This patch is not working well on actual machines. While it worked fine in
qemu, on real machines with multi-gigabytes of memory, attempting to boot
it up with limited memory (while max_paddr is still quite high) is not
working well; I assume it is because the arrays based on max_paddr eg
"struct page" etc is consuming too much memory.

I will try to limit max_paddr and see if that works better.

Does akaros already take a mem= option that I can use instead?

On Mon, Feb 29, 2016 at 12:43 PM, ron minnich <[email protected]> wrote:

> Could we, instead, just have a mem= boot time command line parameter?
>
> ron
>
> On Mon, Feb 29, 2016 at 11:02 AM 'Kanoj Sarcar' via Akaros <
> [email protected]> wrote:
>
>> Developer tool to limit RAM used (eg to restrict to only node0 memory
>> etc).
>>
>> Signed-off-by: Kanoj Sarcar <[email protected]>
>> ---
>>  kern/arch/x86/page_alloc.c | 13 +++++++++++++
>>  1 file changed, 13 insertions(+)
>>
>> diff --git a/kern/arch/x86/page_alloc.c b/kern/arch/x86/page_alloc.c
>> index da3d8b0..ad0aaa4 100644
>> --- a/kern/arch/x86/page_alloc.c
>> +++ b/kern/arch/x86/page_alloc.c
>> @@ -48,6 +48,10 @@ static bool pa64_is_in_kernel(uint64_t paddr)
>>         return (EXTPHYSMEM <= paddr) && (paddr < PADDR(end));
>>  }
>>
>> +#define        MAX_LIMIT_RAM   0xFFFFFFFFFFFFFFFFULL
>> +static uint64_t max_usable_paddr = MAX_LIMIT_RAM;      /* cmdline
>> option? */
>> +#define        USABLE_RAM(_pa_)        ((_pa_) < max_usable_paddr)
>> +
>>  /* Helper.  For every page in the entry, this will determine whether or
>> not the
>>   * page is free, and handle accordingly.  All pages are marked as busy by
>>   * default, and we're just determining which of them could be free. */
>> @@ -89,6 +93,11 @@ static void parse_mboot_region(struct
>> multiboot_mmap_entry *entry, void *data)
>>                  * bootzone.  If it was in the bootzone, we already
>> skipped it. */
>>                 if (pa64_is_in_kernel(i))
>>                         continue;
>> +
>> +               /* If beyond what we want to allocate, give up */
>> +               if (!USABLE_RAM(i))
>> +                       return;
>> +
>>                 track_free_page(pa64_to_page(i));
>>         }
>>  }
>> @@ -98,6 +107,10 @@ static void check_range(uint64_t start, uint64_t end,
>> int expect)
>>         int ref;
>>         if (PGOFF(start))
>>                 printk("Warning: check_range given unaligned addr
>> 0x%016llx\n", start);
>> +
>> +       if ((expect == 0) && (!USABLE_RAM(end)))
>> +               end = max_usable_paddr;
>> +
>>         for (uint64_t i = start; i < end; i += PGSIZE)  {
>>                 ref = kref_refcnt(&pa64_to_page(i)->pg_kref);
>>                 if (ref != expect) {
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "Akaros" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> To post to this group, send email to [email protected].
>> For more options, visit https://groups.google.com/d/optout.
>>
>


-- 
Thanks,

Kanoj

-- 
You received this message because you are subscribed to the Google Groups 
"Akaros" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to