Re: [LKP] [x86] 811565123a: BUG: kernel hang in early-boot stage, last printk: Probing EDD (edd=off to disable)... ok

[Andi Kleen]

On Fri, Oct 14, 2016 at 12:56:00PM +0800, Ye Xiaolong wrote:
> On 10/14, Ye Xiaolong wrote:
> >On 10/13, Andi Kleen wrote:
> >>Andi Kleen <andi@xxxxxxxxxxxxxx> writes:
> >>
> >>Any comments on this?
> >>
> >>I still cannot reproduce the failure unfortunately.
> >>
> >
> >Btw, you can try below commands to reproduce the error on your local
> >host, they will download the necessary images and run QEMU:
> >
> >       git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
> >       cd lkp-tests
> >       bin/lkp qemu -k KERNEL job-script  # job-script is attached in the original report email
> 
> Results show this hang may be related to gcc version, with gcc version
> 4.8.4 (Ubuntu 4.8.4-2ubuntu1~14.04.3), kernel could boot without early
> hang, but if kernel was built with gcc version 6.2.0 20160901 (Debian
> 6.2.0-3), it will stuck in the early stage:

I built a mainline gcc 6.2, unfortunately still doesn't reproduce. The kernel
with your config boots to root.

My guess is that something is broken with paravirt ops on 32bit
on that compiler.

I created a new patch with the probing code moved to a separate
function. Can you test if that works?

-Andi


commit 65c92de6678f04ce14b237d7073e164b98d9a8be
Author: Andi Kleen <ak@xxxxxxxxxxxxxxx>
Date:   Wed May 4 06:07:44 2016 -0700

    x86: Report Intel platform_id in /proc/cpuinfo
    
    We have a need to distinguish systems based on their platform ID.
    For example this is useful to distinguish systems with L4 cache
    versus ones without.
    
    There is a 3 bit identifier (also called processor flags) in
    the IA32_PLATFORM_ID MSR that can give a more fine grained
    identification of the CPU than just the model number/stepping.
    
    IA32_PLATFORM_ID is architectural.
    
    The MSR can be also accessed through /dev/cpu/*/msr, but that
    requires root and is awkward.
    
    The patch moves the reading of PLATFORM_INFO from the
    (late) microcode driver code into the main intel CPU initialization
    path and then also prints it in /proc/cpuinfo
    
    v2: Handle 0 platform_id. Fix commit message.
    v3: Move some code to cpu/intel.c
    v4: Update description too.
    v5: Move msr probe code out of line to w/a potential gcc 6 bug
    Cc: hmh@xxxxxxxxxx
    Signed-off-by: Andi Kleen <ak@xxxxxxxxxxxxxxx>

diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 63def9537a2d..c1313b3f3e59 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -135,6 +135,8 @@ struct cpuinfo_x86 {
 	/* Index into per_cpu list: */
 	u16			cpu_index;
 	u32			microcode;
+	u32			platform_id;
+	u8			has_platform_id;
 };
 
 #define X86_VENDOR_INTEL	0
diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
index fcd484d2bb03..7da7f008cee0 100644
--- a/arch/x86/kernel/cpu/intel.c
+++ b/arch/x86/kernel/cpu/intel.c
@@ -61,6 +61,19 @@ void check_mpx_erratum(struct cpuinfo_x86 *c)
 	}
 }
 
+/* noinline to work around problem with gcc 6.2 */
+static noinline void probe_platformid(struct cpuinfo_x86 *c)
+{
+	if ((c->x86_model >= 5) || (c->x86 > 6)) {
+		unsigned val[2];
+
+		/* get processor flags from MSR 0x17 */
+		rdmsr(MSR_IA32_PLATFORM_ID, val[0], val[1]);
+		c->platform_id = (val[1] >> 18) & 7;
+		c->has_platform_id = true;
+	}
+}
+
 static void early_init_intel(struct cpuinfo_x86 *c)
 {
 	u64 misc_enable;
@@ -211,6 +224,8 @@ static void early_init_intel(struct cpuinfo_x86 *c)
 	}
 
 	check_mpx_erratum(c);
+
+	probe_platformid(c);
 }
 
 #ifdef CONFIG_X86_32
diff --git a/arch/x86/kernel/cpu/microcode/intel.c b/arch/x86/kernel/cpu/microcode/intel.c
index cdc0deab00c9..fab07e49192e 100644
--- a/arch/x86/kernel/cpu/microcode/intel.c
+++ b/arch/x86/kernel/cpu/microcode/intel.c
@@ -855,17 +855,13 @@ static int collect_cpu_info(int cpu_num, struct cpu_signature *csig)
 {
 	static struct cpu_signature prev;
 	struct cpuinfo_x86 *c = &cpu_data(cpu_num);
-	unsigned int val[2];
 
 	memset(csig, 0, sizeof(*csig));
 
 	csig->sig = cpuid_eax(0x00000001);
 
-	if ((c->x86_model >= 5) || (c->x86 > 6)) {
-		/* get processor flags from MSR 0x17 */
-		rdmsr(MSR_IA32_PLATFORM_ID, val[0], val[1]);
-		csig->pf = 1 << ((val[1] >> 18) & 7);
-	}
+	if (c->has_platform_id)
+		csig->pf = 1 << c->platform_id;
 
 	csig->rev = c->microcode;
 
diff --git a/arch/x86/kernel/cpu/proc.c b/arch/x86/kernel/cpu/proc.c
index 18ca99f2798b..5345d50ed709 100644
--- a/arch/x86/kernel/cpu/proc.c
+++ b/arch/x86/kernel/cpu/proc.c
@@ -76,6 +76,8 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 		seq_puts(m, "stepping\t: unknown\n");
 	if (c->microcode)
 		seq_printf(m, "microcode\t: 0x%x\n", c->microcode);
+	if (c->has_platform_id)
+		seq_printf(m, "platform_id\t: %d\n", c->platform_id);
 
 	if (cpu_has(c, X86_FEATURE_TSC)) {
 		unsigned int freq = cpufreq_quick_get(cpu);