1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (C) 2022-2024 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved. 4 */ 5 6 #include <linux/array_size.h> 7 #include <linux/minmax.h> 8 #include <vdso/datapage.h> 9 #include <vdso/getrandom.h> 10 #include <vdso/limits.h> 11 #include <vdso/page.h> 12 #include <vdso/unaligned.h> 13 #include <asm/barrier.h> 14 #include <asm/vdso/getrandom.h> 15 #include <uapi/linux/errno.h> 16 #include <uapi/linux/mman.h> 17 #include <uapi/linux/random.h> 18 19 /* Bring in default accessors */ 20 #include <vdso/vsyscall.h> 21 22 #define MEMCPY_AND_ZERO_SRC(type, dst, src, len) do { \ 23 while (len >= sizeof(type)) { \ 24 __put_unaligned_t(type, __get_unaligned_t(type, src), dst); \ 25 __put_unaligned_t(type, 0, src); \ 26 dst += sizeof(type); \ 27 src += sizeof(type); \ 28 len -= sizeof(type); \ 29 } \ 30 } while (0) 31 32 static void memcpy_and_zero_src(void *dst, void *src, size_t len) 33 { 34 if (IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)) { 35 if (IS_ENABLED(CONFIG_64BIT)) 36 MEMCPY_AND_ZERO_SRC(u64, dst, src, len); 37 MEMCPY_AND_ZERO_SRC(u32, dst, src, len); 38 MEMCPY_AND_ZERO_SRC(u16, dst, src, len); 39 } 40 MEMCPY_AND_ZERO_SRC(u8, dst, src, len); 41 } 42 43 /** 44 * __cvdso_getrandom_data - Generic vDSO implementation of getrandom() syscall. 45 * @rng_info: Describes state of kernel RNG, memory shared with kernel. 46 * @buffer: Destination buffer to fill with random bytes. 47 * @len: Size of @buffer in bytes. 48 * @flags: Zero or more GRND_* flags. 49 * @opaque_state: Pointer to an opaque state area. 50 * @opaque_len: Length of opaque state area. 51 * 52 * This implements a "fast key erasure" RNG using ChaCha20, in the same way that the kernel's 53 * getrandom() syscall does. It periodically reseeds its key from the kernel's RNG, at the same 54 * schedule that the kernel's RNG is reseeded. If the kernel's RNG is not ready, then this always 55 * calls into the syscall. 56 * 57 * If @buffer, @len, and @flags are 0, and @opaque_len is ~0UL, then @opaque_state is populated 58 * with a struct vgetrandom_opaque_params and the function returns 0; if it does not return 0, 59 * this function should not be used. 60 * 61 * @opaque_state *must* be allocated by calling mmap(2) using the mmap_prot and mmap_flags fields 62 * from the struct vgetrandom_opaque_params, and states must not straddle pages. Unless external 63 * locking is used, one state must be allocated per thread, as it is not safe to call this function 64 * concurrently with the same @opaque_state. However, it is safe to call this using the same 65 * @opaque_state that is shared between main code and signal handling code, within the same thread. 66 * 67 * Returns: The number of random bytes written to @buffer, or a negative value indicating an error. 68 */ 69 static __always_inline ssize_t 70 __cvdso_getrandom_data(const struct vdso_rng_data *rng_info, void *buffer, size_t len, 71 unsigned int flags, void *opaque_state, size_t opaque_len) 72 { 73 ssize_t ret = min_t(size_t, INT_MAX & PAGE_MASK /* = MAX_RW_COUNT */, len); 74 struct vgetrandom_state *state = opaque_state; 75 size_t batch_len, nblocks, orig_len = len; 76 bool in_use, have_retried = false; 77 void *orig_buffer = buffer; 78 u64 current_generation; 79 u32 counter[2] = { 0 }; 80 81 if (unlikely(opaque_len == ~0UL && !buffer && !len && !flags)) { 82 struct vgetrandom_opaque_params *params = opaque_state; 83 params->size_of_opaque_state = sizeof(*state); 84 params->mmap_prot = PROT_READ | PROT_WRITE; 85 params->mmap_flags = MAP_DROPPABLE | MAP_ANONYMOUS; 86 for (size_t i = 0; i < ARRAY_SIZE(params->reserved); ++i) 87 params->reserved[i] = 0; 88 return 0; 89 } 90 91 /* The state must not straddle a page, since pages can be zeroed at any time. */ 92 if (unlikely(((unsigned long)opaque_state & ~PAGE_MASK) + sizeof(*state) > PAGE_SIZE)) 93 return -EFAULT; 94 95 /* Handle unexpected flags by falling back to the kernel. */ 96 if (unlikely(flags & ~(GRND_NONBLOCK | GRND_RANDOM | GRND_INSECURE))) 97 goto fallback_syscall; 98 99 /* If the caller passes the wrong size, which might happen due to CRIU, fallback. */ 100 if (unlikely(opaque_len != sizeof(*state))) 101 goto fallback_syscall; 102 103 /* 104 * If the kernel's RNG is not yet ready, then it's not possible to provide random bytes from 105 * userspace, because A) the various @flags require this to block, or not, depending on 106 * various factors unavailable to userspace, and B) the kernel's behavior before the RNG is 107 * ready is to reseed from the entropy pool at every invocation. 108 */ 109 if (unlikely(!READ_ONCE(rng_info->is_ready))) 110 goto fallback_syscall; 111 112 /* 113 * This condition is checked after @rng_info->is_ready, because before the kernel's RNG is 114 * initialized, the @flags parameter may require this to block or return an error, even when 115 * len is zero. 116 */ 117 if (unlikely(!len)) 118 return 0; 119 120 /* 121 * @state->in_use is basic reentrancy protection against this running in a signal handler 122 * with the same @opaque_state, but obviously not atomic wrt multiple CPUs or more than one 123 * level of reentrancy. If a signal interrupts this after reading @state->in_use, but before 124 * writing @state->in_use, there is still no race, because the signal handler will run to 125 * its completion before returning execution. 126 */ 127 in_use = READ_ONCE(state->in_use); 128 if (unlikely(in_use)) 129 /* The syscall simply fills the buffer and does not touch @state, so fallback. */ 130 goto fallback_syscall; 131 WRITE_ONCE(state->in_use, true); 132 133 retry_generation: 134 /* 135 * @rng_info->generation must always be read here, as it serializes @state->key with the 136 * kernel's RNG reseeding schedule. 137 */ 138 current_generation = READ_ONCE(rng_info->generation); 139 140 /* 141 * If @state->generation doesn't match the kernel RNG's generation, then it means the 142 * kernel's RNG has reseeded, and so @state->key is reseeded as well. 143 */ 144 if (unlikely(state->generation != current_generation)) { 145 /* 146 * Write the generation before filling the key, in case of fork. If there is a fork 147 * just after this line, the parent and child will get different random bytes from 148 * the syscall, which is good. However, were this line to occur after the getrandom 149 * syscall, then both child and parent could have the same bytes and the same 150 * generation counter, so the fork would not be detected. Therefore, write 151 * @state->generation before the call to the getrandom syscall. 152 */ 153 WRITE_ONCE(state->generation, current_generation); 154 155 /* 156 * Prevent the syscall from being reordered wrt current_generation. Pairs with the 157 * smp_store_release(&vdso_k_rng_data->generation) in random.c. 158 */ 159 smp_rmb(); 160 161 /* Reseed @state->key using fresh bytes from the kernel. */ 162 if (getrandom_syscall(state->key, sizeof(state->key), 0) != sizeof(state->key)) { 163 /* 164 * If the syscall failed to refresh the key, then @state->key is now 165 * invalid, so invalidate the generation so that it is not used again, and 166 * fallback to using the syscall entirely. 167 */ 168 WRITE_ONCE(state->generation, 0); 169 170 /* 171 * Set @state->in_use to false only after the last write to @state in the 172 * line above. 173 */ 174 WRITE_ONCE(state->in_use, false); 175 176 goto fallback_syscall; 177 } 178 179 /* 180 * Set @state->pos to beyond the end of the batch, so that the batch is refilled 181 * using the new key. 182 */ 183 state->pos = sizeof(state->batch); 184 } 185 186 /* Set len to the total amount of bytes that this function is allowed to read, ret. */ 187 len = ret; 188 more_batch: 189 /* 190 * First use bytes out of @state->batch, which may have been filled by the last call to this 191 * function. 192 */ 193 batch_len = min_t(size_t, sizeof(state->batch) - state->pos, len); 194 if (batch_len) { 195 /* Zeroing at the same time as memcpying helps preserve forward secrecy. */ 196 memcpy_and_zero_src(buffer, state->batch + state->pos, batch_len); 197 state->pos += batch_len; 198 buffer += batch_len; 199 len -= batch_len; 200 } 201 202 if (!len) { 203 /* Prevent the loop from being reordered wrt ->generation. */ 204 barrier(); 205 206 /* 207 * Since @rng_info->generation will never be 0, re-read @state->generation, rather 208 * than using the local current_generation variable, to learn whether a fork 209 * occurred or if @state was zeroed due to memory pressure. Primarily, though, this 210 * indicates whether the kernel's RNG has reseeded, in which case generate a new key 211 * and start over. 212 */ 213 if (unlikely(READ_ONCE(state->generation) != READ_ONCE(rng_info->generation))) { 214 /* 215 * Prevent this from looping forever in case of low memory or racing with a 216 * user force-reseeding the kernel's RNG using the ioctl. 217 */ 218 if (have_retried) { 219 WRITE_ONCE(state->in_use, false); 220 goto fallback_syscall; 221 } 222 223 have_retried = true; 224 buffer = orig_buffer; 225 goto retry_generation; 226 } 227 228 /* 229 * Set @state->in_use to false only when there will be no more reads or writes of 230 * @state. 231 */ 232 WRITE_ONCE(state->in_use, false); 233 return ret; 234 } 235 236 /* Generate blocks of RNG output directly into @buffer while there's enough room left. */ 237 nblocks = len / CHACHA_BLOCK_SIZE; 238 if (nblocks) { 239 __arch_chacha20_blocks_nostack(buffer, state->key, counter, nblocks); 240 buffer += nblocks * CHACHA_BLOCK_SIZE; 241 len -= nblocks * CHACHA_BLOCK_SIZE; 242 } 243 244 BUILD_BUG_ON(sizeof(state->batch_key) % CHACHA_BLOCK_SIZE != 0); 245 246 /* Refill the batch and overwrite the key, in order to preserve forward secrecy. */ 247 __arch_chacha20_blocks_nostack(state->batch_key, state->key, counter, 248 sizeof(state->batch_key) / CHACHA_BLOCK_SIZE); 249 250 /* Since the batch was just refilled, set the position back to 0 to indicate a full batch. */ 251 state->pos = 0; 252 goto more_batch; 253 254 fallback_syscall: 255 return getrandom_syscall(orig_buffer, orig_len, flags); 256 } 257 258 static __always_inline ssize_t 259 __cvdso_getrandom(void *buffer, size_t len, unsigned int flags, void *opaque_state, size_t opaque_len) 260 { 261 return __cvdso_getrandom_data(__arch_get_vdso_u_rng_data(), buffer, len, flags, 262 opaque_state, opaque_len); 263 } 264