Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH 2/3] lightnvm: encapsule rqd dma allocations
To:     javier@cnexlabs.com
Cc:     axboe@kernel.dk, linux-block@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <1535532980-27672-1-git-send-email-javier@cnexlabs.com>
 <1535532980-27672-3-git-send-email-javier@cnexlabs.com>
 <559e39f2-1601-95ea-9305-6a0566943798@lightnvm.io>
 <8B4755CD-3C36-49B4-9F94-1169E19EF090@cnexlabs.com>
 <e5d4678d-814d-e2a5-da3e-66245034743f@lightnvm.io>
 <1CF9AE13-2C0C-426C-8B62-E0B111EDEB5E@cnexlabs.com>
From:   =?UTF-8?Q?Matias_Bj=c3=b8rling?= <mb@lightnvm.io>
Message-ID: <c6414a97-607d-2bd8-c575-4c54e5ee84e1@lightnvm.io>
Date:   Wed, 29 Aug 2018 15:42:57 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.5.0
MIME-Version: 1.0
In-Reply-To: <1CF9AE13-2C0C-426C-8B62-E0B111EDEB5E@cnexlabs.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-GB
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 08/29/2018 03:41 PM, Javier Gonzalez wrote:
> 
>> On 29 Aug 2018, at 15.36, Matias Bjørling <mb@lightnvm.io> wrote:
>>
>> On 08/29/2018 03:18 PM, Javier Gonzalez wrote:
>>>> On 29 Aug 2018, at 15.00, Matias Bjørling <mb@lightnvm.io> wrote:
>>>>
>>>> On 08/29/2018 10:56 AM, Javier González wrote:
>>>>> dma allocations for ppa_list and meta_list in rqd are replicated in
>>>>> several places across the pblk codebase. Make helpers to encapsulate
>>>>> creation and deletion to simplify the code.
>>>>> Signed-off-by: Javier González <javier@cnexlabs.com>
>>>>> ---
>>>>>   drivers/lightnvm/pblk-core.c     | 69 +++++++++++++++++++++++++---------------
>>>>>   drivers/lightnvm/pblk-read.c     | 35 ++++++++++----------
>>>>>   drivers/lightnvm/pblk-recovery.c | 29 ++++++-----------
>>>>>   drivers/lightnvm/pblk-write.c    | 15 ++-------
>>>>>   drivers/lightnvm/pblk.h          |  3 ++
>>>>>   5 files changed, 74 insertions(+), 77 deletions(-)
>>>>> diff --git a/drivers/lightnvm/pblk-core.c b/drivers/lightnvm/pblk-core.c
>>>>> index 09160ec02c5f..767178185f19 100644
>>>>> --- a/drivers/lightnvm/pblk-core.c
>>>>> +++ b/drivers/lightnvm/pblk-core.c
>>>>> @@ -237,6 +237,34 @@ static void pblk_invalidate_range(struct pblk *pblk, sector_t slba,
>>>>>   	spin_unlock(&pblk->trans_lock);
>>>>>   }
>>>>>   +int pblk_setup_rqd(struct pblk *pblk, struct nvm_rq *rqd, gfp_t mem_flags,
>>>>> +		   bool is_vector)
>>>>
>>>>
>>>> The mem_flags argument can be removed. It is GFP_KERNEL from all the
>>>> places it is called.
>>> Thought it was better to have the flexibility in a helper function, but
>>> we can always add it later on if needed...
>>>> is_vector, will be better to do nr_ppas (or similar name). Then
>>>> pblk_submit_read/pblk_submit_read_gc are a bit cleaner.
>>> We can do that too, yes.
>>>>> +{
>>>>> +	struct nvm_tgt_dev *dev = pblk->dev;
>>>>> +
>>>>> +	rqd->meta_list = nvm_dev_dma_alloc(dev->parent, mem_flags,
>>>>> +							&rqd->dma_meta_list);
>>>>> +	if (!rqd->meta_list)
>>>>> +		return -ENOMEM;
>>>>> +
>>>>> +	if (!is_vector)
>>>>> +		return 0;
>>>>> +
>>>>> +	rqd->ppa_list = rqd->meta_list + pblk_dma_meta_size;
>>>>> +	rqd->dma_ppa_list = rqd->dma_meta_list + pblk_dma_meta_size;
>>>>
>>>> Wrt to is_vector, does it matter if we just set ppa_list and
>>>> dma_ppa_list? If we have them, we use them, else leave them alone?
>>> If we only have 1 address then ppa_addr is set and the ppa_list attempt
>>> to free in the completion path interpreting ppa_addr as the dma address.
>>> So I don't think so - unless I'm missing something?
>>
>> In that case, we could drop is_vector/nr_ppas all together? That would be nice.
>>
> 
> The problem is that the metadata region still needs to be used, even if
> the ppa_list is not set. Thing that the oob area can be larger than
> 64bits, so we cannot do the dma address is the actual value trick.
> 
> So if encapsulating, we need to know if we need to allocate the ppa_list
> or not.
> 
> Does it make sense?

Cool. We'll just leave it as is.