serial: spi: add spi-uart driver for Maxim 3110
Subsystems (2)
RFC PATCH 1/1 serial: spi: add spi-uart driver for Maxim 3110
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (drivers/spi: dw_spi.c & dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
As a RFC, the driver doesn't have a valid UART port type yet, but borrow PORT_PXA. I would apply one if there is no main objection for it.
Thanks!
Feng
Quoted 1 line
From
3155eec53d8b1193cae1dce2b0491e86892e6933Mon Sep 17 00:00:00 2001
From: Feng Tang <[email protected]>
Date: Tue, 29 Dec 2009 17:11:17 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110
This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400. It supports both polling and IRQ mode, as well as providing a console for system use
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdf
--- drivers/serial/Kconfig | 17 + drivers/serial/max3110.c | 842 ++++++++++++++++++++++++++++++++++++++++++++++ drivers/serial/max3110.h | 59 ++++ 3 files changed, 918 insertions(+), 0 deletions(-) create mode 100644 drivers/serial/max3110.c create mode 100644 drivers/serial/max3110.h
drivers/serial/Kconfig +17 -0
@@ -688,6 +688,23 @@ config SERIAL_SA1100_CONSOLE your boot loader (lilo or loadlin) about how to pass options to the kernel at boot time.) +config SERIAL_MAX3110 + tristate "SPI UART driver for Max3110" + select SERIAL_CORE + select SERIAL_CORE_CONSOLE + help + This is the UART protocol driver for MAX3110 device + +config MAX3110_DESIGNWARE + boolean "Enable Max3110 to work with Designware controller" + default y + depends on SERIAL_MAX3110 + +config MAX3110_IRQ + boolean "Enable IRQ for Max3110" + default n + depends on SERIAL_MAX3110 + config SERIAL_BFIN tristate "Blackfin serial port support" depends on BLACKFIN
drivers/serial/max3110.c new file +842 -0
842 lines not on this page · raw message
drivers/serial/max3110.h new file +59 -0
59 lines not on this page · raw message
Signature · 1 line
1.6.0.4
Hi Feng,
Quoted 1 line · Feng Tang
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
Is this 3110 device so significantly different from the MAX3100 driver that's already in the mainline kernel (drivers/serial/max3100.c), to require a whole new driver?
Quoted 3 lines
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
baruch
Signature · 3 lines
~. .~ Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
- [email protected] - tel: +972.2.679.5364, http://www.tkos.co.il -Quoted 10 lines
-----Original Message-----
From: Baruch Siach [mailto:[email protected]]
Sent: 2009年12月29日 23:00
To: Tang, Feng
Cc: Greg Kroah-Hartman; David Brownell; Grant Likely; spi-devel-list; [email protected]; [email protected]; Andrew Morton
Subject: Re: [spi-devel-general] [RFC][PATCH] serial: spi: add spi-uart driver for Maxim 3110
Quoted 8 lines
Hi Feng,
Quoted 1 line · Feng Tang
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
Is this 3110 device so significantly different from the MAX3100 driver that's already in the mainline kernel (
drivers/serial/max3100.c), to require a whole new driver?
Yes, I know this question will be asked :) I developed the max3110 before max3100 was posted in public, so the 2 designs differs a lot from the start. I think this driver has 2 good points:
1. It provides a console, which is the main reason that our platform use max3110
2. It utilizes the RX buffer of max3110 that it can reads up to 8 characters in one spi_transfer, and its Tx function can transmit up to 128 chars in one spi_transfer which will save much system load comparing to 1 char per spi transfer
Current max3100.c also has its advantage, like good support in CTS/RTS control, and I think these 2 can merge in the future.
Thanks,
Feng
Quoted 12 lines
Quoted 4 lines
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.baruch
--
~. .~ Tk OpenSystems =}------------------------------------------------ooO--U--Ooo------------{=
- [email protected] - tel: +972.2.679.5364, http://www.tkos.co.il -
Am Mittwoch, den 30.12.2009, 00:05 +0800 schrieb Tang, Feng:
Quoted 26 lines
Quoted 9 lines
-----Original Message-----
From: Baruch Siach [mailto:[email protected]]
Sent: 2009年12月29日 23:00
To: Tang, Feng
Cc: Greg Kroah-Hartman; David Brownell; Grant Likely; spi-devel-list; [email protected]; [email protected]; Andrew Morton
Subject: Re: [spi-devel-general] [RFC][PATCH] serial: spi: add spi-uart driver for Maxim 3110Quoted 8 lines
Hi Feng,
Quoted 1 line · Feng Tang
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
Is this 3110 device so significantly different from the MAX3100 driver that's already in the mainline kernel (
drivers/serial/max3100.c), to require a whole new driver?Yes, I know this question will be asked :) I developed the max3110 before max3100 was posted in public, so the 2 designs differs a lot from the start. I think this driver has 2 good points:
1. It provides a console, which is the main reason that our platform use max3110
2. It utilizes the RX buffer of max3110 that it can reads up to 8 characters in onespi_transfer, and its Tx function can transmit up to 128 chars in onespi_transferwhich will save much system load comparing to 1 char per spi transfer
Quoted 13 lines
Current
max3100.calso has its advantage, like good support in CTS/RTS control, and I think these 2 can merge in the future.Thanks,
FengQuoted 8 lines
Quoted 4 lines
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.baruch
--
Hi,
looking at your driver, there are several points that make me doubt that it is working at all:
* The MAX3110 requires CS going low at the start of each 16-bit transfer, if you look at the datasheet.
* in max3110_read_multi, it looks to me that this will work for big-endian architectures only.
* I can't see how send_circ_buf() should work at all. You are just sending a burst of characters to the MAX3110 without checking the transmit buffer status at all. The MAX3110 has a single TX buffer only, so that will cause transmit characters to be lost.
I think there's no need for a MAX3100 **and** a MAX3110 driver, this is just confusing. The MAX3110 driver is identical to the MAX3100 from the software view, it is simply a MAX3100 with transceivers added to the chip. If there's any improvement, that should be merged into the existing MAX3100 driver.
-Erwin
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, 30 Dec 2009 02:43:48 +0800
Erwin Authried <[email protected]> wrote:
Hi Erwin,
Thanks for the comments.
Quoted 5 lines
Hi,
looking at your driver, there are several points that make me doubt that it is working at all:
You can trust me on this point:) it has worked reliably on our HW platform for long time. I don't know if Alan has played with the platform, but David Woodhouse has, max3110's console is one of the main debug methods for developers.
Actually it can co-work well with other SPI devices connecting to the same SPI controller(our platform use a Designware core based controller)
Quoted 3 lines
* The MAX3110 requires CS going low at the start of each 16-bit transfer, if you look at the datasheet.
Our SPI controller can handle this (drivers/spi/dw_spi.c), it has a dedicated chip select register for handling chip select. And this part is transparent for all SPI slave devices connecting to the controller
Quoted 3 lines
* in
max3110_read_multi, it looks to me that this will work for big-endian architectures only.
We are running on x86 architecture, with little endian. But endian issue is a good point
Quoted 5 lines
* I can't see how
send_circ_buf()should work at all. You are just sending a burst of characters to the MAX3110 without checking the transmit buffer status at all. The MAX3110 has a single TX buffer only, so that will cause transmit characters to be lost.
Yes, I let the SPI controller driver handle this, which has a rather big FIFO. And structure spi_transfer has a member "speed_hz", this driver set it according to current UART baud rate, then SPI controller driver will adjust the bus frequency between controller and 3110 accordingly to prevent its overflow, which will avoid explicit delay in driver
Quoted 6 lines
I think there's no need for a MAX3100 **and** a MAX3110 driver, this is just confusing. The MAX3110 driver is identical to the MAX3100 from the software view, it is simply a MAX3100 with transceivers added to the chip. If there's any improvement, that should be merged into the existing MAX3100 driver.
I agree there should not be 2 drivers cover 1 family of HW, so this is a RFC. I've thinked about merge with current 3100 code, but it depends on one char per spi_transfer, while my driver relys on batch data transfer for efficiency. Another key point is the console, SPI UART can't be directly accessed by CPU, so every spi_transfer will go through tasklet/workqueue, which makes supporting printk a big part of my driver.
Thanks,
Feng
Quoted 4 lines
-Erwin
Quoted 5 lines · Erwin Authried
I think there's no need for a MAX3100 **and** a MAX3110 driver, this is just confusing. The MAX3110 driver is identical to the MAX3100 from the software view, it is simply a MAX3100 with transceivers added to the chip. If there's any improvement, that should be merged into the existing MAX3100 driver.
Assuming that's true ... who will resolve the issue?
Quoted 9 lines · On Thu, 25 Feb 2010 12:47:13 +0800 David Brownell <[email protected]> wrote:
Quoted 5 lines · Erwin Authried
I think there's no need for a MAX3100 **and** a MAX3110 driver, this is just confusing. The MAX3110 driver is identical to the MAX3100 from the software view, it is simply a MAX3100 with transceivers added to the chip. If there's any improvement, that should be merged into the existing MAX3100 driver.
Assuming that's true ... who will resolve the issue?
Hi David,
I've answered Erwin's comments before in v1 submission cycle, following is the quote:
"I agree there should not be 2 drivers cover 1 family of HW, so this is a RFC. I've thinked about merge with current 3100 code, but it depends on one char per spi_transfer, while my driver relys on batch data transfer for efficiency. Another key point is the console, SPI UART can't be directly accessed by CPU, so every spi_transfer will go through tasklet/workqueue, which makes supporting printk a big part of my driver."
I really did consider about that, but has no good clue, so I think better to shape my driver first.
Thanks,
Feng
Quoted 2 lines
As a RFC, the driver doesn't have a valid UART port type yet, but borrow
PORT_PXA. I would apply one if there is no main objection for it.
Please just add a new uart type for it.
drivers/serial/Kconfig config SERIAL_SA1100_CONSOLE · line 707
Quoted 10 lines
drivers/serial/Kconfig +10
-0+config MAX3110_DESIGNWARE + boolean "Enable Max3110 to work with Designware controller" + default y + depends on SERIAL_MAX3110 + +config MAX3110_IRQ + boolean "Enable IRQ for Max3110" + default n + depends on SERIAL_MAX3110 +
These should be runtime
drivers/serial/max3110.c line 91
Quoted 10 lines
drivers/serial/max3110.c new file +10
-0+int max3110_write_then_read (struct uart_max3110 *max, +const u8 *txbuf,u8 *rxbuf,unsigned len,int always_fast) +{ +struct spi_device *spi = max->spi; +struct spi_message message; +struct spi_transfer x; +int ret; + +if (!txbuf || !rxbuf) +return -EINVAL ;
How can the above occur - it seems no user triggered event can cause it and almost all the paths then lose the error before userspace gets it (and silently).
drivers/serial/max3110.c line 144
Quoted 4 lines
drivers/serial/max3110.c new file +4
-0+if (len >MAX_READ_LEN ) { +pr_err (PR_FMT "read len %d is too large\n" , len); +return 0 ; + }
Ditto
drivers/serial/max3110.c line 348
Quoted 7 lines
drivers/serial/max3110.c new file +7
-0+while (len) { + usable =tty_buffer_request_room (tty, len); +if (usable) { +tty_insert_flip_string (tty, str, usable); + str += usable; + port->icount.rx += usable; +tty_flip_buffer_push (tty);
You really only want to push once. tty_insert_flip_string also knows about multiple blocks so all of this can be a single
tty_insert_flip_string(tty, str, len); tty_flip_buffer_push(tty);
Quoted 11 lines
drivers/serial/max3110.c new file +10
-0+static void +serial_m3110_set_termios (struct uart_port *port,struct ktermios *termios, +struct ktermios *old) +{ +struct uart_max3110 *max = + container_of(port, struct uart_max3110, port); +unsigned char cval; +unsigned int baud, parity =0 ; +int clk_div =-1 ; +u16 new_conf = max->cur_conf;
drivers/serial/max3110.c line 562
Quoted 10 lines
drivers/serial/max3110.c new file +10
-0+ +switch (termios->c_cflag &CSIZE ) { +case CS7 : + cval =UART_LCR_WLEN7 ; + new_conf |=WC_7BIT_WORD ; +break ; +default : +case CS8 : + cval =UART_LCR_WLEN8 ; + new_conf |=WC_8BIT_WORD ;
If you can only do CS8 when asked for other values you need to update the termios struct to report that
termios->c_cflag &= ~CSIZE; termios->c_flag |= CS8;
drivers/serial/max3110.c line 643
Quoted 12 lines
drivers/serial/max3110.c new file +12
-0+if (!(termios->c_cflag &PARODD )) + parity |=UART_LCR_EPAR ; + max->parity = parity; + +uart_update_timeout (port, termios->c_cflag, baud); + + new_conf |=WC_TAG ; +if (new_conf != max->cur_conf) { +max3110_out (max, new_conf); + max->cur_conf = new_conf; + max->baud = baud; + }
And as you don't support it
termios->c_cflag &= ~CMSPAR;
I'll take a more serious look over it in the new year when I'm back at work.
------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community Take advantage of Verizon's best-in-class app development support A streamlined, 14 day to market process makes app distribution fast and easy Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
RFC PATCH v2 1/1 serial: spi: add spi-uart driver for Maxim 3110
In reply to Feng TangHi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (drivers/spi: dw_spi.c & dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
change since v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementation
Thanks!
Feng
Quoted 1 line
From
6d14c5d68cdae8d48b6d8a00b6653022f2b100d0Mon Sep 17 00:00:00 2001
From: Feng Tang <[email protected]>
Date: Mon, 8 Feb 2010 16:02:59 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110
This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400, and supports both polling and IRQ mode, as well as providing a console for system use
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdf
--- drivers/serial/Kconfig | 12 + drivers/serial/max3110.c | 848 +++++++++++++++++++++++++++++++++++++++++++ drivers/serial/max3110.h | 59 +++ include/linux/serial_core.h | 3 + 4 files changed, 922 insertions(+), 0 deletions(-) create mode 100644 drivers/serial/max3110.c create mode 100644 drivers/serial/max3110.h
drivers/serial/Kconfig +12 -0
@@ -688,6 +688,18 @@ config SERIAL_SA1100_CONSOLE your boot loader (lilo or loadlin) about how to pass options to the kernel at boot time.) +config SERIAL_MAX3110 + tristate "SPI UART driver for Max3110" + select SERIAL_CORE + select SERIAL_CORE_CONSOLE + help + This is the UART protocol driver for MAX3110 device + +config MAX3110_DESIGNWARE + boolean "Enable Max3110 to work with Designware controller" + default y + depends on SERIAL_MAX3110 + config SERIAL_BFIN tristate "Blackfin serial port support" depends on BLACKFIN
drivers/serial/max3110.c new file +848 -0
848 lines not on this page · raw message
drivers/serial/max3110.h new file +59 -0
59 lines not on this page · raw message
include/linux/serial_core.h +3 -0
@@ -182,6 +182,9 @@/* Aeroflex Gaisler GRLIB APBUART */ #define PORT_APBUART 90 +/* Maxim M3110 */ +#define PORT_MAX3110 91 +#ifdef __KERNEL__#include <linux/compiler.h>
Signature · 1 line
1.6.3.3
Quoted 22 lines · On Mon, 8 Feb 2010 16:59:46 +0800 Feng Tang <[email protected]> wrote:
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.change since v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementationThanks!
FengQuoted 1 line
From
6d14c5d68cdae8d48b6d8a00b6653022f2b100d0Mon Sep 17 00:00:00 2001From: Feng Tang <[email protected]>
Date: Mon, 8 Feb 2010 16:02:59 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110
Quoted 8 lines
This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400, and supports both polling and IRQ mode, as well as providing a console for system use
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdf
I wonder if this is an "spi" subsystem thing or a "serial" subsystem thing. It looks more like a serial driver to me.
Quoted 15 lines
--- drivers/serial/Kconfig | 12 + drivers/serial/max3110.c | 848 +++++++++++++++++++++++++++++++++++++++++++ drivers/serial/max3110.h | 59 +++ include/linux/serial_core.h | 3 +drivers/serial/Kconfig +1
-0+static int use_irq = 1;
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+module_param(use_irq, int, 0444); +MODULE_PARM_DESC(use_irq, "Whether using Max3110's IRQ capability"); + +static void receive_chars(struct uart_max3110 *max, + unsigned char *str, int len); +static int max3110_read_multi(struct uart_max3110 *max, int len, u8 *buf); +static void max3110_con_receive(struct uart_max3110 *max); + +int max3110_write_then_read(struct uart_max3110 *max, + const u8 *txbuf, u8 *rxbuf, unsigned len, int always_fast)
The driver has a number of identifiers which have global scope, apparently unnecessarily. Please review all such identifiers, see if they can be made static.
Quoted 15 lines
drivers/serial/Kconfig +15
-0+{ + struct spi_device *spi = max->spi; + struct spi_message message; + struct spi_transfer x; + int ret; + + spi_message_init(&message); + memset(&x, 0, sizeof x); + x.len = len; + x.tx_buf = txbuf; + x.rx_buf = rxbuf; + spi_message_add_tail(&x, &message); + + if (always_fast) + x.speed_hz = spi->max_speed_hz;
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+ else if (max->baud) + x.speed_hz = max->baud; + + /* Do the i/o */ + ret = spi_sync(spi, &message); + return ret; +} + +/* Write a u16 to the device, and return one u16 read back */ +int max3110_out(struct uart_max3110 *max, const u16 out)
Another.
drivers/serial/Kconfig
Quoted 5 lines
drivers/serial/Kconfig +5
-0+{ + u16 tmp; + int ret; + + ret = max3110_write_then_read(max, (u8 *)&out, (u8 *)&tmp, 2, 1);
Something nasty is happening here. Modifying a u16 via a u8* is to assume a certain endianness, surely. Will the driver break when used on other-endian architectures?
Quoted 21 lines
drivers/serial/Kconfig +20
-0+ if (ret) + return ret; + + /* If some valid data is read back */ + if (tmp & MAX3110_READ_DATA_AVAILABLE) { + tmp &= 0xff; + receive_chars(max, (unsigned char *)&tmp, 1); + } + + return ret; +} + +#define MAX_READ_LEN 20 +/* + * This is usually used to read data from SPIC RX FIFO, which doesn't + * need any delay like flushing character out. + * Returns how many valide bytes are read back + */ +static int max3110_read_multi(struct uart_max3110 *max, int len, u8 *buf) +{
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+ u16 out[MAX_READ_LEN], in[MAX_READ_LEN]; + u8 *pbuf, valid_str[MAX_READ_LEN]; + int i, j, bytelen; + + bytelen = len * 2; + memset(out, 0, bytelen); + memset(in, 0, bytelen); + + if (max3110_write_then_read(max, (u8 *)out, (u8 *)in, bytelen, 1)) + return 0;
max3110_write_then_read() can return an error code (from spi_async()). But here that error code simply gets lost. It should be reported to higher-level code somehow?
Quoted 17 lines
drivers/serial/Kconfig +17
-0+ /* If caller doesn't provide a buffer, then handle received char */ + pbuf = buf ? buf : valid_str; + + for (i = 0, j = 0; i < len; i++) { + if (in[i] & MAX3110_READ_DATA_AVAILABLE) + pbuf[j++] = (u8)(in[i] & 0xff); + } + + if (j && (pbuf == valid_str)) + receive_chars(max, valid_str, j); + + return j; +} + +static void serial_m3110_con_putchar(struct uart_port *port, int ch) +{ + struct uart_max3110 *max =
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+ container_of(port, struct uart_max3110, port); + struct circ_buf *xmit = &max->con_xmit; + + if (uart_circ_chars_free(xmit)) { + xmit->buf[xmit->head] = (char)ch; + xmit->head = (xmit->head + 1) & (PAGE_SIZE - 1); + } + + if (!atomic_read(&max->con_tx_need)) { + atomic_set(&max->con_tx_need, 1);
This manipulation of con_tx_need looks racy on SMP or preempt.
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +7
-0+ wake_up_process(max->main_thread); + } +} + +#define WORDS_PER_XFER 128 +static inline void send_circ_buf(struct uart_max3110 *max, + struct circ_buf *xmit)
I suggest that the `inline' be removed. Modern gcc will take care of this, if it is of benefit.
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+{ + int len, left = 0; + u16 obuf[WORDS_PER_XFER], ibuf[WORDS_PER_XFER]; + u8 valid_str[WORDS_PER_XFER]; + int i, j; + + while (!uart_circ_empty(xmit)) { + left = uart_circ_chars_pending(xmit); + while (left) { + len = (left >= WORDS_PER_XFER) ? WORDS_PER_XFER : left;
Use
len = min(left, WORDS_PER_XFER);
drivers/serial/Kconfig
Quoted 3 lines
drivers/serial/Kconfig +3
-0+ + memset(obuf, 0, len * 2); + memset(ibuf, 0, len * 2);
Using
memset(obuf, 0, sizeof(obuf));
would remove the assumptions that a) obuf is of type u16 and b) that sizeof(u16)==2.
drivers/serial/Kconfig
Quoted 4 lines
drivers/serial/Kconfig +4
-0+ for (i = 0; i < len; i++) { + obuf[i] = (u8)xmit->buf[xmit->tail] | WD_TAG; + xmit->tail = (xmit->tail + 1) & + (UART_XMIT_SIZE - 1);
Could this driver use include/linux/kfifo.h, rather than open-coding it?
drivers/serial/Kconfig
Quoted 3 lines
drivers/serial/Kconfig +3
-0+ } + max3110_write_then_read(max, (u8 *)obuf, + (u8 *)ibuf, len * 2, 0);
Error codes are ignored.
Quoted 28 lines
drivers/serial/Kconfig +25
-0+ for (i = 0, j = 0; i < len; i++) { + if (ibuf[i] & MAX3110_READ_DATA_AVAILABLE) + valid_str[j++] = (u8)(ibuf[i] & 0xff); + } + + if (j) + receive_chars(max, valid_str, j); + + max->port.icount.tx += len; + left -= len; + } + } +} + +static void max3110_con_receive(struct uart_max3110 *max) +{ + int loop = 1, num, total = 0; + u8 recv_buf[512], *pbuf; + + pbuf = recv_buf; + do { + /* 3110 RX buffer is 8 words */ + num = max3110_read_multi(max, 8, pbuf); + + if (num) {
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+ loop = 5; + pbuf += num; + total += num; + + if (total >= 500) { + receive_chars(max, recv_buf, total); + pbuf = recv_buf; + total = 0; + } + }
hm, what do all the magic numbers do? Some comments explaining them would be good.
Quoted 17 lines
drivers/serial/Kconfig +17
-0+ } while (--loop); + + if (total) + receive_chars(max, recv_buf, total); +} + +static int max3110_main_thread(void *_max) +{ + struct uart_max3110 *max = _max; + wait_queue_head_t *wq = &max->wq; + int ret = 0; + struct circ_buf *xmit = &max->con_xmit; + + init_waitqueue_head(wq); + pr_info(PR_FMT "start main thread\n"); + + do {
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+ wait_event_interruptible(*wq, (atomic_read(&max->irq_pending) || + atomic_read(&max->con_tx_need) || + atomic_read(&max->uart_tx_need)) || + kthread_should_stop()); + max->mthread_up = 1; + + if (use_irq && atomic_read(&max->irq_pending)) { + max3110_con_receive(max); + atomic_set(&max->irq_pending, 0); + }
The manipulation of irq_pending looks racy. Perhaps something list test_and_clear_bit() would fix that.
drivers/serial/Kconfig
Quoted 5 lines
drivers/serial/Kconfig +5
-0+ /* First handle console output */ + if (atomic_read(&max->con_tx_need)) { + send_circ_buf(max, xmit); + atomic_set(&max->con_tx_need, 0); + }
Ditto.
drivers/serial/Kconfig
Quoted 5 lines
drivers/serial/Kconfig +5
-0+ /* Handle uart output */ + if (atomic_read(&max->uart_tx_need)) { + transmit_char(max); + atomic_set(&max->uart_tx_need, 0); + }
Ditto.
drivers/serial/Kconfig
Quoted 7 lines
drivers/serial/Kconfig +7
-0+ max->mthread_up = 0; + } while (!kthread_should_stop()); + + return ret; +} + +irqreturn_t static serial_m3110_irq(int irq, void *dev_id)
`static irqreturn_t' would be more typical.
drivers/serial/Kconfig
Quoted 7 lines
drivers/serial/Kconfig +7
-0+{ + struct uart_max3110 *max = dev_id; + + /* max3110's irq is a falling edge, not level triggered, + * so no need to disable the irq */ + if (!atomic_read(&max->irq_pending)) { + atomic_inc(&max->irq_pending);
racy?
Quoted 7 lines
drivers/serial/Kconfig +7
-0+ wake_up_process(max->main_thread); + } + return IRQ_HANDLED; +} + +/* If don't use RX IRQ, then need a thread to polling read */ +static int max3110_read_thread(void *_max)
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+{ + struct uart_max3110 *max = _max; + + pr_info(PR_FMT "start read thread\n"); + do { + if (!max->mthread_up) + max3110_con_receive(max); + + set_current_state(TASK_INTERRUPTIBLE); + schedule_timeout(HZ / 20);
Use schedule_timeout_interruptible()
Quoted 28 lines
drivers/serial/Kconfig +28
-0+ } while (!kthread_should_stop()); + + return 0; +} + +static int serial_m3110_startup(struct uart_port *port) +{ + struct uart_max3110 *max = + container_of(port, struct uart_max3110, port); + u16 config = 0; + int ret = 0; + + if (port->line != 0) + pr_err(PR_FMT "uart port startup failed\n"); + + /* Firstly disable all IRQ and config it to 115200, 8n1 */ + config = WC_TAG | WC_FIFO_ENABLE + | WC_1_STOPBITS + | WC_8BIT_WORD + | WC_BAUD_DR2; + ret = max3110_out(max, config); + + /* As we use thread to handle tx/rx, need set low latency */ + port->state->port.tty->low_latency = 1; + + if (use_irq) { + ret = request_irq(max->irq, serial_m3110_irq, + IRQ_TYPE_EDGE_FALLING, "max3110", max);
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+ if (ret) + return ret; + + /* Enable RX IRQ only */ + config |= WC_RXA_IRQ_ENABLE; + max3110_out(max, config); + } else { + /* if IRQ is disabled, start a read thread for input data */ + max->read_thread = + kthread_run(max3110_read_thread, max, "max3110_read");
Check for kthread_run() failure?
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +7
-0+ } + + max->cur_conf = config; + return 0; +} + +static int serial_m3110_probe(struct spi_device *spi)
Should this be __init or __devinit?
Quoted 74 lines
drivers/serial/Kconfig +71
-0+{ + struct uart_max3110 *max; + int ret; + unsigned char *buffer; + + max = kzalloc(sizeof(*max), GFP_KERNEL); + if (!max) + return -ENOMEM; + + /* set spi info */ + spi->mode = SPI_MODE_0; + spi->bits_per_word = 16; +#ifdef CONFIG_MAX3110_DESIGNWARE + spi->controller_data = &spi_uart; +#endif + spi_setup(spi); + + max->clock = MAX3110_HIGH_CLK; + max->port.type = PORT_MAX3110; + max->port.fifosize = 2; /* only have 16b buffer */ + max->port.ops = &serial_m3110_ops; + max->port.line = 0; + max->port.dev = &spi->dev; + max->port.uartclk = 115200; + + max->spi = spi; + max->name = spi->modalias; /* use spi name as the name */ + max->irq = (u16)spi->irq; + + spin_lock_init(&max->lock); + + max->word_7bits = 0; + max->parity = 0; + max->baud = 0; + + max->cur_conf = 0; + atomic_set(&max->irq_pending, 0); + + buffer = (unsigned char *)__get_free_page(GFP_KERNEL); + if (!buffer) { + ret = -ENOMEM; + goto err_get_page; + } + max->con_xmit.buf = (unsigned char *)buffer; + max->con_xmit.head = max->con_xmit.tail = 0; + + max->main_thread = kthread_run(max3110_main_thread, + max, "max3110_main"); + if (IS_ERR(max->main_thread)) { + ret = PTR_ERR(max->main_thread); + goto err_kthread; + } + + pmax = max; + /* Give membase a psudo value to pass serial_core's check */ + max->port.membase = (void *)0xff110000; + uart_add_one_port(&serial_m3110_reg, &max->port); + + return 0; + +err_kthread: + free_page((unsigned long)buffer); +err_get_page: + pmax = NULL; + kfree(max); + return ret; +} + +static struct spi_driver uart_max3110_driver = { + .driver = { + .name = "spi_max3111",
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+ .bus = &spi_bus_type, + .owner = THIS_MODULE, + }, + .probe = serial_m3110_probe, + .remove = __devexit_p(max3110_remove), + .suspend = serial_m3110_suspend, + .resume = serial_m3110_resume, +}; + +int __init serial_m3110_init(void)
static?
Quoted 5 lines
drivers/serial/Kconfig +5
-0+{ + int ret = 0; + + ret = uart_register_driver(&serial_m3110_reg); + if (ret)
drivers/serial/Kconfig
Quoted 10 lines
drivers/serial/Kconfig +10
-0+ return ret; + + ret = spi_register_driver(&uart_max3110_driver); + if (ret) + uart_unregister_driver(&serial_m3110_reg); + + return ret; +} + +void __exit serial_m3110_exit(void)
static?
Quoted 8 lines
drivers/serial/Kconfig +5
-0+{ + spi_unregister_driver(&uart_max3110_driver); + uart_unregister_driver(&serial_m3110_reg); +} +
Quoted 30 lines · Andrew Morton
Quoted 27 lines · On Mon, 8 Feb 2010 16:59:46 +0800 Feng Tang <[email protected]> wrote:
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.change since v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementationThanks!
FengQuoted 1 line
From
6d14c5d68cdae8d48b6d8a00b6653022f2b100d0Mon Sep 17 00:00:00 2001From: Feng Tang <[email protected]>
Date: Mon, 8 Feb 2010 16:02:59 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400, and supports both polling and IRQ mode, as well as providing a console for system use
Quoted 6 lines
Quoted 3 lines
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdfI wonder if this is an "spi" subsystem thing or a "serial" subsystem thing. It looks more like a serial driver to me.
I'm assuming serial; and hence I haven't picked it up into my tree.
g.
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 15 lines · On Tue, 9 Feb 2010 08:26:35 +0800 Grant Likely <[email protected]>…
Quoted 12 lines · Andrew Morton
Quoted 9 lines · On Mon, 8 Feb 2010 16:59:46 +0800 Feng Tang <[email protected]> wrote:
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
Quoted 6 lines
Quoted 2 lines
I wonder if this is an "spi" subsystem thing or a "serial" subsystem thing. It looks more like a serial driver to me.
I'm assuming serial; and hence I haven't picked it up into my tree.
g.
Yes, it's more related to serial stuff, and thus I put them into drivers/serial
Thanks,
Feng
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 21 lines · Feng Tang
Quoted 18 lines · On Tue, 9 Feb 2010 08:26:35 +0800 Grant Likely…
Quoted 15 lines · Andrew Morton
Quoted 9 lines · On Mon, 8 Feb 2010 16:59:46 +0800 Feng Tang…
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.I wonder if this is an "spi" subsystem thing or a "serial" subsystem thing. It looks more like a serial driver to me.
Quoted 6 lines
Quoted 3 lines
I'm assuming serial; and hence I haven't picked it up into my tree.
g.
Yes, it's more related to serial stuff, and thus I put them into
drivers/serial
That's fine. Care to address Andrew's issues and resend the patch?
thanks,
greg k-h
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
RFC PATCH v3 1/1 serial: spi: add spi-uart driver for Maxim 3110
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (drivers/spi: dw_spi.c & dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
change log:
since v2:
* Address comments from Andrew Morton
* Use test/set_bit to avoid race condition for SMP/preempt case
* Fix bug related with endian order
since v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementation
Thanks!
Feng
Quoted 1 line
From
93455ea6fbf0bfb6494b88ea83296f26f29f7eeeMon Sep 17 00:00:00 2001
From: Feng Tang <[email protected]>
Date: Tue, 23 Feb 2010 17:52:00 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110
This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400, and supports both polling and IRQ mode, as well as providing a console for system use
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdf
--- drivers/serial/Kconfig | 12 + drivers/serial/max3110.c | 879 +++++++++++++++++++++++++++++++++++++++++++ drivers/serial/max3110.h | 60 +++ include/linux/serial_core.h | 3 + 4 files changed, 954 insertions(+), 0 deletions(-) create mode 100644 drivers/serial/max3110.c create mode 100644 drivers/serial/max3110.h
drivers/serial/Kconfig +12 -0
@@ -688,6 +688,18 @@ config SERIAL_SA1100_CONSOLE your boot loader (lilo or loadlin) about how to pass options to the kernel at boot time.) +config SERIAL_MAX3110 + tristate "SPI UART driver for Max3110" + select SERIAL_CORE + select SERIAL_CORE_CONSOLE + help + This is the UART protocol driver for MAX3110 device + +config MAX3110_DESIGNWARE + boolean "Enable Max3110 to work with Designware controller" + default y + depends on SERIAL_MAX3110 + config SERIAL_BFIN tristate "Blackfin serial port support" depends on BLACKFIN
drivers/serial/max3110.c new file +879 -0
879 lines not on this page · raw message
drivers/serial/max3110.h new file +60 -0
60 lines not on this page · raw message
include/linux/serial_core.h +3 -0
@@ -182,6 +182,9 @@/* Aeroflex Gaisler GRLIB APBUART */ #define PORT_APBUART 90 +/* Maxim M3110 */ +#define PORT_MAX3110 91 +#ifdef __KERNEL__#include <linux/compiler.h>
Signature · 1 line
1.6.3.3
Quoted 7 lines · On Wed, 24 Feb 2010 13:11:30 +0800 Feng Tang <[email protected]> wrote:
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
Ack for the serial side
Grant Likely
Acked-byQuoted 7 lines · Alan Cox
Quoted 4 lines · On Wed, 24 Feb 2010 13:11:30 +0800 Feng Tang <[email protected]> wrote:
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
Quoted 7 lines
Quoted 3 lines
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.Ack for the serial side
Acked-by Alan Cox <[email protected]>
Ack for the SPI side
g.
Quoted 78 lines · On Wed, 24 Feb 2010 13:11:30 +0800 Feng Tang <[email protected]> wrote:
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.change log:
since v2:
* Address comments from Andrew Morton
* Use test/set_bit to avoid race condition for SMP/preempt case
* Fix bug related with endian ordersince v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementationThanks!
FengQuoted 1 line
From
93455ea6fbf0bfb6494b88ea83296f26f29f7eeeMon Sep 17 00:00:00 2001From: Feng Tang <[email protected]>
Date: Tue, 23 Feb 2010 17:52:00 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400, and supports both polling and IRQ mode, as well as providing a console for system use
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdf...
drivers/serial/max3110.c new file +40
-0+static int max3110_main_thread (void *_max) +{ +struct uart_max3110 *max = _max; +wait_queue_head_t *wq = &max->wq; +int ret =0 ; +struct circ_buf *xmit = &max->con_xmit; + +init_waitqueue_head (wq); +pr_info (PR_FMT "start main thread\n" ); + +do { +wait_event_interruptible (*wq, + max->flags ||kthread_should_stop ()); +set_bit (0 , &max->mthread_up); + +if (use_irq &&test_bit (M3110_IRQ_PENDING , &max->flags)) { +max3110_con_receive (max); +clear_bit (M3110_IRQ_PENDING , &max->flags); + } + +/* First handle console output */ +if (test_bit (M3110_CON_TX_NEED , &max->flags)) { +send_circ_buf (max, xmit); +clear_bit (M3110_CON_TX_NEED , &max->flags); + } + +/* Handle uart output */ +if (test_bit (M3110_UART_TX_NEED , &max->flags)) { +transmit_char (max); +clear_bit (M3110_UART_TX_NEED , &max->flags); + } +clear_bit (0 , &max->mthread_up); + }while (!kthread_should_stop ()); + +return ret; +} + +static irqreturn_t serial_m3110_irq (int irq,void *dev_id) +{ +struct uart_max3110 *max = dev_id;
drivers/serial/max3110.c line 439
Quoted 10 lines
drivers/serial/max3110.c new file +10
-0+ +/* max3110's irq is a falling edge, not level triggered, +* so no need to disable the irq */ +set_bit (M3110_IRQ_PENDING , &max->flags); + +if (!test_bit (0 , &max->mthread_up)) +wake_up_process (max->main_thread); + +return IRQ_HANDLED ; +}
The manipulation of mthread_up here (and in several other places) is clearly quite racy. If you hit that race, the thread will not wake up and the driver will sit there not doing anything until some other event happens.
This is perhaps fixable with test_and_set_bit() and test_and_clear_bit() (need to think about that) or, of course, by adding locking.
But a simpler fix is just to delete mthread_up altogether. wake_up_process() on a running process is an OK thing to do and hopefully isn't terribly slow.
drivers/serial/max3110.c line 452
Quoted 12 lines
drivers/serial/max3110.c new file +12
-0+/* If don't use RX IRQ, then need a thread to polling read */ +static int max3110_read_thread (void *_max) +{ +struct uart_max3110 *max = _max; + +pr_info (PR_FMT "start read thread\n" ); + do { + if (!test_bit(0, &max->mthread_up)) + max3110_con_receive(max); + + set_current_state(TASK_INTERRUPTIBLE); + schedule_timeout_interruptible(HZ / 20);
The set_current_state(TASK_INTERRUPTIBLE) is unneeded.
Quoted 4 lines
drivers/serial/max3110.c new file +4
-0+ }while (!kthread_should_stop ()); + +return 0 ; +}
Quoted 60 lines · On Thu, 25 Feb 2010 07:18:32 +0800 Andrew Morton <[email protected]>…
Quoted 44 lines
...
(quoted diff) +13
-0+static int max3110_main_thread(void *_max) +{ + struct uart_max3110 *max = _max; + wait_queue_head_t *wq = &max->wq; + int ret = 0; + struct circ_buf *xmit = &max->con_xmit; + + init_waitqueue_head(wq); + pr_info(PR_FMT "start main thread\n"); + + do { + wait_event_interruptible(*wq, + max->flags ||
kthread_should_stop());(quoted diff) +3
-0+ set_bit(0, &max->mthread_up); + + if (use_irq && test_bit(M3110_IRQ_PENDING,
&max->flags)) {(quoted diff) +24
-0+ max3110_con_receive(max); + clear_bit(M3110_IRQ_PENDING, &max->flags); + } + + /* First handle console output */ + if (test_bit(M3110_CON_TX_NEED, &max->flags)) { + send_circ_buf(max, xmit); + clear_bit(M3110_CON_TX_NEED, &max->flags); + } + + /* Handle uart output */ + if (test_bit(M3110_UART_TX_NEED, &max->flags)) { + transmit_char(max); + clear_bit(M3110_UART_TX_NEED, &max->flags); + } + clear_bit(0, &max->mthread_up); + } while (!kthread_should_stop()); + + return ret; +} + +static irqreturn_t serial_m3110_irq(int irq, void *dev_id) +{ + struct uart_max3110 *max = dev_id;Review
Quoted 10 lines
(quoted diff) +10
-0+ + /* max3110's irq is a falling edge, not level triggered, + * so no need to disable the irq */ + set_bit(M3110_IRQ_PENDING, &max->flags); + + if (!test_bit(0, &max->mthread_up)) + wake_up_process(max->main_thread); + + return IRQ_HANDLED; +}The manipulation of
mthread_uphere (and in several other places) is clearly quite racy. If you hit that race, the thread will not wake up and the driver will sit there not doing anything until some other event happens.
Quoted 7 lines
This is perhaps fixable with
test_and_set_bit()andtest_and_clear_bit()(need to think about that) or, of course, by adding locking.But a simpler fix is just to delete
mthread_upaltogether.wake_up_process()on a running process is an OK thing to do and hopefully isn't terribly slow.
Yes, wake_up_process won't harm a running process, our driver's case is a little special, the console's write() func may call wake_up_process() for every character in the buffer, thus we will try to avoid to call it. mthread_up can't be removed as it is also referenced in read_thread.
I prefer to use the test_and_set/clear_bit for "mthread_up".
Thanks,
Feng
drivers/serial/max3110.c +2 -2
@@ -400,7 +400,7 @@ static int max3110_main_thread(void *_max) do { wait_event_interruptible(*wq, max->flags || kthread_should_stop()); - set_bit(0, &max->mthread_up); + test_and_set_bit(0, &max->mthread_up); if (use_irq && test_bit(M3110_IRQ_PENDING, &max->flags)) { max3110_con_receive(max);
@@ -418,7 +418,7 @@ static int max3110_main_thread(void *_max)transmit_char (max);clear_bit (M3110_UART_TX_NEED , &max->flags); } -clear_bit (0 , &max->mthread_up); +test_and_clear_bit (0 , &max->mthread_up); }while (!kthread_should_stop ());return ret;
Quoted 5 lines · Feng Tang
+config
SERIAL_MAX3110
+ tristate "SPI UART driver for Max3110"
+ selectSERIAL_CORE
+ selectSERIAL_CORE_CONSOLE
Shouldn't that depend on SPI_MASTER? As it stands, you're permitting it to build on systems that you *know* don't have a prayer of running this code ... or often, even finishing the build.
drivers/serial/Kconfig config SERIAL_SA1100_CONSOLE · line 698
Quoted 4 lines
+ help
+ This is the UART protocol driver for MAX3110 devicedrivers/serial/Kconfig +2
-0+ +config MAX3110_DESIGNWARE
Having this depend on a specific underlying SPI master controller is not a good thing. It should instead be written so that it runs correctly on *any* controller which exports the standard SPI programming interface.
Quoted 4 lines
+ boolean "Enable Max3110 to work with Designware controller" + default y
+ depends onSERIAL_MAX3110
+
That is, stuff like this, from your max3110 driver:
drivers/serial/max3110.c line 752
Quoted 8 lines
drivers/serial/max3110.c new file +8
-0+ +#ifdef CONFIG_MAX3110_DESIGNWARE +static struct dw_spi_chip spi_uart = { + .poll_mode = 1, + .enable_dma = 0, + .type = SPI_FRF_SPI, +}; +#endif
Is completely irrelevant for other hardware ... or else (as with DMA, which should be "enabled by default") is relevant, but shouldn't be parameterized.
drivers/serial/max3110.c line 769
Quoted 5 lines
+ spi->mode =
SPI_MODE_0;
+ spi->bits_per_word = 16;drivers/serial/max3110.c new file +3
-0+#ifdef CONFIG_MAX3110_DESIGNWARE + spi->controller_data = &spi_uart; +#endif
That all looks fishy too. SPI_MODE should have
been set up as part of device creation. Ditto
any spi->controller_data ... normally, that all
gets set up as part of board-specific setup.
Normally one goes to a lot of effort to keep
such board-specific code out of drivers.
- Dave
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 57 lines · On Thu, 25 Feb 2010 12:43:14 +0800 David Brownell <[email protected]> wrote:
Quoted 5 lines · Feng Tang
+config
SERIAL_MAX3110
+ tristate "SPI UART driver for Max3110"
+ selectSERIAL_CORE
+ selectSERIAL_CORE_CONSOLEShouldn't that depend on
SPI_MASTER? As it stands, you're permitting it to build on systems that you *know* don't have a prayer of running this code ... or often, even finishing the build.drivers/serial/Kconfig config SERIAL_SA1100_CONSOLE · line 698
Quoted 4 lines
+ help
+ This is the UART protocol driver for MAX3110 devicedrivers/serial/Kconfig +2
-0+ +config MAX3110_DESIGNWAREHaving this depend on a specific underlying SPI master controller is not a good thing. It should instead be written so that it runs correctly on *any* controller which exports the standard SPI programming interface.
Quoted 4 lines
+ boolean "Enable Max3110 to work with Designware controller" + default y
+ depends onSERIAL_MAX3110
+That is, stuff like this, from your max3110 driver:
drivers/serial/max3110.c line 752
Quoted 8 lines
drivers/serial/max3110.c new file +8
-0+ +#ifdef CONFIG_MAX3110_DESIGNWARE +static struct dw_spi_chip spi_uart = { + .poll_mode = 1, + .enable_dma = 0, + .type = SPI_FRF_SPI, +}; +#endif Is completely irrelevant for other hardware ... or else (as with DMA, which should be "enabled by default") is relevant, but shouldn't be parameterized.
drivers/serial/max3110.c line 769
Quoted 5 lines
+ spi->mode =
SPI_MODE_0;
+ spi->bits_per_word = 16;drivers/serial/max3110.c new file +3
-0+#ifdef CONFIG_MAX3110_DESIGNWARE + spi->controller_data = &spi_uart; +#endif That all looks fishy too.
SPI_MODEshould have
been set up as part of device creation. Ditto
anyspi->controller_data... normally, that all
gets set up as part of board-specific setup.
Quoted 2 lines
Normally one goes to a lot of effort to keep
such board-specific code out of drivers.
Good point about the DW controller specific data, I'll remove them. For those "bits_per_word" setting, I think we can put it here instead of the board initialization code, as many types of boards can leverage the setting here as it only works in 16b mode.
Thanks,
Feng
Quoted 3 lines
- Dave
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 1 line · Feng Tang
Good point about the DW controller specific data, I'll remove them.
Good.
Quoted 3 lines
For those "
bits_per_word" setting, I think we can put it here instead of the board initialization code, as many types of boards can leverage the setting here as it only works in 16b mode.
Yes. The bits-per-word setting is mostly driver-specific, so that's appropriate as something the driver updates.
- Dave
PATCH v4 1/1 serial: spi: add spi-uart driver for Maxim 3110
In reply to Feng TangHi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (drivers/spi: dw_spi.c & dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
change log:
since v3:
* Remove the Designware controller related stuff
* Remove some initialization code which should be done in the platform setup code
* Add a missing change for drivers/serial/Makefile
since v2:
* Address comments from Andrew Morton
* Use test/set_bit to avoid race condition for SMP/preempt case
* Fix bug related with endian order
since v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementation
Thanks!
Feng
Quoted 1 line
From
4c4fef8784c38e72f6a4578d48688388a426fe67Mon Sep 17 00:00:00 2001
From: Feng Tang <[email protected]>
Date: Fri, 26 Feb 2010 11:37:29 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110
This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400, and supports both polling and IRQ mode, as well as providing a console for system use
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdf
--- drivers/serial/Kconfig | 8 + drivers/serial/Makefile | 1 + drivers/serial/max3110.c | 862 +++++++++++++++++++++++++++++++++++++++++++ drivers/serial/max3110.h | 60 +++ include/linux/serial_core.h | 3 + 5 files changed, 934 insertions(+), 0 deletions(-) create mode 100644 drivers/serial/max3110.c create mode 100644 drivers/serial/max3110.h
drivers/serial/Kconfig +8 -0
@@ -688,6 +688,14 @@ config SERIAL_SA1100_CONSOLE your boot loader (lilo or loadlin) about how to pass options to the kernel at boot time.) +config SERIAL_MAX3110 + tristate "SPI UART driver for Max3110" + depends on SPI_MASTER + select SERIAL_CORE + select SERIAL_CORE_CONSOLE + help + This is the UART protocol driver for MAX3110 device + config SERIAL_BFIN tristate "Blackfin serial port support" depends on BLACKFIN
drivers/serial/Makefile +1 -0
@@ -46,6 +46,7 @@ obj-$(CONFIG_SERIAL_S3C24A0) += s3c24a0.o obj-$(CONFIG_SERIAL_S3C6400) += s3c6400.o obj-$(CONFIG_SERIAL_S5PC100) += s3c6400.o obj-$(CONFIG_SERIAL_MAX3100) += max3100.o +obj-$(CONFIG_SERIAL_MAX3110) += max3110.o obj-$(CONFIG_SERIAL_IP22_ZILOG) += ip22zilog.o obj-$(CONFIG_SERIAL_MUX) += mux.o obj-$(CONFIG_SERIAL_68328) += 68328serial.o
drivers/serial/max3110.c new file +862 -0
862 lines not on this page · raw message
drivers/serial/max3110.h new file +60 -0
60 lines not on this page · raw message
include/linux/serial_core.h +3 -0
@@ -182,6 +182,9 @@/* Aeroflex Gaisler GRLIB APBUART */ #define PORT_APBUART 90 +/* Maxim M3110 */ +#define PORT_MAX3110 91 +#ifdef __KERNEL__#include <linux/compiler.h>
Signature · 1 line
1.6.3.3
Hi,
Quoted 3 lines · On Fri, 26 Feb 2010 11:47:29 +0800 Feng Tang <[email protected]> wrote:
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
<snip>
drivers/serial/max3110.c line 140
Quoted 9 lines
drivers/serial/max3110.c new file +9
-0+/* +* This is usually used to read data from SPIC RX FIFO, which doesn't +* need any delay like flushing character out. +* +* Return how many valide bytes are read back +*/ +static int max3110_read_multi (struct uart_max3110 *max,u8 *buf) +{ +u16 obuf[M3110_RX_FIFO_DEPTH ], ibuf[M3110_RX_FIFO_DEPTH ];
Doing I/O on stack is guaranteed safe for spi functions?
drivers/serial/max3110.c line 145
Quoted 5 lines
drivers/serial/max3110.c new file +5
-0+u8 *pbuf, valid_str[M3110_RX_FIFO_DEPTH ]; +int i, j; + +memset (obuf,0 ,sizeof (obuf)); +memset (obuf,0 ,sizeof (ibuf));
memset(ibuf, 0, sizeof(ibuf))?
Quoted 17 lines
drivers/serial/max3110.c new file +17
-0+ +if (max3110_write_then_read (max, obuf, ibuf, +M3110_RX_FIFO_DEPTH *2 ,1 )) +return 0 ; + +/* If caller doesn't provide a buffer, then handle received char */ + pbuf = buf ? buf : valid_str; + +for (i =0 , j =0 ; i <M3110_RX_FIFO_DEPTH ; i++) { +if (ibuf[i] &MAX3110_READ_DATA_AVAILABLE ) + pbuf[j++] = ibuf[i] &0xff ; + } + +if (j && (pbuf == valid_str)) +receive_chars (max, valid_str, j); + +return j;
Signature · 1 line
Masakazu Mokuno
Quoted 5 lines · Masakazu Mokuno
drivers/serial/max3110.c line 140
Quoted 3 lines
drivers/serial/max3110.c new file +3
-0+static int max3110_read_multi (struct uart_max3110 *max,u8 *buf) +{ + u16 obuf[M3110_RX_FIFO_DEPTH], ibuf[M3110_RX_FIFO_DEPTH];Doing I/O on stack is guaranteed safe for spi functions?
Good catch ... no it's not safe.
No DMA-enabled programming interface allows it, including SPI.
The Documentation/PCI/PCI-DMA-mapping.txt file has a section with the oddly punctuated title "What memory is DMA'able?". That's generic to all uses of DMA. When it was moved to the PCI area, that information became harder to find ... but no less true for non-PCI contexts (sigh). One relevant paragraph being:
This rule also means that you may use neither kernel image addresses (items in data/text/bss segments), nor module image addresses, nor stack addresses for DMA. These could all be mapped somewhere entirely different than the rest of physical memory. Even if those classes of memory could physically work with DMA, you'd need to ensure the I/O buffers were cacheline-aligned. Without that, you'd see cacheline sharing problems (data corruption) on CPUs with DMA-incoherent caches. (The CPU could write to one word, DMA would write to a different one in the same cache line, and one of them could be overwritten.)
Those cacheline sharing problems are particularly bad for the stack, since the data corruption often includes return addresses.
In short, passing a stack-based I/O buffer could work on some machines, but cause perplexing data corruption issues on others. So don't try to do it any driver that claims to be portable.
- Dave
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 10 lines · On Sat, 27 Feb 2010 03:41:42 +0800 David Brownell <[email protected]> wrote:
Quoted 5 lines · Masakazu Mokuno
drivers/serial/max3110.c line 140
Quoted 3 lines
drivers/serial/max3110.c new file +3
-0+static int max3110_read_multi (struct uart_max3110 *max,u8 *buf) +{ + u16 obuf[M3110_RX_FIFO_DEPTH], ibuf[M3110_RX_FIFO_DEPTH];Doing I/O on stack is guaranteed safe for spi functions?
Good catch ... no it's not safe.
No DMA-enabled programming interface allows it, including SPI.
Hi David,
I agree that DMA working mode must not use stack as IO buffer, and spi core code(spi.c/spi.h) has clearly message stating that. But this driver doesn't intend to use DMA by design. In early version of this patch there is some controller specific data which show "enable_dma = 0".
The reason of not using DMA for Maxim 3110 is, it is a low speed SPI UART device, which only has one word (16b) TX buffer and 8 words RX buffer, using DMA may benefit in some use case but will be over-kill for others. This driver implement a console, take one example of editing a file on the console, each char we input will be echoed back on screen, which will use a spi_message and a spi_transfer, if using DMA, that DMA operation length will be 2 bytes, if the file been edited has 4K characters, there will be 4K DMA transactions, which will occupy quiet some system load.
Also whether a 16b word could be DMAed is still a question for some types of DMA controllers, I know some platform whose DMA controllers can only or configured to only work with data at least 32 bits aligned.
So in this driver, it leverages some options provided by SPI controller drivers (pxa2xx and Designware) to chose not using DMA.
Thanks,
Feng
Quoted 18 lines
The
Documentation/PCI/PCI-DMA-mapping.txtfile has a section with the oddly punctuated title "What memory is DMA'able?". That's generic to all uses of DMA. When it was moved to the PCI area, that information became harder to find ... but no less true for non-PCI contexts (sigh). One relevant paragraph being:This rule also means that you may use neither kernel image addresses (items in data/text/bss segments), nor module image addresses, nor stack addresses for DMA. These could all be mapped somewhere
entirely different than the rest of physical memory. Even if those classes of memory could physically work with DMA, you'd need to ensure the I/O buffers were cacheline-aligned. Without that, you'd see cacheline sharing problems (data corruption) on CPUs with DMA-incoherent caches. (The CPU could write to one word, DMA would write to a different one in the same cache line, and one of them could be overwritten.)
Quoted 9 lines
Those cacheline sharing problems are particularly bad for the stack, since the data corruption often includes return addresses.
In short, passing a stack-based I/O buffer could work on some machines, but cause perplexing data corruption issues on others. So don't try to do it any driver that claims to be portable.
- Dave
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 9 lines · On Fri, 26 Feb 2010 17:59:38 +0800 Masakazu Mokuno <[email protected]> wrote:
Hi,
Review
Quoted 5 lines
(quoted diff) +5
-0+ u8 *pbuf, valid_str[M3110_RX_FIFO_DEPTH]; + int i, j; + + memset(obuf, 0, sizeof(obuf)); + memset(obuf, 0, sizeof(ibuf));
memset(ibuf, 0, sizeof(ibuf))?
thanks for the catch, here is a follow on patch
drivers/serial/max3110.c +1 -1
@@ -142,7 +142,7 @@ static int max3110_read_multi(struct uart_max3110 *max, u8 *buf)int i, j;memset (obuf,0 ,sizeof (obuf)); -memset (obuf,0 ,sizeof (ibuf)); +memset (ibuf,0 ,sizeof (ibuf));if (max3110_write_then_read (max, obuf, ibuf,M3110_RX_FIFO_DEPTH *2 ,1 ))
PATCH v4 1/1 serial: spi: add spi-uart driver for Maxim 3110
In reply to Feng TangHi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (drivers/spi: dw_spi.c & dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
change log:
since v3:
* Remove the Designware controller related stuff
* Remove some initialization code which should be done in the platform setup code
* Add a missing change for drivers/serial/Makefile
since v2:
* Address comments from Andrew Morton
* Use test/set_bit to avoid race condition for SMP/preempt case
* Fix bug related with endian order
since v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementation
Thanks!
Feng
Quoted 1 line
From
4c4fef8784c38e72f6a4578d48688388a426fe67Mon Sep 17 00:00:00 2001
From: Feng Tang <[email protected]>
Date: Fri, 26 Feb 2010 11:37:29 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110
This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400, and supports both polling and IRQ mode, as well as providing a console for system use
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdf
--- drivers/serial/Kconfig | 8 + drivers/serial/Makefile | 1 + drivers/serial/max3110.c | 862 +++++++++++++++++++++++++++++++++++++++++++ drivers/serial/max3110.h | 60 +++ include/linux/serial_core.h | 3 + 5 files changed, 934 insertions(+), 0 deletions(-) create mode 100644 drivers/serial/max3110.c create mode 100644 drivers/serial/max3110.h
drivers/serial/Kconfig +8 -0
@@ -688,6 +688,14 @@ config SERIAL_SA1100_CONSOLE your boot loader (lilo or loadlin) about how to pass options to the kernel at boot time.) +config SERIAL_MAX3110 + tristate "SPI UART driver for Max3110" + depends on SPI_MASTER + select SERIAL_CORE + select SERIAL_CORE_CONSOLE + help + This is the UART protocol driver for MAX3110 device + config SERIAL_BFIN tristate "Blackfin serial port support" depends on BLACKFIN
drivers/serial/Makefile +1 -0
@@ -46,6 +46,7 @@ obj-$(CONFIG_SERIAL_S3C24A0) += s3c24a0.o obj-$(CONFIG_SERIAL_S3C6400) += s3c6400.o obj-$(CONFIG_SERIAL_S5PC100) += s3c6400.o obj-$(CONFIG_SERIAL_MAX3100) += max3100.o +obj-$(CONFIG_SERIAL_MAX3110) += max3110.o obj-$(CONFIG_SERIAL_IP22_ZILOG) += ip22zilog.o obj-$(CONFIG_SERIAL_MUX) += mux.o obj-$(CONFIG_SERIAL_68328) += 68328serial.o
drivers/serial/max3110.c new file +862 -0
862 lines not on this page · raw message
drivers/serial/max3110.h new file +60 -0
60 lines not on this page · raw message
include/linux/serial_core.h +3 -0
@@ -182,6 +182,9 @@/* Aeroflex Gaisler GRLIB APBUART */ #define PORT_APBUART 90 +/* Maxim M3110 */ +#define PORT_MAX3110 91 +#ifdef __KERNEL__#include <linux/compiler.h>
Signature · 8 lines
1.6.3.3 ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev
Quoted 6 lines
drivers/serial/max3110.c line 273
Quoted 4 lines
drivers/serial/max3110.c new file +4
-0+for (i =0 ; i < len; i++) { + obuf[i] = (u8 )xmit->buf[xmit->tail] |WD_TAG ; + xmit->tail = (xmit->tail +1 ) & + (UART_XMIT_SIZE -1 );Could this driver use
include/linux/kfifo.h, rather than open-coding it?
The circ buffer stuff comes from the serial layer. The whole serial layer wants switching to kfifo (as and when we know why the kfifo changes seem to have broken USB serial) but the driver can't do that itself and use the serial core code until then.
(Actually most of the serial_core code wants a rewrite but having had a couple of goes at doing it I've not figured out any way to do it without simply breaking all sorts of embedded platform serial drivers that need specific hardware to fix).
Quoted 5 lines
drivers/serial/max3110.c line 276
Quoted 3 lines
drivers/serial/max3110.c new file +3
-0+ } +max3110_write_then_read (max, (u8 *)obuf, + (u8 *)ibuf, len *2 ,0 );Error codes are ignored.
The drivers/serial midlayer has no provision for reporting them.
Alan
PATCH v5 1/1 serial: spi: add spi-uart driver for Maxim 3110
In reply to Feng TangHi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (drivers/spi: dw_spi.c & dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
change log:
since v4:
* Add explanation why not using DMA
* Fix a bug in max3110_read_multi()
since v3:
* Remove the Designware controller related stuff
* Remove some initialization code which should be done in the platform setup code
* Add a missing change for drivers/serial/Makefile
since v2:
* Address comments from Andrew Morton
* Use test/set_bit to avoid race condition for SMP/preempt case
* Fix bug related with endian order
since v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementation
Thanks!
Feng
Quoted 1 line
From
6ea4918365f5eed04f82544f0da3efe93cb2cc20Mon Sep 17 00:00:00 2001
From: Feng Tang <[email protected]>
Date: Fri, 26 Feb 2010 11:37:29 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110
This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400, and supports both polling and IRQ mode, as well as providing a console for system use
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdf
--- drivers/serial/Kconfig | 8 + drivers/serial/Makefile | 1 + drivers/serial/max3110.c | 875 +++++++++++++++++++++++++++++++++++++++++++ drivers/serial/max3110.h | 61 +++ include/linux/serial_core.h | 3 + 5 files changed, 948 insertions(+), 0 deletions(-) create mode 100644 drivers/serial/max3110.c create mode 100644 drivers/serial/max3110.h
drivers/serial/Kconfig +8 -0
@@ -688,6 +688,14 @@ config SERIAL_SA1100_CONSOLE your boot loader (lilo or loadlin) about how to pass options to the kernel at boot time.) +config SERIAL_MAX3110 + tristate "SPI UART driver for Max3110" + depends on SPI_MASTER + select SERIAL_CORE + select SERIAL_CORE_CONSOLE + help + This is the UART protocol driver for MAX3110 device + config SERIAL_BFIN tristate "Blackfin serial port support" depends on BLACKFIN
drivers/serial/Makefile +1 -0
@@ -46,6 +46,7 @@ obj-$(CONFIG_SERIAL_S3C24A0) += s3c24a0.o obj-$(CONFIG_SERIAL_S3C6400) += s3c6400.o obj-$(CONFIG_SERIAL_S5PC100) += s3c6400.o obj-$(CONFIG_SERIAL_MAX3100) += max3100.o +obj-$(CONFIG_SERIAL_MAX3110) += max3110.o obj-$(CONFIG_SERIAL_IP22_ZILOG) += ip22zilog.o obj-$(CONFIG_SERIAL_MUX) += mux.o obj-$(CONFIG_SERIAL_68328) += 68328serial.o
drivers/serial/max3110.c new file +875 -0
875 lines not on this page · raw message
drivers/serial/max3110.h new file +61 -0
61 lines not on this page · raw message
include/linux/serial_core.h +3 -0
@@ -182,6 +182,9 @@/* Aeroflex Gaisler GRLIB APBUART */ #define PORT_APBUART 90 +/* Maxim M3110 */ +#define PORT_MAX3110 91 +#ifdef __KERNEL__#include <linux/compiler.h>
Signature · 1 line
1.6.3.3
Quoted 26 lines · Feng Tang
Hi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (
drivers/spi:dw_spi.c&dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.change log:
since v4:
* Add explanation why not using DMA
* Fix a bug inmax3110_read_multi()
since v3:
* Remove the Designware controller related stuff
* Remove some initialization code which should be done in the platform setup code
* Add a missing change fordrivers/serial/Makefile
since v2:
* Address comments from Andrew Morton
* Use test/set_bit to avoid race condition for SMP/preempt case
* Fix bug related with endian order
since v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementation
[...]
Quoted 4 lines
+ *
+ * 3. This driver doesn't support work with a spi cotnroller in DMA mode, and
+ * the reason for not using DMA is:
+ * a) Maxim 3110 is a low speed UART device, whose tx/rx buffer are very few,
Quoted 8 lines
+ * and using DMA may be over-killing when working as a system console(imaging
+ * we edit a text file on it)
+ * b) Handling only one 16b word transfer will be very common, but non-32b
+ * aligned DMA operation is not supported by all kinds of DMA controllers
+ *
+ * Some spi controller drivers like Pxa2xx, Designware have option for device
+ * driver to chose to work in poll or DMA mode. And platform driver which
+ * enumerates Max3110 device should leverage this option to not use DMA.
Between this, reading through the other comments, and the existence of the max3100 driver, I'm less and less comfortable with this driver. First, if this driver gets merged, then it is quite likely that neither you or the 3100 author will get around to merging them. Second, as others have pointed out, this driver is making assumptions about the SPI bus that it has no business making. The fact that it doesn't work with DMA isn't a design choice, it is a bug and whether or not it is a low speed uart device is completely beside the point.
I'm generally not involved with the serial device drivers, but FWIW, I'm no longer in favor of this driver getting merged. If you really want to get it merged, then I recommend asking Greg to pick it up into staging so that there is some imperative to fix it up and merge it with the max3100 driver.
g.
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 1 line · Feng Tang
+ * 3. This driver doesn't support work with a spi cotnroller in DMA mode,
As Grant said: That's a bug ... one that will randomly kick in based on whether the underlying SPI controller driver happens to use DMA for a given transaction.
Quoted 10 lines · On Wed, 3 Mar 2010 12:51:58 +0800 David Brownell <[email protected]> wrote:
Quoted 2 lines · Feng Tang
+ * 3. This driver doesn't support work with a spi cotnroller in DMA mode,
As Grant said: That's a bug ... one that will randomly kick in based on whether the underlying SPI controller driver happens to use DMA for a given transaction.
From this perspective, yes, it's a bug that this driver use non DMA-safe
IO buffer, which prevents it to be portable for all kinds of controllers, and I can make it DMA safe.
But I still don't think it's a good idea to use DMA for Max3110 for performance reason, I know there is some spi controller who only works in DMA mode.
Here comes another idea, can we add a capability flag in struct spi_master indicating the master supporting poll or dma or both. Also we add similar bits in struct spi_transfer indicating the this transfer wants to be handled in poll or dma mode. For spi controller driver, it can claim its capability when registering as a spi_master, for a spi device driver, it can select the working mode based on spi_master's capability and set the working mode in struct spi_transfer. Then controller and device don't need to guess what the other is capable, and make the info public. Any thoughts?
Thanks,
Feng
Quoted 7 lines · Feng Tang
Quoted 4 lines · On Wed, 3 Mar 2010 12:51:58 +0800 David Brownell <[email protected]> wrote:
Quoted 2 lines · Feng Tang
+ * 3. This driver doesn't support work with a spi cotnroller in DMA mode,
Quoted 7 lines
Quoted 5 lines
As Grant said: That's a bug ... one that will randomly kick in based on whether the underlying SPI controller driver happens to use DMA for a given transaction.
From this perspective, yes, it's a bug that this driver use non DMA-safe IO buffer, which prevents it to be portable for all kinds of controllers,
Yes, that's a bug ... you're using a programming interface which has portability as a core goal. In an operating system kernel which also has such a goal.
Quoted 1 line
and I can make it DMA safe.
That's the right solution.
Quoted 4 lines
Here comes another idea, can we add a capability flag in struct
spi_masterindicating the master supporting poll or dma or both. Also we add similar bits in structspi_transferindicating the this transfer wants to be handled in poll or dma mode.
Let's not do either of those. There's no need to introduce such complexity, or to enable the associated new failure modes and bugs.
Much simpler to just use DMA-safe I/O buffers in the first place ... which is what pretty much every driver stack in Linux already expects or requires.
- Dave
Quoted 9 lines
Quoted 6 lines
Here comes another idea, can we add a capability flag in struct
spi_masterindicating the master supporting poll or dma or both. Also we add similar bits in structspi_transferindicating the this transfer wants to be handled in poll or dma mode.Let's not do either of those. There's no need to introduce such complexity, or to enable the associated new failure modes and bugs.
This idea has nothing to do with the dma-safe problem you pointed out, I will make the buffer dma safe anyway.
What I proposed just wants to provide some flexibility for protocol device drivers, it will use dma-safe buffer always, but it prefer not to use DMA for its one-word transfers, and hope to have a choice to do that. For current existing controller and device drivers, they can simply ignore the new working mode setting in struct spi_master and spi_transfer and leave them as 0.
Thanks,
Feng
Quoted 7 lines
Much simpler to just use DMA-safe I/O buffers in the first place ... which is what pretty much every driver stack in Linux already expects or requires.
- Dave
Quoted 9 lines · Feng Tang
Quoted 7 lines
Quoted 4 lines
Here comes another idea, can we add a capability flag in struct
spi_masterindicating the master supporting poll or dma or both. Also we add similar bits in structspi_transferindicating the this transfer wants to be handled in poll or dma mode.Let's not do either of those. There's no need to introduce such complexity, or to enable the associated new failure modes and bugs.
This idea has nothing to do with the dma-safe problem you pointed out,
Your comment has nothing to do with my response. Are you implying that your suggestions do *NOT* introduce complexity, including new failure modes and thus the possibility of new confusing types of bugs?
Quoted 2 lines
I will
make the buffer dma safe anyway.
Good ...
Quoted 3 lines
What I proposed just wants to provide some flexibility for protocol device drivers, it will use dma-safe buffer always, but it prefer not to use DMA for its one-word transfers,
That kind of heuristic is something the SPI controller driver could already apply if it's a good tradeoff on that hardware. (Considering also the extra added complexity, failure modes, and potential for new flavors of bug...)
Of course, drivers like the max3110 are high enough in the driver stack that they can only guess about such tradeoffs. (And thus most drivers will likely guess wrong...)
Quoted 4 lines
and hope to have a choice to do that. For current
existing controller and device drivers, they can simply ignore the new working mode setting in structspi_masterandspi_transferand leave them as 0.
If someone decided to update a SPI controller driver to avoid DMA in some cases, in favor of PIO, they could already code such heuristics without needing your proposed hinting from upper layers.
The result in the low-level driver would be just to use a different test (maybe "is this a one-word transfer?" instead of checking your per-transfer "use PIO?" hint) before kicking in whatever logic you think would improve performance (by eliminating DMA setup and teardown costs, including cache cleaning).
- Dave
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 7 lines · On Wed, 3 Mar 2010 15:25:52 +0800 David Brownell <[email protected]> wrote:
If someone decided to update a SPI controller driver to avoid DMA in some cases, in favor of PIO, they could already code such heuristics without needing your proposed hinting from upper layers.
The result in the low-level driver would be just to use a different test (maybe "is this a one-word transfer?" instead of checking your per-transfer "use PIO?" hint)
Yup, good point, controller driver can be smarter
Thanks,
Feng
Quoted 6 lines
before kicking in whatever logic you
think would improve performance (by eliminating DMA setup and teardown costs, including cache cleaning).- Dave
RFC PATCH v2 1/1 serial: spi: add spi-uart driver for Maxim 3110
In reply to Feng TangHi All,
Here is a driver for Maxim 3110 SPI-UART device, please help to review.
It has been validated with Designware SPI controller (drivers/spi: dw_spi.c & dw_spi_pci.c). It supports polling and IRQ mode, supports batch read, and provides a console.
change since v1:
* Address comments from Alan Cox
* Use a "use_irq" module parameter to runtime check whether to use IRQ mode
* Add the suspend/resume implementation
Thanks!
Feng
Quoted 1 line
From
6d14c5d68cdae8d48b6d8a00b6653022f2b100d0Mon Sep 17 00:00:00 2001
From: Feng Tang <[email protected]>
Date: Mon, 8 Feb 2010 16:02:59 +0800
Subject: [PATCH] serial: spi: add spi-uart driver for Maxim 3110
This is the driver for Max3110 SPI-UART device, which connect to host with SPI interface. It supports baud rates from 300 to 230400, and supports both polling and IRQ mode, as well as providing a console for system use
Its datasheet could be found here:
http://datasheets.maxim-ic.com/en/ds/MAX3110E-MAX3111E.pdf
--- drivers/serial/Kconfig | 12 + drivers/serial/max3110.c | 848 +++++++++++++++++++++++++++++++++++++++++++ drivers/serial/max3110.h | 59 +++ include/linux/serial_core.h | 3 + 4 files changed, 922 insertions(+), 0 deletions(-) create mode 100644 drivers/serial/max3110.c create mode 100644 drivers/serial/max3110.h
drivers/serial/Kconfig +12 -0
@@ -688,6 +688,18 @@ config SERIAL_SA1100_CONSOLE your boot loader (lilo or loadlin) about how to pass options to the kernel at boot time.) +config SERIAL_MAX3110 + tristate "SPI UART driver for Max3110" + select SERIAL_CORE + select SERIAL_CORE_CONSOLE + help + This is the UART protocol driver for MAX3110 device + +config MAX3110_DESIGNWARE + boolean "Enable Max3110 to work with Designware controller" + default y + depends on SERIAL_MAX3110 + config SERIAL_BFIN tristate "Blackfin serial port support" depends on BLACKFIN
drivers/serial/max3110.c new file +848 -0
848 lines not on this page · raw message
drivers/serial/max3110.h new file +59 -0
59 lines not on this page · raw message
include/linux/serial_core.h +3 -0
@@ -182,6 +182,9 @@/* Aeroflex Gaisler GRLIB APBUART */ #define PORT_APBUART 90 +/* Maxim M3110 */ +#define PORT_MAX3110 91 +#ifdef __KERNEL__#include <linux/compiler.h>
Signature · 8 lines
1.6.3.3 ------------------------------------------------------------------------------ The Planet: dedicated and managed hosting, cloud storage, colocation Stay online with enterprise data centers and the best network in the business Choose flexible plans and management services without long-term contracts Personal 24x7 support from experience hosting pros just a phone call away. http://p.sf.net/sfu/theplanet-com