From: Eran Ben Elisha <eranbe@nvidia.com>
To: <netdev@vger.kernel.org>
Cc: <jiri@resnulli.us>, <saeedm@nvidia.com>,
<andrew.gospodarek@broadcom.com>, <jacob.e.keller@intel.com>,
<guglielmo.morandin@broadcom.com>, <eugenem@fb.com>,
<eranbe@mellanox.com>, Jakub Kicinski <kuba@kernel.org>
Subject: Re: [RFC net-next v2 2/3] devlink: health: add remediation type
Date: Thu, 11 Mar 2021 16:32:44 +0200 [thread overview]
Message-ID: <53f182d0-e1f6-5e18-ac04-ff7f6ec56af8@nvidia.com> (raw)
In-Reply-To: <20210311032613.1533100-2-kuba@kernel.org>
On 3/11/2021 5:26 AM, Jakub Kicinski wrote:
> Currently devlink health does not give user any clear information
> of what kind of remediation ->recover callback will perform. This
> makes it difficult to understand the impact of enabling auto-
> -remediation, and the severity of the error itself.
>
> To allow users to make more informed decision add a new remediation
> type attribute.
>
> Note that we only allow one remediation type per reporter, this
> is intentional. devlink health is not built for mixing issues
> of different severity into one reporter since it only maintains
> one dump, of the first event and a single error counter.
> Nudging vendors towards categorizing issues beyond coarse
> groups is an added bonus.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> include/net/devlink.h | 2 ++
> include/uapi/linux/devlink.h | 25 +++++++++++++++++++++++++
> net/core/devlink.c | 7 ++++++-
> 3 files changed, 33 insertions(+), 1 deletion(-)
>
> diff --git a/include/net/devlink.h b/include/net/devlink.h
> index b424328af658..72b37769761f 100644
> --- a/include/net/devlink.h
> +++ b/include/net/devlink.h
> @@ -659,6 +659,7 @@ struct devlink_health_reporter;
> /**
> * struct devlink_health_reporter_ops - Reporter operations
> * @name: reporter name
> + * remedy: severity of the remediation required
> * @recover: callback to recover from reported error
> * if priv_ctx is NULL, run a full recover
> * @dump: callback to dump an object
> @@ -669,6 +670,7 @@ struct devlink_health_reporter;
>
> struct devlink_health_reporter_ops {
> char *name;
> + enum devlink_health_remedy remedy;
> int (*recover)(struct devlink_health_reporter *reporter,
> void *priv_ctx, struct netlink_ext_ack *extack);
> int (*dump)(struct devlink_health_reporter *reporter,
> diff --git a/include/uapi/linux/devlink.h b/include/uapi/linux/devlink.h
> index 41a6ea3b2256..8cd1508b525b 100644
> --- a/include/uapi/linux/devlink.h
> +++ b/include/uapi/linux/devlink.h
> @@ -534,6 +534,9 @@ enum devlink_attr {
> DEVLINK_ATTR_RELOAD_ACTION_STATS, /* nested */
>
> DEVLINK_ATTR_PORT_PCI_SF_NUMBER, /* u32 */
> +
> + DEVLINK_ATTR_HEALTH_REPORTER_REMEDY, /* u32 */
> +
> /* add new attributes above here, update the policy in devlink.c */
>
> __DEVLINK_ATTR_MAX,
> @@ -620,4 +623,26 @@ enum devlink_health_state {
> DL_HEALTH_STATE_ERROR,
> };
>
> +/**
> + * enum devlink_health_reporter_remedy - severity of remediation procedure
> + * @DL_HEALTH_REMEDY_NONE: transient error, no remediation required
> + * @DL_HEALTH_REMEDY_KICK: device stalled, processing will be re-triggered
> + * @DL_HEALTH_REMEDY_COMP_RESET: associated device component (e.g. device queue)
> + * will be reset
> + * @DL_HEALTH_REMEDY_RESET: full device reset, will result in temporary
> + * unavailability of the device, device configuration
> + * should not be lost
> + * @DL_HEALTH_REMEDY_REINIT: device will be reinitialized and configuration lost
> + *
> + * Used in %DEVLINK_ATTR_HEALTH_REPORTER_REMEDY, categorizes the health reporter
> + * by the severity of the remediation.
> + */
> +enum devlink_health_remedy {
> + DL_HEALTH_REMEDY_NONE = 1,
What is the reason zero is skipped?
> + DL_HEALTH_REMEDY_KICK,
> + DL_HEALTH_REMEDY_COMP_RESET,
> + DL_HEALTH_REMEDY_RESET,
> + DL_HEALTH_REMEDY_REINIT,
> +};
> +
> #endif /* _UAPI_LINUX_DEVLINK_H_ */
> diff --git a/net/core/devlink.c b/net/core/devlink.c
> index 8e4e4bd7bb36..09d77d43ff63 100644
> --- a/net/core/devlink.c
> +++ b/net/core/devlink.c
> @@ -6095,7 +6095,8 @@ __devlink_health_reporter_create(struct devlink *devlink,
> {
> struct devlink_health_reporter *reporter;
>
> - if (WARN_ON(graceful_period && !ops->recover))
> + if (WARN_ON(graceful_period && !ops->recover) ||
> + WARN_ON(ops->recover && !ops->remedy))
It allows drivers to set recover callback and report DL_HEALTH_REMEDY_NONE.
Defining DL_HEALTH_REMEDY_NONE = 0 would make this if clause to catch it.
> return ERR_PTR(-EINVAL);
>
> reporter = kzalloc(sizeof(*reporter), GFP_KERNEL);
> @@ -6265,6 +6266,10 @@ devlink_nl_health_reporter_fill(struct sk_buff *msg,
> if (nla_put_string(msg, DEVLINK_ATTR_HEALTH_REPORTER_NAME,
> reporter->ops->name))
> goto reporter_nest_cancel;
> + if (reporter->ops->remedy &&
> + nla_put_u32(msg, DEVLINK_ATTR_HEALTH_REPORTER_REMEDY,
> + reporter->ops->remedy))
> + goto reporter_nest_cancel;
> if (nla_put_u8(msg, DEVLINK_ATTR_HEALTH_REPORTER_STATE,
> reporter->health_state))
> goto reporter_nest_cancel;
>
next prev parent reply other threads:[~2021-03-11 14:34 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-03-11 3:26 [RFC net-next v2 1/3] devlink: move health state to uAPI Jakub Kicinski
2021-03-11 3:26 ` [RFC net-next v2 2/3] devlink: health: add remediation type Jakub Kicinski
2021-03-11 7:48 ` Jiri Pirko
2021-03-11 14:32 ` Eran Ben Elisha [this message]
2021-03-11 16:45 ` Jakub Kicinski
2021-03-11 3:26 ` [RFC net-next v2 3/3] devlink: add more failure modes Jakub Kicinski
2021-03-11 14:23 ` Eran Ben Elisha
2021-03-11 16:49 ` Jakub Kicinski
2021-03-14 12:33 ` Eran Ben Elisha
2021-03-15 17:06 ` Jakub Kicinski
2021-03-11 7:47 ` [RFC net-next v2 1/3] devlink: move health state to uAPI Jiri Pirko
2021-03-11 16:46 ` Jakub Kicinski
2021-03-12 19:56 ` Keller, Jacob E
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=53f182d0-e1f6-5e18-ac04-ff7f6ec56af8@nvidia.com \
--to=eranbe@nvidia.com \
--cc=andrew.gospodarek@broadcom.com \
--cc=eranbe@mellanox.com \
--cc=eugenem@fb.com \
--cc=guglielmo.morandin@broadcom.com \
--cc=jacob.e.keller@intel.com \
--cc=jiri@resnulli.us \
--cc=kuba@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=saeedm@nvidia.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).