kern.randompid: Random PID modulus

[Dan Lukes]

Miroslav Lachman wrote:
>>>>> Ma tedy ta randomizace nejaky smysl, nebo je to zbytecne?
>>>> Ano ;-)

> Kdyz jsem o tom loni uvazoval, tak me vubec nenapadlo, ze muze existovat 
> pripad, kdy random PID naopak muze zpusobit problemy a to jeste 
> zneuzitelne na dalku:

Je ovsem poctive rict, ze to neni chyba randomizace jako takove, ale 
spis chyba zpusobena nedbalou implementaci generatoru, chyba nezavisla 
na konkretni metode generovani sekvence.

> This flaw relies on the fact that a hello cookie created by the server 
> is generated using the current Unix timestamp (so up to the second) and 
> the PID of the process handling the request.

Jinymi slovy, nejde vubec o skutecne nahodne cislo. To ale neni samo o 
sobe problem (ve vztahu k probiranemu) - ten je jinde.

At uz je metoda generovani jakakoliv, korektni implementace musi 
zajistit aby nedoslo k opakovanemu vygenerovanni stejneho cisla pred 
vyprsenim pouzitelnosti minule inkarnace. Protoze hello cookie na teto 
vlastnosti proste zavisi.

> The author explains that such attack is not realizable on systems using 
> traditionnal sequential PID because it would require more than 65000 
> connections attempts to made in less than one second.

Mozna mi neco unika, ale me 65000 pokusu  spojeni behem jedne sekundy 
nepripada byt nemoznym. Takze i sekvencni generator by mel problem 
duplicit nejak resit.

Dan