From: Maxime Chevallier <maxime.chevallier@bootlin.com>
To: Jijie Shao <shaojijie@huawei.com>
Cc: davem@davemloft.net, "Andrew Lunn" <andrew@lunn.ch>,
"Jakub Kicinski" <kuba@kernel.org>,
"Eric Dumazet" <edumazet@google.com>,
"Paolo Abeni" <pabeni@redhat.com>,
"Russell King" <linux@armlinux.org.uk>,
"Heiner Kallweit" <hkallweit1@gmail.com>,
netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
thomas.petazzoni@bootlin.com,
linux-arm-kernel@lists.infradead.org,
"Christophe Leroy" <christophe.leroy@csgroup.eu>,
"Herve Codina" <herve.codina@bootlin.com>,
"Florian Fainelli" <f.fainelli@gmail.com>,
"Vladimir Oltean" <vladimir.oltean@nxp.com>,
"Köry Maincent" <kory.maincent@bootlin.com>,
"Oleksij Rempel" <o.rempel@pengutronix.de>,
"Simon Horman" <horms@kernel.org>,
"Romain Gantois" <romain.gantois@bootlin.com>,
"shenjian15@huawei.com" <shenjian15@huawei.com>
Subject: Re: [PATCH net-next v5 07/13] net: phy: phy_caps: Allow looking-up link caps based on speed and duplex
Date: Fri, 30 May 2025 09:56:08 +0200 [thread overview]
Message-ID: <20250530095608.63c42399@fedora.home> (raw)
In-Reply-To: <5bbf1201-e803-40ea-a081-c25919f5e89d@huawei.com>
On Thu, 29 May 2025 17:36:11 +0800
Jijie Shao <shaojijie@huawei.com> wrote:
> on 2025/3/8 1:36, Maxime Chevallier wrote:
> > As the link_caps array is efficient for <speed,duplex> lookups,
> > implement a function for speed/duplex lookups that matches a given
> > mask. This replicates to some extent the phy_lookup_settings()
> > behaviour, matching full link_capabilities instead of a single linkmode.
> >
> > phy.c's phy_santize_settings() and phylink's
> > phylink_ethtool_ksettings_set() performs such lookup using the
> > phy_settings table, but are only interested in the actual speed/duplex
> > that were matched, rathet than the individual linkmode.
> >
> > Similar to phy_lookup_settings(), the newly introduced phy_caps_lookup()
> > will run through the link_caps[] array by descending speed/duplex order.
> >
> > If the link_capabilities for a given <speed/duplex> tuple intersects the
> > passed linkmodes, we consider that a match.
> >
> > Similar to phy_lookup_settings(), we also allow passing an 'exact'
> > boolean, allowing non-exact match. Here, we MUST always match the
> > linkmodes mask, but we allow matching on lower speed settings.
> >
> > Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
> > ---
> > drivers/net/phy/phy-caps.h | 4 ++++
> > drivers/net/phy/phy.c | 32 ++++++--------------------
> > drivers/net/phy/phy_caps.c | 47 ++++++++++++++++++++++++++++++++++++++
> > drivers/net/phy/phylink.c | 17 +++++++-------
> > 4 files changed, 67 insertions(+), 33 deletions(-)
> >
> > diff --git a/drivers/net/phy/phy-caps.h b/drivers/net/phy/phy-caps.h
> > index 8833798f141f..aef4b54a8429 100644
> > --- a/drivers/net/phy/phy-caps.h
> > +++ b/drivers/net/phy/phy-caps.h
> > @@ -51,4 +51,8 @@ phy_caps_lookup_by_linkmode(const unsigned long *linkmodes);
> > const struct link_capabilities *
> > phy_caps_lookup_by_linkmode_rev(const unsigned long *linkmodes, bool fdx_only);
> >
> > +const struct link_capabilities *
> > +phy_caps_lookup(int speed, unsigned int duplex, const unsigned long *supported,
> > + bool exact);
> > +
> > #endif /* __PHY_CAPS_H */
> > diff --git a/drivers/net/phy/phy.c b/drivers/net/phy/phy.c
> > index 8df37d221fba..562acde89224 100644
> > --- a/drivers/net/phy/phy.c
> > +++ b/drivers/net/phy/phy.c
> > @@ -213,25 +213,6 @@ int phy_aneg_done(struct phy_device *phydev)
> > }
> > EXPORT_SYMBOL(phy_aneg_done);
> >
> > -/**
> > - * phy_find_valid - find a PHY setting that matches the requested parameters
> > - * @speed: desired speed
> > - * @duplex: desired duplex
> > - * @supported: mask of supported link modes
> > - *
> > - * Locate a supported phy setting that is, in priority order:
> > - * - an exact match for the specified speed and duplex mode
> > - * - a match for the specified speed, or slower speed
> > - * - the slowest supported speed
> > - * Returns the matched phy_setting entry, or %NULL if no supported phy
> > - * settings were found.
> > - */
> > -static const struct phy_setting *
> > -phy_find_valid(int speed, int duplex, unsigned long *supported)
> > -{
> > - return phy_lookup_setting(speed, duplex, supported, false);
> > -}
> > -
> > /**
> > * phy_supported_speeds - return all speeds currently supported by a phy device
> > * @phy: The phy device to return supported speeds of.
> > @@ -274,13 +255,14 @@ EXPORT_SYMBOL(phy_check_valid);
> > */
> > static void phy_sanitize_settings(struct phy_device *phydev)
> > {
> > - const struct phy_setting *setting;
> > + const struct link_capabilities *c;
> > +
> > + c = phy_caps_lookup(phydev->speed, phydev->duplex, phydev->supported,
> > + false);
> >
> > - setting = phy_find_valid(phydev->speed, phydev->duplex,
> > - phydev->supported);
> > - if (setting) {
> > - phydev->speed = setting->speed;
> > - phydev->duplex = setting->duplex;
> > + if (c) {
> > + phydev->speed = c->speed;
> > + phydev->duplex = c->duplex;
> > } else {
> > /* We failed to find anything (no supported speeds?) */
> > phydev->speed = SPEED_UNKNOWN;
> > diff --git a/drivers/net/phy/phy_caps.c b/drivers/net/phy/phy_caps.c
> > index c39f38c12ef2..0366feee2912 100644
> > --- a/drivers/net/phy/phy_caps.c
> > +++ b/drivers/net/phy/phy_caps.c
> > @@ -170,6 +170,53 @@ phy_caps_lookup_by_linkmode_rev(const unsigned long *linkmodes, bool fdx_only)
> > return NULL;
> > }
> >
> > +/**
> > + * phy_caps_lookup() - Lookup capabilities by speed/duplex that matches a mask
> > + * @speed: Speed to match
> > + * @duplex: Duplex to match
> > + * @supported: Mask of linkmodes to match
> > + * @exact: Perform an exact match or not.
> > + *
> > + * Lookup a link_capabilities entry that intersect the supported linkmodes mask,
> > + * and that matches the passed speed and duplex.
> > + *
> > + * When @exact is set, an exact match is performed on speed and duplex, meaning
> > + * that if the linkmodes for the given speed and duplex intersect the supported
> > + * mask, this capability is returned, otherwise we don't have a match and return
> > + * NULL.
> > + *
> > + * When @exact is not set, we return either an exact match, or matching capabilities
> > + * at lower speed, or the lowest matching speed, or NULL.
> > + *
> > + * Returns: a matched link_capabilities according to the above process, NULL
> > + * otherwise.
> > + */
> > +const struct link_capabilities *
> > +phy_caps_lookup(int speed, unsigned int duplex, const unsigned long *supported,
> > + bool exact)
> > +{
> > + const struct link_capabilities *lcap, *last = NULL;
> > +
> > + for_each_link_caps_desc_speed(lcap) {
> > + if (linkmode_intersects(lcap->linkmodes, supported)) {
> > + last = lcap;
> > + /* exact match on speed and duplex*/
> > + if (lcap->speed == speed && lcap->duplex == duplex) {
> > + return lcap;
> > + } else if (!exact) {
> > + if (lcap->speed <= speed)
> > + return lcap;
> > + }
> > + }
> > + }
> > +
> > + if (!exact)
> > + return last;
> > +
> > + return NULL;
> > +}
> > +EXPORT_SYMBOL_GPL(phy_caps_lookup);
> > +
> > /**
> > * phy_caps_linkmode_max_speed() - Clamp a linkmodes set to a max speed
> > * @max_speed: Speed limit for the linkmode set
> > diff --git a/drivers/net/phy/phylink.c b/drivers/net/phy/phylink.c
> > index a3f64b6d2d34..cf9f019382ad 100644
> > --- a/drivers/net/phy/phylink.c
> > +++ b/drivers/net/phy/phylink.c
> > @@ -20,6 +20,7 @@
> > #include <linux/timer.h>
> > #include <linux/workqueue.h>
> >
> > +#include "phy-caps.h"
> > #include "sfp.h"
> > #include "swphy.h"
> >
> > @@ -2852,8 +2853,8 @@ int phylink_ethtool_ksettings_set(struct phylink *pl,
> > const struct ethtool_link_ksettings *kset)
> > {
> > __ETHTOOL_DECLARE_LINK_MODE_MASK(support);
> > + const struct link_capabilities *c;
> > struct phylink_link_state config;
> > - const struct phy_setting *s;
> >
> > ASSERT_RTNL();
> >
> > @@ -2896,23 +2897,23 @@ int phylink_ethtool_ksettings_set(struct phylink *pl,
> > /* Autonegotiation disabled, select a suitable speed and
> > * duplex.
> > */
> > - s = phy_lookup_setting(kset->base.speed, kset->base.duplex,
> > - pl->supported, false);
> > - if (!s)
> > + c = phy_caps_lookup(kset->base.speed, kset->base.duplex,
> > + pl->supported, false);
> > + if (!c)
> > return -EINVAL;
>
>
>
> Hi Maxime, fc81e257d19f ("net: phy: phy_caps: Allow looking-up link caps based on speed and duplex") might have different behavior than the modification.
> My case is set 10M Half with disable autoneg both sides and I expect it is
> link in 10M Half. But now, it is link in 10M Full,which is not what I
> expect.
>
> I used followed command and trace how phy worked.
> ethtool -s eth1 autoneg off speed 10 duplex half
> The log is showed as followed:
> ethtool-13127 [067] 6164.771853: phy_ethtool_ksettings set: (phy_ethtool ksettings set+0x0/0x200) duplex=0 speed=10
> kworker/u322:2-11096 [070] 6164.771853: _phy_start_aneq: ( _phy_start_aneg+0x0/0xb8) duplex=0 speed=10
> kworker/u322:2-11096 [070] 6164.771854: phy_caps_lookup: (phy_caps_lookup+0x0/0xf0) duplex=0 speed=10
> kworker/u322:2-11096 [070] 6164.771855: phy_config_aneg: (phy_config_aneg+0x0/0x70) duplex=1 speed=10
> kworker/u322:2-11096 [070] 6164.771856: genphy_config_aneg: (__genphy_config_aneg+0X0/0X270) duplex=1 speed=10
>
> I also try to fixed it and it works. Do you have any idea about it.
The !exact match logic in the rework is wrong indeed, sorry...
For non-exact matches we return a non-exact match too early without
giving the chance for a potentially exact half-duplex match...
As for the fix, can you try this out :
diff --git a/drivers/net/phy/phy_caps.c b/drivers/net/phy/phy_caps.c
index 703321689726..d80f6a37edf1 100644
--- a/drivers/net/phy/phy_caps.c
+++ b/drivers/net/phy/phy_caps.c
@@ -195,7 +195,7 @@ const struct link_capabilities *
phy_caps_lookup(int speed, unsigned int duplex, const unsigned long *supported,
bool exact)
{
- const struct link_capabilities *lcap, *last = NULL;
+ const struct link_capabilities *lcap, *match = NULL, *last = NULL;
for_each_link_caps_desc_speed(lcap) {
if (linkmode_intersects(lcap->linkmodes, supported)) {
@@ -204,16 +204,19 @@ phy_caps_lookup(int speed, unsigned int duplex, const unsigned long *supported,
if (lcap->speed == speed && lcap->duplex == duplex) {
return lcap;
} else if (!exact) {
- if (lcap->speed <= speed)
- return lcap;
+ if (!match && lcap->speed <= speed)
+ match = lcap;
+
+ if (lcap->speed < speed)
+ break;
}
}
}
- if (!exact)
- return last;
+ if (!match && !exact)
+ match = last;
- return NULL;
+ return match;
}
EXPORT_SYMBOL_GPL(phy_caps_lookup);
next prev parent reply other threads:[~2025-05-30 7:56 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-07 17:35 [PATCH net-next v5 00/13] net: phy: Rework linkmodes handling in a dedicated file Maxime Chevallier
2025-03-07 17:35 ` [PATCH net-next v5 01/13] net: ethtool: Export the link_mode_params definitions Maxime Chevallier
2025-03-07 17:35 ` [PATCH net-next v5 02/13] net: phy: Use an internal, searchable storage for the linkmodes Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 03/13] net: phy: phy_caps: Move phy_speeds to phy_caps Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 04/13] net: phy: phy_caps: Move __set_linkmode_max_speed " Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 05/13] net: phy: phy_caps: Introduce phy_caps_valid Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 06/13] net: phy: phy_caps: Implement link_capabilities lookup by linkmode Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 07/13] net: phy: phy_caps: Allow looking-up link caps based on speed and duplex Maxime Chevallier
2025-05-29 9:36 ` Jijie Shao
2025-05-29 9:40 ` Russell King (Oracle)
2025-05-30 7:56 ` Maxime Chevallier [this message]
2025-06-03 8:25 ` Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 08/13] net: phy: phy_device: Use link_capabilities lookup for PHY aneg config Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 09/13] net: phylink: Use phy_caps_lookup for fixed-link configuration Maxime Chevallier
2025-03-28 1:16 ` Alexander H Duyck
2025-03-28 8:06 ` Maxime Chevallier
2025-03-28 21:03 ` Alexander Duyck
2025-03-28 21:45 ` Andrew Lunn
2025-03-28 23:26 ` Alexander Duyck
2025-03-31 7:35 ` Maxime Chevallier
2025-03-31 14:17 ` Andrew Lunn
2025-03-31 14:54 ` Russell King (Oracle)
2025-03-31 16:20 ` Maxime Chevallier
2025-03-31 16:38 ` Alexander Duyck
2025-04-01 7:41 ` Maxime Chevallier
2025-03-31 22:31 ` Alexander H Duyck
2025-04-01 8:33 ` Maxime Chevallier
2025-03-31 7:14 ` Maxime Chevallier
2025-04-01 15:28 ` Alexander H Duyck
2025-04-01 15:40 ` Maxime Chevallier
2025-04-01 16:14 ` Russell King (Oracle)
2025-04-02 6:47 ` Maxime Chevallier
2025-03-31 12:50 ` Russell King (Oracle)
2025-03-31 22:19 ` Alexander Duyck
2025-03-07 17:36 ` [PATCH net-next v5 10/13] net: phy: drop phy_settings and the associated lookup helpers Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 11/13] net: phylink: Add a mapping between MAC_CAPS and LINK_CAPS Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 12/13] net: phylink: Convert capabilities to linkmodes using phy_caps Maxime Chevallier
2025-03-07 17:36 ` [PATCH net-next v5 13/13] net: phylink: Use phy_caps to get an interface's capabilities and modes Maxime Chevallier
2025-03-13 9:13 ` [PATCH net-next v5 00/13] net: phy: Rework linkmodes handling in a dedicated file Paolo Abeni
2025-03-18 8:10 ` patchwork-bot+netdevbpf
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=20250530095608.63c42399@fedora.home \
--to=maxime.chevallier@bootlin.com \
--cc=andrew@lunn.ch \
--cc=christophe.leroy@csgroup.eu \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=f.fainelli@gmail.com \
--cc=herve.codina@bootlin.com \
--cc=hkallweit1@gmail.com \
--cc=horms@kernel.org \
--cc=kory.maincent@bootlin.com \
--cc=kuba@kernel.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux@armlinux.org.uk \
--cc=netdev@vger.kernel.org \
--cc=o.rempel@pengutronix.de \
--cc=pabeni@redhat.com \
--cc=romain.gantois@bootlin.com \
--cc=shaojijie@huawei.com \
--cc=shenjian15@huawei.com \
--cc=thomas.petazzoni@bootlin.com \
--cc=vladimir.oltean@nxp.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).