Skip to contentSkip to conversation

phy: fsl-lynx-28g: propagate PCVT enable errors

PATCH In review [email protected] 4 participants 4 messages Jul 22 – Jul 23

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 phy: fsl-lynx-28g: propagate PCVT enable errors

From: surendra <[email protected]>

lynx_28g_set_mode() currently ignores failures from lynx_28g_lane_enable_pcvt(). It then updates the lane mode and reports success even though the protocol converter may remain disabled.

Propagate the error and leave the previous lane mode intact so the caller can handle the failed reconfiguration.

Fixes 8f73b37cf3fb ("phy: add support for the Layerscape SerDes 28G")
---
 drivers/phy/freescale/phy-fsl-lynx-28g.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/phy/freescale/phy-fsl-lynx-28g.c +3 -1
index 38afcd081a2a..1fe406adc83c 100644
@@ -1003,7 +1003,9 @@ static int lynx_28g_set_mode(struct phy *phy, enum phy_mode mode, int submode)
 
 	lynx_28g_lane_change_proto_conf(lane, lane_mode);
 	lynx_28g_lane_remap_pll(lane, lane_mode);
-	WARN_ON(lynx_28g_lane_enable_pcvt(lane, lane_mode));
+	err = lynx_28g_lane_enable_pcvt(lane, lane_mode);
+	if (err)
+		goto out;
 
 	lane->mode = lane_mode;
 
Signature · 1 line
2.55.0
maintainer
#1 lore raw

Hi Surendra,

Quoted 31 lines · On Wed, Jul 22, 2026 at 09:06:19AM +0530, [email protected] wrote:

From: surendra <[email protected]>

lynx_28g_set_mode() currently ignores failures from lynx_28g_lane_enable_pcvt(). It then updates the lane mode and reports success even though the protocol converter may remain disabled.

Propagate the error and leave the previous lane mode intact so the caller can handle the failed reconfiguration.

Fixes 8f73b37cf3fb ("phy: add support for the Layerscape SerDes 28G")
---
 drivers/phy/freescale/phy-fsl-lynx-28g.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/phy/freescale/phy-fsl-lynx-28g.c +3 -2
index 38afcd081a2a..1fe406adc83c 100644
@@ -1003,7 +1003,9 @@ static int lynx_28g_set_mode(struct phy *phy, enum phy_mode mode, int submode)
 
        lynx_28g_lane_change_proto_conf(lane, lane_mode);
        lynx_28g_lane_remap_pll(lane, lane_mode);
-       WARN_ON(lynx_28g_lane_enable_pcvt(lane, lane_mode));
+       err = lynx_28g_lane_enable_pcvt(lane, lane_mode);
+       if (err)
+               goto out;
 
        lane->mode = lane_mode;
 
--
Quoted 2 lines

2.55.0

Did you draw any conclusions from the feedback you received on the lynx-10g patch and decided that it doesn't apply to lynx-28g?
https://lore.kernel.org/linux-phy/[email protected]/

Hint: it does.

#2 lore raw
In reply to [email protected]
Quoted 20 lines · On Wed, Jul 22, 2026 at 09:06:19AM +0530, [email protected] wrote:

From: surendra <[email protected]>

lynx_28g_set_mode() currently ignores failures from lynx_28g_lane_enable_pcvt(). It then updates the lane mode and reports success even though the protocol converter may remain disabled.

Propagate the error and leave the previous lane mode intact so the caller can handle the failed reconfiguration.

Fixes 8f73b37cf3fb ("phy: add support for the Layerscape SerDes 28G")
---
 drivers/phy/freescale/phy-fsl-lynx-28g.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/phy/freescale/phy-fsl-lynx-28g.c +0 -0
index 38afcd081a2a..1fe406adc83c 100644

drivers/phy/freescale/phy-fsl-lynx-28g.c static int lynx_28g_set_mode(struct phy *phy, enum phy_mode mode, int submode) · line 1006

Quoted 4 lines
drivers/phy/freescale/phy-fsl-lynx-28g.c +0 -1
@@ -1003,7 +1003,9 @@ static int lynx_28g_set_mode(struct phy *phy, enum phy_mode mode, int submode)
 
 	lynx_28g_lane_change_proto_conf(lane, lane_mode);
 	lynx_28g_lane_remap_pll(lane, lane_mode);
-	WARN_ON(lynx_28g_lane_enable_pcvt(lane, lane_mode));

A WARN_ON() produces a very noticeable splat, which will aid debugging. Just returning the error code makes is easy to miss what causes the issue. Maybe the WARN_ON() should be kept? At least, i would add justification to the commit message for removing it.

Andrew

---
pw-bot: cr

#3 lore raw
In reply to [email protected]
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] Hardware and software state desynchronization on PCVT enable failure prevents mode recovery. --

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Hardware and software state desynchronization on PCVT enable failure prevents mode recovery. --

--- Patch [1]: [PATCH] phy: fsl-lynx-28g: propagate PCVT enable errors --- commit 6193a7adcb18a183b9462da919d480f3a6949562
Author: surendra <[email protected]>

phy: fsl-lynx-28g: propagate PCVT enable errors

This commit updates lynx_28g_set_mode() to check the return value of lynx_28g_lane_enable_pcvt(). If an error occurs, the code now propagates the error and exits before updating the software lane mode tracking.

drivers/phy/freescale/phy-fsl-lynx-28g.c static int lynx_28g_set_mode(struct phy *phy, enum phy_mode mode, int submode) · line 1008

Quoted 12 lines
drivers/phy/freescale/phy-fsl-lynx-28g.c +3 -1
index 38afcd081a2a7..1fe406adc83cb 100644
@@ -1003,7 +1003,9 @@ static int lynx_28g_set_mode(struct phy *phy, enum phy_mode mode, int submode)
 
 	lynx_28g_lane_change_proto_conf(lane, lane_mode);
 	lynx_28g_lane_remap_pll(lane, lane_mode);
-	WARN_ON(lynx_28g_lane_enable_pcvt(lane, lane_mode));
+	err = lynx_28g_lane_enable_pcvt(lane, lane_mode);
+	if (err)
+		goto out;

[Severity: Medium]
Does this new error path leave the software and hardware states desynchronized?

Prior to this point, the hardware has already been reprogrammed to the new mode by lynx_28g_lane_change_proto_conf() and lynx_28g_lane_remap_pll(). If lynx_28g_lane_enable_pcvt() fails and takes the goto out path, lane->mode remains set to the old mode.

If a caller later attempts to recover by setting the PHY back to the old mode, lynx_28g_set_mode() will see:

    if (lane_mode == lane->mode)
        return 0;

and short-circuit, preventing the hardware from actually being reverted.

Should the error path either revert the hardware changes or set lane->mode = LANE_MODE_UNKNOWN to ensure subsequent recovery attempts bypass the short-circuit check?

Quoted 3 lines
drivers/phy/freescale/phy-fsl-lynx-28g.c +0 -0
 	lane->mode = lane_mode;
 
Signature · 6 lines
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1

-- 
linux-phy mailing list
[email protected]
https://lists.infradead.org/mailman/listinfo/linux-phy