> On Oct 10, 2020, at 2:50 PM, Paul Menzel <pmen...@molgen.mpg.de> wrote:
> 
> Dear Linux folks,
> 
> 
> On an Asus F2A85-M PRO with two 2 GB RAM modules installed, and an APU 
> device, building Linux with `ARCH=i386` and `CONFIG_HIGHMEM4G=y` only 3 GB 
> seem to be detected: 2.2 GB according to `free -h` plus the 768 MB for APU 
> graphics memory).
> 
>> [    0.065059] Memory: 2285148K/2324512K available (11785K kernel code, 892K 
>> rwdata, 2748K rodata, 668K init, 544K bss, 39364K reserved, 0K cma-reserved, 
>> 1423796K highmem)
> 
>> [    0.402082] calling  populate_rootfs+0x0/0xa1 @ 1
> 
> 
>> $ free -h
>>              total        used        free      shared  buff/cache   
>> available
>> Mem:          2,2Gi        72Mi       2,0Gi        13Mi       130Mi       
>> 2,0Gi
>> Swap:            0B          0B          0B
> 
> With `CONFIG_HIGHMEM64G=y` the whole 4 GB are used (3.1 GB + 768 MB for APU 
> graphics memory).
> 
>> [    0.121036] Memory: 3229952K/3356700K available (10301K kernel code, 821K 
>> rwdata, 2700K rodata, 708K init, 540K bss, 126748K reserved, 0K 
>> cma-reserved, 2449840K highmem)
> 
>> [    0.450668] calling  populate_rootfs+0x0/0xa1 @ 1
> 
> The Kconfig help text for `HIGHMEM4G` says:
> 
>> Select this if you have a 32-bit processor and between 1 and 4               
>>                                                                  │   
>> gigabytes of physical RAM.
> 
> As I only have 4 GB, I chose that to save 50 ms (maybe only due to less 
> memory detected), and thought non-PAE kernels can use 4 GB of memory.
> 

Your memory map contains:

BIOS-e820: [mem 0x0000000100001000-0x000000013effffff] usable

That’s 0x3effffff bytes mapped at a physical address above 4G.

The 4G limit without PAE isn’t, strictly speaking, about how much RAM can be 
used; it’s about the maximum usable physical address.  One might wonder why 
your firmware set up your memory map like this.

But there’s a much bigger issue: why on Earth are you running a 32-bit kernel 
on a relatively new machine like this?

Reply via email to