The ptrace() bug...
Subsystems (2)
Version 2.2.0 (at least) to 2.4.14-pre7 contain this line in mm/memory.c:copy_page_range:
unsigned long cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
Version 2.4.14-pre8 and later changed this line with:
unsigned long cow = (vma->vm_flags & (VM_SHARED | VM_WRITE)) == VM_WRITE;
This line decides if copy-on-write should be active in a vm_area just after a fork. The latter is more correct IMHO, with one exception:
it breaks ptraced programs, because programs been ptraced can see their pages modified without having the VM_WRITE flag, and this causes that both the parent and the child may see their pages changed (copy-on-write doesn't work).
Reverting that one-line patch solves many problems for me and I think there are no other side-effects...
Could it be included in 2.4?
Thanks,
<====================================================================>
mm/memory.c renamed from linux-2.4.17/mm/memory.c +1 -1
@@ -177,7 +177,7 @@pgd_t * src_pgd, * dst_pgd;unsigned long address = vma->vm_start;unsigned long end = vma->vm_end; -unsigned long cow = (vma->vm_flags & (VM_SHARED |VM_WRITE )) ==VM_WRITE ; +unsigned long cow = (vma->vm_flags & (VM_SHARED |VM_MAYWRITE )) ==VM_MAYWRITE ; src_pgd =pgd_offset (src, address)-1 ; dst_pgd =pgd_offset (dst, address)-1 ;
<====================================================================>
Signature · 3 lines
.+'''+. .+'''+. .+'''+. .+'''+. .+'' Juan Cespedes / \ / \ [email protected] .+' `+...+' `+...+' `+...+' `+...+'