x86/kaslr: Fix off-by-one error in process_gb_huge_pages()
[linux-2.6-microblaze.git] / kernel / bpf / ringbuf.c
1 #include <linux/bpf.h>
2 #include <linux/btf.h>
3 #include <linux/err.h>
4 #include <linux/irq_work.h>
5 #include <linux/slab.h>
6 #include <linux/filter.h>
7 #include <linux/mm.h>
8 #include <linux/vmalloc.h>
9 #include <linux/wait.h>
10 #include <linux/poll.h>
11 #include <uapi/linux/btf.h>
12
13 #define RINGBUF_CREATE_FLAG_MASK (BPF_F_NUMA_NODE)
14
15 /* non-mmap()'able part of bpf_ringbuf (everything up to consumer page) */
16 #define RINGBUF_PGOFF \
17         (offsetof(struct bpf_ringbuf, consumer_pos) >> PAGE_SHIFT)
18 /* consumer page and producer page */
19 #define RINGBUF_POS_PAGES 2
20
21 #define RINGBUF_MAX_RECORD_SZ (UINT_MAX/4)
22
23 /* Maximum size of ring buffer area is limited by 32-bit page offset within
24  * record header, counted in pages. Reserve 8 bits for extensibility, and take
25  * into account few extra pages for consumer/producer pages and
26  * non-mmap()'able parts. This gives 64GB limit, which seems plenty for single
27  * ring buffer.
28  */
29 #define RINGBUF_MAX_DATA_SZ \
30         (((1ULL << 24) - RINGBUF_POS_PAGES - RINGBUF_PGOFF) * PAGE_SIZE)
31
32 struct bpf_ringbuf {
33         wait_queue_head_t waitq;
34         struct irq_work work;
35         u64 mask;
36         struct page **pages;
37         int nr_pages;
38         spinlock_t spinlock ____cacheline_aligned_in_smp;
39         /* Consumer and producer counters are put into separate pages to allow
40          * mapping consumer page as r/w, but restrict producer page to r/o.
41          * This protects producer position from being modified by user-space
42          * application and ruining in-kernel position tracking.
43          */
44         unsigned long consumer_pos __aligned(PAGE_SIZE);
45         unsigned long producer_pos __aligned(PAGE_SIZE);
46         char data[] __aligned(PAGE_SIZE);
47 };
48
49 struct bpf_ringbuf_map {
50         struct bpf_map map;
51         struct bpf_map_memory memory;
52         struct bpf_ringbuf *rb;
53 };
54
55 /* 8-byte ring buffer record header structure */
56 struct bpf_ringbuf_hdr {
57         u32 len;
58         u32 pg_off;
59 };
60
61 static struct bpf_ringbuf *bpf_ringbuf_area_alloc(size_t data_sz, int numa_node)
62 {
63         const gfp_t flags = GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN |
64                             __GFP_ZERO;
65         int nr_meta_pages = RINGBUF_PGOFF + RINGBUF_POS_PAGES;
66         int nr_data_pages = data_sz >> PAGE_SHIFT;
67         int nr_pages = nr_meta_pages + nr_data_pages;
68         struct page **pages, *page;
69         struct bpf_ringbuf *rb;
70         size_t array_size;
71         int i;
72
73         /* Each data page is mapped twice to allow "virtual"
74          * continuous read of samples wrapping around the end of ring
75          * buffer area:
76          * ------------------------------------------------------
77          * | meta pages |  real data pages  |  same data pages  |
78          * ------------------------------------------------------
79          * |            | 1 2 3 4 5 6 7 8 9 | 1 2 3 4 5 6 7 8 9 |
80          * ------------------------------------------------------
81          * |            | TA             DA | TA             DA |
82          * ------------------------------------------------------
83          *                               ^^^^^^^
84          *                                  |
85          * Here, no need to worry about special handling of wrapped-around
86          * data due to double-mapped data pages. This works both in kernel and
87          * when mmap()'ed in user-space, simplifying both kernel and
88          * user-space implementations significantly.
89          */
90         array_size = (nr_meta_pages + 2 * nr_data_pages) * sizeof(*pages);
91         if (array_size > PAGE_SIZE)
92                 pages = vmalloc_node(array_size, numa_node);
93         else
94                 pages = kmalloc_node(array_size, flags, numa_node);
95         if (!pages)
96                 return NULL;
97
98         for (i = 0; i < nr_pages; i++) {
99                 page = alloc_pages_node(numa_node, flags, 0);
100                 if (!page) {
101                         nr_pages = i;
102                         goto err_free_pages;
103                 }
104                 pages[i] = page;
105                 if (i >= nr_meta_pages)
106                         pages[nr_data_pages + i] = page;
107         }
108
109         rb = vmap(pages, nr_meta_pages + 2 * nr_data_pages,
110                   VM_ALLOC | VM_USERMAP, PAGE_KERNEL);
111         if (rb) {
112                 rb->pages = pages;
113                 rb->nr_pages = nr_pages;
114                 return rb;
115         }
116
117 err_free_pages:
118         for (i = 0; i < nr_pages; i++)
119                 __free_page(pages[i]);
120         kvfree(pages);
121         return NULL;
122 }
123
124 static void bpf_ringbuf_notify(struct irq_work *work)
125 {
126         struct bpf_ringbuf *rb = container_of(work, struct bpf_ringbuf, work);
127
128         wake_up_all(&rb->waitq);
129 }
130
131 static struct bpf_ringbuf *bpf_ringbuf_alloc(size_t data_sz, int numa_node)
132 {
133         struct bpf_ringbuf *rb;
134
135         rb = bpf_ringbuf_area_alloc(data_sz, numa_node);
136         if (!rb)
137                 return ERR_PTR(-ENOMEM);
138
139         spin_lock_init(&rb->spinlock);
140         init_waitqueue_head(&rb->waitq);
141         init_irq_work(&rb->work, bpf_ringbuf_notify);
142
143         rb->mask = data_sz - 1;
144         rb->consumer_pos = 0;
145         rb->producer_pos = 0;
146
147         return rb;
148 }
149
150 static struct bpf_map *ringbuf_map_alloc(union bpf_attr *attr)
151 {
152         struct bpf_ringbuf_map *rb_map;
153         u64 cost;
154         int err;
155
156         if (attr->map_flags & ~RINGBUF_CREATE_FLAG_MASK)
157                 return ERR_PTR(-EINVAL);
158
159         if (attr->key_size || attr->value_size ||
160             !is_power_of_2(attr->max_entries) ||
161             !PAGE_ALIGNED(attr->max_entries))
162                 return ERR_PTR(-EINVAL);
163
164 #ifdef CONFIG_64BIT
165         /* on 32-bit arch, it's impossible to overflow record's hdr->pgoff */
166         if (attr->max_entries > RINGBUF_MAX_DATA_SZ)
167                 return ERR_PTR(-E2BIG);
168 #endif
169
170         rb_map = kzalloc(sizeof(*rb_map), GFP_USER);
171         if (!rb_map)
172                 return ERR_PTR(-ENOMEM);
173
174         bpf_map_init_from_attr(&rb_map->map, attr);
175
176         cost = sizeof(struct bpf_ringbuf_map) +
177                sizeof(struct bpf_ringbuf) +
178                attr->max_entries;
179         err = bpf_map_charge_init(&rb_map->map.memory, cost);
180         if (err)
181                 goto err_free_map;
182
183         rb_map->rb = bpf_ringbuf_alloc(attr->max_entries, rb_map->map.numa_node);
184         if (IS_ERR(rb_map->rb)) {
185                 err = PTR_ERR(rb_map->rb);
186                 goto err_uncharge;
187         }
188
189         return &rb_map->map;
190
191 err_uncharge:
192         bpf_map_charge_finish(&rb_map->map.memory);
193 err_free_map:
194         kfree(rb_map);
195         return ERR_PTR(err);
196 }
197
198 static void bpf_ringbuf_free(struct bpf_ringbuf *rb)
199 {
200         /* copy pages pointer and nr_pages to local variable, as we are going
201          * to unmap rb itself with vunmap() below
202          */
203         struct page **pages = rb->pages;
204         int i, nr_pages = rb->nr_pages;
205
206         vunmap(rb);
207         for (i = 0; i < nr_pages; i++)
208                 __free_page(pages[i]);
209         kvfree(pages);
210 }
211
212 static void ringbuf_map_free(struct bpf_map *map)
213 {
214         struct bpf_ringbuf_map *rb_map;
215
216         /* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
217          * so the programs (can be more than one that used this map) were
218          * disconnected from events. Wait for outstanding critical sections in
219          * these programs to complete
220          */
221         synchronize_rcu();
222
223         rb_map = container_of(map, struct bpf_ringbuf_map, map);
224         bpf_ringbuf_free(rb_map->rb);
225         kfree(rb_map);
226 }
227
228 static void *ringbuf_map_lookup_elem(struct bpf_map *map, void *key)
229 {
230         return ERR_PTR(-ENOTSUPP);
231 }
232
233 static int ringbuf_map_update_elem(struct bpf_map *map, void *key, void *value,
234                                    u64 flags)
235 {
236         return -ENOTSUPP;
237 }
238
239 static int ringbuf_map_delete_elem(struct bpf_map *map, void *key)
240 {
241         return -ENOTSUPP;
242 }
243
244 static int ringbuf_map_get_next_key(struct bpf_map *map, void *key,
245                                     void *next_key)
246 {
247         return -ENOTSUPP;
248 }
249
250 static size_t bpf_ringbuf_mmap_page_cnt(const struct bpf_ringbuf *rb)
251 {
252         size_t data_pages = (rb->mask + 1) >> PAGE_SHIFT;
253
254         /* consumer page + producer page + 2 x data pages */
255         return RINGBUF_POS_PAGES + 2 * data_pages;
256 }
257
258 static int ringbuf_map_mmap(struct bpf_map *map, struct vm_area_struct *vma)
259 {
260         struct bpf_ringbuf_map *rb_map;
261         size_t mmap_sz;
262
263         rb_map = container_of(map, struct bpf_ringbuf_map, map);
264         mmap_sz = bpf_ringbuf_mmap_page_cnt(rb_map->rb) << PAGE_SHIFT;
265
266         if (vma->vm_pgoff * PAGE_SIZE + (vma->vm_end - vma->vm_start) > mmap_sz)
267                 return -EINVAL;
268
269         return remap_vmalloc_range(vma, rb_map->rb,
270                                    vma->vm_pgoff + RINGBUF_PGOFF);
271 }
272
273 static unsigned long ringbuf_avail_data_sz(struct bpf_ringbuf *rb)
274 {
275         unsigned long cons_pos, prod_pos;
276
277         cons_pos = smp_load_acquire(&rb->consumer_pos);
278         prod_pos = smp_load_acquire(&rb->producer_pos);
279         return prod_pos - cons_pos;
280 }
281
282 static __poll_t ringbuf_map_poll(struct bpf_map *map, struct file *filp,
283                                  struct poll_table_struct *pts)
284 {
285         struct bpf_ringbuf_map *rb_map;
286
287         rb_map = container_of(map, struct bpf_ringbuf_map, map);
288         poll_wait(filp, &rb_map->rb->waitq, pts);
289
290         if (ringbuf_avail_data_sz(rb_map->rb))
291                 return EPOLLIN | EPOLLRDNORM;
292         return 0;
293 }
294
295 const struct bpf_map_ops ringbuf_map_ops = {
296         .map_alloc = ringbuf_map_alloc,
297         .map_free = ringbuf_map_free,
298         .map_mmap = ringbuf_map_mmap,
299         .map_poll = ringbuf_map_poll,
300         .map_lookup_elem = ringbuf_map_lookup_elem,
301         .map_update_elem = ringbuf_map_update_elem,
302         .map_delete_elem = ringbuf_map_delete_elem,
303         .map_get_next_key = ringbuf_map_get_next_key,
304 };
305
306 /* Given pointer to ring buffer record metadata and struct bpf_ringbuf itself,
307  * calculate offset from record metadata to ring buffer in pages, rounded
308  * down. This page offset is stored as part of record metadata and allows to
309  * restore struct bpf_ringbuf * from record pointer. This page offset is
310  * stored at offset 4 of record metadata header.
311  */
312 static size_t bpf_ringbuf_rec_pg_off(struct bpf_ringbuf *rb,
313                                      struct bpf_ringbuf_hdr *hdr)
314 {
315         return ((void *)hdr - (void *)rb) >> PAGE_SHIFT;
316 }
317
318 /* Given pointer to ring buffer record header, restore pointer to struct
319  * bpf_ringbuf itself by using page offset stored at offset 4
320  */
321 static struct bpf_ringbuf *
322 bpf_ringbuf_restore_from_rec(struct bpf_ringbuf_hdr *hdr)
323 {
324         unsigned long addr = (unsigned long)(void *)hdr;
325         unsigned long off = (unsigned long)hdr->pg_off << PAGE_SHIFT;
326
327         return (void*)((addr & PAGE_MASK) - off);
328 }
329
330 static void *__bpf_ringbuf_reserve(struct bpf_ringbuf *rb, u64 size)
331 {
332         unsigned long cons_pos, prod_pos, new_prod_pos, flags;
333         u32 len, pg_off;
334         struct bpf_ringbuf_hdr *hdr;
335
336         if (unlikely(size > RINGBUF_MAX_RECORD_SZ))
337                 return NULL;
338
339         len = round_up(size + BPF_RINGBUF_HDR_SZ, 8);
340         cons_pos = smp_load_acquire(&rb->consumer_pos);
341
342         if (in_nmi()) {
343                 if (!spin_trylock_irqsave(&rb->spinlock, flags))
344                         return NULL;
345         } else {
346                 spin_lock_irqsave(&rb->spinlock, flags);
347         }
348
349         prod_pos = rb->producer_pos;
350         new_prod_pos = prod_pos + len;
351
352         /* check for out of ringbuf space by ensuring producer position
353          * doesn't advance more than (ringbuf_size - 1) ahead
354          */
355         if (new_prod_pos - cons_pos > rb->mask) {
356                 spin_unlock_irqrestore(&rb->spinlock, flags);
357                 return NULL;
358         }
359
360         hdr = (void *)rb->data + (prod_pos & rb->mask);
361         pg_off = bpf_ringbuf_rec_pg_off(rb, hdr);
362         hdr->len = size | BPF_RINGBUF_BUSY_BIT;
363         hdr->pg_off = pg_off;
364
365         /* pairs with consumer's smp_load_acquire() */
366         smp_store_release(&rb->producer_pos, new_prod_pos);
367
368         spin_unlock_irqrestore(&rb->spinlock, flags);
369
370         return (void *)hdr + BPF_RINGBUF_HDR_SZ;
371 }
372
373 BPF_CALL_3(bpf_ringbuf_reserve, struct bpf_map *, map, u64, size, u64, flags)
374 {
375         struct bpf_ringbuf_map *rb_map;
376
377         if (unlikely(flags))
378                 return 0;
379
380         rb_map = container_of(map, struct bpf_ringbuf_map, map);
381         return (unsigned long)__bpf_ringbuf_reserve(rb_map->rb, size);
382 }
383
384 const struct bpf_func_proto bpf_ringbuf_reserve_proto = {
385         .func           = bpf_ringbuf_reserve,
386         .ret_type       = RET_PTR_TO_ALLOC_MEM_OR_NULL,
387         .arg1_type      = ARG_CONST_MAP_PTR,
388         .arg2_type      = ARG_CONST_ALLOC_SIZE_OR_ZERO,
389         .arg3_type      = ARG_ANYTHING,
390 };
391
392 static void bpf_ringbuf_commit(void *sample, u64 flags, bool discard)
393 {
394         unsigned long rec_pos, cons_pos;
395         struct bpf_ringbuf_hdr *hdr;
396         struct bpf_ringbuf *rb;
397         u32 new_len;
398
399         hdr = sample - BPF_RINGBUF_HDR_SZ;
400         rb = bpf_ringbuf_restore_from_rec(hdr);
401         new_len = hdr->len ^ BPF_RINGBUF_BUSY_BIT;
402         if (discard)
403                 new_len |= BPF_RINGBUF_DISCARD_BIT;
404
405         /* update record header with correct final size prefix */
406         xchg(&hdr->len, new_len);
407
408         /* if consumer caught up and is waiting for our record, notify about
409          * new data availability
410          */
411         rec_pos = (void *)hdr - (void *)rb->data;
412         cons_pos = smp_load_acquire(&rb->consumer_pos) & rb->mask;
413
414         if (flags & BPF_RB_FORCE_WAKEUP)
415                 irq_work_queue(&rb->work);
416         else if (cons_pos == rec_pos && !(flags & BPF_RB_NO_WAKEUP))
417                 irq_work_queue(&rb->work);
418 }
419
420 BPF_CALL_2(bpf_ringbuf_submit, void *, sample, u64, flags)
421 {
422         bpf_ringbuf_commit(sample, flags, false /* discard */);
423         return 0;
424 }
425
426 const struct bpf_func_proto bpf_ringbuf_submit_proto = {
427         .func           = bpf_ringbuf_submit,
428         .ret_type       = RET_VOID,
429         .arg1_type      = ARG_PTR_TO_ALLOC_MEM,
430         .arg2_type      = ARG_ANYTHING,
431 };
432
433 BPF_CALL_2(bpf_ringbuf_discard, void *, sample, u64, flags)
434 {
435         bpf_ringbuf_commit(sample, flags, true /* discard */);
436         return 0;
437 }
438
439 const struct bpf_func_proto bpf_ringbuf_discard_proto = {
440         .func           = bpf_ringbuf_discard,
441         .ret_type       = RET_VOID,
442         .arg1_type      = ARG_PTR_TO_ALLOC_MEM,
443         .arg2_type      = ARG_ANYTHING,
444 };
445
446 BPF_CALL_4(bpf_ringbuf_output, struct bpf_map *, map, void *, data, u64, size,
447            u64, flags)
448 {
449         struct bpf_ringbuf_map *rb_map;
450         void *rec;
451
452         if (unlikely(flags & ~(BPF_RB_NO_WAKEUP | BPF_RB_FORCE_WAKEUP)))
453                 return -EINVAL;
454
455         rb_map = container_of(map, struct bpf_ringbuf_map, map);
456         rec = __bpf_ringbuf_reserve(rb_map->rb, size);
457         if (!rec)
458                 return -EAGAIN;
459
460         memcpy(rec, data, size);
461         bpf_ringbuf_commit(rec, flags, false /* discard */);
462         return 0;
463 }
464
465 const struct bpf_func_proto bpf_ringbuf_output_proto = {
466         .func           = bpf_ringbuf_output,
467         .ret_type       = RET_INTEGER,
468         .arg1_type      = ARG_CONST_MAP_PTR,
469         .arg2_type      = ARG_PTR_TO_MEM,
470         .arg3_type      = ARG_CONST_SIZE_OR_ZERO,
471         .arg4_type      = ARG_ANYTHING,
472 };
473
474 BPF_CALL_2(bpf_ringbuf_query, struct bpf_map *, map, u64, flags)
475 {
476         struct bpf_ringbuf *rb;
477
478         rb = container_of(map, struct bpf_ringbuf_map, map)->rb;
479
480         switch (flags) {
481         case BPF_RB_AVAIL_DATA:
482                 return ringbuf_avail_data_sz(rb);
483         case BPF_RB_RING_SIZE:
484                 return rb->mask + 1;
485         case BPF_RB_CONS_POS:
486                 return smp_load_acquire(&rb->consumer_pos);
487         case BPF_RB_PROD_POS:
488                 return smp_load_acquire(&rb->producer_pos);
489         default:
490                 return 0;
491         }
492 }
493
494 const struct bpf_func_proto bpf_ringbuf_query_proto = {
495         .func           = bpf_ringbuf_query,
496         .ret_type       = RET_INTEGER,
497         .arg1_type      = ARG_CONST_MAP_PTR,
498         .arg2_type      = ARG_ANYTHING,
499 };