backported fixes from 256boss

[bootcensus] / src / libc / string_asm.s

diff --git a/src/libc/string_asm.s b/src/libc/string_asm.s
index 7bb6f40..a15a405 100644 (file)
--- a/src/libc/string_asm.s
+++ b/src/libc/string_asm.s
@@ -22,6 +22,7 @@ memset:
        push %edi
 
        mov 8(%ebp), %edi
+       push %edi
        mov 12(%ebp), %al
        mov %al, %ah
        mov %ax, %cx
@@ -32,25 +33,29 @@ memset:
        cmp $0, %ecx
        jz msdone
 
+       # write 1, 2, or 3 times until we reache a 32bit-aligned dest address
        mov %edi, %edx
        and $3, %edx
+       jz msmain
        jmp *mspre_tab(,%edx,4)
 
 mspre_tab: .long msmain, mspre1, mspre2, mspre3
-mspre1:        stosb
+mspre3:        stosb
        dec %ecx
 mspre2:        stosb
        dec %ecx
-mspre3:        stosb
+mspre1:        stosb
        dec %ecx
        jz msdone
 
+       # edi is 32bit-aligned here, write ecx>>2 32bit values
 msmain:
        push %ecx
        shr $2, %ecx
        rep stosl
        pop %ecx
 
+       # write any trailing bytes
        and $3, %ecx
        jmp *mspost_tab(,%ecx,4)
 
@@ -60,6 +65,7 @@ mspost2:stosb
 mspost1:stosb
 
 msdone:
+       pop %eax
        pop %edi
        pop %ebp
        ret
@@ -74,6 +80,7 @@ memset16:
        push %edi
 
        mov 8(%ebp), %edi
+       push %edi
        mov 12(%ebp), %ax
        shl $16, %eax
        mov 12(%ebp), %ax
@@ -84,6 +91,7 @@ memset16:
 
        mov %edi, %edx
        and $3, %edx
+       jz ms16main
        jmp *ms16pre_tab(,%edx,4)
 
 ms16pre_tab: .long ms16main, ms16pre1, ms16pre2, ms16pre3
@@ -125,7 +133,7 @@ ms16main:
        jz ms16done
        stosw
 ms16done:
-
+       pop %eax
        pop %edi
        pop %ebp
        ret
@@ -145,27 +153,12 @@ memcpy:
        cmp $0, %ecx
        jz mcdone
 
-       mov %edi, %edx
-       and $3, %edx
-       jmp *mcpre_tab(,%edx,4)
-
-mcpre_tab: .long mcmain, mcpre1, mcpre2, mcpre3
-mcpre1:        movsw
-       dec %ecx
-mcpre2:        movsw
-       dec %ecx
-mcpre3:        movsw
-       dec %ecx
-       jz mcdone
-
-mcmain:
-       push %ecx
+       mov %ecx, %edx
        shr $2, %ecx
        rep movsl
-       pop %ecx
 
-       and $3, %ecx
-       jmp *mcpost_tab(,%ecx,4)
+       and $3, %edx
+       jmp *mcpost_tab(,%edx,4)
 
 mcpost_tab: .long mcdone, mcpost1, mcpost2, mcpost3
 mcpost3:movsb