Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH] kdump, vmcoreinfo: Export sme_me_mask value to vmcoreinfo
To:     Borislav Petkov <bp@alien8.de>
Cc:     Baoquan He <bhe@redhat.com>, Petr Tesarik <ptesarik@suse.cz>,
        linux-kernel@vger.kernel.org, x86@kernel.org,
        kexec@lists.infradead.org, mingo@redhat.com, tglx@linutronix.de,
        dyoung@redhat.com
References: <1263471c-a27d-a698-15f0-b5947f13ea93@redhat.com>
 <20181026182440.20a4b107@ezekiel.suse.cz>
 <20181026222517.GB26927@nazgul.tnic> <20181027081343.GA1884@MiWiFi-R3L-srv>
 <20181027091007.GB1046@nazgul.tnic> <20181027093917.GA14493@MiWiFi-R3L-srv>
 <20181027101221.GC1046@nazgul.tnic> <20181027110858.GB14493@MiWiFi-R3L-srv>
 <C5F8ED8E-F361-4C6C-B9E2-E3C8BCD8E050@alien8.de>
 <c55bef2a-b972-7e6e-b698-6ac1bc736046@redhat.com>
 <20181027145151.GA7338@nazgul.tnic>
From:   lijiang <lijiang@redhat.com>
Message-ID: <390b16ba-05d7-3590-c3b2-4b69aa3ae051@redhat.com>
Date:   Mon, 29 Oct 2018 15:59:48 +0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.9.1
MIME-Version: 1.0
In-Reply-To: <20181027145151.GA7338@nazgul.tnic>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

在 2018年10月27日 22:51, Borislav Petkov 写道:
> On Sat, Oct 27, 2018 at 10:41:56PM +0800, lijiang wrote:
>> Actually, the value of 'sme_me_mask' is 0x800000000000 when SME is
>> enabled, otherwise it is 0. That is to say, if the bit 47 is set, the
>> bit number is also 0x800000000000 (1 << 47UL);
> 
> Yes, and you can simply copy the mask into your variable and export
> that. 

Ok, if you agree with the following patch, i will post it again.

diff --git a/arch/x86/kernel/machine_kexec_64.c b/arch/x86/kernel/machine_kexec_64.c
index 4c8acdfdc5a7..de363796ed20 100644
--- a/arch/x86/kernel/machine_kexec_64.c
+++ b/arch/x86/kernel/machine_kexec_64.c
@@ -29,6 +29,9 @@
 #include <asm/setup.h>
 #include <asm/set_memory.h>

+u64 sme_mask;
+EXPORT_SYMBOL(sme_mask);
+
 #ifdef CONFIG_KEXEC_FILE
 const struct kexec_file_ops * const kexec_file_loaders[] = {
                &kexec_bzImage64_ops,
@@ -357,6 +360,9 @@ void arch_crash_save_vmcoreinfo(void)
        vmcoreinfo_append_str("NUMBER(pgtable_l5_enabled)=%d\n",
                        pgtable_l5_enabled());

+       sme_mask = sme_me_mask;
+       VMCOREINFO_NUMBER(sme_mask);
+
 #ifdef CONFIG_NUMA
        VMCOREINFO_SYMBOL(node_data);
        VMCOREINFO_LENGTH(node_data, MAX_NUMNODES);
-- 

> Thinking about it more, though, it might be better if you instead
> export a smaller value - not an u64 - and construct the mask in
> userspace. I.e., a u8 which should be enough for your current purposes.
> 
> I say current because if it turns out we need to export more SME-related
> info to userspace, exporting an u64 and then OR-ing in more information
> in it would allow that. u8 not so much. So doing something like:
> 
> [ misc	   ][ enc bit  ][ other misc SME info 				      ]
> 0000_0000_0000_0000_1000_0000_0000_0000_0000_0000_0000_0000_0000_0000_0000_0000
> 63   59   55   51   47   43   39   35   31   27   23   19   15   11   7    3
> 
> would allow for exporting that additional info.
> 
> Especially if we want to use VMCOREINFO for more than kexec things.
> 

As you mentioned, if need, the definition of this variable(u64 sme_mask) can be extended
conveniently in the future(The bit might be redefined).

Thanks.
Lianbo