Re: v2 [PATCH 1/1] net: phy: marvell: avoid bringing down fibre link when autoneg is bypassed

From: Mateusz Polchlopek
Date: Thu Oct 03 2024 - 02:54:13 EST



On 10/3/2024 6:45 AM, Qingtao Cao wrote:
On 88E151x the SGMII autoneg bypass mode defaults to be enabled. When it is
activated, the device assumes a link-up status with existing configuration
in BMCR, avoid bringing down the fibre link in this case

Test case:
1. Two 88E151x connected with SFP, both enable autoneg, link is up with speed
1000M

checkpatch.pl complains about this line, it exceeds 75 chars allowed for
commit msg. Please adjust.

2. Disable autoneg on one device and explicitly set its speed to 1000M
3. The fibre link can still up with this change, otherwise not.

Signed-off-by: Qingtao Cao <qingtao.cao@xxxxxxxx>
---
drivers/net/phy/marvell.c | 23 ++++++++++++++++++++++-
1 file changed, 22 insertions(+), 1 deletion(-)

diff --git a/drivers/net/phy/marvell.c b/drivers/net/phy/marvell.c
index 9964bf3dea2f..e3a8ad8b08dd 100644
--- a/drivers/net/phy/marvell.c
+++ b/drivers/net/phy/marvell.c
@@ -195,6 +195,10 @@
#define MII_88E1510_MSCR_2 0x15
+#define MII_88E1510_FSCR2 0x1a

Please use GENMASK_ULL for creating mask.

+#define MII_88E1510_FSCR2_BYPASS_ENABLE BIT(6)
+#define MII_88E1510_FSCR2_BYPASS_STATUS BIT(5)
+
#define MII_VCT5_TX_RX_MDI0_COUPLING 0x10
#define MII_VCT5_TX_RX_MDI1_COUPLING 0x11
#define MII_VCT5_TX_RX_MDI2_COUPLING 0x12
@@ -1623,11 +1627,28 @@ static void fiber_lpa_mod_linkmode_lpa_t(unsigned long *advertising, u32 lpa)
static int marvell_read_status_page_an(struct phy_device *phydev,
int fiber, int status)
{
+ int fscr2;
int lpa;
int err;
if (!(status & MII_M1011_PHY_STATUS_RESOLVED)) {
- phydev->link = 0;
+ if (!fiber) {
+ phydev->link = 0;
+ } else {
+ fscr2 = phy_read(phydev, MII_88E1510_FSCR2);
+ if (fscr2 > 0) {
+ if ((fscr2 & MII_88E1510_FSCR2_BYPASS_ENABLE) &&
+ (fscr2 & MII_88E1510_FSCR2_BYPASS_STATUS)) {
+ if (genphy_read_status_fixed(phydev) < 0)
+ phydev->link = 0;
+ } else {
+ phydev->link = 0;
+ }
+ } else {
+ phydev->link = 0;
+ }
+ }
+
return 0;
}

So many levels of indentation... Couldn't it be merged somehow? I do not
know, maybe create local variable, store the current state of
phydev->link, then set phydev->link = 0 and restore it from local
variable only if (fiber && fscr2 > 0 && (fscr2 &
MII_88E1510_FSCR2_BYPASS_ENABLE) && (fscr2 & MII_88E1510_FSCR2_BYPASS_STATUS) && genphy_read_status_fixed(phydev) >=0 ) ...

or other way? Now you have 5 (!) levels of indentation and almost
everywhere you just set phydev->link to 0 depends on the condition.


BTW. We put "v2" inside the tag in the topic and specify the tree, so
instead of:
v2 [PATCH 1/1] net: phy: marvell: avoid bringing down fibre link when autoneg is bypassed

it should be:
[PATCH net-next v2 1/1] net: phy: marvell: avoid bringing down fibre link when autoneg is bypassed