[dpdk-dev] [PATCH 4/6] mem: use address hint for mapping hugepages

Alejandro Lucero alejandro.lucero at netronome.com
Tue Jul 3 12:44:36 CEST 2018


On Tue, Jul 3, 2018 at 10:17 AM, Burakov, Anatoly <anatoly.burakov at intel.com
> wrote:

> On 02-Jul-18 6:27 PM, Alejandro Lucero wrote:
>
>> Linux kernel uses a really high address as starting address for
>> serving mmaps calls. If there exists addressing limitations and
>> IOVA mode is VA, this starting address is likely too high for
>> those devices. However, it is possible to use a lower address in
>> the process virtual address space as with 64 bits there is a lot
>> of available space.
>>
>> This patch adds an address hint as starting address for 64 bits
>> systems.
>>
>> Signed-off-by: Alejandro Lucero <alejandro.lucero at netronome.com>
>> ---
>>
>
> <snip>
>
>
>         long aligned_addr;
>>   -     if (internal_config.base_virtaddr != 0) {
>> -               addr = (void*) (uintptr_t) (internal_config.base_virtaddr
>> +
>> -                               baseaddr_offset);
>> -       }
>> -       else addr = NULL;
>> -
>>         RTE_LOG(DEBUG, EAL, "Ask a virtual area of 0x%zx bytes\n", *size);
>>         fd = open("/dev/zero", O_RDONLY);
>> @@ -278,7 +289,22 @@
>>                 return NULL;
>>         }
>>         do {
>> -               addr = mmap(addr,
>> +               if (internal_config.base_virtaddr != 0) {
>> +                       addr_hint = (void *) (uintptr_t)
>> +                                   (internal_config.base_virtaddr +
>> +                                    baseaddr_offset);
>> +               }
>> +#ifdef RTE_ARCH_64
>> +               else {
>> +                       addr_hint = (void *) (uintptr_t) (baseaddr +
>> +                                       baseaddr_offset);
>> +               }
>> +#else
>> +               else {
>> +                       addr_hint = NULL;
>> +               }
>> +#endif
>>
>
> If my understanding is correct, calculations are all done on static
> variables, only the result is assigned to addr_hint which is local. Can we
> move this into a function and put these #ifdefs there, while keeping this
> code clean?


Ok.


>
>
> +               addr = mmap(addr_hint,
>>                                 (*size) + hugepage_sz, PROT_READ,
>>   #ifdef RTE_ARCH_PPC_64
>>                                 MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB,
>> @@ -286,8 +312,14 @@
>>                                 MAP_PRIVATE,
>>   #endif
>>                                 fd, 0);
>> -               if (addr == MAP_FAILED)
>> +               if (addr == MAP_FAILED) {
>> +                       /* map failed. Let's try with less memory */
>>                         *size -= hugepage_sz;
>> +               } else if (addr_hint && addr != addr_hint) {
>> +                       /* map not using hint. Let's try with another
>> offset */
>>
>
> Comment is slightly misleading - "map not using hint" implies we are about
> to map something without using hint. Suggested rewording:
>
> suggested address hint was not used, try with another offset


What about "hint was not used. Try with another offset" ?

By the way, I forgot to unmap the memory in this case. I will fix this in
next version.


>
>
> +                       addr = MAP_FAILED;
>> +                       baseaddr_offset += 0x100000000;
>> +               }
>>         } while (addr == MAP_FAILED && *size > 0);
>>         if (addr == MAP_FAILED) {
>>
>>
>
> --
> Thanks,
> Anatoly
>


More information about the dev mailing list