| lh | 9ed821d | 2023-04-07 01:36:19 -0700 | [diff] [blame^] | 1 | /* copy no more than N bytes from SRC to DEST, returning the address of |
| 2 | the terminating '\0' in DEST. |
| 3 | For Intel 80x86, x>=3. |
| 4 | Copyright (C) 1994-2015 Free Software Foundation, Inc. |
| 5 | This file is part of the GNU C Library. |
| 6 | Contributed by Ulrich Drepper <drepper@gnu.ai.mit.edu> |
| 7 | Some bug fixes by Alan Modra <Alan@SPRI.Levels.UniSA.Edu.Au> |
| 8 | - original wrote n+1 chars in some cases. |
| 9 | - stpncpy() ought to behave like strncpy() ie. not null-terminate |
| 10 | if limited by n. glibc-1.09 stpncpy() does this. |
| 11 | |
| 12 | The GNU C Library is free software; you can redistribute it and/or |
| 13 | modify it under the terms of the GNU Lesser General Public |
| 14 | License as published by the Free Software Foundation; either |
| 15 | version 2.1 of the License, or (at your option) any later version. |
| 16 | |
| 17 | The GNU C Library is distributed in the hope that it will be useful, |
| 18 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 19 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 20 | Lesser General Public License for more details. |
| 21 | |
| 22 | You should have received a copy of the GNU Lesser General Public |
| 23 | License along with the GNU C Library; if not, see |
| 24 | <http://www.gnu.org/licenses/>. */ |
| 25 | |
| 26 | #include <sysdep.h> |
| 27 | #include "asm-syntax.h" |
| 28 | |
| 29 | #define PARMS 4+4 /* space for 1 saved reg */ |
| 30 | #define RTN PARMS |
| 31 | #define DEST RTN |
| 32 | #define SRC DEST+4 |
| 33 | #define LEN SRC+4 |
| 34 | |
| 35 | .text |
| 36 | ENTRY (__stpncpy) |
| 37 | |
| 38 | pushl %esi |
| 39 | cfi_adjust_cfa_offset (4) |
| 40 | |
| 41 | movl DEST(%esp), %eax |
| 42 | movl SRC(%esp), %esi |
| 43 | cfi_rel_offset (esi, 0) |
| 44 | movl LEN(%esp), %ecx |
| 45 | |
| 46 | subl %eax, %esi /* magic: reduce number of loop variants |
| 47 | to one using addressing mode */ |
| 48 | jmp L(1) /* jump to loop "head" */ |
| 49 | |
| 50 | ALIGN(4) |
| 51 | |
| 52 | /* Four times unfolded loop with two loop counters. We get the |
| 53 | third value (the source address) by using the index+base |
| 54 | addressing mode. */ |
| 55 | L(2): movb (%eax,%esi), %dl /* load current char */ |
| 56 | movb %dl, (%eax) /* and store it */ |
| 57 | testb %dl, %dl /* was it NUL? */ |
| 58 | jz L(7) /* yes, then exit */ |
| 59 | |
| 60 | movb 1(%eax,%esi), %dl /* load current char */ |
| 61 | movb %dl, 1(%eax) /* and store it */ |
| 62 | testb %dl, %dl /* was it NUL? */ |
| 63 | jz L(6) /* yes, then exit */ |
| 64 | |
| 65 | movb 2(%eax,%esi), %dl /* load current char */ |
| 66 | movb %dl, 2(%eax) /* and store it */ |
| 67 | testb %dl, %dl /* was it NUL? */ |
| 68 | jz L(5) /* yes, then exit */ |
| 69 | |
| 70 | movb 3(%eax,%esi), %dl /* load current char */ |
| 71 | movb %dl, 3(%eax) /* and store it */ |
| 72 | testb %dl, %dl /* was it NUL? */ |
| 73 | jz L(4) /* yes, then exit */ |
| 74 | |
| 75 | addl $4, %eax /* increment loop counter for full round */ |
| 76 | |
| 77 | L(1): subl $4, %ecx /* still more than 4 bytes allowed? */ |
| 78 | jae L(2) /* yes, then go to start of loop */ |
| 79 | |
| 80 | /* The maximal remaining 15 bytes are not processed in a loop. */ |
| 81 | |
| 82 | addl $4, %ecx /* correct above subtraction */ |
| 83 | jz L(9) /* maximal allowed char reached => go to end */ |
| 84 | |
| 85 | movb (%eax,%esi), %dl /* load current char */ |
| 86 | movb %dl, (%eax) /* and store it */ |
| 87 | testb %dl, %dl /* was it NUL? */ |
| 88 | jz L(3) /* yes, then exit */ |
| 89 | |
| 90 | incl %eax /* increment pointer */ |
| 91 | decl %ecx /* decrement length counter */ |
| 92 | jz L(9) /* no more allowed => exit */ |
| 93 | |
| 94 | movb (%eax,%esi), %dl /* load current char */ |
| 95 | movb %dl, (%eax) /* and store it */ |
| 96 | testb %dl, %dl /* was it NUL? */ |
| 97 | jz L(3) /* yes, then exit */ |
| 98 | |
| 99 | incl %eax /* increment pointer */ |
| 100 | decl %ecx /* decrement length counter */ |
| 101 | jz L(9) /* no more allowed => exit */ |
| 102 | |
| 103 | movb (%eax,%esi), %dl /* load current char */ |
| 104 | movb %dl, (%eax) /* and store it */ |
| 105 | testb %dl, %dl /* was it NUL? */ |
| 106 | jz L(3) /* yes, then exit */ |
| 107 | |
| 108 | incl %eax /* increment pointer */ |
| 109 | jmp L(9) /* we don't have to test for counter underflow |
| 110 | because we know we had a most 3 bytes |
| 111 | remaining => exit */ |
| 112 | |
| 113 | /* When coming from the main loop we have to adjust the pointer. */ |
| 114 | L(4): decl %ecx /* decrement counter */ |
| 115 | incl %eax /* increment pointer */ |
| 116 | |
| 117 | L(5): decl %ecx /* increment pointer */ |
| 118 | incl %eax /* increment pointer */ |
| 119 | |
| 120 | L(6): decl %ecx /* increment pointer */ |
| 121 | incl %eax /* increment pointer */ |
| 122 | L(7): |
| 123 | |
| 124 | addl $3, %ecx /* correct pre-decrementation of counter |
| 125 | at the beginning of the loop; but why 3 |
| 126 | and not 4? Very simple, we have to count |
| 127 | the NUL char we already wrote. */ |
| 128 | jz L(9) /* counter is also 0 => exit */ |
| 129 | |
| 130 | /* We now have to fill the rest of the buffer with NUL. This |
| 131 | is done in a tricky way. Please note that the addressing mode |
| 132 | used below is not the same we used above. Here we use the |
| 133 | %ecx register. */ |
| 134 | L(8): |
| 135 | movb $0, (%ecx,%eax) /* store NUL char */ |
| 136 | L(3): decl %ecx /* all bytes written? */ |
| 137 | jnz L(8) /* no, then again */ |
| 138 | |
| 139 | L(9): popl %esi /* restore saved register content */ |
| 140 | cfi_adjust_cfa_offset (-4) |
| 141 | cfi_restore (esi) |
| 142 | |
| 143 | ret |
| 144 | END (__stpncpy) |
| 145 | |
| 146 | libc_hidden_def (__stpncpy) |
| 147 | weak_alias (__stpncpy, stpncpy) |