Re: [PATCH] Print a user-friendly message on failed vmentry

[Avi Kivity <avi@redhat.com>]

On 05/20/2010 05:46 PM, Mohammed Gamal wrote:
> On Thu, May 20, 2010 at 5:37 PM, Chris Lalancette<clalance@redhat.com>  wrote:
>    
>> On 05/19/2010 05:16 PM, Mohammed Gamal wrote:
>>      
>>> This patch address bug report in https://bugs.launchpad.net/qemu/+bug/530077.
>>>
>>> Failed vmentries were handled with handle_unhandled() which prints a rather
>>> unfriendly message to the user. This patch separates handling vmentry failures
>>> from unknown exit reasons and prints a friendly message to the user.
>>>
>>> Signed-off-by: Mohammed Gamal<m.gamal005@gmail.com>
>>> ---
>>>   qemu-kvm.c |   16 +++++++++++++++-
>>>   1 files changed, 15 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/qemu-kvm.c b/qemu-kvm.c
>>> index 35a4c8a..deb4df8 100644
>>> --- a/qemu-kvm.c
>>> +++ b/qemu-kvm.c
>>> @@ -106,6 +106,20 @@ static int handle_unhandled(uint64_t reason)
>>>       return -EINVAL;
>>>   }
>>>
>>> +static int handle_failed_vmentry(uint64_t reason)
>>> +{
>>> +    fprintf(stderr, "kvm: vm entry failed with error 0x%" PRIx64 "\n\n", reason);
>>> +    fprintf(stderr, "If you're runnning a guest on an Intel machine, it can be\n");
>>> +    fprintf(stderr, "most-likely due to the guest going into an invalid state\n");
>>> +    fprintf(stderr, "for Intel VT. For example, the guest maybe running in big\n");
>>> +    fprintf(stderr, "real mode which is not supported by Intel VT.\n\n");
>>> +    fprintf(stderr, "You may want to try enabling real mode emulation in KVM.\n");
>>> +    fprintf(stderr, "To Enable it, you may run the following commands as root:\n");
>>> +    fprintf(stderr, "# rmmod kvm_intel\n");
>>> +    fprintf(stderr, "# rmmod kvm\n");
>>> +    fprintf(stderr, "# modprobe kvm_intel emulate_invalid_guest_state=1\n");
>>> +    return -EINVAL;
>>> +}
>>>        
>> The thing is, there are other valid reasons for vmentry failure.  A while ago I tracked
>> down a bug in the Linux kernel that was causing us to vmenter with invalid segments;
>> this message would have been very misleading in that case.  I think you'd have to do
>> more complete analysis of the vmentry failure code to be more certain about the reason
>> for failure.
>>
>>      
> Your point is definitely valid, yet big real mode is usually the most
> likely case, and that's why this message is shown. Note also that it
> says it's _most likely_ a failure caused by an invalid guest state,
> but it doesn't rule out all other reasons. And in any case, it'd be
> better than just printing something along the lines of:
> " kvm: unhandled exit 80000021
>    kvm_run returned -22"
>    

However, we're still giving bad advice.  Currently 
emulate_invalid_guest_state=1 is not going to work well (right?).  Once 
it does, we'll simply make it the default and the message will never appear.

-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.