Blame view
fs/aio.c
47.2 KB
1da177e4c
|
1 2 3 4 5 6 7 8 9 10 |
/* * An async IO implementation for Linux * Written by Benjamin LaHaise <bcrl@kvack.org> * * Implements an efficient asynchronous io interface. * * Copyright 2000, 2001, 2002 Red Hat, Inc. All Rights Reserved. * * See ../COPYING for licensing terms. */ |
caf4167aa
|
11 |
#define pr_fmt(fmt) "%s: " fmt, __func__ |
1da177e4c
|
12 13 14 15 16 |
#include <linux/kernel.h> #include <linux/init.h> #include <linux/errno.h> #include <linux/time.h> #include <linux/aio_abi.h> |
630d9c472
|
17 |
#include <linux/export.h> |
1da177e4c
|
18 |
#include <linux/syscalls.h> |
b9d128f10
|
19 |
#include <linux/backing-dev.h> |
027445c37
|
20 |
#include <linux/uio.h> |
1da177e4c
|
21 |
|
174cd4b1e
|
22 |
#include <linux/sched/signal.h> |
1da177e4c
|
23 24 25 26 |
#include <linux/fs.h> #include <linux/file.h> #include <linux/mm.h> #include <linux/mman.h> |
3d2d827f5
|
27 |
#include <linux/mmu_context.h> |
e1bdd5f27
|
28 |
#include <linux/percpu.h> |
1da177e4c
|
29 30 31 32 33 34 |
#include <linux/slab.h> #include <linux/timer.h> #include <linux/aio.h> #include <linux/highmem.h> #include <linux/workqueue.h> #include <linux/security.h> |
9c3060bed
|
35 |
#include <linux/eventfd.h> |
cfb1e33ee
|
36 |
#include <linux/blkdev.h> |
9d85cba71
|
37 |
#include <linux/compat.h> |
36bc08cc0
|
38 39 |
#include <linux/migrate.h> #include <linux/ramfs.h> |
723be6e39
|
40 |
#include <linux/percpu-refcount.h> |
71ad7490c
|
41 |
#include <linux/mount.h> |
1da177e4c
|
42 43 |
#include <asm/kmap_types.h> |
7c0f6ba68
|
44 |
#include <linux/uaccess.h> |
1da177e4c
|
45 |
|
68d70d03f
|
46 |
#include "internal.h" |
4e179bca6
|
47 48 49 50 51 52 |
#define AIO_RING_MAGIC 0xa10a10a1 #define AIO_RING_COMPAT_FEATURES 1 #define AIO_RING_INCOMPAT_FEATURES 0 struct aio_ring { unsigned id; /* kernel internal index number */ unsigned nr; /* number of io_events */ |
fa8a53c39
|
53 54 |
unsigned head; /* Written to by userland or under ring_lock * mutex by aio_read_events_ring(). */ |
4e179bca6
|
55 56 57 58 59 60 61 62 63 64 65 66 |
unsigned tail; unsigned magic; unsigned compat_features; unsigned incompat_features; unsigned header_length; /* size of aio_ring */ struct io_event io_events[0]; }; /* 128 bytes + ring size */ #define AIO_RING_PAGES 8 |
4e179bca6
|
67 |
|
db446a08c
|
68 |
struct kioctx_table { |
cd21b3400
|
69 70 71 |
struct rcu_head rcu; unsigned nr; struct kioctx __rcu *table[]; |
db446a08c
|
72 |
}; |
e1bdd5f27
|
73 74 75 |
struct kioctx_cpu { unsigned reqs_available; }; |
dc48e56d7
|
76 77 78 79 |
struct ctx_rq_wait { struct completion comp; atomic_t count; }; |
4e179bca6
|
80 |
struct kioctx { |
723be6e39
|
81 |
struct percpu_ref users; |
36f558890
|
82 |
atomic_t dead; |
4e179bca6
|
83 |
|
e34ecee2a
|
84 |
struct percpu_ref reqs; |
4e179bca6
|
85 |
unsigned long user_id; |
4e179bca6
|
86 |
|
e1bdd5f27
|
87 88 89 90 91 92 93 |
struct __percpu kioctx_cpu *cpu; /* * For percpu reqs_available, number of slots we move to/from global * counter at a time: */ unsigned req_batch; |
3e845ce01
|
94 95 96 97 |
/* * This is what userspace passed to io_setup(), it's not used for * anything but counting against the global max_reqs quota. * |
58c85dc20
|
98 |
* The real limit is nr_events - 1, which will be larger (see |
3e845ce01
|
99 100 |
* aio_setup_ring()) */ |
4e179bca6
|
101 |
unsigned max_reqs; |
58c85dc20
|
102 103 |
/* Size of ringbuffer, in units of struct io_event */ unsigned nr_events; |
4e179bca6
|
104 |
|
58c85dc20
|
105 106 107 108 109 |
unsigned long mmap_base; unsigned long mmap_size; struct page **ring_pages; long nr_pages; |
076c7c068
|
110 111 |
struct rcu_head free_rcu; struct work_struct free_work; /* see free_ioctx() */ |
4e23bcaeb
|
112 |
|
e02ba72aa
|
113 114 115 |
/* * signals when all in-flight requests are done */ |
dc48e56d7
|
116 |
struct ctx_rq_wait *rq_wait; |
e02ba72aa
|
117 |
|
4e23bcaeb
|
118 |
struct { |
34e83fc61
|
119 120 121 122 123 |
/* * This counts the number of available slots in the ringbuffer, * so we avoid overflowing it: it's decremented (if positive) * when allocating a kiocb and incremented when the resulting * io_event is pulled off the ringbuffer. |
e1bdd5f27
|
124 125 |
* * We batch accesses to it with a percpu version. |
34e83fc61
|
126 127 |
*/ atomic_t reqs_available; |
4e23bcaeb
|
128 129 130 131 132 133 |
} ____cacheline_aligned_in_smp; struct { spinlock_t ctx_lock; struct list_head active_reqs; /* used for cancellation */ } ____cacheline_aligned_in_smp; |
58c85dc20
|
134 135 |
struct { struct mutex ring_lock; |
4e23bcaeb
|
136 137 |
wait_queue_head_t wait; } ____cacheline_aligned_in_smp; |
58c85dc20
|
138 139 140 |
struct { unsigned tail; |
d856f32a8
|
141 |
unsigned completed_events; |
58c85dc20
|
142 |
spinlock_t completion_lock; |
4e23bcaeb
|
143 |
} ____cacheline_aligned_in_smp; |
58c85dc20
|
144 145 |
struct page *internal_pages[AIO_RING_PAGES]; |
36bc08cc0
|
146 |
struct file *aio_ring_file; |
db446a08c
|
147 148 |
unsigned id; |
4e179bca6
|
149 |
}; |
04b2fa9f8
|
150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 |
/* * We use ki_cancel == KIOCB_CANCELLED to indicate that a kiocb has been either * cancelled or completed (this makes a certain amount of sense because * successful cancellation - io_cancel() - does deliver the completion to * userspace). * * And since most things don't implement kiocb cancellation and we'd really like * kiocb completion to be lockless when possible, we use ki_cancel to * synchronize cancellation and completion - we only set it to KIOCB_CANCELLED * with xchg() or cmpxchg(), see batch_complete_aio() and kiocb_cancel(). */ #define KIOCB_CANCELLED ((void *) (~0ULL)) struct aio_kiocb { struct kiocb common; struct kioctx *ki_ctx; kiocb_cancel_fn *ki_cancel; struct iocb __user *ki_user_iocb; /* user's aiocb */ __u64 ki_user_data; /* user's data for completion */ struct list_head ki_list; /* the aio core uses this * for cancellation */ /* * If the aio_resfd field of the userspace iocb is not zero, * this is the underlying eventfd context to deliver events to. */ struct eventfd_ctx *ki_eventfd; }; |
1da177e4c
|
181 |
/*------ sysctl variables----*/ |
d55b5fdaf
|
182 183 184 |
static DEFINE_SPINLOCK(aio_nr_lock); unsigned long aio_nr; /* current system wide number of aio requests */ unsigned long aio_max_nr = 0x10000; /* system wide maximum number of aio requests */ |
1da177e4c
|
185 |
/*----end sysctl variables---*/ |
e18b890bb
|
186 187 |
static struct kmem_cache *kiocb_cachep; static struct kmem_cache *kioctx_cachep; |
1da177e4c
|
188 |
|
71ad7490c
|
189 190 191 192 193 194 195 196 197 198 199 |
static struct vfsmount *aio_mnt; static const struct file_operations aio_ring_fops; static const struct address_space_operations aio_ctx_aops; static struct file *aio_private_file(struct kioctx *ctx, loff_t nr_pages) { struct qstr this = QSTR_INIT("[aio]", 5); struct file *file; struct path path; struct inode *inode = alloc_anon_inode(aio_mnt->mnt_sb); |
7f62656be
|
200 201 |
if (IS_ERR(inode)) return ERR_CAST(inode); |
71ad7490c
|
202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 |
inode->i_mapping->a_ops = &aio_ctx_aops; inode->i_mapping->private_data = ctx; inode->i_size = PAGE_SIZE * nr_pages; path.dentry = d_alloc_pseudo(aio_mnt->mnt_sb, &this); if (!path.dentry) { iput(inode); return ERR_PTR(-ENOMEM); } path.mnt = mntget(aio_mnt); d_instantiate(path.dentry, inode); file = alloc_file(&path, FMODE_READ | FMODE_WRITE, &aio_ring_fops); if (IS_ERR(file)) { path_put(&path); return file; } file->f_flags = O_RDWR; |
71ad7490c
|
222 223 224 225 226 227 228 229 230 |
return file; } static struct dentry *aio_mount(struct file_system_type *fs_type, int flags, const char *dev_name, void *data) { static const struct dentry_operations ops = { .d_dname = simple_dname, }; |
22f6b4d34
|
231 232 233 234 235 236 |
struct dentry *root = mount_pseudo(fs_type, "aio:", NULL, &ops, AIO_RING_MAGIC); if (!IS_ERR(root)) root->d_sb->s_iflags |= SB_I_NOEXEC; return root; |
71ad7490c
|
237 |
} |
1da177e4c
|
238 239 240 241 242 243 |
/* aio_setup * Creates the slab caches used by the aio routines, panic on * failure as this is done early during the boot sequence. */ static int __init aio_setup(void) { |
71ad7490c
|
244 245 246 247 248 249 250 251 |
static struct file_system_type aio_fs = { .name = "aio", .mount = aio_mount, .kill_sb = kill_anon_super, }; aio_mnt = kern_mount(&aio_fs); if (IS_ERR(aio_mnt)) panic("Failed to create aio fs mount."); |
04b2fa9f8
|
252 |
kiocb_cachep = KMEM_CACHE(aio_kiocb, SLAB_HWCACHE_ALIGN|SLAB_PANIC); |
0a31bd5f2
|
253 |
kioctx_cachep = KMEM_CACHE(kioctx,SLAB_HWCACHE_ALIGN|SLAB_PANIC); |
1da177e4c
|
254 |
|
caf4167aa
|
255 256 |
pr_debug("sizeof(struct page) = %zu ", sizeof(struct page)); |
1da177e4c
|
257 258 259 |
return 0; } |
385773e04
|
260 |
__initcall(aio_setup); |
1da177e4c
|
261 |
|
5e9ae2e5d
|
262 263 264 |
static void put_aio_ring_file(struct kioctx *ctx) { struct file *aio_ring_file = ctx->aio_ring_file; |
de04e7693
|
265 |
struct address_space *i_mapping; |
5e9ae2e5d
|
266 |
if (aio_ring_file) { |
450630975
|
267 |
truncate_setsize(file_inode(aio_ring_file), 0); |
5e9ae2e5d
|
268 269 |
/* Prevent further access to the kioctx from migratepages */ |
450630975
|
270 |
i_mapping = aio_ring_file->f_mapping; |
de04e7693
|
271 272 |
spin_lock(&i_mapping->private_lock); i_mapping->private_data = NULL; |
5e9ae2e5d
|
273 |
ctx->aio_ring_file = NULL; |
de04e7693
|
274 |
spin_unlock(&i_mapping->private_lock); |
5e9ae2e5d
|
275 276 277 278 |
fput(aio_ring_file); } } |
1da177e4c
|
279 280 |
static void aio_free_ring(struct kioctx *ctx) { |
36bc08cc0
|
281 |
int i; |
1da177e4c
|
282 |
|
fa8a53c39
|
283 284 285 286 |
/* Disconnect the kiotx from the ring file. This prevents future * accesses to the kioctx from page migration. */ put_aio_ring_file(ctx); |
36bc08cc0
|
287 |
for (i = 0; i < ctx->nr_pages; i++) { |
8e321fefb
|
288 |
struct page *page; |
36bc08cc0
|
289 290 291 |
pr_debug("pid(%d) [%d] page->count=%d ", current->pid, i, page_count(ctx->ring_pages[i])); |
8e321fefb
|
292 293 294 295 296 |
page = ctx->ring_pages[i]; if (!page) continue; ctx->ring_pages[i] = NULL; put_page(page); |
36bc08cc0
|
297 |
} |
1da177e4c
|
298 |
|
ddb8c45ba
|
299 |
if (ctx->ring_pages && ctx->ring_pages != ctx->internal_pages) { |
58c85dc20
|
300 |
kfree(ctx->ring_pages); |
ddb8c45ba
|
301 302 |
ctx->ring_pages = NULL; } |
36bc08cc0
|
303 |
} |
5477e70a6
|
304 |
static int aio_ring_mremap(struct vm_area_struct *vma) |
e4a0d3e72
|
305 |
{ |
5477e70a6
|
306 |
struct file *file = vma->vm_file; |
e4a0d3e72
|
307 308 |
struct mm_struct *mm = vma->vm_mm; struct kioctx_table *table; |
b2edffdd9
|
309 |
int i, res = -EINVAL; |
e4a0d3e72
|
310 311 312 313 314 315 |
spin_lock(&mm->ioctx_lock); rcu_read_lock(); table = rcu_dereference(mm->ioctx_table); for (i = 0; i < table->nr; i++) { struct kioctx *ctx; |
cd21b3400
|
316 |
ctx = rcu_dereference(table->table[i]); |
e4a0d3e72
|
317 |
if (ctx && ctx->aio_ring_file == file) { |
b2edffdd9
|
318 319 320 321 |
if (!atomic_read(&ctx->dead)) { ctx->user_id = ctx->mmap_base = vma->vm_start; res = 0; } |
e4a0d3e72
|
322 323 324 325 326 327 |
break; } } rcu_read_unlock(); spin_unlock(&mm->ioctx_lock); |
b2edffdd9
|
328 |
return res; |
e4a0d3e72
|
329 |
} |
5477e70a6
|
330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 |
static const struct vm_operations_struct aio_ring_vm_ops = { .mremap = aio_ring_mremap, #if IS_ENABLED(CONFIG_MMU) .fault = filemap_fault, .map_pages = filemap_map_pages, .page_mkwrite = filemap_page_mkwrite, #endif }; static int aio_ring_mmap(struct file *file, struct vm_area_struct *vma) { vma->vm_flags |= VM_DONTEXPAND; vma->vm_ops = &aio_ring_vm_ops; return 0; } |
36bc08cc0
|
345 346 347 |
static const struct file_operations aio_ring_fops = { .mmap = aio_ring_mmap, }; |
0c45355fc
|
348 |
#if IS_ENABLED(CONFIG_MIGRATION) |
36bc08cc0
|
349 350 351 |
static int aio_migratepage(struct address_space *mapping, struct page *new, struct page *old, enum migrate_mode mode) { |
5e9ae2e5d
|
352 |
struct kioctx *ctx; |
36bc08cc0
|
353 |
unsigned long flags; |
fa8a53c39
|
354 |
pgoff_t idx; |
36bc08cc0
|
355 |
int rc; |
2916ecc0f
|
356 357 358 359 360 361 362 |
/* * We cannot support the _NO_COPY case here, because copy needs to * happen under the ctx->completion_lock. That does not work with the * migration workflow of MIGRATE_SYNC_NO_COPY. */ if (mode == MIGRATE_SYNC_NO_COPY) return -EINVAL; |
8e321fefb
|
363 |
rc = 0; |
fa8a53c39
|
364 |
/* mapping->private_lock here protects against the kioctx teardown. */ |
8e321fefb
|
365 366 |
spin_lock(&mapping->private_lock); ctx = mapping->private_data; |
fa8a53c39
|
367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 |
if (!ctx) { rc = -EINVAL; goto out; } /* The ring_lock mutex. The prevents aio_read_events() from writing * to the ring's head, and prevents page migration from mucking in * a partially initialized kiotx. */ if (!mutex_trylock(&ctx->ring_lock)) { rc = -EAGAIN; goto out; } idx = old->index; if (idx < (pgoff_t)ctx->nr_pages) { /* Make sure the old page hasn't already been changed */ if (ctx->ring_pages[idx] != old) rc = -EAGAIN; |
8e321fefb
|
386 387 |
} else rc = -EINVAL; |
8e321fefb
|
388 389 |
if (rc != 0) |
fa8a53c39
|
390 |
goto out_unlock; |
8e321fefb
|
391 |
|
36bc08cc0
|
392 393 |
/* Writeback must be complete */ BUG_ON(PageWriteback(old)); |
8e321fefb
|
394 |
get_page(new); |
36bc08cc0
|
395 |
|
8e321fefb
|
396 |
rc = migrate_page_move_mapping(mapping, new, old, NULL, mode, 1); |
36bc08cc0
|
397 |
if (rc != MIGRATEPAGE_SUCCESS) { |
8e321fefb
|
398 |
put_page(new); |
fa8a53c39
|
399 |
goto out_unlock; |
36bc08cc0
|
400 |
} |
fa8a53c39
|
401 402 403 |
/* Take completion_lock to prevent other writes to the ring buffer * while the old page is copied to the new. This prevents new * events from being lost. |
5e9ae2e5d
|
404 |
*/ |
fa8a53c39
|
405 406 407 408 409 |
spin_lock_irqsave(&ctx->completion_lock, flags); migrate_page_copy(new, old); BUG_ON(ctx->ring_pages[idx] != old); ctx->ring_pages[idx] = new; spin_unlock_irqrestore(&ctx->completion_lock, flags); |
36bc08cc0
|
410 |
|
fa8a53c39
|
411 412 |
/* The old page is no longer accessible. */ put_page(old); |
8e321fefb
|
413 |
|
fa8a53c39
|
414 415 416 417 |
out_unlock: mutex_unlock(&ctx->ring_lock); out: spin_unlock(&mapping->private_lock); |
36bc08cc0
|
418 |
return rc; |
1da177e4c
|
419 |
} |
0c45355fc
|
420 |
#endif |
1da177e4c
|
421 |
|
36bc08cc0
|
422 |
static const struct address_space_operations aio_ctx_aops = { |
835f252c6
|
423 |
.set_page_dirty = __set_page_dirty_no_writeback, |
0c45355fc
|
424 |
#if IS_ENABLED(CONFIG_MIGRATION) |
36bc08cc0
|
425 |
.migratepage = aio_migratepage, |
0c45355fc
|
426 |
#endif |
36bc08cc0
|
427 |
}; |
2a8a98673
|
428 |
static int aio_setup_ring(struct kioctx *ctx, unsigned int nr_events) |
1da177e4c
|
429 430 |
{ struct aio_ring *ring; |
41003a7bc
|
431 |
struct mm_struct *mm = current->mm; |
3dc9acb67
|
432 |
unsigned long size, unused; |
1da177e4c
|
433 |
int nr_pages; |
36bc08cc0
|
434 435 |
int i; struct file *file; |
1da177e4c
|
436 437 438 439 440 441 |
/* Compensate for the ring buffer's head/tail overlap entry */ nr_events += 2; /* 1 is required, 2 for good luck */ size = sizeof(struct aio_ring); size += sizeof(struct io_event) * nr_events; |
1da177e4c
|
442 |
|
36bc08cc0
|
443 |
nr_pages = PFN_UP(size); |
1da177e4c
|
444 445 |
if (nr_pages < 0) return -EINVAL; |
71ad7490c
|
446 |
file = aio_private_file(ctx, nr_pages); |
36bc08cc0
|
447 448 |
if (IS_ERR(file)) { ctx->aio_ring_file = NULL; |
fa8a53c39
|
449 |
return -ENOMEM; |
36bc08cc0
|
450 |
} |
3dc9acb67
|
451 452 453 454 455 456 457 458 459 460 461 462 463 |
ctx->aio_ring_file = file; nr_events = (PAGE_SIZE * nr_pages - sizeof(struct aio_ring)) / sizeof(struct io_event); ctx->ring_pages = ctx->internal_pages; if (nr_pages > AIO_RING_PAGES) { ctx->ring_pages = kcalloc(nr_pages, sizeof(struct page *), GFP_KERNEL); if (!ctx->ring_pages) { put_aio_ring_file(ctx); return -ENOMEM; } } |
36bc08cc0
|
464 465 |
for (i = 0; i < nr_pages; i++) { struct page *page; |
450630975
|
466 |
page = find_or_create_page(file->f_mapping, |
36bc08cc0
|
467 468 469 470 471 472 473 |
i, GFP_HIGHUSER | __GFP_ZERO); if (!page) break; pr_debug("pid(%d) page[%d]->count=%d ", current->pid, i, page_count(page)); SetPageUptodate(page); |
36bc08cc0
|
474 |
unlock_page(page); |
3dc9acb67
|
475 476 |
ctx->ring_pages[i] = page; |
36bc08cc0
|
477 |
} |
3dc9acb67
|
478 |
ctx->nr_pages = i; |
1da177e4c
|
479 |
|
3dc9acb67
|
480 481 |
if (unlikely(i != nr_pages)) { aio_free_ring(ctx); |
fa8a53c39
|
482 |
return -ENOMEM; |
1da177e4c
|
483 |
} |
58c85dc20
|
484 485 486 |
ctx->mmap_size = nr_pages * PAGE_SIZE; pr_debug("attempting mmap of %lu bytes ", ctx->mmap_size); |
36bc08cc0
|
487 |
|
013373e8b
|
488 489 490 491 492 |
if (down_write_killable(&mm->mmap_sem)) { ctx->mmap_size = 0; aio_free_ring(ctx); return -EINTR; } |
36bc08cc0
|
493 494 |
ctx->mmap_base = do_mmap_pgoff(ctx->aio_ring_file, 0, ctx->mmap_size, PROT_READ | PROT_WRITE, |
897ab3e0c
|
495 |
MAP_SHARED, 0, &unused, NULL); |
3dc9acb67
|
496 |
up_write(&mm->mmap_sem); |
58c85dc20
|
497 |
if (IS_ERR((void *)ctx->mmap_base)) { |
58c85dc20
|
498 |
ctx->mmap_size = 0; |
1da177e4c
|
499 |
aio_free_ring(ctx); |
fa8a53c39
|
500 |
return -ENOMEM; |
1da177e4c
|
501 |
} |
58c85dc20
|
502 503 |
pr_debug("mmap address: 0x%08lx ", ctx->mmap_base); |
d6c355c7d
|
504 |
|
58c85dc20
|
505 506 |
ctx->user_id = ctx->mmap_base; ctx->nr_events = nr_events; /* trusted copy */ |
1da177e4c
|
507 |
|
58c85dc20
|
508 |
ring = kmap_atomic(ctx->ring_pages[0]); |
1da177e4c
|
509 |
ring->nr = nr_events; /* user copy */ |
db446a08c
|
510 |
ring->id = ~0U; |
1da177e4c
|
511 512 513 514 515 |
ring->head = ring->tail = 0; ring->magic = AIO_RING_MAGIC; ring->compat_features = AIO_RING_COMPAT_FEATURES; ring->incompat_features = AIO_RING_INCOMPAT_FEATURES; ring->header_length = sizeof(struct aio_ring); |
e8e3c3d66
|
516 |
kunmap_atomic(ring); |
58c85dc20
|
517 |
flush_dcache_page(ctx->ring_pages[0]); |
1da177e4c
|
518 519 520 |
return 0; } |
1da177e4c
|
521 522 523 |
#define AIO_EVENTS_PER_PAGE (PAGE_SIZE / sizeof(struct io_event)) #define AIO_EVENTS_FIRST_PAGE ((PAGE_SIZE - sizeof(struct aio_ring)) / sizeof(struct io_event)) #define AIO_EVENTS_OFFSET (AIO_EVENTS_PER_PAGE - AIO_EVENTS_FIRST_PAGE) |
04b2fa9f8
|
524 |
void kiocb_set_cancel_fn(struct kiocb *iocb, kiocb_cancel_fn *cancel) |
0460fef2a
|
525 |
{ |
04b2fa9f8
|
526 |
struct aio_kiocb *req = container_of(iocb, struct aio_kiocb, common); |
0460fef2a
|
527 528 529 530 531 532 533 534 535 536 537 538 539 |
struct kioctx *ctx = req->ki_ctx; unsigned long flags; spin_lock_irqsave(&ctx->ctx_lock, flags); if (!req->ki_list.next) list_add(&req->ki_list, &ctx->active_reqs); req->ki_cancel = cancel; spin_unlock_irqrestore(&ctx->ctx_lock, flags); } EXPORT_SYMBOL(kiocb_set_cancel_fn); |
04b2fa9f8
|
540 |
static int kiocb_cancel(struct aio_kiocb *kiocb) |
906b973cf
|
541 |
{ |
0460fef2a
|
542 |
kiocb_cancel_fn *old, *cancel; |
906b973cf
|
543 |
|
0460fef2a
|
544 545 546 547 548 549 550 551 |
/* * Don't want to set kiocb->ki_cancel = KIOCB_CANCELLED unless it * actually has a cancel function, hence the cmpxchg() */ cancel = ACCESS_ONCE(kiocb->ki_cancel); do { if (!cancel || cancel == KIOCB_CANCELLED) |
57282d8fd
|
552 |
return -EINVAL; |
906b973cf
|
553 |
|
0460fef2a
|
554 555 556 |
old = cancel; cancel = cmpxchg(&kiocb->ki_cancel, old, KIOCB_CANCELLED); } while (cancel != old); |
906b973cf
|
557 |
|
04b2fa9f8
|
558 |
return cancel(&kiocb->common); |
906b973cf
|
559 |
} |
076c7c068
|
560 561 562 563 564 565 |
/* * free_ioctx() should be RCU delayed to synchronize against the RCU * protected lookup_ioctx() and also needs process context to call * aio_free_ring(), so the double bouncing through kioctx->free_rcu and * ->free_work. */ |
e34ecee2a
|
566 |
static void free_ioctx(struct work_struct *work) |
36f558890
|
567 |
{ |
e34ecee2a
|
568 |
struct kioctx *ctx = container_of(work, struct kioctx, free_work); |
e1bdd5f27
|
569 |
|
e34ecee2a
|
570 571 |
pr_debug("freeing %p ", ctx); |
e1bdd5f27
|
572 |
|
e34ecee2a
|
573 |
aio_free_ring(ctx); |
e1bdd5f27
|
574 |
free_percpu(ctx->cpu); |
9a1049da9
|
575 576 |
percpu_ref_exit(&ctx->reqs); percpu_ref_exit(&ctx->users); |
36f558890
|
577 578 |
kmem_cache_free(kioctx_cachep, ctx); } |
076c7c068
|
579 580 581 582 583 584 585 |
static void free_ioctx_rcufn(struct rcu_head *head) { struct kioctx *ctx = container_of(head, struct kioctx, free_rcu); INIT_WORK(&ctx->free_work, free_ioctx); schedule_work(&ctx->free_work); } |
e34ecee2a
|
586 587 588 |
static void free_ioctx_reqs(struct percpu_ref *ref) { struct kioctx *ctx = container_of(ref, struct kioctx, reqs); |
e02ba72aa
|
589 |
/* At this point we know that there are no any in-flight requests */ |
dc48e56d7
|
590 591 |
if (ctx->rq_wait && atomic_dec_and_test(&ctx->rq_wait->count)) complete(&ctx->rq_wait->comp); |
e02ba72aa
|
592 |
|
076c7c068
|
593 594 |
/* Synchronize against RCU protected table->table[] dereferences */ call_rcu(&ctx->free_rcu, free_ioctx_rcufn); |
e34ecee2a
|
595 |
} |
36f558890
|
596 597 598 599 600 |
/* * When this function runs, the kioctx has been removed from the "hash table" * and ctx->users has dropped to 0, so we know no more kiocbs can be submitted - * now it's safe to cancel any that need to be. */ |
e34ecee2a
|
601 |
static void free_ioctx_users(struct percpu_ref *ref) |
36f558890
|
602 |
{ |
e34ecee2a
|
603 |
struct kioctx *ctx = container_of(ref, struct kioctx, users); |
04b2fa9f8
|
604 |
struct aio_kiocb *req; |
36f558890
|
605 606 607 608 609 |
spin_lock_irq(&ctx->ctx_lock); while (!list_empty(&ctx->active_reqs)) { req = list_first_entry(&ctx->active_reqs, |
04b2fa9f8
|
610 |
struct aio_kiocb, ki_list); |
d52a8f9ea
|
611 |
kiocb_cancel(req); |
6a19487d5
|
612 |
list_del_init(&req->ki_list); |
36f558890
|
613 614 615 |
} spin_unlock_irq(&ctx->ctx_lock); |
e34ecee2a
|
616 617 |
percpu_ref_kill(&ctx->reqs); percpu_ref_put(&ctx->reqs); |
36f558890
|
618 |
} |
db446a08c
|
619 620 621 622 623 624 625 |
static int ioctx_add_table(struct kioctx *ctx, struct mm_struct *mm) { unsigned i, new_nr; struct kioctx_table *table, *old; struct aio_ring *ring; spin_lock(&mm->ioctx_lock); |
855ef0dec
|
626 |
table = rcu_dereference_raw(mm->ioctx_table); |
db446a08c
|
627 628 629 630 |
while (1) { if (table) for (i = 0; i < table->nr; i++) |
cd21b3400
|
631 |
if (!rcu_access_pointer(table->table[i])) { |
db446a08c
|
632 |
ctx->id = i; |
cd21b3400
|
633 |
rcu_assign_pointer(table->table[i], ctx); |
db446a08c
|
634 |
spin_unlock(&mm->ioctx_lock); |
fa8a53c39
|
635 636 637 638 |
/* While kioctx setup is in progress, * we are protected from page migration * changes ring_pages by ->ring_lock. */ |
db446a08c
|
639 640 641 642 643 644 645 |
ring = kmap_atomic(ctx->ring_pages[0]); ring->id = ctx->id; kunmap_atomic(ring); return 0; } new_nr = (table ? table->nr : 1) * 4; |
db446a08c
|
646 647 648 649 650 651 652 653 654 655 |
spin_unlock(&mm->ioctx_lock); table = kzalloc(sizeof(*table) + sizeof(struct kioctx *) * new_nr, GFP_KERNEL); if (!table) return -ENOMEM; table->nr = new_nr; spin_lock(&mm->ioctx_lock); |
855ef0dec
|
656 |
old = rcu_dereference_raw(mm->ioctx_table); |
db446a08c
|
657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 |
if (!old) { rcu_assign_pointer(mm->ioctx_table, table); } else if (table->nr > old->nr) { memcpy(table->table, old->table, old->nr * sizeof(struct kioctx *)); rcu_assign_pointer(mm->ioctx_table, table); kfree_rcu(old, rcu); } else { kfree(table); table = old; } } } |
e34ecee2a
|
672 673 674 675 676 677 678 679 680 |
static void aio_nr_sub(unsigned nr) { spin_lock(&aio_nr_lock); if (WARN_ON(aio_nr - nr > aio_nr)) aio_nr = 0; else aio_nr -= nr; spin_unlock(&aio_nr_lock); } |
1da177e4c
|
681 682 683 684 685 |
/* ioctx_alloc * Allocates and initializes an ioctx. Returns an ERR_PTR if it failed. */ static struct kioctx *ioctx_alloc(unsigned nr_events) { |
41003a7bc
|
686 |
struct mm_struct *mm = current->mm; |
1da177e4c
|
687 |
struct kioctx *ctx; |
e23754f88
|
688 |
int err = -ENOMEM; |
1da177e4c
|
689 |
|
e1bdd5f27
|
690 |
/* |
2a8a98673
|
691 692 693 694 695 696 |
* Store the original nr_events -- what userspace passed to io_setup(), * for counting against the global limit -- before it changes. */ unsigned int max_reqs = nr_events; /* |
e1bdd5f27
|
697 698 699 700 701 702 703 704 705 706 |
* We keep track of the number of available ringbuffer slots, to prevent * overflow (reqs_available), and we also use percpu counters for this. * * So since up to half the slots might be on other cpu's percpu counters * and unavailable, double nr_events so userspace sees what they * expected: additionally, we move req_batch slots to/from percpu * counters at a time, so make sure that isn't 0: */ nr_events = max(nr_events, num_possible_cpus() * 4); nr_events *= 2; |
1da177e4c
|
707 |
/* Prevent overflows */ |
08397acdd
|
708 |
if (nr_events > (0x10000000U / sizeof(struct io_event))) { |
1da177e4c
|
709 710 711 712 |
pr_debug("ENOMEM: nr_events too high "); return ERR_PTR(-EINVAL); } |
2a8a98673
|
713 |
if (!nr_events || (unsigned long)max_reqs > aio_max_nr) |
1da177e4c
|
714 |
return ERR_PTR(-EAGAIN); |
c37622296
|
715 |
ctx = kmem_cache_zalloc(kioctx_cachep, GFP_KERNEL); |
1da177e4c
|
716 717 |
if (!ctx) return ERR_PTR(-ENOMEM); |
2a8a98673
|
718 |
ctx->max_reqs = max_reqs; |
1da177e4c
|
719 |
|
1da177e4c
|
720 |
spin_lock_init(&ctx->ctx_lock); |
0460fef2a
|
721 |
spin_lock_init(&ctx->completion_lock); |
58c85dc20
|
722 |
mutex_init(&ctx->ring_lock); |
fa8a53c39
|
723 724 725 |
/* Protect against page migration throughout kiotx setup by keeping * the ring_lock mutex held until setup is complete. */ mutex_lock(&ctx->ring_lock); |
1da177e4c
|
726 727 728 |
init_waitqueue_head(&ctx->wait); INIT_LIST_HEAD(&ctx->active_reqs); |
1da177e4c
|
729 |
|
2aad2a86f
|
730 |
if (percpu_ref_init(&ctx->users, free_ioctx_users, 0, GFP_KERNEL)) |
fa8a53c39
|
731 |
goto err; |
2aad2a86f
|
732 |
if (percpu_ref_init(&ctx->reqs, free_ioctx_reqs, 0, GFP_KERNEL)) |
fa8a53c39
|
733 |
goto err; |
e1bdd5f27
|
734 735 |
ctx->cpu = alloc_percpu(struct kioctx_cpu); if (!ctx->cpu) |
e34ecee2a
|
736 |
goto err; |
1da177e4c
|
737 |
|
2a8a98673
|
738 |
err = aio_setup_ring(ctx, nr_events); |
fa8a53c39
|
739 |
if (err < 0) |
e34ecee2a
|
740 |
goto err; |
e1bdd5f27
|
741 |
|
34e83fc61
|
742 |
atomic_set(&ctx->reqs_available, ctx->nr_events - 1); |
e1bdd5f27
|
743 |
ctx->req_batch = (ctx->nr_events - 1) / (num_possible_cpus() * 4); |
6878ea72a
|
744 745 |
if (ctx->req_batch < 1) ctx->req_batch = 1; |
34e83fc61
|
746 |
|
1da177e4c
|
747 |
/* limit the number of system wide aios */ |
9fa1cb397
|
748 |
spin_lock(&aio_nr_lock); |
2a8a98673
|
749 750 |
if (aio_nr + ctx->max_reqs > aio_max_nr || aio_nr + ctx->max_reqs < aio_nr) { |
9fa1cb397
|
751 |
spin_unlock(&aio_nr_lock); |
e34ecee2a
|
752 |
err = -EAGAIN; |
d1b943271
|
753 |
goto err_ctx; |
2dd542b7a
|
754 755 |
} aio_nr += ctx->max_reqs; |
9fa1cb397
|
756 |
spin_unlock(&aio_nr_lock); |
1da177e4c
|
757 |
|
1881686f8
|
758 759 |
percpu_ref_get(&ctx->users); /* io_setup() will drop this ref */ percpu_ref_get(&ctx->reqs); /* free_ioctx_users() will drop this */ |
723be6e39
|
760 |
|
da90382c2
|
761 762 |
err = ioctx_add_table(ctx, mm); if (err) |
e34ecee2a
|
763 |
goto err_cleanup; |
da90382c2
|
764 |
|
fa8a53c39
|
765 766 |
/* Release the ring_lock mutex now that all setup is complete. */ mutex_unlock(&ctx->ring_lock); |
caf4167aa
|
767 768 |
pr_debug("allocated ioctx %p[%ld]: mm=%p mask=0x%x ", |
58c85dc20
|
769 |
ctx, ctx->user_id, mm, ctx->nr_events); |
1da177e4c
|
770 |
return ctx; |
e34ecee2a
|
771 772 |
err_cleanup: aio_nr_sub(ctx->max_reqs); |
d1b943271
|
773 |
err_ctx: |
deeb8525f
|
774 775 776 |
atomic_set(&ctx->dead, 1); if (ctx->mmap_size) vm_munmap(ctx->mmap_base, ctx->mmap_size); |
d1b943271
|
777 |
aio_free_ring(ctx); |
e34ecee2a
|
778 |
err: |
fa8a53c39
|
779 |
mutex_unlock(&ctx->ring_lock); |
e1bdd5f27
|
780 |
free_percpu(ctx->cpu); |
9a1049da9
|
781 782 |
percpu_ref_exit(&ctx->reqs); percpu_ref_exit(&ctx->users); |
1da177e4c
|
783 |
kmem_cache_free(kioctx_cachep, ctx); |
caf4167aa
|
784 785 |
pr_debug("error allocating ioctx %d ", err); |
e23754f88
|
786 |
return ERR_PTR(err); |
1da177e4c
|
787 |
} |
36f558890
|
788 789 790 791 792 |
/* kill_ioctx * Cancels all outstanding aio requests on an aio context. Used * when the processes owning a context have all exited to encourage * the rapid destruction of the kioctx. */ |
fb2d44838
|
793 |
static int kill_ioctx(struct mm_struct *mm, struct kioctx *ctx, |
dc48e56d7
|
794 |
struct ctx_rq_wait *wait) |
36f558890
|
795 |
{ |
fa88b6f88
|
796 |
struct kioctx_table *table; |
db446a08c
|
797 |
|
b2edffdd9
|
798 799 800 |
spin_lock(&mm->ioctx_lock); if (atomic_xchg(&ctx->dead, 1)) { spin_unlock(&mm->ioctx_lock); |
fa88b6f88
|
801 |
return -EINVAL; |
b2edffdd9
|
802 |
} |
db446a08c
|
803 |
|
855ef0dec
|
804 |
table = rcu_dereference_raw(mm->ioctx_table); |
cd21b3400
|
805 806 |
WARN_ON(ctx != rcu_access_pointer(table->table[ctx->id])); RCU_INIT_POINTER(table->table[ctx->id], NULL); |
fa88b6f88
|
807 |
spin_unlock(&mm->ioctx_lock); |
4fcc712f5
|
808 |
|
076c7c068
|
809 |
/* free_ioctx_reqs() will do the necessary RCU synchronization */ |
fa88b6f88
|
810 |
wake_up_all(&ctx->wait); |
4fcc712f5
|
811 |
|
fa88b6f88
|
812 813 814 815 816 817 818 819 |
/* * It'd be more correct to do this in free_ioctx(), after all * the outstanding kiocbs have finished - but by then io_destroy * has already returned, so io_setup() could potentially return * -EAGAIN with no ioctxs actually in use (as far as userspace * could tell). */ aio_nr_sub(ctx->max_reqs); |
4fcc712f5
|
820 |
|
fa88b6f88
|
821 822 |
if (ctx->mmap_size) vm_munmap(ctx->mmap_base, ctx->mmap_size); |
fb2d44838
|
823 |
|
dc48e56d7
|
824 |
ctx->rq_wait = wait; |
fa88b6f88
|
825 826 |
percpu_ref_kill(&ctx->users); return 0; |
1da177e4c
|
827 |
} |
36f558890
|
828 829 830 831 832 833 834 |
/* * exit_aio: called when the last user of mm goes away. At this point, there is * no way for any new requests to be submited or any of the io_* syscalls to be * called on the context. * * There may be outstanding kiocbs, but free_ioctx() will explicitly wait on * them. |
1da177e4c
|
835 |
*/ |
fc9b52cd8
|
836 |
void exit_aio(struct mm_struct *mm) |
1da177e4c
|
837 |
{ |
4b70ac5fd
|
838 |
struct kioctx_table *table = rcu_dereference_raw(mm->ioctx_table); |
dc48e56d7
|
839 840 |
struct ctx_rq_wait wait; int i, skipped; |
db446a08c
|
841 |
|
4b70ac5fd
|
842 843 |
if (!table) return; |
db446a08c
|
844 |
|
dc48e56d7
|
845 846 847 848 |
atomic_set(&wait.count, table->nr); init_completion(&wait.comp); skipped = 0; |
4b70ac5fd
|
849 |
for (i = 0; i < table->nr; ++i) { |
cd21b3400
|
850 851 |
struct kioctx *ctx = rcu_dereference_protected(table->table[i], true); |
abf137dd7
|
852 |
|
dc48e56d7
|
853 854 |
if (!ctx) { skipped++; |
4b70ac5fd
|
855 |
continue; |
dc48e56d7
|
856 |
} |
936af1576
|
857 |
/* |
4b70ac5fd
|
858 859 860 861 862 |
* We don't need to bother with munmap() here - exit_mmap(mm) * is coming and it'll unmap everything. And we simply can't, * this is not necessarily our ->mm. * Since kill_ioctx() uses non-zero ->mmap_size as indicator * that it needs to unmap the area, just set it to 0. |
936af1576
|
863 |
*/ |
58c85dc20
|
864 |
ctx->mmap_size = 0; |
dc48e56d7
|
865 866 |
kill_ioctx(mm, ctx, &wait); } |
36f558890
|
867 |
|
dc48e56d7
|
868 |
if (!atomic_sub_and_test(skipped, &wait.count)) { |
6098b45b3
|
869 |
/* Wait until all IO for the context are done. */ |
dc48e56d7
|
870 |
wait_for_completion(&wait.comp); |
1da177e4c
|
871 |
} |
4b70ac5fd
|
872 873 874 |
RCU_INIT_POINTER(mm->ioctx_table, NULL); kfree(table); |
1da177e4c
|
875 |
} |
e1bdd5f27
|
876 877 878 |
static void put_reqs_available(struct kioctx *ctx, unsigned nr) { struct kioctx_cpu *kcpu; |
263782c1c
|
879 |
unsigned long flags; |
e1bdd5f27
|
880 |
|
263782c1c
|
881 |
local_irq_save(flags); |
be6fb451a
|
882 |
kcpu = this_cpu_ptr(ctx->cpu); |
e1bdd5f27
|
883 |
kcpu->reqs_available += nr; |
263782c1c
|
884 |
|
e1bdd5f27
|
885 886 887 888 |
while (kcpu->reqs_available >= ctx->req_batch * 2) { kcpu->reqs_available -= ctx->req_batch; atomic_add(ctx->req_batch, &ctx->reqs_available); } |
263782c1c
|
889 |
local_irq_restore(flags); |
e1bdd5f27
|
890 891 892 893 894 895 |
} static bool get_reqs_available(struct kioctx *ctx) { struct kioctx_cpu *kcpu; bool ret = false; |
263782c1c
|
896 |
unsigned long flags; |
e1bdd5f27
|
897 |
|
263782c1c
|
898 |
local_irq_save(flags); |
be6fb451a
|
899 |
kcpu = this_cpu_ptr(ctx->cpu); |
e1bdd5f27
|
900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 |
if (!kcpu->reqs_available) { int old, avail = atomic_read(&ctx->reqs_available); do { if (avail < ctx->req_batch) goto out; old = avail; avail = atomic_cmpxchg(&ctx->reqs_available, avail, avail - ctx->req_batch); } while (avail != old); kcpu->reqs_available += ctx->req_batch; } ret = true; kcpu->reqs_available--; out: |
263782c1c
|
918 |
local_irq_restore(flags); |
e1bdd5f27
|
919 920 |
return ret; } |
d856f32a8
|
921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 |
/* refill_reqs_available * Updates the reqs_available reference counts used for tracking the * number of free slots in the completion ring. This can be called * from aio_complete() (to optimistically update reqs_available) or * from aio_get_req() (the we're out of events case). It must be * called holding ctx->completion_lock. */ static void refill_reqs_available(struct kioctx *ctx, unsigned head, unsigned tail) { unsigned events_in_ring, completed; /* Clamp head since userland can write to it. */ head %= ctx->nr_events; if (head <= tail) events_in_ring = tail - head; else events_in_ring = ctx->nr_events - (head - tail); completed = ctx->completed_events; if (events_in_ring < completed) completed -= events_in_ring; else completed = 0; if (!completed) return; ctx->completed_events -= completed; put_reqs_available(ctx, completed); } /* user_refill_reqs_available * Called to refill reqs_available when aio_get_req() encounters an * out of space in the completion ring. */ static void user_refill_reqs_available(struct kioctx *ctx) { spin_lock_irq(&ctx->completion_lock); if (ctx->completed_events) { struct aio_ring *ring; unsigned head; /* Access of ring->head may race with aio_read_events_ring() * here, but that's okay since whether we read the old version * or the new version, and either will be valid. The important * part is that head cannot pass tail since we prevent * aio_complete() from updating tail by holding * ctx->completion_lock. Even if head is invalid, the check * against ctx->completed_events below will make sure we do the * safe/right thing. */ ring = kmap_atomic(ctx->ring_pages[0]); head = ring->head; kunmap_atomic(ring); refill_reqs_available(ctx, head, ctx->tail); } spin_unlock_irq(&ctx->completion_lock); } |
1da177e4c
|
982 |
/* aio_get_req |
57282d8fd
|
983 984 |
* Allocate a slot for an aio request. * Returns NULL if no requests are free. |
1da177e4c
|
985 |
*/ |
04b2fa9f8
|
986 |
static inline struct aio_kiocb *aio_get_req(struct kioctx *ctx) |
1da177e4c
|
987 |
{ |
04b2fa9f8
|
988 |
struct aio_kiocb *req; |
a1c8eae75
|
989 |
|
d856f32a8
|
990 991 992 993 994 |
if (!get_reqs_available(ctx)) { user_refill_reqs_available(ctx); if (!get_reqs_available(ctx)) return NULL; } |
a1c8eae75
|
995 |
|
0460fef2a
|
996 |
req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO); |
1da177e4c
|
997 |
if (unlikely(!req)) |
a1c8eae75
|
998 |
goto out_put; |
1da177e4c
|
999 |
|
e34ecee2a
|
1000 |
percpu_ref_get(&ctx->reqs); |
1da177e4c
|
1001 |
req->ki_ctx = ctx; |
080d676de
|
1002 |
return req; |
a1c8eae75
|
1003 |
out_put: |
e1bdd5f27
|
1004 |
put_reqs_available(ctx, 1); |
a1c8eae75
|
1005 |
return NULL; |
1da177e4c
|
1006 |
} |
04b2fa9f8
|
1007 |
static void kiocb_free(struct aio_kiocb *req) |
1da177e4c
|
1008 |
{ |
04b2fa9f8
|
1009 1010 |
if (req->common.ki_filp) fput(req->common.ki_filp); |
133890103
|
1011 1012 |
if (req->ki_eventfd != NULL) eventfd_ctx_put(req->ki_eventfd); |
1da177e4c
|
1013 |
kmem_cache_free(kiocb_cachep, req); |
1da177e4c
|
1014 |
} |
d5470b596
|
1015 |
static struct kioctx *lookup_ioctx(unsigned long ctx_id) |
1da177e4c
|
1016 |
{ |
db446a08c
|
1017 |
struct aio_ring __user *ring = (void __user *)ctx_id; |
abf137dd7
|
1018 |
struct mm_struct *mm = current->mm; |
65c24491b
|
1019 |
struct kioctx *ctx, *ret = NULL; |
db446a08c
|
1020 1021 1022 1023 1024 |
struct kioctx_table *table; unsigned id; if (get_user(id, &ring->id)) return NULL; |
1da177e4c
|
1025 |
|
abf137dd7
|
1026 |
rcu_read_lock(); |
db446a08c
|
1027 |
table = rcu_dereference(mm->ioctx_table); |
abf137dd7
|
1028 |
|
db446a08c
|
1029 1030 |
if (!table || id >= table->nr) goto out; |
1da177e4c
|
1031 |
|
cd21b3400
|
1032 |
ctx = rcu_dereference(table->table[id]); |
f30d704fe
|
1033 |
if (ctx && ctx->user_id == ctx_id) { |
fbcede36b
|
1034 1035 |
if (percpu_ref_tryget_live(&ctx->users)) ret = ctx; |
db446a08c
|
1036 1037 |
} out: |
abf137dd7
|
1038 |
rcu_read_unlock(); |
65c24491b
|
1039 |
return ret; |
1da177e4c
|
1040 |
} |
1da177e4c
|
1041 1042 |
/* aio_complete * Called when the io request on the given iocb is complete. |
1da177e4c
|
1043 |
*/ |
04b2fa9f8
|
1044 |
static void aio_complete(struct kiocb *kiocb, long res, long res2) |
1da177e4c
|
1045 |
{ |
04b2fa9f8
|
1046 |
struct aio_kiocb *iocb = container_of(kiocb, struct aio_kiocb, common); |
1da177e4c
|
1047 |
struct kioctx *ctx = iocb->ki_ctx; |
1da177e4c
|
1048 |
struct aio_ring *ring; |
21b40200c
|
1049 |
struct io_event *ev_page, *event; |
d856f32a8
|
1050 |
unsigned tail, pos, head; |
1da177e4c
|
1051 |
unsigned long flags; |
1da177e4c
|
1052 |
|
70fe2f481
|
1053 1054 1055 1056 1057 1058 1059 |
if (kiocb->ki_flags & IOCB_WRITE) { struct file *file = kiocb->ki_filp; /* * Tell lockdep we inherited freeze protection from submission * thread. */ |
a12f1ae61
|
1060 1061 |
if (S_ISREG(file_inode(file)->i_mode)) __sb_writers_acquired(file_inode(file)->i_sb, SB_FREEZE_WRITE); |
70fe2f481
|
1062 1063 |
file_end_write(file); } |
20dcae324
|
1064 1065 1066 1067 1068 1069 |
/* * Special case handling for sync iocbs: * - events go directly into the iocb for fast handling * - the sync task with the iocb in its stack holds the single iocb * ref, no other paths have a way to get another ref * - the sync task helpfully left a reference to itself in the iocb |
1da177e4c
|
1070 |
*/ |
04b2fa9f8
|
1071 |
BUG_ON(is_sync_kiocb(kiocb)); |
1da177e4c
|
1072 |
|
0460fef2a
|
1073 1074 1075 1076 1077 1078 1079 |
if (iocb->ki_list.next) { unsigned long flags; spin_lock_irqsave(&ctx->ctx_lock, flags); list_del(&iocb->ki_list); spin_unlock_irqrestore(&ctx->ctx_lock, flags); } |
11599ebac
|
1080 |
|
1da177e4c
|
1081 |
/* |
0460fef2a
|
1082 |
* Add a completion event to the ring buffer. Must be done holding |
4b30f07e7
|
1083 |
* ctx->completion_lock to prevent other code from messing with the tail |
0460fef2a
|
1084 1085 1086 |
* pointer since we might be called from irq context. */ spin_lock_irqsave(&ctx->completion_lock, flags); |
58c85dc20
|
1087 |
tail = ctx->tail; |
21b40200c
|
1088 |
pos = tail + AIO_EVENTS_OFFSET; |
58c85dc20
|
1089 |
if (++tail >= ctx->nr_events) |
4bf69b2a0
|
1090 |
tail = 0; |
1da177e4c
|
1091 |
|
58c85dc20
|
1092 |
ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]); |
21b40200c
|
1093 |
event = ev_page + pos % AIO_EVENTS_PER_PAGE; |
04b2fa9f8
|
1094 |
event->obj = (u64)(unsigned long)iocb->ki_user_iocb; |
1da177e4c
|
1095 1096 1097 |
event->data = iocb->ki_user_data; event->res = res; event->res2 = res2; |
21b40200c
|
1098 |
kunmap_atomic(ev_page); |
58c85dc20
|
1099 |
flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]); |
21b40200c
|
1100 1101 1102 |
pr_debug("%p[%u]: %p: %p %Lx %lx %lx ", |
04b2fa9f8
|
1103 |
ctx, tail, iocb, iocb->ki_user_iocb, iocb->ki_user_data, |
caf4167aa
|
1104 |
res, res2); |
1da177e4c
|
1105 1106 1107 1108 1109 |
/* after flagging the request as done, we * must never even look at it again */ smp_wmb(); /* make event visible before updating tail */ |
58c85dc20
|
1110 |
ctx->tail = tail; |
1da177e4c
|
1111 |
|
58c85dc20
|
1112 |
ring = kmap_atomic(ctx->ring_pages[0]); |
d856f32a8
|
1113 |
head = ring->head; |
21b40200c
|
1114 |
ring->tail = tail; |
e8e3c3d66
|
1115 |
kunmap_atomic(ring); |
58c85dc20
|
1116 |
flush_dcache_page(ctx->ring_pages[0]); |
1da177e4c
|
1117 |
|
d856f32a8
|
1118 1119 1120 |
ctx->completed_events++; if (ctx->completed_events > 1) refill_reqs_available(ctx, head, tail); |
0460fef2a
|
1121 |
spin_unlock_irqrestore(&ctx->completion_lock, flags); |
21b40200c
|
1122 1123 |
pr_debug("added to ring %p at [%u] ", iocb, tail); |
8d1c98b0b
|
1124 1125 1126 1127 1128 1129 |
/* * Check if the user asked us to deliver the result through an * eventfd. The eventfd_signal() function is safe to be called * from IRQ context. */ |
87c3a86e1
|
1130 |
if (iocb->ki_eventfd != NULL) |
8d1c98b0b
|
1131 |
eventfd_signal(iocb->ki_eventfd, 1); |
1da177e4c
|
1132 |
/* everything turned out well, dispose of the aiocb. */ |
57282d8fd
|
1133 |
kiocb_free(iocb); |
1da177e4c
|
1134 |
|
6cb2a2104
|
1135 1136 1137 1138 1139 1140 1141 |
/* * We have to order our ring_info tail store above and test * of the wait list below outside the wait lock. This is * like in wake_up_bit() where clearing a bit has to be * ordered with the unlocked test. */ smp_mb(); |
1da177e4c
|
1142 1143 |
if (waitqueue_active(&ctx->wait)) wake_up(&ctx->wait); |
e34ecee2a
|
1144 |
percpu_ref_put(&ctx->reqs); |
1da177e4c
|
1145 |
} |
2be4e7dee
|
1146 |
/* aio_read_events_ring |
a31ad380b
|
1147 1148 |
* Pull an event off of the ioctx's event ring. Returns the number of * events fetched |
1da177e4c
|
1149 |
*/ |
a31ad380b
|
1150 1151 |
static long aio_read_events_ring(struct kioctx *ctx, struct io_event __user *event, long nr) |
1da177e4c
|
1152 |
{ |
1da177e4c
|
1153 |
struct aio_ring *ring; |
5ffac122d
|
1154 |
unsigned head, tail, pos; |
a31ad380b
|
1155 1156 |
long ret = 0; int copy_ret; |
9c9ce763b
|
1157 1158 1159 1160 1161 1162 1163 |
/* * The mutex can block and wake us up and that will cause * wait_event_interruptible_hrtimeout() to schedule without sleeping * and repeat. This should be rare enough that it doesn't cause * peformance issues. See the comment in read_events() for more detail. */ sched_annotate_sleep(); |
58c85dc20
|
1164 |
mutex_lock(&ctx->ring_lock); |
1da177e4c
|
1165 |
|
fa8a53c39
|
1166 |
/* Access to ->ring_pages here is protected by ctx->ring_lock. */ |
58c85dc20
|
1167 |
ring = kmap_atomic(ctx->ring_pages[0]); |
a31ad380b
|
1168 |
head = ring->head; |
5ffac122d
|
1169 |
tail = ring->tail; |
a31ad380b
|
1170 |
kunmap_atomic(ring); |
2ff396be6
|
1171 1172 1173 1174 1175 |
/* * Ensure that once we've read the current tail pointer, that * we also see the events that were stored up to the tail. */ smp_rmb(); |
5ffac122d
|
1176 1177 |
pr_debug("h%u t%u m%u ", head, tail, ctx->nr_events); |
1da177e4c
|
1178 |
|
5ffac122d
|
1179 |
if (head == tail) |
1da177e4c
|
1180 |
goto out; |
edfbbf388
|
1181 1182 |
head %= ctx->nr_events; tail %= ctx->nr_events; |
a31ad380b
|
1183 1184 1185 1186 |
while (ret < nr) { long avail; struct io_event *ev; struct page *page; |
5ffac122d
|
1187 1188 |
avail = (head <= tail ? tail : ctx->nr_events) - head; if (head == tail) |
a31ad380b
|
1189 1190 1191 1192 1193 1194 1195 |
break; avail = min(avail, nr - ret); avail = min_t(long, avail, AIO_EVENTS_PER_PAGE - ((head + AIO_EVENTS_OFFSET) % AIO_EVENTS_PER_PAGE)); pos = head + AIO_EVENTS_OFFSET; |
58c85dc20
|
1196 |
page = ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]; |
a31ad380b
|
1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 |
pos %= AIO_EVENTS_PER_PAGE; ev = kmap(page); copy_ret = copy_to_user(event + ret, ev + pos, sizeof(*ev) * avail); kunmap(page); if (unlikely(copy_ret)) { ret = -EFAULT; goto out; } ret += avail; head += avail; |
58c85dc20
|
1211 |
head %= ctx->nr_events; |
1da177e4c
|
1212 |
} |
1da177e4c
|
1213 |
|
58c85dc20
|
1214 |
ring = kmap_atomic(ctx->ring_pages[0]); |
a31ad380b
|
1215 |
ring->head = head; |
91d80a84b
|
1216 |
kunmap_atomic(ring); |
58c85dc20
|
1217 |
flush_dcache_page(ctx->ring_pages[0]); |
a31ad380b
|
1218 |
|
5ffac122d
|
1219 1220 |
pr_debug("%li h%u t%u ", ret, head, tail); |
a31ad380b
|
1221 |
out: |
58c85dc20
|
1222 |
mutex_unlock(&ctx->ring_lock); |
a31ad380b
|
1223 |
|
1da177e4c
|
1224 1225 |
return ret; } |
a31ad380b
|
1226 1227 |
static bool aio_read_events(struct kioctx *ctx, long min_nr, long nr, struct io_event __user *event, long *i) |
1da177e4c
|
1228 |
{ |
a31ad380b
|
1229 |
long ret = aio_read_events_ring(ctx, event + *i, nr - *i); |
1da177e4c
|
1230 |
|
a31ad380b
|
1231 1232 |
if (ret > 0) *i += ret; |
1da177e4c
|
1233 |
|
a31ad380b
|
1234 1235 |
if (unlikely(atomic_read(&ctx->dead))) ret = -EINVAL; |
1da177e4c
|
1236 |
|
a31ad380b
|
1237 1238 |
if (!*i) *i = ret; |
1da177e4c
|
1239 |
|
a31ad380b
|
1240 |
return ret < 0 || *i >= min_nr; |
1da177e4c
|
1241 |
} |
a31ad380b
|
1242 |
static long read_events(struct kioctx *ctx, long min_nr, long nr, |
1da177e4c
|
1243 1244 1245 |
struct io_event __user *event, struct timespec __user *timeout) { |
2456e8553
|
1246 |
ktime_t until = KTIME_MAX; |
a31ad380b
|
1247 |
long ret = 0; |
1da177e4c
|
1248 |
|
1da177e4c
|
1249 1250 |
if (timeout) { struct timespec ts; |
a31ad380b
|
1251 |
|
1da177e4c
|
1252 |
if (unlikely(copy_from_user(&ts, timeout, sizeof(ts)))) |
a31ad380b
|
1253 |
return -EFAULT; |
1da177e4c
|
1254 |
|
a31ad380b
|
1255 |
until = timespec_to_ktime(ts); |
1da177e4c
|
1256 |
} |
a31ad380b
|
1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 |
/* * Note that aio_read_events() is being called as the conditional - i.e. * we're calling it after prepare_to_wait() has set task state to * TASK_INTERRUPTIBLE. * * But aio_read_events() can block, and if it blocks it's going to flip * the task state back to TASK_RUNNING. * * This should be ok, provided it doesn't flip the state back to * TASK_RUNNING and return 0 too much - that causes us to spin. That * will only happen if the mutex_lock() call blocks, and we then find * the ringbuffer empty. So in practice we should be ok, but it's * something to be aware of when touching this code. */ |
2456e8553
|
1271 |
if (until == 0) |
5f785de58
|
1272 1273 1274 1275 1276 |
aio_read_events(ctx, min_nr, nr, event, &ret); else wait_event_interruptible_hrtimeout(ctx->wait, aio_read_events(ctx, min_nr, nr, event, &ret), until); |
1da177e4c
|
1277 |
|
a31ad380b
|
1278 1279 |
if (!ret && signal_pending(current)) ret = -EINTR; |
1da177e4c
|
1280 |
|
a31ad380b
|
1281 |
return ret; |
1da177e4c
|
1282 |
} |
1da177e4c
|
1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 |
/* sys_io_setup: * Create an aio_context capable of receiving at least nr_events. * ctxp must not point to an aio_context that already exists, and * must be initialized to 0 prior to the call. On successful * creation of the aio_context, *ctxp is filled in with the resulting * handle. May fail with -EINVAL if *ctxp is not initialized, * if the specified nr_events exceeds internal limits. May fail * with -EAGAIN if the specified nr_events exceeds the user's limit * of available events. May fail with -ENOMEM if insufficient kernel * resources are available. May fail with -EFAULT if an invalid * pointer is passed for ctxp. Will fail with -ENOSYS if not * implemented. */ |
002c8976e
|
1296 |
SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp) |
1da177e4c
|
1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 |
{ struct kioctx *ioctx = NULL; unsigned long ctx; long ret; ret = get_user(ctx, ctxp); if (unlikely(ret)) goto out; ret = -EINVAL; |
d55b5fdaf
|
1307 |
if (unlikely(ctx || nr_events == 0)) { |
acd88d4e1
|
1308 1309 |
pr_debug("EINVAL: ctx %lu nr_events %u ", |
d55b5fdaf
|
1310 |
ctx, nr_events); |
1da177e4c
|
1311 1312 1313 1314 1315 1316 1317 |
goto out; } ioctx = ioctx_alloc(nr_events); ret = PTR_ERR(ioctx); if (!IS_ERR(ioctx)) { ret = put_user(ioctx->user_id, ctxp); |
a2e1859ad
|
1318 |
if (ret) |
e02ba72aa
|
1319 |
kill_ioctx(current->mm, ioctx, NULL); |
723be6e39
|
1320 |
percpu_ref_put(&ioctx->users); |
1da177e4c
|
1321 1322 1323 1324 1325 |
} out: return ret; } |
c00d2c7e8
|
1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 |
#ifdef CONFIG_COMPAT COMPAT_SYSCALL_DEFINE2(io_setup, unsigned, nr_events, u32 __user *, ctx32p) { struct kioctx *ioctx = NULL; unsigned long ctx; long ret; ret = get_user(ctx, ctx32p); if (unlikely(ret)) goto out; ret = -EINVAL; if (unlikely(ctx || nr_events == 0)) { pr_debug("EINVAL: ctx %lu nr_events %u ", ctx, nr_events); goto out; } ioctx = ioctx_alloc(nr_events); ret = PTR_ERR(ioctx); if (!IS_ERR(ioctx)) { /* truncating is ok because it's a user address */ ret = put_user((u32)ioctx->user_id, ctx32p); if (ret) kill_ioctx(current->mm, ioctx, NULL); percpu_ref_put(&ioctx->users); } out: return ret; } #endif |
1da177e4c
|
1359 1360 1361 |
/* sys_io_destroy: * Destroy the aio_context specified. May cancel any outstanding * AIOs and block on completion. Will fail with -ENOSYS if not |
642b5123a
|
1362 |
* implemented. May fail with -EINVAL if the context pointed to |
1da177e4c
|
1363 1364 |
* is invalid. */ |
002c8976e
|
1365 |
SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx) |
1da177e4c
|
1366 1367 1368 |
{ struct kioctx *ioctx = lookup_ioctx(ctx); if (likely(NULL != ioctx)) { |
dc48e56d7
|
1369 |
struct ctx_rq_wait wait; |
fb2d44838
|
1370 |
int ret; |
e02ba72aa
|
1371 |
|
dc48e56d7
|
1372 1373 |
init_completion(&wait.comp); atomic_set(&wait.count, 1); |
e02ba72aa
|
1374 1375 1376 1377 |
/* Pass requests_done to kill_ioctx() where it can be set * in a thread-safe way. If we try to set it here then we have * a race condition if two io_destroy() called simultaneously. */ |
dc48e56d7
|
1378 |
ret = kill_ioctx(current->mm, ioctx, &wait); |
723be6e39
|
1379 |
percpu_ref_put(&ioctx->users); |
e02ba72aa
|
1380 1381 1382 1383 1384 |
/* Wait until all IO for the context are done. Otherwise kernel * keep using user-space buffers even if user thinks the context * is destroyed. */ |
fb2d44838
|
1385 |
if (!ret) |
dc48e56d7
|
1386 |
wait_for_completion(&wait.comp); |
e02ba72aa
|
1387 |
|
fb2d44838
|
1388 |
return ret; |
1da177e4c
|
1389 |
} |
acd88d4e1
|
1390 1391 |
pr_debug("EINVAL: invalid context id "); |
1da177e4c
|
1392 1393 |
return -EINVAL; } |
89319d31d
|
1394 1395 |
static int aio_setup_rw(int rw, struct iocb *iocb, struct iovec **iovec, bool vectored, bool compat, struct iov_iter *iter) |
eed4e51fb
|
1396 |
{ |
89319d31d
|
1397 1398 1399 1400 1401 1402 1403 1404 |
void __user *buf = (void __user *)(uintptr_t)iocb->aio_buf; size_t len = iocb->aio_nbytes; if (!vectored) { ssize_t ret = import_single_range(rw, buf, len, *iovec, iter); *iovec = NULL; return ret; } |
9d85cba71
|
1405 1406 |
#ifdef CONFIG_COMPAT if (compat) |
89319d31d
|
1407 1408 |
return compat_import_iovec(rw, buf, len, UIO_FASTIOV, iovec, iter); |
9d85cba71
|
1409 |
#endif |
89319d31d
|
1410 |
return import_iovec(rw, buf, len, UIO_FASTIOV, iovec, iter); |
eed4e51fb
|
1411 |
} |
89319d31d
|
1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 |
static inline ssize_t aio_ret(struct kiocb *req, ssize_t ret) { switch (ret) { case -EIOCBQUEUED: return ret; case -ERESTARTSYS: case -ERESTARTNOINTR: case -ERESTARTNOHAND: case -ERESTART_RESTARTBLOCK: /* * There's no easy way to restart the syscall since other AIO's * may be already running. Just fail this IO with EINTR. */ ret = -EINTR; /*FALLTHRU*/ default: aio_complete(req, ret, 0); return 0; } } static ssize_t aio_read(struct kiocb *req, struct iocb *iocb, bool vectored, bool compat) |
1da177e4c
|
1435 |
{ |
41ef4eb8e
|
1436 |
struct file *file = req->ki_filp; |
00fefb9cf
|
1437 |
struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs; |
293bc9822
|
1438 |
struct iov_iter iter; |
89319d31d
|
1439 |
ssize_t ret; |
1da177e4c
|
1440 |
|
89319d31d
|
1441 1442 1443 1444 |
if (unlikely(!(file->f_mode & FMODE_READ))) return -EBADF; if (unlikely(!file->f_op->read_iter)) return -EINVAL; |
73a7075e3
|
1445 |
|
89319d31d
|
1446 1447 1448 1449 1450 |
ret = aio_setup_rw(READ, iocb, &iovec, vectored, compat, &iter); if (ret) return ret; ret = rw_verify_area(READ, file, &req->ki_pos, iov_iter_count(&iter)); if (!ret) |
bb7462b6f
|
1451 |
ret = aio_ret(req, call_read_iter(file, req, &iter)); |
89319d31d
|
1452 1453 1454 |
kfree(iovec); return ret; } |
73a7075e3
|
1455 |
|
89319d31d
|
1456 1457 1458 1459 1460 1461 1462 |
static ssize_t aio_write(struct kiocb *req, struct iocb *iocb, bool vectored, bool compat) { struct file *file = req->ki_filp; struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs; struct iov_iter iter; ssize_t ret; |
41ef4eb8e
|
1463 |
|
89319d31d
|
1464 1465 1466 |
if (unlikely(!(file->f_mode & FMODE_WRITE))) return -EBADF; if (unlikely(!file->f_op->write_iter)) |
41ef4eb8e
|
1467 |
return -EINVAL; |
1da177e4c
|
1468 |
|
89319d31d
|
1469 1470 1471 1472 1473 |
ret = aio_setup_rw(WRITE, iocb, &iovec, vectored, compat, &iter); if (ret) return ret; ret = rw_verify_area(WRITE, file, &req->ki_pos, iov_iter_count(&iter)); if (!ret) { |
70fe2f481
|
1474 |
req->ki_flags |= IOCB_WRITE; |
89319d31d
|
1475 |
file_start_write(file); |
bb7462b6f
|
1476 |
ret = aio_ret(req, call_write_iter(file, req, &iter)); |
70fe2f481
|
1477 1478 1479 1480 1481 |
/* * We release freeze protection in aio_complete(). Fool lockdep * by telling it the lock got released so that it doesn't * complain about held lock when we return to userspace. */ |
a12f1ae61
|
1482 1483 |
if (S_ISREG(file_inode(file)->i_mode)) __sb_writers_release(file_inode(file)->i_sb, SB_FREEZE_WRITE); |
41ef4eb8e
|
1484 |
} |
89319d31d
|
1485 1486 |
kfree(iovec); return ret; |
1da177e4c
|
1487 |
} |
d5470b596
|
1488 |
static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb, |
a1c8eae75
|
1489 |
struct iocb *iocb, bool compat) |
1da177e4c
|
1490 |
{ |
04b2fa9f8
|
1491 |
struct aio_kiocb *req; |
89319d31d
|
1492 |
struct file *file; |
1da177e4c
|
1493 1494 1495 |
ssize_t ret; /* enforce forwards compatibility on users */ |
9830f4be1
|
1496 |
if (unlikely(iocb->aio_reserved2)) { |
caf4167aa
|
1497 1498 |
pr_debug("EINVAL: reserve field set "); |
1da177e4c
|
1499 1500 1501 1502 1503 1504 1505 1506 1507 |
return -EINVAL; } /* prevent overflows */ if (unlikely( (iocb->aio_buf != (unsigned long)iocb->aio_buf) || (iocb->aio_nbytes != (size_t)iocb->aio_nbytes) || ((ssize_t)iocb->aio_nbytes < 0) )) { |
acd88d4e1
|
1508 1509 |
pr_debug("EINVAL: overflow check "); |
1da177e4c
|
1510 1511 |
return -EINVAL; } |
41ef4eb8e
|
1512 |
req = aio_get_req(ctx); |
1d98ebfcc
|
1513 |
if (unlikely(!req)) |
1da177e4c
|
1514 |
return -EAGAIN; |
1d98ebfcc
|
1515 |
|
89319d31d
|
1516 |
req->common.ki_filp = file = fget(iocb->aio_fildes); |
04b2fa9f8
|
1517 |
if (unlikely(!req->common.ki_filp)) { |
1d98ebfcc
|
1518 1519 |
ret = -EBADF; goto out_put_req; |
1da177e4c
|
1520 |
} |
04b2fa9f8
|
1521 1522 |
req->common.ki_pos = iocb->aio_offset; req->common.ki_complete = aio_complete; |
2ba48ce51
|
1523 |
req->common.ki_flags = iocb_flags(req->common.ki_filp); |
45d06cf70
|
1524 |
req->common.ki_hint = file_write_hint(file); |
1d98ebfcc
|
1525 |
|
9c3060bed
|
1526 1527 1528 1529 1530 1531 1532 |
if (iocb->aio_flags & IOCB_FLAG_RESFD) { /* * If the IOCB_FLAG_RESFD flag of aio_flags is set, get an * instance of the file* now. The file descriptor must be * an eventfd() fd, and will be signaled for each completed * event using the eventfd_signal() function. */ |
133890103
|
1533 |
req->ki_eventfd = eventfd_ctx_fdget((int) iocb->aio_resfd); |
801678c5a
|
1534 |
if (IS_ERR(req->ki_eventfd)) { |
9c3060bed
|
1535 |
ret = PTR_ERR(req->ki_eventfd); |
87c3a86e1
|
1536 |
req->ki_eventfd = NULL; |
9c3060bed
|
1537 1538 |
goto out_put_req; } |
04b2fa9f8
|
1539 1540 |
req->common.ki_flags |= IOCB_EVENTFD; |
9c3060bed
|
1541 |
} |
1da177e4c
|
1542 |
|
9830f4be1
|
1543 1544 1545 1546 1547 1548 |
ret = kiocb_set_rw_flags(&req->common, iocb->aio_rw_flags); if (unlikely(ret)) { pr_debug("EINVAL: aio_rw_flags "); goto out_put_req; } |
8a6608907
|
1549 |
ret = put_user(KIOCB_KEY, &user_iocb->aio_key); |
1da177e4c
|
1550 |
if (unlikely(ret)) { |
caf4167aa
|
1551 1552 |
pr_debug("EFAULT: aio_key "); |
1da177e4c
|
1553 1554 |
goto out_put_req; } |
04b2fa9f8
|
1555 |
req->ki_user_iocb = user_iocb; |
1da177e4c
|
1556 |
req->ki_user_data = iocb->aio_data; |
1da177e4c
|
1557 |
|
89319d31d
|
1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 |
get_file(file); switch (iocb->aio_lio_opcode) { case IOCB_CMD_PREAD: ret = aio_read(&req->common, iocb, false, compat); break; case IOCB_CMD_PWRITE: ret = aio_write(&req->common, iocb, false, compat); break; case IOCB_CMD_PREADV: ret = aio_read(&req->common, iocb, true, compat); break; case IOCB_CMD_PWRITEV: ret = aio_write(&req->common, iocb, true, compat); break; default: pr_debug("invalid aio operation %d ", iocb->aio_lio_opcode); ret = -EINVAL; break; } fput(file); |
1da177e4c
|
1579 |
|
89319d31d
|
1580 1581 |
if (ret && ret != -EIOCBQUEUED) goto out_put_req; |
1da177e4c
|
1582 |
return 0; |
1da177e4c
|
1583 |
out_put_req: |
e1bdd5f27
|
1584 |
put_reqs_available(ctx, 1); |
e34ecee2a
|
1585 |
percpu_ref_put(&ctx->reqs); |
57282d8fd
|
1586 |
kiocb_free(req); |
1da177e4c
|
1587 1588 |
return ret; } |
c00d2c7e8
|
1589 1590 |
static long do_io_submit(aio_context_t ctx_id, long nr, struct iocb __user *__user *iocbpp, bool compat) |
1da177e4c
|
1591 1592 1593 |
{ struct kioctx *ctx; long ret = 0; |
080d676de
|
1594 |
int i = 0; |
9f5b94254
|
1595 |
struct blk_plug plug; |
1da177e4c
|
1596 1597 1598 |
if (unlikely(nr < 0)) return -EINVAL; |
75e1c70fc
|
1599 1600 |
if (unlikely(nr > LONG_MAX/sizeof(*iocbpp))) nr = LONG_MAX/sizeof(*iocbpp); |
1da177e4c
|
1601 1602 1603 1604 1605 |
if (unlikely(!access_ok(VERIFY_READ, iocbpp, (nr*sizeof(*iocbpp))))) return -EFAULT; ctx = lookup_ioctx(ctx_id); if (unlikely(!ctx)) { |
caf4167aa
|
1606 1607 |
pr_debug("EINVAL: invalid context id "); |
1da177e4c
|
1608 1609 |
return -EINVAL; } |
9f5b94254
|
1610 |
blk_start_plug(&plug); |
1da177e4c
|
1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 |
/* * AKPM: should this return a partial result if some of the IOs were * successfully submitted? */ for (i=0; i<nr; i++) { struct iocb __user *user_iocb; struct iocb tmp; if (unlikely(__get_user(user_iocb, iocbpp + i))) { ret = -EFAULT; break; } if (unlikely(copy_from_user(&tmp, user_iocb, sizeof(tmp)))) { ret = -EFAULT; break; } |
a1c8eae75
|
1628 |
ret = io_submit_one(ctx, user_iocb, &tmp, compat); |
1da177e4c
|
1629 1630 1631 |
if (ret) break; } |
9f5b94254
|
1632 |
blk_finish_plug(&plug); |
1da177e4c
|
1633 |
|
723be6e39
|
1634 |
percpu_ref_put(&ctx->users); |
1da177e4c
|
1635 1636 |
return i ? i : ret; } |
9d85cba71
|
1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 |
/* sys_io_submit: * Queue the nr iocbs pointed to by iocbpp for processing. Returns * the number of iocbs queued. May return -EINVAL if the aio_context * specified by ctx_id is invalid, if nr is < 0, if the iocb at * *iocbpp[0] is not properly initialized, if the operation specified * is invalid for the file descriptor in the iocb. May fail with * -EFAULT if any of the data structures point to invalid data. May * fail with -EBADF if the file descriptor specified in the first * iocb is invalid. May fail with -EAGAIN if insufficient resources * are available to queue any iocbs. Will return 0 if nr is 0. Will * fail with -ENOSYS if not implemented. */ SYSCALL_DEFINE3(io_submit, aio_context_t, ctx_id, long, nr, struct iocb __user * __user *, iocbpp) { return do_io_submit(ctx_id, nr, iocbpp, 0); } |
c00d2c7e8
|
1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 |
#ifdef CONFIG_COMPAT static inline long copy_iocb(long nr, u32 __user *ptr32, struct iocb __user * __user *ptr64) { compat_uptr_t uptr; int i; for (i = 0; i < nr; ++i) { if (get_user(uptr, ptr32 + i)) return -EFAULT; if (put_user(compat_ptr(uptr), ptr64 + i)) return -EFAULT; } return 0; } #define MAX_AIO_SUBMITS (PAGE_SIZE/sizeof(struct iocb *)) COMPAT_SYSCALL_DEFINE3(io_submit, compat_aio_context_t, ctx_id, int, nr, u32 __user *, iocb) { struct iocb __user * __user *iocb64; long ret; if (unlikely(nr < 0)) return -EINVAL; if (nr > MAX_AIO_SUBMITS) nr = MAX_AIO_SUBMITS; iocb64 = compat_alloc_user_space(nr * sizeof(*iocb64)); ret = copy_iocb(nr, iocb, iocb64); if (!ret) ret = do_io_submit(ctx_id, nr, iocb64, 1); return ret; } #endif |
1da177e4c
|
1691 1692 |
/* lookup_kiocb * Finds a given iocb for cancellation. |
1da177e4c
|
1693 |
*/ |
04b2fa9f8
|
1694 1695 |
static struct aio_kiocb * lookup_kiocb(struct kioctx *ctx, struct iocb __user *iocb, u32 key) |
1da177e4c
|
1696 |
{ |
04b2fa9f8
|
1697 |
struct aio_kiocb *kiocb; |
d00689af6
|
1698 1699 |
assert_spin_locked(&ctx->ctx_lock); |
8a6608907
|
1700 1701 |
if (key != KIOCB_KEY) return NULL; |
1da177e4c
|
1702 |
/* TODO: use a hash or array, this sucks. */ |
04b2fa9f8
|
1703 1704 |
list_for_each_entry(kiocb, &ctx->active_reqs, ki_list) { if (kiocb->ki_user_iocb == iocb) |
1da177e4c
|
1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 |
return kiocb; } return NULL; } /* sys_io_cancel: * Attempts to cancel an iocb previously passed to io_submit. If * the operation is successfully cancelled, the resulting event is * copied into the memory pointed to by result without being placed * into the completion queue and 0 is returned. May fail with * -EFAULT if any of the data structures pointed to are invalid. * May fail with -EINVAL if aio_context specified by ctx_id is * invalid. May fail with -EAGAIN if the iocb specified was not * cancelled. Will fail with -ENOSYS if not implemented. */ |
002c8976e
|
1720 1721 |
SYSCALL_DEFINE3(io_cancel, aio_context_t, ctx_id, struct iocb __user *, iocb, struct io_event __user *, result) |
1da177e4c
|
1722 |
{ |
1da177e4c
|
1723 |
struct kioctx *ctx; |
04b2fa9f8
|
1724 |
struct aio_kiocb *kiocb; |
1da177e4c
|
1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 |
u32 key; int ret; ret = get_user(key, &iocb->aio_key); if (unlikely(ret)) return -EFAULT; ctx = lookup_ioctx(ctx_id); if (unlikely(!ctx)) return -EINVAL; spin_lock_irq(&ctx->ctx_lock); |
906b973cf
|
1737 |
|
1da177e4c
|
1738 |
kiocb = lookup_kiocb(ctx, iocb, key); |
906b973cf
|
1739 |
if (kiocb) |
d52a8f9ea
|
1740 |
ret = kiocb_cancel(kiocb); |
906b973cf
|
1741 1742 |
else ret = -EINVAL; |
1da177e4c
|
1743 |
spin_unlock_irq(&ctx->ctx_lock); |
906b973cf
|
1744 |
if (!ret) { |
bec68faaf
|
1745 1746 1747 1748 |
/* * The result argument is no longer used - the io_event is * always delivered via the ring buffer. -EINPROGRESS indicates * cancellation is progress: |
906b973cf
|
1749 |
*/ |
bec68faaf
|
1750 |
ret = -EINPROGRESS; |
906b973cf
|
1751 |
} |
1da177e4c
|
1752 |
|
723be6e39
|
1753 |
percpu_ref_put(&ctx->users); |
1da177e4c
|
1754 1755 1756 1757 1758 1759 |
return ret; } /* io_getevents: * Attempts to read at least min_nr events and up to nr events from |
642b5123a
|
1760 1761 1762 1763 1764 1765 1766 1767 |
* the completion queue for the aio_context specified by ctx_id. If * it succeeds, the number of read events is returned. May fail with * -EINVAL if ctx_id is invalid, if min_nr is out of range, if nr is * out of range, if timeout is out of range. May fail with -EFAULT * if any of the memory specified is invalid. May return 0 or * < min_nr if the timeout specified by timeout has elapsed * before sufficient events are available, where timeout == NULL * specifies an infinite timeout. Note that the timeout pointed to by |
6900807c6
|
1768 |
* timeout is relative. Will fail with -ENOSYS if not implemented. |
1da177e4c
|
1769 |
*/ |
002c8976e
|
1770 1771 1772 1773 1774 |
SYSCALL_DEFINE5(io_getevents, aio_context_t, ctx_id, long, min_nr, long, nr, struct io_event __user *, events, struct timespec __user *, timeout) |
1da177e4c
|
1775 1776 1777 1778 1779 |
{ struct kioctx *ioctx = lookup_ioctx(ctx_id); long ret = -EINVAL; if (likely(ioctx)) { |
2e4102559
|
1780 |
if (likely(min_nr <= nr && min_nr >= 0)) |
1da177e4c
|
1781 |
ret = read_events(ioctx, min_nr, nr, events, timeout); |
723be6e39
|
1782 |
percpu_ref_put(&ioctx->users); |
1da177e4c
|
1783 |
} |
1da177e4c
|
1784 1785 |
return ret; } |
c00d2c7e8
|
1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 |
#ifdef CONFIG_COMPAT COMPAT_SYSCALL_DEFINE5(io_getevents, compat_aio_context_t, ctx_id, compat_long_t, min_nr, compat_long_t, nr, struct io_event __user *, events, struct compat_timespec __user *, timeout) { struct timespec t; struct timespec __user *ut = NULL; if (timeout) { if (compat_get_timespec(&t, timeout)) return -EFAULT; ut = compat_alloc_user_space(sizeof(*ut)); if (copy_to_user(ut, &t, sizeof(t))) return -EFAULT; } return sys_io_getevents(ctx_id, min_nr, nr, events, ut); } #endif |