ids"); 
		printk("%s() too many state maps, id: %d, state: %d\n",__FUNCTION__, id, wrapper_state);
	}
}

int cp_lkm_wrapper_send(void* ctxt, int src, int mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    struct cp_lkm_wrapper_context* cpwc = (struct cp_lkm_wrapper_context*)ctxt;
    int res;
	unsigned long flags;

//    DEBUG_ERROR("%s() ctxt:%p", __FUNCTION__, ctxt);
	spin_lock_irqsave(&cpwc->lock, flags);
    res = cpwc->send(ctxt, src, mux_id, skb_in, skb_out);
	spin_unlock_irqrestore(&cpwc->lock, flags); 
    return res;
}

int cp_lkm_wrapper_recv(void* ctxt, int* dst, int* mux_id, struct sk_buff* skb_in, struct sk_buff** skb_out)
{
    struct cp_lkm_wrapper_context* cpwc = (struct cp_lkm_wrapper_context*)ctxt;
    int res;
	unsigned long flags;

	//DEBUG_ERROR("%s() ctxt:%p", __FUNCTION__, ctxt);
	*mux_id = 0; //default this since a lot of wrappers don't set it
	spin_lock_irqsave(&cpwc->lock, flags);
    res = cpwc->recv(ctxt, dst, mux_id, skb_in, skb_out);
	spin_unlock_irqrestore(&cpwc->lock, flags); 
    return res;
}