Fix all debug directives.

The backtrace when a fortify check failed was not correct. This change
adds all of the necessary directives to get a correct backtrace.

Fix the strcmp directives and change all labels to local labels.

Testing:
- Verify that the runtime can decode the stack for __memcpy_chk, __memset_chk,
  __strcpy_chk, __strcat_chk fortify failures.
- Verify that gdb can decode the stack properly when hitting a fortify check.
- Verify that the runtime can decode the stack for a seg fault for all of the
  _chk functions and for memcpy/memset.
- Verify that gdb can decode the stack for a seg fault for all of the _chk
  functions and for memcpy/memset.
- Verify that the runtime can decode the stack for a seg fault for strcmp.
- Verify that gdb can decode the stack for a seg fault in strcmp.

Bug: 10342460
Bug: 10345269

Merge from internal master.

(cherry-picked from 05332f2ce7e542d32ff4d5cd9f60248ad71fbf0d)

Change-Id: Ibc919b117cfe72b9ae97e35bd48185477177c5ca
diff --git a/libc/arch-arm/krait/bionic/__strcat_chk.S b/libc/arch-arm/krait/bionic/__strcat_chk.S
index 4b125c8..ec99077 100644
--- a/libc/arch-arm/krait/bionic/__strcat_chk.S
+++ b/libc/arch-arm/krait/bionic/__strcat_chk.S
@@ -40,13 +40,13 @@
 ENTRY(__strcat_chk)
     .cfi_startproc
     pld     [r0, #0]
-    .save   {r0, lr}
     push    {r0, lr}
+    .save   {r0, lr}
     .cfi_def_cfa_offset 8
     .cfi_rel_offset r0, 0
     .cfi_rel_offset lr, 4
-    .save   {r4, r5}
     push    {r4, r5}
+    .save   {r4, r5}
     .cfi_adjust_cfa_offset 8
     .cfi_rel_offset r4, 0
     .cfi_rel_offset r5, 4
@@ -180,22 +180,29 @@
 .L_strlen_done:
     add     r2, r3, r4
     cmp     r2, lr
-    bgt     .L_fortify_check_failed
+    bgt     __strcat_chk_failed
 
     // Set up the registers for the memcpy code.
     mov     r1, r5
     pld     [r1, #64]
     mov     r2, r4
     add     r0, r0, r3
-    .pad    #-8
     pop     {r4, r5}
-    .cfi_adjust_cfa_offset -8
-    .cfi_restore r4
-    .cfi_restore r5
 
-    #include "memcpy_base.S"
+    .cfi_endproc
+END(__strcat_chk)
 
-.L_fortify_check_failed:
+#define MEMCPY_BASE         __strcat_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcat_chk_memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY(__strcat_chk_failed)
+    .cfi_startproc
+    .save   {r0, lr}
+    .save   {r4, r5}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
     .cfi_adjust_cfa_offset 8
     .cfi_rel_offset r4, 0
     .cfi_rel_offset r5, 4
@@ -211,7 +218,7 @@
     .word   error_string-(1b+4)
 
     .cfi_endproc
-END(__strcat_chk)
+END(__strcat_chk_failed)
 
     .data
 error_string:
diff --git a/libc/arch-arm/krait/bionic/__strcpy_chk.S b/libc/arch-arm/krait/bionic/__strcpy_chk.S
index d5e1db6..7da4d15 100644
--- a/libc/arch-arm/krait/bionic/__strcpy_chk.S
+++ b/libc/arch-arm/krait/bionic/__strcpy_chk.S
@@ -39,8 +39,8 @@
 ENTRY(__strcpy_chk)
     .cfi_startproc
     pld     [r0, #0]
-    .save   {r0, lr}
     push    {r0, lr}
+    .save   {r0, lr}
     .cfi_def_cfa_offset 8
     .cfi_rel_offset r0, 0
     .cfi_rel_offset lr, 4
@@ -151,14 +151,25 @@
     pld     [r1, #64]
     ldr     r0, [sp]
     cmp     r3, lr
-    bge     .L_fortify_check_failed
+    bge     __strcpy_chk_failed
 
     // Add 1 for copy length to get the string terminator.
     add     r2, r3, #1
 
-    #include "memcpy_base.S"
+    .cfi_endproc
+END(__strcpy_chk)
 
-.L_fortify_check_failed:
+#define MEMCPY_BASE         __strcpy_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcpy_chk_memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY(__strcpy_chk_failed)
+    .cfi_startproc
+    .save   {r0, lr}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
+
     ldr     r0, error_message
     ldr     r1, error_code
 1:
@@ -169,7 +180,7 @@
 error_message:
     .word   error_string-(1b+4)
     .cfi_endproc
-END(__strcpy_chk)
+END(__strcpy_chk_failed)
 
     .data
 error_string:
diff --git a/libc/arch-arm/krait/bionic/memcpy.S b/libc/arch-arm/krait/bionic/memcpy.S
index 3b4fca4..e433402 100644
--- a/libc/arch-arm/krait/bionic/memcpy.S
+++ b/libc/arch-arm/krait/bionic/memcpy.S
@@ -55,22 +55,23 @@
 ENTRY(memcpy)
         .cfi_startproc
         pld     [r1, #64]
-        .save   {r0, lr}
         stmfd   sp!, {r0, lr}
+        .save   {r0, lr}
         .cfi_def_cfa_offset 8
         .cfi_rel_offset r0, 0
         .cfi_rel_offset lr, 4
-
-        #include "memcpy_base.S"
         .cfi_endproc
 END(memcpy)
 
-        .fnstart
+#define MEMCPY_BASE         __memcpy_base
+#define MEMCPY_BASE_ALIGNED __memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY(__memcpy_chk_fail)
         .cfi_startproc
-__memcpy_chk_fail:
         // Preserve lr for backtrace.
-        .save   {lr}
         push    {lr}
+        .save   {lr}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset lr, 0
 
@@ -84,7 +85,7 @@
 error_message:
         .word   error_string-(1b+4)
         .cfi_endproc
-        .fnend
+END(__memcpy_chk_fail)
 
         .data
 error_string:
diff --git a/libc/arch-arm/krait/bionic/memcpy_base.S b/libc/arch-arm/krait/bionic/memcpy_base.S
index 48ce477..d87a542 100644
--- a/libc/arch-arm/krait/bionic/memcpy_base.S
+++ b/libc/arch-arm/krait/bionic/memcpy_base.S
@@ -35,6 +35,13 @@
 
 // Assumes neon instructions and a cache line size of 32 bytes.
 
+ENTRY(MEMCPY_BASE)
+        .cfi_startproc
+        .save {r0, lr}
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r0, 0
+        .cfi_rel_offset lr, 4
+
         /* do we have at least 16-bytes to copy (needed for alignment below) */
         cmp         r2, #16
         blo         5f
@@ -115,3 +122,6 @@
 
         ldmfd       sp!, {r0, lr}
         bx          lr
+
+        .cfi_endproc
+END(MEMCPY_BASE)
diff --git a/libc/arch-arm/krait/bionic/strcmp.S b/libc/arch-arm/krait/bionic/strcmp.S
index f26aaf1..d4cf3f4 100644
--- a/libc/arch-arm/krait/bionic/strcmp.S
+++ b/libc/arch-arm/krait/bionic/strcmp.S
@@ -122,7 +122,6 @@
 
         .macro  init
         /* Macro to save temporary registers and prepare magic values.  */
-        .save   {r4-r7}
         subs    sp, sp, #16
         .cfi_def_cfa_offset 16
         strd    r4, r5, [sp, #8]
@@ -178,12 +177,13 @@
         /* Are both strings double-word aligned?  */
         orr     ip, r0, r1
         tst     ip, #7
-        bne     do_align
+        bne     .L_do_align
 
         /* Fast path.  */
+        .save   {r4-r7}
         init
 
-doubleword_aligned:
+.L_doubleword_aligned:
 
         /* Get here when the strings to compare are double-word aligned.  */
         /* Compare two words in every iteration.  */
@@ -196,14 +196,14 @@
         ldrd    r2, r3, [r0], #8
         ldrd    r4, r5, [r1], #8
 
-        magic_compare_and_branch w1=r2, w2=r4, label=return_24
-        magic_compare_and_branch w1=r3, w2=r5, label=return_35
+        magic_compare_and_branch w1=r2, w2=r4, label=.L_return_24
+        magic_compare_and_branch w1=r3, w2=r5, label=.L_return_35
         b       2b
 
-do_align:
+.L_do_align:
         /* Is the first string word-aligned?  */
         ands    ip, r0, #3
-        beq     word_aligned_r0
+        beq     .L_word_aligned_r0
 
         /* Fast compare byte by byte until the first string is word-aligned.  */
         /* The offset of r0 from a word boundary is in ip. Thus, the number of bytes
@@ -211,58 +211,58 @@
         bic     r0, r0, #3
         ldr     r2, [r0], #4
         lsls    ip, ip, #31
-        beq     byte2
-        bcs     byte3
+        beq     .L_byte2
+        bcs     .L_byte3
 
-byte1:
+.L_byte1:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE1_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
+        bne     .L_fast_return
+        m_cbz   reg=r3, label=.L_fast_return
 
-byte2:
+.L_byte2:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE2_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
+        bne     .L_fast_return
+        m_cbz   reg=r3, label=.L_fast_return
 
-byte3:
+.L_byte3:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE3_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbnz  reg=r3, label=word_aligned_r0
+        bne     .L_fast_return
+        m_cbnz  reg=r3, label=.L_word_aligned_r0
 
-fast_return:
+.L_fast_return:
         mov     r0, ip
         bx      lr
 
-word_aligned_r0:
+.L_word_aligned_r0:
         init
         /* The first string is word-aligned.  */
         /* Is the second string word-aligned?  */
         ands    ip, r1, #3
-        bne     strcmp_unaligned
+        bne     .L_strcmp_unaligned
 
-word_aligned:
+.L_word_aligned:
         /* The strings are word-aligned. */
         /* Is the first string double-word aligned?  */
         tst     r0, #4
-        beq     doubleword_aligned_r0
+        beq     .L_doubleword_aligned_r0
 
         /* If r0 is not double-word aligned yet, align it by loading
         and comparing the next word from each string.  */
         ldr     r2, [r0], #4
         ldr     r4, [r1], #4
-        magic_compare_and_branch w1=r2 w2=r4 label=return_24
+        magic_compare_and_branch w1=r2 w2=r4 label=.L_return_24
 
-doubleword_aligned_r0:
+.L_doubleword_aligned_r0:
         /* Get here when r0 is double-word aligned.  */
         /* Is r1 doubleword_aligned?  */
         tst     r1, #4
-        beq     doubleword_aligned
+        beq     .L_doubleword_aligned
 
         /* Get here when the strings to compare are word-aligned,
         r0 is double-word aligned, but r1 is not double-word aligned.  */
@@ -278,9 +278,9 @@
 
         /* Load the next double-word from each string and compare.  */
         ldrd    r2, r3, [r0], #8
-        magic_compare_and_branch w1=r2 w2=r5 label=return_25
+        magic_compare_and_branch w1=r2 w2=r5 label=.L_return_25
         ldrd    r4, r5, [r1], #8
-        magic_compare_and_branch w1=r3 w2=r4 label=return_34
+        magic_compare_and_branch w1=r3 w2=r4 label=.L_return_34
         b       3b
 
         .macro miscmp_word offsetlo offsethi
@@ -304,46 +304,46 @@
         and     r2, r3, r6, S2LOMEM #\offsetlo
         it      eq
         cmpeq   r2, r5
-        bne     return_25
+        bne     .L_return_25
         ldr     r5, [r1], #4
         cmp     ip, #0
         eor r3, r2, r3
         S2HIMEM r2, r5, #\offsethi
         it      eq
         cmpeq   r3, r2
-        bne     return_32
+        bne     .L_return_32
         b       7b
         .endm /* miscmp_word */
 
-strcmp_unaligned:
+.L_strcmp_unaligned:
         /* r0 is word-aligned, r1 is at offset ip from a word.  */
         /* Align r1 to the (previous) word-boundary.  */
         bic     r1, r1, #3
 
         /* Unaligned comparison word by word using LDRs. */
         cmp     ip, #2
-        beq     miscmp_word_16                    /* If ip == 2.  */
-        bge     miscmp_word_24                    /* If ip == 3.  */
+        beq     .L_miscmp_word_16                 /* If ip == 2.  */
+        bge     .L_miscmp_word_24                 /* If ip == 3.  */
         miscmp_word offsetlo=8 offsethi=24        /* If ip == 1.  */
-miscmp_word_24:  miscmp_word offsetlo=24 offsethi=8
+.L_miscmp_word_24:  miscmp_word offsetlo=24 offsethi=8
 
 
-return_32:
+.L_return_32:
         setup_return w1=r3, w2=r2
-        b       do_return
-return_34:
+        b       .L_do_return
+.L_return_34:
         setup_return w1=r3, w2=r4
-        b       do_return
-return_25:
+        b       .L_do_return
+.L_return_25:
         setup_return w1=r2, w2=r5
-        b       do_return
-return_35:
+        b       .L_do_return
+.L_return_35:
         setup_return w1=r3, w2=r5
-        b       do_return
-return_24:
+        b       .L_do_return
+.L_return_24:
         setup_return w1=r2, w2=r4
 
-do_return:
+.L_do_return:
 
 #ifdef __ARMEB__
         mov     r0, ip
@@ -354,7 +354,6 @@
         /* Restore temporaries early, before computing the return value.  */
         ldrd    r6, r7, [sp]
         ldrd    r4, r5, [sp, #8]
-        .pad    #-16
         adds    sp, sp, #16
         .cfi_def_cfa_offset 0
         .cfi_restore r4
@@ -365,7 +364,7 @@
         /* There is a zero or a different byte between r1 and r2.  */
         /* r0 contains a mask of all-zero bytes in r1.  */
         /* Using r0 and not ip here because cbz requires low register.  */
-        m_cbz   reg=r0, label=compute_return_value
+        m_cbz   reg=r0, label=.L_compute_return_value
         clz     r0, r0
         /* r0 contains the number of bits on the left of the first all-zero byte in r1.  */
         rsb     r0, r0, #24
@@ -373,7 +372,7 @@
         lsr     r1, r1, r0
         lsr     r2, r2, r0
 
-compute_return_value:
+.L_compute_return_value:
         movs    r0, #1
         cmp     r1, r2
         /* The return value is computed as follows.
@@ -393,7 +392,7 @@
      * previous version. See bionic/libc/arch-arm/cortex-a15/bionic/strcmp.S
      * for the unedited version of this code.
      */
-miscmp_word_16:
+.L_miscmp_word_16:
 	wp1 .req r0
 	wp2 .req r1
 	b1  .req r2
@@ -465,7 +464,6 @@
     /* Restore registers and stack. */
     ldrd    r6, r7, [sp]
     ldrd    r4, r5, [sp, #8]
-    .pad    #-16
     adds    sp, sp, #16
     .cfi_def_cfa_offset 0
     .cfi_restore r4
@@ -490,7 +488,6 @@
     /* Restore registers and stack. */
     ldrd    r6, r7, [sp]
     ldrd    r4, r5, [sp, #8]
-    .pad    #-16
     adds    sp, sp, #16
     .cfi_def_cfa_offset 0
     .cfi_restore r4