On 7/20/2020 6:20 PM, Tejasree Kondoj wrote:
> Hi Ferruh,
>
> Please see inline.
>
> Thanks
> Tejasree
>
>> -----Original Message-----
>> From: Ferruh Yigit <ferruh.yi...@intel.com>
>> Sent: Monday, July 20, 2020 5:33 PM
>> To: Tejasree Kondoj <ktejas...@marvell.com>; Akhil Goyal
>> <akhil.go...@nxp.com>; Radu Nicolau <radu.nico...@intel.com>
>> Cc: Vamsi Krishna Attunuru <vattun...@marvell.com>; Narayana Prasad
>> Raju Athreya <pathr...@marvell.com>; Anoob Joseph
>> <ano...@marvell.com>; dev@dpdk.org; Thomas Monjalon
>> <tho...@monjalon.net>
>> Subject: [EXT] Re: [dpdk-dev] [PATCH v3 8/8] crypto/octeontx2: add
>> cryptodev sec enqueue and dequeue routines
>>
>> External Email
>>
>> ----------------------------------------------------------------------
>> On 7/16/2020 9:39 AM, Tejasree Kondoj wrote:
>>> From: Vamsi Attunuru <vattun...@marvell.com>
>>>
>>> This patch adds lookaside IPsec enqueue and dequeue routines.
>>>
>>> Signed-off-by: Vamsi Attunuru <vattun...@marvell.com>
>>> Signed-off-by: Tejasree Kondoj <ktejas...@marvell.com>
>>
>> <...>
>>
>>> @@ -33,6 +34,13 @@ struct otx2_cpt_vf {
>>> /**< CPT device capabilities */
>>> };
>>>
>>> +struct cpt_meta_info {
>>> + uint64_t deq_op_info[4];
>>> + uint64_t comp_code_sz;
>>> + union cpt_res_s cpt_res __rte_aligned(16);
>>> + struct cpt_request_info cpt_req __rte_aligned(8); };
>>
>> "struct cpt_request_info" already has cache size alignment
>> (__rte_cache_aligned) [1].
>>
>> Trying to align it to 8 bytes is reducing the alignment, gcc & clang seems
>> silently ignoring this case but icc is giving a warning.
>>
>> Since it is already ignored, and I assume the real intention is not to reduce
>> the alignment, can you please remove the above "__rte_aligned(8)"?
> [Tejasree] 'struct cpt_request_info' need not be cache aligned. It will need
> to be aligned
> to 8 bytes as used in this patch. Will replace __rte_cache_aligned of 'struct
> cpt_request_info'
> with __rte_aligned(8) and remove above __rte_aligned(8). Will submit separate
> patch
> as a fix and mention commit id of merged patch (fab634eb87ca) in fixes
> section.
> Would that be fine?
That sounds fine, thanks.
>>
>>
>> @Thomas, it is still good to have icc around to catch additional issues ;)
>>
>>
>> [1]
>> https://urldefense.proofpoint.com/v2/url?u=http-
>> 3A__lxr.dpdk.org_dpdk_v20.05_source_drivers_common_cpt_cpt-
>> 5Fcommon.h-
>> 23L79&d=DwIDaQ&c=nKjWec2b6R0mOyPaz7xtfQ&r=NjpqGUAf6Xc0ZLzxCvv4i
>> df8zRFeSJHioNlG1Wif1Gs&m=KNBzM9j6nc00ue4kkISJLVTLpkDGjg4A4dUjIVFz
>> Rps&s=Q0RPHY0v3g5f0uOh9vv7SKb61aQuQB2RhCDLiOtLZ7s&e=
