Re: [folded] kernelh-add-printk_ratelimited-and-pr_level_rl-rename.patch removed from -mm tree

Re: [folded] kernelh-add-printk_ratelimited-and-pr_level_rl-rename.patch removed from -mm tree

[Peter Zijlstra]

On Mon, 2009-12-14 at 17:08 -0800, akpm@linux-foundation.org wrote:
> s/_rl/_ratelimited/g

do we feel this pr_* wankery is worth the hassle? I'd as soon send a
patch removing all this crap.

Re: [folded] kernelh-add-printk_ratelimited-and-pr_level_rl-rename.patch removed from -mm tree

[Andrew Morton]

On Tue, 15 Dec 2009 11:28:02 +0100
Peter Zijlstra <peterz@infradead.org> wrote:

> On Mon, 2009-12-14 at 17:08 -0800, akpm@linux-foundation.org wrote:
> > s/_rl/_ratelimited/g
> 
> do we feel this pr_* wankery is worth the hassle? I'd as soon send a
> patch removing all this crap.

I thought it was just me.  No, I find it a net negative, but I don't
work on drivers all day.

otoh, every second driver seems to find it necessary to invent its own
handy printk wrappers, so being able to point people at standardized
ones does offer some improvement over that.

Re: [folded] kernelh-add-printk_ratelimited-and-pr_level_rl-rename.patch removed from -mm tree

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 1010 bytes --]

On Tue, 15 Dec 2009 11:28:02 +0100, Peter Zijlstra said:
> On Mon, 2009-12-14 at 17:08 -0800, akpm@linux-foundation.org wrote:
> > s/_rl/_ratelimited/g
> 
> do we feel this pr_* wankery is worth the hassle? I'd as soon send a
> patch removing all this crap.

pr_foo() instead of printk(KERN_FOO) is probably worth the hassle, as it
allows more selective inclusion of messages if you're trying to build an
embedded kernel.  It's easy to say "I want pr_warning() to stay in, but
lower levels compile to nothing". Trying to keep a 'printk(KERN_WARNING'
while making a printk(KERN_DEBUG go away is just asking for some truly
astounding pre-processor gyrations.

Not an issue for my laptop with 4G of RAM - byt my cell phone says it has
26M of memory, 7M used (presumably all the software, my contact list isn't
*that* big).  No idea what the actual kernel is, but I know whatever it is,
the coders were using that kernel's equivalent of CONFIG_EMBEDDED and tossing
stuff over the side if they didn't need it. ;)



[-- Attachment #2: Type: application/pgp-signature, Size: 227 bytes --]

Re: [folded] kernelh-add-printk_ratelimited-and-pr_level_rl-rename.patch removed from -mm tree

[Peter Zijlstra]

On Tue, 2009-12-15 at 17:44 -0500, Valdis.Kletnieks@vt.edu wrote:
> On Tue, 15 Dec 2009 11:28:02 +0100, Peter Zijlstra said:
> > On Mon, 2009-12-14 at 17:08 -0800, akpm@linux-foundation.org wrote:
> > > s/_rl/_ratelimited/g
> > 
> > do we feel this pr_* wankery is worth the hassle? I'd as soon send a
> > patch removing all this crap.
> 
> pr_foo() instead of printk(KERN_FOO) is probably worth the hassle, as it
> allows more selective inclusion of messages if you're trying to build an
> embedded kernel.  It's easy to say "I want pr_warning() to stay in, but
> lower levels compile to nothing". Trying to keep a 'printk(KERN_WARNING'
> while making a printk(KERN_DEBUG go away is just asking for some truly
> astounding pre-processor gyrations.

So we are depricating printk()?

Last time I asked that the answer was no, at which point there is
absolutely no point in using pr_* wankery.

And I much prefer the printk() thing, because
 1) my fingers know it
 2) it looks like the userspace printf thing
 3) its an easier pattern to grep for