lkml.org 
[lkml]   [2004]   [Jan]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: uid- task_struct --The results of the code sent by you
Hi Robin,

What you say is very reasonable.

I will go bit further into this and come back to you.

What I am going to accomplish is:

At this stage Load Average is collected as a whole.
What I am trying to do is to collect it seperately under respective USERS.

This way load signal(Load average) can be connected to a perticular
user. Then to user behaviour as well.

Thanks a lot for your time
Sena Seneviratne
Computer Engineering Lab
Sydney University



Robin Holt wrote:

>Telnet and login are always going to be uid 0. The telnet
>process is handling socket IO and handling telnet escape
>sequences. When the telnet process is started, the user
>isn't even known to the system.
>
>telnet forks and execs login. Login does the password
>prompting and establishes a session. The login process
>needs to stay around to cleanup the session when the user
>logs off. Once the password is authenticated, login does
>a fork and the child execs the users shell. They child
>is the first process which actually _SHOULD_ be not
>root.
>
>Everything is as you would expect. Telnet will accumulate
>nearly no cpu time. login will accumulate none until
>the user logs off. These will not be significant in
>process accounting or load average. You should be able
>to safely ignore them.
>
>Now, back to the original question. What are you trying
>to accomplish?
>
>Thanks,
>Robin
>
>On Thu, Jan 08, 2004 at 04:16:20AM +1100, sena wrote:
>
>
>>Hi Robin,
>>
>>(As previously I could not copy to linux-kernel I am sending it again)
>>
>>Thanks for your time.
>>
>>I have loaded the module sent with some changes and got some results.
>>
>>I got 2 people to login remotely using telnet.
>>
>>as herft and as herft1
>>
>>herft runs a small executable called ./myprog.
>>
>>The reseults I got is as follows in var/log/messages
>>
>>Jan 8 03:46:23 sena kernel: Pid 1193 uid 0 euid 0
>>Jan 8 03:46:23 sena kernel: Pid 1220 uid 0 euid 0
>>Jan 8 03:46:23 sena kernel: Pid 1251 uid 0 euid 0--->this is telnet
>>Jan 8 03:46:23 sena kernel: Pid 1252 uid 0 euid 0
>>Jan 8 03:46:23 sena kernel: Pid 1253 uid 500 euid 500 --->this is bash
>>Jan 8 03:46:23 sena kernel: Pid 1381 uid 500 euid 500--->This is exec
>>Jan 8 03:46:23 sena kernel: Pid 1383 uid 0 euid 0---->telnet
>>Jan 8 03:46:23 sena kernel: Pid 1384 uid 0 euid 0
>>Jan 8 03:46:23 sena kernel: Pid 1385 uid 501 euid 501---->This is bash
>>Jan 8 03:46:23 sena kernel: Pid 1417 uid 0 euid 0
>>
>>
>>Then I ps -ale
>>
>>The result is as follows
>>UID PID
>>100 S 0 1251 638 0 69 0 - 426 do_sel ? 00:00:00
>>in.telnetd-----------This is telnet
>>100 S 0 1252 1251 0 69 0 - 599 wait4 ? 00:00:00 login
>>100 S 500 1253 1252 0 69 0 - 1009 wait4 pts/4 00:00:00-------- bash
>>000 S 500 1381 1253 0 69 0 - 309 nanosl pts/4 00:00:00----------- myprog
>>100 S 0 1383 638 0 69 0 - 426 do_sel ? 00:00:00 in.telnetd----------This
>>is telnet
>>100 S 0 1384 1383 0 69 0 - 599 wait4 ? 00:00:00 login
>>100 S 501 1385 1384 0 68 0 - 1009 read_c pts/5 00:00:00--------- bash
>>100 R 0 1795 1220 0 74 0 - 762 - pts/3 00:00:00 ps
>>
>>
>>
>>I probably got the same thing as yours. This means task_struct uid has
>>got its owners uids in the processes started through telnet. but telnet
>>servers uid=0
>>
>>This means either I will have to consider telnet thing as root (though
>>it is started by the user login)
>>
>>or
>>
>>What do you think Robin?
>>
>>
>>THE CODE:
>>// count_active_tasks.c -
>>
>>#include <linux/sched.h>
>>#include <linux/module.h> // Needed by all modules
>>#include <linux/kernel.h> // Needed for KERN_ALERT
>>#include <linux/init.h> // Needed for the macros
>>
>>#define DRIVER_AUTHOR "Sena Seneviratne/Robin Holt"
>>
>>
>>#define DRIVER_DESC "A sample Test driver"
>>
>>#define _LOOSE_KERNEL_NAMES
>>/* With some combinations of Linux and gcc, tty.h will not compile without
>>_LOOSE_KERNEL_NAMES.
>>*/
>>#include <linux/tty.h> /* console_print() interface */
>>
>>static int
>>count_active_tasks_init(void)
>>{
>>struct task_struct *p;
>>
>>read_lock(&tasklist_lock);
>>
>>for_each_task(p) {
>>printk(KERN_EMERG "Pid %d uid %d euid %d\n",
>>p->pid, p->uid, p->euid);
>>
>>//console_print("Hellow");
>>
>>}
>>read_unlock(&tasklist_lock);
>>return 0;
>>}
>>
>>static void count_active_tasks_exit(void)
>>{
>>printk(KERN_ALERT "Goodbye, world 2\n");
>>}
>>
>>module_init(count_active_tasks_init);
>>module_exit(count_active_tasks_exit);
>>
>>MODULE_LICENSE("GPL");
>>
>>MODULE_AUTHOR(DRIVER_AUTHOR); // Robin/sena wrote this module
>>MODULE_DESCRIPTION(DRIVER_DESC); // What does this module do?
>>
>>MODULE_SUPPORTED_DEVICE("dummy testdevice");
>>
>>Thanks
>>Sena Seneviratne
>>Computer Engineering Lab
>>Sydney University
>>
>>
>-
>To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>the body of a message to majordomo@vger.kernel.org
>More majordomo info at http://vger.kernel.org/majordomo-info.html
>Please read the FAQ at http://www.tux.org/lkml/
>
>
>



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:59    [W:0.040 / U:0.072 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site