linux-next: build failure after merge of the v4l-dvb-next tree

linux-next: build failure after merge of the v4l-dvb-next tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 971 bytes --]

Hi all,

After merging the v4l-dvb-next tree, today's linux-next build (x86_64
allmodconfig) failed like this:

drivers/media/pci/intel/ipu6/ipu6.c: In function 'request_cpd_fw':
drivers/media/pci/intel/ipu6/ipu6.c:529:21: error: implicit declaration of function 'vmalloc'; did you mean 'kvmalloc'? [-Werror=implicit-function-declaration]
  529 |         dst->data = vmalloc(fw->size);
      |                     ^~~~~~~
      |                     kvmalloc
drivers/media/pci/intel/ipu6/ipu6.c:529:19: error: assignment to 'const u8 *' {aka 'const unsigned char *'} from 'int' makes pointer from integer without a cast [-Werror=int-conversion]
  529 |         dst->data = vmalloc(fw->size);
      |                   ^
cc1: all warnings being treated as errors

Caused by commit

  25fedc021985 ("media: intel/ipu6: add Intel IPU6 PCI device driver")

I have used the vl4-dvb-next tree from next-20240430 for today.

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[PATCH 1/1] media: ipu6: Fix vmalloc memory allocation

[Sakari Ailus]

The driver was calling vmalloc() to allocate memory, something which isn't
available except when particular Kconfig settings are enabled.

Use kvmalloc() instead.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Fixes: 25fedc021985 ("media: intel/ipu6: add Intel IPU6 PCI device driver")
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
---
Hi Stephen,

Thanks for reporting this. I'm a bit surprised this wasn't catched
earlier. But it seems vmalloc() is defined in some configuration.

- Sakari

 drivers/media/pci/intel/ipu6/ipu6.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/media/pci/intel/ipu6/ipu6.c b/drivers/media/pci/intel/ipu6/ipu6.c
index 4b1f69d14d71..082b1d6196be 100644
--- a/drivers/media/pci/intel/ipu6/ipu6.c
+++ b/drivers/media/pci/intel/ipu6/ipu6.c
@@ -526,7 +526,7 @@ static int request_cpd_fw(const struct firmware **firmware_p, const char *name,
 	}
 
 	dst->size = fw->size;
-	dst->data = vmalloc(fw->size);
+	dst->data = kvmalloc(fw->size, GFP_KERNEL);
 	if (!dst->data) {
 		kfree(dst);
 		ret = -ENOMEM;
-- 
2.39.2

Re: [PATCH 1/1] media: ipu6: Fix vmalloc memory allocation

[Mauro Carvalho Chehab]

Em Thu,  2 May 2024 09:15:25 +0300
Sakari Ailus <sakari.ailus@linux.intel.com> escreveu:

> The driver was calling vmalloc() to allocate memory, something which isn't
> available except when particular Kconfig settings are enabled.
> 
> Use kvmalloc() instead.
> 
> Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
> Fixes: 25fedc021985 ("media: intel/ipu6: add Intel IPU6 PCI device driver")
> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
> ---
> Hi Stephen,
> 
> Thanks for reporting this. I'm a bit surprised this wasn't catched
> earlier. But it seems vmalloc() is defined in some configuration.
> 
> - Sakari
> 
>  drivers/media/pci/intel/ipu6/ipu6.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/media/pci/intel/ipu6/ipu6.c b/drivers/media/pci/intel/ipu6/ipu6.c
> index 4b1f69d14d71..082b1d6196be 100644
> --- a/drivers/media/pci/intel/ipu6/ipu6.c
> +++ b/drivers/media/pci/intel/ipu6/ipu6.c
> @@ -526,7 +526,7 @@ static int request_cpd_fw(const struct firmware **firmware_p, const char *name,
>  	}
>  
>  	dst->size = fw->size;
> -	dst->data = vmalloc(fw->size);
> +	dst->data = kvmalloc(fw->size, GFP_KERNEL);

Where are you freeing it? If I understood the code right, you're storing
dst at isp->cpd_fw, but I can't see any code freeing it nor cpw_fw->data
at device removal.

Also, there is a logic there at the same function checking for vmalloc:

	if (is_vmalloc_addr(fw->data)) {
		*firmware_p = fw;
		return 0;
	}

As now the memory may not be inside vmalloc space, shouldn't this
be changed as well?

Regards,
Mauro

>  	if (!dst->data) {
>  		kfree(dst);
>  		ret = -ENOMEM;

[PATCH 1/1] media: intel/ipu6: Don't re-allocate memory for firmware

[Sakari Ailus]

The ipu6 driver allocated vmalloc memory for the firmware if
request_firmware() somehow managed not to use vmalloc to allocate it.

Still how the memory is allocated by request_firmware() is not specified
in its API, so be prepared for kmalloc-allocated firmware, too. Instead of
allocating new vmalloc-backed buffer for the firmware, obtain the pages
from virtual addresses instead.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Fixes: 25fedc021985 ("media: intel/ipu6: add Intel IPU6 PCI device driver")
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
---
Hello everyone,

Mauro preferred not to merge the earlier patch. Admittedly, there are
better ways to fix the problem. Such as this one.

- Sakari

 drivers/media/pci/intel/ipu6/ipu6-buttress.c |  7 +++-
 drivers/media/pci/intel/ipu6/ipu6.c          | 41 +-------------------
 2 files changed, 7 insertions(+), 41 deletions(-)

diff --git a/drivers/media/pci/intel/ipu6/ipu6-buttress.c b/drivers/media/pci/intel/ipu6/ipu6-buttress.c
index dbcf1aa87872..23c537e7ce1e 100644
--- a/drivers/media/pci/intel/ipu6/ipu6-buttress.c
+++ b/drivers/media/pci/intel/ipu6/ipu6-buttress.c
@@ -552,12 +552,16 @@ int ipu6_buttress_reset_authentication(struct ipu6_device *isp)
 int ipu6_buttress_map_fw_image(struct ipu6_bus_device *sys,
 			       const struct firmware *fw, struct sg_table *sgt)
 {
+	bool is_vmalloc = is_vmalloc_addr(fw->data);
 	struct page **pages;
 	const void *addr;
 	unsigned long n_pages;
 	unsigned int i;
 	int ret;
 
+	if (!is_vmalloc && !virt_addr_valid(fw->data))
+		return -EDOM;
+
 	n_pages = PHYS_PFN(PAGE_ALIGN(fw->size));
 
 	pages = kmalloc_array(n_pages, sizeof(*pages), GFP_KERNEL);
@@ -566,7 +570,8 @@ int ipu6_buttress_map_fw_image(struct ipu6_bus_device *sys,
 
 	addr = fw->data;
 	for (i = 0; i < n_pages; i++) {
-		struct page *p = vmalloc_to_page(addr);
+		struct page *p = is_vmalloc ?
+			vmalloc_to_page(addr) : virt_to_page(addr);
 
 		if (!p) {
 			ret = -ENOMEM;
diff --git a/drivers/media/pci/intel/ipu6/ipu6.c b/drivers/media/pci/intel/ipu6/ipu6.c
index 7bcd9c5a381a..2cf04251c9e7 100644
--- a/drivers/media/pci/intel/ipu6/ipu6.c
+++ b/drivers/media/pci/intel/ipu6/ipu6.c
@@ -503,45 +503,6 @@ static void ipu6_configure_vc_mechanism(struct ipu6_device *isp)
 	writel(val, isp->base + BUTTRESS_REG_BTRS_CTRL);
 }
 
-static int request_cpd_fw(const struct firmware **firmware_p, const char *name,
-			  struct device *device)
-{
-	const struct firmware *fw;
-	struct firmware *dst;
-	int ret = 0;
-
-	ret = request_firmware(&fw, name, device);
-	if (ret)
-		return ret;
-
-	if (is_vmalloc_addr(fw->data)) {
-		*firmware_p = fw;
-		return 0;
-	}
-
-	dst = kzalloc(sizeof(*dst), GFP_KERNEL);
-	if (!dst) {
-		ret = -ENOMEM;
-		goto release_firmware;
-	}
-
-	dst->size = fw->size;
-	dst->data = vmalloc(fw->size);
-	if (!dst->data) {
-		kfree(dst);
-		ret = -ENOMEM;
-		goto release_firmware;
-	}
-
-	memcpy((void *)dst->data, fw->data, fw->size);
-	*firmware_p = dst;
-
-release_firmware:
-	release_firmware(fw);
-
-	return ret;
-}
-
 static int ipu6_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
 	struct ipu6_buttress_ctrl *isys_ctrl = NULL, *psys_ctrl = NULL;
@@ -627,7 +588,7 @@ static int ipu6_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	if (ret)
 		return ret;
 
-	ret = request_cpd_fw(&isp->cpd_fw, isp->cpd_fw_name, dev);
+	ret = request_firmware(&isp->cpd_fw, isp->cpd_fw_name, dev);
 	if (ret) {
 		dev_err_probe(&isp->pdev->dev, ret,
 			      "Requesting signed firmware %s failed\n",
-- 
2.39.2