Re: Re:tcp performance tuning using /proc/sys/net/ipv4

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



----- Original Message -----
From: syedk <syedk@intotoinc.com>
Date: 26 Mar 2004 12:22:23 +0530
To: suresh kumar <suresh_vin@indiainfo.com>
Subject: Re: Re:tcp performance tuning using /proc/sys/net/ipv4

> hi suresh,
>    explanation of possible DOS attack :
> if either of the client/server in a TCP connection is advertising a
> large window then it is possible the other end might send only part of
> the data falling withing the window(causing them to be stored in the
> buffers) say window is 64k but server sent everything except first 1k(so
> rest of 63k shud be buffered).. this may result in DOS attack if
> multiple such connections are opened..
Passing data from tcp buffer to the application does not depend 
on the buffer size only, it also depends on the psh flag in the 
tcp header. 
coming to the dos attack, still I didn't see any dos attack. 
cud u be more specific like where the application is being
run and where the dos attack is created wrt client and server and how.
> 
> 
> but now by some googling i came to know that linux stack automatically
> tunes the window sizes if there is memory crunch...
>          now i am facing a problem :
>   in client i am advertising a large window(64k) with setsockopt but in
> the tcpdump the client window is changing b/n 5840 to 14170....
that is what I said that some systems can reset the window size to some
value. I asked u to run getsockopt on the socket desc. itself after client 
connect call is made. 
> how can i make sure that client sends window size equal to 64k..???
what xactly r u looking in to? there are ways to do it.
> when did getsockopt to display rcvbuf it is showing it as 131070(which
> is greater than 64k)... does this rcvbuf has anything do with tcp
> window?? 
what is this rcvbuf.
>    
> On Fri, 2004-03-26 at 11:15, suresh kumar wrote:
> > 
> > 
> > hello list members,
> >    this is regarding linux tcp implementation ..
> > if a tcp client application is requesting large receive window using
> > setsockopt.. 
> > and is opening multiple connections to  different servers...
> > there is a possiblity that tcp buffers might get exhausted(in scenarios
> > where there is lot of congestion or when a server is deliberately 
> > sending only some packets in the advertised window whereby the native
> > tcp has to buffer all these packets before passing to the user,
> > resulting in a DOS attack).
> > 
> > Could u be more specific as how this would be a DOS attack?
> > 
> > ->My question is whether linux tcp implementation deal with such
> > situations by lowering the window sizes outgoing packets so that buffers
> > will not get exhausted....
> > 
> > 
> > When application requested too much of buffer space, during
> > connection establishment time, may be set to some default max 
> > value by systems. Use getsockopt to find the value. 
> > Modiying the window size once the connection is established,
> > doesn't change the buffer space allocated. 
> > 
> > i have seen tcp_adv_win_scale and tcp_app_win of proc by which one can
> > control the window sizes.
> > 1st signifies size of the window+application buffer to isolate 
> > network from application latencies and scheduling. 
> > 2nd signifies size of the application buffer. 
> > 
> > ->how can one use these options to mitigate the aforesaid problems?
> > and does this problem is  automatically taken care of by linux stack...
> > I might be missing some thing, if u are more specific abt the 
> > DOS attack that u see, I will try to explain. Correct me 
> > if I am wrong in ne of the above.
> > cheers
> > Suresh
> > 
> > thanks and regards
> > Syed
> > 
> > 
> > 
> > -- 
> > ______________________________________________
> > IndiaInfo Mail - the free e-mail service with a difference! www.indiainfo.com 
> > Check out our value-added Premium features, such as an extra 20MB for mail storage, POP3, e-mail forwarding, and ads-free mailboxes!
> > 
> > Powered by Outblaze
> > 
> > --
> > Kernelnewbies: Help each other learn about the Linux kernel.
> > Archive:       http://mail.nl.linux.org/kernelnewbies/
> > FAQ:           http://kernelnewbies.org/faq/
> 
> 



V.V Suresh Kumar


-- 
______________________________________________
IndiaInfo Mail - the free e-mail service with a difference! www.indiainfo.com 
Check out our value-added Premium features, such as an extra 20MB for mail storage, POP3, e-mail forwarding, and ads-free mailboxes!

Powered by Outblaze

--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive:       http://mail.nl.linux.org/kernelnewbies/
FAQ:           http://kernelnewbies.org/faq/


[Index of Archives]     [Newbies FAQ]     [Linux Kernel Mentors]     [Linux Kernel Development]     [IETF Annouce]     [Git]     [Networking]     [Security]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux RAID]     [Linux SCSI]     [Linux ACPI]
  Powered by Linux