Qualcomm SSBI bus driver
Subsystems (4)
cover Qualcomm SSBI bus driver
This small series adds the Qualcomm SSBI bus driver. The original driver was developed as part of Android. Kenneth Heitke updated this, and sent it out a while back. This series updates this driver to use DeviceTree, and fixes a few things that have changed since it was originally sent out.
In addition, the 6th patch, the RFC, is a small test driver that reads the version register off of the pm8058 device typically connected to an MSM8660 or MSM8960. This is primarily useful to anyone wanting to work on these pmic drivers.
Updates to the PMIC drivers themselves to use this driver to come.
David
PATCH 1/6 platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
From: Kenneth Heitke <[email protected]>
SSBI is the Qualcomm single-wire serial bus interface used to connect the MSM devices to the PMIC and other devices.
Since SSBI only supports a single slave, the driver gets the name of the slave device passed in from the board file through the master device's platform data.
SSBI registers pretty early (postcore), so that the PMIC can come up before the board init. This is useful if the board init requires the use of gpios that are connected through the PMIC.
Based on a patch by Dima Zavin <[email protected]> that can be found at:
http://android.git.kernel.org/?p=kernel/msm.git;a=commitdiff;h=eb060bac4
This patch adds PMIC Arbiter support for the MSM8660. The PMIC Arbiter is a hardware wrapper around the SSBI 2.0 controller that is designed to overcome concurrency issues and security limitations. A controller_type field is added to the platform data to specify the type of the SSBI controller (1.0, 2.0, or PMIC Arbiter).
[davidb at codeaurora.org:
I've moved this driver into drivers/ssbi/ and added an include for linux/module.h so that it will compile]
--- drivers/Kconfig | 2 + drivers/Makefile | 1 + drivers/ssbi/Kconfig | 17 ++ drivers/ssbi/Makefile | 4 + drivers/ssbi/ssbi.c | 397 +++++++++++++++++++++++++++++++++++++++++++++++ include/linux/msm_ssbi.h | 49 ++++++ 6 files changed, 470 insertions(+) create mode 100644 drivers/ssbi/Kconfig create mode 100644 drivers/ssbi/Makefile create mode 100644 drivers/ssbi/ssbi.c create mode 100644 include/linux/msm_ssbi.h
drivers/Kconfig +2 -0
@@ -52,6 +52,8 @@ source "drivers/i2c/Kconfig" source "drivers/spi/Kconfig" +source "drivers/ssbi/Kconfig" + source "drivers/hsi/Kconfig" source "drivers/pps/Kconfig"
drivers/Makefile +1 -0
@@ -114,6 +114,7 @@ obj-y += firmware/ obj-$(CONFIG_CRYPTO) += crypto/ obj-$(CONFIG_SUPERH) += sh/ obj-$(CONFIG_ARCH_SHMOBILE) += sh/ +obj-$(CONFIG_MSM_SSBI) += ssbi/ ifndef CONFIG_ARCH_USES_GETTIMEOFFSET obj-y += clocksource/ endif
drivers/ssbi/Kconfig new file +17 -0
@@ -0,0 +1,17 @@ +# +# MSM SSBI bus support +# + +menu "Qualcomm MSM SSBI bus support" + depends on ARCH_MSM + +config MSM_SSBI + bool "Qualcomm Single-wire Serial Bus Interface (SSBI)" + help + If you say yes to this option, support will be included for the + built-in SSBI interface on Qualcomm MSM family processors. + + This is required for communicating with Qualcomm PMICs and + other devices that have the SSBI interface. + +endmenu
drivers/ssbi/Makefile new file +4 -0
@@ -0,0 +1,4 @@ +# +# Makefile for the MSM specific device drivers. +# +obj-$(CONFIG_MSM_SSBI) += ssbi.o
drivers/ssbi/ssbi.c new file +397 -0
397 lines not on this page · raw message
include/linux/msm_ssbi.h new file +49 -0
49 lines not on this page · raw message
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
Re: [PATCH 1/6] platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
drivers/ssbi/Kconfig line 6
Quoted 2 lines · David Brown
drivers/ssbi/Kconfig new file +2
-0+menu "Qualcomm MSM SSBI bus support" + depends on ARCH_MSM
Why?
drivers/ssbi/Kconfig line 9
Quoted 2 lines
drivers/ssbi/Kconfig new file +2
-0+config MSM_SSBI + bool "Qualcomm Single-wire Serial Bus Interface (SSBI)"
Why can't this be a module?
The multi-platform (or whatever it's called), requires that things be modules and not built in.
Quoted 14 lines
drivers/ssbi/Kconfig new file +8
-0+ help + If you say yes to this option, support will be included for the + built-in SSBI interface on Qualcomm MSM family processors. + + This is required for communicating with Qualcomm PMICs and + other devices that have the SSBI interface. + +endmenudrivers/ssbi/Makefile new file +0
-0
drivers/ssbi/Makefile line 2
Quoted 2 lines
drivers/ssbi/Makefile new file +2
-0@@ -0,0 +1,4 @@ +# +# Makefile for the MSM specific device drivers.
MSM?
No comment needed at all in this file :)
Quoted 61 lines
drivers/ssbi/ssbi.c new file +58
-0@@ -0,0 +1,397 @@ +/* Copyright (c) 2009-2011, Code Aurora Forum. All rights reserved. +* Copyright (c) 2010, Google Inc. +* +* Original authors: Code Aurora Forum +* +* Author: Dima Zavin <[email protected]> +* - Largely rewritten from original to not be an i2c driver. +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 and +* only version 2 as published by the Free Software Foundation. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. +*/ + +#define pr_fmt (fmt) "%s: " fmt, __func__ + +#include <linux/delay.h> +#include <linux/err.h> +#include <linux/io.h> +#include <linux/kernel.h> +#include <linux/platform_device.h> +#include <linux/slab.h> +#include <linux/msm_ssbi.h> +#include <linux/module.h> + +/* SSBI 2.0 controller registers */ +#define SSBI2_CMD 0x0008 +#define SSBI2_RD 0x0010 +#define SSBI2_STATUS 0x0014 +#define SSBI2_MODE2 0x001C + +/* SSBI_CMD fields */ +#define SSBI_CMD_RDWRN (1 << 24) + +/* SSBI_STATUS fields */ +#define SSBI_STATUS_RD_READY (1 << 2) +#define SSBI_STATUS_READY (1 << 1) +#define SSBI_STATUS_MCHN_BUSY (1 << 0) + +/* SSBI_MODE2 fields */ +#define SSBI_MODE2_REG_ADDR_15_8_SHFT 0x04 +#define SSBI_MODE2_REG_ADDR_15_8_MASK (0x7f << SSBI_MODE2_REG_ADDR_15_8_SHFT) + +#define SET_SSBI_MODE2_REG_ADDR_15_8 (MD ,AD ) \ + (((MD) & 0x0F) | ((((AD) >> 8) << SSBI_MODE2_REG_ADDR_15_8_SHFT) & \ + SSBI_MODE2_REG_ADDR_15_8_MASK)) + +/* SSBI PMIC Arbiter command registers */ +#define SSBI_PA_CMD 0x0000 +#define SSBI_PA_RD_STATUS 0x0004 + +/* SSBI_PA_CMD fields */ +#define SSBI_PA_CMD_RDWRN (1 << 24) +#define SSBI_PA_CMD_ADDR_MASK 0x7fff/* REG_ADDR_7_0, REG_ADDR_8_14*/
drivers/ssbi/ssbi.c line 68
Quoted 10 lines
drivers/ssbi/ssbi.c new file +10
-0+ +/* SSBI_PA_RD_STATUS fields */ +#define SSBI_PA_RD_STATUS_TRANS_DONE (1 << 27) +#define SSBI_PA_RD_STATUS_TRANS_DENIED (1 << 26) + +#define SSBI_TIMEOUT_US 100 + +struct msm_ssbi { +struct device *dev; +struct device *slave;
Really? Two pointers to devices? Why? Shouldn't this have a struct device embedded in it instead of the dev member being a pointer?
drivers/ssbi/ssbi.c line 76
Quoted 8 lines
drivers/ssbi/ssbi.c new file +8
-0+void __iomem *base; +spinlock_t lock; +enum msm_ssbi_controller_type controller_type; +int (*read)(struct msm_ssbi *,u16 addr,u8 *buf,int len); +int (*write)(struct msm_ssbi *,u16 addr,u8 *buf,int len); +}; + +#define to_msm_ssbi (dev) platform_get_drvdata(to_platform_device(dev))
That doesn't work for the above structure, right?
drivers/ssbi/ssbi.c line 86
Quoted 9 lines
drivers/ssbi/ssbi.c new file +9
-0+static inline u32 ssbi_readl (struct msm_ssbi *ssbi,u32 reg) +{ +return readl (ssbi->base + reg); +} + +static inline void ssbi_writel (struct msm_ssbi *ssbi,u32 val,u32 reg) +{ +writel (val, ssbi->base + reg); +}
Did you run sparse on this file?
drivers/ssbi/ssbi.c line 97
Quoted 10 lines
drivers/ssbi/ssbi.c new file +10
-0+static int ssbi_wait_mask (struct msm_ssbi *ssbi,u32 set_mask,u32 clr_mask) +{ +u32 timeout =SSBI_TIMEOUT_US ; +u32 val; + +while (timeout--) { + val =ssbi_readl (ssbi,SSBI2_STATUS ); +if (((val & set_mask) == set_mask) && ((val & clr_mask) ==0 )) +return 0 ; +udelay (1 );
Busy loop? Really?
drivers/ssbi/ssbi.c line 101
Quoted 4 lines
drivers/ssbi/ssbi.c new file +4
-0+ } + +dev_err (ssbi->dev,"%s: timeout (status %x set_mask %x clr_mask %x)\n" , + __func__,ssbi_readl (ssbi,SSBI2_STATUS ), set_mask, clr_mask);
Why polute the log with this? What can a user do with it?
Quoted 32 lines
drivers/ssbi/ssbi.c new file +32
-0+return -ETIMEDOUT ; +} + +static int +msm_ssbi_read_bytes (struct msm_ssbi *ssbi,u16 addr,u8 *buf,int len) +{ +u32 cmd =SSBI_CMD_RDWRN | ((addr &0xff ) <<16 ); +int ret =0 ; + +if (ssbi->controller_type ==MSM_SBI_CTRL_SSBI2 ) { +u32 mode2 =ssbi_readl (ssbi,SSBI2_MODE2 ); + mode2 =SET_SSBI_MODE2_REG_ADDR_15_8 (mode2, addr); +ssbi_writel (ssbi, mode2,SSBI2_MODE2 ); + } + +while (len) { + ret =ssbi_wait_mask (ssbi,SSBI_STATUS_READY ,0 ); +if (ret) + goto err; + +ssbi_writel (ssbi, cmd,SSBI2_CMD ); + ret =ssbi_wait_mask (ssbi,SSBI_STATUS_RD_READY ,0 ); +if (ret) + goto err; + *buf++ =ssbi_readl (ssbi,SSBI2_RD ) &0xff ; + len--; + } + +err: +return ret; +} +
drivers/ssbi/ssbi.c line 143
Quoted 10 lines
drivers/ssbi/ssbi.c new file +10
-0+static int +msm_ssbi_write_bytes (struct msm_ssbi *ssbi,u16 addr,u8 *buf,int len) +{ +int ret =0 ; + +if (ssbi->controller_type ==MSM_SBI_CTRL_SSBI2 ) { +u32 mode2 =ssbi_readl (ssbi,SSBI2_MODE2 ); + mode2 =SET_SSBI_MODE2_REG_ADDR_15_8 (mode2, addr); +ssbi_writel (ssbi, mode2,SSBI2_MODE2 ); + }
Perhaps have a controller type write function that can handle this type of stuff instead of putting it in the "generic" write path?
Quoted 30 lines
drivers/ssbi/ssbi.c new file +30
-0+while (len) { + ret =ssbi_wait_mask (ssbi,SSBI_STATUS_READY ,0 ); +if (ret) + goto err; + +ssbi_writel (ssbi, ((addr &0xff ) <<16 ) | *buf,SSBI2_CMD ); + ret =ssbi_wait_mask (ssbi,0 ,SSBI_STATUS_MCHN_BUSY ); +if (ret) + goto err; + buf++; + len--; + } + +err: +return ret; +} + +static inline int +msm_ssbi_pa_transfer (struct msm_ssbi *ssbi,u32 cmd,u8 *data) +{ +u32 timeout =SSBI_TIMEOUT_US ; +u32 rd_status =0 ; + +ssbi_writel (ssbi, cmd,SSBI_PA_CMD ); + +while (timeout--) { + rd_status =ssbi_readl (ssbi,SSBI_PA_RD_STATUS ); + +if (rd_status &SSBI_PA_RD_STATUS_TRANS_DENIED ) { + dev_err(ssbi->dev, "%s: transaction denied (0x%x)\n",
drivers/ssbi/ssbi.c line 184
Quoted 10 lines
drivers/ssbi/ssbi.c new file +10
-0+ __func__, rd_status); +return -EPERM ; + } + +if (rd_status &SSBI_PA_RD_STATUS_TRANS_DONE ) { +if (data) + *data = rd_status &0xff ; +return 0 ; + } +udelay (1 );
Busy loop again?
drivers/ssbi/ssbi.c line 187
Quoted 3 lines
drivers/ssbi/ssbi.c new file +3
-0+ } + +dev_err (ssbi->dev,"%s: timeout, status 0x%x\n" , __func__, rd_status);
Don't polute the logs unless you want the user to do something with the information.
Quoted 48 lines
drivers/ssbi/ssbi.c new file +48
-0+return -ETIMEDOUT ; +} + +static int +msm_ssbi_pa_read_bytes (struct msm_ssbi *ssbi,u16 addr,u8 *buf,int len) +{ +u32 cmd; +int ret =0 ; + + cmd =SSBI_PA_CMD_RDWRN | (addr &SSBI_PA_CMD_ADDR_MASK ) <<8 ; + +while (len) { + ret =msm_ssbi_pa_transfer (ssbi, cmd, buf); +if (ret) + goto err; + buf++; + len--; + } + +err: +return ret; +} + +static int +msm_ssbi_pa_write_bytes (struct msm_ssbi *ssbi,u16 addr,u8 *buf,int len) +{ +u32 cmd; +int ret =0 ; + +while (len) { + cmd = (addr &SSBI_PA_CMD_ADDR_MASK ) <<8 | *buf; + ret =msm_ssbi_pa_transfer (ssbi, cmd,NULL ); +if (ret) + goto err; + buf++; + len--; + } + +err: +return ret; +} + +int msm_ssbi_read (struct device *dev,u16 addr,u8 *buf,int len) +{ +struct msm_ssbi *ssbi =to_msm_ssbi (dev); +unsigned long flags; +int ret; +
drivers/ssbi/ssbi.c line 245
Quoted 10 lines
drivers/ssbi/ssbi.c new file +10
-0+if (ssbi->dev != dev) +return -ENXIO ; + +spin_lock_irqsave (&ssbi->lock, flags); + ret = ssbi->read (ssbi, addr, buf, len); +spin_unlock_irqrestore (&ssbi->lock, flags); + +return ret; +} +EXPORT_SYMBOL (msm_ssbi_read);
msm_*? Why not just ssbi_*?
EXPORT_SYMBOL_GPL()?
drivers/ssbi/ssbi.c line 375
Quoted 3 lines
drivers/ssbi/ssbi.c new file +3
-0+static struct platform_driver msm_ssbi_driver = { + .probe = msm_ssbi_probe, + .remove =__exit_p (msm_ssbi_remove),
You just oopsed if someone unbound your device from the driver from userspace. Not good.
thanks,
greg k-h
Re: [PATCH 1/6] platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
Quoted 5 lines · Greg Kroah-Hartman
drivers/ssbi/Kconfig line 6
Quoted 2 lines · David Brown
drivers/ssbi/Kconfig new file +2
-0+menu "Qualcomm MSM SSBI bus support" + depends on ARCH_MSMWhy?
In the sense that ARCH_MSM are the only devices that ever were, or ever will be made with this device. It doesn't strictly depend on it, but do we want to clutter the config for everything else.
Quoted 4 lines
drivers/ssbi/Kconfig line 9
Quoted 2 lines
drivers/ssbi/Kconfig new file +2
-0+config MSM_SSBI + bool "Qualcomm Single-wire Serial Bus Interface (SSBI)"Why can't this be a module?
Without this device, the PMIC drivers won't work, regulators can't be turned on, and most of the other devices won't work. I can try if it can still be made a module, and it might be good at exercising the deferred probe mechanism.
So, a deeper question. I've resent this driver with minimal change. I've also made some other changes as patches afterwards. Do we want these squashed into a single patch, or the initial one (not authored by me) followed by updates?
Quoted 6 lines
drivers/ssbi/Makefile line 2
Quoted 2 lines
drivers/ssbi/Makefile new file +2
-0+# +# Makefile for the MSM specific device drivers.MSM?
No comment needed at all in this file :)
Thanks, missed this. I think the original patch was using platform/msm, and the minimalist comment almost made sense in that context.
Quoted 6 lines
drivers/ssbi/ssbi.c line 68
Quoted 3 lines
drivers/ssbi/ssbi.c new file +3
-0+struct msm_ssbi { +struct device *dev; +struct device *slave;Really? Two pointers to devices? Why? Shouldn't this have a struct device embedded in it instead of the dev member being a pointer?
I'll go through this more thoroughly and organize things better.
Quoted 4 lines
Quoted 4 lines
drivers/ssbi/ssbi.c new file +4
-0+static int ssbi_wait_mask (struct msm_ssbi *ssbi,u32 set_mask,u32 clr_mask) +{ +u32 timeout =SSBI_TIMEOUT_US ; +u32 val;
Quoted 8 lines
drivers/ssbi/ssbi.c line 97
Quoted 6 lines
drivers/ssbi/ssbi.c new file +6
-0+ +while (timeout--) { + val =ssbi_readl (ssbi,SSBI2_STATUS ); +if (((val & set_mask) == set_mask) && ((val & clr_mask) ==0 )) +return 0 ; +udelay (1 );Busy loop? Really?
I'm not sure what else to do here. The controller is only slightly more than a bit-banged bus. I think the reason for the longer possibly delay is because there is an arbiter (the PMIC is shared with the radio CPU). I'll investigate further.
Quoted 6 lines
drivers/ssbi/ssbi.c line 101
Quoted 4 lines
drivers/ssbi/ssbi.c new file +4
-0+ } + +dev_err (ssbi->dev,"%s: timeout (status %x set_mask %x clr_mask %x)\n" , + __func__,ssbi_readl (ssbi,SSBI2_STATUS ), set_mask, clr_mask);Why polute the log with this? What can a user do with it?
Nothing in fact, other than possibly learn that things, indeed aren't working. I'll take it and the others out.
Quoted 5 lines
Quoted 5 lines
drivers/ssbi/ssbi.c new file +5
-0+static int +msm_ssbi_write_bytes (struct msm_ssbi *ssbi,u16 addr,u8 *buf,int len) +{ +int ret =0 ; +
Quoted 8 lines
drivers/ssbi/ssbi.c line 115
Quoted 5 lines
drivers/ssbi/ssbi.c new file +5
-0+if (ssbi->controller_type ==MSM_SBI_CTRL_SSBI2 ) { +u32 mode2 =ssbi_readl (ssbi,SSBI2_MODE2 ); + mode2 =SET_SSBI_MODE2_REG_ADDR_15_8 (mode2, addr); +ssbi_writel (ssbi, mode2,SSBI2_MODE2 ); + }Perhaps have a controller type write function that can handle this type of stuff instead of putting it in the "generic" write path?
Yes, that's a better approach.
Quoted 3 lines
Quoted 1 line
+EXPORT_SYMBOL(msm_ssbi_read);
msm_*? Why not just ssbi_*?
I'm fine with ssbi. It is very MSM specific, though.
Quoted 1 line
EXPORT_SYMBOL_GPL()?
Yes, it's definitely a kernel internal API.
Quoted 6 lines
drivers/ssbi/ssbi.c line 375
Quoted 3 lines
drivers/ssbi/ssbi.c new file +3
-0+static struct platform_driver msm_ssbi_driver = { + .probe = msm_ssbi_probe, + .remove =__exit_p (msm_ssbi_remove),You just oopsed if someone unbound your device from the driver from userspace. Not good.
I did catch this one in a later patch :-) I can clean it up in the driver, though, since it looks like some more work needs to go into this.
Thanks,
David
Signature · 3 lines
sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
Re: [PATCH 1/6] platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
Quoted 11 lines · David Brown
Quoted 5 lines · Greg Kroah-Hartman
drivers/ssbi/Kconfig line 6
Quoted 2 lines · David Brown
drivers/ssbi/Kconfig new file +2
-0+menu "Qualcomm MSM SSBI bus support" + depends on ARCH_MSMWhy?
In the sense that
ARCH_MSMare the only devices that ever were, or ever will be made with this device. It doesn't strictly depend on it, but do we want to clutter the config for everything else.
It's not "clutter". You want me to build this on other platforms, to catch api and other types of build breakages. This is the way for almost all Linux drivers.
Quoted 9 lines
Quoted 4 lines
drivers/ssbi/Kconfig line 9
Quoted 2 lines
drivers/ssbi/Kconfig new file +2
-0+config MSM_SSBI + bool "Qualcomm Single-wire Serial Bus Interface (SSBI)"Why can't this be a module?
Without this device, the PMIC drivers won't work, regulators can't be turned on, and most of the other devices won't work. I can try if it can still be made a module, and it might be good at exercising the deferred probe mechanism.
If the PMIC drivers are dependant on the symbols in this module, there should not be any problems, right?
Quoted 4 lines
So, a deeper question. I've resent this driver with minimal change. I've also made some other changes as patches afterwards. Do we want these squashed into a single patch, or the initial one (not authored by me) followed by updates?
To preserve the authorship, you might want to fix this stuff with follow-on patches. As long as the original patch can build properly.
Quoted 10 lines
Quoted 6 lines
drivers/ssbi/ssbi.c line 375
Quoted 3 lines
drivers/ssbi/ssbi.c new file +3
-0+static struct platform_driver msm_ssbi_driver = { + .probe = msm_ssbi_probe, + .remove =__exit_p (msm_ssbi_remove),You just oopsed if someone unbound your device from the driver from userspace. Not good.
I did catch this one in a later patch :-) I can clean it up in the driver, though, since it looks like some more work needs to go into this.
Yes, but it's very close, fix this all up and resend your series and all should be fine.
Nice job,
greg k-h
Re: [PATCH 1/6] platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
Quoted 9 lines · Greg Kroah-Hartman
Quoted 8 lines · David Brown
Quoted 5 lines · Greg Kroah-Hartman
drivers/ssbi/Kconfig line 6
Quoted 2 lines · David Brown
drivers/ssbi/Kconfig new file +2
-0+menu "Qualcomm MSM SSBI bus support" + depends on ARCH_MSMWhy?
Quoted 7 lines
Quoted 3 lines
In the sense that
ARCH_MSMare the only devices that ever were, or ever will be made with this device. It doesn't strictly depend on it, but do we want to clutter the config for everything else.It's not "clutter". You want me to build this on other platforms, to catch api and other types of build breakages. This is the way for almost all Linux drivers.
Not having a depends on helps build coverage, but I have seen objections to showing up irrelevant configurations to users (of x86 for example). See one here from Linus for OMAP_OCP2SCP
http://lkml.indiana.edu/hypermail/linux/kernel/1210.0/00785.html
If this case is different, I am not sure why.
Thanks,
Sekhar
Re: [PATCH 1/6] platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
Quoted 18 lines · Sekhar Nori
Quoted 9 lines · Greg Kroah-Hartman
Quoted 8 lines · David Brown
Quoted 5 lines · Greg Kroah-Hartman
drivers/ssbi/Kconfig line 6
Quoted 2 lines · David Brown
drivers/ssbi/Kconfig new file +2
-0+menu "Qualcomm MSM SSBI bus support" + depends on ARCH_MSMWhy?
Quoted 7 lines
Quoted 3 lines
In the sense that
ARCH_MSMare the only devices that ever were, or ever will be made with this device. It doesn't strictly depend on it, but do we want to clutter the config for everything else.It's not "clutter". You want me to build this on other platforms, to catch api and other types of build breakages. This is the way for almost all Linux drivers.
Quoted 7 lines
Not having a depends on helps build coverage, but I have seen objections to showing up irrelevant configurations to users (of x86 for example). See one here from Linus for
OMAP_OCP2SCPhttp://lkml.indiana.edu/hypermail/linux/kernel/1210.0/00785.html
If this case is different, I am not sure why.
This was, in fact, the reason I put the dependency there. I found it a little annoying being asked about a bunch of OMAP devices when building the x86 kernel. At least they weren't cancer curing options (default y).
David
Signature · 3 lines
sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
Re: [PATCH 1/6] platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
In reply to Greg Kroah-HartmanQuoted 7 lines · Greg Kroah-Hartman
Quoted 4 lines
Without this device, the PMIC drivers won't work, regulators can't be turned on, and most of the other devices won't work. I can try if it can still be made a module, and it might be good at exercising the deferred probe mechanism.
If the PMIC drivers are dependant on the symbols in this module, there should not be any problems, right?
The PMIC drivers directly will be, but the users of gpios/irqs and regulators coming from the PMIC will only depend on the generic APIs for this. In theory, the deferred probe should handle this, and it might be useful for testing.
I can allow it to be a module, although I don't picture that really being a useful configuration.
David
Signature · 3 lines
sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
Re: [PATCH 1/6] platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
Quoted 10 lines · David Brown
Quoted 7 lines · Greg Kroah-Hartman
Quoted 4 lines
Without this device, the PMIC drivers won't work, regulators can't be turned on, and most of the other devices won't work. I can try if it can still be made a module, and it might be good at exercising the deferred probe mechanism.
If the PMIC drivers are dependant on the symbols in this module, there should not be any problems, right?
Quoted 7 lines
The PMIC drivers directly will be, but the users of gpios/irqs and regulators coming from the PMIC will only depend on the generic APIs for this. In theory, the deferred probe should handle this, and it might be useful for testing.
I can allow it to be a module, although I don't picture that really being a useful configuration.
It will allow this to get multi-arch build testing at the very least, a very valuable thing as time goes on. You can always specify this as 'Y' in your defconfig for your board.
thanks,
greg k-h
Re: [PATCH 1/6] platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
In reply to Greg Kroah-HartmanQuoted 4 lines · Greg Kroah-Hartman
Quoted 4 lines
drivers/ssbi/ssbi.c new file +4
-0+static int ssbi_wait_mask (struct msm_ssbi *ssbi,u32 set_mask,u32 clr_mask) +{ +u32 timeout =SSBI_TIMEOUT_US ; +u32 val;
Quoted 8 lines
drivers/ssbi/ssbi.c line 97
Quoted 6 lines
drivers/ssbi/ssbi.c new file +6
-0+ +while (timeout--) { + val =ssbi_readl (ssbi,SSBI2_STATUS ); +if (((val & set_mask) == set_mask) && ((val & clr_mask) ==0 )) +return 0 ; +udelay (1 );Busy loop? Really?
Finally was able to dig up some of the reason for this. The transactions typically take about 5us. In the case of contention with another CPU, it could take as much as 20us.
Would it be sufficient to just explain this in a comment?
David
Re: [PATCH 1/6] platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
Quoted 15 lines · David Brown
Quoted 4 lines · Greg Kroah-Hartman
Quoted 4 lines
drivers/ssbi/ssbi.c new file +4
-0+static int ssbi_wait_mask (struct msm_ssbi *ssbi,u32 set_mask,u32 clr_mask) +{ +u32 timeout =SSBI_TIMEOUT_US ; +u32 val;Quoted 8 lines
drivers/ssbi/ssbi.c line 97
Quoted 6 lines
drivers/ssbi/ssbi.c new file +6
-0+ +while (timeout--) { + val =ssbi_readl (ssbi,SSBI2_STATUS ); +if (((val & set_mask) == set_mask) && ((val & clr_mask) ==0 )) +return 0 ; +udelay (1 );Busy loop? Really?
Quoted 5 lines
Finally was able to dig up some of the reason for this. The transactions typically take about 5us. In the case of contention with another CPU, it could take as much as 20us.
Would it be sufficient to just explain this in a comment?
That would be good to do, especially if it turns out to be a longer delay and people start to wonder why their system load is increasing for no noticeable reason.
thanks,
greg k-h
PATCH 2/6 SSBI: Convert SSBI to device tree
In reply to David BrownThe SSBI bus is exclusive to the Qualcomm MSM targets, and all SoCs using it will be using device tree. Convert this driver to indentify with device tree.
This makes the bus probing a good bit simpler, since the attaching of child nodes can be represented directly in the devicetree, rather than having to be inferred by name.
--- Documentation/devicetree/bindings/arm/msm/ssbi.txt | 18 +++++ arch/arm/boot/dts/msm8660-surf.dts | 6 ++ arch/arm/boot/dts/msm8960-cdp.dts | 6 ++ drivers/ssbi/ssbi.c | 81 +++++++++------------- 4 files changed, 62 insertions(+), 49 deletions(-) create mode 100644 Documentation/devicetree/bindings/arm/msm/ssbi.txt
Documentation/devicetree/bindings/arm/msm/ssbi.txt new file +18 -0
@@ -0,0 +1,18 @@ +* Qualcomm SSBI + +Some Qualcomm MSM devices contain a point-to-point serial bus used to +communicate with a limited range of devices (mostly power management +chips). + +These require the following properties: + +- compatible: "qcom,ssbi" + +- qcom,controller-type + indicates the SSBI bus variant the controller should use to talk + with the slave device. This should be one of "ssbi", "ssbi2", or + "pmic-arbiter". The type chosen is determined by the attached + slave. + +The slave device should be the single child node of the ssbi device +with a compatible field.
arch/arm/boot/dts/msm8660-surf.dts +6 -0
@@ -38,4 +38,10 @@ <0x19c00000 0x1000>; interrupts = <0 195 0x0>; }; + + qcom,ssbi at 500000 { + compatible = "qcom,ssbi"; + reg = <0x500000 0x1000>; + qcom,controller-type = "pmic-arbiter"; + }; };
arch/arm/boot/dts/msm8960-cdp.dts +6 -0
@@ -38,4 +38,10 @@ <0x16400000 0x1000>; interrupts = <0 154 0x0>; }; + + qcom,ssbi at 500000 { + compatible = "qcom,ssbi"; + reg = <0x500000 0x1000>; + qcom,controller-type = "pmic-arbiter"; + }; };
drivers/ssbi/ssbi.c +32 -49
126 lines not on this page · raw message
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH 3/6 ssbi: Fix exit mismatch in remove function
In reply to David Brownmsm_ssbi_remove is referenced with __exit_p, but not declared with __exit. This causes a warning when the driver is not built as a module:
drivers/ssbi/ssbi.c:341:23: warning: 'msm_ssbi_remove' defined but not used [-Wunused-function]
Fix by adding the __exit declaration to the function.
--- drivers/ssbi/ssbi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/ssbi/ssbi.c +1 -1
@@ -337,7 +337,7 @@ err_get_mem_res:return ret; } -static int msm_ssbi_remove (struct platform_device *pdev) +static int __exit msm_ssbi_remove(struct platform_device *pdev) {struct msm_ssbi *ssbi =platform_get_drvdata (pdev);
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
Re: [PATCH 3/6] ssbi: Fix exit mismatch in remove function
Quoted 18 lines · David Brown
msm_ssbi_removeis referenced with__exit_p, but not declared with__exit. This causes a warning when the driver is not built as a module:
drivers/ssbi/ssbi.c:341:23: warning: 'msm_ssbi_remove' defined but not used [-Wunused-function]Fix by adding the
__exitdeclaration to the function.Signed-off-by David Brown <[email protected]>--- drivers/ssbi/ssbi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/ssbi/ssbi.c +0
-0
drivers/ssbi/ssbi.c err_get_mem_res: · line 340
Quoted 5 lines
drivers/ssbi/ssbi.c +1
-1@@ -337,7 +337,7 @@ err_get_mem_res:return ret; } -static int msm_ssbi_remove (struct platform_device *pdev) +static int __exit msm_ssbi_remove(struct platform_device *pdev)
No, remove the __exit_p marking instead, unless you want your kernel to be oopsed :)
thanks,
greg k-h
Re: [PATCH 3/6] ssbi: Fix exit mismatch in remove function
Quoted 22 lines · Greg Kroah-Hartman
Quoted 21 lines · David Brown
msm_ssbi_removeis referenced with__exit_p, but not declared with__exit. This causes a warning when the driver is not built as a module:
drivers/ssbi/ssbi.c:341:23: warning: 'msm_ssbi_remove' defined but not used [-Wunused-function]Fix by adding the
__exitdeclaration to the function.Signed-off-by David Brown <[email protected]>--- drivers/ssbi/ssbi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/ssbi/ssbi.c +0
-0@@ -337,7 +337,7 @@ err_get_mem_res:return ret; }
Quoted 5 lines
drivers/ssbi/ssbi.c err_get_mem_res: · line 340
Quoted 2 lines
drivers/ssbi/ssbi.c +1
-1-static int msm_ssbi_remove (struct platform_device *pdev) +static int __exit msm_ssbi_remove(struct platform_device *pdev)No, remove the
__exit_pmarking instead, unless you want your kernel to be oopsed :)
Thanks. Oopsing is not fun.
David
Signature · 3 lines
sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH 4/6 ssbi: Use regular init level
In reply to David BrownWith device tree, and deferred probe, it is no longer necessary to make sure that the ssbi bus driver is initialized very early. Restore to a regular module_init().
--- drivers/ssbi/ssbi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/ssbi/ssbi.c +1 -1
@@ -366,7 +366,7 @@ static int __init msm_ssbi_init(void) {return platform_driver_register (&msm_ssbi_driver); } -postcore_initcall (msm_ssbi_init); +module_init (msm_ssbi_init); static void __exit msm_ssbi_exit(void) {
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH 5/6 ARM: msm: enable SSBI driver in defconfig
In reply to David Brown--- arch/arm/configs/msm_defconfig | 1 + 1 file changed, 1 insertion(+)
arch/arm/configs/msm_defconfig +1 -0
@@ -84,6 +84,7 @@ CONFIG_HW_RANDOM=y CONFIG_I2C=y CONFIG_I2C_CHARDEV=y CONFIG_SPI=y +CONFIG_MSM_SSBI=y CONFIG_DEBUG_GPIO=y CONFIG_GPIO_SYSFS=y CONFIG_POWER_SUPPLY=y
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH 6/6 RFC: SSBI: Simple pm8058 test driver
In reply to David BrownA very small ssbi device driver that reads the pm8058 version register ;and prints it out.
This isn't really a useful driver, hence the RFC, but it is useful for someone wanting to test the SSBI driver itself. Once the PMIC drivers are finished, this isn't useful at all.
--- arch/arm/boot/dts/msm8660-surf.dts | 4 +++ arch/arm/boot/dts/msm8960-cdp.dts | 4 +++ drivers/ssbi/Makefile | 1 + drivers/ssbi/ssbi-test.c | 61 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 70 insertions(+) create mode 100644 drivers/ssbi/ssbi-test.c
arch/arm/boot/dts/msm8660-surf.dts +4 -0
@@ -43,5 +43,9 @@ compatible = "qcom,ssbi"; reg = <0x500000 0x1000>; qcom,controller-type = "pmic-arbiter"; + + qcom,ssbi-test { + compatible = "qcom,ssbi-test"; + }; }; };
arch/arm/boot/dts/msm8960-cdp.dts +4 -0
@@ -43,5 +43,9 @@ compatible = "qcom,ssbi"; reg = <0x500000 0x1000>; qcom,controller-type = "pmic-arbiter"; + + qcom,ssbi-test { + compatible = "qcom,ssbi-test"; + }; }; };
drivers/ssbi/Makefile +1 -0
@@ -2,3 +2,4 @@ # Makefile for the MSM specific device drivers. # obj-$(CONFIG_MSM_SSBI) += ssbi.o +obj-$(CONFIG_MSM_SSBI) += ssbi-test.o
drivers/ssbi/ssbi-test.c new file +61 -0
61 lines not on this page · raw message
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
v2 cover Qualcomm SSBI bus driver
In reply to David BrownThis small series adds the Qualcomm SSBI bus driver. The original driver was developed as part of Android. Kenneth Heitke updated this, and sent it out a while back. This series updates this driver to use DeviceTree, and fixes a few things that have changed since it was originally sent out.
In addition, the 11th patch, the RFC, is a small test driver that reads the version register off of the pm8058 device typically connected to an MSM8660 or MSM8960. This is primarily useful to anyone wanting to work on these pmic drivers.
Updates to the PMIC drivers themselves to use this driver to come.
Patch version:
v2: Updates from mailing list feedback:
- Use EXPORT_SYMBOL_GPL
- Proper fix for driver remove section mismatch
- Allow compilation as a module
- Remove extraneous dev_err() messages
- Comment use of busywait loop
- Update MAINTAINERS file
PATCH v2 1/11 platform-drivers: msm: add single-wire serial bus interface (SSBI) driver
From: Kenneth Heitke <[email protected]>
SSBI is the Qualcomm single-wire serial bus interface used to connect the MSM devices to the PMIC and other devices.
Since SSBI only supports a single slave, the driver gets the name of the slave device passed in from the board file through the master device's platform data.
SSBI registers pretty early (postcore), so that the PMIC can come up before the board init. This is useful if the board init requires the use of gpios that are connected through the PMIC.
Based on a patch by Dima Zavin <[email protected]> that can be found at:
http://android.git.kernel.org/?p=kernel/msm.git;a=commitdiff;h=eb060bac4
This patch adds PMIC Arbiter support for the MSM8660. The PMIC Arbiter is a hardware wrapper around the SSBI 2.0 controller that is designed to overcome concurrency issues and security limitations. A controller_type field is added to the platform data to specify the type of the SSBI controller (1.0, 2.0, or PMIC Arbiter).
[davidb at codeaurora.org:
I've moved this driver into drivers/ssbi/ and added an include for linux/module.h so that it will compile]
--- drivers/Kconfig | 2 + drivers/Makefile | 1 + drivers/ssbi/Kconfig | 16 ++ drivers/ssbi/Makefile | 1 + drivers/ssbi/ssbi.c | 397 +++++++++++++++++++++++++++++++++++++++++++++++ include/linux/msm_ssbi.h | 49 ++++++ 6 files changed, 466 insertions(+) create mode 100644 drivers/ssbi/Kconfig create mode 100644 drivers/ssbi/Makefile create mode 100644 drivers/ssbi/ssbi.c create mode 100644 include/linux/msm_ssbi.h
drivers/Kconfig +2 -0
@@ -52,6 +52,8 @@ source "drivers/i2c/Kconfig" source "drivers/spi/Kconfig" +source "drivers/ssbi/Kconfig" + source "drivers/hsi/Kconfig" source "drivers/pps/Kconfig"
drivers/Makefile +1 -0
@@ -114,6 +114,7 @@ obj-y += firmware/ obj-$(CONFIG_CRYPTO) += crypto/ obj-$(CONFIG_SUPERH) += sh/ obj-$(CONFIG_ARCH_SHMOBILE) += sh/ +obj-$(CONFIG_MSM_SSBI) += ssbi/ ifndef CONFIG_ARCH_USES_GETTIMEOFFSET obj-y += clocksource/ endif
drivers/ssbi/Kconfig new file +16 -0
@@ -0,0 +1,16 @@ +# +# MSM SSBI bus support +# + +menu "Qualcomm MSM SSBI bus support" + +config MSM_SSBI + bool "Qualcomm Single-wire Serial Bus Interface (SSBI)" + help + If you say yes to this option, support will be included for the + built-in SSBI interface on Qualcomm MSM family processors. + + This is required for communicating with Qualcomm PMICs and + other devices that have the SSBI interface. + +endmenu
drivers/ssbi/Makefile new file +1 -0
@@ -0,0 +1 @@ +obj-$(CONFIG_MSM_SSBI) += ssbi.o
drivers/ssbi/ssbi.c new file +397 -0
397 lines not on this page · raw message
include/linux/msm_ssbi.h new file +49 -0
49 lines not on this page · raw message
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH v2 2/11 fix: Use EXPORT_SYMBOL_GPL
In reply to David Brown--- drivers/ssbi/ssbi.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/ssbi/ssbi.c +2 -2
@@ -242,7 +242,7 @@ int msm_ssbi_read(struct device *dev, u16 addr, u8 *buf, int len)return ret; } -EXPORT_SYMBOL (msm_ssbi_read); +EXPORT_SYMBOL_GPL (msm_ssbi_read);int msm_ssbi_write (struct device *dev,u16 addr,u8 *buf,int len) {
@@ -259,7 +259,7 @@ int msm_ssbi_write(struct device *dev, u16 addr, u8 *buf, int len)return ret; } -EXPORT_SYMBOL (msm_ssbi_write); +EXPORT_SYMBOL_GPL (msm_ssbi_write);static int msm_ssbi_add_slave (struct msm_ssbi *ssbi,const struct msm_ssbi_slave_info *slave)
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH v2 3/11 ssbi: Fix exit mismatch in remove function
In reply to David Brownmsm_ssbi_remove is referenced with __exit_p, but not declared with __exit. This causes a warning when the driver is not built as a module:
drivers/ssbi/ssbi.c:341:23: warning: 'msm_ssbi_remove' defined but not used [-Wunused-function]
The remove is needed for unbinding to work, even if not compiled as a module, so just remove it.
--- drivers/ssbi/ssbi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/ssbi/ssbi.c +1 -1
@@ -372,7 +372,7 @@ static int msm_ssbi_remove(struct platform_device *pdev)static struct platform_driver msm_ssbi_driver = { .probe = msm_ssbi_probe, - .remove =__exit_p (msm_ssbi_remove), + .remove = msm_ssbi_remove, .driver = { .name ="msm_ssbi" , .owner =THIS_MODULE ,
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH v2 4/11 ssbi: Allow compilation as a module
In reply to David BrownThe ssbi driver's read/write entry points are protected with wrappers in the case when the driver isn't enabled. These wrappers don't make any sense, since a client of the SSBI bus won't work without it. Make these just regular functions, so that the SSBI driver can be built as a module.
--- drivers/ssbi/Kconfig | 2 +- include/linux/msm_ssbi.h | 11 ----------- 2 files changed, 1 insertion(+), 12 deletions(-)
drivers/ssbi/Kconfig +1 -1
@@ -5,7 +5,7 @@ menu "Qualcomm MSM SSBI bus support" config MSM_SSBI - bool "Qualcomm Single-wire Serial Bus Interface (SSBI)" + tristate "Qualcomm Single-wire Serial Bus Interface (SSBI)" help If you say yes to this option, support will be included for the built-in SSBI interface on Qualcomm MSM family processors.
include/linux/msm_ssbi.h +0 -11
@@ -33,17 +33,6 @@ struct msm_ssbi_platform_data { enum msm_ssbi_controller_type controller_type; }; -#ifdef CONFIG_MSM_SSBI int msm_ssbi_write (struct device *dev,u16 addr,u8 *buf,int len);int msm_ssbi_read (struct device *dev,u16 addr,u8 *buf,int len); -#else -static inline int msm_ssbi_write (struct device *dev,u16 addr,u8 *buf,int len) -{ -return -ENXIO ; -} -static inline int msm_ssbi_read (struct device *dev,u16 addr,u8 *buf,int len) -{ -return -ENXIO ; -} -#endif #endif
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH v2 5/11 SSBI: Convert SSBI to device tree
In reply to David BrownThe SSBI bus is exclusive to the Qualcomm MSM targets, and all SoCs using it will be using device tree. Convert this driver to indentify with device tree.
This makes the bus probing a good bit simpler, since the attaching of child nodes can be represented directly in the devicetree, rather than having to be inferred by name.
--- Documentation/devicetree/bindings/arm/msm/ssbi.txt | 18 +++++ arch/arm/boot/dts/msm8660-surf.dts | 6 ++ arch/arm/boot/dts/msm8960-cdp.dts | 6 ++ drivers/ssbi/ssbi.c | 81 +++++++++------------- 4 files changed, 62 insertions(+), 49 deletions(-) create mode 100644 Documentation/devicetree/bindings/arm/msm/ssbi.txt
Documentation/devicetree/bindings/arm/msm/ssbi.txt new file +18 -0
@@ -0,0 +1,18 @@ +* Qualcomm SSBI + +Some Qualcomm MSM devices contain a point-to-point serial bus used to +communicate with a limited range of devices (mostly power management +chips). + +These require the following properties: + +- compatible: "qcom,ssbi" + +- qcom,controller-type + indicates the SSBI bus variant the controller should use to talk + with the slave device. This should be one of "ssbi", "ssbi2", or + "pmic-arbiter". The type chosen is determined by the attached + slave. + +The slave device should be the single child node of the ssbi device +with a compatible field.
arch/arm/boot/dts/msm8660-surf.dts +6 -0
@@ -38,4 +38,10 @@ <0x19c00000 0x1000>; interrupts = <0 195 0x0>; }; + + qcom,ssbi at 500000 { + compatible = "qcom,ssbi"; + reg = <0x500000 0x1000>; + qcom,controller-type = "pmic-arbiter"; + }; };
arch/arm/boot/dts/msm8960-cdp.dts +6 -0
@@ -38,4 +38,10 @@ <0x16400000 0x1000>; interrupts = <0 154 0x0>; }; + + qcom,ssbi at 500000 { + compatible = "qcom,ssbi"; + reg = <0x500000 0x1000>; + qcom,controller-type = "pmic-arbiter"; + }; };
drivers/ssbi/ssbi.c +32 -49
126 lines not on this page · raw message
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
Re: [PATCH v2 05/11] SSBI: Convert SSBI to device tree
Quoted 34 lines · David Brown
drivers/ssbi/ssbi.c +0
-30@@ -261,56 +263,13 @@ int msm_ssbi_write(struct device *dev, u16 addr, u8 *buf, int len) }EXPORT_SYMBOL_GPL (msm_ssbi_write); -static int msm_ssbi_add_slave (struct msm_ssbi *ssbi, -const struct msm_ssbi_slave_info *slave) -{ -struct platform_device *slave_pdev; -int ret; - -if (ssbi->slave) { -pr_err ("slave already attached??\n" ); -return -EBUSY ; - } - - slave_pdev =platform_device_alloc (slave->name,-1 ); -if (!slave_pdev) { -pr_err ("cannot allocate pdev for slave '%s'" , slave->name); - ret = -ENOMEM ; - goto err; - } - - slave_pdev->dev.parent = ssbi->dev; - slave_pdev->dev.platform_data = slave->platform_data; - - ret =platform_device_add (slave_pdev); -if (ret) { -pr_err ("cannot add slave platform device for '%s'\n" , - slave->name); - goto err; - } - - ssbi->slave = &slave_pdev->dev; -return 0 ;
drivers/ssbi/ssbi.c int msm_ssbi_write(struct device *dev, u16 addr, u8 *buf, int len) · line 303
Quoted 10 lines
drivers/ssbi/ssbi.c +0
-8- -err: -if (slave_pdev) -platform_device_put (slave_pdev); -return ret; -} -static int msm_ssbi_probe (struct platform_device *pdev) { -const struct msm_ssbi_platform_data *pdata = pdev->dev.platform_data;
Now that all this code is gone do we have a user of the msm_ssbi_platform_data struct? Why not remove it from the header file? If you remove it from the header you can probably move the enum for controller_type into this file too.
drivers/ssbi/ssbi.c static int msm_ssbi_probe(struct platform_device *pdev) · line 298
Quoted 8 lines
drivers/ssbi/ssbi.c +3
-1@@ -334,7 +293,25 @@ static int msm_ssbi_probe(struct platform_device *pdev) ssbi->dev = &pdev->dev;platform_set_drvdata (pdev, ssbi); - ssbi->controller_type = pdata->controller_type; + type =of_get_property (np,"qcom,controller-type" ,NULL ); +if (type ==NULL ) { +pr_err ("Missing qcom,controller-type property\n" );
This could be dev_err() considering you use dev_info() below.
drivers/ssbi/ssbi.c static int msm_ssbi_probe(struct platform_device *pdev) · line 310
Quoted 12 lines
drivers/ssbi/ssbi.c +12
-0+ ret = -EINVAL ; + goto err_ssbi_controller; + } +dev_info (&pdev->dev,"SSBI controller type: '%s'\n" , type); +if (strcmp (type,"ssbi" ) ==0 ) + ssbi->controller_type =MSM_SBI_CTRL_SSBI ; +else if (strcmp (type,"ssbi2" ) ==0 ) + ssbi->controller_type =MSM_SBI_CTRL_SSBI2 ; +else if (strcmp (type,"pmic-arbiter" ) ==0 ) + ssbi->controller_type =MSM_SBI_CTRL_PMIC_ARBITER ; +else { +pr_err ("Unknown qcom,controller-type\n" );
dev_err()?
drivers/ssbi/ssbi.c static int msm_ssbi_remove(struct platform_device *pdev) · line 350
Quoted 5 lines
drivers/ssbi/ssbi.c +1
-0@@ -370,12 +347,18 @@ static int msm_ssbi_remove(struct platform_device *pdev)return 0 ; } +static struct of_device_id ssbi_match_table[] = {
const?
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation
PATCH v2 6/11 ssbi: Comment the use of udelay()
In reply to David BrownThe ssbi driver uses a busywait loop to read its status register. Add a comment explaining the timing of the device itself so that future developers can better understand this delay, and possibly diagnose any problems.
--- drivers/ssbi/ssbi.c | 13 +++++++++++++ 1 file changed, 13 insertions(+)
drivers/ssbi/ssbi.c +13 -0
@@ -87,6 +87,15 @@ static inline void ssbi_writel(struct msm_ssbi *ssbi, u32 val, u32 reg)writel (val, ssbi->base + reg); } +/* +* Via private exchange with one of the original authors, the hardware +* should generally finish a transaction in about 5us. The worst +* case, is when using the arbiter and both other CPUs have just +* started trying to use the SSBI bus will result in a time of about +* 20us. It should never take longer than this. +* +* As such, this wait merely spins, with a udelay. +*/ static int ssbi_wait_mask (struct msm_ssbi *ssbi,u32 set_mask,u32 clr_mask) {u32 timeout =SSBI_TIMEOUT_US ;
@@ -161,6 +170,10 @@ err:return ret; } +/* +* See ssbi_wait_mask for an explanation of the time and the +* busywait. +*/ static inline int msm_ssbi_pa_transfer (struct msm_ssbi *ssbi,u32 cmd,u8 *data) {
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH v2 7/11 ssbi: Use regular init level
In reply to David BrownWith device tree, and deferred probe, it is no longer necessary to make sure that the ssbi bus driver is initialized very early. Restore to a regular module_init().
--- drivers/ssbi/ssbi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/ssbi/ssbi.c +1 -1
@@ -379,7 +379,7 @@ static int __init msm_ssbi_init(void) {return platform_driver_register (&msm_ssbi_driver); } -postcore_initcall (msm_ssbi_init); +module_init (msm_ssbi_init); static void __exit msm_ssbi_exit(void) {
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
Re: [PATCH v2 07/11] ssbi: Use regular init level
Quoted 14 lines · David Brown
With device tree, and deferred probe, it is no longer necessary to make sure that the ssbi bus driver is initialized very early. Restore to a regular
module_init().Signed-off-by David Brown <[email protected]>--- drivers/ssbi/ssbi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/ssbi/ssbi.c +0
-0
drivers/ssbi/ssbi.c static int __init msm_ssbi_init(void) · line 385
Quoted 8 lines
drivers/ssbi/ssbi.c +1
-1@@ -379,7 +379,7 @@ static int __init msm_ssbi_init(void) {return platform_driver_register (&msm_ssbi_driver); } -postcore_initcall (msm_ssbi_init); +module_init (msm_ssbi_init); static void __exit msm_ssbi_exit(void) {
With this change we can use module_platform_driver() too.
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation
PATCH v2 8/11 ssbi: Remove extraneous logging
In reply to David BrownRemove some unhelpful error logs. This also removes the necessity of having a pointer back to the struct device within the ssbi-specific structure
--- drivers/ssbi/ssbi.c | 16 +--------------- 1 file changed, 1 insertion(+), 15 deletions(-)
drivers/ssbi/ssbi.c +1 -15
@@ -66,7 +66,6 @@#define SSBI_TIMEOUT_US 100struct msm_ssbi { -struct device *dev;struct device *slave;void __iomem *base;spinlock_t lock;
@@ -108,8 +107,6 @@ static int ssbi_wait_mask(struct msm_ssbi *ssbi, u32 set_mask, u32 clr_mask)udelay (1 ); } -dev_err (ssbi->dev,"%s: timeout (status %x set_mask %x clr_mask %x)\n" , - __func__,ssbi_readl (ssbi,SSBI2_STATUS ), set_mask, clr_mask);return -ETIMEDOUT ; }
@@ -185,11 +182,8 @@ msm_ssbi_pa_transfer(struct msm_ssbi *ssbi, u32 cmd, u8 *data)while (timeout--) { rd_status =ssbi_readl (ssbi,SSBI_PA_RD_STATUS ); -if (rd_status &SSBI_PA_RD_STATUS_TRANS_DENIED ) { -dev_err (ssbi->dev,"%s: transaction denied (0x%x)\n" , - __func__, rd_status); +if (rd_status &SSBI_PA_RD_STATUS_TRANS_DENIED )return -EPERM ; - }if (rd_status &SSBI_PA_RD_STATUS_TRANS_DONE ) {if (data)
@@ -199,7 +193,6 @@ msm_ssbi_pa_transfer(struct msm_ssbi *ssbi, u32 cmd, u8 *data)udelay (1 ); } -dev_err (ssbi->dev,"%s: timeout, status 0x%x\n" , __func__, rd_status);return -ETIMEDOUT ; }
@@ -248,9 +241,6 @@ int msm_ssbi_read(struct device *dev, u16 addr, u8 *buf, int len)unsigned long flags;int ret; -if (ssbi->dev != dev) -return -ENXIO ; -spin_lock_irqsave (&ssbi->lock, flags); ret = ssbi->read (ssbi, addr, buf, len);spin_unlock_irqrestore (&ssbi->lock, flags);
@@ -265,9 +255,6 @@ int msm_ssbi_write(struct device *dev, u16 addr, u8 *buf, int len)unsigned long flags;int ret; -if (ssbi->dev != dev) -return -ENXIO ; -spin_lock_irqsave (&ssbi->lock, flags); ret = ssbi->write (ssbi, addr, buf, len);spin_unlock_irqrestore (&ssbi->lock, flags);
@@ -303,7 +290,6 @@ static int msm_ssbi_probe(struct platform_device *pdev) ret = -EINVAL ; goto err_ioremap; } - ssbi->dev = &pdev->dev;platform_set_drvdata (pdev, ssbi); type =of_get_property (np,"qcom,controller-type" ,NULL );
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH v2 9/11 SSBI: Remove MSM_ prefix from SSBI drivers
In reply to David BrownAlthough the SSBI sub is currently only used on MSM SoCs, it is still a bus in its own right. Remove this msm_ prefix from the driver and it's symbols. Clients can now refer directly to ssbi_write() and ssbi_read().
---
drivers/Makefile | 2 +-
drivers/mfd/Kconfig | 2 +-
drivers/mfd/pm8921-core.c | 14 +++---
drivers/ssbi/Kconfig | 4 +-
drivers/ssbi/Makefile | 2 +-
drivers/ssbi/ssbi.c | 86 ++++++++++++++++++------------------
include/linux/{msm_ssbi.h => ssbi.h} | 18 ++++----
7 files changed, 64 insertions(+), 64 deletions(-)
rename include/linux/{msm_ssbi.h => ssbi.h} (67%)drivers/Makefile +1 -1
@@ -114,7 +114,7 @@ obj-y += firmware/ obj-$(CONFIG_CRYPTO) += crypto/ obj-$(CONFIG_SUPERH) += sh/ obj-$(CONFIG_ARCH_SHMOBILE) += sh/ -obj-$(CONFIG_MSM_SSBI) += ssbi/ +obj-$(CONFIG_SSBI) += ssbi/ ifndef CONFIG_ARCH_USES_GETTIMEOFFSET obj-y += clocksource/ endif
drivers/mfd/Kconfig +1 -1
@@ -990,7 +990,7 @@ config MFD_PM8XXX config MFD_PM8921_CORE tristate "Qualcomm PM8921 PMIC chip" - depends on MSM_SSBI + depends on SSBI select MFD_CORE select MFD_PM8XXX help
drivers/mfd/pm8921-core.c +7 -7
@@ -17,7 +17,7 @@#include <linux/platform_device.h> #include <linux/slab.h> #include <linux/err.h> -#include <linux/msm_ssbi.h> +#include <linux/ssbi.h> #include <linux/mfd/core.h> #include <linux/mfd/pm8xxx/pm8921.h> #include <linux/mfd/pm8xxx/core.h>
@@ -35,7 +35,7 @@ static int pm8921_readb(const struct device *dev, u16 addr, u8 *val)const struct pm8xxx_drvdata *pm8921_drvdata =dev_get_drvdata (dev);const struct pm8921 *pmic = pm8921_drvdata->pm_chip_data; -return msm_ssbi_read (pmic->dev->parent, addr, val,1 ); +return ssbi_read (pmic->dev->parent, addr, val,1 ); }static int pm8921_writeb (const struct device *dev,u16 addr,u8 val)
@@ -43,7 +43,7 @@ static int pm8921_writeb(const struct device *dev, u16 addr, u8 val)const struct pm8xxx_drvdata *pm8921_drvdata =dev_get_drvdata (dev);const struct pm8921 *pmic = pm8921_drvdata->pm_chip_data; -return msm_ssbi_write (pmic->dev->parent, addr, &val,1 ); +return ssbi_write (pmic->dev->parent, addr, &val,1 ); } static int pm8921_read_buf(const struct device *dev, u16 addr, u8 *buf,
@@ -52,7 +52,7 @@ static int pm8921_read_buf(const struct device *dev, u16 addr, u8 *buf,const struct pm8xxx_drvdata *pm8921_drvdata =dev_get_drvdata (dev);const struct pm8921 *pmic = pm8921_drvdata->pm_chip_data; -return msm_ssbi_read (pmic->dev->parent, addr, buf, cnt); +return ssbi_read (pmic->dev->parent, addr, buf, cnt); } static int pm8921_write_buf(const struct device *dev, u16 addr, u8 *buf,
@@ -61,7 +61,7 @@ static int pm8921_write_buf(const struct device *dev, u16 addr, u8 *buf,const struct pm8xxx_drvdata *pm8921_drvdata =dev_get_drvdata (dev);const struct pm8921 *pmic = pm8921_drvdata->pm_chip_data; -return msm_ssbi_write (pmic->dev->parent, addr, buf, cnt); +return ssbi_write (pmic->dev->parent, addr, buf, cnt); }static int pm8921_read_irq_stat (const struct device *dev,int irq)
@@ -124,7 +124,7 @@ static int pm8921_probe(struct platform_device *pdev) }/* Read PMIC chip revision */ - rc =msm_ssbi_read (pdev->dev.parent,REG_HWREV , &val,sizeof (val)); + rc =ssbi_read (pdev->dev.parent,REG_HWREV , &val,sizeof (val));if (rc) {pr_err ("Failed to read hw rev reg %d:rc=%d\n" ,REG_HWREV , rc); goto err_read_rev;
@@ -133,7 +133,7 @@ static int pm8921_probe(struct platform_device *pdev) rev = val;/* Read PMIC chip revision 2 */ - rc =msm_ssbi_read (pdev->dev.parent,REG_HWREV_2 , &val,sizeof (val)); + rc =ssbi_read (pdev->dev.parent,REG_HWREV_2 , &val,sizeof (val));if (rc) {pr_err ("Failed to read hw rev 2 reg %d:rc=%d\n" ,REG_HWREV_2 , rc);
drivers/ssbi/Kconfig +2 -2
@@ -1,10 +1,10 @@ # -# MSM SSBI bus support +# SSBI bus support # menu "Qualcomm MSM SSBI bus support" -config MSM_SSBI +config SSBI tristate "Qualcomm Single-wire Serial Bus Interface (SSBI)" help If you say yes to this option, support will be included for the
drivers/ssbi/Makefile +1 -1
@@ -1 +1 @@ -obj-$(CONFIG_MSM_SSBI) += ssbi.o +obj-$(CONFIG_SSBI) += ssbi.o
drivers/ssbi/ssbi.c +43 -43
219 lines not on this page · raw message
include/linux/ssbi.h renamed from include/linux/msm_ssbi.h +9 -9
@@ -12,27 +12,27 @@* GNU General Public License for more details. */ -#ifndef _LINUX_MSM_SSBI_H -#define _LINUX_MSM_SSBI_H +#ifndef _LINUX_SSBI_H +#define _LINUX_SSBI_H#include <linux/types.h> -struct msm_ssbi_slave_info { +struct ssbi_slave_info {const char *name;void *platform_data; }; -enum msm_ssbi_controller_type { +enum ssbi_controller_type {MSM_SBI_CTRL_SSBI =0 ,MSM_SBI_CTRL_SSBI2 ,MSM_SBI_CTRL_PMIC_ARBITER , }; -struct msm_ssbi_platform_data { -struct msm_ssbi_slave_info slave; -enum msm_ssbi_controller_type controller_type; +struct ssbi_platform_data { +struct ssbi_slave_info slave; +enum ssbi_controller_type controller_type; }; -int msm_ssbi_write (struct device *dev,u16 addr,u8 *buf,int len); -int msm_ssbi_read (struct device *dev,u16 addr,u8 *buf,int len); +int ssbi_write (struct device *dev,u16 addr,u8 *buf,int len); +int ssbi_read (struct device *dev,u16 addr,u8 *buf,int len);#endif
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
PATCH v2 10/11 MAINTAINERS: add ssbi
In reply to David BrownThe ssbi device is specific to the Qualcomm MSM SoCs.
--- MAINTAINERS | 1 + 1 file changed, 1 insertion(+)
MAINTAINERS +1 -0
@@ -1031,6 +1031,7 @@ F: drivers/mmc/host/msm_sdcc.h F: drivers/tty/serial/msm_serial.h F: drivers/tty/serial/msm_serial.c F: drivers/*/pm8???-* +F: drivers/ssbi/ F: include/linux/mfd/pm8xxx/ T: git git://git.kernel.org/pub/scm/linux/kernel/git/davidb/linux-msm.git S: Maintained
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
Quoted 6 lines · David Brown
This small series adds the Qualcomm SSBI bus driver. The original driver was developed as part of Android. Kenneth Heitke updated this, and sent it out a while back. This series updates this driver to use DeviceTree, and fixes a few things that have changed since it was originally sent out.
Quoted 6 lines
In addition, the 11th patch, the RFC, is a small test driver that reads the version register off of the pm8058 device typically connected to an MSM8660 or MSM8960. This is primarily useful to anyone wanting to work on these pmic drivers.
Updates to the PMIC drivers themselves to use this driver to come.
The first 10 now applied to my char/misc tree, thanks.
greg k-h
PATCH v2 11/11 RFC: SSBI: Simple pm8058 test driver
In reply to David BrownA very small ssbi device driver that reads the pm8058 version register and prints it out.
---
Resending due to address error in header
arch/arm/boot/dts/msm8660-surf.dts | 4 +++ arch/arm/boot/dts/msm8960-cdp.dts | 4 +++ drivers/ssbi/Makefile | 1 + drivers/ssbi/ssbi-test.c | 61 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 70 insertions(+) create mode 100644 drivers/ssbi/ssbi-test.c
arch/arm/boot/dts/msm8660-surf.dts +4 -0
@@ -43,5 +43,9 @@ compatible = "qcom,ssbi"; reg = <0x500000 0x1000>; qcom,controller-type = "pmic-arbiter"; + + qcom,ssbi-test { + compatible = "qcom,ssbi-test"; + }; }; };
arch/arm/boot/dts/msm8960-cdp.dts +4 -0
@@ -43,5 +43,9 @@ compatible = "qcom,ssbi"; reg = <0x500000 0x1000>; qcom,controller-type = "pmic-arbiter"; + + qcom,ssbi-test { + compatible = "qcom,ssbi-test"; + }; }; };
drivers/ssbi/Makefile +1 -0
@@ -1 +1,2 @@ obj-$(CONFIG_SSBI) += ssbi.o +obj-$(CONFIG_SSBI) += ssbi-test.o
drivers/ssbi/ssbi-test.c new file +61 -0
61 lines not on this page · raw message
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation