From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Cooper Subject: Re: [PATCH v2 1/7] x86: Add support for STAC/CLAC instructions Date: Wed, 23 Apr 2014 13:50:15 +0100 Message-ID: <5357B707.5030803@citrix.com> References: <1398263648-31994-1-git-send-email-feng.wu@intel.com> <5357908F.2090607@citrix.com> <5357D161020000780000B763@nat28.tlf.novell.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <5357D161020000780000B763@nat28.tlf.novell.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Jan Beulich Cc: Kevin Tian , Feng Wu , "ian.campbell@citrix.com" , Eddie Dong , "xen-devel@lists.xen.org" , Jun Nakajima List-Id: xen-devel@lists.xenproject.org On 23/04/14 13:42, Jan Beulich wrote: >>>> On 23.04.14 at 14:11, wrote: >>>> +#define ASM_AC(op) \ >>>> + pushq %rax; \ >>>> + leaq boot_cpu_data(%rip), %rax; \ >>>> + btl $X86_FEATURE_SMAP-7*32, CPUINFO86_leaf7_features(%rax); \ >>>> + jnc 881f; \ >>>> + op; \ >>>> +881: popq %rax >>> While it is unlikely to change going forwards, $(X86_FEATURE_SMAP & 31) >>> looks more obviously correct. >> Okay. >>> Also, given that boot_cpu_data(%rip) is an assembler know constant >>> value, and CPUINFO86_leaf7_features is a constant offset from that, is >>> there any way of persuading the assembler to conjoin the two and forgo >>> the push, lea and pop. (I have had a go, but lack sufficient caffeine >>> this early in the day) >>> >>> ~Andrew >> I thought about this idea before posting this version, but I didn't get a >> good >> solution. I will continue to find a better way to handle this. > btl $X86_FEATURE_SMAP & 31, ((X86_FEATURE_SMAP >> 3) & ~3)+CPUINFO86_features+boot_cpu_data(%rip) > > (at once replacing CPUINFO86_ext_features with CPUINFO86_features, > perhaps also changing the [inconsistent] name prefix; likely a suitable > macro could be created usable both here and wherever > CPUINFO86_ext_features is being used currently) > > Jan > This needs careful synchronising with the "no-smap" command line parameter which can shoot that bit out of boot features some time after it being set. It might be easier just to have bool_t __read_mostly smap_in_use; ~Andrew