Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S966874AbeAOMqr (ORCPT + 1 other); Mon, 15 Jan 2018 07:46:47 -0500 Received: from mail.linuxfoundation.org ([140.211.169.12]:49318 "EHLO mail.linuxfoundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S966836AbeAOMqm (ORCPT ); Mon, 15 Jan 2018 07:46:42 -0500 From: Greg Kroah-Hartman To: linux-kernel@vger.kernel.org Cc: Greg Kroah-Hartman , stable@vger.kernel.org, David Woodhouse , Thomas Gleixner , "Peter Zijlstra (Intel)" , gnomes@lxorguk.ukuu.org.uk, Rik van Riel , Andi Kleen , Josh Poimboeuf , thomas.lendacky@amd.com, Linus Torvalds , Jiri Kosina , Andy Lutomirski , Dave Hansen , Kees Cook , Tim Chen , Paul Turner , Greg Kroah-Hartman Subject: [PATCH 4.9 94/96] x86/retpoline: Fill return stack buffer on vmexit Date: Mon, 15 Jan 2018 13:35:33 +0100 Message-Id: <20180115123411.921462518@linuxfoundation.org> X-Mailer: git-send-email 2.15.1 In-Reply-To: <20180115123404.270241256@linuxfoundation.org> References: <20180115123404.270241256@linuxfoundation.org> User-Agent: quilt/0.65 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Return-Path: 4.9-stable review patch. If anyone has any objections, please let me know. ------------------ From: David Woodhouse commit 117cc7a908c83697b0b737d15ae1eb5943afe35b upstream. In accordance with the Intel and AMD documentation, we need to overwrite all entries in the RSB on exiting a guest, to prevent malicious branch target predictions from affecting the host kernel. This is needed both for retpoline and for IBRS. [ak: numbers again for the RSB stuffing labels] Signed-off-by: David Woodhouse Signed-off-by: Thomas Gleixner Tested-by: Peter Zijlstra (Intel) Cc: gnomes@lxorguk.ukuu.org.uk Cc: Rik van Riel Cc: Andi Kleen Cc: Josh Poimboeuf Cc: thomas.lendacky@amd.com Cc: Linus Torvalds Cc: Jiri Kosina Cc: Andy Lutomirski Cc: Dave Hansen Cc: Kees Cook Cc: Tim Chen Cc: Greg Kroah-Hartman Cc: Paul Turner Link: https://lkml.kernel.org/r/1515755487-8524-1-git-send-email-dwmw@amazon.co.uk Signed-off-by: David Woodhouse Signed-off-by: Greg Kroah-Hartman --- arch/x86/include/asm/nospec-branch.h | 78 ++++++++++++++++++++++++++++++++++- arch/x86/kvm/svm.c | 4 + arch/x86/kvm/vmx.c | 4 + 3 files changed, 85 insertions(+), 1 deletion(-) --- a/arch/x86/include/asm/nospec-branch.h +++ b/arch/x86/include/asm/nospec-branch.h @@ -7,6 +7,48 @@ #include #include +/* + * Fill the CPU return stack buffer. + * + * Each entry in the RSB, if used for a speculative 'ret', contains an + * infinite 'pause; jmp' loop to capture speculative execution. + * + * This is required in various cases for retpoline and IBRS-based + * mitigations for the Spectre variant 2 vulnerability. Sometimes to + * eliminate potentially bogus entries from the RSB, and sometimes + * purely to ensure that it doesn't get empty, which on some CPUs would + * allow predictions from other (unwanted!) sources to be used. + * + * We define a CPP macro such that it can be used from both .S files and + * inline assembly. It's possible to do a .macro and then include that + * from C via asm(".include ") but let's not go there. + */ + +#define RSB_CLEAR_LOOPS 32 /* To forcibly overwrite all entries */ +#define RSB_FILL_LOOPS 16 /* To avoid underflow */ + +/* + * Google experimented with loop-unrolling and this turned out to be + * the optimal version — two calls, each with their own speculation + * trap should their return address end up getting used, in a loop. + */ +#define __FILL_RETURN_BUFFER(reg, nr, sp) \ + mov $(nr/2), reg; \ +771: \ + call 772f; \ +773: /* speculation trap */ \ + pause; \ + jmp 773b; \ +772: \ + call 774f; \ +775: /* speculation trap */ \ + pause; \ + jmp 775b; \ +774: \ + dec reg; \ + jnz 771b; \ + add $(BITS_PER_LONG/8) * nr, sp; + #ifdef __ASSEMBLY__ /* @@ -76,6 +118,20 @@ #endif .endm + /* + * A simpler FILL_RETURN_BUFFER macro. Don't make people use the CPP + * monstrosity above, manually. + */ +.macro FILL_RETURN_BUFFER reg:req nr:req ftr:req +#ifdef CONFIG_RETPOLINE + ANNOTATE_NOSPEC_ALTERNATIVE + ALTERNATIVE "jmp .Lskip_rsb_\@", \ + __stringify(__FILL_RETURN_BUFFER(\reg,\nr,%_ASM_SP)) \ + \ftr +.Lskip_rsb_\@: +#endif +.endm + #else /* __ASSEMBLY__ */ #define ANNOTATE_NOSPEC_ALTERNATIVE \ @@ -119,7 +175,7 @@ X86_FEATURE_RETPOLINE) # define THUNK_TARGET(addr) [thunk_target] "rm" (addr) -#else /* No retpoline */ +#else /* No retpoline for C / inline asm */ # define CALL_NOSPEC "call *%[thunk_target]\n" # define THUNK_TARGET(addr) [thunk_target] "rm" (addr) #endif @@ -134,5 +190,25 @@ enum spectre_v2_mitigation { SPECTRE_V2_IBRS, }; +/* + * On VMEXIT we must ensure that no RSB predictions learned in the guest + * can be followed in the host, by overwriting the RSB completely. Both + * retpoline and IBRS mitigations for Spectre v2 need this; only on future + * CPUs with IBRS_ATT *might* it be avoided. + */ +static inline void vmexit_fill_RSB(void) +{ +#ifdef CONFIG_RETPOLINE + unsigned long loops = RSB_CLEAR_LOOPS / 2; + + asm volatile (ANNOTATE_NOSPEC_ALTERNATIVE + ALTERNATIVE("jmp 910f", + __stringify(__FILL_RETURN_BUFFER(%0, RSB_CLEAR_LOOPS, %1)), + X86_FEATURE_RETPOLINE) + "910:" + : "=&r" (loops), ASM_CALL_CONSTRAINT + : "r" (loops) : "memory" ); +#endif +} #endif /* __ASSEMBLY__ */ #endif /* __NOSPEC_BRANCH_H__ */ --- a/arch/x86/kvm/svm.c +++ b/arch/x86/kvm/svm.c @@ -44,6 +44,7 @@ #include #include #include +#include #include #include "trace.h" @@ -4917,6 +4918,9 @@ static void svm_vcpu_run(struct kvm_vcpu #endif ); + /* Eliminate branch target predictions from guest mode */ + vmexit_fill_RSB(); + #ifdef CONFIG_X86_64 wrmsrl(MSR_GS_BASE, svm->host.gs_base); #else --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -48,6 +48,7 @@ #include #include #include +#include #include "trace.h" #include "pmu.h" @@ -9026,6 +9027,9 @@ static void __noclone vmx_vcpu_run(struc #endif ); + /* Eliminate branch target predictions from guest mode */ + vmexit_fill_RSB(); + /* MSR_IA32_DEBUGCTLMSR is zeroed on vmexit. Restore it if needed */ if (debugctlmsr) update_debugctlmsr(debugctlmsr);