Re: Change User on a pconnect

On Sat, 6 Nov 1999, Thies C. Arntzen wrote: > On Sat, 6 Nov 1999, Rasmus Lerdorf wrote: > > > > > By the way, what prevents someone from guessing a persistent link > > > > identifier and using it without authenticating on it right now? I don't > > > > see the code that prevents this, but I should take a closer look. I'll > > > > play with this stuff this afternoon. I'll probably wait until after > > > > 3.0.13/4.0RC2 before I commit it though to avoid last-second breakage. > > > > > > hmm, in PHP 4 the link would be a resource and only c-modules can create > > > resources. in PHP 3 this could be a problem, and thinking a bit longer > > > .... i believe every script could guess a link-number - that's not good. > > > > Yeah, I was only looking at PHP 3 code where the thing is just an > > integer. I'd like to verify/fix this before 3.0.13 and will get right on > > it after lunch here. > > rasmus, > > that sounds non-trivial to me! in php4 we just need to make sure that only > c-modules can create resources. > > the only thing that comes to mind is to offset link-id's by an offset > that's randomly generated in the RINIT() function of the module (or > global). that way link-id's would be "harder" to guess... Hrm.. I was thinking more along the lines of either setting a flag somewhere linked to the connection, or to somehow copy the connection from the plist to the normal list on a connect and not let any of the other functions access the plist directly. > the whole thing is only an issue for ISP with evil customers, isn't > it? (one customer tries to steal data from another customers table in the > same database-instance) Correct. > i think we need to rethink the whole plist/pconnect stuff anyhow - the > do_logon() functions are really, really hard to understand - and a lot of > their complexity is repeated in every module that supports pconnects. (and > now we have the same "bug" in all db-modules;-( a rewrite in that area > would also be needed if we want to share connections between threads in > ZTS mode. so: let's try to "close the hole" for PHP 3 and come up with > something better form PHP 4 *after* B3. I agree that the code is a little hairy. -Rasmus