Module Name: src
Committed By: riastradh
Date: Mon Jul 27 20:52:11 UTC 2020
Modified Files:
src/sys/crypto/aes/arch/arm: aes_neon_32.S
Log Message:
PIC for aes_neon_32.S.
Without this, tests/sys/crypto/aes/t_aes fails to start on armv7
because of R_ARM_ABS32 relocations in a nonwritable text segment for
a PIE -- which atf quietly ignores in the final report! Yikes.
To generate a diff of this commit:
cvs rdiff -u -r1.1 -r1.2 src/sys/crypto/aes/arch/arm/aes_neon_32.S
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
Modified files:
Index: src/sys/crypto/aes/arch/arm/aes_neon_32.S
diff -u src/sys/crypto/aes/arch/arm/aes_neon_32.S:1.1 src/sys/crypto/aes/arch/arm/aes_neon_32.S:1.2
--- src/sys/crypto/aes/arch/arm/aes_neon_32.S:1.1 Mon Jun 29 23:57:56 2020
+++ src/sys/crypto/aes/arch/arm/aes_neon_32.S Mon Jul 27 20:52:10 2020
@@ -1,4 +1,4 @@
-/* $NetBSD: aes_neon_32.S,v 1.1 2020/06/29 23:57:56 riastradh Exp $ */
+/* $NetBSD: aes_neon_32.S,v 1.2 2020/07/27 20:52:10 riastradh Exp $ */
/*-
* Copyright (c) 2020 The NetBSD Foundation, Inc.
@@ -30,8 +30,14 @@
.fpu neon
+ .text
+ .p2align 2
+.Lconstants_addr:
+ .long .Lconstants - .
+
.section .rodata
.p2align 4
+.Lconstants:
.type inv,_ASM_TYPE_OBJECT
inv:
@@ -239,7 +245,7 @@ ENTRY(aes_neon_enc1)
* r3: rmod4
* r4: mc_forward
* r5: mc_backward
- * r6,r7,r8,r10,r11: temporaries
+ * r6,r7,r8,r10,r11,r12: temporaries
* q0={d0-d1}: x/ak/A
* q1={d2-d3}: 0x0f0f...
* q2={d4-d5}: lo/k/j/io
@@ -258,23 +264,30 @@ ENTRY(aes_neon_enc1)
* q15={d30-d31}: A2_B/sr[rmod4]
*/
+ /* r12 := .Lconstants - .Lconstants_addr, r11 := .Lconstants_addr */
+ ldr r12, .Lconstants_addr
+ adr r11, .Lconstants_addr
+
vld1.64 {d28-d29}, [r0 :128]! /* q14 = *rk++ */
movw r3, #0
vmov.i8 q1, #0x0f
+ /* r12 := .Lconstants */
+ add r12, r12, r11
+
/* (q4, q5) := (iptlo, ipthi) */
- ldr r6, =iptlo
- ldr r7, =ipthi
+ add r6, r12, #(iptlo - .Lconstants)
+ add r7, r12, #(ipthi - .Lconstants)
vld1.64 {d8-d9}, [r6 :128]
vld1.64 {d10-d11}, [r7 :128]
/* load the rest of the constants */
- ldr r4, =sb1_0
- ldr r5, =sb1_1
- ldr r6, =sb2_0
- ldr r7, =sb2_1
- ldr r8, =inv
- ldr r10, =inva
+ add r4, r12, #(sb1_0 - .Lconstants)
+ add r5, r12, #(sb1_1 - .Lconstants)
+ add r6, r12, #(sb2_0 - .Lconstants)
+ add r7, r12, #(sb2_1 - .Lconstants)
+ add r8, r12, #(inv - .Lconstants)
+ add r10, r12, #(inva - .Lconstants)
vld1.64 {d12-d13}, [r4 :128] /* q6 = sb1[0] */
vld1.64 {d14-d15}, [r5 :128] /* q7 = sb1[1] */
vld1.64 {d16-d17}, [r6 :128] /* q8 = sb2[0] */
@@ -283,8 +296,8 @@ ENTRY(aes_neon_enc1)
vld1.64 {d22-d23}, [r10 :128] /* q11 = inva */
/* (r4, r5) := (&mc_forward[0], &mc_backward[0]) */
- ldr r4, =mc_forward
- ldr r5, =mc_backward
+ add r4, r12, #(mc_forward - .Lconstants)
+ add r5, r12, #(mc_backward - .Lconstants)
/* (q2, q3) := (lo, hi) */
vshr.u8 q3, q0, #4
@@ -392,9 +405,9 @@ ENTRY(aes_neon_enc1)
bne 1b
/* (q6, q7, q15) := (sbo[0], sbo[1], sr[rmod4]) */
- ldr r8, =sr
- ldr r6, =sbo_0
- ldr r7, =sbo_1
+ add r8, r12, #(sr - .Lconstants)
+ add r6, r12, #(sbo_0 - .Lconstants)
+ add r7, r12, #(sbo_1 - .Lconstants)
add r8, r8, r3, lsl #4
vld1.64 {d12-d13}, [r6 :128]
vld1.64 {d14-d15}, [r7 :128]
@@ -469,23 +482,30 @@ ENTRY(aes_neon_dec1)
* q15={d30-d31}: mc/sr[3 & ~(nrounds - 1)]
*/
+ /* r12 := .Lconstants - .Lconstants_addr, r11 := .Lconstants_addr */
+ ldr r12, .Lconstants_addr
+ adr r11, .Lconstants_addr
+
vld1.64 {d28-d29}, [r0 :128]! /* q14 = *rk++ */
rsb r3, r1, #0 /* r3 := ~(x - 1) = -x */
vmov.i8 q1, #0x0f
and r3, r3, #3 /* r3 := 3 & ~(x - 1) */
+ /* r12 := .Lconstants */
+ add r12, r12, r11
+
/* (q4, q5) := (diptlo, dipthi) */
- ldr r6, =diptlo
- ldr r7, =dipthi
+ add r6, r12, #(diptlo - .Lconstants)
+ add r7, r12, #(dipthi - .Lconstants)
vld1.64 {d8-d9}, [r6 :128]
vld1.64 {d10-d11}, [r7 :128]
/* load the rest of the constants */
- ldr r4, =dsbb_0
- ldr r5, =dsbb_1
- ldr r6, =inv
- ldr r7, =inva
- ldr r8, =.Lmc_forward_3
+ add r4, r12, #(dsbb_0 - .Lconstants)
+ add r5, r12, #(dsbb_1 - .Lconstants)
+ add r6, r12, #(inv - .Lconstants)
+ add r7, r12, #(inva - .Lconstants)
+ add r8, r12, #(.Lmc_forward_3 - .Lconstants)
vld1.64 {d12-d13}, [r4 :128] /* q6 := dsbb[0] */
vld1.64 {d14-d15}, [r5 :128] /* q7 := dsbb[1] */
vld1.64 {d20-d21}, [r6 :128] /* q10 := inv */
@@ -504,8 +524,8 @@ ENTRY(aes_neon_dec1)
vtbl.8 d7, {d10-d11}, d7
/* load dsb9 */
- ldr r4, =dsb9_0
- ldr r5, =dsb9_1
+ add r4, r12, #(dsb9_0 - .Lconstants)
+ add r5, r12, #(dsb9_1 - .Lconstants)
vld1.64 {d8-d9}, [r4 :128] /* q4 := dsb9[0] */
vld1.64 {d10-d11}, [r5 :128] /* q5 := dsb9[1] */
@@ -516,7 +536,7 @@ ENTRY(aes_neon_dec1)
b 2f
1: /* load dsbd */
- ldr r4, =dsbd_0
+ add r4, r12, #(dsbd_0 - .Lconstants)
vld1.64 {d16-d17}, [r4 :128]! /* q8 := dsbd[0] */
vld1.64 {d18-d19}, [r4 :128] /* q9 := dsbd[1] */
@@ -543,7 +563,7 @@ ENTRY(aes_neon_dec1)
veor q0, q0, q13
/* load dsbe */
- ldr r4, =dsbe_0
+ add r4, r12, #(dsbe_0 - .Lconstants)
vld1.64 {d16-d17}, [r4 :128]! /* q8 := dsbe[0] */
vld1.64 {d18-d19}, [r4 :128] /* q9 := dsbe[1] */
@@ -619,9 +639,9 @@ ENTRY(aes_neon_dec1)
bne 1b
/* (q6, q7, q15) := (dsbo[0], dsbo[1], sr[i]) */
- ldr r8, =sr
- ldr r6, =dsbo_0
- ldr r7, =dsbo_1
+ add r8, r12, #(sr - .Lconstants)
+ add r6, r12, #(dsbo_0 - .Lconstants)
+ add r7, r12, #(dsbo_1 - .Lconstants)
add r8, r8, r3, lsl #4
vld1.64 {d12-d13}, [r6 :128]
vld1.64 {d14-d15}, [r7 :128]
