[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Xen-devel] [PATCH 10/19] x86/mce: always write 0 to MSR_IA32_MCG_STATUS on Intel CPU
On 02/17/2017 10:13 AM, Jan Beulich wrote:
On 17.02.17 at 16:01, <boris.ostrovsky@xxxxxxxxxx> wrote:
On 02/17/2017 05:26 AM, Jan Beulich wrote:
On 17.02.17 at 07:39, <haozhong.zhang@xxxxxxxxx> wrote:
--- a/xen/arch/x86/cpu/mcheck/mce.c
+++ b/xen/arch/x86/cpu/mcheck/mce.c
@@ -538,7 +538,14 @@ void mcheck_cmn_handler(const struct cpu_user_regs *regs)
gstatus = mca_rdmsr(MSR_IA32_MCG_STATUS);
if ((gstatus & MCG_STATUS_MCIP) != 0) {
mce_printk(MCE_CRITICAL, "MCE: Clear MCIP@ last step");
- mca_wrmsr(MSR_IA32_MCG_STATUS, gstatus & ~MCG_STATUS_MCIP);
+ if ( boot_cpu_data.x86_vendor == X86_VENDOR_INTEL )
+ /*
+ * Intel SDM 3: An attempt to write to IA32_MCG_STATUS
+ * with any value other than 0 would result in #GP.
+ */
+ mca_wrmsr(MSR_IA32_MCG_STATUS, 0);
+ else
+ mca_wrmsr(MSR_IA32_MCG_STATUS, gstatus & ~MCG_STATUS_MCIP);
I think this wants to be the other way around: Write zero in the
common case, and make an exception for AMD (and that only if
on AMD it's really useful to write other than plain zero). Boris,
Suravee, can you provide some input here please?
I don't see anything stating that writing non-zero values can result in
an exception of any sort. But I am only going by what I see in the
documentation.
But the question was the other way around: Is there a need to
clear only the MCIP bit, or can we as well write plain zero to the
MSR (allowing the conditional to be dropped).
I don't see any indication of problems writing any of the three bits
with any value so assuming the other two bits have been handled by now
(as they were before this series) I don't see any issues.
But I think Suravee would be a better person to answer this, perhaps
after checking with HW people.
-boris
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel
|
