drivers/misc: Aspeed LPC control fix compile error and warning

PATCH In review Cyril Bur 2 participants 3 messages Mar 22, 2017

Subsystems (3)

Linux-next tree Char and misc drivers Arm/Aspeed machine support

lore

Share & tools

lore

Cyril Bur

author

2017-03-22 03:13 UTC #0 lore raw

PATCH 1/1 drivers/misc: Aspeed LPC control fix compile error and warning

pgprot_dmachoerent() is not defined on every architecture. Having COMPILE_TEST set for the driver causes it to be compiled on architectures which do not have pgprot_dmachoerent():
drivers/misc/aspeed-lpc-ctrl.c: In function 'aspeed_lpc_ctrl_mmap':
drivers/misc/aspeed-lpc-ctrl.c:51:9: error: implicit declaration of
function 'pgprot_dmacoherent' [-Werror=implicit-function-declaration] prot = pgprot_dmacoherent(prot);

There are two possible solutions:
1. Remove COMPILE_TEST to ensure the driver is only compiled on ARM
2. Use pgprot_noncached() instead of pgprot_dmachoerent()

The first option results in less compile testing of the LPC control driver which is undesirable.
The second option uses a function that is declared on all architectures and therefore should always build. Currently there is no practical difference between pgprot_noncached() and pgprot_dmachoerent() for the aspeed chips that this driver is compatible with. The reason for pgprot_dmachoerent() was that there may be chips made at some point in the future that could include hardware that pgprot_dmachoerent() could optimise for. As none of this hardware has even been announced there isn't really a need for pgprot_dmachoerent().

Using pgprot_noncached() is completely correct and optimal for all existing hardware on which the LPC control driver will run.

This commit also addresses that phys_addr_t should be printed using %pap rather than %x:
In file included from include/linux/miscdevice.h:6:0,
from drivers/misc/aspeed-lpc-ctrl.c:11:
drivers/misc/aspeed-lpc-ctrl.c: In function 'aspeed_lpc_ctrl_probe':
drivers/misc/aspeed-lpc-ctrl.c:232:17: warning: format '%x' expects
argument of type 'unsigned int', but argument 3 has type 'phys_addr_t {aka long long unsigned int}' [-Wformat=]
dev_info(dev, "Loaded at 0x%08x (0x%08x)\n",

Signed-off-by Cyril Bur <[email protected]>

---
 drivers/misc/aspeed-lpc-ctrl.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

drivers/misc/aspeed-lpc-ctrl.c +3 -3

index f6acbe1d9378..c654651a7b6d 100644
@@ -48,7 +48,7 @@ static int aspeed_lpc_ctrl_mmap(struct file *file, struct vm_area_struct *vma)
 		return -EINVAL;
 
 	/* ast2400/2500 AHB accesses are not cache coherent */
-	prot = pgprot_dmacoherent(prot);
+	prot = pgprot_noncached(prot);
 
 	if (remap_pfn_range(vma, vma->vm_start,
 		(lpc_ctrl->mem_base >> PAGE_SHIFT) + vma->vm_pgoff,

@@ -229,8 +229,8 @@ static int aspeed_lpc_ctrl_probe(struct platform_device *pdev)
 	if (rc)
 		dev_err(dev, "Unable to register device\n");
 	else
-		dev_info(dev, "Loaded at 0x%08x (0x%08x)\n",
-			lpc_ctrl->mem_base, lpc_ctrl->mem_size);
+		dev_info(dev, "Loaded at %pap (0x%08x)\n",
+			&lpc_ctrl->mem_base, lpc_ctrl->mem_size);
 
 	return rc;
 }

Signature · 1 line

2.12.0

Greg KH

2017-03-22 06:07 UTC #1 lore raw

Quoted 36 lines · Cyril Bur

pgprot_dmachoerent() is not defined on every architecture. Having COMPILE_TEST set for the driver causes it to be compiled on architectures which do not have pgprot_dmachoerent():
drivers/misc/aspeed-lpc-ctrl.c: In function 'aspeed_lpc_ctrl_mmap':
drivers/misc/aspeed-lpc-ctrl.c:51:9: error: implicit declaration of
function 'pgprot_dmacoherent' [-Werror=implicit-function-declaration] prot = pgprot_dmacoherent(prot);
There are two possible solutions:
1. Remove COMPILE_TEST to ensure the driver is only compiled on ARM
2. Use pgprot_noncached() instead of pgprot_dmachoerent()
The first option results in less compile testing of the LPC control driver which is undesirable.
The second option uses a function that is declared on all architectures and therefore should always build. Currently there is no practical difference between pgprot_noncached() and pgprot_dmachoerent() for the aspeed chips that this driver is compatible with. The reason for pgprot_dmachoerent() was that there may be chips made at some point in the future that could include hardware that pgprot_dmachoerent() could optimise for. As none of this hardware has even been announced there isn't really a need for pgprot_dmachoerent().
Using pgprot_noncached() is completely correct and optimal for all existing hardware on which the LPC control driver will run.
This commit also addresses that phys_addr_t should be printed using %pap rather than %x:
In file included from include/linux/miscdevice.h:6:0,
from drivers/misc/aspeed-lpc-ctrl.c:11:
drivers/misc/aspeed-lpc-ctrl.c: In function 'aspeed_lpc_ctrl_probe':
drivers/misc/aspeed-lpc-ctrl.c:232:17: warning: format '%x' expects
argument of type 'unsigned int', but argument 3 has type 'phys_addr_t {aka long long unsigned int}' [-Wformat=]
dev_info(dev, "Loaded at 0x%08x (0x%08x)\n",

Quoted 1 line

Signed-off-by Cyril Bur <[email protected]>

You forgot to include a "Reported-by:" line here giving credit to who pointed this out (the kbuild bot). I'll go add it myself...

thanks,

greg k-h

Greg KH

2017-03-22 06:16 UTC #2 lore raw

In reply to Cyril Bur

Quoted 23 lines · Cyril Bur

pgprot_dmachoerent() is not defined on every architecture. Having COMPILE_TEST set for the driver causes it to be compiled on architectures which do not have pgprot_dmachoerent():
drivers/misc/aspeed-lpc-ctrl.c: In function 'aspeed_lpc_ctrl_mmap':
drivers/misc/aspeed-lpc-ctrl.c:51:9: error: implicit declaration of
function 'pgprot_dmacoherent' [-Werror=implicit-function-declaration] prot = pgprot_dmacoherent(prot);
There are two possible solutions:
1. Remove COMPILE_TEST to ensure the driver is only compiled on ARM
2. Use pgprot_noncached() instead of pgprot_dmachoerent()
The first option results in less compile testing of the LPC control driver which is undesirable.
The second option uses a function that is declared on all architectures and therefore should always build. Currently there is no practical difference between pgprot_noncached() and pgprot_dmachoerent() for the aspeed chips that this driver is compatible with. The reason for pgprot_dmachoerent() was that there may be chips made at some point in the future that could include hardware that pgprot_dmachoerent() could optimise for. As none of this hardware has even been announced there isn't really a need for pgprot_dmachoerent().

Quoted 5 lines

Using pgprot_noncached() is completely correct and optimal for all existing hardware on which the LPC control driver will run.
This commit also addresses that phys_addr_t should be printed using %pap rather than %x:

Also in the future, only do one thing per-patch, this is two different things at the same time. I'll take this for now, but please be more careful going forward.

thanks,

greg k-h