Now that fast_mix() has more than one caller, gcc no longer inlines it.
That's fine. But it also doesn't handle the compound literal argument we
pass it very efficiently, nor does it handle the loop as well as it
could. So just expand the code to spell out this function so that it
generates the same code as it did before. Performance-wise, this now
behaves as it did before the last commit. The difference in actual code
size on x86 is 45 bytes, which is less than a cache line.
Signed-off-by: Jason A. Donenfeld <Jason@xxxxxxxxx>
---
drivers/char/random.c | 44 ++++++++++++++++++++++---------------------
1 file changed, 23 insertions(+), 21 deletions(-)
diff --git a/drivers/char/random.c b/drivers/char/random.c
index 818432638c18..a04ecfcb4f1e 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -1101,25 +1101,30 @@ static DEFINE_PER_CPU(struct fast_pool, irq_randomness) = {
* and therefore this has no security on its own. s represents the
* four-word SipHash state, while v represents a two-word input.
*/
-static void fast_mix(unsigned long s[4], const unsigned long v[2])
+static void fast_mix(unsigned long s[4], unsigned long v1, unsigned long v2)
{
- size_t i;
-
- for (i = 0; i < 2; ++i) {
- s[3] ^= v[i];
#ifdef CONFIG_64BIT
- s[0] += s[1]; s[1] = rol64(s[1], 13); s[1] ^= s[0]; s[0] = rol64(s[0], 32);
- s[2] += s[3]; s[3] = rol64(s[3], 16); s[3] ^= s[2];
- s[0] += s[3]; s[3] = rol64(s[3], 21); s[3] ^= s[0];
- s[2] += s[1]; s[1] = rol64(s[1], 17); s[1] ^= s[2]; s[2] = rol64(s[2], 32);
+#define PERM() do { \
+ s[0] += s[1]; s[1] = rol64(s[1], 13); s[1] ^= s[0]; s[0] = rol64(s[0], 32); \
+ s[2] += s[3]; s[3] = rol64(s[3], 16); s[3] ^= s[2]; \
+ s[0] += s[3]; s[3] = rol64(s[3], 21); s[3] ^= s[0]; \
+ s[2] += s[1]; s[1] = rol64(s[1], 17); s[1] ^= s[2]; s[2] = rol64(s[2], 32); \
+} while (0)
#else
- s[0] += s[1]; s[1] = rol32(s[1], 5); s[1] ^= s[0]; s[0] = rol32(s[0], 16);
- s[2] += s[3]; s[3] = rol32(s[3], 8); s[3] ^= s[2];
- s[0] += s[3]; s[3] = rol32(s[3], 7); s[3] ^= s[0];
- s[2] += s[1]; s[1] = rol32(s[1], 13); s[1] ^= s[2]; s[2] = rol32(s[2], 16);
+#define PERM() do { \
+ s[0] += s[1]; s[1] = rol32(s[1], 5); s[1] ^= s[0]; s[0] = rol32(s[0], 16); \
+ s[2] += s[3]; s[3] = rol32(s[3], 8); s[3] ^= s[2]; \
+ s[0] += s[3]; s[3] = rol32(s[3], 7); s[3] ^= s[0]; \
+ s[2] += s[1]; s[1] = rol32(s[1], 13); s[1] ^= s[2]; s[2] = rol32(s[2], 16); \
+} while (0)
#endif
- s[0] ^= v[i];
- }
+
+ s[3] ^= v1;
+ PERM();
+ s[0] ^= v1;
+ s[3] ^= v2;
+ PERM();
+ s[0] ^= v2;
}
#ifdef CONFIG_SMP
@@ -1187,10 +1192,8 @@ void add_interrupt_randomness(int irq)
struct pt_regs *regs = get_irq_regs();
unsigned int new_count;
- fast_mix(fast_pool->pool, (unsigned long[2]){
- entropy,
- (regs ? instruction_pointer(regs) : _RET_IP_) ^ swab(irq)
- });
+ fast_mix(fast_pool->pool, entropy,
+ (regs ? instruction_pointer(regs) : _RET_IP_) ^ swab(irq));
new_count = ++fast_pool->count;
if (new_count & MIX_INFLIGHT)
@@ -1228,8 +1231,7 @@ static void add_timer_randomness(struct timer_rand_state *state, unsigned int nu
long delta, delta2, delta3;
if (in_hardirq()) {
- fast_mix(this_cpu_ptr(&irq_randomness)->pool,
- (unsigned long[2]){ entropy, num });
+ fast_mix(this_cpu_ptr(&irq_randomness)->pool, entropy, num);
return;
}
--
2.35.1
