Skip to contentSkip to conversation

IBM Real-Time "SMI Free" mode driver -v3

RFC PATCH In review Vernon Mauery 4 participants 20 messages Sep 21, 2010 – Sep 25, 2010 x86 platform drivers

lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 IBM Real-Time "SMI Free" mode driver -v3

IBM Real-Time "SMI Free" mode driver

This driver supports the Real-Time Linux (RTL) BIOS feature. The RTL feature allows non-fatal System Management Interrupts (SMIs) to be disabled on supported IBM platforms and is intended to be coupled with a user-space daemon to monitor the hardware in a way that can be prioritized and scheduled to better suit the requirements for the system.

The Device is presented as a special "_RTL_" table to the OS in the Extended BIOS Data Area. There is a simple protocol for entering and exiting the mode at runtime. This driver creates a simple sysfs interface to allow a simple entry and exit from RTL mode in the UFI/BIOS.

This patch has taken the feedback in response to previous posts by Keith Mannthey <[email protected]> in addition to a couple of more improvements, including using a DMI probe to reduce the probability of getting into trouble on non-IBM systems.

drivers/platform/x86/Kconfig +16 -0
index cff7cc2..16ba6e6 100644
@@ -590,4 +590,20 @@ config INTEL_IPS
 	  functionality.  If in doubt, say Y here; it will only load on
 	  supported platforms.
 
+config IBM_RTL
+	tristate "Device driver to enable PRTL support"
+	depends on X86 && PCI
+	---help---
+	  Enable support for IBM Preimium Real Time Mode (PRTM).
+	  This module will allow you the enter and exit PRTM in the BIOS via
+	  sysfs on platforms that support this feature.  System in PRTM will
+	  not recive cpu generated SMIs for recoverable errors.  Use of this
+	  feature without proper support may void your hardware warrenty.
+
+	  If the proper bios support is found the driver will load and create
+	  /sys/devices/system/ibm_rtl/.  The "state" varable will indicate
+	  weather or not the BIOS is in PRTM.
+	  state = 0 (BIOS SMI's on)
+	  state = 1 (BIOS SMI's off)
+
 endif # X86_PLATFORM_DEVICES
drivers/platform/x86/Makefile +1 -0
index 85fb2b8..50fae7a 100644
@@ -30,4 +30,5 @@ obj-$(CONFIG_INTEL_SCU_IPC)	+= intel_scu_ipc.o
 obj-$(CONFIG_RAR_REGISTER)	+= intel_rar_register.o
 obj-$(CONFIG_INTEL_IPS)		+= intel_ips.o
 obj-$(CONFIG_GPIO_INTEL_PMIC)	+= intel_pmic_gpio.o
+obj-$(CONFIG_IBM_RTL)		+= ibm_rtl.o
 
drivers/platform/x86/ibm_rtl.c new file +352 -0

352 lines not on this page · raw message

#1 lore raw
Quoted 9 lines · On Tue, 21 Sep 2010 15:46:10 -0700 Vernon Mauery wrote:

IBM Real-Time "SMI Free" mode driver

drivers/platform/x86/Kconfig +0 -0
index cff7cc2..16ba6e6 100644

drivers/platform/x86/Kconfig config INTEL_IPS · line 597

Quoted 8 lines
drivers/platform/x86/Kconfig +5 -0
@@ -590,4 +590,20 @@ config INTEL_IPS
  	  functionality.  If in doubt, say Y here; it will only load on
  	  supported platforms.
 
+config IBM_RTL
+	tristate "Device driver to enable PRTL support"
+	depends on X86 && PCI
+	---help---
+	  Enable support for IBM Preimium Real Time Mode (PRTM).

What is Preimium?

drivers/platform/x86/Kconfig config INTEL_IPS · line 600

Quoted 3 lines
drivers/platform/x86/Kconfig +3 -0
+	  This module will allow you the enter and exit PRTM in the BIOS via
+	  sysfs on platforms that support this feature.  System in PRTM will
+	  not recive cpu generated SMIs for recoverable errors.  Use of this

receive CPU-generated

drivers/platform/x86/Kconfig config INTEL_IPS · line 601

Quoted 1 line
drivers/platform/x86/Kconfig +1 -0
+	  feature without proper support may void your hardware warrenty.

warranty.

drivers/platform/x86/Kconfig config INTEL_IPS · line 603

Quoted 2 lines
drivers/platform/x86/Kconfig +2 -0
+
+	  If the proper bios support is found the driver will load and create

BIOS

drivers/platform/x86/Kconfig config INTEL_IPS · line 604

Quoted 1 line
drivers/platform/x86/Kconfig +1 -0
+	  /sys/devices/system/ibm_rtl/.  The "state" varable will indicate

variable

drivers/platform/x86/Kconfig config INTEL_IPS · line 605

Quoted 1 line
drivers/platform/x86/Kconfig +1 -0
+	  weather or not the BIOS is in PRTM.

whether

Quoted 33 lines
drivers/platform/x86/Kconfig +3 -0
+	  state = 0 (BIOS SMI's on)
+	  state = 1 (BIOS SMI's off)
+
  endif # X86_PLATFORM_DEVICES
drivers/platform/x86/ibm_rtl.c new file +22 -0
new file mode 100644
index 0000000..37bf610
@@ -0,0 +1,352 @@
+/*
+ * IBM Real-Time Linux driver
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright (C) IBM Corporation, 2010
+ *
+ * Author: Keith Mannthey <[email protected]>
+ *         Vernon Mauery <[email protected]>
+ *

drivers/platform/x86/ibm_rtl.c line 32

Quoted 10 lines
drivers/platform/x86/ibm_rtl.c new file +10 -0
+ */
+
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/module.h>
+#include <linux/io.h>
+#include <linux/sysdev.h>
+#include <linux/dmi.h>
+#include <asm/bios_ebda.h>
+#include <linux/mutex.h>

<asm> after all <linux> headers, please.

Quoted 13 lines
drivers/platform/x86/ibm_rtl.c new file +13 -0
+
+static int force;
+module_param(force, bool, 0);
+MODULE_PARM_DESC(force, "Force driver load, ignore DMI data");
+
+static int debug;
+module_param(debug, bool, 0644);
+MODULE_PARM_DESC(debug, "Show debug output");
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Keith Mannthey <[email protected]>");
+MODULE_AUTHOR("Vernon Mauery <[email protected]>");
+

drivers/platform/x86/ibm_rtl.c line 55

Quoted 10 lines
drivers/platform/x86/ibm_rtl.c new file +10 -0
+enum rtl_addr_type {
+	RTL_ADDR_TYPE_IO = 1,
+	RTL_ADDR_TYPE_MMIO,
+} __attribute__((packed));
+
+enum rtl_cmd_type {
+	RTL_CMD_NOP = 0,
+	RTL_CMD_ENTER_PRTM,
+	RTL_CMD_EXIT_PRTM,
+} __attribute__((packed));

What does packed do to an enum list?

drivers/platform/x86/ibm_rtl.c line 57

Quoted 2 lines
drivers/platform/x86/ibm_rtl.c new file +2 -0
+
+/* The RTL table looks something like: */

Unsure of this?

Quoted 14 lines
drivers/platform/x86/ibm_rtl.c new file +14 -0
+struct ibm_rtl_table {
+	char signature[5];
+	u8 version;
+	u8 rt_status;
+	enum rtl_cmd_type command;
+	u8 command_status;
+	enum rtl_addr_type cmd_address_type;
+	u8 cmd_granularity;
+	u8 cmd_offset;
+	u16 reserve1;
+	u8 cmd_port_address[4]; /* platform dependent address */
+	u8 cmd_port_value[4];   /* platform dependent value */
+};
+

drivers/platform/x86/ibm_rtl.c line 81

Quoted 10 lines
drivers/platform/x86/ibm_rtl.c new file +10 -0
+#define RTL_SIGNATURE (('L'<<24)|('T'<<16)|('R'<<8)|'_')
+
+#define ERROR(A, B...) printk(KERN_ERR "ibm-rtl: " A, ##B )
+#define WARNING(A, B...) printk(KERN_WARNING "ibm-rtl: " A, ##B )
+#define DEBUG(A, B...) do { \
+	if (debug) \
+		printk(KERN_INFO "ibm-rtl: " A, ##B ); \
+} while (0)
+
+DEFINE_MUTEX(rtl_lock);

static DEFINE_MUTEX(rtl_lock);

drivers/platform/x86/ibm_rtl.c line 87

Quoted 6 lines
drivers/platform/x86/ibm_rtl.c new file +6 -0
+static struct ibm_rtl_table __iomem *rtl_table = NULL;
+static void __iomem *ebda_map;
+static void __iomem *rtl_cmd_iomem_addr = NULL;
+static u32 rtl_cmd_port_addr;
+static enum rtl_addr_type rtl_cmd_type;
+static u8 rtl_cmd_width;

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

#2 lore raw
Quoted 14 lines · Randy Dunlap
Quoted 12 lines · On Tue, 21 Sep 2010 15:46:10 -0700 Vernon Mauery wrote:

IBM Real-Time "SMI Free" mode driver

drivers/platform/x86/Kconfig +0 -0
index cff7cc2..16ba6e6 100644
@@ -590,4 +590,20 @@ config INTEL_IPS
  	  functionality.  If in doubt, say Y here; it will only load on
  	  supported platforms.
 
Quoted 7 lines

drivers/platform/x86/Kconfig config INTEL_IPS · line 597

Quoted 5 lines
drivers/platform/x86/Kconfig +5 -0
+config IBM_RTL
+	tristate "Device driver to enable PRTL support"
+	depends on X86 && PCI
+	---help---
+	  Enable support for IBM Preimium Real Time Mode (PRTM).

What is Preimium?

Okay, this is just plain embarrassing. Or should I say embarasing? :) I swear I already fixed those from when you commented on Keith's patch. I must have copied over my changes somehow.

These are fixed in the forthcoming patch. Thanks for your patience.

Quoted 96 lines

drivers/platform/x86/Kconfig config INTEL_IPS · line 600

Quoted 4 lines
drivers/platform/x86/Kconfig +3 -0
+	  This module will allow you the enter and exit PRTM in the BIOS via
+	  sysfs on platforms that support this feature.  System in PRTM will
+	  not recive cpu generated SMIs for recoverable errors.  Use of this

receive CPU-generated

drivers/platform/x86/Kconfig config INTEL_IPS · line 601

Quoted 1 line
drivers/platform/x86/Kconfig +1 -0
+	  feature without proper support may void your hardware warrenty.

warranty.

drivers/platform/x86/Kconfig config INTEL_IPS · line 603

Quoted 2 lines
drivers/platform/x86/Kconfig +2 -0
+
+	  If the proper bios support is found the driver will load and create

BIOS

drivers/platform/x86/Kconfig config INTEL_IPS · line 604

Quoted 1 line
drivers/platform/x86/Kconfig +1 -0
+	  /sys/devices/system/ibm_rtl/.  The "state" varable will indicate

variable

drivers/platform/x86/Kconfig config INTEL_IPS · line 605

Quoted 1 line
drivers/platform/x86/Kconfig +1 -0
+	  weather or not the BIOS is in PRTM.

whether

Quoted 33 lines
drivers/platform/x86/Kconfig +3 -0
+	  state = 0 (BIOS SMI's on)
+	  state = 1 (BIOS SMI's off)
+
  endif # X86_PLATFORM_DEVICES
drivers/platform/x86/ibm_rtl.c new file +22 -0
new file mode 100644
index 0000000..37bf610
@@ -0,0 +1,352 @@
+/*
+ * IBM Real-Time Linux driver
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright (C) IBM Corporation, 2010
+ *
+ * Author: Keith Mannthey <[email protected]>
+ *         Vernon Mauery <[email protected]>
+ *

drivers/platform/x86/ibm_rtl.c line 32

Quoted 10 lines
drivers/platform/x86/ibm_rtl.c new file +10 -0
+ */
+
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/module.h>
+#include <linux/io.h>
+#include <linux/sysdev.h>
+#include <linux/dmi.h>
+#include <asm/bios_ebda.h>
+#include <linux/mutex.h>

<asm> after all <linux> headers, please.

Quoted 13 lines
drivers/platform/x86/ibm_rtl.c new file +13 -0
+
+static int force;
+module_param(force, bool, 0);
+MODULE_PARM_DESC(force, "Force driver load, ignore DMI data");
+
+static int debug;
+module_param(debug, bool, 0644);
+MODULE_PARM_DESC(debug, "Show debug output");
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Keith Mannthey <[email protected]>");
+MODULE_AUTHOR("Vernon Mauery <[email protected]>");
+

drivers/platform/x86/ibm_rtl.c line 55

Quoted 10 lines
drivers/platform/x86/ibm_rtl.c new file +10 -0
+enum rtl_addr_type {
+	RTL_ADDR_TYPE_IO = 1,
+	RTL_ADDR_TYPE_MMIO,
+} __attribute__((packed));
+
+enum rtl_cmd_type {
+	RTL_CMD_NOP = 0,
+	RTL_CMD_ENTER_PRTM,
+	RTL_CMD_EXIT_PRTM,
+} __attribute__((packed));

What does packed do to an enum list?

Quoted 4 lines

drivers/platform/x86/ibm_rtl.c line 57

Quoted 2 lines
drivers/platform/x86/ibm_rtl.c new file +2 -0
+
+/* The RTL table looks something like: */

Unsure of this?

:) Not really. A little less waffling never hurts.

Quoted 19 lines
Quoted 19 lines
drivers/platform/x86/ibm_rtl.c new file +18 -0
+struct ibm_rtl_table {
+	char signature[5];
+	u8 version;
+	u8 rt_status;
+	enum rtl_cmd_type command;
+	u8 command_status;
+	enum rtl_addr_type cmd_address_type;
+	u8 cmd_granularity;
+	u8 cmd_offset;
+	u16 reserve1;
+	u8 cmd_port_address[4]; /* platform dependent address */
+	u8 cmd_port_value[4];   /* platform dependent value */
+};
+
+#define RTL_SIGNATURE (('L'<<24)|('T'<<16)|('R'<<8)|'_')
+
+#define ERROR(A, B...) printk(KERN_ERR "ibm-rtl: " A, ##B )
+#define WARNING(A, B...) printk(KERN_WARNING "ibm-rtl: " A, ##B )
Quoted 8 lines

drivers/platform/x86/ibm_rtl.c line 81

Quoted 6 lines
drivers/platform/x86/ibm_rtl.c new file +6 -0
+#define DEBUG(A, B...) do { \
+	if (debug) \
+		printk(KERN_INFO "ibm-rtl: " A, ##B ); \
+} while (0)
+
+DEFINE_MUTEX(rtl_lock);

static DEFINE_MUTEX(rtl_lock);

Thanks for your review.

--Vernon

Quoted 13 lines

drivers/platform/x86/ibm_rtl.c line 87

Quoted 7 lines
drivers/platform/x86/ibm_rtl.c new file +6 -0
+static struct ibm_rtl_table __iomem *rtl_table = NULL;
+static void __iomem *ebda_map;
+static void __iomem *rtl_cmd_iomem_addr = NULL;
+static u32 rtl_cmd_port_addr;
+static enum rtl_addr_type rtl_cmd_type;
+static u8 rtl_cmd_width;

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

#3 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

IBM Real-Time "SMI Free" mode driver

This driver supports the Real-Time Linux (RTL) BIOS feature. The RTL feature allows non-fatal System Management Interrupts (SMIs) to be disabled on supported IBM platforms and is intended to be coupled with a user-space daemon to monitor the hardware in a way that can be prioritized and scheduled to better suit the requirements for the system.

The Device is presented as a special "_RTL_" table to the OS in the Extended BIOS Data Area. There is a simple protocol for entering and exiting the mode at runtime. This driver creates a simple sysfs interface to allow a simple entry and exit from RTL mode in the UFI/BIOS.

This patch has taken the feedback in response to previous posts by Keith Mannthey <[email protected]> in addition to a couple of more improvements, including using a DMI probe to reduce the probability of getting into trouble on non-IBM systems.

--

Documentation/ABI/testing/sysfs-devices-system-ibm-rtl new file +22 -0
new file mode 100644
index 0000000..b82deea
@@ -0,0 +1,22 @@
+What:           state
+Date:           Sep 2010
+KernelVersion:  2.6.37
+Contact:        Vernon Mauery <[email protected]>
+Description:    The state file allows a means by which to change in and
+                out of Premium Real-Time Mode (PRTM), as well as the
+                ability to query the current state.
+                    0 => PRTM off
+                    1 => PRTM enabled
+Users:          The ibm-prtm userspace daemon uses this interface.
+
+
+What:           version
+Date:           Sep 2010
+KernelVersion:  2.6.37
+Contact:        Vernon Mauery <[email protected]>
+Description:    The version file provides a means by which to query
+                the RTL table version that lives in the Extended
+                BIOS Data Area (EBDA).
+Users:          The ibm-prtm userspace daemon uses this interface.
+
+
drivers/platform/x86/Kconfig +16 -0
index cff7cc2..5fba368 100644
@@ -590,4 +590,20 @@ config INTEL_IPS
 	  functionality.  If in doubt, say Y here; it will only load on
 	  supported platforms.
 
+config IBM_RTL
+	tristate "Device driver to enable PRTL support"
+	depends on X86 && PCI
+	---help---
+	  Enable support for IBM Premium Real Time Mode (PRTM).
+	  This module will allow you the enter and exit PRTM in the BIOS via
+	  sysfs on platforms that support this feature.  System in PRTM will
+	  not recive CPU-generated SMIs for recoverable errors.  Use of this
+	  feature without proper support may void your hardware warranty.
+
+	  If the proper BIOS support is found the driver will load and create
+	  /sys/devices/system/ibm_rtl/.  The "state" variable will indicate
+	  whether or not the BIOS is in PRTM.
+	  state = 0 (BIOS SMIs on)
+	  state = 1 (BIOS SMIs off)
+
 endif # X86_PLATFORM_DEVICES
drivers/platform/x86/Makefile +1 -0
index 85fb2b8..50fae7a 100644
@@ -30,4 +30,5 @@ obj-$(CONFIG_INTEL_SCU_IPC)	+= intel_scu_ipc.o
 obj-$(CONFIG_RAR_REGISTER)	+= intel_rar_register.o
 obj-$(CONFIG_INTEL_IPS)		+= intel_ips.o
 obj-$(CONFIG_GPIO_INTEL_PMIC)	+= intel_pmic_gpio.o
+obj-$(CONFIG_IBM_RTL)		+= ibm_rtl.o
 
drivers/platform/x86/ibm_rtl.c new file +352 -0

352 lines not on this page · raw message

#4 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

drivers/platform/x86/ibm_rtl.c line 55

Quoted 10 lines · Vernon Mauery
drivers/platform/x86/ibm_rtl.c new file +10 -0
+enum rtl_addr_type {
+	RTL_ADDR_TYPE_IO = 1,
+	RTL_ADDR_TYPE_MMIO,
+} __attribute__((packed));
+
+enum rtl_cmd_type {
+	RTL_CMD_NOP = 0,
+	RTL_CMD_ENTER_PRTM,
+	RTL_CMD_EXIT_PRTM,
+} __attribute__((packed));

You didn't reply to Randy's comment about the packed attribute. I think it's rather confusing here.

Quoted 4 lines
drivers/platform/x86/ibm_rtl.c new file +4 -0
+/* The RTL table as presented by the EBDA: */
+struct ibm_rtl_table {
+	char signature[5];
+	u8 version;

drivers/platform/x86/ibm_rtl.c line 70

Quoted 10 lines
drivers/platform/x86/ibm_rtl.c new file +10 -0
+	u8 rt_status;
+	enum rtl_cmd_type command;
+	u8 command_status;
+	enum rtl_addr_type cmd_address_type;
+	u8 cmd_granularity;
+	u8 cmd_offset;
+	u16 reserve1;
+	u8 cmd_port_address[4]; /* platform dependent address */
+	u8 cmd_port_value[4];   /* platform dependent value */
+};

I would recommend marking the member in this structure as packed instead, not the enum.

drivers/platform/x86/ibm_rtl.c line 79

Quoted 8 lines
drivers/platform/x86/ibm_rtl.c new file +8 -0
+#define RTL_SIGNATURE (('L'<<24)|('T'<<16)|('R'<<8)|'_')
+
+#define ERROR(A, B...) printk(KERN_ERR "ibm-rtl: " A, ##B )
+#define WARNING(A, B...) printk(KERN_WARNING "ibm-rtl: " A, ##B )
+#define DEBUG(A, B...) do { \
+	if (debug) \
+		printk(KERN_INFO "ibm-rtl: " A, ##B ); \
+} while (0)

We already have wrappers for these, no need to define your own. Please just use dev_{err,warn,dbg} or pr_{err,warning,debug}.

drivers/platform/x86/ibm_rtl.c line 87

Quoted 7 lines
drivers/platform/x86/ibm_rtl.c new file +7 -0
+static DEFINE_MUTEX(rtl_lock);
+static struct ibm_rtl_table __iomem *rtl_table = NULL;
+static void __iomem *ebda_map;
+static void __iomem *rtl_cmd_iomem_addr = NULL;
+static u32 rtl_cmd_port_addr;
+static enum rtl_addr_type rtl_cmd_type;
+static u8 rtl_cmd_width;

This is somewhat inconsistent, some of these are implicitly initialized, others have an explicit "= NULL". I would recommend leaving out the initialization, which is the historic way to do this in the kernel.

Some people find it cleaner to define a structure containing all the driver specific data. Since there can be only one of these devices in your case, it's probably not important.

drivers/platform/x86/ibm_rtl.c line 114

Quoted 4 lines
drivers/platform/x86/ibm_rtl.c new file +4 -0
+			if (rtl_cmd_type == RTL_ADDR_TYPE_MMIO)
+				iowrite8((u8)cmd_port_val, rtl_cmd_iomem_addr);
+			else
+				outb((u8)cmd_port_val, rtl_cmd_port_addr);

ioread/iowrite already has the capability to use both mmio and pio addresses. You can use ioport_map() to create an __iomem token that corresponds to your rtl_cmd_port_addr and get rid of the rtl_cmd_type variable.

Arnd

#5 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

Quoted 14 lines · Arnd Bergmann

drivers/platform/x86/ibm_rtl.c line 55

Quoted 10 lines · Vernon Mauery
drivers/platform/x86/ibm_rtl.c new file +10 -0
+enum rtl_addr_type {
+     RTL_ADDR_TYPE_IO = 1,
+     RTL_ADDR_TYPE_MMIO,
+} __attribute__((packed));
+
+enum rtl_cmd_type {
+     RTL_CMD_NOP = 0,
+     RTL_CMD_ENTER_PRTM,
+     RTL_CMD_EXIT_PRTM,
+} __attribute__((packed));

You didn't reply to Randy's comment about the packed attribute. I think it's rather confusing here.

Sorry. I missed that comment. The packed attribute on an enum forces it into the smallest int type it can be. enums are normally the size of an int, but this enum in the RTL table must fit in an 8-bit int.

Quoted 9 lines
Quoted 9 lines
drivers/platform/x86/ibm_rtl.c new file +9 -0
+/* The RTL table as presented by the EBDA: */
+struct ibm_rtl_table {
+     char signature[5];
+     u8 version;
+     u8 rt_status;
+     enum rtl_cmd_type command;
+     u8 command_status;
+     enum rtl_addr_type cmd_address_type;
+     u8 cmd_granularity;
Quoted 8 lines
Quoted 5 lines

+     u8 cmd_offset;
+     u16 reserve1;
+     u8 cmd_port_address[4]; /* platform dependent address */
+     u8 cmd_port_value[4];   /* platform dependent value */ +};

I would recommend marking the member in this structure as packed instead, not the enum.

It does not have the same effect. Without the packed attribute on the enums, they end up to be the wrong size and then we would be reading from the wrong location in memory.

Quoted 11 lines

drivers/platform/x86/ibm_rtl.c line 79

Quoted 8 lines
drivers/platform/x86/ibm_rtl.c new file +8 -0
+#define RTL_SIGNATURE (('L'<<24)|('T'<<16)|('R'<<8)|'_')
+
+#define ERROR(A, B...) printk(KERN_ERR "ibm-rtl: " A, ##B )
+#define WARNING(A, B...) printk(KERN_WARNING "ibm-rtl: " A, ##B )
+#define DEBUG(A, B...) do { \
+     if (debug) \
+             printk(KERN_INFO "ibm-rtl: " A, ##B ); \
+} while (0)

We already have wrappers for these, no need to define your own. Please just use dev_{err,warn,dbg} or pr_{err,warning,debug}.

Ack.

Quoted 11 lines

drivers/platform/x86/ibm_rtl.c line 87

Quoted 7 lines
drivers/platform/x86/ibm_rtl.c new file +7 -0
+static DEFINE_MUTEX(rtl_lock);
+static struct ibm_rtl_table __iomem *rtl_table = NULL;
+static void __iomem *ebda_map;
+static void __iomem *rtl_cmd_iomem_addr = NULL;
+static u32 rtl_cmd_port_addr;
+static enum rtl_addr_type rtl_cmd_type;
+static u8 rtl_cmd_width;

This is somewhat inconsistent, some of these are implicitly initialized, others have an explicit "= NULL". I would recommend leaving out the initialization, which is the historic way to do this in the kernel.

The variables that have an explicit initialization are the variables that I read before I write. It just looks funny to me to read a variable that hasn't been initialized. But I can axe the initializations for the sake of consistency.

Quoted 9 lines

Some people find it cleaner to define a structure containing all the driver specific data. Since there can be only one of these devices in your case, it's probably not important.

Quoted 4 lines

+                     if (rtl_cmd_type == RTL_ADDR_TYPE_MMIO)
+                             iowrite8((u8)cmd_port_val, rtl_cmd_iomem_addr);
+                     else
+                             outb((u8)cmd_port_val, rtl_cmd_port_addr);

Quoted 4 lines

ioread/iowrite already has the capability to use both mmio and pio addresses. You can use ioport_map() to create an __iomem token that corresponds to your rtl_cmd_port_addr and get rid of the rtl_cmd_type variable.

Thank you for that tip. I will look into it and roll out another version.

--Vernon

Quoted 7 lines

       Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

#6 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

Quoted 12 lines · Vernon Mauery
Quoted 11 lines

drivers/platform/x86/ibm_rtl.c line 87

Quoted 7 lines
drivers/platform/x86/ibm_rtl.c new file +7 -0
+static DEFINE_MUTEX(rtl_lock);
+static struct ibm_rtl_table __iomem *rtl_table = NULL;
+static void __iomem *ebda_map;
+static void __iomem *rtl_cmd_iomem_addr = NULL;
+static u32 rtl_cmd_port_addr;
+static enum rtl_addr_type rtl_cmd_type;
+static u8 rtl_cmd_width;

This is somewhat inconsistent, some of these are implicitly initialized, others have an explicit "= NULL". I would recommend leaving out the initialization, which is the historic way to do this in the kernel.

Quoted 4 lines

The variables that have an explicit initialization are the variables that I read before I write. It just looks funny to me to read a variable that hasn't been initialized. But I can axe the initializations for the sake of consistency.

Ok. Global variables are by definition initialized to zero. Old gcc versions used to allocate space in the .data section if you had explicit = NULL statements while they would end up in .bss otherwise, resulting in a smaller binary image.

With newer gcc versions, it doesn't make a difference, but many people have the "don't initialize globals to zero" meme hardcoded to their brains now ;-)

Arnd

#7 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

In reply to Vernon Mauery
Quoted 15 lines · On Fri, 24 Sep 2010 07:14:27 -0700 Vernon Mauery wrote:
Quoted 14 lines · Arnd Bergmann

drivers/platform/x86/ibm_rtl.c line 55

Quoted 10 lines · Vernon Mauery
drivers/platform/x86/ibm_rtl.c new file +10 -0
+enum rtl_addr_type {
+     RTL_ADDR_TYPE_IO = 1,
+     RTL_ADDR_TYPE_MMIO,
+} __attribute__((packed));
+
+enum rtl_cmd_type {
+     RTL_CMD_NOP = 0,
+     RTL_CMD_ENTER_PRTM,
+     RTL_CMD_EXIT_PRTM,
+} __attribute__((packed));

You didn't reply to Randy's comment about the packed attribute. I think it's rather confusing here.

thanks.

Quoted 3 lines

Sorry. I missed that comment. The packed attribute on an enum forces it into the smallest int type it can be. enums are normally the size of an int, but this enum in the RTL table must fit in an 8-bit int.

I think you would be better off making 'command' below be a u8 then. Otherwise someone else might add an enum value that is > 255 and the struct size will change. I.e., the current way is error-prone. Just using u8 for command's type is safer.

Quoted 18 lines
Quoted 9 lines
Quoted 9 lines
drivers/platform/x86/ibm_rtl.c new file +9 -0
+/* The RTL table as presented by the EBDA: */
+struct ibm_rtl_table {
+     char signature[5];
+     u8 version;
+     u8 rt_status;
+     enum rtl_cmd_type command;
+     u8 command_status;
+     enum rtl_addr_type cmd_address_type;
+     u8 cmd_granularity;
Quoted 8 lines
Quoted 5 lines

+     u8 cmd_offset;
+     u16 reserve1;
+     u8 cmd_port_address[4]; /* platform dependent address */
+     u8 cmd_port_value[4];   /* platform dependent value */ +};

I would recommend marking the member in this structure as packed instead, not the enum.

Quoted 3 lines

It does not have the same effect. Without the packed attribute on the enums, they end up to be the wrong size and then we would be reading from the wrong location in memory.

Thanks for the explanation.

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

#8 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v5

Quoted 20 lines · Randy Dunlap
Quoted 15 lines · On Fri, 24 Sep 2010 07:14:27 -0700 Vernon Mauery wrote:
Quoted 14 lines · Arnd Bergmann

Review

Quoted 10 lines · Vernon Mauery
(quoted diff) +10 -0
+enum rtl_addr_type {
+     RTL_ADDR_TYPE_IO = 1,
+     RTL_ADDR_TYPE_MMIO,
+} __attribute__((packed));
+
+enum rtl_cmd_type {
+     RTL_CMD_NOP = 0,
+     RTL_CMD_ENTER_PRTM,
+     RTL_CMD_EXIT_PRTM,
+} __attribute__((packed));

You didn't reply to Randy's comment about the packed attribute. I think it's rather confusing here.

thanks.

Quoted 8 lines
Quoted 3 lines

Sorry. I missed that comment. The packed attribute on an enum forces it into the smallest int type it can be. enums are normally the size of an int, but this enum in the RTL table must fit in an 8-bit int.

I think you would be better off making 'command' below be a u8 then. Otherwise someone else might add an enum value that is > 255 and the struct size will change. I.e., the current way is error-prone. Just using u8 for command's type is safer.

Here is a new and improved version that includes your changes as well as Arnd's suggestions. Thank you both for your reviews.

--Vernon

-------------------------------

IBM Real-Time "SMI Free" mode driver

This driver supports the Real-Time Linux (RTL) BIOS feature. The RTL feature allows non-fatal System Management Interrupts (SMIs) to be disabled on supported IBM platforms and is intended to be coupled with a user-space daemon to monitor the hardware in a way that can be prioritized and scheduled to better suit the requirements for the system.

The Device is presented as a special "_RTL_" table to the OS in the Extended BIOS Data Area. There is a simple protocol for entering and exiting the mode at runtime. This driver creates a simple sysfs interface to allow a simple entry and exit from RTL mode in the UFI/BIOS.

Since the driver is specific to IBM SystemX hardware (x86- based servers) it only builds on x86 builds. To reduce the risk of loading on the wrong hardware, the module uses DMI information and checks a list of servers that are known to work.

Documentation/ABI/testing/sysfs-devices-system-ibm-rtl new file +22 -0
new file mode 100644
index 0000000..b82deea
@@ -0,0 +1,22 @@
+What:           state
+Date:           Sep 2010
+KernelVersion:  2.6.37
+Contact:        Vernon Mauery <[email protected]>
+Description:    The state file allows a means by which to change in and
+                out of Premium Real-Time Mode (PRTM), as well as the
+                ability to query the current state.
+                    0 => PRTM off
+                    1 => PRTM enabled
+Users:          The ibm-prtm userspace daemon uses this interface.
+
+
+What:           version
+Date:           Sep 2010
+KernelVersion:  2.6.37
+Contact:        Vernon Mauery <[email protected]>
+Description:    The version file provides a means by which to query
+                the RTL table version that lives in the Extended
+                BIOS Data Area (EBDA).
+Users:          The ibm-prtm userspace daemon uses this interface.
+
+
drivers/platform/x86/Kconfig +0 -0
index cff7cc2..5fba368 100644
@@ -590,4 +590,20 @@ config INTEL_IPS

functionality. If in doubt, say Y here; it will only load on supported platforms.

+config IBM_RTL
+ tristate "Device driver to enable PRTL support"
+ depends on X86 && PCI
+ ---help---

+	  Enable support for IBM Premium Real Time Mode (PRTM).
+	  This module will allow you the enter and exit PRTM in the BIOS via
+	  sysfs on platforms that support this feature.  System in PRTM will
+	  not recive CPU-generated SMIs for recoverable errors.  Use of this
+	  feature without proper support may void your hardware warranty.
+
+	  If the proper BIOS support is found the driver will load and create
+	  /sys/devices/system/ibm_rtl/.  The "state" variable will indicate
+	  whether or not the BIOS is in PRTM.
+	  state = 0 (BIOS SMIs on)
+	  state = 1 (BIOS SMIs off)

+
endif # X86_PLATFORM_DEVICES

drivers/platform/x86/Makefile +0 -0
index 85fb2b8..50fae7a 100644
@@ -30,4 +30,5 @@ obj-$(CONFIG_INTEL_SCU_IPC)	+= intel_scu_ipc.o

obj-$(CONFIG_RAR_REGISTER) += intel_rar_register.o
obj-$(CONFIG_INTEL_IPS) += intel_ips.o
obj-$(CONFIG_GPIO_INTEL_PMIC) += intel_pmic_gpio.o
+obj-$(CONFIG_IBM_RTL) += ibm_rtl.o

drivers/platform/x86/ibm_rtl.c new file +339 -0

339 lines not on this page · raw message

#9 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v5

Quoted 6 lines · On Fri, 24 Sep 2010 14:06:42 -0700 Vernon Mauery wrote:
drivers/platform/x86/Kconfig +0 -0
index cff7cc2..5fba368 100644
@@ -590,4 +590,20 @@ config INTEL_IPS
 	  functionality.  If in doubt, say Y here; it will only load on

drivers/platform/x86/Kconfig config INTEL_IPS · line 600

Quoted 10 lines
drivers/platform/x86/Kconfig +8 -0
 	  supported platforms.
 
+config IBM_RTL
+	tristate "Device driver to enable PRTL support"
+	depends on X86 && PCI
+	---help---
+	  Enable support for IBM Premium Real Time Mode (PRTM).
+	  This module will allow you the enter and exit PRTM in the BIOS via
+	  sysfs on platforms that support this feature.  System in PRTM will
+	  not recive CPU-generated SMIs for recoverable errors.  Use of this

receive (again/still)

Quoted 40 lines
drivers/platform/x86/Kconfig +8 -0
+	  feature without proper support may void your hardware warranty.
+
+	  If the proper BIOS support is found the driver will load and create
+	  /sys/devices/system/ibm_rtl/.  The "state" variable will indicate
+	  whether or not the BIOS is in PRTM.
+	  state = 0 (BIOS SMIs on)
+	  state = 1 (BIOS SMIs off)
+
 endif # X86_PLATFORM_DEVICES
drivers/platform/x86/ibm_rtl.c new file +24 -0
new file mode 100644
index 0000000..2a0a390
@@ -0,0 +1,339 @@
+/*
+ * IBM Real-Time Linux driver
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright (C) IBM Corporation, 2010
+ *
+ * Author: Keith Mannthey <[email protected]>
+ *         Vernon Mauery <[email protected]>
+ *
+ */
+

drivers/platform/x86/ibm_rtl.c line 34

Quoted 10 lines
drivers/platform/x86/ibm_rtl.c new file +10 -0
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/module.h>
+#include <linux/io.h>
+#include <linux/sysdev.h>
+#include <linux/dmi.h>
+#include <linux/mutex.h>
+#include <asm/bios_ebda.h>
+
+static int force;

s/int/bool/

drivers/platform/x86/ibm_rtl.c line 38

Quoted 4 lines
drivers/platform/x86/ibm_rtl.c new file +4 -0
+module_param(force, bool, 0);
+MODULE_PARM_DESC(force, "Force driver load, ignore DMI data");
+
+static int debug;

s/int/bool/

Quoted 6 lines
drivers/platform/x86/ibm_rtl.c new file +6 -0
+module_param(debug, bool, 0644);
+MODULE_PARM_DESC(debug, "Show debug output");
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Keith Mannthey <[email protected]>");
+MODULE_AUTHOR("Vernon Mauery <[email protected]>");

drivers/platform/x86/ibm_rtl.c line 54

Quoted 10 lines
drivers/platform/x86/ibm_rtl.c new file +10 -0
+
+#define RTL_ADDR_TYPE_IO    1
+#define RTL_ADDR_TYPE_MMIO  2
+
+#define RTL_CMD_ENTER_PRTM  1
+#define RTL_CMD_EXIT_PRTM   2
+
+/* The RTL table as presented by the EBDA: */
+struct ibm_rtl_table {
+	char signature[5];

Question: are all 5 bytes of signature[] used? It looks like the search code only checks for 32 bits of signature (and the last byte can be garbage?).

Quoted 25 lines
drivers/platform/x86/ibm_rtl.c new file +25 -0
+	u8 version;
+	u8 rt_status;
+	u8 command;
+	u8 command_status;
+	u8 cmd_address_type;
+	u8 cmd_granularity;
+	u8 cmd_offset;
+	u16 reserve1;
+	u32 cmd_port_address; /* platform dependent address */
+	u32 cmd_port_value;   /* platform dependent value */
+} __attribute__((packed));
+
+#define RTL_SIGNATURE (('L'<<24)|('T'<<16)|('R'<<8)|'_')
+
+#define RTL_DEBUG(A, ...) do { \
+	if (debug) \
+		pr_info("ibm-rtl: " A, ##__VA_ARGS__ ); \
+} while (0)
+
+static DEFINE_MUTEX(rtl_lock);
+static struct ibm_rtl_table __iomem *rtl_table;
+static void __iomem *ebda_map;
+static void __iomem *rtl_cmd_addr;
+static u8 rtl_cmd_type;
+static u8 rtl_cmd_width;

...

drivers/platform/x86/ibm_rtl.c line 256

Quoted 7 lines
drivers/platform/x86/ibm_rtl.c new file +7 -0
+static int __init ibm_rtl_init(void) {
+	unsigned long ebda_addr, ebda_size;
+	unsigned int ebda_kb;
+	int ret = -ENODEV, i;
+
+	if (force)
+		pr_warning("ibm-rtl: module loaded by force!\n");

Please drop the '!'.

Quoted 57 lines
drivers/platform/x86/ibm_rtl.c new file +57 -0
+	/* first ensure that we are running on IBM HW */
+	else if (!dmi_check_system(ibm_rtl_dmi_table))
+		return -ENODEV;
+
+	/* Get the address for the Extended BIOS Data Area */
+	ebda_addr = get_bios_ebda();
+	if (!ebda_addr) {
+		RTL_DEBUG("no BIOS EBDA found\n");
+		return -ENODEV;
+	}
+
+	ebda_map = ioremap(ebda_addr, 4);
+	if (!ebda_map)
+		return -ENOMEM;
+
+	/* First word in the EDBA is the Size in KB */
+	ebda_kb = ioread16(ebda_map);
+	RTL_DEBUG("EBDA is %d kB\n", ebda_kb);
+
+	if (ebda_kb == 0)
+		goto out;
+
+	iounmap(ebda_map);
+	ebda_size = ebda_kb*1024;
+
+	/* Remap the whole table */
+	ebda_map = ioremap(ebda_addr, ebda_size);
+	if (!ebda_map)
+		return -ENOMEM;
+
+	/* search for the _RTL_ signature at the start of the table */
+	for (i = 0 ; i < ebda_size/sizeof(unsigned int); i++) {
+		struct ibm_rtl_table __iomem * tmp;
+		tmp = (struct ibm_rtl_table __iomem *) (ebda_map+i);
+		if (ioread32(&tmp->signature) == RTL_SIGNATURE) {
+			phys_addr_t addr;
+			unsigned int plen;
+			RTL_DEBUG("found RTL_SIGNATURE at %#llx\n", (u64)tmp);
+			rtl_table = tmp;
+			/* The address, value, width and offset are platform
+			 * dependent and found in the ibm_rtl_table */
+			rtl_cmd_width = ioread8(&rtl_table->cmd_granularity);
+			rtl_cmd_type = ioread8(&rtl_table->cmd_address_type);
+			RTL_DEBUG("rtl_cmd_width = %u, rtl_cmd_type = %u\n",
+			      rtl_cmd_width, rtl_cmd_type);
+			addr = ioread32(&rtl_table->cmd_port_address);
+			RTL_DEBUG("addr = %#llx\n", addr);
+			plen = rtl_cmd_width/sizeof(char);
+			rtl_cmd_addr = rtl_port_map(addr, plen);
+			RTL_DEBUG("rtl_cmd_addr = %#llx\n", (u64)rtl_cmd_addr);
+			if (!rtl_cmd_addr) {
+				ret = -ENOMEM;
+				break;
+			}
+			ret = rtl_setup_sysfs();
+			break;
+		}

drivers/platform/x86/ibm_rtl.c line 323

Quoted 10 lines
drivers/platform/x86/ibm_rtl.c new file +10 -0
+	}
+
+out:
+	if (ret) {
+		iounmap(ebda_map);
+		rtl_port_unmap(rtl_cmd_addr);
+	}
+
+	return ret;
+}

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

#10 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v5

Quoted 10 lines · Randy Dunlap
Quoted 8 lines · On Fri, 24 Sep 2010 14:06:42 -0700 Vernon Mauery wrote:
drivers/platform/x86/Kconfig +0 -0
index cff7cc2..5fba368 100644
@@ -590,4 +590,20 @@ config INTEL_IPS
 	  functionality.  If in doubt, say Y here; it will only load on
 	  supported platforms.
 
Quoted 10 lines

drivers/platform/x86/Kconfig config INTEL_IPS · line 600

Quoted 8 lines
drivers/platform/x86/Kconfig +8 -0
+config IBM_RTL
+	tristate "Device driver to enable PRTL support"
+	depends on X86 && PCI
+	---help---
+	  Enable support for IBM Premium Real Time Mode (PRTM).
+	  This module will allow you the enter and exit PRTM in the BIOS via
+	  sysfs on platforms that support this feature.  System in PRTM will
+	  not recive CPU-generated SMIs for recoverable errors.  Use of this

receive (again/still)

Ack. Again.

Quoted 5 lines
Quoted 5 lines
drivers/platform/x86/Kconfig +5 -0
+	  feature without proper support may void your hardware warranty.
+
+	  If the proper BIOS support is found the driver will load and create
+	  /sys/devices/system/ibm_rtl/.  The "state" variable will indicate
+	  whether or not the BIOS is in PRTM.

[snip]

Quoted 12 lines

drivers/platform/x86/Kconfig

Quoted 9 lines
drivers/platform/x86/Kconfig +9 -0
+#include <linux/delay.h>
+#include <linux/module.h>
+#include <linux/io.h>
+#include <linux/sysdev.h>
+#include <linux/dmi.h>
+#include <linux/mutex.h>
+#include <asm/bios_ebda.h>
+
+static int force;

s/int/bool/

Quoted 6 lines

drivers/platform/x86/Kconfig

Quoted 4 lines
drivers/platform/x86/Kconfig +4 -0
+module_param(force, bool, 0);
+MODULE_PARM_DESC(force, "Force driver load, ignore DMI data");
+
+static int debug;

s/int/bool/

Ack.

Quoted 11 lines
Quoted 11 lines
drivers/platform/x86/Kconfig +11 -0
+module_param(debug, bool, 0644);
+MODULE_PARM_DESC(debug, "Show debug output");
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Keith Mannthey <[email protected]>");
+MODULE_AUTHOR("Vernon Mauery <[email protected]>");
+
+#define RTL_ADDR_TYPE_IO    1
+#define RTL_ADDR_TYPE_MMIO  2
+
+#define RTL_CMD_ENTER_PRTM  1
Quoted 8 lines

drivers/platform/x86/Kconfig

Quoted 5 lines
drivers/platform/x86/Kconfig +5 -0
+#define RTL_CMD_EXIT_PRTM   2
+
+/* The RTL table as presented by the EBDA: */
+struct ibm_rtl_table {
+	char signature[5];

Question: are all 5 bytes of signature[] used? It looks like the search code only checks for 32 bits of signature (and the last byte can be garbage?).

Technically, the 5 bytes are supposed to be "_RTL_", but it is a lot easier to just check _RTL since we can do that with a single read. I suppose I could do a 64-bit read and mask the bits that matter. Do you have an opinion on the matter?

Quoted 28 lines
Quoted 25 lines
drivers/platform/x86/Kconfig +25 -0
+	u8 version;
+	u8 rt_status;
+	u8 command;
+	u8 command_status;
+	u8 cmd_address_type;
+	u8 cmd_granularity;
+	u8 cmd_offset;
+	u16 reserve1;
+	u32 cmd_port_address; /* platform dependent address */
+	u32 cmd_port_value;   /* platform dependent value */
+} __attribute__((packed));
+
+#define RTL_SIGNATURE (('L'<<24)|('T'<<16)|('R'<<8)|'_')
+
+#define RTL_DEBUG(A, ...) do { \
+	if (debug) \
+		pr_info("ibm-rtl: " A, ##__VA_ARGS__ ); \
+} while (0)
+
+static DEFINE_MUTEX(rtl_lock);
+static struct ibm_rtl_table __iomem *rtl_table;
+static void __iomem *ebda_map;
+static void __iomem *rtl_cmd_addr;
+static u8 rtl_cmd_type;
+static u8 rtl_cmd_width;

...

Quoted 9 lines

drivers/platform/x86/Kconfig

Quoted 7 lines
drivers/platform/x86/Kconfig +7 -0
+static int __init ibm_rtl_init(void) {
+	unsigned long ebda_addr, ebda_size;
+	unsigned int ebda_kb;
+	int ret = -ENODEV, i;
+
+	if (force)
+		pr_warning("ibm-rtl: module loaded by force!\n");

Please drop the '!'.

Ack.

--Vernon

#11 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v5

Quoted 11 lines · On Fri, 24 Sep 2010 14:30:09 -0700 Vernon Mauery wrote:
Quoted 6 lines

drivers/platform/x86/ibm_rtl.c line 54

Quoted 3 lines
drivers/platform/x86/ibm_rtl.c new file +3 -0
+/* The RTL table as presented by the EBDA: */
+struct ibm_rtl_table {
+	char signature[5];

Question: are all 5 bytes of signature[] used? It looks like the search code only checks for 32 bits of signature (and the last byte can be garbage?).

Technically, the 5 bytes are supposed to be "_RTL_", but it is a lot easier to just check _RTL since we can do that with a single read. I suppose I could do a 64-bit read and mask the bits that matter. Do you have an opinion on the matter?

I would just comment that the first 4 bytes are checked.

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

#12 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v6

Quoted 9 lines · Randy Dunlap
Quoted 7 lines · On Fri, 24 Sep 2010 14:30:09 -0700 Vernon Mauery wrote:
Quoted 6 lines

Review

Quoted 3 lines
(quoted diff) +3 -0
+/* The RTL table as presented by the EBDA: */
+struct ibm_rtl_table {
+	char signature[5];

Question: are all 5 bytes of signature[] used? It looks like the search code only checks for 32 bits of signature (and the last byte can be garbage?).

Quoted 6 lines
Quoted 4 lines

Technically, the 5 bytes are supposed to be "_RTL_", but it is a lot easier to just check _RTL since we can do that with a single read. I suppose I could do a 64-bit read and mask the bits that matter. Do you have an opinion on the matter?

I would just comment that the first 4 bytes are checked.

Okay. I think I have something that should pass. Thanks again for your review.

--Vernon

-----------------------------------------------------

IBM Real-Time "SMI Free" mode driver

This driver supports the Real-Time Linux (RTL) BIOS feature. The RTL feature allows non-fatal System Management Interrupts (SMIs) to be disabled on supported IBM platforms and is intended to be coupled with a user-space daemon to monitor the hardware in a way that can be prioritized and scheduled to better suit the requirements for the system.

The Device is presented as a special "_RTL_" table to the OS in the Extended BIOS Data Area. There is a simple protocol for entering and exiting the mode at runtime. This driver creates a simple sysfs interface to allow a simple entry and exit from RTL mode in the UFI/BIOS.

Since the driver is specific to IBM SystemX hardware (x86- based servers) it only builds on x86 builds. To reduce the risk of loading on the wrong hardware, the module uses DMI information and checks a list of servers that are known to work.

Documentation/ABI/testing/sysfs-devices-system-ibm-rtl new file +22 -0
new file mode 100644
index 0000000..b82deea
@@ -0,0 +1,22 @@
+What:           state
+Date:           Sep 2010
+KernelVersion:  2.6.37
+Contact:        Vernon Mauery <[email protected]>
+Description:    The state file allows a means by which to change in and
+                out of Premium Real-Time Mode (PRTM), as well as the
+                ability to query the current state.
+                    0 => PRTM off
+                    1 => PRTM enabled
+Users:          The ibm-prtm userspace daemon uses this interface.
+
+
+What:           version
+Date:           Sep 2010
+KernelVersion:  2.6.37
+Contact:        Vernon Mauery <[email protected]>
+Description:    The version file provides a means by which to query
+                the RTL table version that lives in the Extended
+                BIOS Data Area (EBDA).
+Users:          The ibm-prtm userspace daemon uses this interface.
+
+
drivers/platform/x86/Kconfig +16 -0
index cff7cc2..55c6b86 100644
@@ -590,4 +590,20 @@ config INTEL_IPS
 	  functionality.  If in doubt, say Y here; it will only load on
 	  supported platforms.
 
+config IBM_RTL
+	tristate "Device driver to enable PRTL support"
+	depends on X86 && PCI
+	---help---
+	  Enable support for IBM Premium Real Time Mode (PRTM).
+	  This module will allow you the enter and exit PRTM in the BIOS via
+	  sysfs on platforms that support this feature.  System in PRTM will
+	  not receive CPU-generated SMIs for recoverable errors.  Use of this
+	  feature without proper support may void your hardware warranty.
+
+	  If the proper BIOS support is found the driver will load and create
+	  /sys/devices/system/ibm_rtl/.  The "state" variable will indicate
+	  whether or not the BIOS is in PRTM.
+	  state = 0 (BIOS SMIs on)
+	  state = 1 (BIOS SMIs off)
+
 endif # X86_PLATFORM_DEVICES
drivers/platform/x86/Makefile +1 -0
index 85fb2b8..50fae7a 100644
@@ -30,4 +30,5 @@ obj-$(CONFIG_INTEL_SCU_IPC)	+= intel_scu_ipc.o
 obj-$(CONFIG_RAR_REGISTER)	+= intel_rar_register.o
 obj-$(CONFIG_INTEL_IPS)		+= intel_ips.o
 obj-$(CONFIG_GPIO_INTEL_PMIC)	+= intel_pmic_gpio.o
+obj-$(CONFIG_IBM_RTL)		+= ibm_rtl.o
 
drivers/platform/x86/ibm_rtl.c new file +341 -0

341 lines not on this page · raw message

#13 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v5

In reply to Vernon Mauery
Quoted 10 lines · Vernon Mauery
Quoted 6 lines

drivers/platform/x86/ibm_rtl.c line 54

Quoted 3 lines
drivers/platform/x86/ibm_rtl.c new file +3 -0
+/* The RTL table as presented by the EBDA: */
+struct ibm_rtl_table {
+	char signature[5];

Question: are all 5 bytes of signature[] used? It looks like the search code only checks for 32 bits of signature (and the last byte can be garbage?).

Technically, the 5 bytes are supposed to be "_RTL_", but it is a lot easier to just check _RTL since we can do that with a single read. I suppose I could do a 64-bit read and mask the bits that matter.

Well, as long as you guys write in stone there at IBM that you will remember to not have a possible future _RTL2_ that would clash horribly with an non-up-to-date version of the driver that has no clue about RTL2 and still would match the signature because it only looks at the first 32 bits...

Signature · 4 lines
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh
#14 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v6

Quoted 12 lines · Henrique de Moraes Holschuh
Quoted 10 lines · Vernon Mauery
Quoted 6 lines

Review

Quoted 3 lines
(quoted diff) +3 -0
+/* The RTL table as presented by the EBDA: */
+struct ibm_rtl_table {
+	char signature[5];

Question: are all 5 bytes of signature[] used? It looks like the search code only checks for 32 bits of signature (and the last byte can be garbage?).

Technically, the 5 bytes are supposed to be "_RTL_", but it is a lot easier to just check _RTL since we can do that with a single read. I suppose I could do a 64-bit read and mask the bits that matter.

Quoted 5 lines

Well, as long as you guys write in stone there at IBM that you will remember to not have a possible future _RTL2_ that would clash horribly with an non-up-to-date version of the driver that has no clue about RTL2 and still would match the signature because it only looks at the first 32 bits...

I do not have a strong opinion either way. Yes, a 5-byte compare is more precise, but 4-bytes is very likely to be unique. There is a spec (written in stone, as it were) that tells what the table looks like. It says a 5-byte signature followed by a 1-byte version number. Currently there is only one version, but rather than change the signature on future versions, we will change the version number.

That said, here is a new version that does a 5-byte masked quad-word compare to find the signature.

--Vernon

IBM Real-Time "SMI Free" mode driver

This driver supports the Real-Time Linux (RTL) BIOS feature. The RTL feature allows non-fatal System Management Interrupts (SMIs) to be disabled on supported IBM platforms and is intended to be coupled with a user-space daemon to monitor the hardware in a way that can be prioritized and scheduled to better suit the requirements for the system.

The Device is presented as a special "_RTL_" table to the OS in the Extended BIOS Data Area. There is a simple protocol for entering and exiting the mode at runtime. This driver creates a simple sysfs interface to allow a simple entry and exit from RTL mode in the UFI/BIOS.

Since the driver is specific to IBM SystemX hardware (x86- based servers) it only builds on x86 builds. To reduce the risk of loading on the wrong hardware, the module uses DMI information and checks a list of servers that are known to work.

Documentation/ABI/testing/sysfs-devices-system-ibm-rtl new file +22 -0
new file mode 100644
index 0000000..b82deea
@@ -0,0 +1,22 @@
+What:           state
+Date:           Sep 2010
+KernelVersion:  2.6.37
+Contact:        Vernon Mauery <[email protected]>
+Description:    The state file allows a means by which to change in and
+                out of Premium Real-Time Mode (PRTM), as well as the
+                ability to query the current state.
+                    0 => PRTM off
+                    1 => PRTM enabled
+Users:          The ibm-prtm userspace daemon uses this interface.
+
+
+What:           version
+Date:           Sep 2010
+KernelVersion:  2.6.37
+Contact:        Vernon Mauery <[email protected]>
+Description:    The version file provides a means by which to query
+                the RTL table version that lives in the Extended
+                BIOS Data Area (EBDA).
+Users:          The ibm-prtm userspace daemon uses this interface.
+
+
drivers/platform/x86/Kconfig +16 -0
index cff7cc2..55c6b86 100644
@@ -590,4 +590,20 @@ config INTEL_IPS
 	  functionality.  If in doubt, say Y here; it will only load on
 	  supported platforms.
 
+config IBM_RTL
+	tristate "Device driver to enable PRTL support"
+	depends on X86 && PCI
+	---help---
+	  Enable support for IBM Premium Real Time Mode (PRTM).
+	  This module will allow you the enter and exit PRTM in the BIOS via
+	  sysfs on platforms that support this feature.  System in PRTM will
+	  not receive CPU-generated SMIs for recoverable errors.  Use of this
+	  feature without proper support may void your hardware warranty.
+
+	  If the proper BIOS support is found the driver will load and create
+	  /sys/devices/system/ibm_rtl/.  The "state" variable will indicate
+	  whether or not the BIOS is in PRTM.
+	  state = 0 (BIOS SMIs on)
+	  state = 1 (BIOS SMIs off)
+
 endif # X86_PLATFORM_DEVICES
drivers/platform/x86/Makefile +1 -0
index 85fb2b8..50fae7a 100644
@@ -30,4 +30,5 @@ obj-$(CONFIG_INTEL_SCU_IPC)	+= intel_scu_ipc.o
 obj-$(CONFIG_RAR_REGISTER)	+= intel_rar_register.o
 obj-$(CONFIG_INTEL_IPS)		+= intel_ips.o
 obj-$(CONFIG_GPIO_INTEL_PMIC)	+= intel_pmic_gpio.o
+obj-$(CONFIG_IBM_RTL)		+= ibm_rtl.o
 
drivers/platform/x86/ibm_rtl.c new file +341 -0

341 lines not on this page · raw message

#15 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

In reply to Vernon Mauery
Quoted 2 lines · Vernon Mauery
Quoted 1 line · Arnd Bergmann

On Friday 24 September 2010, Vernon Mauery wrote:

[snip]

Quoted 11 lines
Quoted 9 lines
Quoted 4 lines

+                     if (rtl_cmd_type == RTL_ADDR_TYPE_MMIO)
+                             iowrite8((u8)cmd_port_val, rtl_cmd_iomem_addr);
+                     else
+                             outb((u8)cmd_port_val, rtl_cmd_port_addr);

ioread/iowrite already has the capability to use both mmio and pio addresses. You can use ioport_map() to create an __iomem token that corresponds to your rtl_cmd_port_addr and get rid of the rtl_cmd_type variable.

Thank you for that tip. I will look into it and roll out another version.

I looked into this and tested it on some hardware, but it doesn't work. After more digging and poking, it looks like the reason is that the port IO address is not within the x86 standard port IO range.

I tried something like this:

			addr = ioread32(&rtl_table->cmd_port_address);
			plen = rtl_cmd_width/8;
			if (rtl_cmd_type == RTL_ADDR_TYPE_MMIO)
				rtl_cmd_addr = ioremap(addr, plen);
			else
				rtl_cmd_addr = ioport_map(addr, plen);
			RTL_DEBUG("rtl_cmd_addr = %#llx\n", (u64)rtl_cmd_addr);

It printed out that rtl_cmd_addr was 0, meaning the ioport_map failed. After more digging, it turns out that on at least one of the machines this code is targeted for, the port IO address (from the first line above) is 0x40000. Even if this did get mapped, the IO_COND macro would target it for MMIO access instead of PIO access. So I don't think I can use this method (even though it did make my code *a lot* nicer to read).

Any suggestions?

--Vernon

#16 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

Quoted 14 lines · On Friday 24 September 2010 19:09:43 Vernon Mauery wrote:

I looked into this and tested it on some hardware, but it doesn't work. After more digging and poking, it looks like the reason is that the port IO address is not within the x86 standard port IO range.

I tried something like this:

                        addr = ioread32(&rtl_table->cmd_port_address);
                        plen = rtl_cmd_width/8;
                        if (rtl_cmd_type == RTL_ADDR_TYPE_MMIO)
                                rtl_cmd_addr = ioremap(addr, plen);
                        else
                                rtl_cmd_addr = ioport_map(addr, plen);
                        RTL_DEBUG("rtl_cmd_addr = %#llx\n", (u64)rtl_cmd_addr);
Quoted 8 lines

It printed out that rtl_cmd_addr was 0, meaning the ioport_map failed. After more digging, it turns out that on at least one of the machines this code is targeted for, the port IO address (from the first line above) is 0x40000. Even if this did get mapped, the IO_COND macro would target it for MMIO access instead of PIO access. So I don't think I can use this method (even though it did make my code a lot nicer to read).

Any suggestions?

That seems really strange. I thought the inb/outb instructions could not actually operate on addresses above 0x10000 at all, since they take a 16 bit address operand (DX register). Passing 0x40000 into inb should have the same effect as zero AFAICT, which means that your existing code should not work either.

For non-x86 architectures, I would recommend defining HAVE_ARCH_PIO_SIZE and setting PIO_RESERVED to a higher value.

Arnd

#17 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

Quoted 22 lines · Arnd Bergmann
Quoted 21 lines · On Friday 24 September 2010 19:09:43 Vernon Mauery wrote:

I looked into this and tested it on some hardware, but it doesn't work. After more digging and poking, it looks like the reason is that the port IO address is not within the x86 standard port IO range.

I tried something like this:

                        addr = ioread32(&rtl_table->cmd_port_address);
                        plen = rtl_cmd_width/8;
                        if (rtl_cmd_type == RTL_ADDR_TYPE_MMIO)
                                rtl_cmd_addr = ioremap(addr, plen);
                        else
                                rtl_cmd_addr = ioport_map(addr, plen);
                        RTL_DEBUG("rtl_cmd_addr = %#llx\n", (u64)rtl_cmd_addr);

It printed out that rtl_cmd_addr was 0, meaning the ioport_map failed. After more digging, it turns out that on at least one of the machines this code is targeted for, the port IO address (from the first line above) is 0x40000. Even if this did get mapped, the IO_COND macro would target it for MMIO access instead of PIO access. So I don't think I can use this method (even though it did make my code a lot nicer to read).

Quoted 7 lines
Quoted 1 line

Any suggestions?

That seems really strange. I thought the inb/outb instructions could not actually operate on addresses above 0x10000 at all, since they take a 16 bit address operand (DX register). Passing 0x40000 into inb should have the same effect as zero AFAICT, which means that your existing code should not work either.

No, inb/outb have address as as unsigned long, which would explain why 0x40000 works with PIO. When doing inb/outb via the ioread/iowrite macros, the port value is the address (minus the offset) masked off to PIO_MASK, which is 16 bits on x86.

So it looks like this really not going to work and I will have to go back to how it was before. Would it be tacky to write my own little macro?

Quoted 2 lines

For non-x86 architectures, I would recommend defining HAVE_ARCH_PIO_SIZE and setting PIO_RESERVED to a higher value.

Unfortunately, this is targeted only for x86 platforms.

--Vernon

#18 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

Quoted 4 lines · On Friday 24 September 2010 20:23:27 Vernon Mauery wrote:

No, inb/outb have address as as unsigned long, which would explain why 0x40000 works with PIO. When doing inb/outb via the ioread/iowrite macros, the port value is the address (minus the offset) masked off to PIO_MASK, which is 16 bits on x86.

What I meant is that the inb/outb *instructions* only take a 16 bit argument, not the kernel functions. You should really try to find out what is going on there.

Arnd

#19 lore raw

Re: [RFC][Patch] IBM Real-Time "SMI Free" mode driver -v4

In reply to Vernon Mauery
Quoted 31 lines · Vernon Mauery
Quoted 22 lines · Arnd Bergmann
Quoted 21 lines · On Friday 24 September 2010 19:09:43 Vernon Mauery wrote:

I looked into this and tested it on some hardware, but it doesn't work. After more digging and poking, it looks like the reason is that the port IO address is not within the x86 standard port IO range.

I tried something like this:

                       addr = ioread32(&rtl_table->cmd_port_address);
                       plen = rtl_cmd_width/8;
                       if (rtl_cmd_type == RTL_ADDR_TYPE_MMIO)
                               rtl_cmd_addr = ioremap(addr, plen);
                       else
                               rtl_cmd_addr = ioport_map(addr, plen);
                       RTL_DEBUG("rtl_cmd_addr = %#llx\n", (u64)rtl_cmd_addr);

It printed out that rtl_cmd_addr was 0, meaning the ioport_map failed. After more digging, it turns out that on at least one of the machines this code is targeted for, the port IO address (from the first line above) is 0x40000. Even if this did get mapped, the IO_COND macro would target it for MMIO access instead of PIO access. So I don't think I can use this method (even though it did make my code a lot nicer to read).

Quoted 7 lines
Quoted 1 line

Any suggestions?

That seems really strange. I thought the inb/outb instructions could not actually operate on addresses above 0x10000 at all, since they take a 16 bit address operand (DX register). Passing 0x40000 into inb should have the same effect as zero AFAICT, which means that your existing code should not work either.

Quoted 3 lines

No, inb/outb have address as as unsigned long, which would explain why 0x40000 works with PIO. When doing inb/outb via the ioread/iowrite macros, the port value is the address (minus the offset) masked off to PIO_MASK, which is 16 bits on x86.

So it looks like this really not going to work and I will have to go back to how it was before. Would it be tacky to write my own little macro?

Okay. I finally have it. In one of my changes, I inadvertently changed the shape of the struct ibm_rtl_table, which is where I read the ioport address from. This explains why the weird port. Once I fixed my table so it was correct again, the port is 0x600, which is completely reasonable and will work just fine with the iowrite functions. Thanks for bringing this up -- it greatly simplifies things.

--Vernon