Re: grub failure with mem > 8gb
Christophe Prud'homme <email@example.com> writes:
> [ Monday 09 August 2004 17:23 ]
> | Hi
> | I just got an amd64/smp with 8GB of ram
> | the latest install iso worked perfect except that grub failed, I had to enter the master block
> | (/dev/sda) by hand in the intercace and it worked.
> | After the reboot, I get the grub menu and the error message that was previously mentionned on this
> | list(I just subscribed today but checked the archives).
> | I read from an email last thursday from Goswin that
> | http://debian-amd64.alioth.debian.org/grub_0.95+cvs20040624-18.104.22.168.pure64.1_amd64.deb
> replying to myself
> I tested this package. here is what I get when I run it after
> an install from CD + chroot /target+ add amd64 mirror to /etc/apt/sources.list+apt-get install grub
That would be 0.95+cvs20040624-22.214.171.124.pure64 or
0.95+cvs20040624-126.96.36.199.pure64 then. Earlierversions didn't have the
lines below. :)
Please check which one since the later should only print the lines on
error and the numbers don't match an error condition.
> D: addr=0x555cd200
> D: mbi.mem_lower = 0x280
> D: RAW_ADDR (mbi.mem_lower * 10214) = 0x555f5000
> D: len = 0x1
> D: (addr - 0x100000) + len = 0x554cd201
> D: RAW_ADDR (0x1000) = 0x55556000 RAW_ADDR(0x100000)=0x55655000
Lets compare that to the code:
| if ((addr < RAW_ADDR (0x1000))
0x555cd200 < 0x55556000 => false
| || (addr < RAW_ADDR (0x100000)
0x555cd200 < 0x55655000 => true
| && RAW_ADDR (mbi.mem_lower * 1024) < (addr + len))
0x555f5000 < 0x555cd201 => false
| || (addr >= RAW_ADDR (0x100000)
0x555cd200 >= 0x55655000 => false
| && RAW_ADDR (mbi.mem_upper * 1024) < ((addr - 0x100000) + len)))
0x555f5000 < 0x554cd201 => false
| errnum = ERR_WONT_FIT;
errnum will NOT be set here. The later version should not print the
> does that help ?
This confuses me even more.
Are you sure there is no 'D: ERR_WONT_FIT <number>' printed?
> In the bios the mttr is set to continuous (I tested discrete but it was the same, at least to me)
> PS btw when running lilo from cd install I see a seg fault. is that supposed to happen ?
> Debian Developer
> Scienfific computing packages maintainer
> Fingerprint = 3703 50DE 7A9F 024E 0F26 0D07 A18F B40B D4BE 1450