Running on a GIGABYTE motherboard with an AMD Athlon 650Mhz processor and 640MB RAM, lilo does not recognise the amount of memory. Adding 'append = "mem=640M" ' in lilo.conf does not help. Only if you type 'linux mem=640M ' at boot time helps to use the full 640MB. Is there any patch for this problem.
Where did you add the append=? In the stanza of your image? Can you attach your lilo.conf to this bug report? In all actuality, the kernel itself is to blame for not properly finding the correct amount of available memory, and the function of passing a mem= parameter to your kernel is a kludge. Lilo itself is only functioning to pass a specified parameter. Also, if you do have the append= in the image stanza, try using: append="mem=640M" the lack of spaces may help (never tried w/ the extra whitespace, so we'll see)
Lilo isn't responsible for telling the kernel how much memory there is, the kernel detects the memory for itself. The problem you describe is consistent with a BIOS that doesn't implement the 0xe820 BIOS extended memory information call (which keeps the kernel from detecting the actual amount of RAM in your machine) and with a mis-placement of the append= stanza in the lilo.conf file as mentioned by jmm. Since it isn't feasible to correct the lack of a BIOS 0xe820 call, the best option is to correct the lilo.conf file so that it will pass the information at each boot (keeping in mind that you must re-run the lilo program each time you modify the lilo.conf file for the changes to take effect, just in case you forgot that step and that's why lilo isn't passing the mem= parameter to the kernel). It's possible that later versions of the OS with later kernels might get around the problem by enabling different memory detection routines, such as ACPI based ones, but that's not a feasible solution to your problem.