looking for non-loaded, non-writable, non-allocated elf segment

looking for non-loaded, non-writable, non-allocated elf segment

[Andrew Brown]

what i'm trying to do is embed some configuration data that describes
my program in the actual binary itself so that i have less "things" i
need to keep track of.  what seemed best to me was to stuff the data
into a non-loadable, non-allocated, and also read-only (although i
supposed that's moot if it's not loaded or allocated) elf segment via,
eg:

static const char config[]
#ifdef __ELF__
	__attribute__((section(".netbsd.config")))
#endif /* __ELF__ */
	__attribute__((__unused__)) = "...";

running objdump on my binary shows at this point shows me this:

% objdump -h netbsd 
...
Sections:
Idx Name          Size      VMA       LMA       File off  Algn
...
  6 .netbsd.config 0000b140  00000000  00000000  002bca40  2**5
                  CONTENTS, ALLOC, LOAD, READONLY, DATA

which isn't exactly what i was looking for.  after a little random
poking around, i happened upon this syntax:

	__attribute__((section(".netbsd.config#")))

which changes the attributes of the given section to "CONTENTS,
READONLY" which is exactly what i was looking for, however, after
skimming through gas to see if that's the "correct" behavior, i have
reached the conclusion that it might perhaps be a happy bug.  that
__attribute__() turns into this "assembly":

	#NO_APP
	.section	.netbsd.config#,"a",@progbits
		.align 32
		.type	 config,@object
		.size	 config,45068
	config:
		.ascii "..."

so it seems that the ,"a",@progbits piece of the .section is just
being ignored.  it seems, however, that since i'm toying with this
idea on an i386 platform, this "just works" because the # is the
comment character, which isn't exactly portable, nor is it an actual
feature.

what i'm looking for is a way to specify that a given section is not
to be loaded or allocated, which is also what i get if i remove the
"a" from the ,"a", sequence on the .section line.  is there a way to
specify this with some other __attribute__ flags or carefully crafted
section names?  in a way that's likely to "remain functional"?

-- 
|-----< "CODE WARRIOR" >-----|
codewarrior@daemon.org             * "ah!  i see you have the internet
twofsonet@graffiti.com (Andrew Brown)                that goes *ping*!"
andrew@crossbar.com       * "information is power -- share the wealth."

Re: looking for non-loaded, non-writable, non-allocated elf segment

[Philippe GIACINTI]

on January 3 2002 Andrew Brown <atatat@atatdot.net> wrote:
 > what i'm trying to do is embed some configuration data that describes
 > my program in the actual binary itself so that i have less "things" i
 > need to keep track of.  what seemed best to me was to stuff the data
 > into a non-loadable, non-allocated, and also read-only (although i
 > supposed that's moot if it's not loaded or allocated) elf segment via,
 > eg:
 > 
 > 
 > ...
 >
 >
 > so it seems that the ,"a",@progbits piece of the .section is just
 > being ignored.  it seems, however, that since i'm toying with this
 > idea on an i386 platform, this "just works" because the # is the
 > comment character, which isn't exactly portable, nor is it an actual
 > feature.
 > 
 > what i'm looking for is a way to specify that a given section is not
 > to be loaded or allocated, which is also what i get if i remove the
 > "a" from the ,"a", sequence on the .section line.  is there a way to
 > specify this with some other __attribute__ flags or carefully crafted
 > section names?  in a way that's likely to "remain functional"?
 > 

Hello,

I already faced this kind of problem, and, after looking at the code, it
seems there is no elegant way to achieve your goal (the `a,@progbits'
is actually hardcoded somewhere ...).
To solve my problem, I used a dirty workaround: I forced the definition
of the section with inline assembly before letting the compiler define
it...

For example, I have this kind of code (I want a section without content):

__asm__(".section my_section,\"aw\",@nobits\n");
static int my_variable __attribute__ ((section ("my_section")));

The assembler issues a warning because of the redefinition of the
section, but ignores the second definition (i.e: generated by the
compiler ...).

Just my 0,02 Euro...
Best regards.
-- 
Philippe Giacinti - Alcatel Business Systems
mailto:Philippe.Giacinti@sxb.bsf.alcatel.fr