On Wed, Jun 14, 2017 at 12:55:51AM +0200, Martin Wilck wrote:
> The logic applied here should match the logic in select_features().
> If no_path_retry is anything but "undef", queue_if_no_path can be
> removed from the feature string, assemble_map() will infer it
> correctly.
> The case where no_path_retry is undefined and "queue_if_no_path"
> is set is treated as if "no_path_retry queue" had been set.
>
> Signed-off-by: Martin Wilck <mwilck@xxxxxxxx>
> ---
> libmultipath/config.c | 20 ++++++++++++++++----
> 1 file changed, 16 insertions(+), 4 deletions(-)
>
> diff --git a/libmultipath/config.c b/libmultipath/config.c
> index 61bbba91..b928fbe7 100644
> --- a/libmultipath/config.c
> +++ b/libmultipath/config.c
> @@ -355,12 +355,24 @@ merge_hwe (struct hwentry * dst, struct hwentry * src)
>
> /*
> * Make sure features is consistent with
> - * no_path_retry
> + * no_path_retry.
> + * The logic should be consistent with select_features().
> + * The actual queue_if_no_path feature is set in assemble_map().
> */
> - if (dst->no_path_retry == NO_PATH_RETRY_FAIL)
> + if (dst->no_path_retry == NO_PATH_RETRY_UNDEF &&
> + strstr(dst->features, "queue_if_no_path")) {
> + condlog(3, "%s/%s: 'queue_if_no_path' is set, assuming no_path_retry='queue'",
> + dst->vendor, dst->product);
> + dst->no_path_retry = NO_PATH_RETRY_QUEUE;
> + }
> + else if (dst->no_path_retry != NO_PATH_RETRY_UNDEF) {
> + condlog(3, "%s/%s: 'no_path_retry' is set, ignoring 'queue_if_no_path'",
> + dst->vendor, dst->product);
> remove_feature(&dst->features, "queue_if_no_path");
This has the same nit as [PATCH 3/7], and it's just as unimportant to
fix here as there.
-Ben
> - else if (dst->no_path_retry != NO_PATH_RETRY_UNDEF)
> - add_feature(&dst->features, "queue_if_no_path");
> + }
> +
> + if (dst->retain_hwhandler != RETAIN_HWHANDLER_UNDEF)
> + remove_feature(&dst->features, "retain_attached_hw_handler");
>
> return 0;
> }
> --
> 2.13.0
--
dm-devel mailing list
dm-devel@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/dm-devel
