Re: [RFC-v5] Fix .text section offset for windows DLL (was Calling __stdcall functions in the inferior)

[Pedro Alves <palves@redhat.com>]

On 12/13/2012 11:07 AM, Pedro Alves wrote:
>> >   Pedro,
>> > you are right that my just committed patch does not 
>> > fix the issue for windows gdbserver...
>> >   So your idea to share the new function between gdb and gdbserver
>> > seemed indeed appealing, but when I looked at the new function pe_text_section_offset
>> > it doesn't seem so easy to me to remove bfd dependency...

It actually looks quite easy to me.

>> > 
>> >   I will try to come up with a fix for gdbserver,
>> > but I am not sure it will be soon...
> Thanks.


> CORE_ADDR
> pe_text_section_offset (struct bfd *abfd)
>
> {
>   unsigned long pe_header_offset, opthdr_ofs, num_entries, i;
>   unsigned long export_rva, export_size, nsections, secptr, expptr;
>   unsigned long exp_funcbase;
>   unsigned char *expdata, *erva;
>   unsigned long name_rvas, ordinals, nexp, ordbase;
>   char *dll_name;
>   int is_pe64 = 0;
>   int is_pe32 = 0;
>   char const *target;
>
>   if (!abfd)
>     return DEFAULT_COFF_PE_TEXT_SECTION_OFFSET;
>
>   target = bfd_get_target (abfd);
>
>   is_pe64 = (strcmp (target, "pe-x86-64") == 0
> 	     || strcmp (target, "pei-x86-64") == 0);
>   is_pe32 = (strcmp (target, "pe-i386") == 0
> 	     || strcmp (target, "pei-i386") == 0
> 	     || strcmp (target, "pe-arm-wince-little") == 0
> 	     || strcmp (target, "pei-arm-wince-little") == 0);
>
>   if (!is_pe32 && !is_pe64)
>     {
>       /* This is not a recognized PE format file.  Abort now, because
> 	 the code is untested on anything else.  *FIXME* test on
> 	 further architectures and loosen or remove this test.  */
>       return DEFAULT_COFF_PE_TEXT_SECTION_OFFSET;
>     }

Just drop all these is_pe64/is_pe32 checks.  They're not really necessary.

>   /* Get pe_header, optional header and numbers of sections.  */
>   pe_header_offset = pe_get32 (abfd, 0x3c);
>   opthdr_ofs = pe_header_offset + 4 + 20;
>   nsections = pe_get16 (abfd, pe_header_offset + 4 + 2);
>   secptr = (pe_header_offset + 4 + 20 +
> 	    pe_get16 (abfd, pe_header_offset + 4 + 16));
>
>   /* Get the rva and size of the export section.  */
>   for (i = 0; i < nsections; i++)
>     {
>       char sname[8];
>       unsigned long secptr1 = secptr + 40 * i;
>       unsigned long vaddr = pe_get32 (abfd, secptr1 + 12);
>
>       bfd_seek (abfd, (file_ptr) secptr1, SEEK_SET);
>       bfd_bread (sname, (bfd_size_type) 8, abfd);
>       if (strcmp (sname, ".text") == 0)
> 	return vaddr;
>     }

All the bfd_seek/bfd_bread calls can be trivially replaced
by the equivalent Win32 file handling functions -- note how
win32_add_one_solib already has a handy handle open.
We can just make handle_load_dll not add the 0x1000 itself,
but leave win32_add_one_solib to do it instead.  Alternatively, to
make the code portable and usable from GDB, use fopen/fread, etc.
instead.

>   return DEFAULT_COFF_PE_TEXT_SECTION_OFFSET;
> }

-- 
Pedro Alves