[Dwarf-Discuss] _starting address_ in eh_frame FDE encoding

Francesco Zappa Nardelli francesco.zappa.nardelli at gmail.com
Thu Mar 31 01:43:45 PDT 2016


Dear all

> I am still confused though, can we check this example together?

Sorry for the noise, I made a silly mistake and the addresses in the
hexdump, which are off by 0x10.  Problem solved, thank you for the help!

-francesco


> Here is an annotated hexdump for the first CIE and FDE of an eh_frame
> table:
>
> $ hexdump -C -s 0x580 -n 658 foo2
> 00000580  14 00 00 00 00 00 00 00  01 7a 52 00 01 78 10 01
>           length    | ID        |  v  aug    | af df rr la
>
> 00000590  1b 0c 07 08 90 01 07 10  14 00 00 00 1c 00 00 00
>           *| opcds  .     .     || length    | ptr to CIE|
> *FDE encoding (aug = R) : 1b = PC_REL, sdata4
>
> 000005a0  00 fe ff ff 2a 00 00 00  00 00 00 00 00 00 00 00
>           strt_addr*|                                   ||
>
> The CIE augmentation data state states:
>
>   DW_EH_PE_pcrel, DW_EH_PE_sdata4
>
> The untranslated starting address in the FDE (noted strt_addr) is
>
>   fffffe00 (that is -200 in hex, -512 in dec)
>
> which, according to readelf must be translated to
>
>   pc=00000000004003a0
>
> The offset since the beginning of the eh_frame of the strt_addr field is
> 32, and
>
>  $ readelf -S foo2
> There are 34 section headers, starting at offset 0xc20:
>
> Section Headers:
>   [Nr] Name              Type             Address           Offset
>        Size              EntSize          Flags  Link  Info  Align
> ...
>   [16] .eh_frame         PROGBITS         0000000000400590  00000590
>        0000000000000114  0000000000000000   A       0     0     8
>
> In hex,  400590 + 20 = 4005b0, is this the “PC address” that we should
> consider when we parse the PCREL start address?  If yes, 4005b0 - 200 =
> 4003b0, but we should have got 4003a0.  Can you point me to where I am
> wrong?
>
> Thank you again.
>
> Best
> -francesco
>
>
>
> On Thu, Mar 24, 2016 at 3:47 PM, Cary Coutant <ccoutant at gmail.com> wrote:
>
>> > I have a question about the FDE encoding in eh_frame (and not
>> dwarf_frame)
>> > tables.
>> >
>> > This document (I could not find a better description of eh_frame)
>> >
>> >   http://www.airs.com/blog/archives/460
>>
>> The official description of the .eh_frame section, with a description
>> of the augmentation string and what the various DW_EH_PE values mean,
>> can be found in the relevant psABI document for your architecture.
>>
>> > states that:
>> >
>> >   "An FDE starts with the length and ID described above, and then
>> continues
>> > as follows.   1 The _starting address_ to which this FDE applies. This
>> is
>> > encoded using the FDE encoding specified by the associated CIE.”
>> >
>> > In a sample elf file, the associated CIE augmentation data state
>> reports:
>> >
>> >    DW_EH_PE_pcrel, DW_EH_PE_sdata4
>> >
>> > and the untranslated starting address value in the FDE is fffffe88
>> > (according to readelf this should be translated to 4004d0).  I do not
>> > understand what “value is PC relative” means here, as we are actually
>> > defining the range of PC values.
>> >
>> > Can somebody explain me how to interpret this _starting address_ field?
>>
>> PC-relative simply means that the value is not an absolute address,
>> but rather an offset relative to itself (i.e., the address of the
>> field that contains the value. The value you read there was
>> 0xfffffe88, which, according to the augmentation string, is a signed
>> 4-byte value, or -0x178. If you add that to the address of the
>> starting address field itself, you'll get the actual starting address,
>> which readelf has computed as 0x4004d0. Working backwards, that
>> implies that your FDE started at 400640, with the starting address
>> field at 0x400648.
>>
>> -cary
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.dwarfstd.org/private.cgi/dwarf-discuss-dwarfstd.org/attachments/20160331/fa70197f/attachment.htm>


More information about the Dwarf-Discuss mailing list