All of lore.kernel.org
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Christian Marangi" <ansuelsmth@gmail.com>, devicetree@vger.kernel.org
Cc: robh@kernel.org, conor+dt@kernel.org
Subject: Re: [PATCH net-next v22 08/15] net: phylink: support PCS provider release
Date: Wed, 18 Nov 2026 12:42:50 +0000	[thread overview]
Message-ID: <sashiko-outbox-220928@kernel.org> (raw)
In-Reply-To: <20261117123956.11981-9-ansuelsmth@gmail.com>

Thank you for your contribution! Sashiko AI review found 4 potential
issues to consider:
- [Medium] net: phylink: broken capability recalculation in
  phylink_update_capabilities()
- [Medium] net: phylink: mishandled force_major_config with
  phylink_disable_state
- [Medium] net: phylink: race between pcs_provider_notify() and
  phylink_create()
- [Medium] net: phylink: missing hardware advertisement update in
  pcs_provider_notify()
--

commit 43bb91371f247055f063dc4ec828173d12f85adf
Author: Christian Marangi <ansuelsmth@gmail.com>

net: phylink: support PCS provider release

This commit adds a blocking notifier chain for fwnode PCS provider removal
and updates phylink to detach removed PCS instances, recalculate supported
interfaces and capabilities, and trigger link resolution.

> diff --git a/drivers/net/phy/phylink.c b/drivers/net/phy/phylink.c
> index 921c227a5b3c0..094a15f3dbcd0 100644
> --- a/drivers/net/phy/phylink.c
> +++ b/drivers/net/phy/phylink.c

[ ... ]

> @@ -2350,6 +2371,11 @@ static void phylink_update_capabilities(struct phylink *pl,
>  							config->interface);
>  			linkmode_copy(pl->supported, supported);
>  			linkmode_copy(pl->advertising, config->advertising);
> +		} else {
> +			/* No SFP modes remain usable with the current interfaces. */
> +			linkmode_zero(pl->supported);
> +			linkmode_zero(pl->advertising);
> +			linkmode_zero(config->advertising);
>  		}
>  
>  		goto exit;

[Severity: Medium]
Can this else branch ever be reached when a PCS is removed, and what
happens to the non-SFP branches in phylink_update_capabilities() when
validation fails?

First, phylink_update_capabilities() passes unmasked pl->sfp_interfaces to
__phylink_validate_mask(), and in phylink_validate_mac_and_pcs() a missing
PCS in pl->pcs_list leaves pcs as NULL without returning -EINVAL:

drivers/net/phy/phylink.c:phylink_validate_mac_and_pcs() {
    ...
	} else if (test_bit(state->interface, pl->config->pcs_interfaces)) {
		struct phylink_pcs *tmp;

		list_for_each_entry(tmp, &pl->pcs_list, list) {
			if (!phylink_validate_pcs_interface(tmp, state->interface)) {
				pcs = tmp;
				break;
			}
		}
	}

	if (pcs) {
    ...
}

Because pcs is NULL, phylink_validate_mac_and_pcs() skips PCS validation
and succeeds against the MAC capabilities alone, so
__phylink_validate_mask() still returns 0 after the PCS is removed.

Second, when validation does return -EINVAL (for example, when
state->interface is no longer in pl->supported_interfaces), none of the
other branches in phylink_update_capabilities() clear the capability masks:
the generic fallback presets pl->supported and config->advertising with
linkmode_fill(), ignores the -EINVAL return from __phylink_validate(), and
leaves all linkmode bits set in pl->supported and pl->advertising;
similarly, the pl->phydev, MLO_AN_FIXED, and MLO_AN_INBAND paths leave
stale or unvalidated linkmodes in pl->supported and pl->advertising.

> @@ -2425,6 +2451,133 @@ void phylink_update_mac_pause_capabilities(struct phylink *pl, unsigned long mac
>  }
>  EXPORT_SYMBOL_GPL(phylink_update_mac_pause_capabilities);
>  
> +static void phylink_del_pcs(struct phylink *pl, struct phylink_pcs *pcs)
> +{
> +	ASSERT_RTNL();
> +
> +	lockdep_assert_held(&pl->state_mutex);
> +
> +	rcu_assign_pointer(pcs->phylink, NULL);
> +	synchronize_rcu();
> +	list_del(&pcs->list);
> +
> +	/*
> +	 * Check if we are removing the PCS currently
> +	 * in use by this phylink instance. If this is the case,
> +	 * tear down the link, force phylink resolve to reconfigure the
> +	 * interface mode, disable the current PCS and set the
> +	 * phylink PCS to NULL.
> +	 */
> +	if (pl->pcs == pcs) {
> +		if (pl->old_link_state) {
> +			phylink_link_down(pl);
> +			pl->old_link_state = false;
> +		}
> +		if (pl->cfg_link_an_mode == MLO_AN_INBAND)
> +			timer_delete_sync(&pl->link_poll);
> +
> +		if (pl->pcs_state != PCS_STATE_DOWN)
> +			phylink_pcs_disable(pl->pcs);
> +
> +		if (!pl->phylink_disable_state)
> +			pl->force_major_config = true;
> +
> +		WRITE_ONCE(pl->pcs, NULL);
> +	}
> +}

[Severity: Medium]
Does setting pl->force_major_config based on !pl->phylink_disable_state
interact safely with phylink_stop() and PHYLINK_DISABLE_LINK?

If pl->phylink_disable_state is 0 when phylink_del_pcs() sets
pl->force_major_config = true, and phylink_stop() runs after
pcs_provider_notify() drops rtnl_lock() but before pl->resolve executes,
phylink_stop() calls timer_delete_sync(&pl->link_poll) and then flushes
pl->resolve via phylink_run_resolve_and_disable().  In phylink_resolve(),
pl->force_major_config triggers phylink_major_config() even when
pl->phylink_disable_state is non-zero, reconfiguring stopped hardware and
potentially re-arming pl->link_poll via phylink_pcs_poll_start() after
phylink_stop() already deleted it.

Conversely, if the active PCS is removed while the instance is started
(pl->pcs_state == PCS_STATE_STARTED) but PHYLINK_DISABLE_LINK is set (such
as when SFP LOS is asserted), !pl->phylink_disable_state is false, so
pl->pcs is set to NULL without setting pl->force_major_config.  When
PHYLINK_DISABLE_LINK is later cleared on the same interface,
phylink_resolve() skips phylink_major_config() and calls phylink_link_up()
with pl->pcs == NULL.

> +
> +static int pcs_provider_notify(struct notifier_block *self,
> +			       unsigned long val, void *data)
> +{
> +	struct phylink *pl = container_of(self, struct phylink, fwnode_pcs_nb);
> +	struct fwnode_pcs_provider *pp = data;
> +	struct phylink_pcs *pcs, *tmp;
> +	struct phy_device *phydev;
> +	bool resolve = false;
> +
> +	rtnl_lock();
> +	mutex_lock(&pl->phydev_mutex);
> +	phydev = pl->phydev;
> +	if (phydev)
> +		mutex_lock(&phydev->lock);
> +	mutex_lock(&pl->state_mutex);
> +
> +	/* phylink instance is being destroyed, ignore any event */
> +	if (pl->pcs_state == PCS_STATE_DETACHING)
> +		goto exit;
> +
> +	/*
> +	 * Loop all the PCS for phylink instance and check if
> +	 * this notification is relevant for some of them.
> +	 */
> +	list_for_each_entry_safe(pcs, tmp, &pl->pcs_list, list) {
> +		if (!fwnode_pcs_matches_provider(pp, pl->fwnode, pcs))
> +			continue;
> +
> +		phylink_del_pcs(pl, pcs);
> +		resolve = true;
> +	}
> +
> +	/* Exit early if nothing has changed */
> +	if (!resolve)
> +		goto exit;
> +
> +	/* Refresh supported interfaces */
> +	phy_interface_copy(pl->supported_interfaces,
> +			   pl->config->supported_interfaces);
> +	list_for_each_entry(pcs, &pl->pcs_list, list)
> +		phy_interface_or(pl->supported_interfaces,
> +				 pl->supported_interfaces,
> +				 pcs->supported_interfaces);
> +
> +	/* Recalculate capabilities */
> +	phylink_update_capabilities(pl, &pl->link_config);
> +
> +exit:
> +	mutex_unlock(&pl->state_mutex);
> +	if (phydev)
> +		mutex_unlock(&phydev->lock);
> +	mutex_unlock(&pl->phydev_mutex);
> +	rtnl_unlock();
> +
> +	if (resolve)
> +		phylink_run_resolve(pl);
> +
> +	return resolve ? NOTIFY_OK : NOTIFY_DONE;
> +}

[Severity: Medium]
When phylink_update_capabilities() restricts phydev->advertising or
pl->link_config.advertising after a PCS is removed, what updates the
hardware advertisement?

phylink_phy_update_capabilities() copies the restricted mask into
phy->advertising in memory, but once the PHY state machine has moved past
PHY_UP to PHY_NOLINK or PHY_RUNNING, phylib will not reprogram the PHY
autonegotiation registers unless _phy_start_aneg() is called.  Similarly,
if an unused PCS is removed in in-band mode without setting
pl->force_major_config, phylink_resolve() does not call
phylink_change_inband_advert().  An attached PHY can then continue
advertising and negotiating link modes that require the removed PCS,
causing phylink_major_config() to fail when the link comes up.

[ ... ]

> @@ -2515,10 +2668,18 @@ struct phylink *phylink_create(struct phylink_config *config,
>  	__set_bit(PHYLINK_DISABLE_STOPPED, &pl->phylink_disable_state);
>  	timer_setup(&pl->link_poll, phylink_fixed_poll, 0);
>  
> +	/* First register notifier for hotplug PCS events */
> +	if (!phy_interface_empty(config->pcs_interfaces)) {
> +		pl->fwnode_pcs_nb.notifier_call = pcs_provider_notify;
> +		register_fwnode_pcs_notifier(&pl->fwnode_pcs_nb);
> +	}
> +
>  	/* Fill the PCS list with available PCS from phylink config */
>  	ret = phylink_fill_available_pcs(pl, config);
>  	if (ret < 0)
> -		goto free_pl;
> +		goto unregister_pcs_notify;
> +
> +	mutex_lock(&pl->state_mutex);
>  
>  	/* Link available PCS to phylink */
>  	list_for_each_entry(pcs, &pl->pcs_list, list)
> @@ -2533,13 +2694,13 @@ struct phylink *phylink_create(struct phylink_config *config,
>  				 pl->supported_interfaces,
>  				 pcs->supported_interfaces);
>  
> -	mutex_lock(&pl->state_mutex);
>  	phylink_update_capabilities(pl, &pl->link_config);
> +
>  	mutex_unlock(&pl->state_mutex);
>  
>  	ret = phylink_parse_mode(pl, fwnode);
>  	if (ret < 0)
> -		goto unlink_pcs_list;
> +		goto unregister_pcs_notify;

[Severity: Medium]
Can pcs_provider_notify() race with phylink_create() before pl
initialization is finished?

phylink_create() is called without rtnl_lock() held and registers
pl->fwnode_pcs_nb before pl->pcs_list, pl->supported_interfaces,
pl->cfg_link_an_mode, and pl->link_config are initialized, dropping
pl->state_mutex between steps:

CPU0                                  CPU1
phylink_create()
  register_fwnode_pcs_notifier()
  phylink_fill_available_pcs()
    mutex_lock(&pl->state_mutex)
    ...
    mutex_unlock(&pl->state_mutex)
  mutex_lock(&pl->state_mutex)
  ...
  mutex_unlock(&pl->state_mutex)
  phylink_parse_mode()
    pl->cfg_link_an_mode = ...
                                      fwnode_pcs_del_provider()
                                        pcs_provider_notify()
                                          mutex_lock(&pl->state_mutex)
                                          phylink_update_capabilities()
  phylink_parse_fixedlink()
    pl->link_config.speed = ...
    __set_bit(..., lp_advertising)        <-- concurrent access to link_config

If pcs_provider_notify() runs after phylink_fill_available_pcs() unlocks
pl->state_mutex, it can also observe entries in pl->pcs_list before their
pcs->phylink pointer or pl->supported_interfaces is initialized.
Registering the notifier after pl is initialized (or keeping initialization
under pl->state_mutex before publishing the notifier) would avoid these
races.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117123956.11981-1-ansuelsmth@gmail.com?part=8

  reply	other threads:[~2026-11-18 12:42 UTC|newest]

Thread overview: 26+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-17 12:39 [PATCH net-next v22 00/15] net: pcs: Introduce support for fwnode PCS Christian Marangi
2026-11-17 12:39 ` [PATCH net-next v22 01/15] net: phylink: make PCS phylink instance access RCU protected Christian Marangi
2026-11-18 12:42   ` sashiko-bot
2026-11-17 12:39 ` [PATCH net-next v22 02/15] net: phylink: keep and use MAC supported_interfaces in phylink struct Christian Marangi
2026-11-17 12:39 ` [PATCH net-next v22 03/15] net: phylink: introduce internal phylink PCS handling Christian Marangi
2026-11-18 12:42   ` sashiko-bot
2026-11-17 12:39 ` [PATCH net-next v22 04/15] net: pcs: implement Firmware node support for PCS driver Christian Marangi
2026-11-18 12:42   ` sashiko-bot
2026-11-17 12:39 ` [PATCH net-next v22 05/15] net: phylink: save phylink instance fwnode on phylink_create Christian Marangi
2026-11-17 12:39 ` [PATCH net-next v22 06/15] net: phylink: make phylink_validate state locked Christian Marangi
2026-11-17 12:39 ` [PATCH net-next v22 07/15] net: phylink: add capabilities helpers and track modified advertising Christian Marangi
2026-11-18 12:42   ` sashiko-bot
2026-11-17 12:39 ` [PATCH net-next v22 08/15] net: phylink: support PCS provider release Christian Marangi
2026-11-18 12:42   ` sashiko-bot [this message]
2026-11-17 12:39 ` [PATCH net-next v22 09/15] net: phylink: support late PCS provider attach Christian Marangi
2026-11-18 12:42   ` sashiko-bot
2026-11-17 12:39 ` [PATCH net-next v22 10/15] net: Document PCS subsystem Christian Marangi
2026-11-17 12:39 ` [PATCH net-next v22 11/15] MAINTAINERS: add myself as PCS subsystem maintainer Christian Marangi
2026-11-17 12:39 ` [PATCH net-next v22 12/15] net: phylink: add .pcs_link_down PCS OP Christian Marangi
2026-11-17 12:39 ` [PATCH net-next v22 13/15] dt-bindings: net: pcs: Document support for Airoha Ethernet PCS Christian Marangi
2026-11-18 12:42   ` sashiko-bot
2026-11-17 12:39 ` [PATCH net-next v22 14/15] net: pcs: airoha: add PCS driver for Airoha AN7581 SoC Christian Marangi
2026-11-18 12:42   ` sashiko-bot
2026-11-17 12:39 ` [PATCH net-next v22 15/15] net: airoha: add phylink support Christian Marangi
2026-11-17 12:43 ` [PATCH net-next v22 00/15] net: pcs: Introduce support for fwnode PCS netdev-bot+sinfo
2026-11-17 12:47   ` Christian Marangi

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=sashiko-outbox-220928@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=ansuelsmth@gmail.com \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=robh@kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.