diff mbox series

[12/13] crypto: x86/chacha - yield FPU context only when needed

Message ID 20221219220223.3982176-13-elliott@hpe.com
State New
Headers
Series crypto: x86 - yield FPU context during long loops |

Commit Message

Elliott, Robert (Servers) Dec. 19, 2022, 10:02 p.m. UTC 
  The x86 assembly language implementations using SIMD process data
between kernel_fpu_begin() and kernel_fpu_end() calls. That
disables scheduler preemption, so prevents the CPU core from being
used by other threads.

Rather than break the processing into 4 KiB passes, each of which
unilaterally calls kernel_fpu_begin() and kernel_fpu_end(),
periodically check if the kernel scheduler wants to run something
else on the CPU. If so, yield the kernel FPU context and let the
scheduler intervene.

Suggested-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Robert Elliott <elliott@hpe.com>
---
 arch/x86/crypto/chacha_glue.c | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)
diff mbox series

Patch

diff --git a/arch/x86/crypto/chacha_glue.c b/arch/x86/crypto/chacha_glue.c
index 7b3a1cf0984b..892cbae958b8 100644
--- a/arch/x86/crypto/chacha_glue.c
+++ b/arch/x86/crypto/chacha_glue.c
@@ -146,17 +146,21 @@  void chacha_crypt_arch(u32 *state, u8 *dst, const u8 *src, unsigned int bytes,
 	    bytes <= CHACHA_BLOCK_SIZE)
 		return chacha_crypt_generic(state, dst, src, bytes, nrounds);
 
-	do {
-		unsigned int todo = min_t(unsigned int, bytes, SZ_4K);
+	kernel_fpu_begin();
+	for (;;) {
+		const unsigned int chunk = min(bytes, 4096U);
 
-		kernel_fpu_begin();
-		chacha_dosimd(state, dst, src, todo, nrounds);
-		kernel_fpu_end();
+		chacha_dosimd(state, dst, src, chunk, nrounds);
 
-		bytes -= todo;
-		src += todo;
-		dst += todo;
-	} while (bytes);
+		bytes -= chunk;
+		if (!bytes)
+			break;
+
+		src += chunk;
+		dst += chunk;
+		kernel_fpu_yield();
+	}
+	kernel_fpu_end();
 }
 EXPORT_SYMBOL(chacha_crypt_arch);