https://sourceware.org/bugzilla/show_bug.cgi?id=28824
--- Comment #20 from Alan Modra <amodra at gmail dot com> --- (In reply to Alan Modra from comment #19) > It might be even better to put the SHT_NOBITS padding on the header before the > relro segment. No, that can't work. Since the end of the relro segment must end on a maxpagesize boundary that fixes the start of the on-disk relro segment. The end of the previous segment is also fixed, so padding there doesn't help fit the two segments together on disk. Padding at the end works, because it simply extends the end of the relro segment to a maxpagesize boundary. However there is a small problem with padding at the end, because the padding separates parts of the GOT. For example, on x86_64-linux without -z now, 24 bytes at the start of .got.plt is supposed to be relro. We lose that protection when padding. -- You are receiving this mail because: You are on the CC list for the bug.