Re: [PATCH v2 1/2] memblock: drop for_each_memblock_type() and open code its users

[tarunsahu@google.com]

Pratyush Yadav <pratyush@kernel.org> writes:

> On Fri, Sep 18 2026, Tarun Sahu wrote:
>
>> for_each_memblock_type() always starts the iteration at index 0 and its
>> body is a trivial 'for' loop, so it hides very little. It also gets in
>> the way of iterating from an arbitrary index, which the next patch
>> needs.
>>
>> Remove the macro and open code its three users: memblock_add_range(),
>> memblock_isolate_range() and memblock_dump(). While at it, move the
>> region pointer into the loop body scope.
>>
>> No functional change.
>>
>> Signed-off-by: Tarun Sahu <tarunsahu@google.com>
>> ---
>>  mm/memblock.c | 14 ++++++--------
>>  1 file changed, 6 insertions(+), 8 deletions(-)
>>
>> diff --git a/mm/memblock.c b/mm/memblock.c
>> index 9ce86349a29f..5aae75d34dec 100644
>> --- a/mm/memblock.c
>> +++ b/mm/memblock.c
>> @@ -155,11 +155,6 @@ struct memblock_type physmem = {
>>   */
>>  static __refdata struct memblock_type *memblock_memory = &memblock.memory;
>>  
>> -#define for_each_memblock_type(i, memblock_type, rgn)			\
>> -	for (i = 0, rgn = &memblock_type->regions[0];			\
>> -	     i < memblock_type->cnt;					\
>> -	     i++, rgn = &memblock_type->regions[i])
>> -
>>  #define memblock_dbg(fmt, ...)						\
>>  	do {								\
>>  		if (memblock_debug)					\
>> @@ -651,7 +646,8 @@ static int __init_memblock memblock_add_range(struct memblock_type *type,
>>  	base = obase;
>>  	nr_new = 0;
>>  
>> -	for_each_memblock_type(idx, type, rgn) {
>> +	for (idx = 0; idx < type->cnt; idx++) {
>> +		rgn = &type->regions[idx];
>
> Nit: this is really odd to read. You assign rgn and then declare rbase
> and rend. Doesn't checkpatch complain?

I kept it there to minimize the changes in the patch as the variable was
already defined there. and unfortunately, checkpatch did not throw any error.

Regrardless, Good idea. I updated in v3 here:
https://lore.kernel.org/all/20260926092448.4090401-1-tarunsahu@google.com/

>
> Regardless, I reckon it would be a lot nicer to read of you move the
> declaration of rgn into the loop. Same for the others below.
>
> LGTM otherwise, so after fixing this up, feel free to add:
>
> Reviewed-by: Pratyush Yadav <pratyush@kernel.org>

Done! Thanks for reviewing it.

>
>>  		phys_addr_t rbase = rgn->base;
>>  		phys_addr_t rend = rbase + rgn->size;
>>  
>> @@ -827,7 +823,8 @@ static int __init_memblock memblock_isolate_range(struct memblock_type *type,
>>  		if (memblock_double_array(type, base, size) < 0)
>>  			return -ENOMEM;
>>  
>> -	for_each_memblock_type(idx, type, rgn) {
>> +	for (idx = 0; idx < type->cnt; idx++) {
>> +		rgn = &type->regions[idx];
>>  		phys_addr_t rbase = rgn->base;
>>  		phys_addr_t rend = rbase + rgn->size;
>>  
>> @@ -2198,7 +2195,8 @@ static void __init_memblock memblock_dump(struct memblock_type *type)
>>  
>>  	pr_info(" %s.cnt  = 0x%lx\n", type->name, type->cnt);
>>  
>> -	for_each_memblock_type(idx, type, rgn) {
>> +	for (idx = 0; idx < type->cnt; idx++) {
>> +		rgn = &type->regions[idx];
>>  		char nid_buf[32] = "";
>>  
>>  		base = rgn->base;
>
> -- 
> Regards,
> Pratyush Yadav