ICP: gcm: Allocate hash subkey table separately

While evaluating other assembler implementations it turns out that
the precomputed hash subkey tables vary in size, from 8*16 bytes
(avx2/avx512) up to 48*16 bytes (avx512-vaes), depending on the
implementation.

To be able to handle the size differences later, allocate
`gcm_Htable` dynamically rather then having a fixed size array, and
adapt consumers.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Attila Fülöp <attila@fueloep.org>
Closes #11102

module/icp/algs/modes/gcm.c

@@ -59,10 +59,12 @@ boolean_t gcm_avx_can_use_movbe = B_FALSE;
 static boolean_t gcm_use_avx = B_FALSE;
 #define	GCM_IMPL_USE_AVX	(*(volatile boolean_t *)&gcm_use_avx)
 
+extern boolean_t atomic_toggle_boolean_nv(volatile boolean_t *);
+
 static inline boolean_t gcm_avx_will_work(void);
 static inline void gcm_set_avx(boolean_t);
 static inline boolean_t gcm_toggle_avx(void);
-extern boolean_t atomic_toggle_boolean_nv(volatile boolean_t *);
+static inline size_t gcm_simd_get_htab_size(boolean_t);
 
 static int gcm_mode_encrypt_contiguous_blocks_avx(gcm_ctx_t *, char *, size_t,
     crypto_data_t *, size_t);
@@ -629,6 +631,21 @@ gcm_init_ctx(gcm_ctx_t *gcm_ctx, char *param, size_t block_size,
 			    (volatile boolean_t *)&gcm_avx_can_use_movbe);
 		}
 	}
+	/* Allocate Htab memory as needed. */
+	if (gcm_ctx->gcm_use_avx == B_TRUE) {
+		size_t htab_len = gcm_simd_get_htab_size(gcm_ctx->gcm_use_avx);
+
+		if (htab_len == 0) {
+			return (CRYPTO_MECHANISM_PARAM_INVALID);
+		}
+		gcm_ctx->gcm_htab_len = htab_len;
+		gcm_ctx->gcm_Htable =
+		    (uint64_t *)kmem_alloc(htab_len, gcm_ctx->gcm_kmflag);
+
+		if (gcm_ctx->gcm_Htable == NULL) {
+			return (CRYPTO_HOST_MEMORY);
+		}
+	}
 	/* Avx and non avx context initialization differs from here on. */
 	if (gcm_ctx->gcm_use_avx == B_FALSE) {
 #endif /* ifdef CAN_USE_GCM_ASM */
@@ -689,6 +706,22 @@ gmac_init_ctx(gcm_ctx_t *gcm_ctx, char *param, size_t block_size,
 	if (ks->ops->needs_byteswap == B_TRUE) {
 		gcm_ctx->gcm_use_avx = B_FALSE;
 	}
+	/* Allocate Htab memory as needed. */
+	if (gcm_ctx->gcm_use_avx == B_TRUE) {
+		size_t htab_len = gcm_simd_get_htab_size(gcm_ctx->gcm_use_avx);
+
+		if (htab_len == 0) {
+			return (CRYPTO_MECHANISM_PARAM_INVALID);
+		}
+		gcm_ctx->gcm_htab_len = htab_len;
+		gcm_ctx->gcm_Htable =
+		    (uint64_t *)kmem_alloc(htab_len, gcm_ctx->gcm_kmflag);
+
+		if (gcm_ctx->gcm_Htable == NULL) {
+			return (CRYPTO_HOST_MEMORY);
+		}
+	}
+
 	/* Avx and non avx context initialization differs from here on. */
 	if (gcm_ctx->gcm_use_avx == B_FALSE) {
 #endif	/* ifdef CAN_USE_GCM_ASM */
@@ -1018,7 +1051,7 @@ MODULE_PARM_DESC(icp_gcm_impl, "Select gcm implementation.");
 /* Clear the FPU registers since they hold sensitive internal state. */
 #define	clear_fpu_regs() clear_fpu_regs_avx()
 #define	GHASH_AVX(ctx, in, len) \
-    gcm_ghash_avx((ctx)->gcm_ghash, (const uint64_t (*)[2])(ctx)->gcm_Htable, \
+    gcm_ghash_avx((ctx)->gcm_ghash, (const uint64_t *)(ctx)->gcm_Htable, \
     in, len)
 
 #define	gcm_incr_counter_block(ctx) gcm_incr_counter_block_by(ctx, 1)
@@ -1036,8 +1069,8 @@ extern void gcm_xor_avx(const uint8_t *src, uint8_t *dst);
 extern void aes_encrypt_intel(const uint32_t rk[], int nr,
     const uint32_t pt[4], uint32_t ct[4]);
 
-extern void gcm_init_htab_avx(uint64_t Htable[16][2], const uint64_t H[2]);
+extern void gcm_init_htab_avx(uint64_t *Htable, const uint64_t H[2]);
-extern void gcm_ghash_avx(uint64_t ghash[2], const uint64_t Htable[16][2],
+extern void gcm_ghash_avx(uint64_t ghash[2], const uint64_t *Htable,
     const uint8_t *in, size_t len);
 
 extern size_t aesni_gcm_encrypt(const uint8_t *, uint8_t *, size_t,
@@ -1073,6 +1106,18 @@ gcm_toggle_avx(void)
 	}
 }
 
+static inline size_t
+gcm_simd_get_htab_size(boolean_t simd_mode)
+{
+	switch (simd_mode) {
+	case B_TRUE:
+		return (2 * 6 * 2 * sizeof (uint64_t));
+
+	default:
+		return (0);
+	}
+}
+
 /*
  * Clear sensitive data in the context.
  *
@@ -1088,7 +1133,6 @@ gcm_clear_ctx(gcm_ctx_t *ctx)
 {
 	bzero(ctx->gcm_remainder, sizeof (ctx->gcm_remainder));
 	bzero(ctx->gcm_H, sizeof (ctx->gcm_H));
-	bzero(ctx->gcm_Htable, sizeof (ctx->gcm_Htable));
 	bzero(ctx->gcm_J0, sizeof (ctx->gcm_J0));
 	bzero(ctx->gcm_tmp, sizeof (ctx->gcm_tmp));
 }

module/icp/algs/modes/modes.c

@@ -152,6 +152,14 @@ crypto_free_mode_ctx(void *ctx)
 			vmem_free(((gcm_ctx_t *)ctx)->gcm_pt_buf,
 			    ((gcm_ctx_t *)ctx)->gcm_pt_buf_len);
 
+#ifdef CAN_USE_GCM_ASM
+		if (((gcm_ctx_t *)ctx)->gcm_Htable != NULL) {
+			gcm_ctx_t *gcm_ctx = (gcm_ctx_t *)ctx;
+			bzero(gcm_ctx->gcm_Htable, gcm_ctx->gcm_htab_len);
+			kmem_free(gcm_ctx->gcm_Htable, gcm_ctx->gcm_htab_len);
+		}
+#endif
+
 		kmem_free(ctx, sizeof (gcm_ctx_t));
 	}
 }

module/icp/asm-x86_64/modes/aesni-gcm-x86_64.S

@@ -718,6 +718,8 @@ aesni_gcm_decrypt:
 .cfi_offset	%r14,-48
 	pushq	%r15
 .cfi_offset	%r15,-56
+	pushq	%r9
+.cfi_offset	%r9,-64
 	vzeroupper
 
 	vmovdqu	(%r8),%xmm1
@@ -730,7 +732,8 @@ aesni_gcm_decrypt:
 	andq	$-128,%rsp
 	vmovdqu	(%r11),%xmm0
 	leaq	128(%rcx),%rcx
-	leaq	32+32(%r9),%r9
+	movq	32(%r9),%r9
+	leaq	32(%r9),%r9
 	movl	504-128(%rcx),%ebp	// ICP has a larger offset for rounds.
 	vpshufb	%xmm0,%xmm8,%xmm8
 
@@ -786,7 +789,9 @@ aesni_gcm_decrypt:
 	vmovups	%xmm14,-16(%rsi)
 
 	vpshufb	(%r11),%xmm8,%xmm8
-	vmovdqu	%xmm8,-64(%r9)
+	movq	-56(%rax),%r9
+.cfi_restore	%r9
+	vmovdqu	%xmm8,(%r9)
 
 	vzeroupper
 	movq	-48(%rax),%r15
@@ -924,6 +929,8 @@ aesni_gcm_encrypt:
 .cfi_offset	%r14,-48
 	pushq	%r15
 .cfi_offset	%r15,-56
+	pushq	%r9
+.cfi_offset	%r9,-64
 	vzeroupper
 
 	vmovdqu	(%r8),%xmm1
@@ -966,7 +973,8 @@ aesni_gcm_encrypt:
 	call	_aesni_ctr32_6x
 
 	vmovdqu	(%r9),%xmm8
-	leaq	32+32(%r9),%r9
+	movq	32(%r9),%r9
+	leaq	32(%r9),%r9
 	subq	$12,%rdx
 	movq	$192,%r10
 	vpshufb	%xmm0,%xmm8,%xmm8
@@ -1157,7 +1165,9 @@ aesni_gcm_encrypt:
 	vpxor	%xmm7,%xmm2,%xmm2
 	vpxor	%xmm2,%xmm8,%xmm8
 	vpshufb	(%r11),%xmm8,%xmm8
-	vmovdqu	%xmm8,-64(%r9)
+	movq	-56(%rax),%r9
+.cfi_restore	%r9
+	vmovdqu	%xmm8,(%r9)
 
 	vzeroupper
 	movq	-48(%rax),%r15

module/icp/include/modes/modes.h

@@ -219,14 +219,14 @@ typedef struct gcm_ctx {
 	size_t gcm_pt_buf_len;
 	uint32_t gcm_tmp[4];
 	/*
-	 * The relative positions of gcm_ghash, gcm_H and pre-computed
+	 * The offset of gcm_Htable relative to gcm_ghash, (32), is hard coded
-	 * gcm_Htable are hard coded in aesni-gcm-x86_64.S and ghash-x86_64.S,
+	 * in aesni-gcm-x86_64.S, so please don't change (or adjust there).
-	 * so please don't change (or adjust accordingly).
 	 */
 	uint64_t gcm_ghash[2];
 	uint64_t gcm_H[2];
 #ifdef CAN_USE_GCM_ASM
-	uint64_t gcm_Htable[12][2];
+	uint64_t *gcm_Htable;
+	size_t gcm_htab_len;
 #endif
 	uint64_t gcm_J0[2];
 	uint64_t gcm_len_a_len_c[2];

module/icp/io/aes.c

@@ -1051,6 +1051,16 @@ out:
 		bzero(aes_ctx.ac_keysched, aes_ctx.ac_keysched_len);
 		kmem_free(aes_ctx.ac_keysched, aes_ctx.ac_keysched_len);
 	}
+#ifdef CAN_USE_GCM_ASM
+	if (aes_ctx.ac_flags & (GCM_MODE|GMAC_MODE) &&
+	    ((gcm_ctx_t *)&aes_ctx)->gcm_Htable != NULL) {
+
+		gcm_ctx_t *ctx = (gcm_ctx_t *)&aes_ctx;
+
+		bzero(ctx->gcm_Htable, ctx->gcm_htab_len);
+		kmem_free(ctx->gcm_Htable, ctx->gcm_htab_len);
+	}
+#endif
 
 	return (ret);
 }
@@ -1209,6 +1219,14 @@ out:
 			vmem_free(((gcm_ctx_t *)&aes_ctx)->gcm_pt_buf,
 			    ((gcm_ctx_t *)&aes_ctx)->gcm_pt_buf_len);
 		}
+#ifdef CAN_USE_GCM_ASM
+		if (((gcm_ctx_t *)&aes_ctx)->gcm_Htable != NULL) {
+			gcm_ctx_t *ctx = (gcm_ctx_t *)&aes_ctx;
+
+			bzero(ctx->gcm_Htable, ctx->gcm_htab_len);
+			kmem_free(ctx->gcm_Htable, ctx->gcm_htab_len);
+		}
+#endif
 	}
 
 	return (ret);