Choose your poison

Posted Dec 24, 2015 15:02 UTC (Thu) by abatters (✭ supporter ✭, #6932)
Parent article: Two PaX features move toward the mainline

> (0xff for all architectures except x86-64, which uses 0xfe)

What is the reason for setting SLAB_MEMORY_SANITIZE_VALUE differently for x86-64?

Choose your poison

Posted Dec 28, 2015 18:14 UTC (Mon) by iq-0 (subscriber, #36655) [Link] (1 responses)

I think it's partially meant as a value that can't sensibly be used as a pointer and is also unusable for most other purposes when used accidentally. By using 0xfe on x86-64 you also run afoul of the sign extension for pointers, which gives even beter protection for pointers.

Choose your poison

Posted Dec 29, 2015 9:13 UTC (Tue) by renox (guest, #23785) [Link]

Thanks for this clarification, I had the same reaction as barryascott: this patch seemed to conflate security and debugability, now with your explanation it seems that this is not necessarily the case.