Skip to contentSkip to conversation

drivers-core: beware dev_printk() from printing nonsense

PATCH In review Sebastian Andrzej Siewior 3 participants 4 messages Aug 20, 2012 Driver core

lore
Share & tools
lore

Sebastian Andrzej Siewior

author
#0 lore raw

PATCH 1/1 drivers-core: beware dev_printk() from printing nonsense

on boot I saw plenty of rubbish like:

|ACPI: PCI Root Bridge [PCI0] (domain 0000 [bus 00-ff])
|pci_root PNP0A03:00: Force enabled HPET at 0x%lx
|host bridge window [io  0x0000-0x0cf7] (ignored)
|pci_root PNP0A03:00: Force enabled HPET at 0x%lx
|host bridge window [io  0x0d00-0xffff] (ignored)
|pci_root PNP0A03:00: Force enabled HPET at 0x%lx
|ACPI: PCI Interrupt Link [LNKA] enabled at IRQ 10
|virtio-pci 0000:00:05.0: Force enabled HPET at 0x%lx

First I wonder myself why the kernel Force enables more than one HPET and why the address isn't printed. Then I wasn't sure why virtio-pci does the same thing.
As it turns out commit af7f2158fdee ("drivers-core: make structured logging play nice with dynamic-debug") has probably an off by one error. Since commit 314ba3520 ("printk: add kern_levels.h to make KERN_<LEVEL> available for asm use") the whole shrunk by one byte. And for some reason the compiler put the HPET string just after one of the kernel log levels so level[3] was always true.

This should make the rubish go away, I am not sure if the dynamic debug is still working. Jim could you please try?

Cc Jim Cromie <[email protected]>
Cc Jason Baron <[email protected]>
Cc stable <[email protected]>
Cc Kay Sievers <[email protected]>
Cc Joe Perches <[email protected]>
Signed-off-by Sebastian Andrzej Siewior <[email protected]>
---
 drivers/base/core.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/base/core.c +1 -1
index cdd01c5..5a864c3 100644
@@ -1912,7 +1912,7 @@ int __dev_printk(const char *level, const struct device *dev,
 				    "DEVICE=+%s:%s", subsys, dev_name(dev));
 	}
 skip:
-	if (level[3])
+	if (level[sizeof(KERN_ERR) - 1])
 		level_extra = &level[3]; /* skip past "<L>" */
 
 	return printk_emit(0, level[1] - '0',
Signature · 1 line
1.7.10.4
maintainer
#1 lore raw
Quoted 39 lines · Sebastian Andrzej Siewior

on boot I saw plenty of rubbish like:

|ACPI: PCI Root Bridge [PCI0] (domain 0000 [bus 00-ff])
|pci_root PNP0A03:00: Force enabled HPET at 0x%lx
|host bridge window [io  0x0000-0x0cf7] (ignored)
|pci_root PNP0A03:00: Force enabled HPET at 0x%lx
|host bridge window [io  0x0d00-0xffff] (ignored)
|pci_root PNP0A03:00: Force enabled HPET at 0x%lx
|ACPI: PCI Interrupt Link [LNKA] enabled at IRQ 10
|virtio-pci 0000:00:05.0: Force enabled HPET at 0x%lx

First I wonder myself why the kernel Force enables more than one HPET and why the address isn't printed. Then I wasn't sure why virtio-pci does the same thing.
As it turns out commit af7f2158fdee ("drivers-core: make structured logging play nice with dynamic-debug") has probably an off by one error. Since commit 314ba3520 ("printk: add kern_levels.h to make KERN_<LEVEL> available for asm use") the whole shrunk by one byte. And for some reason the compiler put the HPET string just after one of the kernel log levels so level[3] was always true.

This should make the rubish go away, I am not sure if the dynamic debug is still working. Jim could you please try?

Cc Jim Cromie <[email protected]>
Cc Jason Baron <[email protected]>
Cc stable <[email protected]>
Cc Kay Sievers <[email protected]>
Cc Joe Perches <[email protected]>
Signed-off-by Sebastian Andrzej Siewior <[email protected]>
---
 drivers/base/core.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/base/core.c +0 -0
index cdd01c5..5a864c3 100644

drivers/base/core.c int __dev_printk(const char *level, const struct device *dev, · line 1915

Quoted 5 lines
drivers/base/core.c +1 -1
@@ -1912,7 +1912,7 @@ int __dev_printk(const char *level, const struct device *dev,
 				    "DEVICE=+%s:%s", subsys, dev_name(dev));
 	}
 skip:
-	if (level[3])
+	if (level[sizeof(KERN_ERR) - 1])

This has already been fixed in my driver-core tree and will go to Linus in a few hours.

thanks,

greg k-h

author
#2 lore raw
Quoted 6 lines · Greg Kroah-Hartman

drivers/base/core.c int __dev_printk(const char *level, const struct device *dev, · line 1915

Quoted 3 lines

skip:

drivers/base/core.c +1 -1
-	if (level[3])
+	if (level[sizeof(KERN_ERR) - 1])

This has already been fixed in my driver-core tree and will go to Linus in a few hours.

Ah. So I'm too late. Care to send a patch which replace [2] with the sizeof() operator like above?

Quoted 4 lines

thanks,

greg k-h

Sebastian

#3 lore raw
Quoted 10 lines · Sebastian Andrzej Siewior
Quoted 6 lines · Greg Kroah-Hartman

drivers/base/core.c int __dev_printk(const char *level, const struct device *dev, · line 1915

Quoted 3 lines

skip:

drivers/base/core.c +1 -1
-	if (level[3])
+	if (level[sizeof(KERN_ERR) - 1])

This has already been fixed in my driver-core tree and will go to Linus in a few hours.

Ah. So I'm too late. Care to send a patch which replace [2] with the sizeof() operator like above?

Nope, the better fix is to use a different mechanism.

See commit acc8fa41ad31c5
("printk: add generic functions to find KERN_<LEVEL> headers")

and

commit 04d2c8c83d0e3a
("printk: convert the format for KERN_<LEVEL> to a 2 byte pattern")

I'll submit that "better fix" shortly after Greg's update has been pulled/merged by Linus.

cheers, Joe