Hi all- This adds a new "xonly" mode for vsyscalls and makes it the default. xonly is a bit more secure -- Kees knows about an exploit that relied on read access to the vsyscall page. It's also nicer from a paging perspective, as it doesn't require user access to any of the kernel address space as far as the CPU is concerned. This would, for example, allow a much simpler implementation of per-process vsyscall disabling.
I will follow up with two non-x86 changes that are related but have no dependencies. Changes from v1: - Minor cleanups (Kees) - Add a searchable message when a vsyscall read is denied (Kees) - The test case is vastly improved - Get rid of the extra gate vma object - Add the __ro_after_init patch Andy Lutomirski (8): x86/vsyscall: Remove the vsyscall=native documentation x86/vsyscall: Add a new vsyscall=xonly mode x86/vsyscall: Show something useful on a read fault x86/vsyscall: Document odd SIGSEGV error code for vsyscalls selftests/x86/vsyscall: Verify that vsyscall=none blocks execution x86/vsyscall: Change the default vsyscall mode to xonly x86/vsyscall: Add __ro_after_init to global variables selftests/x86: Add a test for process_vm_readv() on the vsyscall page .../admin-guide/kernel-parameters.txt | 11 +- arch/x86/Kconfig | 35 +++-- arch/x86/entry/vsyscall/vsyscall_64.c | 37 +++++- arch/x86/include/asm/vsyscall.h | 6 +- arch/x86/mm/fault.c | 18 ++- tools/testing/selftests/x86/test_vsyscall.c | 120 ++++++++++++++---- 6 files changed, 174 insertions(+), 53 deletions(-) -- 2.21.0
