Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 27 Nov 1997 16:07:00 -0800
From:      John Polstra <jdp@polstra.com>
To:        Charles Mott <cmott@srv.net>
Cc:        hackers@FreeBSD.ORG
Subject:   Re: Shared Libraries and debugging 
Message-ID:  <199711280007.QAA18537@austin.polstra.com>
In-Reply-To: Your message of "Thu, 27 Nov 1997 16:40:35 MST." <Pine.BSF.3.96.971127163806.28870A-100000@darkstar.home> 

next in thread | previous in thread | raw e-mail | index | archive | help
> So memory allocated with mmap() is not dumped in a core file?

Right.  You get the "u" area and the data segment and the stack
segment.  The data segment is the contiguous region from "etext" up to
the sbrk(2) level.  I.e., its data + bss + the heap.

> Is this not possible

Practically anything would be possible. :-)

> or just not desirable?

Sometimes it might be desirable, and sometimes not.  It would require
a new core file format that permitted multiple noncontiguous segments.
(ELF core files could handle this -- they have the same format as ELF
object files.)  And of course it would complicate the kernel code that
generates core files.  And you could end up with some mighty large
core files at times.

Most mmapped areas are just shared libraries, which still exist on
disk.  True, they have some data areas that might have been changed
by the program, but the data areas in shared libraries are usually
relatively small.

If you dumped the mmapped areas, you'd probably want to limit it to
only the writable regions, on the assumption that the read-only
regions could be gotten from the underlying files on disk.

John
--
   John Polstra                                       jdp@polstra.com
   John D. Polstra & Co., Inc.                Seattle, Washington USA
   "Self-knowledge is always bad news."                 -- John Barth



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199711280007.QAA18537>