On 10/28/20 7:24 PM, Philippe Mathieu-Daudé wrote:
> On 10/28/20 4:16 PM, Stefan Hajnoczi wrote:
>> On Tue, Oct 27, 2020 at 02:55:35PM +0100, Philippe Mathieu-Daudé wrote:
>>> Rename Submission Queue flags with 'Sq' and introduce
>>> Completion Queue flag definitions.
>>>
>>> Signed-off-by: Philippe Mathieu-Daudé <phi...@redhat.com>
>>> ---
>>> include/block/nvme.h | 17 +++++++++++------
>>> 1 file changed, 11 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/include/block/nvme.h b/include/block/nvme.h
>>> index 65e68a82c89..079f884a2d3 100644
>>> --- a/include/block/nvme.h
>>> +++ b/include/block/nvme.h
>>> @@ -491,6 +491,11 @@ typedef struct QEMU_PACKED NvmeCreateCq {
>>> #define NVME_CQ_FLAGS_PC(cq_flags) (cq_flags & 0x1)
>>> #define NVME_CQ_FLAGS_IEN(cq_flags) ((cq_flags >> 1) & 0x1)
>>>
>>> +enum NvmeFlagsCq {
>>> + NVME_CQ_PC = 1,
>>> + NVME_CQ_IEN = 2,
>>> +};
>>> +
>>> typedef struct QEMU_PACKED NvmeCreateSq {
>>> uint8_t opcode;
>>> uint8_t flags;
>>> @@ -508,12 +513,12 @@ typedef struct QEMU_PACKED NvmeCreateSq {
>>> #define NVME_SQ_FLAGS_PC(sq_flags) (sq_flags & 0x1)
>>> #define NVME_SQ_FLAGS_QPRIO(sq_flags) ((sq_flags >> 1) & 0x3)
>>>
>>> -enum NvmeQueueFlags {
>>> - NVME_Q_PC = 1,
>>> - NVME_Q_PRIO_URGENT = 0,
>>> - NVME_Q_PRIO_HIGH = 1,
>>> - NVME_Q_PRIO_NORMAL = 2,
>>> - NVME_Q_PRIO_LOW = 3,
>>> +enum NvmeFlagsSq {
>>> + NVME_SQ_PC = 1,
>>> + NVME_SQ_PRIO_URGENT = 0,
>>> + NVME_SQ_PRIO_HIGH = 1,
>>> + NVME_SQ_PRIO_NORMAL = 2,
>>> + NVME_SQ_PRIO_LOW = 3,
>>> };
>>
>> There is also:
>>
>> #define NVME_SQ_FLAGS_PC(sq_flags) (sq_flags & 0x1)
>> #define NVME_SQ_FLAGS_QPRIO(sq_flags) ((sq_flags >> 1) & 0x3)
>>
>> These macros should use the new constants.
SQ_PC is bit#0, NVME_SQ_PC is "bit SQ_PC set (PC enabled)",
SQ_PRIO are bits #1-2 (shift by 1, mask 2 bits),
NVME_SQ_PRIO_xxx is the enum of these 2 bits.
The NVME_SQ_FLAGS_X() macros extract the flags.
So the macros can not use the new constants.
>>
>> I didn't check if there are additional magic numbers in hw/block/nvme.c
>> that should be converted.
>
> FYI we discussed with Klaus and might convert "block/nvme.h" to
> use the registerfields API during the 6.0 dev cycle.
