On Mon, Mar 7, 2022 at 12:46 PM zhenwei pi <pizhen...@bytedance.com> wrote:

> After assigning a NVMe/SCSI controller to guest by VFIO, we lose
> everything on the host side. A guest uses these devices exclusively,
> we usually don't care the actions on these devices. But there is a
> low probability that hitting physical hardware warning, we need a
> chance to get the basic smart log info.
>
> Introduce disk smart, and implement NVMe smart on linux.
>
> Thanks to Keith and Marc-André.
>
> CC: Keith Busch <kbu...@kernel.org>
> Signed-off-by: zhenwei pi <pizhen...@bytedance.com>
>

Reviewed-by: Marc-André Lureau <marcandre.lur...@redhat.com>


> ---
>  qga/commands-posix.c | 73 ++++++++++++++++++++++++++++++++++++++++++++
>  qga/qapi-schema.json | 53 +++++++++++++++++++++++++++++++-
>  2 files changed, 125 insertions(+), 1 deletion(-)
>
> diff --git a/qga/commands-posix.c b/qga/commands-posix.c
> index 4ec83bbfbc..383bbebe42 100644
> --- a/qga/commands-posix.c
> +++ b/qga/commands-posix.c
> @@ -27,6 +27,7 @@
>  #include "qemu/base64.h"
>  #include "qemu/cutils.h"
>  #include "commands-common.h"
> +#include "block/nvme.h"
>
>  #ifdef HAVE_UTMPX
>  #include <utmpx.h>
> @@ -49,6 +50,7 @@ extern char **environ;
>  #include <sys/socket.h>
>  #include <net/if.h>
>  #include <sys/statvfs.h>
> +#include <linux/nvme_ioctl.h>
>
>  #ifdef CONFIG_LIBUDEV
>  #include <libudev.h>
> @@ -1390,6 +1392,76 @@ static GuestDiskInfoList *get_disk_partitions(
>      return ret;
>  }
>
> +static void get_nvme_smart(GuestDiskInfo *disk)
> +{
> +    int fd;
> +    GuestNVMeSmart *smart;
> +    NvmeSmartLog log = {0};
> +    struct nvme_admin_cmd cmd = {
> +        .opcode = NVME_ADM_CMD_GET_LOG_PAGE,
> +        .nsid = NVME_NSID_BROADCAST,
> +        .addr = (uint64_t)&log,
> +        .data_len = sizeof(log),
> +        .cdw10 = NVME_LOG_SMART_INFO | (1 << 15) /* RAE bit */
> +                 | (((sizeof(log) >> 2) - 1) << 16)
> +    };
> +
> +    fd = qemu_open_old(disk->name, O_RDONLY);
> +    if (fd == -1) {
> +        g_debug("Failed to open device: %s: %s", disk->name,
> g_strerror(errno));
> +        return;
> +    }
> +
> +    if (ioctl(fd, NVME_IOCTL_ADMIN_CMD, &cmd)) {
> +        g_debug("Failed to get smart: %s: %s", disk->name,
> g_strerror(errno));
> +        close(fd);
> +        return;
> +    }
> +
> +    disk->has_smart = true;
> +    disk->smart = g_new0(GuestDiskSmart, 1);
> +    disk->smart->type = GUEST_DISK_BUS_TYPE_NVME;
> +
> +    smart = &disk->smart->u.nvme;
> +    smart->critical_warning = log.critical_warning;
> +    smart->temperature = lduw_le_p(&log.temperature); /* unaligned field
> */
> +    smart->available_spare = log.available_spare;
> +    smart->available_spare_threshold = log.available_spare_threshold;
> +    smart->percentage_used = log.percentage_used;
> +    smart->data_units_read_lo = le64_to_cpu(log.data_units_read[0]);
> +    smart->data_units_read_hi = le64_to_cpu(log.data_units_read[1]);
> +    smart->data_units_written_lo = le64_to_cpu(log.data_units_written[0]);
> +    smart->data_units_written_hi = le64_to_cpu(log.data_units_written[1]);
> +    smart->host_read_commands_lo = le64_to_cpu(log.host_read_commands[0]);
> +    smart->host_read_commands_hi = le64_to_cpu(log.host_read_commands[1]);
> +    smart->host_write_commands_lo =
> le64_to_cpu(log.host_write_commands[0]);
> +    smart->host_write_commands_hi =
> le64_to_cpu(log.host_write_commands[1]);
> +    smart->controller_busy_time_lo =
> le64_to_cpu(log.controller_busy_time[0]);
> +    smart->controller_busy_time_hi =
> le64_to_cpu(log.controller_busy_time[1]);
> +    smart->power_cycles_lo = le64_to_cpu(log.power_cycles[0]);
> +    smart->power_cycles_hi = le64_to_cpu(log.power_cycles[1]);
> +    smart->power_on_hours_lo = le64_to_cpu(log.power_on_hours[0]);
> +    smart->power_on_hours_hi = le64_to_cpu(log.power_on_hours[1]);
> +    smart->unsafe_shutdowns_lo = le64_to_cpu(log.unsafe_shutdowns[0]);
> +    smart->unsafe_shutdowns_hi = le64_to_cpu(log.unsafe_shutdowns[1]);
> +    smart->media_errors_lo = le64_to_cpu(log.media_errors[0]);
> +    smart->media_errors_hi = le64_to_cpu(log.media_errors[1]);
> +    smart->number_of_error_log_entries_lo =
> +        le64_to_cpu(log.number_of_error_log_entries[0]);
> +    smart->number_of_error_log_entries_hi =
> +        le64_to_cpu(log.number_of_error_log_entries[1]);
> +
> +    close(fd);
> +}
> +
> +static void get_disk_smart(GuestDiskInfo *disk)
> +{
> +    if (disk->has_address
> +        && (disk->address->bus_type == GUEST_DISK_BUS_TYPE_NVME)) {
> +        get_nvme_smart(disk);
> +    }
> +}
> +
>  GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
>  {
>      GuestDiskInfoList *ret = NULL;
> @@ -1463,6 +1535,7 @@ GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
>          }
>
>          get_disk_deps(disk_dir, disk);
> +        get_disk_smart(disk);
>          ret = get_disk_partitions(ret, de->d_name, disk_dir, dev_name);
>      }
>
> diff --git a/qga/qapi-schema.json b/qga/qapi-schema.json
> index 8f73770210..bbfb817347 100644
> --- a/qga/qapi-schema.json
> +++ b/qga/qapi-schema.json
> @@ -888,6 +888,55 @@
>             '*serial': 'str', '*dev': 'str',
>             '*ccw-address': 'GuestCCWAddress'} }
>
> +##
> +# @GuestNVMeSmart:
> +#
> +# NVMe smart informations, based on NVMe specification,
> +# section <SMART / Health Information (Log Identifier 02h)>
> +#
> +# Since: 7.0
> +##
> +{ 'struct': 'GuestNVMeSmart',
> +  'data': {'critical-warning': 'int',
> +           'temperature': 'int',
> +           'available-spare': 'int',
> +           'available-spare-threshold': 'int',
> +           'percentage-used': 'int',
> +           'data-units-read-lo': 'uint64',
> +           'data-units-read-hi': 'uint64',
> +           'data-units-written-lo': 'uint64',
> +           'data-units-written-hi': 'uint64',
> +           'host-read-commands-lo': 'uint64',
> +           'host-read-commands-hi': 'uint64',
> +           'host-write-commands-lo': 'uint64',
> +           'host-write-commands-hi': 'uint64',
> +           'controller-busy-time-lo': 'uint64',
> +           'controller-busy-time-hi': 'uint64',
> +           'power-cycles-lo': 'uint64',
> +           'power-cycles-hi': 'uint64',
> +           'power-on-hours-lo': 'uint64',
> +           'power-on-hours-hi': 'uint64',
> +           'unsafe-shutdowns-lo': 'uint64',
> +           'unsafe-shutdowns-hi': 'uint64',
> +           'media-errors-lo': 'uint64',
> +           'media-errors-hi': 'uint64',
> +           'number-of-error-log-entries-lo': 'uint64',
> +           'number-of-error-log-entries-hi': 'uint64' } }
> +
> +##
> +# @GuestDiskSmart:
> +#
> +# Disk type related smart information.
> +#
> +# - @nvme: NVMe disk smart
> +#
> +# Since: 7.0
> +##
> +{ 'union': 'GuestDiskSmart',
> +  'base': { 'type': 'GuestDiskBusType' },
> +  'discriminator': 'type',
> +  'data': { 'nvme': 'GuestNVMeSmart' } }
> +
>  ##
>  # @GuestDiskInfo:
>  #
> @@ -899,12 +948,14 @@
>  # @address: disk address information (only for non-virtual devices)
>  # @alias: optional alias assigned to the disk, on Linux this is a name
> assigned
>  #         by device mapper
> +# @smart: disk smart information (Since 7.0)
>  #
>  # Since 5.2
>  ##
>  { 'struct': 'GuestDiskInfo',
>    'data': {'name': 'str', 'partition': 'bool', '*dependencies': ['str'],
> -           '*address': 'GuestDiskAddress', '*alias': 'str'} }
> +           '*address': 'GuestDiskAddress', '*alias': 'str',
> +           '*smart': 'GuestDiskSmart'} }
>
>  ##
>  # @guest-get-disks:
> --
> 2.20.1
>
>

-- 
Marc-André Lureau

Reply via email to