On 2018/1/24 21:21, Leon Romanovsky wrote:
> On Wed, Jan 24, 2018 at 08:40:42PM +0800, Liuyixian (Eason) wrote:
>>
>>
>> On 2018/1/23 19:07, Leon Romanovsky wrote:
>>> On Thu, Jan 18, 2018 at 10:34:30AM +0800, Yixian Liu wrote:
>>>> This patch updates to support rq record doorbell for
>>>> the kernel space.
>>>>
>>>> Signed-off-by: Yixian Liu <liuyixian@xxxxxxxxxx>
>>>> Signed-off-by: Lijun Ou <oulijun@xxxxxxxxxx>
>>>> Signed-off-by: Wei Hu (Xavier) <xavier.huwei@xxxxxxxxxx>
>>>> Signed-off-by: Shaobo Xu <xushaobo2@xxxxxxxxxx>
>>>> ---
>>>> drivers/infiniband/hw/hns/hns_roce_db.c | 112 ++++++++++++++++++++++++++++
>>>> drivers/infiniband/hw/hns/hns_roce_device.h | 6 ++
>>>> drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 15 +---
>>>> drivers/infiniband/hw/hns/hns_roce_main.c | 5 ++
>>>> drivers/infiniband/hw/hns/hns_roce_qp.c | 18 ++++-
>>>> 5 files changed, 143 insertions(+), 13 deletions(-)
>>>>
>>>> diff --git a/drivers/infiniband/hw/hns/hns_roce_db.c b/drivers/infiniband/hw/hns/hns_roce_db.c
>>>> index 0908208..de3fa94 100644
>>>> --- a/drivers/infiniband/hw/hns/hns_roce_db.c
>>>> +++ b/drivers/infiniband/hw/hns/hns_roce_db.c
>>>> @@ -92,3 +92,115 @@ void hns_roce_db_unmap_user(struct hns_roce_ucontext *context,
>>>> mutex_unlock(&context->db_page_mutex);
>>>> }
>>>> EXPORT_SYMBOL(hns_roce_db_unmap_user);
>>>> +
>>>> +static struct hns_roce_db_pgdir *hns_roce_alloc_db_pgdir(
>>>> + struct device *dma_device, gfp_t gfp)
>>>> +{
>>>> + struct hns_roce_db_pgdir *pgdir;
>>>> +
>>>> + pgdir = kzalloc(sizeof(*pgdir), gfp);
>>>> + if (!pgdir)
>>>> + return NULL;
>>>> +
>>>> + bitmap_fill(pgdir->order1, HNS_ROCE_DB_PER_PAGE / 2);
>>>> + pgdir->bits[0] = pgdir->order0;
>>>
>>> order0? isn't it equal to zero?
>>>
>>
>> No. Here order0 is the name of a bitmap array and bits[0] points to it.
>
> pgdir was allocated with kzalloc() -> all fields are zero ->
> pgdir->order0 == 0 or pgdir->order0 == NULL -> What is the point in such
> assignment?
The definition of hns_roce_db_pgdir is in patch [1/4] as following:
struct hns_roce_db_pgdir {
struct list_head list;
DECLARE_BITMAP(order0, HNS_ROCE_DB_PER_PAGE);
DECLARE_BITMAP(order1, HNS_ROCE_DB_PER_PAGE / 2);
unsigned long *bits[2];
u32 *db_page;
dma_addr_t db_dma;
};
and,
#define DECLARE_BITMAP(name,bits) \
unsigned long name[BITS_TO_LONGS(bits)]
Therefore, pgdir->order0 is the base address of an array, not NULL.
Thanks.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
