Re: [PATCH] WARN(): add a \n to the message printk

[Linus Torvalds]

On Mon, 15 Jun 2009, Linus Torvalds wrote:
> 
> NOTE! This is, of course, totally untested. And we're bound to have 
> continuation printk's that don't have the KERN_CONT at front, and need 
> them added, but I think this is generally a saner model than what we have 
> now, or your suggested explicit addition of '\n'.

Ok, it's tested now.

It works, and yes, it does show cases of insanity: both missing KERN_CONT 
(common), and _extra_ KERN_CONT (odd).

For example, the ACPI printk's seem to have pointless KERN_CONT's in them, 
an I get printouts like:

	[    0.000000] <c>ACPI: RSDP 00000000000fe020 00024 (v02 INTEL )
	[    0.000000] <c>ACPI: XSDT 00000000bf7fe120 0006C (v01 INTEL  DX58SO   0000084F      01000013)
	[    0.000000] <c>ACPI: FACP 00000000bf7fd000 000F4 (v03 INTEL  DX58SO   0000084F MSFT 0100000D)
	...

where that "<c>" is just because ACPI does something odd and pointless.

The lack of KERN_CONT shows up in printouts like

	[   26.626492] CPU: L1 I cache: 32K
	[   26.626492] , L1 D cache: 32K
	...
	[   26.826201] ACPI: (supports S0
	[   26.826243]  S5
	[   26.826326] )
	...
	[   26.839617] ACPI: PCI Interrupt Link [LNKA] (IRQs
	[   26.839660]  3
	[   26.839741]  4
	[   26.839823]  5
	[   26.839904]  6
	[   26.839985]  7
	[   26.840067]  9
	[   26.840148]  10
	[   26.840230]  *11
	[   26.840313]  12
	[   26.840395]  14
	[   26.840476]  15
	[   26.840558] )
	...
	[   26.964999] ACPI: CPU0 (power states:
	[   26.965040]  C1[C1]
	[   26.965123]  C2[C3]
	[   26.965205] )
	...
	[   27.231268] ata_piix 0000:00:1f.5: MAP [
	[   27.231309]  P0
	[   27.231390]  --
	[   27.231472]  P1
	[   27.231553]  --
	[   27.231635]  ]
	...
	[   28.092534]  sda:
	[   28.092820]  sda1
	[   28.092910]  sda2
	...

where the kernel now added a newline because they were separate printk's 
and didn't have KERN_CONT on the continuation.

But despite seeing these kinds of things, I do think the patch is the 
right thing to do. It just shows that since KERN_CONT didn't use to _do_ 
anything, people obviously mis-used it. It's the old "if it wasn't tested, 
it's buggy" thing, but none of these look in the least like serious 
problems to the approach.

Comments? We could make it remove the unnecessary '<c>' things (so that 
you can always add KERN_CONT if you _want_ to), but the patch as-is will 
show them because I think it's useful to see them. 

			Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/