Re: [PATCH-RFC] 4 of 4 - New problem logging macros, SCSI RAID device

[Jeff Garzik]

Larry Kessler wrote:
> Please see [PATCH-RFC] README 1st note.
> 
> Also note that this patch includes pci_problem.h, as does the eepro100.c
> device driver patch included in the 'README 1st' note.
>  
> Summary of this patch...
>  
>  drivers/scsi/ips.c
>     Device Driver for the IBM ServeRAID controller, with use of new 
>     macros replacing prink() for error conditions.
>  
>  include/linux/scsi_problem.h
>   -  scsi_host_detail() macro providing common information of interest
>      for scsi-class devices.    
>   -  scsi_host_problem and scsi_host_introduce macros   
> 
>  include/linux/pci_problem.h
> 
>   -  pci_detail() macro providing common information on a per class
>      basis when problems are being reported for devices of that class. 
>   -  pci_problem and pci_introduce macros

Bloat, bloat, and more bloat.  This API is not scalable at all, if we 
have to add a new header and new foo_problem() macros for every little 
subsystem in the kernel.

If you actually want to standardize some diagnostic messages, it is a 
huge mistake [as your scsi driver example shows] to continue to use 
random text strings followed by a typed attribute list.  If you really 
wanted to standardize logging, why continue to allow driver authors to 
printk driver-specific text strings in lieu of a standard string that 
applies to the same situation in N drivers.

I do encourage the clean-up of drivers logging and can see the utility 
of it, but you are really using a sledgehammer to drive in a carpet nail 
here...

	Jeff