/****************************************************************************** * * Copyright(c) 2007 - 2017 Realtek Corporation. * * This program is free software; you can redistribute it and/or modify it * under the terms of version 2 of the GNU General Public License 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. * *****************************************************************************/ #ifndef __OSDEP_LINUX_SERVICE_H_ #define __OSDEP_LINUX_SERVICE_H_ #include #include #include #include #include #include #include #include #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 5)) #include #endif /* #include */ #include #include #include #include #include #include #include #include #if (LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 26)) #include #else #include #endif #include #include #include #include #include #include #include #include #include #include /* for struct tasklet_struct */ #include #include #include #if (LINUX_VERSION_CODE <= KERNEL_VERSION(2, 5, 41)) #include #endif #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3, 7, 0)) #include #else #include #endif #ifdef RTK_DMP_PLATFORM #if (LINUX_VERSION_CODE > KERNEL_VERSION(2, 6, 12)) #include #endif #include #endif #ifdef CONFIG_NET_RADIO #define CONFIG_WIRELESS_EXT #endif /* Monitor mode */ #include #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 24)) #include #endif #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 25) && \ LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)) #define CONFIG_IEEE80211_HT_ADDT_INFO #endif #ifdef CONFIG_IOCTL_CFG80211 /* #include */ #include #endif /* CONFIG_IOCTL_CFG80211 */ #ifdef CONFIG_HAS_EARLYSUSPEND #include #endif /* CONFIG_HAS_EARLYSUSPEND */ #ifdef CONFIG_EFUSE_CONFIG_FILE #include #endif #ifdef CONFIG_USB_HCI #include #if (LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 21)) #include #else #include #endif #endif #ifdef CONFIG_BT_COEXIST_SOCKET_TRX #include #include #include #include #include #endif /* CONFIG_BT_COEXIST_SOCKET_TRX */ #ifdef CONFIG_USB_HCI typedef struct urb *PURB; #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)) #ifdef CONFIG_USB_SUSPEND #define CONFIG_AUTOSUSPEND 1 #endif #endif #endif #if defined(CONFIG_RTW_GRO) && (!defined(CONFIG_RTW_NAPI)) #error "Enable NAPI before enable GRO\n" #endif #if (KERNEL_VERSION(2, 6, 29) > LINUX_VERSION_CODE && defined(CONFIG_RTW_NAPI)) #undef CONFIG_RTW_NAPI /*#warning "Linux Kernel version too old to support NAPI (should newer than 2.6.29)\n"*/ #endif #if (KERNEL_VERSION(2, 6, 33) > LINUX_VERSION_CODE && defined(CONFIG_RTW_GRO)) #undef CONFIG_RTW_GRO /*#warning "Linux Kernel version too old to support GRO(should newer than 2.6.33)\n"*/ #endif typedef struct semaphore _sema; typedef spinlock_t _lock; #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)) typedef struct mutex _mutex; #else typedef struct semaphore _mutex; #endif struct rtw_timer_list { struct timer_list timer; void (*function)(void *); void *arg; }; typedef struct rtw_timer_list _timer; typedef struct completion _completion; struct __queue { struct list_head queue; _lock lock; }; typedef struct sk_buff _pkt; typedef unsigned char _buffer; typedef struct __queue _queue; typedef struct list_head _list; /* hlist */ typedef struct hlist_head rtw_hlist_head; typedef struct hlist_node rtw_hlist_node; /* RCU */ typedef struct rcu_head rtw_rcu_head; #define rtw_rcu_dereference(p) rcu_dereference((p)) #define rtw_rcu_dereference_protected(p, c) rcu_dereference_protected(p, c) #define rtw_rcu_assign_pointer(p, v) rcu_assign_pointer((p), (v)) #define rtw_rcu_read_lock() rcu_read_lock() #define rtw_rcu_read_unlock() rcu_read_unlock() #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 34)) #define rtw_rcu_access_pointer(p) rcu_access_pointer(p) #endif /* rhashtable */ #include "../os_dep/linux/rtw_rhashtable.h" typedef int _OS_STATUS; /* typedef u32 _irqL; */ typedef unsigned long _irqL; typedef struct net_device *_nic_hdl; typedef void *_thread_hdl_; typedef int thread_return; typedef void *thread_context; typedef void timer_hdl_return; typedef void *timer_hdl_context; #if (LINUX_VERSION_CODE > KERNEL_VERSION(2, 5, 41)) typedef struct work_struct _workitem; #else typedef struct tq_struct _workitem; #endif #if (LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 24)) #define DMA_BIT_MASK(n) (((n) == 64) ? ~0ULL : ((1ULL<<(n))-1)) #endif typedef unsigned long systime; #if (LINUX_VERSION_CODE < KERNEL_VERSION(5, 15, 0)) /* Porting from linux kernel v5.15-rc1 48eab831ae8b9f7002a533fa4235eed63ea1f1a3 */ static inline void eth_hw_addr_set(struct net_device *dev, const u8 *addr) { memcpy(dev->dev_addr, addr, ETH_ALEN); } #endif #if (LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 22)) /* Porting from linux kernel, for compatible with old kernel. */ static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) { return skb->tail; } static inline void skb_reset_tail_pointer(struct sk_buff *skb) { skb->tail = skb->data; } static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) { skb->tail = skb->data + offset; } static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) { return skb->end; } #endif __inline static void rtw_list_delete(_list *plist) { list_del_init(plist); } __inline static _list *get_next(_list *list) { return list->next; } #define LIST_CONTAINOR(ptr, type, member) \ ((type *)((char *)(ptr)-(SIZE_T)(&((type *)0)->member))) #define rtw_list_first_entry(ptr, type, member) list_first_entry(ptr, type, member) #define rtw_hlist_for_each_entry(pos, head, member) hlist_for_each_entry(pos, head, member) #define rtw_hlist_for_each_safe(pos, n, head) hlist_for_each_safe(pos, n, head) #define rtw_hlist_entry(ptr, type, member) hlist_entry(ptr, type, member) #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3, 9, 0)) #define rtw_hlist_for_each_entry_safe(pos, np, n, head, member) hlist_for_each_entry_safe(pos, n, head, member) #define rtw_hlist_for_each_entry_rcu(pos, node, head, member) hlist_for_each_entry_rcu(pos, head, member) #else #define rtw_hlist_for_each_entry_safe(pos, np, n, head, member) hlist_for_each_entry_safe(pos, np, n, head, member) #define rtw_hlist_for_each_entry_rcu(pos, node, head, member) hlist_for_each_entry_rcu(pos, node, head, member) #endif __inline static void _enter_critical(_lock *plock, _irqL *pirqL) { spin_lock_irqsave(plock, *pirqL); } __inline static void _exit_critical(_lock *plock, _irqL *pirqL) { spin_unlock_irqrestore(plock, *pirqL); } __inline static void _enter_critical_ex(_lock *plock, _irqL *pirqL) { spin_lock_irqsave(plock, *pirqL); } __inline static void _exit_critical_ex(_lock *plock, _irqL *pirqL) { spin_unlock_irqrestore(plock, *pirqL); } __inline static void _enter_critical_bh(_lock *plock, _irqL *pirqL) { spin_lock_bh(plock); } __inline static void _exit_critical_bh(_lock *plock, _irqL *pirqL) { spin_unlock_bh(plock); } __inline static void enter_critical_bh(_lock *plock) { spin_lock_bh(plock); } __inline static void exit_critical_bh(_lock *plock) { spin_unlock_bh(plock); } __inline static int _enter_critical_mutex(_mutex *pmutex, _irqL *pirqL) { int ret = 0; #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)) /* mutex_lock(pmutex); */ ret = mutex_lock_interruptible(pmutex); #else ret = down_interruptible(pmutex); #endif return ret; } __inline static int _enter_critical_mutex_lock(_mutex *pmutex, _irqL *pirqL) { int ret = 0; #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)) mutex_lock(pmutex); #else down(pmutex); #endif return ret; } __inline static void _exit_critical_mutex(_mutex *pmutex, _irqL *pirqL) { #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)) mutex_unlock(pmutex); #else up(pmutex); #endif } __inline static _list *get_list_head(_queue *queue) { return &(queue->queue); } #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4, 14, 0)) static inline void timer_hdl(struct timer_list *in_timer) #else static inline void timer_hdl(unsigned long cntx) #endif { #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4, 14, 0)) _timer *ptimer = from_timer(ptimer, in_timer, timer); #else _timer *ptimer = (_timer *)cntx; #endif ptimer->function(ptimer->arg); } __inline static void _init_timer(_timer *ptimer, _nic_hdl nic_hdl, void *pfunc, void *cntx) { ptimer->function = pfunc; ptimer->arg = cntx; #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4, 14, 0)) timer_setup(&ptimer->timer, timer_hdl, 0); #else /* setup_timer(ptimer, pfunc,(u32)cntx); */ ptimer->timer.function = timer_hdl; ptimer->timer.data = (unsigned long)ptimer; init_timer(&ptimer->timer); #endif } __inline static void _set_timer(_timer *ptimer, u32 delay_time) { mod_timer(&ptimer->timer , (jiffies + (delay_time * HZ / 1000))); } __inline static void _cancel_timer(_timer *ptimer, u8 *bcancelled) { *bcancelled = del_timer_sync(&ptimer->timer) == 1 ? 1 : 0; } static inline void _init_workitem(_workitem *pwork, void *pfunc, void *cntx) { #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)) INIT_WORK(pwork, pfunc); #elif (LINUX_VERSION_CODE > KERNEL_VERSION(2, 5, 41)) INIT_WORK(pwork, pfunc, pwork); #else INIT_TQUEUE(pwork, pfunc, pwork); #endif } __inline static void _set_workitem(_workitem *pwork) { #if (LINUX_VERSION_CODE > KERNEL_VERSION(2, 5, 41)) schedule_work(pwork); #else schedule_task(pwork); #endif } __inline static void _cancel_workitem_sync(_workitem *pwork) { #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)) cancel_work_sync(pwork); #elif (LINUX_VERSION_CODE > KERNEL_VERSION(2, 5, 41)) flush_scheduled_work(); #else flush_scheduled_tasks(); #endif } /* * Global Mutex: can only be used at PASSIVE level. * */ #define ACQUIRE_GLOBAL_MUTEX(_MutexCounter) \ { \ while (atomic_inc_return((atomic_t *)&(_MutexCounter)) != 1) { \ atomic_dec((atomic_t *)&(_MutexCounter)); \ msleep(10); \ } \ } #define RELEASE_GLOBAL_MUTEX(_MutexCounter) \ { \ atomic_dec((atomic_t *)&(_MutexCounter)); \ } static inline int rtw_netif_queue_stopped(struct net_device *pnetdev) { #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 35)) return (netif_tx_queue_stopped(netdev_get_tx_queue(pnetdev, 0)) && netif_tx_queue_stopped(netdev_get_tx_queue(pnetdev, 1)) && netif_tx_queue_stopped(netdev_get_tx_queue(pnetdev, 2)) && netif_tx_queue_stopped(netdev_get_tx_queue(pnetdev, 3))); #else return netif_queue_stopped(pnetdev); #endif } static inline void rtw_netif_wake_queue(struct net_device *pnetdev) { #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 35)) netif_tx_wake_all_queues(pnetdev); #else netif_wake_queue(pnetdev); #endif } static inline void rtw_netif_start_queue(struct net_device *pnetdev) { #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 35)) netif_tx_start_all_queues(pnetdev); #else netif_start_queue(pnetdev); #endif } static inline void rtw_netif_stop_queue(struct net_device *pnetdev) { #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 35)) netif_tx_stop_all_queues(pnetdev); #else netif_stop_queue(pnetdev); #endif } static inline void rtw_netif_device_attach(struct net_device *pnetdev) { netif_device_attach(pnetdev); } static inline void rtw_netif_device_detach(struct net_device *pnetdev) { netif_device_detach(pnetdev); } static inline void rtw_netif_carrier_on(struct net_device *pnetdev) { netif_carrier_on(pnetdev); } static inline void rtw_netif_carrier_off(struct net_device *pnetdev) { netif_carrier_off(pnetdev); } static inline int rtw_merge_string(char *dst, int dst_len, const char *src1, const char *src2) { int len = 0; len += snprintf(dst + len, dst_len - len, "%s", src1); len += snprintf(dst + len, dst_len - len, "%s", src2); return len; } #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27)) #define rtw_signal_process(pid, sig) kill_pid(find_vpid((pid)), (sig), 1) #else /* (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27)) */ #define rtw_signal_process(pid, sig) kill_proc((pid), (sig), 1) #endif /* (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27)) */ /* Suspend lock prevent system from going suspend */ #ifdef CONFIG_WAKELOCK #include #elif defined(CONFIG_ANDROID_POWER) #include #endif /* limitation of path length */ #define PATH_LENGTH_MAX PATH_MAX /* Atomic integer operations */ #define atomic_t atomic_t #define rtw_netdev_priv(netdev) (((struct rtw_netdev_priv_indicator *)netdev_priv(netdev))->priv) #define NDEV_FMT "%s" #define NDEV_ARG(ndev) ndev->name #define ADPT_FMT "%s" #define ADPT_ARG(adapter) (adapter->pnetdev ? adapter->pnetdev->name : NULL) #define FUNC_NDEV_FMT "%s(%s)" #define FUNC_NDEV_ARG(ndev) __func__, ndev->name #define FUNC_ADPT_FMT "%s(%s)" #define FUNC_ADPT_ARG(adapter) __func__, (adapter->pnetdev ? adapter->pnetdev->name : NULL) struct rtw_netdev_priv_indicator { void *priv; u32 sizeof_priv; }; struct net_device *rtw_alloc_etherdev_with_old_priv(int sizeof_priv, void *old_priv); extern struct net_device *rtw_alloc_etherdev(int sizeof_priv); #define STRUCT_PACKED __attribute__ ((packed)) #endif