Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <e306700c-3153-9422-974c-1f5f10e232d6@gmail.com>
Date:   Tue, 1 Mar 2022 14:43:55 +0900
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux aarch64; rv:91.0) Gecko/20100101
 Thunderbird/91.5.0
Subject: Re: [PATCH] virtio-blk: Assign discard_granularity
Content-Language: en-US
To:     Stefan Hajnoczi <stefanha@redhat.com>
Cc:     linux-kernel@vger.kernel.org, linux-block@vger.kernel.org,
        virtualization@lists.linux-foundation.org,
        Jens Axboe <axboe@kernel.dk>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Christoph Hellwig <hch@lst.de>
References: <20220224093802.11348-1-akihiko.odaki@gmail.com>
 <YhypTNtWpcgh3gb2@stefanha-x1.localdomain>
From:   Akihiko Odaki <akihiko.odaki@gmail.com>
In-Reply-To: <YhypTNtWpcgh3gb2@stefanha-x1.localdomain>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2022/02/28 19:51, Stefan Hajnoczi wrote:
> On Thu, Feb 24, 2022 at 06:38:02PM +0900, Akihiko Odaki wrote:
>> Virtual I/O Device (VIRTIO) Version 1.1
>> https://docs.oasis-open.org/virtio/virtio/v1.1/csprd01/virtio-v1.1-csprd01.html
>>> discard_sector_alignment can be used by OS when splitting a request
>>> based on alignment.
>>
>> According to Documentation/ABI/stable/sysfs-block, the corresponding
>> field in the kernel is, confusingly, discard_granularity, not
>> discard_alignment.
> 
> Good catch, struct virtio_blk_config->discard_sector_alignment is Linux
> q->limits.discard_granularity.
> 
>>
>> Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
>> ---
>>   drivers/block/virtio_blk.c | 4 +---
>>   1 file changed, 1 insertion(+), 3 deletions(-)
>>
>> diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
>> index c443cd64fc9b..1fb3c89900e3 100644
>> --- a/drivers/block/virtio_blk.c
>> +++ b/drivers/block/virtio_blk.c
>> @@ -913,11 +913,9 @@ static int virtblk_probe(struct virtio_device *vdev)
>>   		blk_queue_io_opt(q, blk_size * opt_io_size);
>>   
>>   	if (virtio_has_feature(vdev, VIRTIO_BLK_F_DISCARD)) {
>> -		q->limits.discard_granularity = blk_size;
>> -
>>   		virtio_cread(vdev, struct virtio_blk_config,
>>   			     discard_sector_alignment, &v);
>> -		q->limits.discard_alignment = v ? v << SECTOR_SHIFT : 0;
> 
> Should we use struct virtio_blk_config->topology.alignment_offset
> ("offset of first aligned logical block" and used for Linux
> blk_queue_alignment_offset()) for q->limits.discard_alignment?

Maybe but I'm not sure. I had looked at the code of QEMU
(commit 5c1ee569660d4a205dced9cb4d0306b907fb7599) but it apparently 
always sets 0 for virtio_blk_config->topology.alignment_offset.
I don't have a hardware which requires discard_alignment either so I 
cannot test it.

I'd like to leave this patch as is since I cannot deny the possibility 
that the host has a different alignment offset for discarding and other 
operations.