RFR: 8248876: LoadObject with bad base address created for exec file on linux

[Kevin Walls]

On Wed, 3 Feb 2021 22:17:16 GMT, Chris Plummer <cjplummer at openjdk.org> wrote:

>> We could see NPE on `LinuxCDebugger::loadObjectContainingPC` which is used mainly by the clhsdb `findpc` command. (See [JDK-8248876](https://bugs.openjdk.java.net/browse/JDK-8248876) for more details)
>> 
>> It is caused by PIE executable handling.
>> 
>> Currently we use virtual address of top of PT_LOAD in executable as base address - it is valid for ET_EXEC binary.
>> However, in case of ET_DYN binary (PIE binary), we should handle it like a shared library.
>> In GDB, base address of executable would be calculated by dynamic section. SA should also do so.
>
> Marked as reviewed by cjplummer (Reviewer).

Hi!
I found the "svr4_exec_displacement() in GDB" more confusing than informative, maybe this is the same technique GDB uses...  But I followed our dynamic_addr setting and agree with the new changes.

An alternative comment could be this, which would also removes the "substract" typo (which was not yours!):

ps_core.c 509

    } else { // ET_DYN
      // dynamic_addr was initially set to entry point (AT_ENTRY).
      // Now subtract the entry point offset to get load address:
      result = ph->core->dynamic_addr - exec_ehdr->e_entry;
      // Find actual dynamic_addr:
      ph->core->dynamic_addr += exec_php->p_vaddr - exec_ehdr->e_entry;
    }

-------------

PR: https://git.openjdk.java.net/jdk/pull/2366