Re: [PATCHv2 0/2] log guest name and memory error type AO, AR for MCEs

[Paolo Bonzini]

On 09/10/19 18:44, Mario Smarduch wrote:
> In a large VPC environment we want to log memory error occurrences
> and log them with guest name and type - there are few use cases
> 
> 
> - if VM crashes on AR mce inform the user about the reason and resolve the 
> case
> - if VM hangs notify the user to reboot and resume processing
> - if VM continues to run let the user know, he/she maybe able to correlate
>   to vm internal outage
> - Rawhammer attacks - isolate/determine the attacker possible migrating it off
>   the hypervisor
> - In general track memory errors on a hyperviosr over time to determine trends
> 
> Monitoring our fleet we come across quite a few of these and been
> able to take action where before there were no clues to the causes.
> 
> When memory error occurs we get a log entry in qemu log:
> 
> Guest [Droplet-12345678] 2019-08-02T05:00:11.940270Z qemu-system-x86_64:
> Guest MCE Memory Error at QEMU addr 0x7f3c7622f000 and GUEST 0x78e42f000
> addr of type BUS_MCEERR_AR injected
> 
> with enterprise logging environment we can to take further actions.
> 
> v1 -> v2:
> - split into two patches one to get the gustname second to log MCEs 
> - addressed comments for MCE logging
> 
> Mario Smarduch (2):
>   util/qemu-error: add guest name helper with -msg options
>   target/i386: log MCE guest and host addresses
> 
>  include/qemu/error-report.h |  1 +
>  qemu-options.hx             | 10 ++++++----
>  target/i386/kvm.c           | 29 ++++++++++++++++++++++++-----
>  util/qemu-error.c           | 31 +++++++++++++++++++++++++++++++
>  vl.c                        |  5 +++++
>  5 files changed, 67 insertions(+), 9 deletions(-)
> 

Queued, thanks.

Paolo