include/linux/aio.h

   1 #ifndef __LINUX__AIO_H
   2 #define __LINUX__AIO_H
   3
   4 #include <linux/list.h>
   5 #include <linux/workqueue.h>
   6 #include <linux/aio_abi.h>
   7 #include <linux/uio.h>
   8 #include <linux/rcupdate.h>
   9
  10 #include <linux/atomic.h>
  11
  12 #define AIO_MAXSEGS             4
  13 #define AIO_KIOGRP_NR_ATOMIC    8
  14
  15 struct kioctx;
  16
  17 /* Notes on cancelling a kiocb:
  18  *      If a kiocb is cancelled, aio_complete may return 0 to indicate
  19  *      that cancel has not yet disposed of the kiocb.  All cancel
  20  *      operations *must* call aio_put_req to dispose of the kiocb
  21  *      to guard against races with the completion code.
  22  */
  23 #define KIOCB_C_CANCELLED       0x01
  24 #define KIOCB_C_COMPLETE        0x02
  25
  26 #define KIOCB_SYNC_KEY          (~0U)
  27
  28 /* ki_flags bits */
  29 /*
  30  * This may be used for cancel/retry serialization in the future, but
  31  * for now it's unused and we probably don't want modules to even
  32  * think they can use it.
  33  */
  34 /* #define KIF_LOCKED           0 */
  35 #define KIF_KICKED              1
  36 #define KIF_CANCELLED           2
  37
  38 #define kiocbTryLock(iocb)      test_and_set_bit(KIF_LOCKED, &(iocb)->ki_flags)
  39 #define kiocbTryKick(iocb)      test_and_set_bit(KIF_KICKED, &(iocb)->ki_flags)
  40
  41 #define kiocbSetLocked(iocb)    set_bit(KIF_LOCKED, &(iocb)->ki_flags)
  42 #define kiocbSetKicked(iocb)    set_bit(KIF_KICKED, &(iocb)->ki_flags)
  43 #define kiocbSetCancelled(iocb) set_bit(KIF_CANCELLED, &(iocb)->ki_flags)
  44
  45 #define kiocbClearLocked(iocb)  clear_bit(KIF_LOCKED, &(iocb)->ki_flags)
  46 #define kiocbClearKicked(iocb)  clear_bit(KIF_KICKED, &(iocb)->ki_flags)
  47 #define kiocbClearCancelled(iocb)       clear_bit(KIF_CANCELLED, &(iocb)->ki_flags)
  48
  49 #define kiocbIsLocked(iocb)     test_bit(KIF_LOCKED, &(iocb)->ki_flags)
  50 #define kiocbIsKicked(iocb)     test_bit(KIF_KICKED, &(iocb)->ki_flags)
  51 #define kiocbIsCancelled(iocb)  test_bit(KIF_CANCELLED, &(iocb)->ki_flags)
  52
  53 /* is there a better place to document function pointer methods? */
  54 /**
  55  * ki_retry     -       iocb forward progress callback
  56  * @kiocb:      The kiocb struct to advance by performing an operation.
  57  *
  58  * This callback is called when the AIO core wants a given AIO operation
  59  * to make forward progress.  The kiocb argument describes the operation
  60  * that is to be performed.  As the operation proceeds, perhaps partially,
  61  * ki_retry is expected to update the kiocb with progress made.  Typically
  62  * ki_retry is set in the AIO core and it itself calls file_operations
  63  * helpers.
  64  *
  65  * ki_retry's return value determines when the AIO operation is completed
  66  * and an event is generated in the AIO event ring.  Except the special
  67  * return values described below, the value that is returned from ki_retry
  68  * is transferred directly into the completion ring as the operation's
  69  * resulting status.  Once this has happened ki_retry *MUST NOT* reference
  70  * the kiocb pointer again.
  71  *
  72  * If ki_retry returns -EIOCBQUEUED it has made a promise that aio_complete()
  73  * will be called on the kiocb pointer in the future.  The AIO core will
  74  * not ask the method again -- ki_retry must ensure forward progress.
  75  * aio_complete() must be called once and only once in the future, multiple
  76  * calls may result in undefined behaviour.
  77  *
  78  * If ki_retry returns -EIOCBRETRY it has made a promise that kick_iocb()
  79  * will be called on the kiocb pointer in the future.  This may happen
  80  * through generic helpers that associate kiocb->ki_wait with a wait
  81  * queue head that ki_retry uses via current->io_wait.  It can also happen
  82  * with custom tracking and manual calls to kick_iocb(), though that is
  83  * discouraged.  In either case, kick_iocb() must be called once and only
  84  * once.  ki_retry must ensure forward progress, the AIO core will wait
  85  * indefinitely for kick_iocb() to be called.
  86  */
  87 struct kiocb {
  88         struct list_head        ki_run_list;
  89         unsigned long           ki_flags;
  90         int                     ki_users;
  91         unsigned                ki_key;         /* id of this request */
  92
  93         struct file             *ki_filp;
  94         struct kioctx           *ki_ctx;        /* may be NULL for sync ops */
  95         int                     (*ki_cancel)(struct kiocb *, struct io_event *);
  96         ssize_t                 (*ki_retry)(struct kiocb *);
  97         void                    (*ki_dtor)(struct kiocb *);
  98
  99         union {
 100                 void __user             *user;
 101                 struct task_struct      *tsk;
 102         } ki_obj;
 103
 104         __u64                   ki_user_data;   /* user's data for completion */
 105         loff_t                  ki_pos;
 106
 107         void                    *private;
 108         /* State that we remember to be able to restart/retry  */
 109         unsigned short          ki_opcode;
 110         size_t                  ki_nbytes;      /* copy of iocb->aio_nbytes */
 111         char                    __user *ki_buf; /* remaining iocb->aio_buf */
 112         size_t                  ki_left;        /* remaining bytes */
 113         struct iovec            ki_inline_vec;  /* inline vector */
 114         struct iovec            *ki_iovec;
 115         unsigned long           ki_nr_segs;
 116         unsigned long           ki_cur_seg;
 117
 118         struct list_head        ki_list;        /* the aio core uses this
 119                                                  * for cancellation */
 120         struct list_head        ki_batch;       /* batch allocation */
 121
 122         /*
 123          * If the aio_resfd field of the userspace iocb is not zero,
 124          * this is the underlying eventfd context to deliver events to.
 125          */
 126         struct eventfd_ctx      *ki_eventfd;
 127 };
 128
 129 static inline bool is_sync_kiocb(struct kiocb *kiocb)
 130 {
 131         return kiocb->ki_key == KIOCB_SYNC_KEY;
 132 }
 133
 134 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
 135 {
 136         *kiocb = (struct kiocb) {
 137                         .ki_users = 1,
 138                         .ki_key = KIOCB_SYNC_KEY,
 139                         .ki_filp = filp,
 140                         .ki_obj.tsk = current,
 141                 };
 142 }
 143
 144 #define AIO_RING_MAGIC                  0xa10a10a1
 145 #define AIO_RING_COMPAT_FEATURES        1
 146 #define AIO_RING_INCOMPAT_FEATURES      0
 147 struct aio_ring {
 148         unsigned        id;     /* kernel internal index number */
 149         unsigned        nr;     /* number of io_events */
 150         unsigned        head;
 151         unsigned        tail;
 152
 153         unsigned        magic;
 154         unsigned        compat_features;
 155         unsigned        incompat_features;
 156         unsigned        header_length;  /* size of aio_ring */
 157
 158
 159         struct io_event         io_events[0];
 160 }; /* 128 bytes + ring size */
 161
 162 #define AIO_RING_PAGES  8
 163 struct aio_ring_info {
 164         unsigned long           mmap_base;
 165         unsigned long           mmap_size;
 166
 167         struct page             **ring_pages;
 168         spinlock_t              ring_lock;
 169         long                    nr_pages;
 170
 171         unsigned                nr, tail;
 172
 173         struct page             *internal_pages[AIO_RING_PAGES];
 174 };
 175
 176 static inline unsigned aio_ring_avail(struct aio_ring_info *info,
 177                                         struct aio_ring *ring)
 178 {
 179         return (ring->head + info->nr - 1 - ring->tail) % info->nr;
 180 }
 181
 182 struct kioctx {
 183         atomic_t                users;
 184         int                     dead;
 185         struct mm_struct        *mm;
 186
 187         /* This needs improving */
 188         unsigned long           user_id;
 189         struct hlist_node       list;
 190
 191         wait_queue_head_t       wait;
 192
 193         spinlock_t              ctx_lock;
 194
 195         int                     reqs_active;
 196         struct list_head        active_reqs;    /* used for cancellation */
 197         struct list_head        run_list;       /* used for kicked reqs */
 198
 199         /* sys_io_setup currently limits this to an unsigned int */
 200         unsigned                max_reqs;
 201
 202         struct aio_ring_info    ring_info;
 203
 204         struct delayed_work     wq;
 205
 206         struct rcu_head         rcu_head;
 207 };
 208
 209 /* prototypes */
 210 extern unsigned aio_max_size;
 211
 212 #ifdef CONFIG_AIO
 213 extern ssize_t wait_on_sync_kiocb(struct kiocb *iocb);
 214 extern int aio_put_req(struct kiocb *iocb);
 215 extern void kick_iocb(struct kiocb *iocb);
 216 extern int aio_complete(struct kiocb *iocb, long res, long res2);
 217 struct mm_struct;
 218 extern void exit_aio(struct mm_struct *mm);
 219 extern long do_io_submit(aio_context_t ctx_id, long nr,
 220                          struct iocb __user *__user *iocbpp, bool compat);
 221 #else
 222 static inline ssize_t wait_on_sync_kiocb(struct kiocb *iocb) { return 0; }
 223 static inline int aio_put_req(struct kiocb *iocb) { return 0; }
 224 static inline void kick_iocb(struct kiocb *iocb) { }
 225 static inline int aio_complete(struct kiocb *iocb, long res, long res2) { return 0; }
 226 struct mm_struct;
 227 static inline void exit_aio(struct mm_struct *mm) { }
 228 static inline long do_io_submit(aio_context_t ctx_id, long nr,
 229                                 struct iocb __user * __user *iocbpp,
 230                                 bool compat) { return 0; }
 231 #endif /* CONFIG_AIO */
 232
 233 static inline struct kiocb *list_kiocb(struct list_head *h)
 234 {
 235         return list_entry(h, struct kiocb, ki_list);
 236 }
 237
 238 /* for sysctl: */
 239 extern unsigned long aio_nr;
 240 extern unsigned long aio_max_nr;
 241
 242 #endif /* __LINUX__AIO_H */