wifi: mt76: mt7915: add full system reset into debugfs

Trigger firmware crash and enable full system recovery through debugfs.
This also renames knob "fw_ser" to a clear-cut name "sys_recovery".

Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

5 files changed, 80 insertions, 14 deletions

diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/debugfs.c b/drivers/net/wireless/mediatek/mt76/mt7915/debugfs.c
index cd774c1bc49f..096cb8a4db3d 100644
--- a/drivers/net/wireless/mediatek/mt76/mt7915/debugfs.c
+++ b/drivers/net/wireless/mediatek/mt76/mt7915/debugfs.c
@@ -46,8 +46,8 @@ DEFINE_DEBUGFS_ATTRIBUTE(fops_implicit_txbf, mt7915_implicit_txbf_get,
 
 /* test knob of system error recovery */
 static ssize_t
-mt7915_fw_ser_set(struct file *file, const char __user *user_buf,
-		  size_t count, loff_t *ppos)
+mt7915_sys_recovery_set(struct file *file, const char __user *user_buf,
+			size_t count, loff_t *ppos)
 {
 	struct mt7915_phy *phy = file->private_data;
 	struct mt7915_dev *dev = phy->dev;
@@ -71,8 +71,18 @@ mt7915_fw_ser_set(struct file *file, const char __user *user_buf,
 		return -EINVAL;
 
 	switch (val) {
+	/*
+	 * 0: grab firmware current SER state.
+	 * 1: trigger & enable system error L1 recovery.
+	 * 2: trigger & enable system error L2 recovery.
+	 * 3: trigger & enable system error L3 rx abort.
+	 * 4: trigger & enable system error L3 tx abort
+	 * 5: trigger & enable system error L3 tx disable.
+	 * 6: trigger & enable system error L3 bf recovery.
+	 * 7: trigger & enable system error full recovery.
+	 * 8: trigger firmware crash.
+	 */
 	case SER_QUERY:
-		/* grab firmware SER stats */
 		ret = mt7915_mcu_set_ser(dev, 0, 0, ext_phy);
 		break;
 	case SER_SET_RECOVER_L1:
@@ -87,6 +97,23 @@ mt7915_fw_ser_set(struct file *file, const char __user *user_buf,
 
 		ret = mt7915_mcu_set_ser(dev, SER_RECOVER, val, ext_phy);
 		break;
+
+	/* enable full chip reset */
+	case SER_SET_RECOVER_FULL:
+		mt76_set(dev, MT_WFDMA0_MCU_HOST_INT_ENA, MT_MCU_CMD_WDT_MASK);
+		ret = mt7915_mcu_set_ser(dev, 1, 3, ext_phy);
+		if (ret)
+			return ret;
+
+		dev->recovery.state |= MT_MCU_CMD_WDT_MASK;
+		mt7915_reset(dev);
+		break;
+
+	/* WARNING: trigger firmware crash */
+	case SER_SET_SYSTEM_ASSERT:
+		mt76_wr(dev, MT_MCU_WM_CIRQ_EINT_MASK_CLR_ADDR, BIT(18));
+		mt76_wr(dev, MT_MCU_WM_CIRQ_EINT_SOFT_ADDR, BIT(18));
+		break;
 	default:
 		break;
 	}
@@ -95,20 +122,45 @@ mt7915_fw_ser_set(struct file *file, const char __user *user_buf,
 }
 
 static ssize_t
-mt7915_fw_ser_get(struct file *file, char __user *user_buf,
-		  size_t count, loff_t *ppos)
+mt7915_sys_recovery_get(struct file *file, char __user *user_buf,
+			size_t count, loff_t *ppos)
 {
 	struct mt7915_phy *phy = file->private_data;
 	struct mt7915_dev *dev = phy->dev;
 	char *buff;
 	int desc = 0;
 	ssize_t ret;
-	static const size_t bufsz = 400;
+	static const size_t bufsz = 1024;
 
 	buff = kmalloc(bufsz, GFP_KERNEL);
 	if (!buff)
 		return -ENOMEM;
 
+	/* HELP */
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "Please echo the correct value ...\n");
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "0: grab firmware transient SER state\n");
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "1: trigger system error L1 recovery\n");
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "2: trigger system error L2 recovery\n");
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "3: trigger system error L3 rx abort\n");
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "4: trigger system error L3 tx abort\n");
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "5: trigger system error L3 tx disable\n");
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "6: trigger system error L3 bf recovery\n");
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "7: trigger system error full recovery\n");
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "8: trigger firmware crash\n");
+
+	/* SER statistics */
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "\nlet's dump firmware SER statistics...\n");
 	desc += scnprintf(buff + desc, bufsz - desc,
 			  "::E  R , SER_STATUS        = 0x%08x\n",
 			  mt76_rr(dev, MT_SWDEF_SER_STATS));
@@ -139,15 +191,19 @@ mt7915_fw_ser_get(struct file *file, char __user *user_buf,
 	desc += scnprintf(buff + desc, bufsz - desc,
 			  "::E  R , SER_LMAC_WISR7_B1 = 0x%08x\n",
 			  mt76_rr(dev, MT_SWDEF_LAMC_WISR7_BN1_STATS));
+	desc += scnprintf(buff + desc, bufsz - desc,
+			  "\nSYS_RESET_COUNT: WM %d, WA %d\n",
+			  dev->recovery.wm_reset_count,
+			  dev->recovery.wa_reset_count);
 
 	ret = simple_read_from_buffer(user_buf, count, ppos, buff, desc);
 	kfree(buff);
 	return ret;
 }
 
-static const struct file_operations mt7915_fw_ser_ops = {
-	.write = mt7915_fw_ser_set,
-	.read = mt7915_fw_ser_get,
+static const struct file_operations mt7915_sys_recovery_ops = {
+	.write = mt7915_sys_recovery_set,
+	.read = mt7915_sys_recovery_get,
 	.open = simple_open,
 	.llseek = default_llseek,
 };
@@ -598,10 +654,6 @@ mt7915_fw_util_wm_show(struct seq_file *file, void *data)
 	struct mt7915_dev *dev = file->private;
 
 	seq_printf(file, "Program counter: 0x%x\n", mt76_rr(dev, MT_WM_MCU_PC));
-	seq_printf(file, "Exception state: 0x%x\n",
-		   is_mt7915(&dev->mt76) ?
-		   (u32)mt76_get_field(dev, MT_FW_EXCEPTION, GENMASK(15, 8)) :
-		   (u32)mt76_get_field(dev, MT_FW_EXCEPTION, GENMASK(7, 0)));
 
 	if (dev->fw.debug_wm) {
 		seq_printf(file, "Busy: %u%%  Peak busy: %u%%\n",
@@ -1009,7 +1061,8 @@ int mt7915_init_debugfs(struct mt7915_phy *phy)
 	debugfs_create_file("xmit-queues", 0400, dir, phy,
 			    &mt7915_xmit_queues_fops);
 	debugfs_create_file("tx_stats", 0400, dir, phy, &mt7915_tx_stats_fops);
-	debugfs_create_file("fw_ser", 0600, dir, phy, &mt7915_fw_ser_ops);
+	debugfs_create_file("sys_recovery", 0600, dir, phy,
+			    &mt7915_sys_recovery_ops);
 	debugfs_create_file("fw_debug_wm", 0600, dir, dev, &fops_fw_debug_wm);
 	debugfs_create_file("fw_debug_wa", 0600, dir, dev, &fops_fw_debug_wa);
 	debugfs_create_file("fw_debug_bin", 0600, dir, dev, &fops_fw_debug_bin);
diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/mac.c b/drivers/net/wireless/mediatek/mt76/mt7915/mac.c
index 383b260eb0aa..3ac143a9e804 100644
--- a/drivers/net/wireless/mediatek/mt76/mt7915/mac.c
+++ b/drivers/net/wireless/mediatek/mt76/mt7915/mac.c
@@ -1526,6 +1526,11 @@ void mt7915_mac_reset_work(struct work_struct *work)
 		mt76_clear(dev, MT_WFDMA0_MCU_HOST_INT_ENA,
 			   MT_MCU_CMD_WDT_MASK);
 
+		if (READ_ONCE(dev->recovery.state) & MT_MCU_CMD_WA_WDT)
+			dev->recovery.wa_reset_count++;
+		else
+			dev->recovery.wm_reset_count++;
+
 		mt7915_mac_full_reset(dev);
 
 		/* enable mcu irq */
diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/mcu.h b/drivers/net/wireless/mediatek/mt76/mt7915/mcu.h
index a4f903f88078..c19b5d66c0e1 100644
--- a/drivers/net/wireless/mediatek/mt76/mt7915/mcu.h
+++ b/drivers/net/wireless/mediatek/mt76/mt7915/mcu.h
@@ -450,6 +450,8 @@ enum {
 	SER_SET_RECOVER_L3_TX_ABORT,
 	SER_SET_RECOVER_L3_TX_DISABLE,
 	SER_SET_RECOVER_L3_BF,
+	SER_SET_RECOVER_FULL,
+	SER_SET_SYSTEM_ASSERT,
 	/* action */
 	SER_ENABLE = 2,
 	SER_RECOVER
diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/mt7915.h b/drivers/net/wireless/mediatek/mt76/mt7915/mt7915.h
index fce1fde72116..8615a38bec06 100644
--- a/drivers/net/wireless/mediatek/mt76/mt7915/mt7915.h
+++ b/drivers/net/wireless/mediatek/mt76/mt7915/mt7915.h
@@ -310,6 +310,8 @@ struct mt7915_dev {
 
 	struct {
 		u32 state;
+		u32 wa_reset_count;
+		u32 wm_reset_count;
 		bool hw_full_reset:1;
 		bool hw_init_done:1;
 		bool restart:1;
diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/regs.h b/drivers/net/wireless/mediatek/mt76/mt7915/regs.h
index 16214cee5d93..dc360d52f49b 100644
--- a/drivers/net/wireless/mediatek/mt76/mt7915/regs.h
+++ b/drivers/net/wireless/mediatek/mt76/mt7915/regs.h
@@ -733,6 +733,8 @@ enum offs_rev {
 #define MT_MCU_CMD_NORMAL_STATE		BIT(5)
 #define MT_MCU_CMD_ERROR_MASK		GENMASK(5, 1)
 
+#define MT_MCU_CMD_WA_WDT		BIT(31)
+#define MT_MCU_CMD_WM_WDT		BIT(30)
 #define MT_MCU_CMD_WDT_MASK		GENMASK(31, 30)
 
 /* TOP RGU */
@@ -1151,5 +1153,7 @@ enum offs_rev {
 #define MT_MCU_WM_CIRQ(ofs)			(MT_MCU_WM_CIRQ_BASE + (ofs))
 #define MT_MCU_WM_CIRQ_IRQ_MASK_CLR_ADDR	MT_MCU_WM_CIRQ(0x80)
 #define MT_MCU_WM_CIRQ_IRQ_SOFT_ADDR		MT_MCU_WM_CIRQ(0xc0)
+#define MT_MCU_WM_CIRQ_EINT_MASK_CLR_ADDR	MT_MCU_WM_CIRQ(0x108)
+#define MT_MCU_WM_CIRQ_EINT_SOFT_ADDR		MT_MCU_WM_CIRQ(0x118)
 
 #endif