[patch] mlxsw: i2c: Prevent transaction execution for special chip st…

…ates

This patch fixes an issue encountered during SONiC warm-reboot: during
ISSU start the kernel reports that CPU stall is detected:

```
INFO: rcu_sched detected stalls on CPUs/tasks
```

This patch won't be upstreamed due to the upstream driver not supporting
ISSU.

Signed-off-by: Stepan Blyschak <stepanb@nvidia.com>

patch/0080-TMP-mlxsw-i2c-Prevent-transaction-execution-for-spec.patch

@@ -0,0 +1,126 @@
+From 392127480c91a26dae2449288229d4957213dcc1 Mon Sep 17 00:00:00 2001
+From: Vadim Pasternak <vadimp@nvidia.com>
+Date: Tue, 31 May 2022 10:27:38 +0300
+Subject: [PATCH v4.9 ISSU WA 1/1] TMP: mlxsw: i2c: Prevent transaction
+ execution for special chip states
+
+Do not run transaction in cases chip is in reset or in-service update
+states.
+In such case firmware is not accessible and will reject transaction
+with the relevant status "RUNNING_RESET" or "FW_ISSU_ONGOING".
+In case transaction is failed do to one of these reasons, stop sending
+transactions. In such case driver is about to be removed since it
+cannot continue running after reset or in-service update. And
+re-probed again after reset or in-service update is completed.
+
+Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
+---
+ drivers/net/ethernet/mellanox/mlxsw/cmd.h |  4 ++++
+ drivers/net/ethernet/mellanox/mlxsw/i2c.c | 29 ++++++++++++++++++++---
+ 2 files changed, 30 insertions(+), 3 deletions(-)
+
+diff --git a/drivers/net/ethernet/mellanox/mlxsw/cmd.h b/drivers/net/ethernet/mellanox/mlxsw/cmd.h
+index 0772e4339b33..6329ebb09523 100644
+--- a/drivers/net/ethernet/mellanox/mlxsw/cmd.h
++++ b/drivers/net/ethernet/mellanox/mlxsw/cmd.h
+@@ -149,6 +149,8 @@ enum mlxsw_cmd_status {
+ 	MLXSW_CMD_STATUS_BAD_NVMEM	= 0x0B,
+ 	/* Device is currently running reset */
+ 	MLXSW_CMD_STATUS_RUNNING_RESET	= 0x26,
++	/* FW ISSU ongoing. */
++	MLXSW_CMD_STATUS_FW_ISSU	= 0x27,
+ 	/* Bad management packet (silently discarded). */
+ 	MLXSW_CMD_STATUS_BAD_PKT	= 0x30,
+ };
+@@ -180,6 +182,8 @@ static inline const char *mlxsw_cmd_status_str(u8 status)
+ 		return "BAD_NVMEM";
+ 	case MLXSW_CMD_STATUS_RUNNING_RESET:
+ 		return "RUNNING_RESET";
++	case MLXSW_CMD_STATUS_FW_ISSU:
++		return "FW_ISSU_ONGOING";
+ 	case MLXSW_CMD_STATUS_BAD_PKT:
+ 		return "BAD_PKT";
+ 	default:
+diff --git a/drivers/net/ethernet/mellanox/mlxsw/i2c.c b/drivers/net/ethernet/mellanox/mlxsw/i2c.c
+index e04d521d9376..1205c6111b0b 100644
+--- a/drivers/net/ethernet/mellanox/mlxsw/i2c.c
++++ b/drivers/net/ethernet/mellanox/mlxsw/i2c.c
+@@ -62,6 +62,7 @@
+  * @core: switch core pointer;
+  * @bus_info: bus info block;
+  * @block_size: maximum block size allowed to pass to under layer;
++ * @status: status to indicate chip reset or in-service update;
+  */
+ struct mlxsw_i2c {
+ 	struct {
+@@ -75,6 +76,7 @@ struct mlxsw_i2c {
+ 	struct mlxsw_core *core;
+ 	struct mlxsw_bus_info bus_info;
+ 	u16 block_size;
++	u8 status;
+ };
+
+ #define MLXSW_I2C_READ_MSG(_client, _addr_buf, _buf, _len) {	\
+@@ -221,6 +223,19 @@ static int mlxsw_i2c_write_cmd(struct i2c_client *client,
+ 	return 0;
+ }
+
++static bool
++mlxsw_i2c_cmd_status_verify(struct device *dev, struct mlxsw_i2c *mlxsw_i2c,
++			    u8 status)
++{
++	if (status == MLXSW_CMD_STATUS_FW_ISSU ||
++	    status == MLXSW_CMD_STATUS_RUNNING_RESET) {
++		mlxsw_i2c->status = status;
++		dev_info(dev, "FW status=%x(%s))\n", status, mlxsw_cmd_status_str(status));
++		return true;
++	}
++	return false;
++}
++
+ /* Routine posts initialization command to ASIC through mail box. */
+ static int
+ mlxsw_i2c_write_init_cmd(struct i2c_client *client,
+@@ -404,6 +419,10 @@ mlxsw_i2c_cmd(struct device *dev, u16 opcode, u32 in_mod, size_t in_mbox_size,
+
+ 	WARN_ON(in_mbox_size % sizeof(u32) || out_mbox_size % sizeof(u32));
+
++	/* Do not run transaction if chip is in reset or in-service update state. */
++	if (mlxsw_i2c->status)
++		return 0;
++
+ 	if (in_mbox) {
+ 		reg_size = mlxsw_i2c_get_reg_size(in_mbox);
+ 		num = reg_size / mlxsw_i2c->block_size;
+@@ -478,6 +497,8 @@ mlxsw_i2c_cmd(struct device *dev, u16 opcode, u32 in_mod, size_t in_mbox_size,
+
+ cmd_fail:
+ 	mutex_unlock(&mlxsw_i2c->cmd.lock);
++	if (mlxsw_i2c_cmd_status_verify(&client->dev, mlxsw_i2c, *status))
++		err = 0;
+ 	return err;
+ }
+
+@@ -607,14 +628,16 @@ static int mlxsw_i2c_probe(struct i2c_client *client,
+ 	/* Wait until go bit is cleared. */
+ 	err = mlxsw_i2c_wait_go_bit(client, mlxsw_i2c, &status);
+ 	if (err) {
+-		dev_err(&client->dev, "HW semaphore is not released");
++		if (!mlxsw_i2c_cmd_status_verify(&client->dev, mlxsw_i2c, status))
++			dev_err(&client->dev, "HW semaphore is not released");
+ 		goto errout;
+ 	}
+
+ 	/* Validate transaction completion status. */
+ 	if (status) {
+-		dev_err(&client->dev, "Bad transaction completion status %x\n",
+-			status);
++		if (!mlxsw_i2c_cmd_status_verify(&client->dev, mlxsw_i2c, status))
++			dev_err(&client->dev, "Bad transaction completion status %x\n",
++				status);
+ 		err = -EIO;
+ 		goto errout;
+ 	}
+--
+2.20.1
+

patch/series

@@ -119,6 +119,7 @@ linux-4.13-thermal-intel_pch_thermal-Fix-enable-check-on.patch
 0077-mlxsw-core-Increase-critical-threshold-for-ASIC-ther.patch
 0078-mlxsw-core-Add-validation-of-transceiver-temperature.patch
 0079-mlxsw-core-Remove-critical-trip-point-from-thermal-z.patch
+0080-TMP-mlxsw-i2c-Prevent-transaction-execution-for-spec.patch
 linux-4.16-firmware-dmi-handle-missing-DMI-data-gracefully.patch
 mellanox-backport-introduce-psample-a-new-genetlink-channel.patch
 mellanox-backport-introduce-tc-sample-action.patch