Re: Considering segfaults: how to protect other requests?

On 01/21/2012 10:16 AM, Kārlis Repsons wrote: > Hi there... > > Hopefully I've chosen a good place to start with a problem like this: > using apache2 prefork, I've confirmed that the same PHP extension can > receive multiple requests through browser and if one of those requests > causes a segfault, the other request, which is still in process during > the segfault, will be completed. Good... But how to be sure if this is > just a particular case or there is some way to know that ONLY those > requests will get terminated, which are themselves causing a segfault? > Or make it that way? > > What should I read to understand (sufficiently well) what is going on > inside of Apache and PHP to be sure of questions like this? I know short > is in a conflict with complete or even sufficient, but anyway... Have to > get started. Your advices are welcome! One of the big benefits of not relying on a multi-threaded architecture is that each preforked httpd (or php-fpm) process has its own separate memory space managed by the kernel. In an MT architecture you have to do your own memory management. I much prefer letting the kernel keep me safe. So yes, if something corrupts memory and an httpd process segfaults, there is no bleed-over into other httpd processes. This is an architectural feature, not a fluke. -Rasmus