Re: [PATCH 3/3] virtio-blk: Add bio-based IO path for virtio-blk

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


On Mon, Jun 18, 2012 at 11:21 AM, Michael S. Tsirkin <mst@xxxxxxxxxx> wrote:
> On Mon, Jun 18, 2012 at 02:53:10PM +0800, Asias He wrote:
>> +static void virtblk_make_request(struct request_queue *q, struct bio *bio)
>> +{
>> +     struct virtio_blk *vblk = q->queuedata;
>> +     unsigned int num, out = 0, in = 0;
>> +     struct virtblk_req *vbr;
>> +
>> +     BUG_ON(bio->bi_phys_segments + 2 > vblk->sg_elems);
>> +     BUG_ON(bio->bi_rw & (REQ_FLUSH | REQ_FUA));
>> +
>> +     vbr = virtblk_alloc_req(vblk, GFP_NOIO);
>> +     if (!vbr) {
>> +             bio_endio(bio, -ENOMEM);
>> +             return;
>> +     }
>> +
>> +     vbr->bio = bio;
>> +     vbr->req = NULL;
>> +     vbr->out_hdr.type = 0;
>> +     vbr->out_hdr.sector = bio->bi_sector;
>> +     vbr->out_hdr.ioprio = bio_prio(bio);
>> +
>> +     sg_set_buf(&vbr->sg[out++], &vbr->out_hdr, sizeof(vbr->out_hdr));
>> +
>> +     num = blk_bio_map_sg(q, bio, vbr->sg + out);
>> +
>> +     sg_set_buf(&vbr->sg[num + out + in++], &vbr->status,
>> +                sizeof(vbr->status));
>> +
>> +     if (num) {
>> +             if (bio->bi_rw & REQ_WRITE) {
>> +                     vbr->out_hdr.type |= VIRTIO_BLK_T_OUT;
>> +                     out += num;
>> +             } else {
>> +                     vbr->out_hdr.type |= VIRTIO_BLK_T_IN;
>> +                     in += num;
>> +             }
>> +     }
>> +
>> +     spin_lock_irq(vblk->disk->queue->queue_lock);
>> +     if (virtqueue_add_buf(vblk->vq, vbr->sg, out, in, vbr,
>> +                           GFP_ATOMIC) < 0) {
>> +             spin_unlock_irq(vblk->disk->queue->queue_lock);
>
> Any implications of dropping lock like that?
> E.g. for suspend. like we are still discussing with
> unlocked kick?

Since we aquired the lock in this function there should be no problem.
 Whatever protects against vblk or vblk->disk disappearing upon
entering this function also protects after unlocking queue_lock.
Otherwise all .make_request_fn() functions would be broken.

I'd still like to understand the details though.

Stefan
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


[KVM ARM]     [KVM ia64]     [KVM ppc]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Photo]     [Yosemite News]    [Yosemite Photos]    [Linux Kernel]     [Linux SCSI]     [XFree86]

Add to Google Powered by Linux