RE: [RFC PATCH 1/5] printk: implement pr_cont_t

[David Laight]

From: Joe Perches
> Sent: 20 August 2020 01:34
> 
> On Thu, 2020-08-20 at 01:32 +0206, John Ogness wrote:
> > Implement a new buffering mechanism for pr_cont messages.
> >
> > Old mechanism syntax:
> >
> >     printk(KERN_INFO "text");
> >     printk(KERN_CONT " continued");
> >     printk(KERN_CONT "\n");
> >
> > New mechanism syntax:
> >
> >     pr_cont_t c;
> >
> >     pr_cont_begin(&c, KERN_INFO "text");
> 
> bikeshed:
> 
> I suggest:
> 
> 	printk_begin(&printk_context, fmt, ...)
> 	printk_continue(&printk_context, fmt, ...) (maybe printk_next())
> 	printk_end(&printk_context, fmt, ...)

I see some very long source lines looming ...

Blue bikeshed:

You'd probably want printk_end(&ctx, NULL) to work.
Although the example doesn't show where the '\n' comes from.
Although I guess it is now inferred and actually deleted
from 'normal' printk() call.

I've no idea how you'd 'size' the number of buffers.
You could use kmalloc(), perhaps falling back on a local buffer.
While might lead to:
	pr_init(&ctx, level, GFP_KERNEL);
	pr_cont(&ctx, fmt, ...);
	pr_flush(&ctx);

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)