a little less jank salsa20

RustCrypto · Jul 12, 2024 · 7079e5d · 7079e5d
1 parent 353384d
commit 7079e5d
Show file tree

Hide file tree

Showing 2 changed files with 27 additions and 47 deletions.
diff --git a/yescrypt/src/lib.rs b/yescrypt/src/lib.rs
@@ -1665,10 +1665,7 @@ unsafe fn blockmix_pwxform(mut B: *mut uint32_t, mut ctx: *mut PwxformCtx, mut r
         .wrapping_sub(1 as libc::c_int as libc::c_ulong)
         .wrapping_mul((4 as libc::c_int * 2 as libc::c_int * 8 as libc::c_int) as libc::c_ulong)
         .wrapping_div(64 as libc::c_int as libc::c_ulong);
-    salsa20::salsa20(
-        &mut *B.offset(i.wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize),
-        2 as libc::c_int as uint32_t,
-    );
+    salsa20::salsa20_2(&mut *B.offset(i.wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize));
     i = i.wrapping_add(1);
     i;
     while i < (2 as libc::c_int as libc::c_ulong).wrapping_mul(r) {
@@ -1680,9 +1677,8 @@ unsafe fn blockmix_pwxform(mut B: *mut uint32_t, mut ctx: *mut PwxformCtx, mut r
             ),
             16 as libc::c_int as size_t,
         );
-        salsa20::salsa20(
+        salsa20::salsa20_2(
             &mut *B.offset(i.wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize),
-            2 as libc::c_int as uint32_t,
         );
         i = i.wrapping_add(1);
         i;

diff --git a/yescrypt/src/salsa20.rs b/yescrypt/src/salsa20.rs
@@ -1,9 +1,15 @@
+use salsa20::cipher::Unsigned;
+
 use crate::{
     common::{blkcpy, blkxor},
     size_t, uint32_t,
 };
 
-pub(crate) unsafe fn salsa20(mut B: *mut uint32_t, mut rounds: uint32_t) {
+pub(crate) unsafe fn salsa20_2(mut B: *mut uint32_t) {
+    salsa20::<salsa20::cipher::consts::U1>(B);
+}
+
+unsafe fn salsa20<R: Unsigned>(mut B: *mut uint32_t) {
     let mut x: [uint32_t; 16] = [0; 16];
     for i in 0..16 {
         x[i * 5 % 16] = *B.offset(i as isize);
@@ -12,14 +18,7 @@ pub(crate) unsafe fn salsa20(mut B: *mut uint32_t, mut rounds: uint32_t) {
     use salsa20::cipher::StreamCipherCore;
 
     let mut block = [0u8; 64];
-
-    if rounds == 2 {
-        salsa20::SalsaCore::<salsa20::cipher::consts::U1>::from_raw_state(x)
-            .write_keystream_block((&mut block).into());
-    } else if rounds == 8 {
-        salsa20::SalsaCore::<salsa20::cipher::consts::U4>::from_raw_state(x)
-            .write_keystream_block((&mut block).into());
-    }
+    salsa20::SalsaCore::<R>::from_raw_state(x).write_keystream_block((&mut block).into());
 
     for (c, b) in block.chunks_exact(4).zip(x.iter_mut()) {
         *b = u32::from_le_bytes(c.try_into().expect("4 bytes is 1 u32")).wrapping_sub(*b);
@@ -36,56 +35,41 @@ pub(crate) unsafe fn blockmix_salsa8(mut B: *mut uint32_t, mut Y: *mut uint32_t,
     let mut i: size_t = 0;
     blkcpy(
         X.as_mut_ptr(),
-        &mut *B.offset(
-            (2 as libc::c_int as libc::c_ulong)
-                .wrapping_mul(r)
-                .wrapping_sub(1 as libc::c_int as libc::c_ulong)
-                .wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize,
-        ),
-        16 as libc::c_int as size_t,
+        &mut *B.offset(2_u64.wrapping_mul(r).wrapping_sub(1).wrapping_mul(16) as isize),
+        16,
     );
-    i = 0 as libc::c_int as size_t;
-    while i < (2 as libc::c_int as libc::c_ulong).wrapping_mul(r) {
+    i = 0;
+    while i < 2_u64.wrapping_mul(r) {
         blkxor(
             X.as_mut_ptr(),
-            &mut *B.offset(i.wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize),
-            16 as libc::c_int as size_t,
+            &mut *B.offset(i.wrapping_mul(16) as isize),
+            16,
         );
-        salsa20(X.as_mut_ptr(), 8);
+        salsa20::<salsa20::cipher::consts::U4>(X.as_mut_ptr());
         blkcpy(
-            &mut *Y.offset(i.wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize),
+            &mut *Y.offset(i.wrapping_mul(16) as isize),
             X.as_mut_ptr(),
-            16 as libc::c_int as size_t,
+            16,
         );
         i = i.wrapping_add(1);
         i;
     }
-    i = 0 as libc::c_int as size_t;
+    i = 0;
     while i < r {
         blkcpy(
-            &mut *B.offset(i.wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize),
-            &mut *Y.offset(
-                i.wrapping_mul(2 as libc::c_int as libc::c_ulong)
-                    .wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize,
-            ),
-            16 as libc::c_int as size_t,
+            &mut *B.offset(i.wrapping_mul(16) as isize),
+            &mut *Y.offset(i.wrapping_mul(2).wrapping_mul(16) as isize),
+            16,
         );
         i = i.wrapping_add(1);
         i;
     }
-    i = 0 as libc::c_int as size_t;
+    i = 0;
     while i < r {
         blkcpy(
-            &mut *B.offset(
-                i.wrapping_add(r)
-                    .wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize,
-            ),
-            &mut *Y.offset(
-                i.wrapping_mul(2 as libc::c_int as libc::c_ulong)
-                    .wrapping_add(1 as libc::c_int as libc::c_ulong)
-                    .wrapping_mul(16 as libc::c_int as libc::c_ulong) as isize,
-            ),
-            16 as libc::c_int as size_t,
+            &mut *B.offset(i.wrapping_add(r).wrapping_mul(16) as isize),
+            &mut *Y.offset(i.wrapping_mul(2).wrapping_add(1).wrapping_mul(16) as isize),
+            16,
         );
         i = i.wrapping_add(1);
         i;