This series fixes some issues in the Lynx 28G SerDes driver, namely an
oops when unloading the module, a race between the periodic workqueue
and the PHY API, and a race between phy_set_mode_ext() calls on multiple
lanes on the same SerDes.
Ioana Ciornei (1):
phy: lynx-28g: cancel the CDR check work item on the remove path
Vladimir Oltean (2):
phy: lynx-28g: lock PHY while performing CDR lock workaround
phy: lynx-28g: serialize concurrent phy_set_mode_ext() calls to shared
registers
drivers/phy/freescale/phy-fsl-lynx-28g.c | 27 +++++++++++++++++++++---
1 file changed, 24 insertions(+), 3 deletions(-)
---
I have some more work pending on this driver, hopefully for this
development cycle, which is entangled with net-next. Can this series go
through the networking tree, with the PHY maintainers' Ack? The original
driver submission also went through net-next.
--
2.34.1
lynx_28g_cdr_lock_check() runs once per second in a workqueue to reset
the lane receiver if the CDR has not locked onto bit transitions in the
RX stream. But the PHY consumer may do stuff with the PHY simultaneously,
and that isn't okay. Block concurrent generic PHY calls by holding the
PHY mutex from this workqueue.
Fixes: 8f73b37cf3fb ("phy: add support for the Layerscape SerDes 28G")
Signed-off-by: Vladimir Oltean <[email protected]>
---
drivers/phy/freescale/phy-fsl-lynx-28g.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/drivers/phy/freescale/phy-fsl-lynx-28g.c b/drivers/phy/freescale/phy-fsl-lynx-28g.c
index c6323669f119..d5385d36735d 100644
--- a/drivers/phy/freescale/phy-fsl-lynx-28g.c
+++ b/drivers/phy/freescale/phy-fsl-lynx-28g.c
@@ -508,11 +508,12 @@ static void lynx_28g_cdr_lock_check(struct work_struct *work)
for (i = 0; i < LYNX_28G_NUM_LANE; i++) {
lane = &priv->lane[i];
- if (!lane->init)
- continue;
+ mutex_lock(&lane->phy->mutex);
- if (!lane->powered_up)
+ if (!lane->init || !lane->powered_up) {
+ mutex_unlock(&lane->phy->mutex);
continue;
+ }
rrstctl = lynx_28g_lane_read(lane, LNaRRSTCTL);
if (!(rrstctl & LYNX_28G_LNaRRSTCTL_CDR_LOCK)) {
@@ -521,6 +522,8 @@ static void lynx_28g_cdr_lock_check(struct work_struct *work)
rrstctl = lynx_28g_lane_read(lane, LNaRRSTCTL);
} while (!(rrstctl & LYNX_28G_LNaRRSTCTL_RST_DONE));
}
+
+ mutex_unlock(&lane->phy->mutex);
}
queue_delayed_work(system_power_efficient_wq, &priv->cdr_check,
msecs_to_jiffies(1000));
--
2.34.1
The protocol converter configuration registers PCC8, PCCC, PCCD
(implemented by the driver), as well as others, control protocol
converters from multiple lanes (each represented as a different
struct phy). So, if there are simultaneous calls to phy_set_mode_ext()
to lanes sharing the same PCC register (either for the "old" or for the
"new" protocol), corruption of the values programmed to hardware is
possible, because lynx_28g_rmw() has no locking.
Add a spinlock in the struct lynx_28g_priv shared by all lanes, and take
the global spinlock from the phy_ops :: set_mode() implementation. There
are no other callers which modify PCC registers.
Fixes: 8f73b37cf3fb ("phy: add support for the Layerscape SerDes 28G")
Signed-off-by: Vladimir Oltean <[email protected]>
---
drivers/phy/freescale/phy-fsl-lynx-28g.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/drivers/phy/freescale/phy-fsl-lynx-28g.c b/drivers/phy/freescale/phy-fsl-lynx-28g.c
index d5385d36735d..e2187767ce00 100644
--- a/drivers/phy/freescale/phy-fsl-lynx-28g.c
+++ b/drivers/phy/freescale/phy-fsl-lynx-28g.c
@@ -127,6 +127,10 @@ struct lynx_28g_lane {
struct lynx_28g_priv {
void __iomem *base;
struct device *dev;
+ /* Serialize concurrent access to registers shared between lanes,
+ * like PCCn
+ */
+ spinlock_t pcc_lock;
struct lynx_28g_pll pll[LYNX_28G_NUM_PLL];
struct lynx_28g_lane lane[LYNX_28G_NUM_LANE];
@@ -397,6 +401,8 @@ static int lynx_28g_set_mode(struct phy *phy, enum phy_mode mode, int submode)
if (powered_up)
lynx_28g_power_off(phy);
+ spin_lock(&priv->pcc_lock);
+
switch (submode) {
case PHY_INTERFACE_MODE_SGMII:
case PHY_INTERFACE_MODE_1000BASEX:
@@ -413,6 +419,8 @@ static int lynx_28g_set_mode(struct phy *phy, enum phy_mode mode, int submode)
lane->interface = submode;
out:
+ spin_unlock(&priv->pcc_lock);
+
/* Power up the lane if necessary */
if (powered_up)
lynx_28g_power_on(phy);
@@ -596,6 +604,7 @@ static int lynx_28g_probe(struct platform_device *pdev)
dev_set_drvdata(dev, priv);
+ spin_lock_init(&priv->pcc_lock);
INIT_DELAYED_WORK(&priv->cdr_check, lynx_28g_cdr_lock_check);
queue_delayed_work(system_power_efficient_wq, &priv->cdr_check,
--
2.34.1
From: Ioana Ciornei <[email protected]>
The blamed commit added the CDR check work item but didn't cancel it on
the remove path. Fix this by adding a remove function which takes care
of it.
Fixes: 8f73b37cf3fb ("phy: add support for the Layerscape SerDes 28G")
Signed-off-by: Ioana Ciornei <[email protected]>
Signed-off-by: Vladimir Oltean <[email protected]>
---
drivers/phy/freescale/phy-fsl-lynx-28g.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/drivers/phy/freescale/phy-fsl-lynx-28g.c b/drivers/phy/freescale/phy-fsl-lynx-28g.c
index 4f036c77284e..c6323669f119 100644
--- a/drivers/phy/freescale/phy-fsl-lynx-28g.c
+++ b/drivers/phy/freescale/phy-fsl-lynx-28g.c
@@ -604,6 +604,14 @@ static int lynx_28g_probe(struct platform_device *pdev)
return PTR_ERR_OR_ZERO(provider);
}
+static void lynx_28g_remove(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ struct lynx_28g_priv *priv = dev_get_drvdata(dev);
+
+ cancel_delayed_work_sync(&priv->cdr_check);
+}
+
static const struct of_device_id lynx_28g_of_match_table[] = {
{ .compatible = "fsl,lynx-28g" },
{ },
@@ -612,6 +620,7 @@ MODULE_DEVICE_TABLE(of, lynx_28g_of_match_table);
static struct platform_driver lynx_28g_driver = {
.probe = lynx_28g_probe,
+ .remove_new = lynx_28g_remove,
.driver = {
.name = "lynx-28g",
.of_match_table = lynx_28g_of_match_table,
--
2.34.1
Hello:
This series was applied to netdev/net.git (main)
by David S. Miller <[email protected]>:
On Wed, 4 Oct 2023 14:17:05 +0300 you wrote:
> This series fixes some issues in the Lynx 28G SerDes driver, namely an
> oops when unloading the module, a race between the periodic workqueue
> and the PHY API, and a race between phy_set_mode_ext() calls on multiple
> lanes on the same SerDes.
>
> Ioana Ciornei (1):
> phy: lynx-28g: cancel the CDR check work item on the remove path
>
> [...]
Here is the summary with links:
- [net,1/3] phy: lynx-28g: cancel the CDR check work item on the remove path
https://git.kernel.org/netdev/net/c/f200bab3756f
- [net,2/3] phy: lynx-28g: lock PHY while performing CDR lock workaround
https://git.kernel.org/netdev/net/c/0ac87fe54a17
- [net,3/3] phy: lynx-28g: serialize concurrent phy_set_mode_ext() calls to shared registers
https://git.kernel.org/netdev/net/c/139ad1143151
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html