Symbol file scope?

Symbol file scope?

[Andrei Warkentin]

Hi,

This is my first time posting on a GDB mailing list, so excuse me in
advance if this has been covered before...

I am using GDB to debug TianoCore UEFI firmware
(http://sourceforge.net/apps/mediawiki/tianocore/index.php?title=Welcome)
running in a virtual machine using an ICE-like gdb remote stub (gdbsx
for Xen, or just QEMU gdb support). UEFI consists of a number of
separate modules - UEFI core, various drivers and UEFI application
(not applications, just one, like the EFI shell or a bootloader). All
of these modules live in the
same indentity-mapped address space, and interact with each other via
"protocols", which are effectively just function dispatch tables
published with the UEFI core. There is no concept of "processes", a
backtrace weaves through many modulules (i.e. EFI shell called into
Core, which called into fs driver, etc).

The problem is globals. GDB expects globals to be unique, since it
thinks it is debugging a program. Global search is done is "first
fit". So it will start with the symtabs in the first object file that
was "add-symbol-file"d and return in. It would be nice to be able to
have globals be searched for first in the object file to which the
current scope is in. Additionally, it would be nice to be able to
implement "module scope" in the same fashion as "file scope", so you
can look at globals in any module like this -

(gdb) print &'DxeCore.dll'::gST
$15 = (EFI_SYSTEM_TABLE **) 0x17fca6f0
(gdb) print &'IsaBusDxe.dll'::gST
$16 = (EFI_SYSTEM_TABLE **) 0x17e21da4

I have a patch for 7.1 (I know...old) that I put together that basically -
1) Modifies lookup_symbol_global to use
lookup_global_symbol_from_objfile before lookup_symbol_aux_symatbs or
solib_global_lookup
2) Modifies lookup_symtab to match symbol object file as well. It
returns the first symtab, which is good enough for
lookup_symbol_global logic to work. (I basically wanted to avoid
modifying c-exp, p-exp, etc, parsers)
3) Fixes location completion to match object file names as well.
4) Adds a list-symbol-files command to list all objfiles and their
locations, or list only those that match objfile/source file names.
I.e -
gdb) list-symbol-files DiskIo.c Dispatcher.c EhciDxe.dll
DiskIo.c:
       /home/fjnh84/mine/edk2/Build/OvmfIa32/DEBUG_GCC44/IA32/MdeModulePkg/Universal/Disk/DiskIoDxe/DiskIoDxe/DEBUG/DiskIoDxe.dll
               .text, [0x17e4c240-0x17e4f440)
Dispatcher.c:
       /home/fjnh84/mine/edk2/Build/OvmfIa32/DEBUG_GCC44/IA32/MdeModulePkg/Core/Dxe/DxeMain/DEBUG/DxeCore.dll
               .text, [0x17fb1240-0x17fcb8c0)
EhciDxe.dll:
       /home/fjnh84/mine/edk2/Build/OvmfIa32/DEBUG_GCC44/IA32/MdeModulePkg/Bus/Pci/EhciDxe/EhciDxe/DEBUG/EhciDxe.dll
               .text, [0x17deb240-0x17df3920)

I have a patch here -
https://github.com/andreiw/andreiw-wip/blob/master/gdb/0001-GDB-Initial-prototype-of-symbol-file-scope-module-sc.patch

I realize I need to rebase it on latest gdb. I wanted to get the GDB
folks' feedback before I did anything else with this :-).

Thank you for your time,
A

Re: Symbol file scope?

[Tom Tromey]

>>>>> "Andrei" == Andrei Warkentin <andreiw@motorola.com> writes:

Andrei> It would be nice to be able to
Andrei> have globals be searched for first in the object file to which the
Andrei> current scope is in. Additionally, it would be nice to be able to
Andrei> implement "module scope" in the same fashion as "file scope", so you
Andrei> can look at globals in any module like this -

Andrei> (gdb) print &'DxeCore.dll'::gST
Andrei> $15 = (EFI_SYSTEM_TABLE **) 0x17fca6f0
Andrei> (gdb) print &'IsaBusDxe.dll'::gST
Andrei> $16 = (EFI_SYSTEM_TABLE **) 0x17e21da4

This seems like a good feature.

Andrei> 1) Modifies lookup_symbol_global to use
Andrei> lookup_global_symbol_from_objfile before lookup_symbol_aux_symatbs or
Andrei> solib_global_lookup
Andrei> 2) Modifies lookup_symtab to match symbol object file as well. It
Andrei> returns the first symtab, which is good enough for
Andrei> lookup_symbol_global logic to work. (I basically wanted to avoid
Andrei> modifying c-exp, p-exp, etc, parsers)

I'd have to see it and hear more rationale for the placement.
Offhand it seems reasonable.

Andrei> 3) Fixes location completion to match object file names as well.

Thanks.

Andrei> 4) Adds a list-symbol-files command to list all objfiles and their
Andrei> locations, or list only those that match objfile/source file names.
Andrei> I.e -
Andrei> gdb) list-symbol-files DiskIo.c Dispatcher.c EhciDxe.dll

This should be an "info" subcommand instead.

Andrei> I have a patch here -
Andrei> https://github.com/andreiw/andreiw-wip/blob/master/gdb/0001-GDB-Initial-prototype-of-symbol-file-scope-module-sc.patch

I glanced at it but didn't try to do any serious review.
I'd suggest: rebase, write a ChangeLog entry, and submit here.
You'll also have to update the docs and write some tests.

If you don't have copyright assignment paperwork in place, please
contact me off-list so we can get you started on that.  It is best to
start this early, because sometimes it takes a while.

Tom

Re: Symbol file scope?

[Andrei Warkentin]

Hi,

On Tue, May 31, 2011 at 12:08 PM, Tom Tromey <tromey@redhat.com> wrote:
>>>>>> "Andrei" == Andrei Warkentin <andreiw@motorola.com> writes:
>
> Andrei> It would be nice to be able to
> Andrei> have globals be searched for first in the object file to which the
> Andrei> current scope is in. Additionally, it would be nice to be able to
> Andrei> implement "module scope" in the same fashion as "file scope", so you
> Andrei> can look at globals in any module like this -
>
> Andrei> (gdb) print &'DxeCore.dll'::gST
> Andrei> $15 = (EFI_SYSTEM_TABLE **) 0x17fca6f0
> Andrei> (gdb) print &'IsaBusDxe.dll'::gST
> Andrei> $16 = (EFI_SYSTEM_TABLE **) 0x17e21da4
>
> This seems like a good feature.
>
> Andrei> 1) Modifies lookup_symbol_global to use
> Andrei> lookup_global_symbol_from_objfile before lookup_symbol_aux_symatbs or
> Andrei> solib_global_lookup
> Andrei> 2) Modifies lookup_symtab to match symbol object file as well. It
> Andrei> returns the first symtab, which is good enough for
> Andrei> lookup_symbol_global logic to work. (I basically wanted to avoid
> Andrei> modifying c-exp, p-exp, etc, parsers)
>
> I'd have to see it and hear more rationale for the placement.
> Offhand it seems reasonable.
>
> Andrei> 3) Fixes location completion to match object file names as well.
>
> Thanks.
>
> Andrei> 4) Adds a list-symbol-files command to list all objfiles and their
> Andrei> locations, or list only those that match objfile/source file names.
> Andrei> I.e -
> Andrei> gdb) list-symbol-files DiskIo.c Dispatcher.c EhciDxe.dll
>
> This should be an "info" subcommand instead.
>
> Andrei> I have a patch here -
> Andrei> https://github.com/andreiw/andreiw-wip/blob/master/gdb/0001-GDB-Initial-prototype-of-symbol-file-scope-module-sc.patch
>
> I glanced at it but didn't try to do any serious review.
> I'd suggest: rebase, write a ChangeLog entry, and submit here.
> You'll also have to update the docs and write some tests.
>
> If you don't have copyright assignment paperwork in place, please
> contact me off-list so we can get you started on that.  It is best to
> start this early, because sometimes it takes a while.
>

Thanks for the feedback. Will start cleaning up the patch into
something that can be submitted.

A