From mboxrd@z Thu Jan 1 00:00:00 1970 From: Catalin Marinas Subject: Re: [PATCH v7 4/4] arm64/syscalls: Architecture specific pre-usermode check Date: Mon, 10 Apr 2017 18:12:40 +0100 Message-ID: <20170410171240.GE30647@e104818-lin.cambridge.arm.com> References: <20170410164420.64003-1-thgarnie@google.com> <20170410164420.64003-4-thgarnie@google.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: List-Post: List-Help: List-Unsubscribe: List-Subscribe: Content-Disposition: inline In-Reply-To: <20170410164420.64003-4-thgarnie@google.com> To: Thomas Garnier Cc: Martin Schwidefsky , Heiko Carstens , Arnd Bergmann , Dave Hansen , Andrew Morton , David Howells , =?iso-8859-1?Q?Ren=E9?= Nyffenegger , "Paul E . McKenney" , Ingo Molnar , Thomas Gleixner , Oleg Nesterov , Stephen Smalley , Pavel Tikhomirov , Ingo Molnar , "H . Peter Anvin" , Andy Lutomirski , Paolo Bonzini , Kees Cook , Rik van Riel , Josh Poimboeuf , Borislav Petkov , Brian Gerst , "Kirill A . Shutemov" List-Id: linux-api@vger.kernel.org On Mon, Apr 10, 2017 at 09:44:20AM -0700, Thomas Garnier wrote: > Disable the generic pre-usermode check in favor of an optimized > implementation. This patch adds specific checks on user-mode return path > to make it faster and smaller. > > The address limit is checked on each syscall return path to user-mode. > If it was changed, a generic handler is called to stop the kernel on an > explicit check. > > Signed-off-by: Thomas Garnier > --- > Based on next-20170410 > --- > arch/arm64/Kconfig | 1 + > arch/arm64/kernel/entry.S | 13 +++++++++++++ > 2 files changed, 14 insertions(+) > > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig > index 9b8fcab7da56..3f9e8e7d9376 100644 > --- a/arch/arm64/Kconfig > +++ b/arch/arm64/Kconfig > @@ -24,6 +24,7 @@ config ARM64 > select ARCH_WANT_COMPAT_IPC_PARSE_VERSION > select ARCH_WANT_FRAME_POINTERS > select ARCH_HAS_UBSAN_SANITIZE_ALL > + select ARCH_NO_SYSCALL_VERIFY_PRE_USERMODE_STATE > select ARM_AMBA > select ARM_ARCH_TIMER > select ARM_GIC > diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S > index 43512d4d7df2..bdd094c7837a 100644 > --- a/arch/arm64/kernel/entry.S > +++ b/arch/arm64/kernel/entry.S > @@ -744,6 +744,8 @@ ENDPROC(cpu_switch_to) > ret_fast_syscall: > disable_irq // disable interrupts > str x0, [sp, #S_X0] // returned x0 > + ldr x2, [tsk, #TSK_TI_ADDR_LIMIT] // check addr limit change > + tbnz x2, #63, addr_limit_fail > ldr x1, [tsk, #TSK_TI_FLAGS] // re-check for syscall tracing > and x2, x1, #_TIF_SYSCALL_WORK > cbnz x2, ret_fast_syscall_trace > @@ -771,6 +773,9 @@ work_pending: > */ > ret_to_user: > disable_irq // disable interrupts > + ldr x2, [tsk, #TSK_TI_ADDR_LIMIT] // check addr limit change > + tbnz x2, #63, addr_limit_fail > + Nitpick: unnecessary empty line added. > ldr x1, [tsk, #TSK_TI_FLAGS] > and x2, x1, #_TIF_WORK_MASK > cbnz x2, work_pending > @@ -780,6 +785,14 @@ finish_ret_to_user: > ENDPROC(ret_to_user) > > /* > + * Address limit was incorrect before returning in user-mode which could be > + * used to elevate privileges. Call the generic handler to stop the kernel on > + * the appropriate check. This function does not return. > + */ > +addr_limit_fail: > + bl address_limit_check_failed You could do a "b address_limit_check_failed" rather than "bl" since we don't care about returning here (and a nitpick: insert tab between "b" and "address_limit_check_failed"). With the above: Reviewed-by: Catalin Marinas