The fancy ideas of aggressively stripping the vdso and placing the vvar data after the vdso have been considerably more painful than I hoped. This reverts to using objcopy -S to strip the vdso (thus eliminating the whole fake section mechanism and all of its attendent build-time fragility, I hope) and places the vvar data before the vdso.
The latter is a bit messy: ELF wasn't really designed to have symbols at negative virtual addresses. I think I got all the bitness issues and cross-compilation stuff right, but I'm not well-equipped to test that. The kbuild robot hasn't complained yet, though. I don't know whether this is 3.16 material. I'm not aware of an actual problem at this point, but the current state of affairs is quite fragile. Applies on top of tip/x86/urgent as of a few minutes ago. Andy Lutomirski (2): x86,vdso: Move the vvar area before the vdso text x86,vdso: Get rid of the fake section mechanism arch/x86/include/asm/vdso.h | 18 +-- arch/x86/vdso/Makefile | 16 ++- arch/x86/vdso/vdso-fakesections.c | 21 ---- arch/x86/vdso/vdso-layout.lds.S | 44 ++++---- arch/x86/vdso/vdso2c.c | 128 ++++++++++++++++----- arch/x86/vdso/vdso2c.h | 227 +++++++------------------------------- arch/x86/vdso/vma.c | 20 ++-- 7 files changed, 188 insertions(+), 286 deletions(-) delete mode 100644 arch/x86/vdso/vdso-fakesections.c -- 1.9.3 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
