Skip to contentSkip to conversation

quiet MMCONFIG related printks

PATCH NAK Jesse Barnes 4 participants 9 messages Jan 2, 2007 – Jan 3, 2007

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 quiet MMCONFIG related printks

Using MMCONFIG for PCI config space access is simply an optimization, not a requirement. Therefore, when it can't be used, there's no need for KERN_ERR level message. This patch makes the message a KERN_INFO instead to reduce some of the noise in a kernel boot with the 'quiet' option. (Note that this has no effect on a normal boot, which is ridiculously verbose these days.)

Thanks,
Jesse

arch/i386/pci/mmconfig.c +2 -2
index e2616a2..b95e7f3 100644
@@ -210,9 +210,9 @@ void __init pci_mmcfg_init(int type)
 	if (type == 1 && !e820_all_mapped(pci_mmcfg_config[0].base_address,
 			pci_mmcfg_config[0].base_address + MMCONFIG_APER_MIN,
 			E820_RESERVED)) {
-		printk(KERN_ERR "PCI: BIOS Bug: MCFG area at %x is not E820-reserved\n",
+		printk(KERN_INFO "PCI: BIOS Bug: MCFG area at %x is not E820-reserved\n",
 				pci_mmcfg_config[0].base_address);
-		printk(KERN_ERR "PCI: Not using MMCONFIG.\n");
+		printk(KERN_INFO "PCI: Not using MMCONFIG.\n");
 		return;
 	}
 
NAK
#1 lore raw
Quoted 2 lines · On Mon, 1 Jan 2007 21:01:38 -0800 Jesse Barnes <[email protected]> wrote:

Using MMCONFIG for PCI config space access is simply an optimization, not a requirement. Therefore, when it can't be used, there's no need for

Some hardware reqires MCFG. In addition this is an error, a real error on the vendors part or ours and could indicate there are many other BIOS problems outstanding.

We shouldn't keep quiet about serious errors in tables, we need people to know and be able to take appropriate action (eg new BIOSen, refusing certifications etc).

NAK

#2 lore raw
Quoted 10 lines · Alan

On Mon, 1 Jan 2007 21:01:38 -0800

Quoted 3 lines · Jesse Barnes

Using MMCONFIG for PCI config space access is simply an optimization, not a requirement. Therefore, when it can't be used, there's no need for

Some hardware reqires MCFG. In addition this is an error, a real error on the vendors part or ours and could indicate there are many other BIOS problems outstanding.

Ok, I was mistaken then. However, I see this on several boxes, and the broken BIOSen out in the wild are unlikely to be fixed. Maybe this should really be a KERN_WARNING instead since it may indicate that some devices will fail to work?

Quoted 3 lines

We shouldn't keep quiet about serious errors in tables, we need people to know and be able to take appropriate action (eg new BIOSen, refusing certifications etc).

Ok.

Jesse

#3 lore raw
In reply to Alan
Quoted 6 lines · Alan
Quoted 2 lines · On Mon, 1 Jan 2007 21:01:38 -0800 Jesse Barnes <[email protected]> wrote:

Using MMCONFIG for PCI config space access is simply an optimization, not a requirement. Therefore, when it can't be used, there's no need for

Quoted 7 lines

Some hardware reqires MCFG. In addition this is an error, a real error on the vendors part or ours and could indicate there are many other BIOS problems outstanding.

We shouldn't keep quiet about serious errors in tables, we need people to know and be able to take appropriate action (eg new BIOSen, refusing certifications etc).

to some degree that is what the firmware kit is for ;)

Signature · 2 lines
if you want to mail me at work (you don't), use arjan (at) linux.intel.com
Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org
#4 lore raw
In reply to Jesse Barnes
Quoted 6 lines · Jesse Barnes

Using MMCONFIG for PCI config space access is simply an optimization, not a requirement. Therefore, when it can't be used, there's no need for KERN_ERR level message. This patch makes the message a KERN_INFO instead to reduce some of the noise in a kernel boot with the 'quiet' option. (Note that this has no effect on a normal boot, which is ridiculously verbose these days.)

this is wrong, please leave this loud complaint in...

Signature · 2 lines
if you want to mail me at work (you don't), use arjan (at) linux.intel.com
Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org
#5 lore raw
Quoted 9 lines · Arjan van de Ven
Quoted 6 lines · Jesse Barnes

Using MMCONFIG for PCI config space access is simply an optimization, not a requirement. Therefore, when it can't be used, there's no need for KERN_ERR level message. This patch makes the message a KERN_INFO instead to reduce some of the noise in a kernel boot with the 'quiet' option. (Note that this has no effect on a normal boot, which is ridiculously verbose these days.)

this is wrong, please leave this loud complaint in...

So the issues as I understand them:
o some BIOSes are broken and don't properly map MCFG space (though according to Petr V. reserving MCFG space in e820 is optional, so the test may be slightly wrong as-is)
o MCFG space is required for (many) PCIe devices (any regular PCI devices?)
o often, there's nothing the user can do to address the points above

So where does that leave us? I've got what I consider to be a stupid error message in my log. My system behavior isn't affected in any way (at least that I can tell), yet I get a loud complaint at boot time.

I guess I just have to live with it?

Thanks,
Jesse

#6 lore raw
Quoted 9 lines · Jesse Barnes
Quoted 8 lines · Arjan van de Ven
Quoted 6 lines · Jesse Barnes

Using MMCONFIG for PCI config space access is simply an optimization, not a requirement. Therefore, when it can't be used, there's no need for KERN_ERR level message. This patch makes the message a KERN_INFO instead to reduce some of the noise in a kernel boot with the 'quiet' option. (Note that this has no effect on a normal boot, which is ridiculously verbose these days.)

Quoted 6 lines
Quoted 1 line

this is wrong, please leave this loud complaint in...

So the issues as I understand them:
o some BIOSes are broken and don't properly map MCFG space (though according to Petr V. reserving MCFG space in e820 is optional, so the test may be slightly wrong as-is)

it's optional but it's the best test we have for "is the bios total crap" ;(

Quoted 2 lines

o MCFG space is required for (many) PCIe devices (any regular PCI devices?)

it's not required for *many* (it can't be, windows XP doesn't use MCFG), but it's required for some of the advanced PCI-E features

Quoted 1 line

o often, there's nothing the user can do to address the points above

other than complain to the vendor.

Quoted 3 lines

So where does that leave us? I've got what I consider to be a stupid error message in my log.

contact your bios vendor.

Quoted 4 lines

My system behavior isn't affected in any way
(at least that I can tell), yet I get a loud complaint at boot time.

I guess I just have to live with it?

We really really should complain about bios issues. If only to make sure vendors who do pay attention to linux have a chance of finding and fixing them (and via the firmware kit, several big vendors pay attention early on nowadays)

Signature · 2 lines
if you want to mail me at work (you don't), use arjan (at) linux.intel.com
Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org
#7 lore raw
Quoted 61 lines · Arjan van de Ven
Quoted 14 lines · Jesse Barnes
Quoted 11 lines · Arjan van de Ven
Quoted 8 lines · Jesse Barnes

Using MMCONFIG for PCI config space access is simply an optimization, not a requirement. Therefore, when it can't be used, there's no need for KERN_ERR level message. This patch makes the message a KERN_INFO instead to reduce some of the noise in a kernel boot with the 'quiet' option. (Note that this has no effect on a normal boot, which is ridiculously verbose these days.)

Quoted 9 lines
Quoted 3 lines

this is wrong, please leave this loud complaint in...

So the issues as I understand them:
o some BIOSes are broken and don't properly map MCFG space (though according to Petr V. reserving MCFG space in e820 is optional, so the test may be slightly wrong as-is)

it's optional but it's the best test we have for "is the bios total crap" ;(

Quoted 4 lines

o MCFG space is required for (many) PCIe devices (any regular PCI devices?)

it's not required for *many* (it can't be, windows XP doesn't use MCFG), but it's required for some of the advanced PCI-E features

Quoted 3 lines

o often, there's nothing the user can do to address the points above

other than complain to the vendor.

Quoted 3 lines

So where does that leave us? I've got what I consider to be a stupid error message in my log.

contact your bios vendor.

Quoted 3 lines

My system behavior isn't affected in any way
(at least that I can tell), yet I get a loud complaint at boot time.

Quoted 11 lines
Quoted 3 lines

I guess I just have to live with it?

We really really should complain about bios issues. If only to make sure vendors who do pay attention to linux have a chance of finding and fixing them (and via the firmware kit, several big vendors pay attention early on nowadays)

Hi Arjan,

Do you have a list of E-Mail addresses for the people we should be complaining to. I have an Asus VBI laptop that spews all kinds of error messages when I boot the firmware test kit cd, but I have no idea who to complain to about this, ASUS? Intel? Heck Intel puts out the test kit, if they are going to say it is VBI why don't the make the vendors do the bios correctly?

Regards,
Steve

Signature · 3 lines
"They that give up essential liberty to obtain temporary safety, deserve neither liberty nor safety."  (Ben Franklin)

"The course of history shows that as a government grows, liberty decreases."  (Thomas Jefferson)
#8 lore raw
In reply to Arjan van de Ven
Quoted 9 lines · Arjan van de Ven
Quoted 6 lines · Jesse Barnes

Using MMCONFIG for PCI config space access is simply an optimization, not a requirement. Therefore, when it can't be used, there's no need for KERN_ERR level message. This patch makes the message a KERN_INFO instead to reduce some of the noise in a kernel boot with the 'quiet' option. (Note that this has no effect on a normal boot, which is ridiculously verbose these days.)

this is wrong, please leave this loud complaint in...

Or maybe the test is just wrong. I'll try out the PCI MMConfig per-chipset patches to see if they work. That seems like a better long term solution anyway.

Thanks,
Jesse