Sorry for my confusion. Checking other sources I see MADV_DONTNEED results in modifications to private mappings being lost (the section on zero-filling non-file-backed pages address this, in retrospect) which clearly changes semantics. I had noticed that MADV_DONTFORK can result in different behavior after a fork (eg, a parent mmaps a file, madvises dontfork, then forks, and the child writes to the page -- with madvise, the page is updated, visible to both processes, as is the backing file, without it the child will segfault) and thought there might be a simple typo. Alex On 04/01/2014 02:18 PM, Michael Kerrisk (man-pages) wrote: > On 03/28/2014 12:16 AM, Alex Roper wrote: >> Version: Current git repo >> >> The first paragraph of DESCRIPTION states that madvise does not change >> the semantics of an application except in the case of MADV_DONTNEED, >> which I believe should read MADV_DONTFORK. > > Alex, > > MADV_DONTNEED does change the semantics. But, could you say more > about wht you think MADV_DONTFORK should have been written here > instead? (Perhaps some fix is needed to the page.) > > Thanks, > > Michael > -- To unsubscribe from this list: send the line "unsubscribe linux-man" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
