[Patch v6 0/4] memory: tegra: Add MC channels and error logging

[Patch v6 0/4] memory: tegra: Add MC channels and error logging

[Ashish Mhetre]

From tegra186 onward, memory controllers support multiple channels.
Add memory controller channels in device tree and add support to map
address spaces of these channels in tegra MC driver.
When memory controller interrupt occurs, registers from these channels
are required to be read in order to get error information.
Add error logging support from tegra186 onward for memory controller
interrupts.

Ashish Mhetre (4):
  memory: tegra: Add memory controller channels support
  memory: tegra: Add MC error logging on tegra186 onward
  dt-bindings: memory: Update reg maxitems for tegra186
  arm64: tegra: Add memory controller channels

---
Changes in v6:
- Added reg-names for each reg item of memory controller node
- Added logging for interrupts on multiple memory controller channels
- Added clearing interrupt support for global intstatus
- Updated DT binding documentation to work with existing DTS as well
- Updated function to get MC channels
- Updated variable names

Changes in v5:
- Updated patch sequence such that driver patches are before DT patches
- Fixed DT ABI break from v4
- Fixed smatch bug
- Updated description in DT binding documentation
- Updated variable names

Changes in v4:
- Added memory controller channels support
- Added newlines after every break statement of all switch cases
- Fixed compile error with W=1 build
- Fixed the interrupt mask bit logic

Changes in v3:
- Removed unnecessary ifdefs
- Grouped newly added MC registers with existing MC registers
- Removed unnecessary initialization of variables
- Updated code to use newly added field 'has_addr_hi_reg' instead of ifdefs

Changes in v2:
- Updated patch subject and commit message
- Removed separate irq handlers
- Updated tegra30_mc_handle_irq to be used for tegra186 onwards as well

 .../nvidia,tegra186-mc.yaml                   |  14 +-
 arch/arm64/boot/dts/nvidia/tegra186.dtsi      |   7 +-
 arch/arm64/boot/dts/nvidia/tegra194.dtsi      |  21 ++-
 arch/arm64/boot/dts/nvidia/tegra234.dtsi      |  21 ++-
 drivers/memory/tegra/mc.c                     | 120 +++++++++++++++---
 drivers/memory/tegra/mc.h                     |  37 +++++-
 drivers/memory/tegra/tegra186.c               |  98 ++++++++++++++
 drivers/memory/tegra/tegra194.c               |  45 +++++++
 drivers/memory/tegra/tegra234.c               |  64 ++++++++++
 include/soc/tegra/mc.h                        |  12 ++
 10 files changed, 412 insertions(+), 27 deletions(-)

-- 
2.17.1

[Patch v6 1/4] memory: tegra: Add memory controller channels support

[Ashish Mhetre]

From tegra186 onwards, memory controller support multiple channels.
Add support for mapping address spaces of these channels.
Make sure that number of channels are as expected on each SOC.
During error interrupts from memory controller, appropriate registers
from these channels need to be accessed for logging error info.

Signed-off-by: Ashish Mhetre <amhetre@nvidia.com>
---
 drivers/memory/tegra/mc.c       |  6 ++++
 drivers/memory/tegra/tegra186.c | 54 +++++++++++++++++++++++++++++++++
 drivers/memory/tegra/tegra194.c |  1 +
 drivers/memory/tegra/tegra234.c |  1 +
 include/soc/tegra/mc.h          |  7 +++++
 5 files changed, 69 insertions(+)

diff --git a/drivers/memory/tegra/mc.c b/drivers/memory/tegra/mc.c
index bf3abb6d8354..3cda1d9ad32a 100644
--- a/drivers/memory/tegra/mc.c
+++ b/drivers/memory/tegra/mc.c
@@ -749,6 +749,12 @@ static int tegra_mc_probe(struct platform_device *pdev)
 	if (IS_ERR(mc->regs))
 		return PTR_ERR(mc->regs);
 
+	if (mc->soc->ops && mc->soc->ops->map_regs) {
+		err = mc->soc->ops->map_regs(mc, pdev);
+		if (err < 0)
+			return err;
+	}
+
 	mc->debugfs.root = debugfs_create_dir("mc", NULL);
 
 	if (mc->soc->ops && mc->soc->ops->probe) {
diff --git a/drivers/memory/tegra/tegra186.c b/drivers/memory/tegra/tegra186.c
index 3d153881abc1..2ca8ce349188 100644
--- a/drivers/memory/tegra/tegra186.c
+++ b/drivers/memory/tegra/tegra186.c
@@ -139,11 +139,64 @@ static int tegra186_mc_probe_device(struct tegra_mc *mc, struct device *dev)
 	return 0;
 }
 
+static int tegra186_mc_map_regs(struct tegra_mc *mc,
+				struct platform_device *pdev)
+{
+	struct device_node *np = pdev->dev.parent->of_node;
+	int num_dt_channels, reg_cells = 0;
+	int i, ret;
+	u32 val;
+
+	ret = of_property_read_u32(np, "#address-cells", &val);
+	if (ret) {
+		dev_err(&pdev->dev, "missing #address-cells property\n");
+		return ret;
+	}
+
+	reg_cells = val;
+
+	ret = of_property_read_u32(np, "#size-cells", &val);
+	if (ret) {
+		dev_err(&pdev->dev, "missing #size-cells property\n");
+		return ret;
+	}
+
+	reg_cells += val;
+
+	num_dt_channels = of_property_count_elems_of_size(pdev->dev.of_node, "reg",
+							  reg_cells * sizeof(u32));
+	/*
+	 * On tegra186 onwards, memory controller support multiple channels.
+	 * Apart from regular memory controller channels, there is one broadcast
+	 * channel and one for stream-id registers.
+	 */
+	if (num_dt_channels < mc->soc->num_channels + 2) {
+		dev_warn(&pdev->dev, "MC channels are missing, please update memory controller DT node with MC channels\n");
+		return 0;
+	}
+
+	mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev, "mc-broadcast");
+	if (IS_ERR(mc->bcast_ch_regs))
+		return PTR_ERR(mc->bcast_ch_regs);
+
+	for (i = 0; i < mc->soc->num_channels; i++) {
+		char name[4];
+
+		sprintf(name, "mc%u", i);
+		mc->ch_regs[i] = devm_platform_ioremap_resource_byname(pdev, name);
+		if (IS_ERR(mc->ch_regs[i]))
+			return PTR_ERR(mc->ch_regs[i]);
+	}
+
+	return 0;
+}
+
 const struct tegra_mc_ops tegra186_mc_ops = {
 	.probe = tegra186_mc_probe,
 	.remove = tegra186_mc_remove,
 	.resume = tegra186_mc_resume,
 	.probe_device = tegra186_mc_probe_device,
+	.map_regs = tegra186_mc_map_regs,
 };
 
 #if defined(CONFIG_ARCH_TEGRA_186_SOC)
@@ -875,6 +928,7 @@ const struct tegra_mc_soc tegra186_mc_soc = {
 	.num_clients = ARRAY_SIZE(tegra186_mc_clients),
 	.clients = tegra186_mc_clients,
 	.num_address_bits = 40,
+	.num_channels = 4,
 	.ops = &tegra186_mc_ops,
 };
 #endif
diff --git a/drivers/memory/tegra/tegra194.c b/drivers/memory/tegra/tegra194.c
index cab998b8bd5c..94001174deaf 100644
--- a/drivers/memory/tegra/tegra194.c
+++ b/drivers/memory/tegra/tegra194.c
@@ -1347,5 +1347,6 @@ const struct tegra_mc_soc tegra194_mc_soc = {
 	.num_clients = ARRAY_SIZE(tegra194_mc_clients),
 	.clients = tegra194_mc_clients,
 	.num_address_bits = 40,
+	.num_channels = 16,
 	.ops = &tegra186_mc_ops,
 };
diff --git a/drivers/memory/tegra/tegra234.c b/drivers/memory/tegra/tegra234.c
index e22824a79f45..6335a132be2d 100644
--- a/drivers/memory/tegra/tegra234.c
+++ b/drivers/memory/tegra/tegra234.c
@@ -97,5 +97,6 @@ const struct tegra_mc_soc tegra234_mc_soc = {
 	.num_clients = ARRAY_SIZE(tegra234_mc_clients),
 	.clients = tegra234_mc_clients,
 	.num_address_bits = 40,
+	.num_channels = 16,
 	.ops = &tegra186_mc_ops,
 };
diff --git a/include/soc/tegra/mc.h b/include/soc/tegra/mc.h
index 1066b1194a5a..c3c121fbfbb7 100644
--- a/include/soc/tegra/mc.h
+++ b/include/soc/tegra/mc.h
@@ -13,6 +13,9 @@
 #include <linux/irq.h>
 #include <linux/reset-controller.h>
 #include <linux/types.h>
+#include <linux/platform_device.h>
+
+#define MC_MAX_CHANNELS 16
 
 struct clk;
 struct device;
@@ -181,6 +184,7 @@ struct tegra_mc_ops {
 	int (*resume)(struct tegra_mc *mc);
 	irqreturn_t (*handle_irq)(int irq, void *data);
 	int (*probe_device)(struct tegra_mc *mc, struct device *dev);
+	int (*map_regs)(struct tegra_mc *mc, struct platform_device *pdev);
 };
 
 struct tegra_mc_soc {
@@ -194,6 +198,7 @@ struct tegra_mc_soc {
 	unsigned int atom_size;
 
 	u8 client_id_mask;
+	u8 num_channels;
 
 	const struct tegra_smmu_soc *smmu;
 
@@ -212,6 +217,8 @@ struct tegra_mc {
 	struct tegra_smmu *smmu;
 	struct gart_device *gart;
 	void __iomem *regs;
+	void __iomem *bcast_ch_regs;
+	void __iomem *ch_regs[MC_MAX_CHANNELS];
 	struct clk *clk;
 	int irq;
 
-- 
2.17.1

[Patch v6 2/4] memory: tegra: Add MC error logging on tegra186 onward

[Ashish Mhetre]

Add support for reading MC_GLOBAL_INTSTATUS register which points to the
memory controller channels on which interrupts have occurred.
Add new function 'global_intstatus_to_channel' which returns the channel
which should be used to get the information of interrupts.
Remove static from tegra30_mc_handle_irq and use it as interrupt handler
for MC interrupts on tegra186, tegra194 and tegra234 to log the errors.
Add error specific MC status and address register bits and use them on
tegra186, tegra194 and tegra234.
Add error logging for generalized carveout interrupt on tegra186, tegra194
and tegra234.
Add error logging for route sanity interrupt on tegra194 an tegra234.
Add register for higher bits of error address which is available on
tegra194 and tegra234.
Add a boolean variable 'has_addr_hi_reg' in tegra_mc_soc struture which
will be true if soc has register for higher bits of memory controller
error address. Set it true for tegra194 and tegra234.
Clear the global_intstatus bit corresponding to channel of which
interrupts are logged.

Signed-off-by: Ashish Mhetre <amhetre@nvidia.com>
---
 drivers/memory/tegra/mc.c       | 128 ++++++++++++++++++++++++++++----
 drivers/memory/tegra/mc.h       |  37 ++++++++-
 drivers/memory/tegra/tegra186.c |   9 +++
 drivers/memory/tegra/tegra194.c |   8 ++
 drivers/memory/tegra/tegra234.c |   7 ++
 include/soc/tegra/mc.h          |   3 +
 6 files changed, 175 insertions(+), 17 deletions(-)

diff --git a/drivers/memory/tegra/mc.c b/drivers/memory/tegra/mc.c
index 3cda1d9ad32a..6f4e29d4bd33 100644
--- a/drivers/memory/tegra/mc.c
+++ b/drivers/memory/tegra/mc.c
@@ -508,14 +508,48 @@ int tegra30_mc_probe(struct tegra_mc *mc)
 	return 0;
 }
 
-static irqreturn_t tegra30_mc_handle_irq(int irq, void *data)
+const struct tegra_mc_ops tegra30_mc_ops = {
+	.probe = tegra30_mc_probe,
+	.handle_irq = tegra30_mc_handle_irq,
+};
+#endif
+
+static int global_intstatus_to_channel(const struct tegra_mc *mc, u32 status,
+				       unsigned int *mc_channel)
+{
+	if ((status & mc->soc->ch_intmask) == 0)
+		return -EINVAL;
+
+	*mc_channel = __ffs((status & mc->soc->ch_intmask) >>
+			    mc->soc->status_reg_chan_shift);
+
+	return 0;
+}
+
+irqreturn_t tegra30_mc_handle_irq(int irq, void *data)
 {
 	struct tegra_mc *mc = data;
+	unsigned int bit, channel;
 	unsigned long status;
-	unsigned int bit;
 
-	/* mask all interrupts to avoid flooding */
-	status = mc_readl(mc, MC_INTSTATUS) & mc->soc->intmask;
+	if (mc->soc->num_channels) {
+		u32 global_status;
+		int err;
+
+		global_status = mc_ch_readl(mc, MC_BROADCAST_CHANNEL, MC_GLOBAL_INTSTATUS);
+		err = global_intstatus_to_channel(mc, global_status, &channel);
+		if (err < 0) {
+			dev_err_ratelimited(mc->dev, "unknown interrupt channel 0x%08x\n",
+					    global_status);
+			return IRQ_NONE;
+		}
+
+		/* mask all interrupts to avoid flooding */
+		status = mc_ch_readl(mc, channel, MC_INTSTATUS) & mc->soc->intmask;
+	} else {
+		status = mc_readl(mc, MC_INTSTATUS) & mc->soc->intmask;
+	}
+
 	if (!status)
 		return IRQ_NONE;
 
@@ -523,18 +557,70 @@ static irqreturn_t tegra30_mc_handle_irq(int irq, void *data)
 		const char *error = tegra_mc_status_names[bit] ?: "unknown";
 		const char *client = "unknown", *desc;
 		const char *direction, *secure;
+		u32 status_reg, addr_reg;
+		u32 intmask = BIT(bit);
 		phys_addr_t addr = 0;
+#ifdef CONFIG_PHYS_ADDR_T_64BIT
+		u32 addr_hi_reg = 0;
+#endif
 		unsigned int i;
 		char perm[7];
 		u8 id, type;
 		u32 value;
 
-		value = mc_readl(mc, MC_ERR_STATUS);
+		switch (intmask) {
+		case MC_INT_DECERR_VPR:
+			status_reg = MC_ERR_VPR_STATUS;
+			addr_reg = MC_ERR_VPR_ADR;
+			break;
+
+		case MC_INT_SECERR_SEC:
+			status_reg = MC_ERR_SEC_STATUS;
+			addr_reg = MC_ERR_SEC_ADR;
+			break;
+
+		case MC_INT_DECERR_MTS:
+			status_reg = MC_ERR_MTS_STATUS;
+			addr_reg = MC_ERR_MTS_ADR;
+			break;
+
+		case MC_INT_DECERR_GENERALIZED_CARVEOUT:
+			status_reg = MC_ERR_GENERALIZED_CARVEOUT_STATUS;
+			addr_reg = MC_ERR_GENERALIZED_CARVEOUT_ADR;
+			break;
+
+		case MC_INT_DECERR_ROUTE_SANITY:
+			status_reg = MC_ERR_ROUTE_SANITY_STATUS;
+			addr_reg = MC_ERR_ROUTE_SANITY_ADR;
+			break;
+
+		default:
+			status_reg = MC_ERR_STATUS;
+			addr_reg = MC_ERR_ADR;
+
+#ifdef CONFIG_PHYS_ADDR_T_64BIT
+			if (mc->soc->has_addr_hi_reg)
+				addr_hi_reg = MC_ERR_ADR_HI;
+#endif
+			break;
+		}
+
+		if (mc->soc->num_channels)
+			value = mc_ch_readl(mc, channel, status_reg);
+		else
+			value = mc_readl(mc, status_reg);
 
 #ifdef CONFIG_PHYS_ADDR_T_64BIT
 		if (mc->soc->num_address_bits > 32) {
-			addr = ((value >> MC_ERR_STATUS_ADR_HI_SHIFT) &
-				MC_ERR_STATUS_ADR_HI_MASK);
+			if (addr_hi_reg) {
+				if (mc->soc->num_channels)
+					addr = mc_ch_readl(mc, channel, addr_hi_reg);
+				else
+					addr = mc_readl(mc, addr_hi_reg);
+			} else {
+				addr = ((value >> MC_ERR_STATUS_ADR_HI_SHIFT) &
+					MC_ERR_STATUS_ADR_HI_MASK);
+			}
 			addr <<= 32;
 		}
 #endif
@@ -591,7 +677,10 @@ static irqreturn_t tegra30_mc_handle_irq(int irq, void *data)
 			break;
 		}
 
-		value = mc_readl(mc, MC_ERR_ADR);
+		if (mc->soc->num_channels)
+			value = mc_ch_readl(mc, channel, addr_reg);
+		else
+			value = mc_readl(mc, addr_reg);
 		addr |= value;
 
 		dev_err_ratelimited(mc->dev, "%s: %s%s @%pa: %s (%s%s)\n",
@@ -600,17 +689,18 @@ static irqreturn_t tegra30_mc_handle_irq(int irq, void *data)
 	}
 
 	/* clear interrupts */
-	mc_writel(mc, status, MC_INTSTATUS);
+	if (mc->soc->num_channels) {
+		u32 status_chan_bit;
+
+		mc_ch_writel(mc, channel, status, MC_INTSTATUS);
+		status_chan_bit = BIT(channel) << mc->soc->status_reg_chan_shift;
+		mc_ch_writel(mc, MC_BROADCAST_CHANNEL, status_chan_bit, MC_GLOBAL_INTSTATUS);
+	} else
+		mc_writel(mc, status, MC_INTSTATUS);
 
 	return IRQ_HANDLED;
 }
 
-const struct tegra_mc_ops tegra30_mc_ops = {
-	.probe = tegra30_mc_probe,
-	.handle_irq = tegra30_mc_handle_irq,
-};
-#endif
-
 const char *const tegra_mc_status_names[32] = {
 	[ 1] = "External interrupt",
 	[ 6] = "EMEM address decode error",
@@ -622,6 +712,8 @@ const char *const tegra_mc_status_names[32] = {
 	[12] = "VPR violation",
 	[13] = "Secure carveout violation",
 	[16] = "MTS carveout violation",
+	[17] = "Generalized carveout violation",
+	[20] = "Route Sanity error",
 };
 
 const char *const tegra_mc_error_names[8] = {
@@ -770,7 +862,11 @@ static int tegra_mc_probe(struct platform_device *pdev)
 
 		WARN(!mc->soc->client_id_mask, "missing client ID mask for this SoC\n");
 
-		mc_writel(mc, mc->soc->intmask, MC_INTMASK);
+		if (mc->soc->num_channels && mc->bcast_ch_regs)
+			mc_ch_writel(mc, MC_BROADCAST_CHANNEL, mc->soc->intmask,
+				     MC_INTMASK);
+		else
+			mc_writel(mc, mc->soc->intmask, MC_INTMASK);
 
 		err = devm_request_irq(&pdev->dev, mc->irq, mc->soc->ops->handle_irq, 0,
 				       dev_name(&pdev->dev), mc);
diff --git a/drivers/memory/tegra/mc.h b/drivers/memory/tegra/mc.h
index 062886e94c04..77b3873e245c 100644
--- a/drivers/memory/tegra/mc.h
+++ b/drivers/memory/tegra/mc.h
@@ -43,7 +43,21 @@
 #define MC_EMEM_ARB_OVERRIDE				0xe8
 #define MC_TIMING_CONTROL_DBG				0xf8
 #define MC_TIMING_CONTROL				0xfc
-
+#define MC_ERR_VPR_STATUS				0x654
+#define MC_ERR_VPR_ADR					0x658
+#define MC_ERR_SEC_STATUS				0x67c
+#define MC_ERR_SEC_ADR					0x680
+#define MC_ERR_MTS_STATUS				0x9b0
+#define MC_ERR_MTS_ADR					0x9b4
+#define MC_ERR_ROUTE_SANITY_STATUS			0x9c0
+#define MC_ERR_ROUTE_SANITY_ADR				0x9c4
+#define MC_ERR_GENERALIZED_CARVEOUT_STATUS		0xc00
+#define MC_ERR_GENERALIZED_CARVEOUT_ADR			0xc04
+#define MC_GLOBAL_INTSTATUS				0xf24
+#define MC_ERR_ADR_HI					0x11fc
+
+#define MC_INT_DECERR_ROUTE_SANITY			BIT(20)
+#define MC_INT_DECERR_GENERALIZED_CARVEOUT		BIT(17)
 #define MC_INT_DECERR_MTS				BIT(16)
 #define MC_INT_SECERR_SEC				BIT(13)
 #define MC_INT_DECERR_VPR				BIT(12)
@@ -78,6 +92,8 @@
 
 #define MC_TIMING_UPDATE				BIT(0)
 
+#define MC_BROADCAST_CHANNEL				~0
+
 static inline u32 tegra_mc_scale_percents(u64 val, unsigned int percents)
 {
 	val = val * percents;
@@ -92,6 +108,24 @@ icc_provider_to_tegra_mc(struct icc_provider *provider)
 	return container_of(provider, struct tegra_mc, provider);
 }
 
+static inline u32 mc_ch_readl(const struct tegra_mc *mc, int ch,
+			       unsigned long offset)
+{
+	if (ch == MC_BROADCAST_CHANNEL)
+		return readl_relaxed(mc->bcast_ch_regs + offset);
+
+	return readl_relaxed(mc->ch_regs[ch] + offset);
+}
+
+static inline void mc_ch_writel(const struct tegra_mc *mc, int ch,
+				u32 value, unsigned long offset)
+{
+	if (ch == MC_BROADCAST_CHANNEL)
+		writel_relaxed(value, mc->bcast_ch_regs + offset);
+	else
+		writel_relaxed(value, mc->ch_regs[ch] + offset);
+}
+
 static inline u32 mc_readl(const struct tegra_mc *mc, unsigned long offset)
 {
 	return readl_relaxed(mc->regs + offset);
@@ -156,6 +190,7 @@ extern const struct tegra_mc_ops tegra30_mc_ops;
 extern const struct tegra_mc_ops tegra186_mc_ops;
 #endif
 
+irqreturn_t tegra30_mc_handle_irq(int irq, void *data);
 extern const char * const tegra_mc_status_names[32];
 extern const char * const tegra_mc_error_names[8];
 
diff --git a/drivers/memory/tegra/tegra186.c b/drivers/memory/tegra/tegra186.c
index 2ca8ce349188..bb2cc405eb20 100644
--- a/drivers/memory/tegra/tegra186.c
+++ b/drivers/memory/tegra/tegra186.c
@@ -16,6 +16,8 @@
 #include <dt-bindings/memory/tegra186-mc.h>
 #endif
 
+#include "mc.h"
+
 #define MC_SID_STREAMID_OVERRIDE_MASK GENMASK(7, 0)
 #define MC_SID_STREAMID_SECURITY_WRITE_ACCESS_DISABLED BIT(16)
 #define MC_SID_STREAMID_SECURITY_OVERRIDE BIT(8)
@@ -197,6 +199,7 @@ const struct tegra_mc_ops tegra186_mc_ops = {
 	.resume = tegra186_mc_resume,
 	.probe_device = tegra186_mc_probe_device,
 	.map_regs = tegra186_mc_map_regs,
+	.handle_irq = tegra30_mc_handle_irq,
 };
 
 #if defined(CONFIG_ARCH_TEGRA_186_SOC)
@@ -929,6 +932,12 @@ const struct tegra_mc_soc tegra186_mc_soc = {
 	.clients = tegra186_mc_clients,
 	.num_address_bits = 40,
 	.num_channels = 4,
+	.client_id_mask = 0xff,
+	.intmask = MC_INT_DECERR_GENERALIZED_CARVEOUT | MC_INT_DECERR_MTS |
+		   MC_INT_SECERR_SEC | MC_INT_DECERR_VPR |
+		   MC_INT_SECURITY_VIOLATION | MC_INT_DECERR_EMEM,
 	.ops = &tegra186_mc_ops,
+	.ch_intmask = 0x0000000f,
+	.status_reg_chan_shift = 0,
 };
 #endif
diff --git a/drivers/memory/tegra/tegra194.c b/drivers/memory/tegra/tegra194.c
index 94001174deaf..963d6085a247 100644
--- a/drivers/memory/tegra/tegra194.c
+++ b/drivers/memory/tegra/tegra194.c
@@ -1348,5 +1348,13 @@ const struct tegra_mc_soc tegra194_mc_soc = {
 	.clients = tegra194_mc_clients,
 	.num_address_bits = 40,
 	.num_channels = 16,
+	.client_id_mask = 0xff,
+	.intmask = MC_INT_DECERR_ROUTE_SANITY |
+		   MC_INT_DECERR_GENERALIZED_CARVEOUT | MC_INT_DECERR_MTS |
+		   MC_INT_SECERR_SEC | MC_INT_DECERR_VPR |
+		   MC_INT_SECURITY_VIOLATION | MC_INT_DECERR_EMEM,
+	.has_addr_hi_reg = true,
 	.ops = &tegra186_mc_ops,
+	.ch_intmask = 0x00000f00,
+	.status_reg_chan_shift = 8,
 };
diff --git a/drivers/memory/tegra/tegra234.c b/drivers/memory/tegra/tegra234.c
index 6335a132be2d..44884fb78b5c 100644
--- a/drivers/memory/tegra/tegra234.c
+++ b/drivers/memory/tegra/tegra234.c
@@ -98,5 +98,12 @@ const struct tegra_mc_soc tegra234_mc_soc = {
 	.clients = tegra234_mc_clients,
 	.num_address_bits = 40,
 	.num_channels = 16,
+	.intmask = MC_INT_DECERR_ROUTE_SANITY |
+		   MC_INT_DECERR_GENERALIZED_CARVEOUT | MC_INT_DECERR_MTS |
+		   MC_INT_SECERR_SEC | MC_INT_DECERR_VPR |
+		   MC_INT_SECURITY_VIOLATION | MC_INT_DECERR_EMEM,
+	.has_addr_hi_reg = true,
 	.ops = &tegra186_mc_ops,
+	.ch_intmask = 0x0000ff00,
+	.status_reg_chan_shift = 8,
 };
diff --git a/include/soc/tegra/mc.h b/include/soc/tegra/mc.h
index c3c121fbfbb7..284ad5988d10 100644
--- a/include/soc/tegra/mc.h
+++ b/include/soc/tegra/mc.h
@@ -203,6 +203,9 @@ struct tegra_mc_soc {
 	const struct tegra_smmu_soc *smmu;
 
 	u32 intmask;
+	u32 ch_intmask;
+	u32 status_reg_chan_shift;
+	bool has_addr_hi_reg;
 
 	const struct tegra_mc_reset_ops *reset_ops;
 	const struct tegra_mc_reset *resets;
-- 
2.17.1

[Patch v6 3/4] dt-bindings: memory: Update reg maxitems for tegra186

[Ashish Mhetre]

From tegra186 onwards, memory controller support multiple channels.
Reg items are updated with address and size of these channels.
Tegra186 has overall 5 memory controller channels. Tegra194 and tegra234
have overall 17 memory controller channels each.
There is 1 reg item for memory controller stream-id registers.
So update the reg maxItems to 18 in tegra186 devicetree documentation.

Signed-off-by: Ashish Mhetre <amhetre@nvidia.com>
---
 .../memory-controllers/nvidia,tegra186-mc.yaml    | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/Documentation/devicetree/bindings/memory-controllers/nvidia,tegra186-mc.yaml b/Documentation/devicetree/bindings/memory-controllers/nvidia,tegra186-mc.yaml
index 13c4c82fd0d3..0fe396a2e162 100644
--- a/Documentation/devicetree/bindings/memory-controllers/nvidia,tegra186-mc.yaml
+++ b/Documentation/devicetree/bindings/memory-controllers/nvidia,tegra186-mc.yaml
@@ -35,7 +35,7 @@ properties:
 
   reg:
     minItems: 1
-    maxItems: 3
+    maxItems: 18
 
   interrupts:
     items:
@@ -142,7 +142,8 @@ allOf:
     then:
       properties:
         reg:
-          maxItems: 1
+          maxItems: 6
+          description: 5 memory controller channels and 1 for stream-id registers
 
   - if:
       properties:
@@ -152,6 +153,7 @@ allOf:
       properties:
         reg:
           minItems: 3
+          description: 17 memory controller channels and 1 for stream-id registers
 
   - if:
       properties:
@@ -161,6 +163,7 @@ allOf:
       properties:
         reg:
           minItems: 3
+          description: 17 memory controller channels and 1 for stream-id registers
 
 additionalProperties: false
 
@@ -182,7 +185,13 @@ examples:
 
         memory-controller@2c00000 {
             compatible = "nvidia,tegra186-mc";
-            reg = <0x0 0x02c00000 0x0 0xb0000>;
+            reg = <0x0 0x02c00000 0x0 0x10000>,    /* MC-SID */
+                  <0x0 0x02c10000 0x0 0x10000>,    /* Broadcast channel */
+                  <0x0 0x02c20000 0x0 0x10000>,    /* MC0 */
+                  <0x0 0x02c30000 0x0 0x10000>,    /* MC1 */
+                  <0x0 0x02c40000 0x0 0x10000>,    /* MC2 */
+                  <0x0 0x02c50000 0x0 0x10000>;    /* MC3 */
+            reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1", "mc2", "mc3";
             interrupts = <GIC_SPI 223 IRQ_TYPE_LEVEL_HIGH>;
 
             #address-cells = <2>;
-- 
2.17.1

[Patch v6 4/4] arm64: tegra: Add memory controller channels

[Ashish Mhetre]

From tegra186 onwards, memory controller support multiple channels.
During the error interrupts from memory controller, corresponding
channels need to be accessed for logging error info and clearing the
interrupt.
So add address and size of these channels in device tree node of
tegra186, tegra194 and tegra234 memory controller. Also add reg-names
for each of these reg items which are used by driver for mapping.

Signed-off-by: Ashish Mhetre <amhetre@nvidia.com>
---
 arch/arm64/boot/dts/nvidia/tegra186.dtsi |  8 +++++++-
 arch/arm64/boot/dts/nvidia/tegra194.dtsi | 24 +++++++++++++++++++++---
 arch/arm64/boot/dts/nvidia/tegra234.dtsi | 24 +++++++++++++++++++++---
 3 files changed, 49 insertions(+), 7 deletions(-)

diff --git a/arch/arm64/boot/dts/nvidia/tegra186.dtsi b/arch/arm64/boot/dts/nvidia/tegra186.dtsi
index e9b40f5d79ec..e4499db46339 100644
--- a/arch/arm64/boot/dts/nvidia/tegra186.dtsi
+++ b/arch/arm64/boot/dts/nvidia/tegra186.dtsi
@@ -521,7 +521,13 @@
 
 	mc: memory-controller@2c00000 {
 		compatible = "nvidia,tegra186-mc";
-		reg = <0x0 0x02c00000 0x0 0xb0000>;
+		reg = <0x0 0x02c00000 0x0 0x10000>,    /* MC-SID */
+		      <0x0 0x02c10000 0x0 0x10000>,    /* Broadcast channel */
+		      <0x0 0x02c20000 0x0 0x10000>,    /* MC0 */
+		      <0x0 0x02c30000 0x0 0x10000>,    /* MC1 */
+		      <0x0 0x02c40000 0x0 0x10000>,    /* MC2 */
+		      <0x0 0x02c50000 0x0 0x10000>;    /* MC3 */
+		reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1", "mc2", "mc3";
 		interrupts = <GIC_SPI 223 IRQ_TYPE_LEVEL_HIGH>;
 		status = "disabled";
 
diff --git a/arch/arm64/boot/dts/nvidia/tegra194.dtsi b/arch/arm64/boot/dts/nvidia/tegra194.dtsi
index 751ebe5e9506..88a1a5e426ff 100644
--- a/arch/arm64/boot/dts/nvidia/tegra194.dtsi
+++ b/arch/arm64/boot/dts/nvidia/tegra194.dtsi
@@ -604,9 +604,27 @@
 
 		mc: memory-controller@2c00000 {
 			compatible = "nvidia,tegra194-mc";
-			reg = <0x02c00000 0x100000>,
-			      <0x02b80000 0x040000>,
-			      <0x01700000 0x100000>;
+			reg = <0x02c00000 0x10000>,   /* MC-SID */
+			      <0x02c10000 0x10000>,   /* MC Broadcast*/
+			      <0x02c20000 0x10000>,   /* MC0 */
+			      <0x02c30000 0x10000>,   /* MC1 */
+			      <0x02c40000 0x10000>,   /* MC2 */
+			      <0x02c50000 0x10000>,   /* MC3 */
+			      <0x02b80000 0x10000>,   /* MC4 */
+			      <0x02b90000 0x10000>,   /* MC5 */
+			      <0x02ba0000 0x10000>,   /* MC6 */
+			      <0x02bb0000 0x10000>,   /* MC7 */
+			      <0x01700000 0x10000>,   /* MC8 */
+			      <0x01710000 0x10000>,   /* MC9 */
+			      <0x01720000 0x10000>,   /* MC10 */
+			      <0x01730000 0x10000>,   /* MC11 */
+			      <0x01740000 0x10000>,   /* MC12 */
+			      <0x01750000 0x10000>,   /* MC13 */
+			      <0x01760000 0x10000>,   /* MC14 */
+			      <0x01770000 0x10000>;   /* MC15 */
+			reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1", "mc2", "mc3",
+				    "mc4", "mc5", "mc6", "mc7", "mc8", "mc9", "mc10",
+				    "mc11", "mc12", "mc13", "mc14", "mc15";
 			interrupts = <GIC_SPI 223 IRQ_TYPE_LEVEL_HIGH>;
 			#interconnect-cells = <1>;
 			status = "disabled";
diff --git a/arch/arm64/boot/dts/nvidia/tegra234.dtsi b/arch/arm64/boot/dts/nvidia/tegra234.dtsi
index aaace605bdaa..216a079ba569 100644
--- a/arch/arm64/boot/dts/nvidia/tegra234.dtsi
+++ b/arch/arm64/boot/dts/nvidia/tegra234.dtsi
@@ -507,9 +507,27 @@
 
 		mc: memory-controller@2c00000 {
 			compatible = "nvidia,tegra234-mc";
-			reg = <0x02c00000 0x100000>,
-			      <0x02b80000 0x040000>,
-			      <0x01700000 0x100000>;
+			reg = <0x02c00000 0x10000>,   /* MC-SID */
+			      <0x02c10000 0x10000>,   /* MC Broadcast*/
+			      <0x02c20000 0x10000>,   /* MC0 */
+			      <0x02c30000 0x10000>,   /* MC1 */
+			      <0x02c40000 0x10000>,   /* MC2 */
+			      <0x02c50000 0x10000>,   /* MC3 */
+			      <0x02b80000 0x10000>,   /* MC4 */
+			      <0x02b90000 0x10000>,   /* MC5 */
+			      <0x02ba0000 0x10000>,   /* MC6 */
+			      <0x02bb0000 0x10000>,   /* MC7 */
+			      <0x01700000 0x10000>,   /* MC8 */
+			      <0x01710000 0x10000>,   /* MC9 */
+			      <0x01720000 0x10000>,   /* MC10 */
+			      <0x01730000 0x10000>,   /* MC11 */
+			      <0x01740000 0x10000>,   /* MC12 */
+			      <0x01750000 0x10000>,   /* MC13 */
+			      <0x01760000 0x10000>,   /* MC14 */
+			      <0x01770000 0x10000>;   /* MC15 */
+			reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1", "mc2", "mc3",
+				    "mc4", "mc5", "mc6", "mc7", "mc8", "mc9", "mc10",
+				    "mc11", "mc12", "mc13", "mc14", "mc15";
 			interrupts = <GIC_SPI 223 IRQ_TYPE_LEVEL_HIGH>;
 			#interconnect-cells = <1>;
 			status = "okay";
-- 
2.17.1

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Dmitry Osipenko]

06.04.2022 08:24, Ashish Mhetre пишет:
> +	num_dt_channels = of_property_count_elems_of_size(pdev->dev.of_node, "reg",
> +							  reg_cells * sizeof(u32));
> +	/*
> +	 * On tegra186 onwards, memory controller support multiple channels.
> +	 * Apart from regular memory controller channels, there is one broadcast
> +	 * channel and one for stream-id registers.
> +	 */
> +	if (num_dt_channels < mc->soc->num_channels + 2) {
> +		dev_warn(&pdev->dev, "MC channels are missing, please update memory controller DT node with MC channels\n");
> +		return 0;
> +	}
> +
> +	mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev, "mc-broadcast");
> +	if (IS_ERR(mc->bcast_ch_regs))
> +		return PTR_ERR(mc->bcast_ch_regs);

Looks to me that you don't need to use of_property_count_elems_of_size()
and could only check the "mc-broadcast" presence to decide whether this
is an older DT.

mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
"broadcast");
if (IS_ERR(mc->bcast_ch_regs)) {
	dev_warn(&pdev->dev, "Broadcast channel is missing, please update your
device-tree\n");
	return PTR_ERR(mc->bcast_ch_regs);
}

Re: [Patch v6 3/4] dt-bindings: memory: Update reg maxitems for tegra186

[Dmitry Osipenko]

06.04.2022 08:24, Ashish Mhetre пишет:
>          memory-controller@2c00000 {
>              compatible = "nvidia,tegra186-mc";
> -            reg = <0x0 0x02c00000 0x0 0xb0000>;
> +            reg = <0x0 0x02c00000 0x0 0x10000>,    /* MC-SID */
> +                  <0x0 0x02c10000 0x0 0x10000>,    /* Broadcast channel */
> +                  <0x0 0x02c20000 0x0 0x10000>,    /* MC0 */
> +                  <0x0 0x02c30000 0x0 0x10000>,    /* MC1 */
> +                  <0x0 0x02c40000 0x0 0x10000>,    /* MC2 */
> +                  <0x0 0x02c50000 0x0 0x10000>;    /* MC3 */
> +            reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1", "mc2", "mc3";

The "mc-" prefix feels redundant to me, I'd name the regs like this:

  "sid", "broadcast", "ch0", "ch1", "ch2", "ch3"


You should also add validation of the regs/reg-names to the yaml based
on SoC version. I.e. it's not enough to only bump the maxItems.

Re: [Patch v6 2/4] memory: tegra: Add MC error logging on tegra186 onward

[Dmitry Osipenko]

06.04.2022 08:24, Ashish Mhetre пишет:
>  	/* clear interrupts */
> -	mc_writel(mc, status, MC_INTSTATUS);
> +	if (mc->soc->num_channels) {
> +		u32 status_chan_bit;
> +
> +		mc_ch_writel(mc, channel, status, MC_INTSTATUS);
> +		status_chan_bit = BIT(channel) << mc->soc->status_reg_chan_shift;

s/status_reg_chan_shift/global_intstatus_channel_shift/

> +		mc_ch_writel(mc, MC_BROADCAST_CHANNEL, status_chan_bit, MC_GLOBAL_INTSTATUS);
and maybe add a one-line mc_channel_to_global_intstatus(mc, channel) helper

mc_ch_writel(mc, MC_BROADCAST_CHANNEL,
	     mc_channel_to_global_intstatus(mc, chan), 
	     MC_GLOBAL_INTSTATUS);

> +	} else
> +		mc_writel(mc, status, MC_INTSTATUS);

Missing braces after "else", "./scripts/checkpatch.pl --strict" should warn about this.

Otherwise this patch looks okay at a quick glance.

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Dmitry Osipenko]

06.04.2022 08:24, Ashish Mhetre пишет:
> diff --git a/include/soc/tegra/mc.h b/include/soc/tegra/mc.h
> index 1066b1194a5a..c3c121fbfbb7 100644
> --- a/include/soc/tegra/mc.h
> +++ b/include/soc/tegra/mc.h
> @@ -13,6 +13,9 @@
>  #include <linux/irq.h>
>  #include <linux/reset-controller.h>
>  #include <linux/types.h>
> +#include <linux/platform_device.h>
> +
> +#define MC_MAX_CHANNELS 16
>  
>  struct clk;
>  struct device;
> @@ -181,6 +184,7 @@ struct tegra_mc_ops {
>  	int (*resume)(struct tegra_mc *mc);
>  	irqreturn_t (*handle_irq)(int irq, void *data);
>  	int (*probe_device)(struct tegra_mc *mc, struct device *dev);
> +	int (*map_regs)(struct tegra_mc *mc, struct platform_device *pdev);

Use to_platform_device(mc->dev) instead of passing the pdev argument.

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Dmitry Osipenko]

06.04.2022 08:24, Ashish Mhetre пишет:
> diff --git a/include/soc/tegra/mc.h b/include/soc/tegra/mc.h
> index 1066b1194a5a..c3c121fbfbb7 100644
> --- a/include/soc/tegra/mc.h
> +++ b/include/soc/tegra/mc.h
> @@ -13,6 +13,9 @@
>  #include <linux/irq.h>
>  #include <linux/reset-controller.h>
>  #include <linux/types.h>
> +#include <linux/platform_device.h>
> +
> +#define MC_MAX_CHANNELS 16
>  
>  struct clk;
>  struct device;
> @@ -181,6 +184,7 @@ struct tegra_mc_ops {
>  	int (*resume)(struct tegra_mc *mc);
>  	irqreturn_t (*handle_irq)(int irq, void *data);
>  	int (*probe_device)(struct tegra_mc *mc, struct device *dev);
> +	int (*map_regs)(struct tegra_mc *mc, struct platform_device *pdev);
>  };
>  
>  struct tegra_mc_soc {
> @@ -194,6 +198,7 @@ struct tegra_mc_soc {
>  	unsigned int atom_size;
>  
>  	u8 client_id_mask;
> +	u8 num_channels;
>  
>  	const struct tegra_smmu_soc *smmu;
>  
> @@ -212,6 +217,8 @@ struct tegra_mc {
>  	struct tegra_smmu *smmu;
>  	struct gart_device *gart;
>  	void __iomem *regs;
> +	void __iomem *bcast_ch_regs;
> +	void __iomem *ch_regs[MC_MAX_CHANNELS];

Why not to allocate ch_regs at runtime?

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Ashish Mhetre]

On 4/10/2022 7:48 PM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> 06.04.2022 08:24, Ashish Mhetre пишет:
>> +     num_dt_channels = of_property_count_elems_of_size(pdev->dev.of_node, "reg",
>> +                                                       reg_cells * sizeof(u32));
>> +     /*
>> +      * On tegra186 onwards, memory controller support multiple channels.
>> +      * Apart from regular memory controller channels, there is one broadcast
>> +      * channel and one for stream-id registers.
>> +      */
>> +     if (num_dt_channels < mc->soc->num_channels + 2) {
>> +             dev_warn(&pdev->dev, "MC channels are missing, please update memory controller DT node with MC channels\n");
>> +             return 0;
>> +     }
>> +
>> +     mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev, "mc-broadcast");
>> +     if (IS_ERR(mc->bcast_ch_regs))
>> +             return PTR_ERR(mc->bcast_ch_regs);
> 
> Looks to me that you don't need to use of_property_count_elems_of_size()
> and could only check the "mc-broadcast" presence to decide whether this
> is an older DT.
> 
Now that we are using reg-names in new DT, yes it'd be fine to just
check mc-broadcast to decide it's a new or old DT.

> mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
> "broadcast");
> if (IS_ERR(mc->bcast_ch_regs)) {
>          dev_warn(&pdev->dev, "Broadcast channel is missing, please update your
> device-tree\n");
>          return PTR_ERR(mc->bcast_ch_regs);
> }

return 0;

to avoid DT ABI break, right?

Re: [Patch v6 3/4] dt-bindings: memory: Update reg maxitems for tegra186

[Ashish Mhetre]

On 4/10/2022 7:51 PM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> 06.04.2022 08:24, Ashish Mhetre пишет:
>>           memory-controller@2c00000 {
>>               compatible = "nvidia,tegra186-mc";
>> -            reg = <0x0 0x02c00000 0x0 0xb0000>;
>> +            reg = <0x0 0x02c00000 0x0 0x10000>,    /* MC-SID */
>> +                  <0x0 0x02c10000 0x0 0x10000>,    /* Broadcast channel */
>> +                  <0x0 0x02c20000 0x0 0x10000>,    /* MC0 */
>> +                  <0x0 0x02c30000 0x0 0x10000>,    /* MC1 */
>> +                  <0x0 0x02c40000 0x0 0x10000>,    /* MC2 */
>> +                  <0x0 0x02c50000 0x0 0x10000>;    /* MC3 */
>> +            reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1", "mc2", "mc3";
> 
> The "mc-" prefix feels redundant to me, I'd name the regs like this:
> 
>    "sid", "broadcast", "ch0", "ch1", "ch2", "ch3"
> 
Okay, I'll update the reg-names by omitting "mc-".
> 
> You should also add validation of the regs/reg-names to the yaml based
> on SoC version. I.e. it's not enough to only bump the maxItems.

Okay, I'll add validation of reg-names as well.

Re: [Patch v6 2/4] memory: tegra: Add MC error logging on tegra186 onward

[Ashish Mhetre]

On 4/10/2022 8:25 PM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> 06.04.2022 08:24, Ashish Mhetre пишет:
>>        /* clear interrupts */
>> -     mc_writel(mc, status, MC_INTSTATUS);
>> +     if (mc->soc->num_channels) {
>> +             u32 status_chan_bit;
>> +
>> +             mc_ch_writel(mc, channel, status, MC_INTSTATUS);
>> +             status_chan_bit = BIT(channel) << mc->soc->status_reg_chan_shift;
> 
> s/status_reg_chan_shift/global_intstatus_channel_shift/
> 
Okay, I'll update in v7.

>> +             mc_ch_writel(mc, MC_BROADCAST_CHANNEL, status_chan_bit, MC_GLOBAL_INTSTATUS);
> and maybe add a one-line mc_channel_to_global_intstatus(mc, channel) helper
> 
> mc_ch_writel(mc, MC_BROADCAST_CHANNEL,
>               mc_channel_to_global_intstatus(mc, chan),
>               MC_GLOBAL_INTSTATUS);
> 
Okay, I'll add helper for getting status bit from channel and use it.

>> +     } else
>> +             mc_writel(mc, status, MC_INTSTATUS);
> 
> Missing braces after "else", "./scripts/checkpatch.pl --strict" should warn about this.
> 
I didn't run checkpatch.pl with "--strict". I will run from next time
and address this.

> Otherwise this patch looks okay at a quick glance.

Thanks Dmitry. I'll address these comments.

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Ashish Mhetre]

On 4/10/2022 8:31 PM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> 06.04.2022 08:24, Ashish Mhetre пишет:
>> diff --git a/include/soc/tegra/mc.h b/include/soc/tegra/mc.h
>> index 1066b1194a5a..c3c121fbfbb7 100644
>> --- a/include/soc/tegra/mc.h
>> +++ b/include/soc/tegra/mc.h
>> @@ -13,6 +13,9 @@
>>   #include <linux/irq.h>
>>   #include <linux/reset-controller.h>
>>   #include <linux/types.h>
>> +#include <linux/platform_device.h>
>> +
>> +#define MC_MAX_CHANNELS 16
>>
>>   struct clk;
>>   struct device;
>> @@ -181,6 +184,7 @@ struct tegra_mc_ops {
>>        int (*resume)(struct tegra_mc *mc);
>>        irqreturn_t (*handle_irq)(int irq, void *data);
>>        int (*probe_device)(struct tegra_mc *mc, struct device *dev);
>> +     int (*map_regs)(struct tegra_mc *mc, struct platform_device *pdev);
> 
> Use to_platform_device(mc->dev) instead of passing the pdev argument.

Okay, I'll update in v7.

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Ashish Mhetre]

On 4/10/2022 8:34 PM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> 06.04.2022 08:24, Ashish Mhetre пишет:
>> diff --git a/include/soc/tegra/mc.h b/include/soc/tegra/mc.h
>> index 1066b1194a5a..c3c121fbfbb7 100644
>> --- a/include/soc/tegra/mc.h
>> +++ b/include/soc/tegra/mc.h
>> @@ -13,6 +13,9 @@
>>   #include <linux/irq.h>
>>   #include <linux/reset-controller.h>
>>   #include <linux/types.h>
>> +#include <linux/platform_device.h>
>> +
>> +#define MC_MAX_CHANNELS 16
>>
>>   struct clk;
>>   struct device;
>> @@ -181,6 +184,7 @@ struct tegra_mc_ops {
>>        int (*resume)(struct tegra_mc *mc);
>>        irqreturn_t (*handle_irq)(int irq, void *data);
>>        int (*probe_device)(struct tegra_mc *mc, struct device *dev);
>> +     int (*map_regs)(struct tegra_mc *mc, struct platform_device *pdev);
>>   };
>>
>>   struct tegra_mc_soc {
>> @@ -194,6 +198,7 @@ struct tegra_mc_soc {
>>        unsigned int atom_size;
>>
>>        u8 client_id_mask;
>> +     u8 num_channels;
>>
>>        const struct tegra_smmu_soc *smmu;
>>
>> @@ -212,6 +217,8 @@ struct tegra_mc {
>>        struct tegra_smmu *smmu;
>>        struct gart_device *gart;
>>        void __iomem *regs;
>> +     void __iomem *bcast_ch_regs;
>> +     void __iomem *ch_regs[MC_MAX_CHANNELS];
> 
> Why not to allocate ch_regs at runtime?

Yes, we can do that. I'll make necessary changes in v7.

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Dmitry Osipenko]

On 4/11/22 09:05, Ashish Mhetre wrote:
> 
> 
> On 4/10/2022 7:48 PM, Dmitry Osipenko wrote:
>> External email: Use caution opening links or attachments
>>
>>
>> 06.04.2022 08:24, Ashish Mhetre пишет:
>>> +     num_dt_channels =
>>> of_property_count_elems_of_size(pdev->dev.of_node, "reg",
>>> +                                                       reg_cells *
>>> sizeof(u32));
>>> +     /*
>>> +      * On tegra186 onwards, memory controller support multiple
>>> channels.
>>> +      * Apart from regular memory controller channels, there is one
>>> broadcast
>>> +      * channel and one for stream-id registers.
>>> +      */
>>> +     if (num_dt_channels < mc->soc->num_channels + 2) {
>>> +             dev_warn(&pdev->dev, "MC channels are missing, please
>>> update memory controller DT node with MC channels\n");
>>> +             return 0;
>>> +     }
>>> +
>>> +     mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
>>> "mc-broadcast");
>>> +     if (IS_ERR(mc->bcast_ch_regs))
>>> +             return PTR_ERR(mc->bcast_ch_regs);
>>
>> Looks to me that you don't need to use of_property_count_elems_of_size()
>> and could only check the "mc-broadcast" presence to decide whether this
>> is an older DT.
>>
> Now that we are using reg-names in new DT, yes it'd be fine to just
> check mc-broadcast to decide it's a new or old DT.
> 
>> mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
>> "broadcast");
>> if (IS_ERR(mc->bcast_ch_regs)) {
>>          dev_warn(&pdev->dev, "Broadcast channel is missing, please
>> update your
>> device-tree\n");
>>          return PTR_ERR(mc->bcast_ch_regs);
>> }
> 
> return 0;
> 
> to avoid DT ABI break, right?

Yes, it should be "return 0".

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Ashish Mhetre]

On 4/11/2022 12:03 PM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> On 4/11/22 09:05, Ashish Mhetre wrote:
>>
>>
>> On 4/10/2022 7:48 PM, Dmitry Osipenko wrote:
>>> External email: Use caution opening links or attachments
>>>
>>>
>>> 06.04.2022 08:24, Ashish Mhetre пишет:
>>>> +     num_dt_channels =
>>>> of_property_count_elems_of_size(pdev->dev.of_node, "reg",
>>>> +                                                       reg_cells *
>>>> sizeof(u32));
>>>> +     /*
>>>> +      * On tegra186 onwards, memory controller support multiple
>>>> channels.
>>>> +      * Apart from regular memory controller channels, there is one
>>>> broadcast
>>>> +      * channel and one for stream-id registers.
>>>> +      */
>>>> +     if (num_dt_channels < mc->soc->num_channels + 2) {
>>>> +             dev_warn(&pdev->dev, "MC channels are missing, please
>>>> update memory controller DT node with MC channels\n");
>>>> +             return 0;
>>>> +     }
>>>> +
>>>> +     mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
>>>> "mc-broadcast");
>>>> +     if (IS_ERR(mc->bcast_ch_regs))
>>>> +             return PTR_ERR(mc->bcast_ch_regs);
>>>
>>> Looks to me that you don't need to use of_property_count_elems_of_size()
>>> and could only check the "mc-broadcast" presence to decide whether this
>>> is an older DT.
>>>
>> Now that we are using reg-names in new DT, yes it'd be fine to just
>> check mc-broadcast to decide it's a new or old DT.
>>
>>> mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
>>> "broadcast");
>>> if (IS_ERR(mc->bcast_ch_regs)) {
>>>           dev_warn(&pdev->dev, "Broadcast channel is missing, please
>>> update your
>>> device-tree\n");
>>>           return PTR_ERR(mc->bcast_ch_regs);
>>> }
>>
>> return 0;
>>
>> to avoid DT ABI break, right?
> 
> Yes, it should be "return 0".

But if we "return 0" from here, then what about the case when ioremap()
actually fails with new DT i.e. when broadcast reg is present in DT?
In that case error should be returned and probe should be failed, right?

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Dmitry Osipenko]

On 4/11/22 10:28, Ashish Mhetre wrote:
> 
> 
> On 4/11/2022 12:03 PM, Dmitry Osipenko wrote:
>> External email: Use caution opening links or attachments
>>
>>
>> On 4/11/22 09:05, Ashish Mhetre wrote:
>>>
>>>
>>> On 4/10/2022 7:48 PM, Dmitry Osipenko wrote:
>>>> External email: Use caution opening links or attachments
>>>>
>>>>
>>>> 06.04.2022 08:24, Ashish Mhetre пишет:
>>>>> +     num_dt_channels =
>>>>> of_property_count_elems_of_size(pdev->dev.of_node, "reg",
>>>>> +                                                       reg_cells *
>>>>> sizeof(u32));
>>>>> +     /*
>>>>> +      * On tegra186 onwards, memory controller support multiple
>>>>> channels.
>>>>> +      * Apart from regular memory controller channels, there is one
>>>>> broadcast
>>>>> +      * channel and one for stream-id registers.
>>>>> +      */
>>>>> +     if (num_dt_channels < mc->soc->num_channels + 2) {
>>>>> +             dev_warn(&pdev->dev, "MC channels are missing, please
>>>>> update memory controller DT node with MC channels\n");
>>>>> +             return 0;
>>>>> +     }
>>>>> +
>>>>> +     mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
>>>>> "mc-broadcast");
>>>>> +     if (IS_ERR(mc->bcast_ch_regs))
>>>>> +             return PTR_ERR(mc->bcast_ch_regs);
>>>>
>>>> Looks to me that you don't need to use
>>>> of_property_count_elems_of_size()
>>>> and could only check the "mc-broadcast" presence to decide whether this
>>>> is an older DT.
>>>>
>>> Now that we are using reg-names in new DT, yes it'd be fine to just
>>> check mc-broadcast to decide it's a new or old DT.
>>>
>>>> mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
>>>> "broadcast");
>>>> if (IS_ERR(mc->bcast_ch_regs)) {
>>>>           dev_warn(&pdev->dev, "Broadcast channel is missing, please
>>>> update your
>>>> device-tree\n");
>>>>           return PTR_ERR(mc->bcast_ch_regs);
>>>> }
>>>
>>> return 0;
>>>
>>> to avoid DT ABI break, right?
>>
>> Yes, it should be "return 0".
> 
> But if we "return 0" from here, then what about the case when ioremap()
> actually fails with new DT i.e. when broadcast reg is present in DT?
> In that case error should be returned and probe should be failed, right?

You should check for the -ENOENT.

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Ashish Mhetre]

On 4/11/2022 1:05 PM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> On 4/11/22 10:28, Ashish Mhetre wrote:
>>
>>
>> On 4/11/2022 12:03 PM, Dmitry Osipenko wrote:
>>> External email: Use caution opening links or attachments
>>>
>>>
>>> On 4/11/22 09:05, Ashish Mhetre wrote:
>>>>
>>>>
>>>> On 4/10/2022 7:48 PM, Dmitry Osipenko wrote:
>>>>> External email: Use caution opening links or attachments
>>>>>
>>>>>
>>>>> 06.04.2022 08:24, Ashish Mhetre пишет:
>>>>>> +     num_dt_channels =
>>>>>> of_property_count_elems_of_size(pdev->dev.of_node, "reg",
>>>>>> +                                                       reg_cells *
>>>>>> sizeof(u32));
>>>>>> +     /*
>>>>>> +      * On tegra186 onwards, memory controller support multiple
>>>>>> channels.
>>>>>> +      * Apart from regular memory controller channels, there is one
>>>>>> broadcast
>>>>>> +      * channel and one for stream-id registers.
>>>>>> +      */
>>>>>> +     if (num_dt_channels < mc->soc->num_channels + 2) {
>>>>>> +             dev_warn(&pdev->dev, "MC channels are missing, please
>>>>>> update memory controller DT node with MC channels\n");
>>>>>> +             return 0;
>>>>>> +     }
>>>>>> +
>>>>>> +     mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
>>>>>> "mc-broadcast");
>>>>>> +     if (IS_ERR(mc->bcast_ch_regs))
>>>>>> +             return PTR_ERR(mc->bcast_ch_regs);
>>>>>
>>>>> Looks to me that you don't need to use
>>>>> of_property_count_elems_of_size()
>>>>> and could only check the "mc-broadcast" presence to decide whether this
>>>>> is an older DT.
>>>>>
>>>> Now that we are using reg-names in new DT, yes it'd be fine to just
>>>> check mc-broadcast to decide it's a new or old DT.
>>>>
>>>>> mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
>>>>> "broadcast");
>>>>> if (IS_ERR(mc->bcast_ch_regs)) {
>>>>>            dev_warn(&pdev->dev, "Broadcast channel is missing, please
>>>>> update your
>>>>> device-tree\n");
>>>>>            return PTR_ERR(mc->bcast_ch_regs);
>>>>> }
>>>>
>>>> return 0;
>>>>
>>>> to avoid DT ABI break, right?
>>>
>>> Yes, it should be "return 0".
>>
>> But if we "return 0" from here, then what about the case when ioremap()
>> actually fails with new DT i.e. when broadcast reg is present in DT?
>> In that case error should be returned and probe should be failed, right?
> 
> You should check for the -ENOENT.

I checked __devm_ioremap_resource(), it returns -EINVAL if given
resource is not present. So should we check for -EINVAL instead?

Re: [Patch v6 1/4] memory: tegra: Add memory controller channels support

[Dmitry Osipenko]

On 4/11/22 12:18, Ashish Mhetre wrote:
> 
> 
> On 4/11/2022 1:05 PM, Dmitry Osipenko wrote:
>> External email: Use caution opening links or attachments
>>
>>
>> On 4/11/22 10:28, Ashish Mhetre wrote:
>>>
>>>
>>> On 4/11/2022 12:03 PM, Dmitry Osipenko wrote:
>>>> External email: Use caution opening links or attachments
>>>>
>>>>
>>>> On 4/11/22 09:05, Ashish Mhetre wrote:
>>>>>
>>>>>
>>>>> On 4/10/2022 7:48 PM, Dmitry Osipenko wrote:
>>>>>> External email: Use caution opening links or attachments
>>>>>>
>>>>>>
>>>>>> 06.04.2022 08:24, Ashish Mhetre пишет:
>>>>>>> +     num_dt_channels =
>>>>>>> of_property_count_elems_of_size(pdev->dev.of_node, "reg",
>>>>>>> +                                                       reg_cells *
>>>>>>> sizeof(u32));
>>>>>>> +     /*
>>>>>>> +      * On tegra186 onwards, memory controller support multiple
>>>>>>> channels.
>>>>>>> +      * Apart from regular memory controller channels, there is one
>>>>>>> broadcast
>>>>>>> +      * channel and one for stream-id registers.
>>>>>>> +      */
>>>>>>> +     if (num_dt_channels < mc->soc->num_channels + 2) {
>>>>>>> +             dev_warn(&pdev->dev, "MC channels are missing, please
>>>>>>> update memory controller DT node with MC channels\n");
>>>>>>> +             return 0;
>>>>>>> +     }
>>>>>>> +
>>>>>>> +     mc->bcast_ch_regs =
>>>>>>> devm_platform_ioremap_resource_byname(pdev,
>>>>>>> "mc-broadcast");
>>>>>>> +     if (IS_ERR(mc->bcast_ch_regs))
>>>>>>> +             return PTR_ERR(mc->bcast_ch_regs);
>>>>>>
>>>>>> Looks to me that you don't need to use
>>>>>> of_property_count_elems_of_size()
>>>>>> and could only check the "mc-broadcast" presence to decide whether
>>>>>> this
>>>>>> is an older DT.
>>>>>>
>>>>> Now that we are using reg-names in new DT, yes it'd be fine to just
>>>>> check mc-broadcast to decide it's a new or old DT.
>>>>>
>>>>>> mc->bcast_ch_regs = devm_platform_ioremap_resource_byname(pdev,
>>>>>> "broadcast");
>>>>>> if (IS_ERR(mc->bcast_ch_regs)) {
>>>>>>            dev_warn(&pdev->dev, "Broadcast channel is missing, please
>>>>>> update your
>>>>>> device-tree\n");
>>>>>>            return PTR_ERR(mc->bcast_ch_regs);
>>>>>> }
>>>>>
>>>>> return 0;
>>>>>
>>>>> to avoid DT ABI break, right?
>>>>
>>>> Yes, it should be "return 0".
>>>
>>> But if we "return 0" from here, then what about the case when ioremap()
>>> actually fails with new DT i.e. when broadcast reg is present in DT?
>>> In that case error should be returned and probe should be failed, right?
>>
>> You should check for the -ENOENT.
> 
> I checked __devm_ioremap_resource(), it returns -EINVAL if given
> resource is not present. So should we check for -EINVAL instead?

Yes

Re: [Patch v6 3/4] dt-bindings: memory: Update reg maxitems for tegra186

[Ashish Mhetre]

On 4/10/2022 7:51 PM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> 06.04.2022 08:24, Ashish Mhetre пишет:
>>           memory-controller@2c00000 {
>>               compatible = "nvidia,tegra186-mc";
>> -            reg = <0x0 0x02c00000 0x0 0xb0000>;
>> +            reg = <0x0 0x02c00000 0x0 0x10000>,    /* MC-SID */
>> +                  <0x0 0x02c10000 0x0 0x10000>,    /* Broadcast channel */
>> +                  <0x0 0x02c20000 0x0 0x10000>,    /* MC0 */
>> +                  <0x0 0x02c30000 0x0 0x10000>,    /* MC1 */
>> +                  <0x0 0x02c40000 0x0 0x10000>,    /* MC2 */
>> +                  <0x0 0x02c50000 0x0 0x10000>;    /* MC3 */
>> +            reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1", "mc2", "mc3";
> 
> The "mc-" prefix feels redundant to me, I'd name the regs like this:
> 
>    "sid", "broadcast", "ch0", "ch1", "ch2", "ch3"
> 
> 
> You should also add validation of the regs/reg-names to the yaml based
> on SoC version. I.e. it's not enough to only bump the maxItems.

Okay, I will add validation of reg-names as following:

   reg-names:
     minItems: 0
     maxItems: 6
     items:
       - const: sid
       - const: broadcast
       - const: ch0
       - const: ch1
       - const: ch2
       - const: ch3


We will have to keep minItems to 0 in order to make it compatible with
old DT, right?

Re: [Patch v6 3/4] dt-bindings: memory: Update reg maxitems for tegra186

[Dmitry Osipenko]

On 4/11/22 18:02, Ashish Mhetre wrote:
> 
> 
> On 4/10/2022 7:51 PM, Dmitry Osipenko wrote:
>> External email: Use caution opening links or attachments
>>
>>
>> 06.04.2022 08:24, Ashish Mhetre пишет:
>>>           memory-controller@2c00000 {
>>>               compatible = "nvidia,tegra186-mc";
>>> -            reg = <0x0 0x02c00000 0x0 0xb0000>;
>>> +            reg = <0x0 0x02c00000 0x0 0x10000>,    /* MC-SID */
>>> +                  <0x0 0x02c10000 0x0 0x10000>,    /* Broadcast
>>> channel */
>>> +                  <0x0 0x02c20000 0x0 0x10000>,    /* MC0 */
>>> +                  <0x0 0x02c30000 0x0 0x10000>,    /* MC1 */
>>> +                  <0x0 0x02c40000 0x0 0x10000>,    /* MC2 */
>>> +                  <0x0 0x02c50000 0x0 0x10000>;    /* MC3 */
>>> +            reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1",
>>> "mc2", "mc3";
>>
>> The "mc-" prefix feels redundant to me, I'd name the regs like this:
>>
>>    "sid", "broadcast", "ch0", "ch1", "ch2", "ch3"
>>
>>
>> You should also add validation of the regs/reg-names to the yaml based
>> on SoC version. I.e. it's not enough to only bump the maxItems.
> 
> Okay, I will add validation of reg-names as following:
> 
>   reg-names:
>     minItems: 0
>     maxItems: 6
>     items:
>       - const: sid
>       - const: broadcast
>       - const: ch0
>       - const: ch1
>       - const: ch2
>       - const: ch3
> 
> 
> We will have to keep minItems to 0 in order to make it compatible with
> old DT, right?

Bindings are about the latest DTs. In general older dtbs must be updated
and you must get error from the schema checker for older DTs. It's only
drivers that should care about older dtbs.

Re: [Patch v6 3/4] dt-bindings: memory: Update reg maxitems for tegra186

[Ashish Mhetre]

On 4/11/2022 8:59 PM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> On 4/11/22 18:02, Ashish Mhetre wrote:
>>
>>
>> On 4/10/2022 7:51 PM, Dmitry Osipenko wrote:
>>> External email: Use caution opening links or attachments
>>>
>>>
>>> 06.04.2022 08:24, Ashish Mhetre пишет:
>>>>            memory-controller@2c00000 {
>>>>                compatible = "nvidia,tegra186-mc";
>>>> -            reg = <0x0 0x02c00000 0x0 0xb0000>;
>>>> +            reg = <0x0 0x02c00000 0x0 0x10000>,    /* MC-SID */
>>>> +                  <0x0 0x02c10000 0x0 0x10000>,    /* Broadcast
>>>> channel */
>>>> +                  <0x0 0x02c20000 0x0 0x10000>,    /* MC0 */
>>>> +                  <0x0 0x02c30000 0x0 0x10000>,    /* MC1 */
>>>> +                  <0x0 0x02c40000 0x0 0x10000>,    /* MC2 */
>>>> +                  <0x0 0x02c50000 0x0 0x10000>;    /* MC3 */
>>>> +            reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1",
>>>> "mc2", "mc3";
>>>
>>> The "mc-" prefix feels redundant to me, I'd name the regs like this:
>>>
>>>     "sid", "broadcast", "ch0", "ch1", "ch2", "ch3"
>>>
>>>
>>> You should also add validation of the regs/reg-names to the yaml based
>>> on SoC version. I.e. it's not enough to only bump the maxItems.
>>
>> Okay, I will add validation of reg-names as following:
>>
>>    reg-names:
>>      minItems: 0
>>      maxItems: 6
>>      items:
>>        - const: sid
>>        - const: broadcast
>>        - const: ch0
>>        - const: ch1
>>        - const: ch2
>>        - const: ch3
>>
>>
>> We will have to keep minItems to 0 in order to make it compatible with
>> old DT, right?
> 
> Bindings are about the latest DTs. In general older dtbs must be updated
> and you must get error from the schema checker for older DTs. It's only
> drivers that should care about older dtbs.

On v5 Krzysztof mentioned that old DTS will start failing with new
bindings https://lkml.org/lkml/2022/3/22/907.
So I just wanted to confirm whether it's fine if updated bindings
start to fail with old DTS?

Re: [Patch v6 3/4] dt-bindings: memory: Update reg maxitems for tegra186

[Dmitry Osipenko]

On 4/11/22 18:41, Ashish Mhetre wrote:
> 
> 
> On 4/11/2022 8:59 PM, Dmitry Osipenko wrote:
>> External email: Use caution opening links or attachments
>>
>>
>> On 4/11/22 18:02, Ashish Mhetre wrote:
>>>
>>>
>>> On 4/10/2022 7:51 PM, Dmitry Osipenko wrote:
>>>> External email: Use caution opening links or attachments
>>>>
>>>>
>>>> 06.04.2022 08:24, Ashish Mhetre пишет:
>>>>>            memory-controller@2c00000 {
>>>>>                compatible = "nvidia,tegra186-mc";
>>>>> -            reg = <0x0 0x02c00000 0x0 0xb0000>;
>>>>> +            reg = <0x0 0x02c00000 0x0 0x10000>,    /* MC-SID */
>>>>> +                  <0x0 0x02c10000 0x0 0x10000>,    /* Broadcast
>>>>> channel */
>>>>> +                  <0x0 0x02c20000 0x0 0x10000>,    /* MC0 */
>>>>> +                  <0x0 0x02c30000 0x0 0x10000>,    /* MC1 */
>>>>> +                  <0x0 0x02c40000 0x0 0x10000>,    /* MC2 */
>>>>> +                  <0x0 0x02c50000 0x0 0x10000>;    /* MC3 */
>>>>> +            reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1",
>>>>> "mc2", "mc3";
>>>>
>>>> The "mc-" prefix feels redundant to me, I'd name the regs like this:
>>>>
>>>>     "sid", "broadcast", "ch0", "ch1", "ch2", "ch3"
>>>>
>>>>
>>>> You should also add validation of the regs/reg-names to the yaml based
>>>> on SoC version. I.e. it's not enough to only bump the maxItems.
>>>
>>> Okay, I will add validation of reg-names as following:
>>>
>>>    reg-names:
>>>      minItems: 0
>>>      maxItems: 6
>>>      items:
>>>        - const: sid
>>>        - const: broadcast
>>>        - const: ch0
>>>        - const: ch1
>>>        - const: ch2
>>>        - const: ch3
>>>
>>>
>>> We will have to keep minItems to 0 in order to make it compatible with
>>> old DT, right?
>>
>> Bindings are about the latest DTs. In general older dtbs must be updated
>> and you must get error from the schema checker for older DTs. It's only
>> drivers that should care about older dtbs.
> 
> On v5 Krzysztof mentioned that old DTS will start failing with new
> bindings https://lkml.org/lkml/2022/3/22/907.
> So I just wanted to confirm whether it's fine if updated bindings
> start to fail with old DTS?

Since the older DT was incorrect, it's fine that the DT check will fail
for it.

Re: [Patch v6 3/4] dt-bindings: memory: Update reg maxitems for tegra186

[Ashish Mhetre]

On 4/12/2022 3:58 AM, Dmitry Osipenko wrote:
> External email: Use caution opening links or attachments
> 
> 
> On 4/11/22 18:41, Ashish Mhetre wrote:
>>
>>
>> On 4/11/2022 8:59 PM, Dmitry Osipenko wrote:
>>> External email: Use caution opening links or attachments
>>>
>>>
>>> On 4/11/22 18:02, Ashish Mhetre wrote:
>>>>
>>>>
>>>> On 4/10/2022 7:51 PM, Dmitry Osipenko wrote:
>>>>> External email: Use caution opening links or attachments
>>>>>
>>>>>
>>>>> 06.04.2022 08:24, Ashish Mhetre пишет:
>>>>>>             memory-controller@2c00000 {
>>>>>>                 compatible = "nvidia,tegra186-mc";
>>>>>> -            reg = <0x0 0x02c00000 0x0 0xb0000>;
>>>>>> +            reg = <0x0 0x02c00000 0x0 0x10000>,    /* MC-SID */
>>>>>> +                  <0x0 0x02c10000 0x0 0x10000>,    /* Broadcast
>>>>>> channel */
>>>>>> +                  <0x0 0x02c20000 0x0 0x10000>,    /* MC0 */
>>>>>> +                  <0x0 0x02c30000 0x0 0x10000>,    /* MC1 */
>>>>>> +                  <0x0 0x02c40000 0x0 0x10000>,    /* MC2 */
>>>>>> +                  <0x0 0x02c50000 0x0 0x10000>;    /* MC3 */
>>>>>> +            reg-names = "mc-sid", "mc-broadcast", "mc0", "mc1",
>>>>>> "mc2", "mc3";
>>>>>
>>>>> The "mc-" prefix feels redundant to me, I'd name the regs like this:
>>>>>
>>>>>      "sid", "broadcast", "ch0", "ch1", "ch2", "ch3"
>>>>>
>>>>>
>>>>> You should also add validation of the regs/reg-names to the yaml based
>>>>> on SoC version. I.e. it's not enough to only bump the maxItems.
>>>>
>>>> Okay, I will add validation of reg-names as following:
>>>>
>>>>     reg-names:
>>>>       minItems: 0
>>>>       maxItems: 6
>>>>       items:
>>>>         - const: sid
>>>>         - const: broadcast
>>>>         - const: ch0
>>>>         - const: ch1
>>>>         - const: ch2
>>>>         - const: ch3
>>>>
>>>>
>>>> We will have to keep minItems to 0 in order to make it compatible with
>>>> old DT, right?
>>>
>>> Bindings are about the latest DTs. In general older dtbs must be updated
>>> and you must get error from the schema checker for older DTs. It's only
>>> drivers that should care about older dtbs.
>>
>> On v5 Krzysztof mentioned that old DTS will start failing with new
>> bindings https://lkml.org/lkml/2022/3/22/907.
>> So I just wanted to confirm whether it's fine if updated bindings
>> start to fail with old DTS?
> 
> Since the older DT was incorrect, it's fine that the DT check will fail
> for it.

Thanks for confirming Dmitry. I will incorporate all comments and send
v7.