IRC, freenode, #hurd, 2012-07-04
<braunr> yes, we have integer overflows on resident_page_count, but luckily, the member is rarely used
See also gnumach vm object resident page count.
IRC, freenode, #hurd, 2013-06-04
<elmig> this is declared as int on vm_object.h <elmig> and as it as counter it's always positive <braunr> yes it should be unsigned <elmig> ok <braunr> but leave it as it is for consistency with the rest <elmig> i send patch :) <braunr> please no <braunr> unless you've fully determined the side effects <elmig> i've grepped the vars and saw only comparisons > and = 0 <elmig> never less than 0 <braunr> > 0 is the same <braunr> well <braunr> > not, but >= would be a problem <elmig> http://paste.debian.net/plain/8527 <elmig> asctually no >=0 <braunr> still, i don't want to change that unless it's strictly necessary <braunr> hum, you're grepping ref_count, not resident_page_count <elmig> i did both <elmig> on resident_page_count theres resident_page_count >= 0 <elmig> = 0, == 0 <braunr> this isn't the only possible issue <braunr> anyway <braunr> for now there is no reason to change anything unless you do a full review <elmig> only place i see resdent_page_count and page_count being decreased it's on vm/vm_resident.c <elmig> vm_page_remove() and vm_page_replace()