Re: [Bug-grub] [PATCH] Re: address@hidden: Bug#110793: should not secretly pass options to the kernel]

[Matt Wilson]

On Sat, Sep 01, 2001 at 10:40:45PM -0600, Derrik Pates wrote:
> 
> Why is it hurting anything? I was having a problem with Grub getting a
> 0-length mem range at 4 GB passed from the BIOS, but I have submitted a
> patch to the maintainers (now in CVS) to prevent that situation. It's
> probably not the worst thing for a sane bootloader to pass mem info to the
> kernel it loads. And unless the info it's passing is known to be bad, why
> prevent it from doing so?

The kernel does its own e820 call to determine the memory map after
the bootloader boots it.  If you pass mem=, the memory holes for
things like ACPI are trampled over (they're marked as usable memory as
far as the kernel is concerned.).  If GRUB passed mem=exactmap
address@hidden address@hidden, etc to give the memory map as detected by
GRUB, this would be OK.  Otherwise it's totally wrong to pass a whole
number to the kernel.

GRUB had this hack to always pass mem= for the old kernels that could
not detect >64MB properly.  This is ancient history and the hack is no
longer needed.

Cheers,

Matt