Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752161AbdGDNmA (ORCPT ); Tue, 4 Jul 2017 09:42:00 -0400 Received: from Galois.linutronix.de ([146.0.238.70]:48321 "EHLO Galois.linutronix.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751536AbdGDNl7 (ORCPT ); Tue, 4 Jul 2017 09:41:59 -0400 Date: Tue, 4 Jul 2017 15:41:10 +0200 (CEST) From: Thomas Gleixner To: Mikulas Patocka cc: Ingo Molnar , Bernhard Held , Andy Lutomirski , Toshi Kani , Borislav Petkov , Andrew Morton , Brian Gerst , Linus Torvalds , "H. Peter Anvin" , Peter Zijlstra , "Luis R. Rodriguez" , Denys Vlasenko , Josh Poimboeuf , "linux-kernel@vger.kernel.org" Subject: Re: [PATCH v2] X86: don't report PAT on CPUs that don't support it In-Reply-To: Message-ID: References: <3d69fb9d-651a-8266-8e00-789fedd74659@gmx.de> User-Agent: Alpine 2.20 (DEB 67 2015-01-07) MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4663 Lines: 163 On Mon, 3 Jul 2017, Mikulas Patocka wrote: > Is there any progress with this patch? Will you accept it or do you want > some changes to it? Aside of the unparseable changelog, that patch is mostly duct tape. 1) __pat_enabled should be renamed to pat_disabled, as that is the actual purpose of that variable 2) Making the call to init_cache_modes() conditional in setup_arch() is pointless. init_cache_modes() has it's own protection against multiple invocations. 3) It adds yet another invocation of init_cache_modes() instead of getting rid of the ones in pat_disable() and the pat disabled case in pat_init(). I've reworked the whole thing into the patch below. Thanks, tglx 8<--------------------- Subject: x86/mm/pat: Don't report PAT on CPUs that don't support it From: Mikulas Patocka Date: Tue, 6 Jun 2017 18:49:39 -0400 (EDT) The pat_enabled() logic is broken on CPUs which do not support PAT and where the initialization code fails to call pat_init(). Due to that the enabled flag stays true and pat_enabled() returns true wrongfully. As a consequence the mappings, e.g. for Xorg, are set up with the wrong caching mode and the required MTRR setups are omitted. To cure this the following changes are required: 1) Make pat_enabled() return true only if PAT initialization was invoked and successful. 2) Invoke init_cache_modes() unconditionally in setup_arch() and remove the extra callsites in pat_disable() and the pat disabled code path in pat_init(). Also rename __pat_enabled to pat_disabled to reflect the real purpose of this variable. Signed-off-by: Mikulas Patocka Cc: Bernhard Held Cc: Toshi Kani Cc: Peter Zijlstra Cc: Brian Gerst Cc: "Luis R. Rodriguez" Cc: Borislav Petkov Cc: Andy Lutomirski Cc: Josh Poimboeuf Cc: Denys Vlasenko Cc: Andrew Morton Cc: Linus Torvalds --- arch/x86/include/asm/pat.h | 1 + arch/x86/kernel/setup.c | 7 +++++++ arch/x86/mm/pat.c | 22 +++++++++------------- 3 files changed, 17 insertions(+), 13 deletions(-) --- a/arch/x86/include/asm/pat.h +++ b/arch/x86/include/asm/pat.h @@ -7,6 +7,7 @@ bool pat_enabled(void); void pat_disable(const char *reason); extern void pat_init(void); +extern void init_cache_modes(void); extern int reserve_memtype(u64 start, u64 end, enum page_cache_mode req_pcm, enum page_cache_mode *ret_pcm); --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -1076,6 +1076,13 @@ void __init setup_arch(char **cmdline_p) max_possible_pfn = max_pfn; /* + * This call is required when the CPU does not support PAT. If + * mtrr_bp_init() invoked it already via pat_init() the call has no + * effect. + */ + init_cache_modes(); + + /* * Define random base addresses for memory sections after max_pfn is * defined and before each memory section base is used. */ --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -37,14 +37,13 @@ #undef pr_fmt #define pr_fmt(fmt) "" fmt -static bool boot_cpu_done; - -static int __read_mostly __pat_enabled = IS_ENABLED(CONFIG_X86_PAT); -static void init_cache_modes(void); +static bool __read_mostly boot_cpu_done; +static bool __read_mostly pat_disabled = !IS_ENABLED(CONFIG_X86_PAT); +static bool __read_mostly pat_initialized; void pat_disable(const char *reason) { - if (!__pat_enabled) + if (pat_disabled) return; if (boot_cpu_done) { @@ -52,10 +51,8 @@ void pat_disable(const char *reason) return; } - __pat_enabled = 0; + pat_disabled = true; pr_info("x86/PAT: %s\n", reason); - - init_cache_modes(); } static int __init nopat(char *str) @@ -67,7 +64,7 @@ early_param("nopat", nopat); bool pat_enabled(void) { - return !!__pat_enabled; + return pat_initialized; } EXPORT_SYMBOL_GPL(pat_enabled); @@ -225,6 +222,7 @@ static void pat_bsp_init(u64 pat) } wrmsrl(MSR_IA32_CR_PAT, pat); + __pat_initialized = true; __init_cache_modes(pat); } @@ -242,7 +240,7 @@ static void pat_ap_init(u64 pat) wrmsrl(MSR_IA32_CR_PAT, pat); } -static void init_cache_modes(void) +void init_cache_modes(void) { u64 pat = 0; static int init_cm_done; @@ -306,10 +304,8 @@ void pat_init(void) u64 pat; struct cpuinfo_x86 *c = &boot_cpu_data; - if (!pat_enabled()) { - init_cache_modes(); + if (pat_disabled) return; - } if ((c->x86_vendor == X86_VENDOR_INTEL) && (((c->x86 == 0x6) && (c->x86_model <= 0xd)) ||