On 10.08.2021 13:47:32, Matt Kline wrote: > On Wed, Aug 04, 2021 at 11:18:58AM +0200, Marc Kleine-Budde wrote: > > > > > > - cdev->ops->write_fifo(cdev, addr_offset, val); > > > + result = cdev->ops->write_fifo(cdev, addr_offset, val, val_count); > > > + WARN_ON(result != 0); > > > > What about converting all read/write functions to return an error, and > > handle the error in the caller? > > Yeah, that would be cleaner. In the mcp251xfd (another SPI-CAN controller) driver I have the same problem. I've basically implemented error checking everywhere. If there is an error in the interrupt handler, I shut down the driver, see: https://elixir.bootlin.com/linux/v5.13/source/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c#L2298 > > > /* acknowledge rx fifo 0 */ > > > @@ -1546,8 +1548,8 @@ static netdev_tx_t m_can_tx_handler(struct m_can_classdev *cdev) > > > struct net_device *dev = cdev->net; > > > struct sk_buff *skb = cdev->tx_skb; > > > u32 id, cccr, fdflags; > > > - int i; > > > int putidx; > > > + u32 id_and_dlc[2]; > > > > Can you create a struct for this? > > Ditto, sure! A struct can easily extended to hold the data, too. > > > > > > cdev->tx_skb = NULL; > > > > > > @@ -1563,18 +1565,16 @@ static netdev_tx_t m_can_tx_handler(struct m_can_classdev *cdev) > > > if (cf->can_id & CAN_RTR_FLAG) > > > id |= TX_BUF_RTR; > > > > > > + id_and_dlc[0] = id; > > > + > > > if (cdev->version == 30) { > > > netif_stop_queue(dev); > > > > > > - /* message ram configuration */ > > > - m_can_fifo_write(cdev, 0, M_CAN_FIFO_ID, id); > > > - m_can_fifo_write(cdev, 0, M_CAN_FIFO_DLC, > > > - can_fd_len2dlc(cf->len) << 16); > > > + id_and_dlc[1] = can_fd_len2dlc(cf->len) << 16; > > > > > > - for (i = 0; i < cf->len; i += 4) > > > - m_can_fifo_write(cdev, 0, > > > - M_CAN_FIFO_DATA(i / 4), > > > - *(u32 *)(cf->data + i)); > > > + /* Write the frame ID, DLC, and payload to the FIFO element. */ > > > + m_can_fifo_write(cdev, 0, M_CAN_FIFO_ID, id_and_dlc, ARRAY_SIZE(id_and_dlc)); > > > + m_can_fifo_write(cdev, 0, M_CAN_FIFO_DATA, cf->data, DIV_ROUND_UP(cf->len, 4)); > > > > Does it make sense to combine these, too? Same for the v3.1 variant. > > I think that's the eventual goal, but since the ID, DLC, and frame data would > have to be contiguous for a single m_can_fifo_write(), you'd end up copying > things around. Yes, but at least for the SPI this is a neglectable overhead. > I wanted to start with this smaller, simpler patch first. Is that > alright? Fine with me! > I'll try to send a v3 up shortly. regards, Marc -- Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
Attachment:
signature.asc
Description: PGP signature
