[Dwarf-Discuss] dwarf stack operator for byte swap.

Michael Eager eager at eagercon.com
Mon Oct 28 08:07:11 PDT 2019


Hi Chirag --

A byte-swap operator would be a reasonable addition to the DWARF 
interpreter.

In addition to writing a proposal as Paul suggests, one option I would 
recommend you consider is to create a user-defined operation to perform 
this byte-swap operation and give it a value in the range between 
DW_OP_lo_user and DW_OP_hi_user.  Of course, this user-defined would 
need to be implemented not only in the producer (i.e., compiler) but in 
any consumers (i.e., debugger).

When a new version of the DWARF Standard is released, this operation 
would be assigned a new value.

On 10/28/19 6:53 AM, Robinson, Paul via Dwarf-Discuss wrote:
> Hello Chirag,
> 
> Regarding a byte-swap operation, it seems that you have a reasonable 
> use-case on a bi-endian machine.  Feel free to request a new operator on 
> the “public comments” page at http://dwarfstd.org/Comment.php
> 
> Note that a byte-swap operator would swap all bytes in the top-of-stack 
> value, which on your 64-bit machine would of course be a 64-bit value.  
> As you want a 32-bit swapped value, you would still need to do a shift 
> afterward, but even so, “DW_OP_byte_swap DW_OP_const1u 32 DW_OP_shr” 
> would be considerably shorter than what you have to do now.
> 
> Of course a new operator would be introduced in a new DWARF revision, 
> which is likely to be years away.  In the meantime let me suggest a 
> shorter expression for doing the byte-swap operation.  The book 
> “Hacker’s Delight” shows a straightforward 32-bit byte swap with masks 
> no wider than 16 bits, as follows:
> 
>                 x = (x << 24) | ((x & 0xff00) << 8) | ((x >> 8) & 
> 0xff00) | (x >> 24);
> 
> Your 64-bit machine will of course use 64-bit values on the expression 
> stack, so to keep the result “32-bit clean” we want to do one additional 
> mask:
> 
>                 x = ((x & 0xff) << 24) | ((x & 0xff00) << 8) | ((x >> 8) 
> & 0xff00) | (x >> 24);
> 
> Translating this into a DWARF expression, I get the following:
> 
>                 DW_OP_dup, DW_OP_const1u 0xff, DW_OP_and, DW_OP_lit24, 
> DW_OP_shl, DW_OP_swap, DW_OP_dup, DW_OP_const2u 0xff00, DW_OP_and, 
> DW_OP_lit8, DW_OP_shl, DW_OP_swap, DW_OP_dup, DW_OP_lit8, DW_OP_shr, 
> DW_OP_const2u 0xff00, DW_OP_and, DW_OP_swap, DW_OP_lit24, DW_OP_shr, 
> DW_OP_or, DW_OP_or, DW_OP_or
> 
> I hope this is helpful to you.
> 
> --paulr
> 
> *From:* Dwarf-Discuss <dwarf-discuss-bounces at lists.dwarfstd.org> *On 
> Behalf Of *Chirag Patel via Dwarf-Discuss
> *Sent:* Monday, October 28, 2019 12:47 AM
> *To:* dwarf-discuss at lists.dwarfstd.org
> *Subject:* [Dwarf-Discuss] dwarf stack operator for byte swap.
> 
> Hello Dwarf experts.
> 
> I am currently working trying to encode dwarf of binaries with having 
> bi-endian variables marked with DW_AT_endianity attribute.
> 
> The location calculation for some  variable depends on other variable 
> with different endianity, also the value of this other variable is known 
> at runtime.
> 
> At the moment I am using location list to calculate the correct location 
> of first variable and list of dwarf operators to reverse the endianity 
> of variable “__gbloffset__” in below case (I only needed lower 32 bits 
> on 64 bit machine).
> 
> 0x000001e5: DW_TAG_base_type
> 
>   DW_AT_byte_size  (0x04)
> 
>    DW_AT_encoding  (DW_ATE_signed)
> 
>                  DW_AT_name      ("int")
> 
>                  DW_AT_endianity (DW_END_big)
> 
>> 
> 0x00000057:   DW_TAG_variable
> 
>                  DW_AT_name      ("__gbloffset__")
> 
>                  DW_AT_type      (0x000001e5 "int")
> 
>                  DW_AT_external  (true)
> 
>                  DW_AT_decl_file ("…")
> 
>                  DW_AT_decl_line (8)
> 
>                  DW_AT_location  (DW_OP_addr 0) // pre linkage
> 
>                  DW_AT_linkage_name      ("_gblsection__")
> 
> 0x00000170:   DW_TAG_variable
> 
>                  DW_AT_name      ("VAR1")
> 
>                  DW_AT_type      (0x0000010b "fixed.dec.display.72")
> 
>                  DW_AT_decl_file ("…")
> 
>                  DW_AT_decl_line (10)
> 
>                  DW_AT_location  (DW_OP_addr 0x0, DW_OP_call4 0x57, 
> DW_OP_deref_size, 4,
> 
> DW_OP_dup, DW_OP_constu 0xff, DW_OP_lit0, DW_OP_shl, DW_OP_and, 
> DW_OP_lit24, DW_OP_shl, DW_OP_swap, DW_OP_dup, DW_OP_constu 0xff, 
> DW_OP_lit8, DW_OP_shl, DW_OP_and, DW_OP_lit8, DW_OP_shl, DW_OP_swap, 
> DW_OP_dup, DW_OP_constu 0xff, DW_OP_lit16, DW_OP_shl, DW_OP_and, 
> DW_OP_lit8, DW_OP_shr, DW_OP_swap, DW_OP_constu 0xff, DW_OP_lit24, 
> DW_OP_shl, DW_OP_and, DW_OP_lit24, DW_OP_shr, DW_OP_swap, DW_OP_or, 
> DW_OP_or, DW_OP_or, DW_OP_plus)
> 
>                  DW_AT_linkage_name      ("VAR1")
> 
> In above snippet of dwarf dump, I am using yellow highlighted list of 
> operators to swap the bytes.
> 
> I think there should be a support for DW_OP_byte_swap simple operator to 
> accomplice this simple task. Does this idea looks like it can be useful? 
> Is there any specific reason why dwarf spec does not have it or I am 
> missing something subtle.
> 
> I hope I conveyed the idea properly, apologies in advanced as English is 
> not my first language.
> 
> **
> 
> *Thanks and regards,*
> 
> **
> 
> *Chirag Patel*
> 
> Software Engineer | Raincode Labs India
> 
> *Tel*: (+91) 080 41159811
> 
> *Mob*: (+91) 9049336744
> 
> www.raincodelabs.com <http://www.raincodelabs.com/>__
> 
> linkedin-button <https://in.linkedin.com/in/chirag-patel->
> 
> 
> _______________________________________________
> Dwarf-Discuss mailing list
> Dwarf-Discuss at lists.dwarfstd.org
> http://lists.dwarfstd.org/listinfo.cgi/dwarf-discuss-dwarfstd.org
> 

-- 
Michael Eager    eager at eagercon.com
1960 Park Blvd., Palo Alto, CA 94306


More information about the Dwarf-Discuss mailing list