Re: [PATCH 1/4] stringbuf: A string buffer implementation

["Pekka Enberg" <penberg@cs.helsinki.fi>]

Hi Rusty,

On 10/26/07, Rusty Russell <rusty@rustcorp.com.au> wrote:
>        This just seems like more optimization and complexity that we need.  Interfaces
> using vsnprintf don't seem like good candidates for optimization.
>
> How about this?  It's as simple as I could make it...

FWIW I like this patch better.

On 10/26/07, Rusty Russell <rusty@rustcorp.com.au> wrote:
> +void sb_printf_append(struct stringbuf **sb, gfp_t gfp, const char *fmt, ...)
> +{
> +       unsigned int fmtlen, len;
> +       va_list args;
> +       struct stringbuf *oldsb = *sb;
> +
> +       if (oldsb->buf == enomem_string)
> +               return;
> +
> +       va_start(args, fmt);
> +       fmtlen = vsnprintf(NULL, 0, fmt, args);
> +       va_end(args);
> +
> +       len = oldsb ? strlen(oldsb->buf) : 0;
> +       *sb = krealloc(oldsb, len + fmtlen + 1, gfp);
> +       if (!*sb) {
> +               kfree(oldsb);
> +               *sb = (struct stringbuf *)enomem_string;

Why don't we just return -ENOMEM here just like all other APIs in the
kernel? And I wonder if it makes more sense to store gfp_flags in
struct stringbuf and always use that? I mean, why would you want to
sometimes do GFP_ATOMIC and GFP_KERNEL allocations for the same
buffer?

                                  Pekka