Initrd file not loaded

Initrd file not loaded

[Jesús Velazquez]

[-- Attachment #1: Type: text/plain, Size: 407 bytes --]

Hi guys:

We are trying to use Grub2 for booting RedHat Enterprise Linux 4 update 2 on
Xeon Platforms. But, we found that the initrd file is not loaded by grub and
the kernel can't mount /. The linux kernel doesn't recognize the boot disk
and It dies.

 Do you know how the Linux Kernel knows where the initrd file is in memory?.
Is this a known issue?, Has it happened before?.

Thanks

Jesús

[-- Attachment #2: Type: text/html, Size: 1562 bytes --]

Re: Initrd file not loaded

[Yoshinori K. Okuji]

On Wednesday 31 May 2006 03:18, Jesús Velazquez wrote:
> We are trying to use Grub2 for booting RedHat Enterprise Linux 4 update 2
> on Xeon Platforms. But, we found that the initrd file is not loaded by grub
> and the kernel can't mount /. The linux kernel doesn't recognize the boot
> disk and It dies.

Please describe how you boot the linux.

Thanks,
Okuji

Re: Initrd file not loaded

[Jesús Velazquez]

[-- Attachment #1: Type: text/plain, Size: 1882 bytes --]

Hi:

I posted an issue loading the initrd file in a Xeon Platform, the
problem was the following:

>On Wednesday 31 May 2006 03:18, Jesús Velazquez wrote:
>>*  We are trying to use Grub2 for booting RedHat Enterprise Linux 4 update 2
*
>>*  on Xeon Platforms. But, we found that the initrd file is not
loaded by grub*
>>*  and the kernel can't mount /. The linux kernel doesn't recognize the boot*
>>* disk and It dies.
*

>Please describe how you boot the linux.

>Thanks,
>Okuji


We found the following misbehavior:

In the function grub_rescue_cmd_initrd (at file loader/i386/pc/linux.c), the
following lines of code (lines 335-336)

 if (!linux_mem_size && linux_mem_size < addr_max)
    addr_max = linux_mem_size;

Setup the addr_max to linux_mem_size when "mem=" option is not specified. I
mean the condition is inverted. That means
that if somebody doesn't specify the "mem=" option, linux_mem_size is 0 and
addr_max = linux_mem_size. If we remove
the inverted condition (!linux_mem_size), the initrd file is loaded without
any problem, and the system boots ok. We think that
the logic of this condition is if someone specifies the "mem=" option we
need to adjust the addr_max variable to point to the upper memory boundary,
right?

We are working with a 8-Way Xeon Platform with 8GB of memory. The weird
thing is that the original code works without any issues on Desktop
configurations (< 4GB of memory). We also forced addr_max to
GRUB_LINUX_INITRD_MAX_ADDRESS (lines 324-329 at file
loader/i386/pc/linux.c).

#if 0
  if (grub_le_to_cpu16 (lh->version) >= 0x0203)
    addr_max = grub_cpu_to_le32 (lh->initrd_addr_max);
#else
    addr_max = GRUB_LINUX_INITRD_MAX_ADDRESS;
#endif

We have 8 GB of memory (4G available for protected mode), so we don't bother
about it, we have enough memory.

Regards,

Jesus & Oscar

[-- Attachment #2: Type: text/html, Size: 2154 bytes --]

Re: Initrd file not loaded

[Yoshinori K. Okuji]

On Monday 05 June 2006 21:44, Jesús Velazquez wrote:
> In the function grub_rescue_cmd_initrd (at file loader/i386/pc/linux.c),
> the following lines of code (lines 335-336)
>
>  if (!linux_mem_size && linux_mem_size < addr_max)
>     addr_max = linux_mem_size;
>
> Setup the addr_max to linux_mem_size when "mem=" option is not specified. I
> mean the condition is inverted.

Uggh, you are right. I will fix this.

> #if 0
>   if (grub_le_to_cpu16 (lh->version) >= 0x0203)
>     addr_max = grub_cpu_to_le32 (lh->initrd_addr_max);
> #else
>     addr_max = GRUB_LINUX_INITRD_MAX_ADDRESS;
> #endif

Hmm, this is strange. The boot protocl 2.03 is supposed to put a right value 
there. What value is written in your kernel?

Thanks,
Okuji