RE: [PATCH v14] platform/mellanox: Add TmFifo driver for Mellanox BlueField Soc

[Liming Sun]

Thanks Andy! I'll try to post v15 to address these comments this weekend.
(Please also see responses to each comments below).

> -----Original Message-----
> From: Andy Shevchenko <andy.shevchenko@xxxxxxxxx>
> Sent: Thursday, April 11, 2019 10:10 AM
> To: Liming Sun <lsun@xxxxxxxxxxxx>
> Cc: David Woods <dwoods@xxxxxxxxxxxx>; Andy Shevchenko <andy@xxxxxxxxxxxxx>; Darren Hart <dvhart@xxxxxxxxxxxxx>; Vadim
> Pasternak <vadimp@xxxxxxxxxxxx>; Linux Kernel Mailing List <linux-kernel@xxxxxxxxxxxxxxx>; Platform Driver <platform-driver-
> x86@xxxxxxxxxxxxxxx>
> Subject: Re: [PATCH v14] platform/mellanox: Add TmFifo driver for Mellanox BlueField Soc
> 
> On Sun, Apr 7, 2019 at 5:03 AM Liming Sun <lsun@xxxxxxxxxxxx> wrote:
> >
> > This commit adds the TmFifo platform driver for Mellanox BlueField
> > Soc. TmFifo is a shared FIFO which enables external host machine
> > to exchange data with the SoC via USB or PCIe. The driver is based
> > on virtio framework and has console and network access enabled.
> 
> Thanks for an update, my comments below.

Thanks for the comments!

> 
> > +/**
> > + * mlxbf_tmfifo_vdev - Structure of the TmFifo virtual device
> > + * @vdev: virtio device, in which the vdev.id.device field has the
> > + *        VIRTIO_ID_xxx id to distinguish the virtual device.
> > + * @status: status of the device
> > + * @features: supported features of the device
> > + * @vrings: array of tmfifo vrings of this device
> > + * @config.cons: virtual console config -
> > + *               select if vdev.id.device is VIRTIO_ID_CONSOLE
> > + * @config.net: virtual network config -
> > + *              select if vdev.id.device is VIRTIO_ID_NET
> > + * @tx_buf: tx buffer used to buffer data before writing into the FIFO
> > + */
> > +struct mlxbf_tmfifo_vdev {
> > +       struct virtio_device vdev;
> > +       u8 status;
> > +       u64 features;
> > +       struct mlxbf_tmfifo_vring vrings[MLXBF_TMFIFO_VRING_MAX];
> > +       union {
> > +               struct virtio_console_config cons;
> > +               struct virtio_net_config net;
> > +       } config;
> > +       struct circ_buf tx_buf;
> > +};
> 
> (1)
> 
> > +/**
> > + * mlxbf_tmfifo_msg_hdr - Structure of the TmFifo message header
> > + * @type: message type
> > + * @len: payload length
> > + * @data: 64-bit data used to write the message header into the TmFifo register.
> > + *
> > + * This message header is a union of struct and u64 data. The 'struct' has
> > + * type and length field which are used to encode & decode the message. The
> > + * 'data' field is used to read/write the message header from/to the FIFO.
> > + */
> > +union mlxbf_tmfifo_msg_hdr {
> > +       struct {
> > +               u8 type;
> > +               __be16 len;
> > +               u8 unused[5];
> > +       } __packed;
> > +       u64 data;
> > +};
> 
> This union misses a type. See, for example, above structure (1) where
> union is used correctly.

This union seems causing confusion. I'll try to remove the union in v15 
and "construct this on-the-fly" just like you mentioned in another email. 
So instead of " writeq(hdr.data, ...)" we could simply do 
"writeq(*(u64 *)&hdr, ...)", thus no need for a union.

> 
> > +/* Allocate vrings for the FIFO. */
> > +static int mlxbf_tmfifo_alloc_vrings(struct mlxbf_tmfifo *fifo,
> > +                                    struct mlxbf_tmfifo_vdev *tm_vdev)
> > +{
> > +       struct mlxbf_tmfifo_vring *vring;
> > +       struct device *dev;
> > +       dma_addr_t dma;
> > +       int i, size;
> > +       void *va;
> > +
> > +       for (i = 0; i < ARRAY_SIZE(tm_vdev->vrings); i++) {
> > +               vring = &tm_vdev->vrings[i];
> > +               vring->fifo = fifo;
> > +               vring->num = MLXBF_TMFIFO_VRING_SIZE;
> > +               vring->align = SMP_CACHE_BYTES;
> > +               vring->index = i;
> > +               vring->vdev_id = tm_vdev->vdev.id.device;
> > +               dev = &tm_vdev->vdev.dev;
> > +
> > +               size = vring_size(vring->num, vring->align);
> > +               va = dma_alloc_coherent(dev->parent, size, &dma, GFP_KERNEL);
> > +               if (!va) {
> 
> > +                       dev_err(dev->parent, "dma_alloc_coherent failed\n");
> > +                       mlxbf_tmfifo_free_vrings(fifo, tm_vdev);
> 
> First do things, then report about what has been done.

Will update it in v15.

> 
> > +                       return -ENOMEM;
> > +               }
> > +
> > +               vring->va = va;
> > +               vring->dma = dma;
> > +       }
> > +
> > +       return 0;
> > +}
> 
> > +/* Interrupt handler. */
> > +static irqreturn_t mlxbf_tmfifo_irq_handler(int irq, void *arg)
> > +{
> > +       struct mlxbf_tmfifo_irq_info *irq_info = arg;
> > +
> 
> > +       if (irq_info->index < MLXBF_TM_MAX_IRQ &&
> 
> On which circumstances this is possible?

Yes, Not needed at all. Will update it in v15.

> 
> > +           !test_and_set_bit(irq_info->index, &irq_info->fifo->pend_events))
> > +               schedule_work(&irq_info->fifo->work);
> > +
> > +       return IRQ_HANDLED;
> > +}
> 
> > +static void mlxbf_tmfifo_release_pending_pkt(struct mlxbf_tmfifo_vring *vring)
> > +{
> > +       struct vring_desc *desc_head;
> > +       u32 len = 0;
> > +
> > +       if (vring->desc_head) {
> > +               desc_head = vring->desc_head;
> > +               len = vring->pkt_len;
> > +       } else {
> > +               desc_head = mlxbf_tmfifo_get_next_desc(vring);
> 
> > +               if (desc_head)
> 
> Redundant...

Will update it in v15.

> 
> > +                       len = mlxbf_tmfifo_get_pkt_len(vring, desc_head);
> 
> ...this is NULL-aware AFAICS.

Yes, it is.

> 
> > +       }
> > +
> > +       if (desc_head)
> > +               mlxbf_tmfifo_release_desc(vring, desc_head, len);
> > +
> > +       vring->pkt_len = 0;
> > +       vring->desc = NULL;
> > +       vring->desc_head = NULL;
> > +}
> 
> > +/* The notify function is called when new buffers are posted. */
> > +static bool mlxbf_tmfifo_virtio_notify(struct virtqueue *vq)
> > +{
> > +       struct mlxbf_tmfifo_vring *vring = vq->priv;
> > +       struct mlxbf_tmfifo_vdev *tm_vdev;
> > +       struct mlxbf_tmfifo *fifo;
> > +       unsigned long flags;
> > +
> > +       fifo = vring->fifo;
> > +
> > +       /*
> > +        * Virtio maintains vrings in pairs, even number ring for Rx
> > +        * and odd number ring for Tx.
> > +        */
> 
> > +       if (!(vring->index & BIT(0))) {
> 
> Perhaps positive conditional is better.

Will update it in v15.

> 
> > +               if (test_and_set_bit(MLXBF_TM_RX_HWM_IRQ, &fifo->pend_events))
> > +                       return true;
> > +       } else {
> > +               /*
> > +                * Console could make blocking call with interrupts disabled.
> > +                * In such case, the vring needs to be served right away. For
> > +                * other cases, just set the TX LWM bit to start Tx in the
> > +                * worker handler.
> > +                */
> > +               if (vring->vdev_id == VIRTIO_ID_CONSOLE) {
> > +                       spin_lock_irqsave(&fifo->spin_lock, flags);
> > +                       tm_vdev = fifo->vdev[VIRTIO_ID_CONSOLE];
> > +                       mlxbf_tmfifo_console_output(tm_vdev, vring);
> > +                       spin_unlock_irqrestore(&fifo->spin_lock, flags);
> > +               } else if (test_and_set_bit(MLXBF_TM_TX_LWM_IRQ,
> > +                                           &fifo->pend_events)) {
> > +                       return true;
> > +               }
> > +       }
> > +
> > +       schedule_work(&fifo->work);
> > +
> > +       return true;
> > +}
> 
> > +/* Read the value of a configuration field. */
> > +static void mlxbf_tmfifo_virtio_get(struct virtio_device *vdev,
> > +                                   unsigned int offset,
> > +                                   void *buf,
> > +                                   unsigned int len)
> > +{
> > +       struct mlxbf_tmfifo_vdev *tm_vdev = mlxbf_vdev_to_tmfifo(vdev);
> > +
> 
> > +       if (offset + len > sizeof(tm_vdev->config))
> > +               return;
> 
> This doesn't protect against too big len and offset.
> Same for other similar checks.

Will revise it in v15 like "if ((u64)offset + len > sizeof(tm_vdev->config))"

> 
> > +
> > +       memcpy(buf, (u8 *)&tm_vdev->config + offset, len);
> > +}
> 
> > +/* Read the configured network MAC address from efi variable. */
> > +static void mlxbf_tmfifo_get_cfg_mac(u8 *mac)
> > +{
> > +       efi_guid_t guid = EFI_GLOBAL_VARIABLE_GUID;
> > +       unsigned long size = ETH_ALEN;
> > +       efi_status_t status;
> > +       u8 buf[ETH_ALEN];
> > +
> 
> > +       status = efi.get_variable(mlxbf_tmfifo_efi_name, &guid, NULL, &size,
> > +                                 buf);
> 
> Use one line.

Will update it in v15.

> 
> > +       if (status == EFI_SUCCESS && size == ETH_ALEN)
> > +               ether_addr_copy(mac, buf);
> > +       else
> > +               ether_addr_copy(mac, mlxbf_tmfifo_net_default_mac);
> > +}
> 
> > +/* Probe the TMFIFO. */
> > +static int mlxbf_tmfifo_probe(struct platform_device *pdev)
> > +{
> 
> > +       res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > +       fifo->rx_base = devm_ioremap_resource(dev, res);
> 
> There is new helper devm_platform_ioremap_resource().
> Please, use it instead.

Will update it in v15.

> 
> > +       if (IS_ERR(fifo->rx_base))
> > +               return PTR_ERR(fifo->rx_base);
> 
> > +       res = platform_get_resource(pdev, IORESOURCE_MEM, 1);
> > +       fifo->tx_base = devm_ioremap_resource(dev, res);
> 
> Ditto.

Will update it in v15.

> 
> > +       if (IS_ERR(fifo->tx_base))
> > +               return PTR_ERR(fifo->tx_base);
> 
> > +}
> 
> --
> With Best Regards,
> Andy Shevchenko