Re: [Nasm-users] Talk, don't spam

[Frank K. <fbk...@ve...>]

H. Peter Anvin wrote:
> Frank Kotler wrote:
> 
>>That's the option I was looking for. I'll look again. What I've done 
>>seems to be asking me for approval after the fact...
> 
> That's what "moderation" means...

After they post it they ask me if it's okay??? That's *not* what's 
happening... I just saw the "reply to poster" from you well before I got 
the "message waits your approval" on the "reply to list", and Thought* 
that's what they were doing.

Anyway... bad idea, I'll undo it and look more for that "subscribers 
only" option... I think I just screwed up! I got a "waits your approval" 
on two messages - one from you, and one selling watches. I intended to 
approve one and reject the other... but I haven't seen the one from you 
show up. Maybe they're just slow... But I think I blew it somehow. :(

>>On a completely different topic, but still "talk", I'm wondering if you, 
>>or someone, can help me. There's an issue come up on alt.lang.asm (yeah, 
>>I know...) under the subject "Why is my nasm program killing itself". A 
>>guy has got Slackware 12.0 - kernel 2.6.21.5, apparently, and is having 
>>apparently good programs being "killed". Looks like they're killing the 
>>loader, actually... His disassembly suggests it's not a Nasm or ld 
>>problem...
>>
>>We saw something in kernels 2.6.10 - 2.6.16 (approximately) where 
>>programs without a writeable section last were causing a SIGSEGV in the 
>>loader (fs/binfmt_elf.c, IIRC) - original patch threw SIGKILL. I suspect 
>>something similar here.
> 
> Well, I can't even log into Usenet at this time, it seems.

Okay. I probably shouldn't have pestered you with it anyway. I thought 
you might know a "workaround" I could tell the guy. So far, it looks 
like just not having a .bss at all is working for him.

Looking at the loader in 2.6.23.9 (the latest one he encountered the 
problem on), I see several places we could get killed. Since it seems 
.bss might have something to do with it, I'm suspicious of this one...

...
if (unlikely (elf_brk > elf_bss)) {
			unsigned long nbyte;
	
			/* There was a PT_LOAD segment with p_memsz > p_filesz
			   before this one. Map anonymous pages, if needed,
			   and clear the area.  */
			retval = set_brk (elf_bss + load_bias,
					  elf_brk + load_bias);
			if (retval) {
				send_sig(SIGKILL, current, 0);
				goto out_free_dentry;
			}
...

... but I'm not "oriented" enough to know if we even hit that code, much 
less what to do to avoid it. Quite possibly a red herring... I'll study 
it some more (ought to anyway), but I don't have a lot of hope...

Meantime... if anyone is seeing mysterious kills... try removing .bss 
(if any), and see if that helps... and let me know, I guess, so I can 
advise the guy. Or, if anyone knows on which kernel this *stopped* 
happening... (I understand it's been "fixed", but can't confirm it), I'd 
pass that info along, too...
		
Sorry I deleted your message... as appears to have happened... could ya 
repost?

Best,
Frank