Memory Usage of Kernel Driver Module in Linux

Memory Usage of Kernel Driver Module in Linux

A nice tool memstrack .

kthread.c ( example for memory allocation of 50mb)

#include
#include
#include
#include
#include
#include


static int thread_init(void){

    char *buffer;
    int i =0;
    for(i=0;i<50 br="" i="">    {
        buffer = (char *)kmalloc(1000*1000, GFP_KERNEL);
    }
    if(buffer == NULL)
        printk(KERN_ERR "low memory...");
    else
        printk(KERN_ERR "Allocation succedded...\n");
    return 0;
}

void thread_exit(void){
        printk(KERN_INFO "done.");
}

2. run 

./memstrack --report module_summary,proc_slab_static --notui -o mem.txt

(https://github.com/ryncsn/memstrack)

3. cat mem.txt

======== Report format module_summary: ========
Module kthread using 50.0MB (12800 pages), peak allocation 50.0MB (12800 pages)
Module xfs using 0.1MB (31 pages), peak allocation 0.1MB (31 pages)
Module tg3 using 0.1MB (16 pages), peak allocation 0.1MB (16 pages)
Module sr_mod using 0.0MB (1 pages), peak allocation 0.0MB (1 pages)
Module cdrom using 0.0MB (0 pages), peak allocation 0.0MB (0 pages)
======== Report format module_summary END ========

you can cleary see 50mb tracked by the tool.

Forcing Packet to go through Wire using Two Ports of Same Card or 2 NIC on Single HOST Linux Machine

Forcing Packet to go through Wire using Two Ports of Same Card or 2 NIC on Single HOST Linux Machine

Src
https://wiki.psuter.ch/doku.php?id=force_local_traffic_through_external_ethernet_cable_by_using_ip_namespaces
https://serverfault.com/questions/127636/force-local-ip-traffic-to-an-external-interface




We have one interface which is called as loopback interface (lo). When we ping or send traffic to test local
interface it is the loopback interface which replies.

Lets say we have three interfaces on Linux PC eth1, eth2 and lo (loopback interface).
so whatver br the ip of eth1 and eth2. you can always ping them and packet will not actually go over wire.

To Force packet over wire we use either of approach
1.Iptables modification
2.Netns

This blog will use netns as it much clearner methid.

Normally in OS there is only one instance of Network stack and related sets of Table ( Arp , routing table etc).
With Namespace you logically have seperate have copy of All of Above.





ip netns add ns_server
ip netns add ns_client


ip link set ens1f0 netns ns_server
ip netns exec ns_server ip addr add dev ens1f0 192.168.1.1/24
ip netns exec ns_server ip link set dev ens1f0 up

ip link set ens1f1 netns ns_client
ip netns exec ns_client ip addr add dev ens1f1 192.168.1.2/24
ip netns exec ns_client ip link set dev ens1f1 up


ip netns exec ns_server iperf -s -B 192.168.1.1
ip netns exec ns_client iperf -c 192.168.1.1 -B 192.168.1.2






ethtool shows actual hardware stats ( dont rely on ifconfig/ip command output they are from kernel stats)

root@hp-p70:/home/fastlinq# ip netns exec ns_server ethtool -S ens1f0 | grep rcv
           rcv_pkts: 187171024
root@hp-p70:/home/fastlinq# ip netns exec ns_server ethtool -S ens1f0 | grep xmit
           xmit_pkts: 98899174