thanx for clearing that jon. (btw. don't read my post regarding my explaination on how NULL deref. is handled...that was a horribly wrong statement i made :P) this might be off-topic...(it's non-linux!) while testing some add-on cards for the product we work on, we came across a problem. we had two legacy boards - one with PPC and the other with mips. somewhere in the code of the OS that runs on them there was a null pointer deref. On the PPC board the value cud be read but the mips one gave a processor exception (both these boards were using MMU-less config's) while trying to do the same. it turned out that actually there was a memeory attached to the 0 address but in mips we configured some region to be protected from write's and that's why we were getting that exception, whereas this fxn-ality was not available on the PPC board so it was going through. I wonder how uCLinux handles such cases??? -mandeep On Fri, 2004-11-19 at 05:53, Jon Masters wrote: > On Thu, 18 Nov 2004 10:27:49 +0530, Mandeep Sandhu > <mandeep_sandhu@xxxxxxxxxxx> wrote: > > > a basic question. How does the kernel handle NULL pointer > > dereferencing inside kernel space. > > Indeed. Since everyone else jumped on to a fantastic discussion of > NULL vs. 0 and other pointerisms but failed to answer your original > question concisely, I'll pipe up now. > > This varies by architecture, but the kernel essentially contains a > deliberately bad NULL page which is mapped at zero so that any > dereferences to that virtual address will result in a processor > exception calling the page_fault_handler, and having an oops error > result. It's just a simple case of throwing away a single page to try > to handle bad code. > > > I think on PowerPC 0 is a valid address > > On some processors, zero is not a valid address, on PowerPC processors > which substantially follow the specification then it is valid to have > a mapping at that location in virtual memory and zero is certainly a > valid physical memory address (as it usually is on most processors > Linux supports). So it's valid but deliberately triggers a page fault. > > Jon. -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
