RE: [PATCH v3] crypto: add virtio-crypto driver

[Gonglei (Arei)]

Hi Halil,

> 
> On 11/28/2016 06:19 PM, Michael S. Tsirkin wrote:
> >>> +static int virtio_crypto_alg_ablkcipher_init_session(
> >>> > > +		struct virtio_crypto_ablkcipher_ctx *ctx,
> >>> > > +		uint32_t alg, const uint8_t *key,
> >>> > > +		unsigned int keylen,
> >>> > > +		int encrypt)
> >>> > > +{
> >>> > > +	struct scatterlist outhdr, key_sg, inhdr, *sgs[3];
> >>> > > +	unsigned int tmp;
> >>> > > +	struct virtio_crypto *vcrypto = ctx->vcrypto;
> >>> > > +	int op = encrypt ? VIRTIO_CRYPTO_OP_ENCRYPT :
> VIRTIO_CRYPTO_OP_DECRYPT;
> >>> > > +	int err;
> >>> > > +	unsigned int num_out = 0, num_in = 0;
> >>> > > +
> >>> > > +	/*
> >>> > > +	 * Avoid to do DMA from the stack, switch to using
> >>> > > +	 * dynamically-allocated for the key
> >>> > > +	 */
> >>> > > +	uint8_t *cipher_key = kmalloc(keylen, GFP_ATOMIC);
> >>> > > +
> >>> > > +	if (!cipher_key)
> >>> > > +		return -ENOMEM;
> >>> > > +
> >>> > > +	memcpy(cipher_key, key, keylen);
> >>> > > +
> >>> > > +	spin_lock(&vcrypto->ctrl_lock);
> >>> > > +	/* Pad ctrl header */
> >>> > > +	vcrypto->ctrl.header.opcode =
> >>> > > +		cpu_to_le32(VIRTIO_CRYPTO_CIPHER_CREATE_SESSION);
> >>> > > +	vcrypto->ctrl.header.algo = cpu_to_le32(alg);
> >>> > > +	/* Set the default dataqueue id to 0 */
> >>> > > +	vcrypto->ctrl.header.queue_id = 0;
> >>> > > +
> >>> > > +	vcrypto->input.status = cpu_to_le32(VIRTIO_CRYPTO_ERR);
> >>> > > +	/* Pad cipher's parameters */
> >>> > > +	vcrypto->ctrl.u.sym_create_session.op_type =
> >>> > > +		cpu_to_le32(VIRTIO_CRYPTO_SYM_OP_CIPHER);
> >>> > > +	vcrypto->ctrl.u.sym_create_session.u.cipher.para.algo =
> >>> > > +		vcrypto->ctrl.header.algo;
> >>> > > +	vcrypto->ctrl.u.sym_create_session.u.cipher.para.keylen =
> >>> > > +		cpu_to_le32(keylen);
> >>> > > +	vcrypto->ctrl.u.sym_create_session.u.cipher.para.op =
> >>> > > +		cpu_to_le32(op);
> >>> > > +
> >>> > > +	sg_init_one(&outhdr, &vcrypto->ctrl, sizeof(vcrypto->ctrl));
> >>> > > +	sgs[num_out++] = &outhdr;
> >>> > > +
> >>> > > +	/* Set key */
> >>> > > +	sg_init_one(&key_sg, cipher_key, keylen);
> >>> > > +	sgs[num_out++] = &key_sg;
> >>> > > +
> >>> > > +	/* Return status and session id back */
> >>> > > +	sg_init_one(&inhdr, &vcrypto->input, sizeof(vcrypto->input));
> >>> > > +	sgs[num_out + num_in++] = &inhdr;
> >>> > > +
> >>> > > +	err = virtqueue_add_sgs(vcrypto->ctrl_vq, sgs, num_out,
> >>> > > +				num_in, vcrypto, GFP_ATOMIC);
> >>> > > +	if (err < 0) {
> >>> > > +		spin_unlock(&vcrypto->ctrl_lock);
> >>> > > +		kfree(cipher_key);
> >>> > > +		return err;
> >>> > > +	}
> >>> > > +	virtqueue_kick(vcrypto->ctrl_vq);
> >>> > > +
> >>> > > +	/*
> >>> > > +	 * Spin for a response, the kick causes an ioport write, trapping
> >>> > > +	 * into the hypervisor, so the request should be handled immediately.
> >>> > > +	 */
> 
> I have my doubts about this comment (and about the code below too). Is
> 'kick causes an ioport write' true for every transport/architecture?
> If we relay on this property maybe the documentation of notify should
> mention it.
> 
Actually it isn't true for every transport, see the call trace:
 /**
 * virtqueue_kick - update after add_buf
 * @vq: the struct virtqueue
 *
 * After one or more virtqueue_add_* calls, invoke this to kick
 * the other side.
 *
 * Caller must ensure we don't call this with other virtqueue
 * operations at the same time (except where noted).
 *
 * Returns false if kick failed, otherwise true.
 */
bool virtqueue_kick(struct virtqueue *vq)
{
	if (virtqueue_kick_prepare(vq))
		return virtqueue_notify(vq);
	return true;
}

If virtqueue_kick_prepare return ture, then notify which causes an ioport write.

Let me remove the comments avoid confusions. 
 	
> I know we have the same message in virtio-net.
> 
Yes, I just migrated it from virtio-net. ;)

> >>> > > +	while (!virtqueue_get_buf(vcrypto->ctrl_vq, &tmp) &&
> >>> > > +	       !virtqueue_is_broken(vcrypto->ctrl_vq))
> >>> > > +		cpu_relax();
> > this spin under lock is kind of ugly.
> > Why do we need to hold it while spinning?
> > to prevent submitting more than one request?
> > Isn't there a way to control this within crypto core?
> >
> > unlock
> > relax
> > lock
> >
> > would be better.
> >
> >>> > > +
> >>> > > +	if (le32_to_cpu(vcrypto->input.status) != VIRTIO_CRYPTO_OK) {
> >>> > > +		spin_unlock(&vcrypto->ctrl_lock);
> >>> > > +		pr_err("virtio_crypto: Create session failed status: %u\n",
> >>> > > +			le32_to_cpu(vcrypto->input.status));
> >>> > > +		kfree(cipher_key);
> >>> > > +		return -EINVAL;
> >>> > > +	}
> >>> > > +	spin_unlock(&vcrypto->ctrl_lock);
> >>> > > +
> > You drop lock here. If someone is trying to submit multiple
> > requests, then the below will be racy as it might overwrite
> > new result with previous data.
> >
> 
> Was going to object on this too but Michael was faster.
> 
Will fix.

Thanks,
-Gonglei

> Halil
> 
> >>> > > +	spin_lock(&ctx->lock);
> >>> > > +	if (encrypt)
> >>> > > +		ctx->enc_sess_info.session_id =
> >>> > > +			le64_to_cpu(vcrypto->input.session_id);
> >>> > > +	else
> >>> > > +		ctx->dec_sess_info.session_id =
> >>> > > +			le64_to_cpu(vcrypto->input.session_id);
> >>> > > +	spin_unlock(&ctx->lock);
> >>> > > +
> >>> > > +	kfree(cipher_key);
> >>> > > +	return 0;
> >>> > > +}
> >>> > > +