i2c: add support for microchip fpga i2c controllers
Subsystems (2)
PATCH v3 1/1 i2c: add support for microchip fpga i2c controllers
Add Microchip CoreI2C i2c controller support. This driver supports the "hard" i2c controller on the Microchip PolarFire SoC & the basic feature set for "soft" i2c controller implemtations in the FPGA fabric.
---
Changes from v2:
- Fixed whitespace in defines that I removed accidentally
Changes from v1:
- Use byte write and read functions
- Return IRQ_NONE and don't warn if not this drivers interrupt
- Add a clk_disable_unprepare to avoid leaking a clock in the probe
--- drivers/i2c/busses/Kconfig | 11 + drivers/i2c/busses/Makefile | 1 + drivers/i2c/busses/i2c-microchip-core.c | 488 ++++++++++++++++++++++++ 3 files changed, 500 insertions(+) create mode 100644 drivers/i2c/busses/i2c-microchip-core.c
drivers/i2c/busses/Kconfig +11 -0
@@ -781,6 +781,17 @@ config I2C_MESON If you say yes to this option, support will be included for the I2C interface on the Amlogic Meson family of SoCs. +config I2C_MICROCHIP_CORE + tristate "Microchip FPGA I2C controller" + depends on SOC_MICROCHIP_POLARFIRE || COMPILE_TEST + depends on OF + help + If you say yes to this option, support will be included for the + I2C interface on Microchip FPGAs. + + This driver can also be built as a module. If so, the module will be + called i2c-microchip-core. + config I2C_MPC tristate "MPC107/824x/85xx/512x/52xx/83xx/86xx" depends on PPC
drivers/i2c/busses/Makefile +1 -0
@@ -78,6 +78,7 @@ obj-$(CONFIG_I2C_JZ4780) += i2c-jz4780.o obj-$(CONFIG_I2C_KEMPLD) += i2c-kempld.o obj-$(CONFIG_I2C_LPC2K) += i2c-lpc2k.o obj-$(CONFIG_I2C_MESON) += i2c-meson.o +obj-$(CONFIG_I2C_MICROCHIP_CORE) += i2c-microchip-core.o obj-$(CONFIG_I2C_MPC) += i2c-mpc.o obj-$(CONFIG_I2C_MT65XX) += i2c-mt65xx.o obj-$(CONFIG_I2C_MT7621) += i2c-mt7621.o
drivers/i2c/busses/i2c-microchip-core.c new file +488 -0
488 lines not on this page · raw message
Signature · 1 line
2.36.1
Hi Conor,
driver looks mostly good, but some comments:
Quoted 7 lines
+/*
+ *mchp_corei2c_dev- I2C device context
+ * @base: pointer to register struct
+ * @msg: pointer to current message
+ * @msg_len: number of bytes transferred in msg
+ * @msg_err: error code for completed message
+ * @msg_complete: xfer completion object
Quoted 8 lines
+ * @dev: device reference
+ * @adapter: core i2c abstraction
+ * @i2c_clk: clock reference for i2c input clock
+ * @bus_clk_rate: current i2c bus clock rate
+ * @buf: ptr to msg buffer for easier use.
+ * @isr_status: cached copy of local ISR status.
+ * @lock: spinlock for IRQ synchronization.
+ */
This seems outdated, e.g. msg is not in the struct but addr is missing. Also, proper kdoc header is missing?
drivers/i2c/busses/i2c-microchip-core.c line 110
Quoted 3 lines
drivers/i2c/busses/i2c-microchip-core.c new file +3
-0+struct mchp_corei2c_dev { +void __iomem *base; +size_t msg_len;
Maybe u16 to match the original type from struct i2c_msg?
drivers/i2c/busses/i2c-microchip-core.c line 118
Quoted 8 lines
drivers/i2c/busses/i2c-microchip-core.c new file +8
-0+int msg_err; +struct completion msg_complete; +struct device *dev; +struct i2c_adapter adapter; +struct clk *i2c_clk; +spinlock_t lock;/* IRQ synchronization */ +u32 bus_clk_rate; +u32 msg_read;
This is initialized but never used?
Quoted 5 lines
drivers/i2c/busses/i2c-microchip-core.c new file +5
-0+u32 isr_status; +u8 *buf; +u8 addr; +}; +
...
drivers/i2c/busses/i2c-microchip-core.c line 193
Quoted 4 lines
drivers/i2c/busses/i2c-microchip-core.c new file +4
-0+static int mchp_corei2c_init (struct mchp_corei2c_dev *idev) +{ +u32 clk_rate =clk_get_rate (idev->i2c_clk); +u32 divisor = clk_rate / idev->bus_clk_rate;
I don't see a protection against division by zero?
Quoted 6 lines
drivers/i2c/busses/i2c-microchip-core.c new file +6
-0+int ret; + + ret =mchp_corei2c_set_divisor (divisor, idev); +if (ret) +return ret; +
drivers/i2c/busses/i2c-microchip-core.c line 209
Quoted 10 lines
drivers/i2c/busses/i2c-microchip-core.c new file +10
-0+mchp_corei2c_reset (idev); + +return 0 ; +} + +static void mchp_corei2c_transfer (struct mchp_corei2c_dev *idev,u32 data) +{ +if (idev->msg_len >0 ) +writeb (data, idev->base +CORE_I2C_DATA ); +}
Minor: this is very finegrained and only called once. I'd put it into the calling function.
...
Quoted 6 lines
drivers/i2c/busses/i2c-microchip-core.c new file +6
-0+static irqreturn_t mchp_corei2c_handle_isr (struct mchp_corei2c_dev *idev) +{ +u32 status = idev->isr_status; +u8 ctrl; + +if (!idev->buf)
drivers/i2c/busses/i2c-microchip-core.c line 250
Quoted 10 lines
drivers/i2c/busses/i2c-microchip-core.c new file +10
-0+return IRQ_NONE ; + +switch (status) { +case STATUS_M_START_SENT : +case STATUS_M_REPEATED_START_SENT : + ctrl =readb (idev->base +CORE_I2C_CTRL ); + ctrl &= ~CTRL_STA ; +writeb (idev->addr, idev->base +CORE_I2C_DATA ); +writeb (ctrl, idev->base +CORE_I2C_CTRL ); +if (idev->msg_len <=0 )
msg_len is unsigned, can't be < 0?
drivers/i2c/busses/i2c-microchip-core.c line 261
Quoted 11 lines
drivers/i2c/busses/i2c-microchip-core.c new file +11
-0+ goto finished; +break ; +case STATUS_M_ARB_LOST : + idev->msg_err = -EAGAIN ; + goto finished; +case STATUS_M_SLAW_ACK : +case STATUS_M_TX_DATA_ACK : +if (idev->msg_len >0 ) +mchp_corei2c_fill_tx (idev); +else + goto last_byte;
IMO this is a bit too much of goto here. Can't we have a flag for last_byte and handle it at the end of the handler?
Quoted 39 lines
drivers/i2c/busses/i2c-microchip-core.c new file +39
-0+break ; +case STATUS_M_TX_DATA_NACK : +case STATUS_M_SLAR_NACK : +case STATUS_M_SLAW_NACK : + idev->msg_err = -ENXIO ; + goto last_byte; +case STATUS_M_SLAR_ACK : + ctrl =readb (idev->base +CORE_I2C_CTRL ); +if (idev->msg_len ==1u ) { + ctrl &= ~CTRL_AA ; +writeb (ctrl, idev->base +CORE_I2C_CTRL ); + }else { + ctrl |=CTRL_AA ; +writeb (ctrl, idev->base +CORE_I2C_CTRL ); + } +if (idev->msg_len <1u ) + goto last_byte; +break ; +case STATUS_M_RX_DATA_ACKED : +mchp_corei2c_empty_rx (idev); +break ; +case STATUS_M_RX_DATA_NACKED : +mchp_corei2c_empty_rx (idev); +if (idev->msg_len ==0 ) + goto last_byte; +break ; +default : +break ; + } + +return IRQ_HANDLED ; + +last_byte: +/* On the last byte to be transmitted, send STOP */ +mchp_corei2c_stop (idev); +finished: +complete (&idev->msg_complete); +return IRQ_HANDLED ; +}
...
drivers/i2c/busses/i2c-microchip-core.c line 329
Quoted 9 lines
drivers/i2c/busses/i2c-microchip-core.c new file +9
-0+ +static int mchp_corei2c_xfer_msg (struct mchp_corei2c_dev *idev, +struct i2c_msg *msg) +{ +u8 ctrl; +unsigned long time_left; + +if (msg->len ==0 ) +return -EINVAL ;
If your hardware cannot do zero length messages, you need to set up an i2c_adapter_quirk struct with I2C_AQ_NO_ZERO_LEN.
Quoted 7 lines
drivers/i2c/busses/i2c-microchip-core.c new file +7
-0+ + idev->addr =i2c_8bit_addr_from_msg (msg); + idev->msg_len = msg->len; + idev->buf = msg->buf; + idev->msg_err =0 ; + idev->msg_read = (msg->flags &I2C_M_RD ); +
drivers/i2c/busses/i2c-microchip-core.c line 346
Quoted 10 lines
drivers/i2c/busses/i2c-microchip-core.c new file +10
-0+reinit_completion (&idev->msg_complete); + +mchp_corei2c_core_enable (idev); + + ctrl =readb (idev->base +CORE_I2C_CTRL ); + ctrl |=CTRL_STA ; +writeb (ctrl, idev->base +CORE_I2C_CTRL ); + + time_left =wait_for_completion_timeout (&idev->msg_complete, +MICROCHIP_I2C_TIMEOUT );
You should use adapter.timeout here, so it can be changed from userspace.
Quoted 15 lines
drivers/i2c/busses/i2c-microchip-core.c new file +15
-0+if (!time_left) +return -ETIMEDOUT ; + +return idev->msg_err; +} + +static int mchp_corei2c_xfer (struct i2c_adapter *adap,struct i2c_msg *msgs, +int num) +{ +struct mchp_corei2c_dev *idev =i2c_get_adapdata (adap); +int i, ret; + +for (i =0 ; i < num; i++) { + ret =mchp_corei2c_xfer_msg (idev, msgs++); +if (ret)
drivers/i2c/busses/i2c-microchip-core.c line 371
Quoted 10 lines
drivers/i2c/busses/i2c-microchip-core.c new file +10
-0+return ret; + } + +return num; +} + +static u32 mchp_corei2c_func (struct i2c_adapter *adap) +{ +return I2C_FUNC_I2C |I2C_FUNC_SMBUS_EMUL ; +}
If you can't handle zero length messages, you need to mask I2C_FUNC_SMBUS_QUICK out.
drivers/i2c/busses/i2c-microchip-core.c line 380
Quoted 9 lines
drivers/i2c/busses/i2c-microchip-core.c new file +9
-0+ +static const struct i2c_algorithm mchp_corei2c_algo = { + .master_xfer = mchp_corei2c_xfer, + .functionality = mchp_corei2c_func, +}; + +static int mchp_corei2c_probe (struct platform_device *pdev) +{ +struct mchp_corei2c_dev *idev =NULL ;
Unneeded initialization.
Quoted 18 lines
drivers/i2c/busses/i2c-microchip-core.c new file +18
-0+struct resource *res; +int irq, ret; +u32 val; + + idev =devm_kzalloc (&pdev->dev,sizeof (*idev),GFP_KERNEL ); +if (!idev) +return -ENOMEM ; + + idev->base =devm_platform_get_and_ioremap_resource (pdev,0 , &res); +if (IS_ERR (idev->base)) +return PTR_ERR (idev->base); + + irq =platform_get_irq (pdev,0 ); +if (irq <0 ) +return dev_err_probe (&pdev->dev, irq, +"missing interrupt resource\n" ); + + idev->i2c_clk =devm_clk_get (&pdev->dev,NULL );
drivers/i2c/busses/i2c-microchip-core.c line 408
Quoted 10 lines
drivers/i2c/busses/i2c-microchip-core.c new file +10
-0+if (IS_ERR (idev->i2c_clk)) +return dev_err_probe (&pdev->dev,PTR_ERR (idev->i2c_clk), +"missing clock\n" ); + + idev->dev = &pdev->dev; +init_completion (&idev->msg_complete); +spin_lock_init (&idev->lock); + + val =device_property_read_u32 (idev->dev,"clock-frequency" , + &idev->bus_clk_rate);
This functions returns an int. So, 'ret' please which is also more readable.
drivers/i2c/busses/i2c-microchip-core.c line 409
Quoted 1 line
drivers/i2c/busses/i2c-microchip-core.c new file +1
-0+if (val) {
I think the missing check against 'division by zero' should go here.
drivers/i2c/busses/i2c-microchip-core.c line 420
Quoted 11 lines
drivers/i2c/busses/i2c-microchip-core.c new file +11
-0+dev_info (&pdev->dev,"default to 100kHz\n" ); + idev->bus_clk_rate =100000 ; + } + +if (idev->bus_clk_rate >400000 ) +return dev_err_probe (&pdev->dev, -EINVAL , +"clock-frequency too high: %d\n" , + idev->bus_clk_rate); + + ret =devm_request_irq (&pdev->dev, irq, mchp_corei2c_isr,IRQF_SHARED , + pdev->name, idev);
Really SHARED?
Quoted 8 lines
drivers/i2c/busses/i2c-microchip-core.c new file +8
-0+if (ret) +return dev_err_probe (&pdev->dev, ret, +"failed to claim irq %d\n" , irq); + + ret =clk_prepare_enable (idev->i2c_clk); +if (ret) +return dev_err_probe (&pdev->dev, ret, +"failed to enable clock\n" );
drivers/i2c/busses/i2c-microchip-core.c line 438
Quoted 10 lines
drivers/i2c/busses/i2c-microchip-core.c new file +10
-0+ + ret =mchp_corei2c_init (idev); +if (ret) { +clk_disable_unprepare (idev->i2c_clk); +return dev_err_probe (&pdev->dev, ret,"failed to program clock divider\n" ); + } + +i2c_set_adapdata (&idev->adapter, idev); +snprintf (idev->adapter.name,sizeof (idev->adapter.name), +"Microchip I2C hw bus" );
Most people add something like the base address here to distinguish multiple instances.
Quoted 7 lines
drivers/i2c/busses/i2c-microchip-core.c new file +7
-0+ idev->adapter.owner =THIS_MODULE ; + idev->adapter.algo = &mchp_corei2c_algo; + idev->adapter.dev.parent = &pdev->dev; + idev->adapter.dev.of_node = pdev->dev.of_node; + +platform_set_drvdata (pdev, idev); +
drivers/i2c/busses/i2c-microchip-core.c line 455
Quoted 10 lines
drivers/i2c/busses/i2c-microchip-core.c new file +10
-0+ ret =i2c_add_adapter (&idev->adapter); +if (ret) { +clk_disable_unprepare (idev->i2c_clk); +return ret; + } + +dev_info (&pdev->dev,"Microchip I2C Probe Complete\n" ); + +return 0 ; +}
Rest looks good, Thank you for the submission.
All the best,
Wolfram
Quoted 6 lines · Wolfram Sang
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
Quoted 1 line
Hi Conor,
driver looks mostly good, but some comments:
Hey Wolfram, thanks for the review.
It'll be a few days before I can actually get back to you, but:
---8<---
Quoted 14 lines
drivers/i2c/busses/i2c-microchip-core.c line 420
Quoted 12 lines
drivers/i2c/busses/i2c-microchip-core.c new file +11
-0+dev_info (&pdev->dev,"default to 100kHz\n" ); + idev->bus_clk_rate =100000 ; + } + +if (idev->bus_clk_rate >400000 ) +return dev_err_probe (&pdev->dev, -EINVAL , +"clock-frequency too high: %d\n" , + idev->bus_clk_rate); + + ret =devm_request_irq (&pdev->dev, irq, mchp_corei2c_isr,IRQF_SHARED , + pdev->name, idev);Really SHARED?
Yeah, this driver supports both the hard peripherals & soft FPGA cores. The hard peripherals do not have shared IRQs, but we don't really have any control over what way the interrupts are wired for the soft cores. Worth adding a comment to explain the reason for it being shared?
Thanks,
Conor.
Quoted 43 lines
Quoted 9 lines
drivers/i2c/busses/i2c-microchip-core.c new file +8
-0+if (ret) +return dev_err_probe (&pdev->dev, ret, +"failed to claim irq %d\n" , irq); + + ret =clk_prepare_enable (idev->i2c_clk); +if (ret) +return dev_err_probe (&pdev->dev, ret, +"failed to enable clock\n" );drivers/i2c/busses/i2c-microchip-core.c line 438
Quoted 10 lines
drivers/i2c/busses/i2c-microchip-core.c new file +10
-0+ + ret =mchp_corei2c_init (idev); +if (ret) { +clk_disable_unprepare (idev->i2c_clk); +return dev_err_probe (&pdev->dev, ret,"failed to program clock divider\n" ); + } + +i2c_set_adapdata (&idev->adapter, idev); +snprintf (idev->adapter.name,sizeof (idev->adapter.name), +"Microchip I2C hw bus" );Most people add something like the base address here to distinguish multiple instances.
Quoted 7 lines
drivers/i2c/busses/i2c-microchip-core.c new file +7
-0+ idev->adapter.owner =THIS_MODULE ; + idev->adapter.algo = &mchp_corei2c_algo; + idev->adapter.dev.parent = &pdev->dev; + idev->adapter.dev.of_node = pdev->dev.of_node; + +platform_set_drvdata (pdev, idev); +drivers/i2c/busses/i2c-microchip-core.c line 455
Quoted 10 lines
drivers/i2c/busses/i2c-microchip-core.c new file +10
-0+ ret =i2c_add_adapter (&idev->adapter); +if (ret) { +clk_disable_unprepare (idev->i2c_clk); +return ret; + } + +dev_info (&pdev->dev,"Microchip I2C Probe Complete\n" ); + +return 0 ; +}Rest looks good, Thank you for the submission.
Quoted 3 lines
All the best,
Wolfram
_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Quoted 10 lines
Quoted 5 lines
drivers/i2c/busses/i2c-microchip-core.c line 420
Quoted 3 lines
drivers/i2c/busses/i2c-microchip-core.c new file +2
-0+ ret =devm_request_irq (&pdev->dev, irq, mchp_corei2c_isr,IRQF_SHARED , + pdev->name, idev);Really SHARED?
Yeah, this driver supports both the hard peripherals & soft FPGA cores. The hard peripherals do not have shared IRQs, but we don't really have any control over what way the interrupts are wired for the soft cores. Worth adding a comment to explain the reason for it being shared?
Yes. Such information never hurst, I'd say.
Thanks, have a nice weekend,
Wolfram
_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv