Re: [PATCH v3 2/5] memblock: Improve memblock to support allocationfrom lower address.

[Zhang Yanfei]

Hello toshi-san,

On 09/14/2013 05:53 AM, Toshi Kani wrote:
> On Fri, 2013-09-13 at 17:30 +0800, Tang Chen wrote:
>  :
>> @@ -100,8 +180,7 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t start,
>>  					phys_addr_t end, phys_addr_t size,
>>  					phys_addr_t align, int nid)
>>  {
>> -	phys_addr_t this_start, this_end, cand;
>> -	u64 i;
>> +	phys_addr_t ret;
>>  
>>  	/* pump up @end */
>>  	if (end == MEMBLOCK_ALLOC_ACCESSIBLE)
>> @@ -111,18 +190,22 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t start,
>>  	start = max_t(phys_addr_t, start, PAGE_SIZE);
>>  	end = max(start, end);
>>  
>> -	for_each_free_mem_range_reverse(i, nid, &this_start, &this_end, NULL) {
>> -		this_start = clamp(this_start, start, end);
>> -		this_end = clamp(this_end, start, end);
>> +	if (memblock_direction_bottom_up()) {
>> +		/*
>> +		 * MEMBLOCK_ALLOC_ACCESSIBLE is 0, which is less than the end
>> +		 * of kernel image. So callers specify MEMBLOCK_ALLOC_ACCESSIBLE
>> +		 * as @start is OK.
>> +		 */
>> +		start =	max(start, __pa_symbol(_end)); /* End of kernel image. */
>>  
>> -		if (this_end < size)
>> -			continue;
>> +		ret = __memblock_find_range(start, end, size, align, nid);
>> +		if (ret)
>> +			return ret;
>>  
>> -		cand = round_down(this_end - size, align);
>> -		if (cand >= this_start)
>> -			return cand;
>> +		pr_warn("memblock: Failed to allocate memory in bottom up direction. Now try top down direction.\n");
> 
> Is there any chance that this retry will succeed given that start and
> end are still the same?

Thanks for pointing this. We've made a mistake here. If the original start
is less than __pa_symbol(_end), and if the bottom up search fails, then
the top down search deserves a try with the original start argument.

> 
> Thanks,
> -Toshi
> 
> 
>>  	}
>> -	return 0;
>> +
>> +	return __memblock_find_range_rev(start, end, size, align, nid);
>>  }
>>  
>>  /**
> 
> 
> 


-- 
Thanks.
Zhang Yanfei
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/