Re: kernel too big

["H. Peter Anvin" <hpa@transmeta.com>]

Andries.Brouwer@cwi.nl wrote:
> In 2.5.61 hpa changed
> 
> -      /* 0x28000*16 = 2.5 MB, conservative estimate for the current maximum */
> -      if (sys_size > (is_big_kernel ? 0x28000 : DEF_SYSSIZE))
> +      /* 0x40000*16 = 4.0 MB, reasonable estimate for the current maximum */
> +      if (sys_size > (is_big_kernel ? 0x40000 : DEF_SYSSIZE))
>               die("System is too big. ...");
> 
> (with comment "bootsect removal").
> 
> Today I find for a 2.6.3 machine that it boots with 2994 kB and
> crashes at boot time with 3005 kB.
> 
> Thus, it looks like this "reasonable estimate" is too optimistic.
> 
> If I understand correctly, the real requirement is that
> _end must stay below 8MB (unless the initial page tables
> in head.S are made larger). A crash occurs with _end = c07fcf8c.
> 
> Maybe these "conservative" or "reasonable" estimates should be
> replaced by a text on _end?
> 

What we should do is to actually look at the limit that matters.

Also, if the limit really is around 3 MB compressed we probably need to
extend it to 16 MB uncompressed; we're a little too close for comfort.