| Message ID | 20190319051512.14950-3-qiangqing.zhang@nxp.com |
|---|---|
| State | Awaiting Upstream |
| Delegated to: | David Miller |
| Headers | show |
| Series | can: flexcan: add CAN FD support on i.MX8QM | expand |
Hello Joakim, I successfully validated your CAN FD support on an S32V234 based board. Here are some points which are not very clear to me: Joakim Zhang <qiangqing.zhang@nxp.com> writes: > + priv->write(reg_mcr | FLEXCAN_MCR_FDEN, ®s->mcr); According to the S32V234 reference manual (not sure if this applies to i.MX8QM as well), CAN_CTRL1[SMP] cannot be asserted when CAN FD is enabled. Is it safe to configure the interface with both "fd on" and "triple-sampling on"? If not, flexcan_open should return an error when both CAN_CTRLMODE_FD and CAN_CTRLMODE_3_SAMPLES are set in ctrlmode. Joakim Zhang <qiangqing.zhang@nxp.com> writes: > + priv->offload.is_canfd = true; Shouldn't is_canfd be assigned false in the "else" branch? Otherwise, is_canfd will stay true after the following sequence: root@s32v234evb:~# ip link set can0 type can bitrate 500000 dbitrate 2000000 fd on root@s32v234evb:~# ip link set can0 up root@s32v234evb:~# ip link set can0 down root@s32v234evb:~# ip link set can0 type can bitrate 500000 fd off root@s32v234evb:~# ip link set can0 up Moreover, I wonder if we even need is_canfd, because frames should be fed to the kernel as classic or FD based only on the EDL bit in the C/S word of the Rx MB, to keep their original format. That is, you may either read EDL from within a new callback (but I think that mailbox_read should remain the only one which both locks and unlocks the mailboxes) or let mailbox_read call alloc_can(fd)_skb itself. Regards, Stefan
Hi Stefan, Thanks for your validation! Could you add your test tag if you can successfully validated? Comments in-lined about your points. Best Regards, Joakim Zhang > -----Original Message----- > From: Stefan-gabriel Mirea > Sent: 2019年4月9日 1:46 > To: Joakim Zhang <qiangqing.zhang@nxp.com>; linux-can@vger.kernel.org; > netdev@vger.kernel.org; linux-kernel@vger.kernel.org > Subject: Re: [PATCH V2 2/5] can: flexcan: add CAN FD mode support > > Hello Joakim, > > I successfully validated your CAN FD support on an S32V234 based board. > Here are some points which are not very clear to me: > > Joakim Zhang <qiangqing.zhang@nxp.com> writes: > > + priv->write(reg_mcr | FLEXCAN_MCR_FDEN, ®s->mcr); > > According to the S32V234 reference manual (not sure if this applies to > i.MX8QM as well), CAN_CTRL1[SMP] cannot be asserted when CAN FD is > enabled. Is it safe to configure the interface with both "fd on" and > "triple-sampling on"? If not, flexcan_open should return an error when both > CAN_CTRLMODE_FD and CAN_CTRLMODE_3_SAMPLES are set in ctrlmode. [Joakim Zhang] Yes, you are right. I checked in i.MX8QM reference manual, CAN_CTRLMODE_FD and CAN_CTRLMODE_3_SAMPLES should not set together. I will fix it in V3. > Joakim Zhang <qiangqing.zhang@nxp.com> writes: > > + priv->offload.is_canfd = true; > > Shouldn't is_canfd be assigned false in the "else" branch? Otherwise, is_canfd > will stay true after the following sequence: > > root@s32v234evb:~# ip link set can0 type can bitrate 500000 dbitrate > 2000000 fd on > root@s32v234evb:~# ip link set can0 up > root@s32v234evb:~# ip link set can0 down root@s32v234evb:~# ip link set > can0 type can bitrate 500000 fd off root@s32v234evb:~# ip link set can0 up [Joakim Zhang] Good catch! I will fix it in V3. > Moreover, I wonder if we even need is_canfd, because frames should be fed to > the kernel as classic or FD based only on the EDL bit in the C/S word of the Rx > MB, to keep their original format. That is, you may either read EDL from within > a new callback (but I think that mailbox_read should remain the only one > which both locks and unlocks the > mailboxes) or let mailbox_read call alloc_can(fd)_skb itself. [Joakim Zhang] We added can fd support in 4.9 kernel which let mailbox_read call alloc_can(fd)_skb in the past. Now the driver allocate skb before mailbox_read in rx_offload and also read the overflow frames. I add the "is_canfd" since I don't want to change the rx_offload framework too much. @mkl@pengutronix.de, could you give some advice, which solution is better? > Regards, > Stefan
> From: Joakim Zhang > Sent: Tuesday, April 9, 2019 5:07 AM > Hi Stefan, > > Thanks for your validation! Could you add your test tag if you can > successfully validated? Sure, no problem. Please note that I needed to replace "flexcan_read" and "flexcan_write" with "priv->read" and "priv->write" respectively, in PATCH V2 4/5. > [Joakim Zhang] We added can fd support in 4.9 kernel which let > mailbox_read call alloc_can(fd)_skb in the past. Now the driver > allocate skb before mailbox_read in rx_offload and also read the overflow frames. > I add the "is_canfd" since I don't want to change the > rx_offload framework too much. @mkl@pengutronix.de, could you give > some advice, which solution is better? This is more of a functionality issue. For example, candump from canutils 4.0.6 never shows CAN FD frames, but should still be able to show the CAN 2.0 ones regardless of whether "fd on" was supplied. I suggest the following separation: can_rx_offload_offload_one: struct sk_buff *skb = NULL; ... bool drop = unlikely(skb_queue_len(&offload->skb_queue) > offload->skb_queue_len_max); if (offload->mailbox_read(offload, drop, &skb, ×tamp, n) && !skb) offload->dev->stats.rx_dropped++; if (skb) { struct can_rx_offload_cb *cb = can_rx_offload_get_cb(skb); cb->timestamp = timestamp; } return skb; flexcan_mailbox_read: ... if (!drop) { if (reg_ctrl & FLEXCAN_MB_CNT_EDL) *skb = alloc_canfd_skb(offload->dev, &cf); else *skb = alloc_can_skb(offload->dev, (struct can_frame **)&cf); } if (*skb) { /* use cf */ ... } /* mark as read */ Although not visible in the FlexCAN case, the socket buffers would be also freed from inside mailbox_read if errors were encountered after allocation. Regards, Stefan
> -----Original Message----- > From: Stefan-gabriel Mirea > Sent: 2019年4月9日 15:05 > To: Joakim Zhang <qiangqing.zhang@nxp.com>; Marc Kleine-Budde > <mkl@pengutronix.de> > Cc: linux-can@vger.kernel.org; netdev@vger.kernel.org; > linux-kernel@vger.kernel.org > Subject: RE: [PATCH V2 2/5] can: flexcan: add CAN FD mode support > > > From: Joakim Zhang > > Sent: Tuesday, April 9, 2019 5:07 AM > > Hi Stefan, > > > > Thanks for your validation! Could you add your test tag if you can > > successfully validated? > > Sure, no problem. Please note that I needed to replace "flexcan_read" and > "flexcan_write" with "priv->read" and "priv->write" respectively, in PATCH V2 > 4/5. Hi Stefan, I made a mistake here, thank you for pointing out. I will fix it in V3. > > [Joakim Zhang] We added can fd support in 4.9 kernel which let > > mailbox_read call alloc_can(fd)_skb in the past. Now the driver > > allocate skb before mailbox_read in rx_offload and also read the overflow > frames. > > I add the "is_canfd" since I don't want to change the > > rx_offload framework too much. @mkl@pengutronix.de, could you give > > some advice, which solution is better? > > This is more of a functionality issue. For example, candump from canutils 4.0.6 > never shows CAN FD frames, but should still be able to show the CAN 2.0 ones > regardless of whether "fd on" was supplied. I suggest the following separation: > > can_rx_offload_offload_one: > struct sk_buff *skb = NULL; > ... > bool drop = unlikely(skb_queue_len(&offload->skb_queue) > > offload->skb_queue_len_max); > > if (offload->mailbox_read(offload, drop, &skb, ×tamp, n) && !skb) > offload->dev->stats.rx_dropped++; > > if (skb) { > struct can_rx_offload_cb *cb = can_rx_offload_get_cb(skb); > > cb->timestamp = timestamp; > } > > return skb; > > flexcan_mailbox_read: > ... > if (!drop) { > if (reg_ctrl & FLEXCAN_MB_CNT_EDL) > *skb = alloc_canfd_skb(offload->dev, &cf); > else > *skb = alloc_can_skb(offload->dev, > (struct can_frame **)&cf); > } > if (*skb) { > /* use cf */ > ... > } > /* mark as read */ > > Although not visible in the FlexCAN case, the socket buffers would be also > freed from inside mailbox_read if errors were encountered after allocation. This is a good solution, I will add it in V3. Thank you very much. Best Regards, Joakim Zhang > Regards, > Stefan
diff --git a/drivers/net/can/flexcan.c b/drivers/net/can/flexcan.c index e35083ff31ee..f28a4c3e8087 100644 --- a/drivers/net/can/flexcan.c +++ b/drivers/net/can/flexcan.c @@ -52,6 +52,7 @@ #define FLEXCAN_MCR_IRMQ BIT(16) #define FLEXCAN_MCR_LPRIO_EN BIT(13) #define FLEXCAN_MCR_AEN BIT(12) +#define FLEXCAN_MCR_FDEN BIT(11) /* MCR_MAXMB: maximum used MBs is MAXMB + 1 */ #define FLEXCAN_MCR_MAXMB(x) ((x) & 0x7f) #define FLEXCAN_MCR_IDAM_A (0x0 << 8) @@ -137,6 +138,20 @@ FLEXCAN_ESR_BOFF_INT | FLEXCAN_ESR_ERR_INT | \ FLEXCAN_ESR_WAK_INT) +/* FLEXCAN FD control register (FDCTRL) bits */ +#define FLEXCAN_FDCTRL_FDRATE BIT(31) +#define FLEXCAN_FDCTRL_MBDSR3(x) (((x) & 0x3) << 25) +#define FLEXCAN_FDCTRL_MBDSR2(x) (((x) & 0x3) << 22) +#define FLEXCAN_FDCTRL_MBDSR1(x) (((x) & 0x3) << 19) +#define FLEXCAN_FDCTRL_MBDSR0(x) (((x) & 0x3) << 16) + +/* FLEXCAN FD Bit Timing register (FDCBT) bits */ +#define FLEXCAN_FDCBT_FPRESDIV(x) (((x) & 0x3ff) << 20) +#define FLEXCAN_FDCBT_FRJW(x) (((x) & 0x07) << 16) +#define FLEXCAN_FDCBT_FPROPSEG(x) (((x) & 0x1f) << 10) +#define FLEXCAN_FDCBT_FPSEG1(x) (((x) & 0x07) << 5) +#define FLEXCAN_FDCBT_FPSEG2(x) ((x) & 0x07) + /* FLEXCAN interrupt flag register (IFLAG) bits */ /* Errata ERR005829 step7: Reserve first valid MB */ #define FLEXCAN_TX_MB_RESERVED_OFF_FIFO 8 @@ -148,6 +163,10 @@ #define FLEXCAN_IFLAG_RX_FIFO_AVAILABLE BIT(5) /* FLEXCAN message buffers */ +#define FLEXCAN_MB_CNT_EDL BIT(31) +#define FLEXCAN_MB_CNT_BRS BIT(30) +#define FLEXCAN_MB_CNT_ESI BIT(29) + #define FLEXCAN_MB_CODE_MASK (0xf << 24) #define FLEXCAN_MB_CODE_RX_BUSY_BIT (0x1 << 24) #define FLEXCAN_MB_CODE_RX_INACTIVE (0x0 << 24) @@ -192,6 +211,7 @@ #define FLEXCAN_QUIRK_BROKEN_PERR_STATE BIT(6) /* No interrupt for error passive */ #define FLEXCAN_QUIRK_DEFAULT_BIG_ENDIAN BIT(7) /* default to BE register access */ #define FLEXCAN_QUIRK_SETUP_STOP_MODE BIT(8) /* Setup stop mode to support wakeup */ +#define FLEXCAN_QUIRK_TIMESTAMP_SUPPORT_FD BIT(9) /* Use timestamp then support can fd mode */ /* Structure of the message buffer */ struct flexcan_mb { @@ -250,6 +270,9 @@ struct flexcan_regs { u32 rerrdr; /* 0xaf4 */ u32 rerrsynr; /* 0xaf8 */ u32 errsr; /* 0xafc */ + u32 _reserved7[64]; /* 0xb00 */ + u32 fdctrl; /* 0xc00 */ + u32 fdcbt; /* 0xc04 */ }; struct flexcan_devtype_data { @@ -337,6 +360,18 @@ static const struct can_bittiming_const flexcan_bittiming_const = { .brp_inc = 1, }; +static const struct can_bittiming_const flexcan_fd_data_bittiming_const = { + .name = DRV_NAME, + .tseg1_min = 1, + .tseg1_max = 39, + .tseg2_min = 1, + .tseg2_max = 8, + .sjw_max = 8, + .brp_min = 1, + .brp_max = 1024, + .brp_inc = 1, +}; + /* FlexCAN module is essentially modelled as a little-endian IP in most * SoCs, i.e the registers as well as the message buffer areas are * implemented in a little-endian fashion. @@ -609,10 +644,10 @@ static int flexcan_get_berr_counter(const struct net_device *dev, static netdev_tx_t flexcan_start_xmit(struct sk_buff *skb, struct net_device *dev) { const struct flexcan_priv *priv = netdev_priv(dev); - struct can_frame *cf = (struct can_frame *)skb->data; + struct canfd_frame *cf = (struct canfd_frame *)skb->data; u32 can_id; u32 data; - u32 ctrl = FLEXCAN_MB_CODE_TX_DATA | (cf->can_dlc << 16); + u32 ctrl = FLEXCAN_MB_CODE_TX_DATA | ((can_len2dlc(cf->len)) << 16); int i; if (can_dropped_invalid_skb(dev, skb)) @@ -630,7 +665,10 @@ static netdev_tx_t flexcan_start_xmit(struct sk_buff *skb, struct net_device *de if (cf->can_id & CAN_RTR_FLAG) ctrl |= FLEXCAN_MB_CNT_RTR; - for (i = 0; i < cf->can_dlc; i += sizeof(u32)) { + if (can_is_canfd_skb(skb)) + ctrl |= FLEXCAN_MB_CNT_EDL; + + for (i = 0; i < cf->len; i += sizeof(u32)) { data = be32_to_cpup((__be32 *)&cf->data[i]); priv->write(data, &priv->tx_mb->data[i / sizeof(u32)]); } @@ -790,7 +828,7 @@ static inline struct flexcan_priv *rx_offload_to_priv(struct can_rx_offload *off } static unsigned int flexcan_mailbox_read(struct can_rx_offload *offload, - struct can_frame *cf, + struct canfd_frame *cf, u32 *timestamp, unsigned int n) { struct flexcan_priv *priv = rx_offload_to_priv(offload); @@ -836,11 +874,21 @@ static unsigned int flexcan_mailbox_read(struct can_rx_offload *offload, else cf->can_id = (reg_id >> 18) & CAN_SFF_MASK; - if (reg_ctrl & FLEXCAN_MB_CNT_RTR) - cf->can_id |= CAN_RTR_FLAG; - cf->can_dlc = get_can_dlc((reg_ctrl >> 16) & 0xf); + if (reg_ctrl & FLEXCAN_MB_CNT_EDL) { + cf->len = can_dlc2len((reg_ctrl >> 16) & 0x0F); + } else { + cf->len = get_can_dlc((reg_ctrl >> 16) & 0x0F); + + if (reg_ctrl & FLEXCAN_MB_CNT_RTR) + cf->can_id |= CAN_RTR_FLAG; + } + + if (reg_ctrl & FLEXCAN_MB_CNT_ESI) { + cf->flags |= CANFD_ESI; + netdev_warn(priv->can.dev, "ESI Error\n"); + } - for (i = 0; i < cf->can_dlc; i += sizeof(u32)) { + for (i = 0; i < cf->len; i += sizeof(u32)) { __be32 data = cpu_to_be32(priv->read(&mb->data[i / sizeof(u32)])); *(__be32 *)(cf->data + i) = data; } @@ -985,6 +1033,7 @@ static void flexcan_set_bittiming(struct net_device *dev) { const struct flexcan_priv *priv = netdev_priv(dev); const struct can_bittiming *bt = &priv->can.bittiming; + const struct can_bittiming *dbt = &priv->can.data_bittiming; struct flexcan_regs __iomem *regs = priv->regs; u32 reg; @@ -1014,6 +1063,15 @@ static void flexcan_set_bittiming(struct net_device *dev) netdev_dbg(dev, "writing ctrl=0x%08x\n", reg); priv->write(reg, ®s->ctrl); + if (priv->can.ctrlmode & CAN_CTRLMODE_FD) { + reg = FLEXCAN_FDCBT_FPRESDIV(dbt->brp - 1) | + FLEXCAN_FDCBT_FPSEG1(dbt->phase_seg1 - 1) | + FLEXCAN_FDCBT_FPSEG2(dbt->phase_seg2 - 1) | + FLEXCAN_FDCBT_FRJW(dbt->sjw - 1) | + FLEXCAN_FDCBT_FPROPSEG(dbt->prop_seg); + priv->write(reg, ®s->fdcbt); + } + /* print chip status */ netdev_dbg(dev, "%s: mcr=0x%08x ctrl=0x%08x\n", __func__, priv->read(®s->mcr), priv->read(®s->ctrl)); @@ -1028,7 +1086,7 @@ static int flexcan_chip_start(struct net_device *dev) { struct flexcan_priv *priv = netdev_priv(dev); struct flexcan_regs __iomem *regs = priv->regs; - u32 reg_mcr, reg_ctrl, reg_ctrl2, reg_mecr; + u32 reg_mcr, reg_ctrl, reg_ctrl2, reg_mecr, reg_fdctrl; u64 reg_imask; int err, i; struct flexcan_mb __iomem *mb; @@ -1125,6 +1183,24 @@ static int flexcan_chip_start(struct net_device *dev) netdev_dbg(dev, "%s: writing ctrl=0x%08x", __func__, reg_ctrl); priv->write(reg_ctrl, ®s->ctrl); + /* FDCTRL + * + * support BRS when set CAN FD mode + * 64 bytes payload per MB and 7 MBs per RAM block by default + * enable CAN FD mode + */ + if (priv->can.ctrlmode & CAN_CTRLMODE_FD) { + reg_fdctrl = priv->read(®s->fdctrl); + reg_fdctrl |= FLEXCAN_FDCTRL_FDRATE; + reg_fdctrl |= FLEXCAN_FDCTRL_MBDSR3(3) | FLEXCAN_FDCTRL_MBDSR2(3) | + FLEXCAN_FDCTRL_MBDSR1(3) | FLEXCAN_FDCTRL_MBDSR0(3); + priv->write(reg_fdctrl, ®s->fdctrl); + reg_mcr = priv->read(®s->mcr); + priv->write(reg_mcr | FLEXCAN_MCR_FDEN, ®s->mcr); + + priv->offload.is_canfd = true; + } + if ((priv->devtype_data->quirks & FLEXCAN_QUIRK_ENABLE_EACEN_RRS)) { reg_ctrl2 = priv->read(®s->ctrl2); reg_ctrl2 |= FLEXCAN_CTRL2_EACEN | FLEXCAN_CTRL2_RRS; @@ -1261,7 +1337,10 @@ static int flexcan_open(struct net_device *dev) if (err) goto out_close; - priv->mb_size = sizeof(struct flexcan_mb) + CAN_MAX_DLEN; + if (priv->can.ctrlmode & CAN_CTRLMODE_FD) + priv->mb_size = sizeof(struct flexcan_mb) + CANFD_MAX_DLEN; + else + priv->mb_size = sizeof(struct flexcan_mb) + CAN_MAX_DLEN; priv->mb_count = (sizeof(priv->regs->mb[0]) / priv->mb_size) + (sizeof(priv->regs->mb[1]) / priv->mb_size); @@ -1600,6 +1679,17 @@ static int flexcan_probe(struct platform_device *pdev) priv->clk_src = clk_src; priv->devtype_data = devtype_data; priv->reg_xceiver = reg_xceiver; + priv->offload.is_canfd = false; + + if (priv->devtype_data->quirks & FLEXCAN_QUIRK_TIMESTAMP_SUPPORT_FD) { + if (priv->devtype_data->quirks & FLEXCAN_QUIRK_USE_OFF_TIMESTAMP) { + priv->can.ctrlmode_supported |= CAN_CTRLMODE_FD; + priv->can.data_bittiming_const = &flexcan_fd_data_bittiming_const; + } else { + dev_err(&pdev->dev, "canfd mode can't work on fifo mode\n"); + err = -EINVAL; + } + } pm_runtime_get_noresume(&pdev->dev); pm_runtime_set_active(&pdev->dev);