[PATCH 05/13] crypto: x86/sm3 - yield FPU context during long loops

[Robert Elliott]

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.

The update() and finup() functions might be called to process
large quantities of data, which can result in RCU stalls and
soft lockups.

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.

Fixes: 930ab34d906d ("crypto: x86/sm3 - add AVX assembly implementation")
Suggested-by: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx>
Signed-off-by: Robert Elliott <elliott@xxxxxxx>
---
 arch/x86/crypto/sm3_avx_glue.c | 34 +++++++++++++++++++++++++++++-----
 1 file changed, 29 insertions(+), 5 deletions(-)

diff --git a/arch/x86/crypto/sm3_avx_glue.c b/arch/x86/crypto/sm3_avx_glue.c
index 661b6f22ffcd..9e4b21c0e748 100644
--- a/arch/x86/crypto/sm3_avx_glue.c
+++ b/arch/x86/crypto/sm3_avx_glue.c
@@ -25,8 +25,7 @@ static int sm3_avx_update(struct shash_desc *desc, const u8 *data,
 {
 	struct sm3_state *sctx = shash_desc_ctx(desc);
 
-	if (!crypto_simd_usable() ||
-			(sctx->count % SM3_BLOCK_SIZE) + len < SM3_BLOCK_SIZE) {
+	if (((sctx->count % SM3_BLOCK_SIZE) + len < SM3_BLOCK_SIZE) || !crypto_simd_usable()) {
 		sm3_update(sctx, data, len);
 		return 0;
 	}
@@ -38,7 +37,19 @@ static int sm3_avx_update(struct shash_desc *desc, const u8 *data,
 	BUILD_BUG_ON(offsetof(struct sm3_state, state) != 0);
 
 	kernel_fpu_begin();
-	sm3_base_do_update(desc, data, len, sm3_transform_avx);
+	for (;;) {
+		const unsigned int chunk = min(len, 4096U);
+
+		sm3_base_do_update(desc, data, chunk, sm3_transform_avx);
+
+		len -= chunk;
+
+		if (!len)
+			break;
+
+		data += chunk;
+		kernel_fpu_yield();
+	}
 	kernel_fpu_end();
 
 	return 0;
@@ -58,8 +69,21 @@ static int sm3_avx_finup(struct shash_desc *desc, const u8 *data,
 	}
 
 	kernel_fpu_begin();
-	if (len)
-		sm3_base_do_update(desc, data, len, sm3_transform_avx);
+	if (len) {
+		for (;;) {
+			const unsigned int chunk = min(len, 4096U);
+
+			sm3_base_do_update(desc, data, chunk, sm3_transform_avx);
+			len -= chunk;
+
+			if (!len)
+				break;
+
+			data += chunk;
+			kernel_fpu_yield();
+		}
+	}
+
 	sm3_base_do_finalize(desc, sm3_transform_avx);
 	kernel_fpu_end();
 
-- 
2.38.1