Q: -fpic and $_GLOBAL_OFFSET_TABLE_

Q: -fpic and $_GLOBAL_OFFSET_TABLE_

[Oleg Nesterov]

Hello.

I thought, that &GOT == $_GLOBAL_OFFSET_TABLE_ + .

$ cat test.c

void test(void)
{
	asm volatile (
		"addl $_GLOBAL_OFFSET_TABLE_, %eax\n"
		"addl $_GLOBAL_OFFSET_TABLE_, %ecx\n"
	);
}

$ cc -c -fpic -fomit-frame-pointer test.c && ld -shared -o test.so test.o && objdump -d test.so

test.so:     file format elf32-i386

Disassembly of section .text:

0000018c <test>:
 18c:	05 0d 10 00 00       	addl   $0x100d,%eax	==> &GOT = 0x18c + 0x100d = 0x1199
 191:	81 c1 07 10 00 00    	addl   $0x1007,%ecx	==> &GOT = 0x191 + 0x1007 = 0x1198
 197:	c3                   	ret    

How can it be?

Oleg.

Re: Q: -fpic and $_GLOBAL_OFFSET_TABLE_

[hp]

Oleg Nesterov am Dienstag, 15. Juli 2003 14:19:
> Hello.
>
> I thought, that &GOT == $_GLOBAL_OFFSET_TABLE_ + .
						^^^^^^^^^^^^^^^^^^^^^
thus your GOT would be at a different place whenever fetched at a different 
address in your code - not very likely...

and, more in the ELF standard description, re links at
		http://www.lxhp.in-berlin.de/lhplinks.html#elf

>
> $ cat test.c
>
> void test(void)
> {
> 	asm volatile (
> 		"addl $_GLOBAL_OFFSET_TABLE_, %eax\n"
> 		"addl $_GLOBAL_OFFSET_TABLE_, %ecx\n"
> 	);
> }
>
> $ cc -c -fpic -fomit-frame-pointer test.c && ld -shared -o test.so test.o
> && objdump -d test.so
>
> test.so:     file format elf32-i386
>
> Disassembly of section .text:
>
> 0000018c <test>:
>  18c:	05 0d 10 00 00       	addl   $0x100d,%eax	==> &GOT = 0x18c + 0x100d =
> 0x1199 191:	81 c1 07 10 00 00    	addl   $0x1007,%ecx	==> &GOT = 0x191 +
> 0x1007 = 0x1198 197:	c3                   	ret
>
> How can it be?
>
> Oleg.
> -
> To unsubscribe from this list: send the line "unsubscribe linux-assembly"
> in the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 
Linux,Assembly,Forth: http://www.lxhp.in-berlin.de/index-lx.shtml en/de
	FAQ(s) + DOCs at http://linuxassembly.org
  pse, reply to << lx -at- lxhp -dot- in-berlin -dot- de >>

Keyboard and Mouse library

[jeff]

Hello,
I'm trying to create a linux library and running into
trouble with the mouse.  The intent was to use kernel
calls only and run from the user level.  Is this
possible?

I've looked at Gpm which works with terminals
but find the "c" code difficult.  Assembler for me
is much easier to understand.

Also, it would be nice to get the raw keyboard scan
codes.  I've found programs that claim to do this
but they don't work on my 9.1 Mandrake version
of Linux?

Any suggestions or ideas?

Jeff Owens

Re: Keyboard and Mouse library

[Luciano Miguel Ferreira Rocha]

Yes. Don't use Linux.

Linux is an OS kernel. It's very purpose is to remove the requirement for
people to directly program the keyboard controler and/or serial/psaux port and
the attached mouse. All devices are presented as files.

The hardware abstraction in Linux allows programs to be used remotely and
still assume a keyboard and a mouse.

Anyway, you can access serial mice as a normal serial device via /dev/ttySx,
and keycodes by setting the terminal to raw.

PS/2 mice go to /dev/psaux and USB to /dev/input/mouse*

Also, gpm can be used to convert from one protocol to another (bus mouse ->
serial, etc.).

Regarding keyboard, you can read characters by reading from stdin (file
descriptor 0), and control it with ioctl calls.

Relevant man pages: console_codes console_ioctl mouse ttyS ...

I recomend you get a C programming book and another for Unix.

You'll see C has its benefits over assembly.

Regards,
Luciano Rocha

On Wed, Jul 16, 2003 at 03:03:15AM -0700, jeff wrote:
> Hello,
> I'm trying to create a linux library and running into
> trouble with the mouse.  The intent was to use kernel
> calls only and run from the user level.  Is this
> possible?
> 
> I've looked at Gpm which works with terminals
> but find the "c" code difficult.  Assembler for me
> is much easier to understand.
> 
> Also, it would be nice to get the raw keyboard scan
> codes.  I've found programs that claim to do this
> but they don't work on my 9.1 Mandrake version
> of Linux?
> 
> Any suggestions or ideas?
> 
> Jeff Owens
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-assembly" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Keyboard and Mouse library

[hp]

jeff am Mittwoch, 16. Juli 2003 11:03:
> Hello,
> I'm trying to create a linux library and running into
> trouble with the mouse.  The intent was to use kernel
> calls only and run from the user level.  Is this
> possible?

very difficult. simple ioctls won't work.

> I've looked at Gpm which works with terminals
> but find the "c" code difficult.  Assembler for me
> is much easier to understand.

yes, but (most of) those C-'programmers' don't understand (and ignore) 
assembly, so you might need to learn their 'language'...

> Also, it would be nice to get the raw keyboard scan
> codes.  I've found programs that claim to do this
> but they don't work on my 9.1 Mandrake version
> of Linux?

re library of 'biew' program, file 'keyboard.c', a link to the source can be 
found on http://linuxassembly.org

mandrake might be difficult, I'd find even some signals actions different (and 
falsely modified!) , in an old version.

best,
	hp
-- 
Linux,Assembly,Forth: http://www.lxhp.in-berlin.de/index-lx.shtml en/de
  pse, reply to << lx -at- lxhp -dot- in-berlin -dot- de >>

Re: Keyboard and Mouse library

[hp]

hp am Mittwoch, 16. Juli 2003 00:15:
> > but find the "c" code difficult.  Assembler for me
> > is much easier to understand.
>
> yes, but (most of) those C-'programmers' don't understand (and ignore)
...and are ignorant of...
> assembly, so you might need to learn their 'language'...

even in this group...

 	hp

-- 
Linux,Assembly,Forth: http://www.lxhp.in-berlin.de/index-lx.shtml en/de
  pse, reply to << lx -at- lxhp -dot- in-berlin -dot- de >>

Re: Keyboard and Mouse library

[jeff]

On Tuesday 15 July 2003 04:15 pm, hp wrote:
> yes, but (most of) those C-'programmers' don't understand (and ignore)
> assembly, so you might need to learn their 'language'...

True, most of Linux seems to be documented from a "C" programmers
perspective.  One has to learn a long list of symbolic names for
things and then look in the .h files to see what codes they create.
Then you run into a lot of ifdef statements and have trouble
deciding if this symbol code is for BSD or AIX or ?   Eventually,
it is necessary to compile the code and use GDB to see
what code is generated.

Sigh... I sometimes wish we had a simple assembler interface
to the kernel.  If a call can only be accessed as root then
that would be handled etc.  Everything clearly documented
for assembler code.  No macros.   No includes.  Just
registers, dwords, words, and bytes.  If common operations
require several kernel functions they could be added.

Enough dreaming... back to work <grin>  Jeff Owens

Re: Keyboard and Mouse library

[linuxassembly]

On Wed, 16 Jul 2003 03:03:15 -0700, jeff <jko@save-net.com> wrote:

> Also, it would be nice to get the raw keyboard scan
> codes.  I've found programs that claim to do this
> but they don't work on my 9.1 Mandrake version
> of Linux?

sys_ioctl STDIN, KDSKBMODE, K_RAW

Afterwards you'll be in raw mode.  Unfortunatly, whatever is in the kernel 
that the stty command controls is still mucking things up, and continutes 
to translate byte 13 into a byte 10, and byte 127 into a byte 8, and keeps 
all input from you until a byte 13 comes along which may be causing you to 
think it isn't working since you can't read anything from stdin.

There are some mysterious ioctls that control this as well, but they don't 
seem to be documented anywhere at all.  However, you can simply call the 
stty program to do this for you, which sounds bad at first, but if you do 
it this way you're guaranteed it'll work right, and if the secret ioctls 
are changed, you don't have to care as long as the stty program is updated 
with them.  This is the way it is done with my program Softer, and you 
can't even tell it's doing it.

First run the command:  stty -g

It'll spit out a long string of stuff.  This stuff is the current stty 
settings.  Save them somewhere.

Then do: stty raw -echo

Now the kernel will stop it's character translation, allowing you to read 
data as soon as it's available rather than waiting on a byte 13, and it 
won't echo everything to the screen either, which makes no sense in raw 
mode.  In fact, the whole deal makes no sense with the keyboard in raw 
mode, so I don't see why it isn't all disabled automatically.  Anyway, this 
is what you have to do, unfortunatly.

Upon exit, do this:  stty [that_string_of_crap_from_earlier]

Then the settings will be back to what they used to be.  Just call the 
keyboard ioctl to switch it back out of raw mode, and you're done.

Now when in raw mode, there's the little matter of not being able to hit 
Atl-Fn to switch consoles anymore.  Your program should fix this by 
watching for such key combinations and switching the console for the user.  
You don't need to switch out of raw mode or re-run stty when this is done, 
the kernel keeps keyboard states seperate for each console.  Also, Control- 
C won't end your program anymore, so make sure there's always some way to 
end your program.

You might want to have a look at my program Softer which is at

http://www.evobsyniva.com/softer/

It does raw keyboard and video access.  It runs other programs under it, 
and gives them scancodes, in addition to figuring out what the user is 
typing (so that you can ignore the scancodes if you don't want them).  
Also, it's scancodes aren't the keyboard scan codes, but some of it's very 
own, so that each key has it's own code and there are no prefix codes to 
worry about.  I don't know your reasons for writing this library, but if 
it's just to get around linux's lousy terminal interface, you might be 
happy just using Softer.  That's why I wrote it.  It doesn't do mouse, 
however, which isn't to say that if you figure out a way to get mouse 
input, and tell me about it, that I won't add it in.  I just don't care 
enough about mouse support to figure it out on my own.

At the moment it requires users to execute softer to run softer programs, 
but sooner or later I'm going to make it so that programs can detect if 
softer is already running, and if not, start it thenselves, making it 
possible for users to run softer programs without having to start softer 
first.

Re: Keyboard and Mouse library

[hp]

linuxassembly@evobsyniva.com am Mittwoch, 16. Juli 2003 04:09:
> On Wed, 16 Jul 2003 03:03:15 -0700, jeff <jko@save-net.com> wrote:
>
> There are some mysterious ioctls that control this as well, but they don't
> seem to be documented anywhere at all.  However, you can simply call the

man 4 console_ioctl (or console_ioctls, old vers.)

hp
-- 
Linux,Assembly,Forth: http://www.lxhp.in-berlin.de/index-lx.shtml en/de
	FAQ(s) + DOCs at http://linuxassembly.org
  pse, reply to << lx -at- lxhp -dot- in-berlin -dot- de >>

Re: Keyboard and Mouse library

[linuxassembly]

On Wed, 16 Jul 2003 07:46:30 +0100, hp <lx@lxhp.in-berlin.de> wrote:

> linuxassembly@evobsyniva.com am Mittwoch, 16. Juli 2003 04:09:
>> On Wed, 16 Jul 2003 03:03:15 -0700, jeff <jko@save-net.com> wrote:
>>
>> There are some mysterious ioctls that control this as well, but they 
>> don't
>> seem to be documented anywhere at all.  However, you can simply call the
>
> man 4 console_ioctl (or console_ioctls, old vers.)

The only thing I see that might apply is KDGKBENT and KDSKBSENT.  They 
might be related, but I can't tell, and suspect they are not.  There's no 
mention of anything that might remove the line buffering, or the byte 3 
leading to a sigkill thing.  As I mentioned, setting the keyboard in raw 
mode doesn't remove these things, it just mucks them up.  Instead of 
control-C sending a sigkill, it becomes the number 2 key, who's scan code 
is 3.

Doing an strace on stty reveals that it calls two mystery ioctls 0x5401 and 
0x5403, and looking around the kernel source I figured out they are TCGETS 
and TCSETSW, however I've been unsuccessful in tracking down any more 
information than that.  The two ioctls lead to drivers/char/tty_ioctl.c, 
which calls a function set_termios, which calls change_termios, which calls 
???.set_termios, and that's about where my ability to read C code comes to 
an end.

Re: Keyboard and Mouse library

[Maciej Hrebien]

linuxassembly@evobsyniva.com wrote:
> 
> On Wed, 16 Jul 2003 07:46:30 +0100, hp <lx@lxhp.in-berlin.de> wrote:
> 
> > linuxassembly@evobsyniva.com am Mittwoch, 16. Juli 2003 04:09:
> >> On Wed, 16 Jul 2003 03:03:15 -0700, jeff <jko@save-net.com> wrote:
> >>
> >> There are some mysterious ioctls that control this as well, but they
> >> don't
> >> seem to be documented anywhere at all.  However, you can simply call the
> >
> > man 4 console_ioctl (or console_ioctls, old vers.)
> 
> The only thing I see that might apply is KDGKBENT and KDSKBSENT.  They
> might be related, but I can't tell, and suspect they are not.  There's no
> mention of anything that might remove the line buffering, or the byte 3
> leading to a sigkill thing.  As I mentioned, setting the keyboard in raw
> mode doesn't remove these things, it just mucks them up.  Instead of
> control-C sending a sigkill, it becomes the number 2 key, who's scan code
> is 3.
> 
> Doing an strace on stty reveals that it calls two mystery ioctls 0x5401 and
> 0x5403, and looking around the kernel source I figured out they are TCGETS
> and TCSETSW, however I've been unsuccessful in tracking down any more
> information than that.  The two ioctls lead to drivers/char/tty_ioctl.c,
> which calls a function set_termios, which calls change_termios, which calls
> ???.set_termios, and that's about where my ability to read C code comes to
> an end.

I don't know if i understand you right, but if want to switch off line
buffering on terminal see man 3 termios routines or do ioctl on stdin.
TCGETS gets the current termios structure, TCSETS sets. You modify
c_*flag(s) and here you are. I also think c_lflag&=~ICANON may be
usefull for you in this case. Is this what you wanted?

-- 
Maciej Hrebien

Re: Keyboard and Mouse library

[linuxassembly]

On Thu, 17 Jul 2003 18:45:37 +0200, Maciej Hrebien <m_hrebien@wp.pl> wrote:

>> Doing an strace on stty reveals that it calls two mystery ioctls 0x5401 
>> and
>> 0x5403, and looking around the kernel source I figured out they are 
>> TCGETS
>> and TCSETSW, however I've been unsuccessful in tracking down any more
>> information than that.  The two ioctls lead to drivers/char/tty_ioctl.c,
>> which calls a function set_termios, which calls change_termios, which 
>> calls
>> ???.set_termios, and that's about where my ability to read C code comes 
>> to
>> an end.
>
> I don't know if i understand you right, but if want to switch off line
> buffering on terminal see man 3 termios routines or do ioctl on stdin.
> TCGETS gets the current termios structure, TCSETS sets. You modify
> c_*flag(s) and here you are. I also think c_lflag&=~ICANON may be
> usefull for you in this case. Is this what you wanted?

That's it.  I knew it was an ioctl, and that it was TCSETS, but I couldn't 
find that sturcture anywhere.  I just didn't know it was called termios, 
and so I didn't know what to look up in man.  

Re: Keyboard and Mouse library

[Maciej Hrebien]

linuxassembly@evobsyniva.com wrote:
> 
> > I don't know if i understand you right, but if want to switch off line
> > buffering on terminal see man 3 termios routines or do ioctl on stdin.
> > TCGETS gets the current termios structure, TCSETS sets. You modify
> > c_*flag(s) and here you are. I also think c_lflag&=~ICANON may be
> > usefull for you in this case. Is this what you wanted?
> 
> That's it.  I knew it was an ioctl, and that it was TCSETS, but I couldn't
> find that sturcture anywhere.  I just didn't know it was called termios,
> and so I didn't know what to look up in man.

For sure: <asm/termbits.h>

Regards,

-- 
Maciej Hrebien

Re: Keyboard and Mouse library

[hp]

this is a bit frustrating.
<http://www.lxhp.in-berlin.de/lhpioctl.html#termios.h> takes you directly to 
the required reference,  which was also found by a simple text search in the 
browser for TCSETS. the link points you to the structure and all those items' 
definitions...

what else could I do? - i.e. I give up.

linuxassembly@evobsyniva.com am Donnerstag, 17. Juli 2003 20:26:
> On Thu, 17 Jul 2003 18:45:37 +0200, Maciej Hrebien <m_hrebien@wp.pl> wrote:
>
> That's it.  I knew it was an ioctl, and that it was TCSETS, but I couldn't
> find that sturcture anywhere.  I just didn't know it was called termios,
> and so I didn't know what to look up in man.
> -

-- 
Linux,Assembly,Forth: http://www.lxhp.in-berlin.de/index-lx.shtml en/de
	FAQ(s) + DOCs at http://linuxassembly.org
  pse, reply to << lx -at- lxhp -dot- in-berlin -dot- de >>

Re: Keyboard and Mouse library

[jeff]

On Tuesday 15 July 2003 08:09 pm, linuxassembly@evobsyniva.com wrote:

> sys_ioctl STDIN, KDSKBMODE, K_RAW
>
> Afterwards you'll be in raw mode.  Unfortunatly, whatever is in the kernel
> that the stty command controls is still mucking things up, and continutes
> to translate byte 13 into a byte 10, and byte 127 into a byte 8, and keeps
> all input from you until a byte 13 comes along which may be causing you to
> think it isn't working since you can't read anything from stdin.i

It is possible to turn the buffering of characters off in termios and this
works for normal reads.  My problem is an error code when trying
to set raw mode.  Here is the code

 mov  eax,54      ;ioctl
 mov ebx,0         ;stdin
 mov ecx,0x4b45  ;KDSKBMODE
 mov edx,0        ;raw mode
 int   0x80

returns code -22 which i assume is "invalid arguement"

In some sample code they opened stdin and used the
returned handle for stdin.  This did not work either.

 Cheers, jeff

Re: Keyboard and Mouse library

[Konstantin Boldyshev]

On Thu, 17 Jul 2003, jeff wrote:

> > sys_ioctl STDIN, KDSKBMODE, K_RAW
> >
> > Afterwards you'll be in raw mode.  Unfortunatly, whatever is in the
> kernel
> > that the stty command controls is still mucking things up, and
> continutes
> > to translate byte 13 into a byte 10, and byte 127 into a byte 8, and
> keeps
> > all input from you until a byte 13 comes along which may be causing
> you to
> > think it isn't working since you can't read anything from stdin.i
>
> It is possible to turn the buffering of characters off in termios and
> this
> works for normal reads.  My problem is an error code when trying
> to set raw mode.  Here is the code
>
>  mov  eax,54      ;ioctl
>  mov ebx,0         ;stdin
>  mov ecx,0x4b45  ;KDSKBMODE
>  mov edx,0        ;raw mode
>  int   0x80
>
> returns code -22 which i assume is "invalid arguement"
>
> In some sample code they opened stdin and used the
> returned handle for stdin.  This did not work either.

I think that this article by Karten answers most of "raw keyboard"
questions: http://linuxassembly.org/articles/rawkb.html.

-- 
Regards,
Konstantin

Re: Q: -fpic and $_GLOBAL_OFFSET_TABLE_

[Oleg Nesterov]

Hello.

hp <lx () lxhp ! in-berlin ! de> wrote:
>
> Oleg Nesterov am Dienstag, 15. Juli 2003 14:19:
> > Hello.
> >
> > I thought, that &GOT == $_GLOBAL_OFFSET_TABLE_ + .
>						^^^^^^^^^^^^^^^^^^^^^
> thus your GOT would be at a different place whenever fetched at a different 
> address in your code - not very likely...

There is R_386_GOTPC relocation item for _GLOBAL_OFFSET_TABLE_.

Look at this code produced by gcc to load &GOT in ebx:

	call .L3
.L3:
	popl %ebx
	addl $_GLOBAL_OFFSET_TABLE_+[.-.L3],%ebx

So &GOT = .L3 + $_GLOBAL_OFFSET_TABLE_+ . - .L3 = $_GLOBAL_OFFSET_TABLE_+ .

Thank you for response,

Oleg.

Q: -fpic and $_GLOBAL_OFFSET_TABLE_

[Oleg Nesterov]

Hello.

I thought, that &GOT == $_GLOBAL_OFFSET_TABLE_ + .

$ cat test.c

void test(void)
{
	asm volatile (
		"addl $_GLOBAL_OFFSET_TABLE_, %eax\n"
		"addl $_GLOBAL_OFFSET_TABLE_, %ecx\n"
	);
}

$ cc -c -fpic -fomit-frame-pointer test.c && ld -shared -o test.so test.o && objdump -d test.so

test.so:     file format elf32-i386

Disassembly of section .text:

0000018c <test>:
 18c:	05 0d 10 00 00       	addl   $0x100d,%eax	==> &GOT = 0x18c + 0x100d = 0x1199
 191:	81 c1 07 10 00 00    	addl   $0x1007,%ecx	==> &GOT = 0x191 + 0x1007 = 0x1198
 197:	c3                   	ret    

How can it be?

Oleg.