where 's the location of the implementation of mlock metho..

On Aug 22, 10:40 am, Paul Pluzhnikov <ppluzhnikov-....TakeThisOut@charter.net>
wrote:
> sods <skyofdre....TakeThisOut@gmail.com> writes:
> > I'm using the mlock() to lock pages,
> > and search the glibc-2.3.6 source code for its implementation.
> > but i get nothing expect a general dummy definition and a hurd
> > implementation.
>
> > Where is the linux i386 version of mlock implementation ?
>
> From ChangeLog.5 in glibc-2.5 sources:
>
> Thu Jan 18 00:32:43 1996 Roland McGrath <rol....TakeThisOut@churchy.gnu.ai.mit.edu>
> ...
> Replaced all simple system call files *.S throughout sysdeps/unix
> with syscalls.list files to be processed by make-syscalls.sh.
> ...
> * sysdeps/unix/sysv/linux/mlock.S: File removed.
>
Thanks.

it's a good point.
and
this implementation is really a little magic.

> > is it in kernel?
>
> There is sys_mlock() in the kernel, but that has nothing whatsoever to do
> with glibc mlock syscall wrapper.

I do some findings.
all method in syscalls.list are "compiled" into object file in
glibc-build/misc/
--
glibc-build/misc/mlock.o
glibc-build/misc/mlockall.o
--

and finally link into libc.os.6, so
libc.so.6 have mlock in .text section.
--
00bc0d0 T mlock
000bc150 T mlockall
--

so , the mlock in libc.so.6 is a wrapper? should be,
but the mechanism of this wrapper is still unclear.

-sods

sods <skyofdreams.TakeThisOut@gmail.com> writes:

> so , the mlock in libc.so.6 is a wrapper?

Yes, it's a wrapper for the syscall:

0x000ba000 <mlock+0>: mov %ebx,%edx # %ebx must be preserved
0x000ba002 <mlock+2>: mov 0x8(%esp),%ecx # load second arg into %ecx
0x000ba006 <mlock+6>: mov 0x4(%esp),%ebx # load first arg into %ebx
0x000ba00a <mlock+10>: mov $0x96,%eax # 0x96 == 150 == __NR_mlock /usr/include/asm/unistd.h
0x000ba00f <mlock+15>: int $0x80 # syscall

> but the mechanism of this wrapper is still unclear.

It's unclear to me what's unclear to you.

Did you read make-syscalls.sh? Did you understand what it does?

Cheers,
--
In order to understand recursion you must first understand recursion.
Remove /-nsp/ for email.

On Aug 23, 10:30 am, Paul Pluzhnikov <ppluzhnikov-....TakeThisOut@charter.net>
wrote:
> sods <skyofdre....TakeThisOut@gmail.com> writes:
> > so , the mlock in libc.so.6 is a wrapper?
>
> Yes, it's a wrapper for the syscall:
>
> 0x000ba000 <mlock+0>: mov %ebx,%edx # %ebx must be preserved
> 0x000ba002 <mlock+2>: mov 0x8(%esp),%ecx # load second arg into %ecx
> 0x000ba006 <mlock+6>: mov 0x4(%esp),%ebx # load first arg into %ebx
> 0x000ba00a <mlock+10>: mov $0x96,%eax # 0x96 == 150 == __NR_mlock /usr/include/asm/unistd.h
> 0x000ba00f <mlock+15>: int $0x80 # syscall
>

yes, it's more clear from asm code.
just a simple wrapper.

> > but the mechanism of this wrapper is still unclear.
>
> It's unclear to me what's unclear to you.
>
> Did you read make-syscalls.sh? Did you understand what it does?
>

I have readed it now.

it's the make-syscalls.sh where the magic is in.

Makefile use it to create another script call sysd-syscallT, which
then do the actual operations.
sysd-syscallT seems generate the c source wrapper code, and pass them
to gcc through pipe. (.. pipe so no mlock.c file.)

but it all doesn't matter ,
I want to know what logic is processed when i call mlock.
If the glibc just a simple wrapper, all actual case should happen in
kernel.

now, i know it's in kernel, i 'll dip into kernel a little , e.g.
learn how trap works
to contnue.

thanks, for the help.
-sods

> Cheers,
> --
> In order to understand recursion you must first understand recursion.
> Remove /-nsp/ for email.

sods <skyofdreams RemoveThis @gmail.com> writes:

> I want to know what logic is processed when i call mlock.

If you want to understand the logic for *any* system call, you have
to look in the kernel -- that's where the code for them resides
(and that's what makes them 'system call', as opposed to libc calls).

> now, i know it's in kernel, i 'll dip into kernel a little , e.g.
> learn how trap works to contnue.

If you want to understand what mlock does, start with sys_mlock
in /usr/src/linux-X.Y/mm/mlock.c (it's somewhat long way from how
trap works, you can just assume that int80 with %eax==150 eventually
gets to sys_mlock in the kernel).

Cheers,
--
In order to understand recursion you must first understand recursion.
Remove /-nsp/ for email.

On Aug 23, 10:19 pm, Paul Pluzhnikov <ppluzhnikov-....TakeThisOut@charter.net>
wrote:
> sods <skyofdre....TakeThisOut@gmail.com> writes:
> > I want to know what logic is processed when i call mlock.
>
> If you want to understand the logic for *any* system call, you have
> to look in the kernel -- that's where the code for them resides
> (and that's what makes them 'system call', as opposed to libc calls).

Yeah, you're right,

Several day ago, i thought libc calls have some 'real' logic to
wrap the system call.
but seems this doesn't happen to mlock.

>
> > now, i know it's in kernel, i 'll dip into kernel a little , e.g.
> > learn how trap works to contnue.
>
> If you want to understand what mlock does, start with sys_mlock
> in /usr/src/linux-X.Y/mm/mlock.c (it's somewhat long way from how
> trap works, you can just assume that int80 with %eax==150 eventually
> gets to sys_mlock in the kernel).

Thanks, i'll try to read the mm/mlock.c@sys_mlock
i guess if i meet more problems, the help topic should appear on
comp.os.linux.system.

Thank for your help, again.
-sods


>
> Cheers,
> --
> In order to understand recursion you must first understand recursion.
> Remove /-nsp/ for email.