The MASM Forum

64 bit assembler => UASM Assembler Development => Topic started by: mabdelouahab on September 10, 2020, 09:35:24 PM

Title: BSS
Post by: mabdelouahab on September 10, 2020, 09:35:24 PM

Code: [Select]

.data?
data01 db 1073741824 dup(?)
data02 db 1073741824 dup(?)
.code
main PROC SYSTEMV      _argc:DWORD, _argv:QWORD
xor rax,rax
ret
main ENDP
end
linux: 6963 Segmentation fault   

Code: [Select]

.data?
data01 db 1073741824 dup(?)
data02 db 1073741824 dup(?)
.code
entry_point proc
XOR RAX,RAX
ret
entry_point endp
end
windows: 00000005

Are there limits to BSS, How can I use 4 GB?

Title: Re: BSS
Post by: Vortex on September 10, 2020, 10:38:42 PM

Hi mabdelouahab,

Why not to use dynamic memory allocation functions?

Title: Re: BSS
Post by: BugCatcher on September 11, 2020, 12:11:12 AM

.data?
 data01 QWORD ?

invoke   GlobalAlloc,GMEM_ZEROINIT,1073741824
mov data01,rax

Title: Re: BSS
Post by: mabdelouahab on September 11, 2020, 12:27:09 AM

Thank you Vortex, I know the allocation dynamically , But my question is about the limits of BSS

Code: [Select]

invoke g_malloc,16032385536
mov view,rax
mov rcx,16032385536-32
add rax,rcx
mov dword ptr [rax],"DCBA"
mov dword ptr [rax+4],"HGFE"
mov qword ptr [rax+8],0
invoke printf,str$(<13,10,"0%Xh='%s' ">),rax   ,rax


Quote

07D5C69F0h='ABCDEFGH'

16GB

Title: Re: BSS
Post by: BugCatcher on September 11, 2020, 12:38:11 AM

Its been my experience that large amount bss can be compiled, but it takes forever for the assembler to do it.

Title: Re: BSS
Post by: jj2007 on September 11, 2020, 12:59:15 AM

Quote from: BugCatcher on September 11, 2020, 12:38:11 AM

Its been my experience that large amount bss can be compiled, but it takes forever for the assembler to do it.

This was true for older versions of MASM.

Title: Re: BSS
Post by: Vortex on September 11, 2020, 03:28:09 AM

Hi mabdelouahab,

No any error messages in my test :

Code: [Select]

.data?
data01 db 1073741824 dup(?)
data02 db 1073741824 dup(?)
.code
entry_point proc
XOR RAX,RAX
ret
entry_point endp
end

Code: [Select]

uasm64.exe -win64 test.asm
UASM v2.49, Jun 21 2019, Masm-compatible assembler.
Portions Copyright (c) 1992-2002 Sybase, Inc. All Rights Reserved.
Source code is available under the Sybase Open Watcom Public License.

test.asm: 9 lines, 2 passes, 2 ms, 0 warnings, 0 errors

Title: Re: BSS
Post by: mabdelouahab on September 11, 2020, 03:54:55 AM

Vortex ,
After running: 00000005 (windows) or 6963 Segmentation fault  (linux)

Title: Re: BSS
Post by: mineiro on September 11, 2020, 11:03:13 AM

hello sir mabdelouahab, these are my tests in linux 64
test2.uasm

Code: [Select]

;uasm -elf64 -pie test2.uasm
;ld test2.o -o test2
;./test2 ; echo $?

.X64
option casemap:none

.data?
data01 db 7FFFFFFFh dup(?)
data02 db 7FFFFFFFh dup(?)

.code

_start:
mov rdi,0   ;return code
mov rax,60  ;sysexit
syscall

end _start


Code: [Select]

;---first try
; data01 db 80000000h dup(?)       ;Error A2209: Count must be positive or zero

;---second try
; data01 db 7FFFFFFFh dup(?)       ;OK to assemble, link and execute

;---third try
; data01 db 7FFFFFFFh dup(?)       ;pseudo OK to assemble, link and execute
; data02 db 7FFFFFFFh dup(?)       
;readelf -S test2.o
; [ 3] .bss              NOBITS           0000000000000000  00000210
;      fffffffffffffffe  0000000000000000  WA       0     0     16
;readelf -S test2
;   No bss section created in executable
;size test2
;   text    data     bss     dec     hex filename
;     16       0       0      16      10 test2
;size test2.o
;   text    data     bss     dec     hex filename
;     16       0 18446744073709551614         14       e test2.o


;fourth try, using previous test.o and changing with hexadecimal editor fffffffffffffffeh at offset 120h
;to FF FF FF FF │ 00 00 00 00
;link OK, execute OK
; [ 3] .bss              NOBITS           0000000000000000  00000210            object file
;       00000000ffffffff  0000000000000000  WA       0     0     16
;[ 2] .bss              NOBITS           0000000000402000  00002000             executable
;       0000000100000000  0000000000000000  WA       0     0     16


;fifth try, using previous test.o and changing with hexadecimal editor FF FF FF FF │ 00 00 00 00 at offset 120h
;to 00 00 00 00 │ 10 00 00 00
;link OK, execute return 139 segmentation fault
;[ 3] .bss              NOBITS           0000000000000000  00000210
;       0000001000000000  0000000000000000  WA       0     0     16
;strace ./test2 ; echo $?
;execve("./test2", ["./test2"], 0x7ffc150b9e40 /* 65 vars */) = -1 ENOMEM (Not possible to alloc memory)
;+++ killed by SIGSEGV +++

So I try with as assembler same tests and increasing allocation data:
test3.s

Code: [Select]

#as test3.s -o test3.o
#ld -e _start test3.o -o test3
#./test3 ; echo$?

.code64
.intel_syntax noprefix
.bss
data01: .space  0x40fffffff

.text
.global _start
_start:
mov rdi,0
mov rax,60
syscall

I was thinking in elf64 field but thats 64 bits, so not a problem. After I think in 48 bits addressing of some machines or hardware limitation of some machines. After I think in some link switches but was not able to find anything; so I check ld loader, ... . But when I tried with as assembler I see that can be done.
Maybe some field in object file.