Skip to contentSkip to conversation

media: ipu6: fix the wrong type cast and 64-bit division

PATCH Superseded [email protected] 3 participants 6 messages Oct 8, 2024 – Oct 9, 2024

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 media: ipu6: fix the wrong type cast and 64-bit division

From: Bingbu Cao <[email protected]>

This patch fixes the build errors with i386-allmodconfig, the errors are caused by wrong type cast and 64-bit division.

---
 drivers/media/pci/intel/ipu6/ipu6-cpd.c          | 6 +++---
 drivers/media/pci/intel/ipu6/ipu6-fw-com.c       | 8 ++++----
 drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c | 8 ++++----
 drivers/media/pci/intel/ipu6/ipu6-isys.c         | 2 +-
 drivers/media/pci/intel/ipu6/ipu6.c              | 3 ++-
 5 files changed, 14 insertions(+), 13 deletions(-)
drivers/media/pci/intel/ipu6/ipu6-cpd.c +3 -3
index 715b21ab4b8e..9b46b91cfc1a 100644
@@ -43,9 +43,9 @@
  * 63:56        55      54:48   47:32   31:24   23:0
  * Rsvd         Rsvd    Type    Version Rsvd    Size
  */
-#define PKG_DIR_SIZE_MASK	GENMASK(23, 0)
-#define PKG_DIR_VERSION_MASK	GENMASK(47, 32)
-#define PKG_DIR_TYPE_MASK	GENMASK(54, 48)
+#define PKG_DIR_SIZE_MASK	GENMASK_ULL(23, 0)
+#define PKG_DIR_VERSION_MASK	GENMASK_ULL(47, 32)
+#define PKG_DIR_TYPE_MASK	GENMASK_ULL(54, 48)
 
 static inline const struct ipu6_cpd_ent *ipu6_cpd_get_entry(const void *cpd,
 							    u8 idx)
drivers/media/pci/intel/ipu6/ipu6-fw-com.c +4 -4
index 0b33fe9e703d..6da2aca69bd3 100644
@@ -30,7 +30,7 @@
 
 /* Shared structure between driver and FW - do not modify */
 struct ipu6_fw_sys_queue {
-	u64 host_address;
+	uintptr_t host_address;
 	u32 vied_address;
 	u32 size;
 	u32 token_size;
@@ -40,7 +40,7 @@ struct ipu6_fw_sys_queue {
 } __packed;
 
 struct ipu6_fw_sys_queue_res {
-	u64 host_address;
+	uintptr_t host_address;
 	u32 vied_address;
 	u32 reg;
 } __packed;
@@ -242,7 +242,7 @@ void *ipu6_fw_com_prepare(struct ipu6_fw_com_cfg *cfg,
 	/* initialize input queues */
 	offset += specific_size;
 	res.reg = SYSCOM_QPR_BASE_REG;
-	res.host_address = (u64)(ctx->dma_buffer + offset);
+	res.host_address = (uintptr_t)(ctx->dma_buffer + offset);
 	res.vied_address = ctx->dma_addr + offset;
 	for (i = 0; i < cfg->num_input_queues; i++)
 		ipu6_sys_queue_init(ctx->input_queue + i,
@@ -251,7 +251,7 @@ void *ipu6_fw_com_prepare(struct ipu6_fw_com_cfg *cfg,
 
 	/* initialize output queues */
 	offset += sizeinput;
-	res.host_address = (u64)(ctx->dma_buffer + offset);
+	res.host_address = (uintptr_t)(ctx->dma_buffer + offset);
 	res.vied_address = ctx->dma_addr + offset;
 	for (i = 0; i < cfg->num_output_queues; i++) {
 		ipu6_sys_queue_init(ctx->output_queue + i,
drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c +4 -4
index 1715275e6776..1cfc9a45bcea 100644
@@ -67,8 +67,8 @@ static void dwc_dphy_write(struct ipu6_isys *isys, u32 phy_id, u32 addr,
 	void __iomem *isys_base = isys->pdata->base;
 	void __iomem *base = isys_base + IPU6_DWC_DPHY_BASE(phy_id);
 
-	dev_dbg(dev, "write: reg 0x%lx = data 0x%x", base + addr - isys_base,
-		data);
+	dev_dbg(dev, "write: reg 0x%lx = data 0x%x",
+		(ulong)(base + addr - isys_base), data);
 	writel(data, base + addr);
 }
 
@@ -80,8 +80,8 @@ static u32 dwc_dphy_read(struct ipu6_isys *isys, u32 phy_id, u32 addr)
 	u32 data;
 
 	data = readl(base + addr);
-	dev_dbg(dev, "read: reg 0x%lx = data 0x%x", base + addr - isys_base,
-		data);
+	dev_dbg(dev, "read: reg 0x%lx = data 0x%x",
+		(ulong)(base + addr - isys_base), data);
 
 	return data;
 }
drivers/media/pci/intel/ipu6/ipu6-isys.c +1 -1
index c4aff2e2009b..51c1a567eff8 100644
@@ -576,7 +576,7 @@ void update_watermark_setting(struct ipu6_isys *isys)
 	}
 
 	enable_iwake(isys, true);
-	calc_fill_time_us = max_sram_size / isys_pb_datarate_mbs;
+	calc_fill_time_us = div64_u64(max_sram_size, isys_pb_datarate_mbs);
 
 	if (isys->pdata->ipdata->enhanced_iwake) {
 		ltr = isys->pdata->ipdata->ltr;
drivers/media/pci/intel/ipu6/ipu6.c +2 -1
index 7fb707d35309..797d11d1d3b1 100644
@@ -247,7 +247,8 @@ ipu6_pkg_dir_configure_spc(struct ipu6_device *isp,
 		dma_addr = sg_dma_address(isp->psys->fw_sgt.sgl);
 
 	pg_offset = server_fw_addr - dma_addr;
-	prog = (struct ipu6_cell_program *)((u64)isp->cpd_fw->data + pg_offset);
+	prog = (struct ipu6_cell_program *)((uintptr_t)isp->cpd_fw->data +
+					    pg_offset);
 	spc_base = base + prog->regs_addr;
 	if (spc_base != (base + hw_variant->spc_offset))
 		dev_warn(&isp->pdev->dev,
Signature · 1 line
2.34.1
#1 lore raw

PATCH 2/2 media: ipu6: remove redundant dependency in Kconfig

From: Bingbu Cao <[email protected]>

IPU6 driver simply depends on X86, X86 and 64BIT cover the X86_64, redundant X86_64 dependency in Kconfig could be removed.

Fixes c70281cc83d6 ("media: intel/ipu6: add Kconfig and Makefile")
---
 drivers/media/pci/intel/ipu6/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/media/pci/intel/ipu6/Kconfig +1 -1
index 49e4fb696573..624d8e44c53b 100644
@@ -2,7 +2,7 @@ config VIDEO_INTEL_IPU6
 	tristate "Intel IPU6 driver"
 	depends on ACPI || COMPILE_TEST
 	depends on VIDEO_DEV
-	depends on X86 && X86_64 && HAS_DMA
+	depends on X86 && HAS_DMA
 	depends on IPU_BRIDGE || !IPU_BRIDGE
 	#
 	# This driver incorrectly tries to override the dma_ops.  It should
Signature · 1 line
2.34.1
Reviewed-by on v1 2/2
#2 lore raw

Re: [PATCH 2/2] media: ipu6: remove redundant dependency in Kconfig

Quoted 4 lines · On Tue, Oct 08, 2024 at 02:19:16PM +0800, [email protected] wrote:

From: Bingbu Cao <[email protected]>

IPU6 driver simply depends on X86, X86 and 64BIT cover the X86_64, redundant X86_64 dependency in Kconfig could be removed.

Please, fix my email address in the other tag to be @linux,intel.com.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw
In reply to [email protected]
Quoted 4 lines · On Tue, Oct 08, 2024 at 02:19:15PM +0800, [email protected] wrote:

From: Bingbu Cao <[email protected]>

This patch fixes the build errors with i386-allmodconfig, the errors are caused by wrong type cast and 64-bit division.

Thanks for the change, my comments below.

...

Review

Quoted 4 lines

/* Shared structure between driver and FW - do not modify */ struct ipu6_fw_sys_queue {

(quoted diff) +1 -1
-	u64 host_address;
+	uintptr_t host_address;

Okay, in the given semantic this probably should be phys_addr_t. BUT, is this address somehow is going to be used by IPU6 hardware? If "yes", the type shall not be changed.

Looking at types used I hope the answer is "no", otherwise the types in the structures should be properly choose WRT endianess (and what __packed is doing here? Is it part of the protocol?).

drivers/media/pci/intel/ipu6/ipu6-fw-com.c struct ipu6_fw_sys_queue { · line 43

Quoted 9 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c +1 -1
 	u32 vied_address;
 	u32 size;
 	u32 token_size;
@@ -40,7 +40,7 @@ struct ipu6_fw_sys_queue {
 } __packed;
 
 struct ipu6_fw_sys_queue_res {
-	u64 host_address;
+	uintptr_t host_address;

Ditto.

Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c +0 -0
 	u32 vied_address;
 	u32 reg;
 } __packed;

...

drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c static void dwc_dphy_write(struct ipu6_isys *isys, u32 phy_id, u32 addr, · line 71

Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c +2 -2
-	dev_dbg(dev, "write: reg 0x%lx = data 0x%x", base + addr - isys_base,
-		data);
+	dev_dbg(dev, "write: reg 0x%lx = data 0x%x",
+		(ulong)(base + addr - isys_base), data);

No, one should use proper specifiers for this. And what the heck 'ulong' is? Where is it being defined?

...

drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c static u32 dwc_dphy_read(struct ipu6_isys *isys, u32 phy_id, u32 addr) · line 84

Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c +2 -2
-	dev_dbg(dev, "read: reg 0x%lx = data 0x%x", base + addr - isys_base,
-		data);
+	dev_dbg(dev, "read: reg 0x%lx = data 0x%x",
+		(ulong)(base + addr - isys_base), data);

Ditto.

...

drivers/media/pci/intel/ipu6/ipu6.c ipu6_pkg_dir_configure_spc(struct ipu6_device *isp, · line 251

Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6.c +2 -1
 	pg_offset = server_fw_addr - dma_addr;
-	prog = (struct ipu6_cell_program *)((u64)isp->cpd_fw->data + pg_offset);
+	prog = (struct ipu6_cell_program *)((uintptr_t)isp->cpd_fw->data +
+					    pg_offset);

Side Q: What are the alignment requirements for the prog pointer?

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#4 lore raw

Andy,

Thanks for the comments.

Quoted 19 lines · Andy Shevchenko
Quoted 4 lines · On Tue, Oct 08, 2024 at 02:19:15PM +0800, [email protected] wrote:

From: Bingbu Cao <[email protected]>

This patch fixes the build errors with i386-allmodconfig, the errors are caused by wrong type cast and 64-bit division.

Thanks for the change, my comments below.

...

Review

Quoted 4 lines

/* Shared structure between driver and FW - do not modify */ struct ipu6_fw_sys_queue {

(quoted diff) +1 -1
-	u64 host_address;
+	uintptr_t host_address;

Okay, in the given semantic this probably should be phys_addr_t. BUT, is this address somehow is going to be used by IPU6 hardware? If "yes", the type shall not be changed.

Quoted 3 lines

Looking at types used I hope the answer is "no", otherwise the types in the structures should be properly choose WRT endianess (and what __packed is doing here? Is it part of the protocol?).

You are correct, the type here should not be changed.

Quoted 36 lines

drivers/media/pci/intel/ipu6/ipu6-fw-com.c struct ipu6_fw_sys_queue { · line 43

Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c +1 -1
 	u32 vied_address;
 	u32 size;
 	u32 token_size;
@@ -40,7 +40,7 @@ struct ipu6_fw_sys_queue {
 } __packed;
 
 struct ipu6_fw_sys_queue_res {
-	u64 host_address;
+	uintptr_t host_address;

Ditto.

Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c +0 -0
 	u32 vied_address;
 	u32 reg;
 } __packed;

...

drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c static void dwc_dphy_write(struct ipu6_isys *isys, u32 phy_id, u32 addr, · line 71

Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c +2 -2
-	dev_dbg(dev, "write: reg 0x%lx = data 0x%x", base + addr - isys_base,
-		data);
+	dev_dbg(dev, "write: reg 0x%lx = data 0x%x",
+		(ulong)(base + addr - isys_base), data);

No, one should use proper specifiers for this. And what the heck 'ulong' is? Where is it being defined?

...

drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c static u32 dwc_dphy_read(struct ipu6_isys *isys, u32 phy_id, u32 addr) · line 84

Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c +2 -2
-	dev_dbg(dev, "read: reg 0x%lx = data 0x%x", base + addr - isys_base,
-		data);
+	dev_dbg(dev, "read: reg 0x%lx = data 0x%x",
+		(ulong)(base + addr - isys_base), data);

Ditto.

Quoted 8 lines

...

drivers/media/pci/intel/ipu6/ipu6.c ipu6_pkg_dir_configure_spc(struct ipu6_device *isp, · line 251

Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6.c +2 -1
 	pg_offset = server_fw_addr - dma_addr;
-	prog = (struct ipu6_cell_program *)((u64)isp->cpd_fw->data + pg_offset);
+	prog = (struct ipu6_cell_program *)((uintptr_t)isp->cpd_fw->data +
+					    pg_offset);

Side Q: What are the alignment requirements for the prog pointer?

It should be 64 bytes alignment.

Quoted 1 line
Signature · 2 lines
Best regards,
Bingbu Cao
#5 lore raw
Quoted 3 lines · Bingbu Cao
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · On Tue, Oct 08, 2024 at 02:19:15PM +0800, [email protected] wrote:

From: Bingbu Cao <[email protected]>

...

Quoted 9 lines
Quoted 9 lines

Review

Quoted 4 lines

/* Shared structure between driver and FW - do not modify */ struct ipu6_fw_sys_queue {

(quoted diff) +1 -1
-	u64 host_address;
+	uintptr_t host_address;

Okay, in the given semantic this probably should be phys_addr_t. BUT, is this address somehow is going to be used by IPU6 hardware? If "yes", the type shall not be changed.

Quoted 5 lines
Quoted 3 lines

Looking at types used I hope the answer is "no", otherwise the types in the structures should be properly choose WRT endianess (and what __packed is doing here? Is it part of the protocol?).

You are correct, the type here should not be changed.

Okay, so perhaps for now we can amend the code to work around 32-bit compilation issues while leaving the type the same. Ideally this struct should be modified to follow endianess (all those types with __le prefixes), but that is another story.

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c +0 -0
 	u32 vied_address;
 	u32 size;
 	u32 token_size;
 
 } __packed;

...

Quoted 3 lines
Quoted 1 line

Side Q: What are the alignment requirements for the prog pointer?

It should be 64 bytes alignment.

Okay, so in this case it's good to cast like this. Perhaps a comment should be added at some point, because it sounds like I already asked the very same question a while ago.

Signature · 2 lines
With Best Regards,
Andy Shevchenko