aes_encrypt() now uses AES instructions when available instead of always
using table-based code.  AES instructions are constant-time and don't
benefit from disabling IRQs as a constant-time hardening measure.

In fact, on two architectures (arm and riscv) disabling IRQs is
counterproductive because it prevents the AES instructions from being
used.  (See the may_use_simd() implementation on those architectures.)

Therefore, let's remove the IRQ disabling/enabling and leave the choice
of constant-time hardening measures to the AES library code.

Note that currently the arm table-based AES code (which runs on arm
kernels that don't have ARMv8 CE) disables IRQs, while the generic
table-based AES code does not.  So this does technically regress in
constant-time hardening when that generic code is used.  But as
discussed in commit a22fd0e3c495 ("lib/crypto: aes: Introduce improved
AES library") I think just leaving IRQs enabled is the right choice.
Disabling them is slow and can cause problems, and AES instructions
(which modern CPUs have) solve the problem in a much better way anyway.

Signed-off-by: Eric Biggers <ebiggers@kernel.org>
---
 lib/crypto/aescfb.c | 25 +++----------------------
 1 file changed, 3 insertions(+), 22 deletions(-)

diff --git a/lib/crypto/aescfb.c b/lib/crypto/aescfb.c
index 147e5211728f..e38848d101e3 100644
--- a/lib/crypto/aescfb.c
+++ b/lib/crypto/aescfb.c
@@ -7,29 +7,10 @@
 
 #include <crypto/aes.h>
 #include <crypto/algapi.h>
 #include <linux/export.h>
 #include <linux/module.h>
-#include <asm/irqflags.h>
-
-static void aescfb_encrypt_block(const struct aes_enckey *key, void *dst,
-				 const void *src)
-{
-	unsigned long flags;
-
-	/*
-	 * In AES-CFB, the AES encryption operates on known 'plaintext' (the IV
-	 * and ciphertext), making it susceptible to timing attacks on the
-	 * encryption key. The AES library already mitigates this risk to some
-	 * extent by pulling the entire S-box into the caches before doing any
-	 * substitutions, but this strategy is more effective when running with
-	 * interrupts disabled.
-	 */
-	local_irq_save(flags);
-	aes_encrypt(key, dst, src);
-	local_irq_restore(flags);
-}
 
 /**
  * aescfb_encrypt - Perform AES-CFB encryption on a block of data
  *
  * @key:	The AES-CFB key schedule
@@ -43,11 +24,11 @@ void aescfb_encrypt(const struct aes_enckey *key, u8 *dst, const u8 *src,
 {
 	u8 ks[AES_BLOCK_SIZE];
 	const u8 *v = iv;
 
 	while (len > 0) {
-		aescfb_encrypt_block(key, ks, v);
+		aes_encrypt(key, ks, v);
 		crypto_xor_cpy(dst, src, ks, min(len, AES_BLOCK_SIZE));
 		v = dst;
 
 		dst += AES_BLOCK_SIZE;
 		src += AES_BLOCK_SIZE;
@@ -70,20 +51,20 @@ EXPORT_SYMBOL(aescfb_encrypt);
 void aescfb_decrypt(const struct aes_enckey *key, u8 *dst, const u8 *src,
 		    int len, const u8 iv[AES_BLOCK_SIZE])
 {
 	u8 ks[2][AES_BLOCK_SIZE];
 
-	aescfb_encrypt_block(key, ks[0], iv);
+	aes_encrypt(key, ks[0], iv);
 
 	for (int i = 0; len > 0; i ^= 1) {
 		if (len > AES_BLOCK_SIZE)
 			/*
 			 * Generate the keystream for the next block before
 			 * performing the XOR, as that may update in place and
 			 * overwrite the ciphertext.
 			 */
-			aescfb_encrypt_block(key, ks[!i], src);
+			aes_encrypt(key, ks[!i], src);
 
 		crypto_xor_cpy(dst, src, ks[i], min(len, AES_BLOCK_SIZE));
 
 		dst += AES_BLOCK_SIZE;
 		src += AES_BLOCK_SIZE;

base-commit: d2a68aba8505ce88b39c34ecb3b707c776af79d4
-- 
2.53.0

On Tue, 31 Mar 2026, at 04:44, Eric Biggers wrote:
> aes_encrypt() now uses AES instructions when available instead of always
> using table-based code.  AES instructions are constant-time and don't
> benefit from disabling IRQs as a constant-time hardening measure.
>
> In fact, on two architectures (arm and riscv) disabling IRQs is
> counterproductive because it prevents the AES instructions from being
> used.  (See the may_use_simd() implementation on those architectures.)
>
> Therefore, let's remove the IRQ disabling/enabling and leave the choice
> of constant-time hardening measures to the AES library code.
>
> Note that currently the arm table-based AES code (which runs on arm
> kernels that don't have ARMv8 CE) disables IRQs, while the generic
> table-based AES code does not.  So this does technically regress in
> constant-time hardening when that generic code is used.  But as
> discussed in commit a22fd0e3c495 ("lib/crypto: aes: Introduce improved
> AES library") I think just leaving IRQs enabled is the right choice.
> Disabling them is slow and can cause problems, and AES instructions
> (which modern CPUs have) solve the problem in a much better way anyway.
>
> Signed-off-by: Eric Biggers <ebiggers@kernel.org>
> ---
>  lib/crypto/aescfb.c | 25 +++----------------------
>  1 file changed, 3 insertions(+), 22 deletions(-)
>

Reviewed-by: Ard Biesheuvel <ardb@kernel.org>