[9fans] malloc/ (long)((void*)v)

[9fans] malloc/ (long)((void*)v)

[Noah Evans]

One last question before I go to bed.

I've got the Data and BSS sections set up reasonably it seems, but  
malloc blows the kernel to shreds. I think I've traced the problem  
down to PADDR which tries to get a variable's physical address by  
coercing it to a void pointer and then back to a long. xalloc seems  
to be looking for the physical address of the memory segment it's  
being sent but it getting null instead. This is actually quite  
amusing because xalloc ends up tromping through graphics memory. Also  
this only happens to variables that are in the data or bss sections.  
PADDR returns the address of 'end' just fine.

What should I do to fix this?

Noah

Re: [9fans] malloc/ (long)((void*)v)

[Charles Forsyth]

>>What should I do to fix this?

there are a few things that need to be set up consistently.
i'll switch to private e-mail for that.

Re: [9fans] malloc/ (long)((void*)v)

[Ronald G Minnich]

Noah Evans wrote:
> One last question before I go to bed.
> 
> I've got the Data and BSS sections set up reasonably it seems, but  
> malloc blows the kernel to shreds.

it might be what you said, but when I've had this before it's because I 
screwed up the location of the data segment, in which case initialized 
data was wrong.

So, for test, I did this:

int something = 0x55aaaa55;

main(){
.
.
.
	if (something != 0x55aaaa55)
		die()




ron

Re: [9fans] malloc/ (long)((void*)v)

[Charles Forsyth]

[-- Attachment #1: Type: text/plain, Size: 227 bytes --]

yes, and don't forget that
	print("hello world\n");
might not be a good equivalent test if strings are in the text segment.
(it's amazing how far the code can get into and sometimes out of things like print
with data wrong)

[-- Attachment #2: Type: message/rfc822, Size: 3213 bytes --]

From: Ronald G Minnich <rminnich@lanl.gov>
To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
Subject: Re: [9fans] malloc/ (long)((void*)v)
Date: Tue, 31 Jan 2006 12:33:12 -0700
Message-ID: <43DFBB78.4030006@lanl.gov>

Noah Evans wrote:
> One last question before I go to bed.
> 
> I've got the Data and BSS sections set up reasonably it seems, but  
> malloc blows the kernel to shreds.

it might be what you said, but when I've had this before it's because I 
screwed up the location of the data segment, in which case initialized 
data was wrong.

So, for test, I did this:

int something = 0x55aaaa55;

main(){
.
.
.
	if (something != 0x55aaaa55)
		die()




ron

Re: [9fans] malloc/ (long)((void*)v)

[Noah Evans]

Bingo:

ulong test = 0xAFAFAFAF;

void
main(void)
{

	ulong i = 0, j = 0;
	memcpy(etext,bdata,edata-bdata);
	memset(edata, 0, end-edata);		/* clear the BSS */
	
	while(1) {
		i++; /* change a lot so I can see which reg its in */

Watching the registers(no screen yet) the value is getting hosed.

Noah

On Feb 1, 2006, at 4:33 AM, Ronald G Minnich wrote:

> Noah Evans wrote:
>> One last question before I go to bed.
>> I've got the Data and BSS sections set up reasonably it seems,  
>> but  malloc blows the kernel to shreds.
>
> it might be what you said, but when I've had this before it's  
> because I screwed up the location of the data segment, in which  
> case initialized data was wrong.
>
> So, for test, I did this:
>
> int something = 0x55aaaa55;
>
> main(){
> .
> .
> .
> 	if (something != 0x55aaaa55)
> 		die()
>
>
>
>
> ron