From 5209aed5137880fa229746cb521f715e55596460 Mon Sep 17 00:00:00 2001 From: "Jason A. Donenfeld" Date: Thu, 7 Apr 2022 21:23:08 +0200 Subject: random: allow partial reads if later user copies fail Rather than failing entirely if a copy_to_user() fails at some point, instead we should return a partial read for the amount that succeeded prior, unless none succeeded at all, in which case we return -EFAULT as before. This makes it consistent with other reader interfaces. For example, the following snippet for /dev/zero outputs "4" followed by "1": int fd; void *x = mmap(NULL, 4096, PROT_WRITE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0); assert(x != MAP_FAILED); fd = open("/dev/zero", O_RDONLY); assert(fd >= 0); printf("%zd\n", read(fd, x, 4)); printf("%zd\n", read(fd, x + 4095, 4)); close(fd); This brings that same standard behavior to the various RNG reader interfaces. While we're at it, we can streamline the loop logic a little bit. Suggested-by: Linus Torvalds Cc: Jann Horn Signed-off-by: Jason A. Donenfeld --- drivers/char/random.c | 22 ++++++++++++---------- 1 file changed, 12 insertions(+), 10 deletions(-) (limited to 'drivers/char') diff --git a/drivers/char/random.c b/drivers/char/random.c index e15063d61460..df43c5060f00 100644 --- a/drivers/char/random.c +++ b/drivers/char/random.c @@ -523,8 +523,7 @@ EXPORT_SYMBOL(get_random_bytes); static ssize_t get_random_bytes_user(void __user *buf, size_t nbytes) { - ssize_t ret = 0; - size_t len; + size_t len, left, ret = 0; u32 chacha_state[CHACHA_STATE_WORDS]; u8 output[CHACHA_BLOCK_SIZE]; @@ -543,37 +542,40 @@ static ssize_t get_random_bytes_user(void __user *buf, size_t nbytes) * the user directly. */ if (nbytes <= CHACHA_KEY_SIZE) { - ret = copy_to_user(buf, &chacha_state[4], nbytes) ? -EFAULT : nbytes; + ret = nbytes - copy_to_user(buf, &chacha_state[4], nbytes); goto out_zero_chacha; } - do { + for (;;) { chacha20_block(chacha_state, output); if (unlikely(chacha_state[12] == 0)) ++chacha_state[13]; len = min_t(size_t, nbytes, CHACHA_BLOCK_SIZE); - if (copy_to_user(buf, output, len)) { - ret = -EFAULT; + left = copy_to_user(buf, output, len); + if (left) { + ret += len - left; break; } - nbytes -= len; buf += len; ret += len; + nbytes -= len; + if (!nbytes) + break; BUILD_BUG_ON(PAGE_SIZE % CHACHA_BLOCK_SIZE != 0); - if (!(ret % PAGE_SIZE) && nbytes) { + if (ret % PAGE_SIZE == 0) { if (signal_pending(current)) break; cond_resched(); } - } while (nbytes); + } memzero_explicit(output, sizeof(output)); out_zero_chacha: memzero_explicit(chacha_state, sizeof(chacha_state)); - return ret; + return ret ? ret : -EFAULT; } /* -- cgit v1.2.3-71-gd317 From b0c3e796f24b588b862b61ce235d3c9417dc8983 Mon Sep 17 00:00:00 2001 From: "Jason A. Donenfeld" Date: Fri, 8 Apr 2022 18:14:57 +0200 Subject: random: make random_get_entropy() return an unsigned long Some implementations were returning type `unsigned long`, while others that fell back to get_cycles() were implicitly returning a `cycles_t` or an untyped constant int literal. That makes for weird and confusing code, and basically all code in the kernel already handled it like it was an `unsigned long`. I recently tried to handle it as the largest type it could be, a `cycles_t`, but doing so doesn't really help with much. Instead let's just make random_get_entropy() return an unsigned long all the time. This also matches the commonly used `arch_get_random_long()` function, so now RDRAND and RDTSC return the same sized integer, which means one can fallback to the other more gracefully. Cc: Dominik Brodowski Cc: Theodore Ts'o Acked-by: Thomas Gleixner Signed-off-by: Jason A. Donenfeld --- drivers/char/random.c | 20 +++++++------------- include/linux/timex.h | 2 +- 2 files changed, 8 insertions(+), 14 deletions(-) (limited to 'drivers/char') diff --git a/drivers/char/random.c b/drivers/char/random.c index df43c5060f00..6b01b2be9dd4 100644 --- a/drivers/char/random.c +++ b/drivers/char/random.c @@ -1018,7 +1018,7 @@ int __init rand_initialize(void) */ void add_device_randomness(const void *buf, size_t size) { - cycles_t cycles = random_get_entropy(); + unsigned long cycles = random_get_entropy(); unsigned long flags, now = jiffies; if (crng_init == 0 && size) @@ -1049,8 +1049,7 @@ struct timer_rand_state { */ static void add_timer_randomness(struct timer_rand_state *state, unsigned int num) { - cycles_t cycles = random_get_entropy(); - unsigned long flags, now = jiffies; + unsigned long cycles = random_get_entropy(), now = jiffies, flags; long delta, delta2, delta3; spin_lock_irqsave(&input_pool.lock, flags); @@ -1339,8 +1338,7 @@ static void mix_interrupt_randomness(struct work_struct *work) void add_interrupt_randomness(int irq) { enum { MIX_INFLIGHT = 1U << 31 }; - cycles_t cycles = random_get_entropy(); - unsigned long now = jiffies; + unsigned long cycles = random_get_entropy(), now = jiffies; struct fast_pool *fast_pool = this_cpu_ptr(&irq_randomness); struct pt_regs *regs = get_irq_regs(); unsigned int new_count; @@ -1353,16 +1351,12 @@ void add_interrupt_randomness(int irq) if (cycles == 0) cycles = get_reg(fast_pool, regs); - if (sizeof(cycles) == 8) + if (sizeof(unsigned long) == 8) { irq_data.u64[0] = cycles ^ rol64(now, 32) ^ irq; - else { + irq_data.u64[1] = regs ? instruction_pointer(regs) : _RET_IP_; + } else { irq_data.u32[0] = cycles ^ irq; irq_data.u32[1] = now; - } - - if (sizeof(unsigned long) == 8) - irq_data.u64[1] = regs ? instruction_pointer(regs) : _RET_IP_; - else { irq_data.u32[2] = regs ? instruction_pointer(regs) : _RET_IP_; irq_data.u32[3] = get_reg(fast_pool, regs); } @@ -1409,7 +1403,7 @@ static void entropy_timer(struct timer_list *t) static void try_to_generate_entropy(void) { struct { - cycles_t cycles; + unsigned long cycles; struct timer_list timer; } stack; diff --git a/include/linux/timex.h b/include/linux/timex.h index 059b18eb1f1f..5745c90c8800 100644 --- a/include/linux/timex.h +++ b/include/linux/timex.h @@ -75,7 +75,7 @@ * By default we use get_cycles() for this purpose, but individual * architectures may override this in their asm/timex.h header file. */ -#define random_get_entropy() get_cycles() +#define random_get_entropy() ((unsigned long)get_cycles()) #endif /* -- cgit v1.2.3-71-gd317 From 35a33ff3807d3adb9daaf937f5bca002ffa9f84e Mon Sep 17 00:00:00 2001 From: "Jason A. Donenfeld" Date: Thu, 14 Apr 2022 01:50:38 +0200 Subject: random: use memmove instead of memcpy for remaining 32 bytes In order to immediately overwrite the old key on the stack, before servicing a userspace request for bytes, we use the remaining 32 bytes of block 0 as the key. This means moving indices 8,9,a,b,c,d,e,f -> 4,5,6,7,8,9,a,b. Since 4 < 8, for the kernel implementations of memcpy(), this doesn't actually appear to be a problem in practice. But relying on that characteristic seems a bit brittle. So let's change that to a proper memmove(), which is the by-the-books way of handling overlapping memory copies. Reviewed-by: Dominik Brodowski Signed-off-by: Jason A. Donenfeld --- drivers/char/random.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/char') diff --git a/drivers/char/random.c b/drivers/char/random.c index 6b01b2be9dd4..3a293f919af9 100644 --- a/drivers/char/random.c +++ b/drivers/char/random.c @@ -333,7 +333,7 @@ static void crng_fast_key_erasure(u8 key[CHACHA_KEY_SIZE], chacha20_block(chacha_state, first_block); memcpy(key, first_block, CHACHA_KEY_SIZE); - memcpy(random_data, first_block + CHACHA_KEY_SIZE, random_data_len); + memmove(random_data, first_block + CHACHA_KEY_SIZE, random_data_len); memzero_explicit(first_block, sizeof(first_block)); } -- cgit v1.2.3-71-gd317 From 8717627d6ac53251ee012c3c7aca392f29f38a42 Mon Sep 17 00:00:00 2001 From: "Jason A. Donenfeld" Date: Mon, 18 Apr 2022 20:57:31 +0200 Subject: random: document crng_fast_key_erasure() destination possibility This reverts 35a33ff3807d ("random: use memmove instead of memcpy for remaining 32 bytes"), which was made on a totally bogus basis. The thing it was worried about overlapping came from the stack, not from one of its arguments, as Eric pointed out. But the fact that this confusion even happened draws attention to the fact that it's a bit non-obvious that the random_data parameter can alias chacha_state, and in fact should do so when the caller can't rely on the stack being cleared in a timely manner. So this commit documents that. Reported-by: Eric Biggers Reviewed-by: Eric Biggers Signed-off-by: Jason A. Donenfeld --- drivers/char/random.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'drivers/char') diff --git a/drivers/char/random.c b/drivers/char/random.c index 3a293f919af9..4c9adb4f3d5d 100644 --- a/drivers/char/random.c +++ b/drivers/char/random.c @@ -318,6 +318,13 @@ static void crng_reseed(bool force) * the resultant ChaCha state to the user, along with the second * half of the block containing 32 bytes of random data that may * be used; random_data_len may not be greater than 32. + * + * The returned ChaCha state contains within it a copy of the old + * key value, at index 4, so the state should always be zeroed out + * immediately after using in order to maintain forward secrecy. + * If the state cannot be erased in a timely manner, then it is + * safer to set the random_data parameter to &chacha_state[4] so + * that this function overwrites it before returning. */ static void crng_fast_key_erasure(u8 key[CHACHA_KEY_SIZE], u32 chacha_state[CHACHA_STATE_WORDS], @@ -333,7 +340,7 @@ static void crng_fast_key_erasure(u8 key[CHACHA_KEY_SIZE], chacha20_block(chacha_state, first_block); memcpy(key, first_block, CHACHA_KEY_SIZE); - memmove(random_data, first_block + CHACHA_KEY_SIZE, random_data_len); + memcpy(random_data, first_block + CHACHA_KEY_SIZE, random_data_len); memzero_explicit(first_block, sizeof(first_block)); } -- cgit v1.2.3-71-gd317