| |||||||||||||
Can user-space bugs be kernel regressions?Can user-space bugs be kernel regressions?Posted Aug 7, 2008 16:26 UTC (Thu) by NRArnot (subscriber, #3033)Parent article: Can user-space bugs be kernel regressions?
Should it not be standard practice that if a user passes an N-byte buffer and the kernel has less than N bytes to store in it, the rest should nevertheless be written (for example zeroed)? This would avoid the possibility of latent bugs such as this one biting years later. Instead, they'd probably bite while the original code was being debugged, and even if not, the bug would clearly be seen to be a latent memory-corruption bug in userspace and not anything that's the kernel's fault.
(Log in to post comments)
Can user-space bugs be kernel regressions? Posted Aug 7, 2008 20:30 UTC (Thu) by davecb (subscriber, #1574) [Link] That might be an interesting thing to propose to the LKML as a janitorial project. Of course, it would need to default off for the interfaces we find broken right now (;-)) but over time it would close this ABI hole... --dave (ex-ABI team guy) c-b
|
|||||||||||||