Thread (20 messages) 20 messages, 3 authors, 2021-06-02

Re: [RFC PATCH v2 net-next 9/9] net: pcs: xpcs: convert to phylink_pcs_ops

From: Vladimir Oltean <olteanv@gmail.com>
Date: 2021-06-02 15:15:09

On Wed, Jun 02, 2021 at 11:10:56PM +0800, Wong Vee Khee wrote:
On Wed, Jun 02, 2021 at 05:57:30PM +0300, Vladimir Oltean wrote:
quoted
On Wed, Jun 02, 2021 at 10:43:36PM +0800, Wong Vee Khee wrote:
quoted
On Wed, Jun 02, 2021 at 05:02:33PM +0300, Vladimir Oltean wrote:
quoted
On Wed, Jun 02, 2021 at 02:47:49PM +0100, Russell King (Oracle) wrote:
quoted
On Wed, Jun 02, 2021 at 04:43:21PM +0300, Vladimir Oltean wrote:
quoted
On Tue, Jun 01, 2021 at 01:10:33PM +0100, Russell King (Oracle) wrote:
quoted
On Tue, Jun 01, 2021 at 03:33:25AM +0300, Vladimir Oltean wrote:
quoted
 static const struct phylink_mac_ops stmmac_phylink_mac_ops = {
 	.validate = stmmac_validate,
-	.mac_pcs_get_state = stmmac_mac_pcs_get_state,
-	.mac_config = stmmac_mac_config,
mac_config is still a required function.
This is correct, thanks.

VK, would you mind testing again with this extra patch added to the mix?
If it works, I will add it to the series in v3, ordered properly.

-----------------------------[ cut here]-----------------------------
From a79863027998451c73d5bbfaf1b77cf6097a110c Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <vladimir.oltean@nxp.com>
Date: Wed, 2 Jun 2021 16:35:55 +0300
Subject: [PATCH] net: phylink: allow the mac_config method to be missing if
 pcs_ops are provided

The pcs_config method from struct phylink_pcs_ops does everything that
the mac_config method from struct phylink_mac_ops used to do in the
legacy approach of driving a MAC PCS. So allow drivers to not implement
the mac_config method if there is nothing to do. Keep the method
required for setups that do not provide pcs_ops.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
---
 drivers/net/phy/phylink.c | 9 +++++++++
 1 file changed, 9 insertions(+)
diff --git a/drivers/net/phy/phylink.c b/drivers/net/phy/phylink.c
index 96d8e88b4e46..a8842c6ce3a2 100644
--- a/drivers/net/phy/phylink.c
+++ b/drivers/net/phy/phylink.c
@@ -415,6 +415,9 @@ static void phylink_resolve_flow(struct phylink_link_state *state)
 static void phylink_mac_config(struct phylink *pl,
 			       const struct phylink_link_state *state)
 {
+	if (!pl->mac_ops->mac_config)
+		return;
+
 	phylink_dbg(pl,
 		    "%s: mode=%s/%s/%s/%s adv=%*pb pause=%02x link=%u an=%u\n",
 		    __func__, phylink_an_mode_str(pl->cur_link_an_mode),
@@ -1192,6 +1195,12 @@ void phylink_start(struct phylink *pl)
 
 	ASSERT_RTNL();
 
+	/* The mac_ops::mac_config method may be absent only if the
+	 * pcs_ops are present.
+	 */
+	if (WARN_ON_ONCE(!pl->mac_ops->mac_config && !pl->pcs_ops))
+		return;
+
 	phylink_info(pl, "configuring for %s/%s link mode\n",
 		     phylink_an_mode_str(pl->cur_link_an_mode),
 		     phy_modes(pl->link_config.interface));
I would rather we didn't do that - I suspect your case is not the
common scenario, so please add a dummy function to stmmac instead.
Ok, in that case the only delta to be applied should be:

-----------------------------[ cut here]-----------------------------
quoted
From 998569108392befc591f790e46fe5dcd1d0b6278 Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <vladimir.oltean@nxp.com>
Date: Wed, 2 Jun 2021 17:00:12 +0300
Subject: [PATCH] fixup! net: pcs: xpcs: convert to phylink_pcs_ops

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
---
 drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 6 ++++++
 1 file changed, 6 insertions(+)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index d5685a74f3b7..704aa91b145a 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1000,6 +1000,12 @@ static void stmmac_validate(struct phylink_config *config,
 		xpcs_validate(priv->hw->xpcs, supported, state);
 }
 
+static void stmmac_mac_config(struct phylink_config *config, unsigned int mode,
+			      const struct phylink_link_state *state)
+{
+	/* Nothing to do, xpcs_config() handles everything */
+}
+
 static void stmmac_fpe_link_state_handle(struct stmmac_priv *priv, bool is_up)
 {
 	struct stmmac_fpe_cfg *fpe_cfg = priv->plat->fpe_cfg;
-----------------------------[ cut here]-----------------------------
No luck.. I am still seeing the same kernel panic on my side with the
additional patch applied:

[   12.513652] intel-eth-pci 0000:00:1e.4 enp0s30f4: FPE workqueue start
[   12.520079] intel-eth-pci 0000:00:1e.4 enp0s30f4: configuring for inband/sgmii link mode
[   12.528141] BUG: kernel NULL pointer dereference, address: 0000000000000000
[   12.535072] #PF: supervisor instruction fetch in kernel mode
[   12.540719] #PF: error_code(0x0010) - not-present page
[   12.545843] PGD 0 P4D 0
[   12.548382] Oops: 0010 [#1] PREEMPT SMP NOPTI
[   12.552733] CPU: 3 PID: 2657 Comm: connmand Tainted: G     U            5.13.0-rc3-intel-lts #79
[   12.561485] Hardware name: Intel Corporation Tiger Lake Client Platform/TigerLake U DDR4 SODIMM RVP, BIOS TGLIFUI1.R00.3373.AF0.2009230546 09/23/2020
[   12.574803] RIP: 0010:0x0
[   12.577436] Code: Unable to access opcode bytes at RIP 0xffffffffffffffd6.
[   12.584280] RSP: 0018:ffffaa05404efb78 EFLAGS: 00010246
[   12.589489] RAX: 0000000000000000 RBX: ffffa0fb91758c00 RCX: 0000000000000000
[   12.596600] RDX: ffffaa05404efba0 RSI: 0000000000000002 RDI: ffffa0facc63c4d8
[   12.603711] RBP: ffffaa05404efba0 R08: 0000000000000000 R09: ffffaa05404ef888
[   12.610825] R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000001
[   12.617935] R13: ffffa0facc6388c0 R14: 0000000000000006 R15: 0000000000000001
[   12.625040] FS:  00007f4a6d5b87c0(0000) GS:ffffa0fc57f80000(0000) knlGS:0000000000000000
[   12.633096] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   12.638825] CR2: ffffffffffffffd6 CR3: 000000010e3f6004 CR4: 0000000000770ee0
[   12.645932] PKRU: 55555554
[   12.648641] Call Trace:
[   12.651095]  phylink_major_config+0x5e/0x1a0 [phylink]
[   12.656219]  phylink_start+0x204/0x2c0 [phylink]
[   12.660833]  stmmac_open+0x3d0/0x9f0 [stmmac]
[   12.665186]  __dev_open+0xe7/0x180
[   12.668594]  __dev_change_flags+0x174/0x1d0
[   12.672773]  ? __thaw_task+0x40/0x40
[   12.676348]  ? arch_stack_walk+0x9e/0xf0
[   12.680266]  dev_change_flags+0x21/0x60
[   12.684100]  devinet_ioctl+0x5e8/0x750
[   12.687847]  inet_ioctl+0x190/0x1c0
[   12.691335]  ? dev_ioctl+0x26d/0x4c0
[   12.694907]  sock_do_ioctl+0x44/0x140
[   12.698569]  ? alloc_empty_file+0x61/0xb0
[   12.702572]  sock_ioctl+0x22c/0x320
[   12.706061]  __x64_sys_ioctl+0x80/0xb0
[   12.709808]  do_syscall_64+0x42/0x80
[   12.713381]  entry_SYSCALL_64_after_hwframe+0x44/0xae

My git log:-
dc25bc46886e (HEAD -> master) phy: maxlinear: add Maxlinear GPY115/21x/24x driver
d059f32fcef1 fixup! net: pcs: xpcs: convert to phylink_pcs_ops
1a1a9c7267c1 net: pcs: xpcs: convert to phylink_pcs_ops
ec25f3c80108 net: pcs: xpcs: convert to mdio_device
912a23a5768a net: pcs: xpcs: use mdiobus_c45_addr in xpcs_{read,write}
af2bd0a4e32c net: pcs: xpcs: export xpcs_probe
7e4eea51132f net: pcs: xpcs: export xpcs_config_eee
37d5e086fc83 net: pcs: xpcs: export xpcs_validate
d58487edb1a8 net: pcs: xpcs: make the checks related to the PHY interface mode stateless
2f0eba94af8d net: pcs: xpcs: there is only one PHY ID
67148dd06896 net: pcs: xpcs: delete shim definition for mdio_xpcs_get_ops()
5fe8e519e44f (origin/master, origin/HEAD) Merge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-next
Yikes, stupid me, I forgot to add this piece to the delta, the new
callback wasn't doing anything, but it also didn't warn me that the
static function is unused in my build:

-----------------------------[ cut here]-----------------------------
quoted
From 7a2b56091fd480ae1018ac964756368904a41a0c Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <vladimir.oltean@nxp.com>
Date: Wed, 2 Jun 2021 17:54:58 +0300
Subject: [PATCH] fixup! fixup! net: pcs: xpcs: convert to phylink_pcs_ops

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
---
 drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 1 +
 1 file changed, 1 insertion(+)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 704aa91b145a..5b5edfdccab3 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1137,6 +1137,7 @@ static void stmmac_mac_link_up(struct phylink_config *config,
 
 static const struct phylink_mac_ops stmmac_phylink_mac_ops = {
 	.validate = stmmac_validate,
+	.mac_config = stmmac_mac_config,
 	.mac_link_down = stmmac_mac_link_down,
 	.mac_link_up = stmmac_mac_link_up,
 };
-----------------------------[ cut here]-----------------------------
It works on my setup now after applying the last fixup patch! :)
Thanks again for testing.
Just for clarity, "it works" includes traffic at 10/100/1000, right?
If I'm not mistaken, you tested with sgmii and MLO_INBAND_AN, correct?
On the sja1105 I am testing with sgmii and no MLO_INBAND_AN, so not
exactly the same setup (that, plus my hardware needs a few more changes
still).
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help