Skip to contentSkip to conversation

i5k_amb: Support Intel 5400 chipset

PATCH In review Darrick J. Wong 3 participants 11 messages May 3, 2008 – May 7, 2008

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 i5k_amb: Support Intel 5400 chipset

Minor rework to support the Intel 5400 chipset.

---

 drivers/hwmon/i5k_amb.c |   39 ++++++++++++++++++++++++++++++++++-----
 include/linux/pci_ids.h |    3 +++
 2 files changed, 37 insertions(+), 5 deletions(-)
drivers/hwmon/i5k_amb.c +34 -5
index 6ac5c6f..f9e2ed6 100644
@@ -111,6 +111,7 @@ struct i5k_amb_data {
 	void __iomem *amb_mmio;
 	struct i5k_device_attribute *attrs;
 	unsigned int num_attrs;
+	unsigned long chipset_id;
 };
 
 static ssize_t show_name(struct device *dev, struct device_attribute *devattr,
@@ -382,7 +383,8 @@ err:
 	return res;
 }
 
-static int __devinit i5k_find_amb_registers(struct i5k_amb_data *data)
+static int __devinit i5k_find_amb_registers(struct i5k_amb_data *data,
+					    unsigned long devid)
 {
 	struct pci_dev *pcidev;
 	u32 val32;
@@ -390,7 +392,7 @@ static int __devinit i5k_find_amb_registers(struct i5k_amb_data *data)
 
 	/* Find AMB register memory space */
 	pcidev = pci_get_device(PCI_VENDOR_ID_INTEL,
-				PCI_DEVICE_ID_INTEL_5000_ERR,
+				devid,
 				NULL);
 	if (!pcidev)
 		return -ENODEV;
@@ -409,6 +411,8 @@ static int __devinit i5k_find_amb_registers(struct i5k_amb_data *data)
 		goto out;
 	}
 
+	data->chipset_id = devid;
+
 	res = 0;
 out:
 	pci_dev_put(pcidev);
@@ -441,10 +445,30 @@ out:
 	return res;
 }
 
+static unsigned long i5k_channel_pci_id(struct i5k_amb_data *data,
+					unsigned long channel)
+{
+	switch (data->chipset_id) {
+	case PCI_DEVICE_ID_INTEL_5000_ERR:
+		return PCI_DEVICE_ID_INTEL_5000_FBD0 + channel;
+	case PCI_DEVICE_ID_INTEL_5400_ERR:
+		return PCI_DEVICE_ID_INTEL_5400_FBD0 + channel;
+	default:
+		BUG();
+	}
+}
+
+static unsigned long chipset_ids[] = {
+	PCI_DEVICE_ID_INTEL_5000_ERR,
+	PCI_DEVICE_ID_INTEL_5400_ERR,
+	0
+};
+
 static int __devinit i5k_amb_probe(struct platform_device *pdev)
 {
 	struct i5k_amb_data *data;
 	struct resource *reso;
+	int i;
 	int res = -ENODEV;
 
 	data = kzalloc(sizeof(*data), GFP_KERNEL);
@@ -452,19 +476,24 @@ static int __devinit i5k_amb_probe(struct platform_device *pdev)
 		return -ENOMEM;
 
 	/* Figure out where the AMB registers live */
-	res = i5k_find_amb_registers(data);
+	i = 0;
+	do {
+		res = i5k_find_amb_registers(data, chipset_ids[i]);
+		i++;
+	} while (res && chipset_ids[i]);
+
 	if (res)
 		goto err;
 
 	/* Copy the DIMM presence map for the first two channels */
 	res = i5k_channel_probe(&data->amb_present[0],
-				PCI_DEVICE_ID_INTEL_5000_FBD0);
+				i5k_channel_pci_id(data, 0));
 	if (res)
 		goto err;
 
 	/* Copy the DIMM presence map for the optional second two channels */
 	i5k_channel_probe(&data->amb_present[2],
-			  PCI_DEVICE_ID_INTEL_5000_FBD1);
+			  i5k_channel_pci_id(data, 1));
 
 	/* Set up resource regions */
 	reso = request_mem_region(data->amb_base, data->amb_len, DRVNAME);
include/linux/pci_ids.h +3 -0
index 70eb3c8..bb6e5e2 100644
@@ -2382,6 +2382,9 @@
 #define PCI_DEVICE_ID_INTEL_ICH10_4	0x3a30
 #define PCI_DEVICE_ID_INTEL_ICH10_5	0x3a60
 #define PCI_DEVICE_ID_INTEL_IOAT_SNB	0x402f
+#define PCI_DEVICE_ID_INTEL_5400_ERR	0x4030
+#define PCI_DEVICE_ID_INTEL_5400_FBD0	0x4035
+#define PCI_DEVICE_ID_INTEL_5400_FBD1	0x4036
 #define PCI_DEVICE_ID_INTEL_IOAT_SCNB	0x65ff
 #define PCI_DEVICE_ID_INTEL_TOLAPAI_0	0x5031
 #define PCI_DEVICE_ID_INTEL_TOLAPAI_1	0x5032
#1 lore raw
Quoted 1 line · On Fri, 2 May 2008 21:55:59 -0700 "Darrick J. Wong" <[email protected]> wrote:

Minor rework to support the Intel 5400 chipset.

Thanks, I added this to the for-git-hwmon part of the -mm queue.

I haven't heard from Mark for quite some time so soon I might declare him absent and I'll go through my hwmon patches with a view to merging them into 2.6.26 directly.

That queue currently consists of:

hdaps-invert-the-axes-for-hdaps-on-lenovo-r61i-thinkpads.patch adt746x-logical-bitwise-confusion-in-set_max_duty_at_crit.patch adt7473-minor-documentation-update.patch i5k_amb-support-intel-5400-chipset.patch

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#2 lore raw

Hi Andrew,

Quoted 10 lines · Andrew Morton
Quoted 1 line · On Fri, 2 May 2008 21:55:59 -0700 "Darrick J. Wong" <[email protected]> wrote:

Minor rework to support the Intel 5400 chipset.

Thanks, I added this to the for-git-hwmon part of the -mm queue.

I haven't heard from Mark for quite some time so soon I might declare him absent and I'll go through my hwmon patches with a view to merging them into 2.6.26 directly.

For Mark's status see:
http://lists.lm-sensors.org/pipermail/lm-sensors/2008-May/023006.html

I guess he'll be happy if you merge a couple patches for him.

Quoted 5 lines

That queue currently consists of:

hdaps-invert-the-axes-for-hdaps-on-lenovo-r61i-thinkpads.patch adt746x-logical-bitwise-confusion-in-set_max_duty_at_crit.patch

The adt7473:set_max_duty_at_crit() code was originally confusing but correct. After this patch the code is clear but broken. Is it considered an improvement?

(I would have told you before if that patch had been submitted to the lm-sensors list as it should have, given that it modifies a hardware monitoring driver... Not to mention that the patch file name doesn't match the driver file name, just to add to the confusion.)

Quoted 1 line

adt7473-minor-documentation-update.patch

I reviewed and acked that one, it's ready to go.

Quoted 1 line

i5k_amb-support-intel-5400-chipset.patch

Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
#3 lore raw
Quoted 10 lines · On Mon, 5 May 2008 23:22:45 +0200 Jean Delvare <[email protected]> wrote:

Hi Andrew,

Quoted 7 lines · Andrew Morton
Quoted 1 line · On Fri, 2 May 2008 21:55:59 -0700 "Darrick J. Wong" <[email protected]> wrote:

Minor rework to support the Intel 5400 chipset.

Thanks, I added this to the for-git-hwmon part of the -mm queue.

Quoted 8 lines
Quoted 3 lines

I haven't heard from Mark for quite some time so soon I might declare him absent and I'll go through my hwmon patches with a view to merging them into 2.6.26 directly.

For Mark's status see:
http://lists.lm-sensors.org/pipermail/lm-sensors/2008-May/023006.html

I guess he'll be happy if you merge a couple patches for him.

hm, OK.

Quoted 9 lines
Quoted 5 lines

That queue currently consists of:

hdaps-invert-the-axes-for-hdaps-on-lenovo-r61i-thinkpads.patch adt746x-logical-bitwise-confusion-in-set_max_duty_at_crit.patch

The adt7473:set_max_duty_at_crit() code was originally confusing but correct. After this patch the code is clear but broken. Is it considered an improvement?

I'll drop it.

Quoted 4 lines

(I would have told you before if that patch had been submitted to the lm-sensors list as it should have, given that it modifies a hardware monitoring driver... Not to mention that the patch file name doesn't match the driver file name, just to add to the confusion.)

Well, y'know, stuff happens. Stuff like you forgetting to cc that patch's author on this email.

Quoted 3 lines
Quoted 1 line

adt7473-minor-documentation-update.patch

I reviewed and acked that one, it's ready to go.

OK, thanks.

Quoted 1 line
Quoted 1 line

i5k_amb-support-intel-5400-chipset.patch

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#4 lore raw

ibmaem driver status?

In reply to Andrew Morton
Quoted 10 lines · Andrew Morton
Quoted 1 line · On Fri, 2 May 2008 21:55:59 -0700 "Darrick J. Wong" <[email protected]> wrote:

Minor rework to support the Intel 5400 chipset.

Thanks, I added this to the for-git-hwmon part of the -mm queue.

I haven't heard from Mark for quite some time so soon I might declare him absent and I'll go through my hwmon patches with a view to merging them into 2.6.26 directly.

Jean/Mark:

Any thoughts on merging the ibmaem driver? I haven't heard anything in a good week and a half; if it's in Mark's backlog that's not a problem. (Or at least not as big of one as "silently bitrotting" :))

--D

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#5 lore raw

Re: ibmaem driver status?

Quoted 12 lines · On Mon, 5 May 2008 16:04:40 -0700 "Darrick J. Wong" <[email protected]> wrote:
Quoted 10 lines · Andrew Morton
Quoted 1 line · On Fri, 2 May 2008 21:55:59 -0700 "Darrick J. Wong" <[email protected]> wrote:

Minor rework to support the Intel 5400 chipset.

Thanks, I added this to the for-git-hwmon part of the -mm queue.

I haven't heard from Mark for quite some time so soon I might declare him absent and I'll go through my hwmon patches with a view to merging them into 2.6.26 directly.

Quoted 6 lines

Jean/Mark:

Any thoughts on merging the ibmaem driver? I haven't heard anything in a good week and a half; if it's in Mark's backlog that's not a problem. (Or at least not as big of one as "silently bitrotting" :))

Am just looking at it now. Your email was mysteriously truncated partway through the patch. Resend, please.

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#6 lore raw

PATCH RESEND v2 1/1 ibmaem: New driver for power/energy meters in IBM System X hardware

[resend due to truncation]
Refactor the registration function to shrink the macros. Let me know if more aggressive de-macroing is desirable.
---
New driver for power meters in IBM System X hardware, with a few cleanups suggested by Anthony Liguori.

---

 Documentation/hwmon/ibmaem |   37 +
 drivers/hwmon/Kconfig      |   14 +
 drivers/hwmon/Makefile     |    1 
 drivers/hwmon/ibmaem.c     | 1157 ++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 1209 insertions(+), 0 deletions(-)
Documentation/hwmon/ibmaem new file +37 -0
new file mode 100644
index 0000000..2fefaf5
@@ -0,0 +1,37 @@
+Kernel driver ibmaem
+===========
+
+Supported systems:
+  * Any recent IBM System X server with Active Energy Manager support.
+    This includes the x3350, x3550, x3650, x3655, x3755, x3850 M2,
+    x3950 M2, and certain HS2x/LS2x/QS2x blades.  The IPMI host interface
+    driver ("ipmi-si") needs to be loaded for this driver to do anything.
+    Prefix: 'ibmaem'
+    Datasheet: Not available
+
+Author: Darrick J. Wong
+
+Description
+-----------
+
+This driver implements sensor reading support for the energy and power
+meters available on various IBM System X hardware through the BMC.  All
+sensor banks will be exported as platform devices; this driver can talk
+to both v1 and v2 interfaces.  This driver is completely separate from the
+older ibmpex driver.
+
+The v1 AEM interface has a simple set of features to monitor energy use.
+There is a register that displays an estimate of raw energy consumption
+since the last BMC reset, and a power sensor that returns average power
+use over a configurable interval.
+
+The v2 AEM interface is a bit more sophisticated, being able to present
+a wider range of energy and power use registers, the power cap as
+set by the AEM software, and temperature sensors.
+
+Special Features
+----------------
+
+The "power_cap" value displays the current system power cap, as set by
+the Active Energy Manager software.  Setting the power cap from the host
+is not currently supported.
drivers/hwmon/Kconfig +14 -0
index 4dc76bc..00ff533 100644
@@ -330,6 +330,20 @@ config SENSORS_CORETEMP
 	  sensor inside your CPU. Supported all are all known variants
 	  of Intel Core family.
 
+config SENSORS_IBMAEM
+	tristate "IBM Active Energy Manager temperature/power sensors and control"
+	select IPMI_SI
+	depends on IPMI_HANDLER
+	help
+	  If you say yes here you get support for the temperature and
+	  power sensors and capping hardware in various IBM System X
+	  servers that support Active Energy Manager.  This includes
+	  the x3350, x3550, x3650, x3655, x3755, x3850 M2, x3950 M2,
+	  and certain HS2x/LS2x/QS2x blades.
+
+	  This driver can also be built as a module.  If so, the module
+	  will be called ibmaem.
+
 config SENSORS_IBMPEX
 	tristate "IBM PowerExecutive temperature/power sensors"
 	select IPMI_SI
drivers/hwmon/Makefile +1 -0
index 3bdb05a..d098677 100644
@@ -41,6 +41,7 @@ obj-$(CONFIG_SENSORS_GL518SM)	+= gl518sm.o
 obj-$(CONFIG_SENSORS_GL520SM)	+= gl520sm.o
 obj-$(CONFIG_SENSORS_HDAPS)	+= hdaps.o
 obj-$(CONFIG_SENSORS_I5K_AMB)	+= i5k_amb.o
+obj-$(CONFIG_SENSORS_IBMAEM)	+= ibmaem.o
 obj-$(CONFIG_SENSORS_IBMPEX)	+= ibmpex.o
 obj-$(CONFIG_SENSORS_IT87)	+= it87.o
 obj-$(CONFIG_SENSORS_K8TEMP)	+= k8temp.o
drivers/hwmon/ibmaem.c new file +1157 -0

1,157 lines not on this page · raw message

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#7 lore raw

Re: [resend] [PATCH v2] ibmaem: New driver for power/energy meters in IBM System X hardware

Quoted 56 lines · On Tue, 6 May 2008 15:38:13 -0700 "Darrick J. Wong" <[email protected]> wrote:

[resend due to truncation]
Refactor the registration function to shrink the macros. Let me know if more aggressive de-macroing is desirable.
---
New driver for power meters in IBM System X hardware, with a few cleanups suggested by Anthony Liguori.

---

index 0000000..2fefaf5

Documentation/hwmon/ibmaem new file +37 -0
@@ -0,0 +1,37 @@
+Kernel driver ibmaem
+===========
+
+Supported systems:
+  * Any recent IBM System X server with Active Energy Manager support.
+    This includes the x3350, x3550, x3650, x3655, x3755, x3850 M2,
+    x3950 M2, and certain HS2x/LS2x/QS2x blades.  The IPMI host interface
+    driver ("ipmi-si") needs to be loaded for this driver to do anything.
+    Prefix: 'ibmaem'
+    Datasheet: Not available
+
+Author: Darrick J. Wong
+
+Description
+-----------
+
+This driver implements sensor reading support for the energy and power
+meters available on various IBM System X hardware through the BMC.  All
+sensor banks will be exported as platform devices; this driver can talk
+to both v1 and v2 interfaces.  This driver is completely separate from the
+older ibmpex driver.
+
+The v1 AEM interface has a simple set of features to monitor energy use.
+There is a register that displays an estimate of raw energy consumption
+since the last BMC reset, and a power sensor that returns average power
+use over a configurable interval.
+
+The v2 AEM interface is a bit more sophisticated, being able to present
+a wider range of energy and power use registers, the power cap as
+set by the AEM software, and temperature sensors.
+
+Special Features
+----------------
+
+The "power_cap" value displays the current system power cap, as set by
+the Active Energy Manager software.  Setting the power cap from the host
+is not currently supported.
drivers/hwmon/Kconfig +0 -0
index 4dc76bc..00ff533 100644

drivers/hwmon/Kconfig config SENSORS_CORETEMP · line 336

Quoted 7 lines
drivers/hwmon/Kconfig +4 -0
@@ -330,6 +330,20 @@ config SENSORS_CORETEMP
 	  sensor inside your CPU. Supported all are all known variants
 	  of Intel Core family.
 
+config SENSORS_IBMAEM
+	tristate "IBM Active Energy Manager temperature/power sensors and control"
+	select IPMI_SI
+	depends on IPMI_HANDLER

hm. IMPI_SI depends on IPMI_HANDLER, so this looks OK. But this stuff explodes in our facs so often...

Quoted 23 lines
drivers/hwmon/Kconfig +20 -0
+	help
+	  If you say yes here you get support for the temperature and
+	  power sensors and capping hardware in various IBM System X
+	  servers that support Active Energy Manager.  This includes
+	  the x3350, x3550, x3650, x3655, x3755, x3850 M2, x3950 M2,
+	  and certain HS2x/LS2x/QS2x blades.
+
+	  This driver can also be built as a module.  If so, the module
+	  will be called ibmaem.
+
 
...
 
+static DEFINE_IDR(aem1_idr);
+static DEFINE_SPINLOCK(aem1_idr_lock);
+static DEFINE_IDR(aem2_idr);
+static DEFINE_SPINLOCK(aem2_idr_lock);
+
+struct aem_ipmi_data;
+struct aem1_data;
+struct aem2_data;
+
+static void aem_register_bmc(int iface, struct device *dev);

drivers/hwmon/Kconfig config SENSORS_CORETEMP

Quoted 10 lines
drivers/hwmon/Kconfig +10 -0
+static void aem_bmc_gone(int iface);
+static void aem_msg_handler(struct ipmi_recv_msg *msg, void *user_msg_data);
+static void aem_init_aem1(struct aem_ipmi_data *probe);
+static void aem1_delete(struct aem1_data *data);
+static void aem_init_aem2(struct aem_ipmi_data *probe);
+static void aem2_delete(struct aem2_data *data);
+static void aem1_remove_sensors(struct aem1_data *data);
+static int aem1_find_sensors(struct aem1_data *data);
+static void aem2_remove_sensors(struct aem2_data *data);
+static int aem2_find_sensors(struct aem2_data *data);

If these were moved below the struct definitions, we wouldn't need the forward declarations of the structs.

Quoted 60 lines
drivers/hwmon/Kconfig +60 -0
+static struct device_driver aem_driver = {
+	.name = DRVNAME,
+	.bus = &platform_bus_type,
+};
+
+struct aem_ipmi_data {
+	struct completion	read_complete;
+	struct ipmi_addr	address;
+	ipmi_user_t		user;
+	int			interface;
+
+	struct kernel_ipmi_msg	tx_message;
+	long			tx_msgid;
+
+	void			*rx_msg_data;
+	unsigned short		rx_msg_len;
+	unsigned char		rx_result;
+	int			rx_recv_type;
+
+	struct device		*bmc_device;
+};
+
+struct aem_fw_data {
+	struct device		*hwmon_dev;
+	struct platform_device	*pdev;
+	struct mutex		lock;
+	char			valid;
+	unsigned long		last_updated;	/* In jiffies */
+	u8			ver_major;
+	u8			ver_minor;
+	u8			module_handle;
+	int			id;
+	struct aem_ipmi_data	ipmi;
+};
+
+struct aem_ro_sensor_template {
+	char *label;
+	ssize_t (*show)(struct device *dev,
+			struct device_attribute *devattr,
+			char *buf);
+	int index;
+};
+
+struct aem_rw_sensor_template {
+	char *label;
+	ssize_t (*show)(struct device *dev,
+			struct device_attribute *devattr,
+			char *buf);
+	ssize_t (*set)(struct device *dev,
+		       struct device_attribute *devattr,
+		       const char *buf, size_t count);
+	int index;
+};
+
+struct aem1_data {
+	struct aem_fw_data	fw;
+	struct list_head	list;
+
+	/*
+	 * Available sensors:

drivers/hwmon/Kconfig

Quoted 10 lines
drivers/hwmon/Kconfig +10 -0
+	 * Energy meter
+	 * Power meter
+	 */
+	struct sensor_device_attribute	sensors[AEM1_NUM_SENSORS];
+
+	/* energy use */
+	u64			energy[AEM1_NUM_ENERGY_REGS];
+
+	int			power_period[AEM1_NUM_ENERGY_REGS];
+};

It'd be nice to document the units in which these physical quantities are recorded. BTUs/fortnight?

Quoted 15 lines
drivers/hwmon/Kconfig +15 -0
+struct aem2_data {
+	struct aem_fw_data	fw;
+	struct list_head	list;
+
+	/*
+	 * Available sensors:
+	 * Two energy meters
+	 * Two power meters
+	 * Two temperature sensors
+	 * Six powercap registers
+	 */
+	struct sensor_device_attribute	sensors[AEM2_NUM_SENSORS];
+
+	/* energy use */
+	u64			energy[AEM2_NUM_ENERGY_REGS];

drivers/hwmon/Kconfig

Quoted 10 lines
drivers/hwmon/Kconfig +10 -0
+
+	/* power caps */
+	u16			pcap[AEM2_NUM_PCAP_REGS];
+
+	/* exhaust temperature */
+	u8			temp[AEM2_NUM_TEMP_REGS];
+
+	/* power sampling interval */
+	int			power_period[AEM2_NUM_ENERGY_REGS];
+};

ditto.

Quoted 5 lines
drivers/hwmon/Kconfig +5 -0
+/* Data structures returned by the AEM firmware */
+struct aem_iana_id {
+	u8			bytes[3];
+};
+static struct aem_iana_id system_x_id = {

drivers/hwmon/Kconfig

Quoted 10 lines
drivers/hwmon/Kconfig +10 -0
+	.bytes = {0x4D, 0x4F, 0x00}
+};
+
+/* These are used to find AEM1 instances */
+struct aem_find_firmware_req {
+	struct aem_iana_id	id;
+	u8			rsvd;
+	u16			index;
+	u16			module_type_id;
+} __attribute__ ((packed));

We have a __packed helper in compiler-gcc.h

Quoted 9 lines
drivers/hwmon/Kconfig +9 -0
+struct aem_find_firmware_resp {
+	struct aem_iana_id	id;
+	u8			num_instances;
+} __attribute__ ((packed));
+
+/* These are used to find AEM2 instances */
+struct aem_find_instance_req {
+	struct aem_iana_id	id;
+	u8			instance_number;

drivers/hwmon/Kconfig

Quoted 10 lines
drivers/hwmon/Kconfig +7 -0
+	u16			module_type_id;
+} __attribute__ ((packed));
+
 
...
 
+/* Probe a BMC for AEM firmware instances */
+static void aem_register_bmc(int iface, struct device *dev)
+{
+	struct aem_ipmi_data probe;

184 bytes of stack. OK, but worth keeping an eye on.

Quoted 7 lines
drivers/hwmon/Kconfig +7 -0
+	if (aem_init_ipmi_data(&probe, iface, dev))
+		return;
+
+	aem_init_aem1(&probe);
+	aem_init_aem2(&probe);
+
+	ipmi_destroy_user(probe.user);

drivers/hwmon/Kconfig

Quoted 10 lines
drivers/hwmon/Kconfig +7 -0
+}
+
 
...
 
+/* Dispatch IPMI messages to callers */
+static void aem_msg_handler(struct ipmi_recv_msg *msg, void *user_msg_data)
+{
+	unsigned short rx_len;
+	struct aem_ipmi_data *data = (struct aem_ipmi_data *)user_msg_data;

unneeded and undesirable cast of void*.

Quoted 52 lines
drivers/hwmon/Kconfig +49 -0
+	if (msg->msgid != data->tx_msgid) {
+		dev_err(data->bmc_device, "Mismatch between received msgid "
+			"(%02x) and transmitted msgid (%02x)!\n",
+			(int)msg->msgid,
+			(int)data->tx_msgid);
+		ipmi_free_recv_msg(msg);
+		return;
+	}
+
+	data->rx_recv_type = msg->recv_type;
+	if (msg->msg.data_len > 0)
+		data->rx_result = msg->msg.data[0];
+	else
+		data->rx_result = IPMI_UNKNOWN_ERR_COMPLETION_CODE;
+
+	if (msg->msg.data_len > 1) {
+		rx_len = msg->msg.data_len - 1;
+		if (data->rx_msg_len < rx_len)
+			rx_len = data->rx_msg_len;
+		data->rx_msg_len = rx_len;
+		memcpy(data->rx_msg_data, msg->msg.data + 1, data->rx_msg_len);
+	} else
+		data->rx_msg_len = 0;
+
+	ipmi_free_recv_msg(msg);
+	complete(&data->read_complete);
+}
+
 
...
 
+/* Obtain an id */
+#define AEM_IDR_GET(type) \
+static int type##_idr_get(int *id) \
+{ \
+	int i, err; \
+\
+again: \
+	if (unlikely(idr_pre_get(&type##_idr, GFP_KERNEL) = 0)) \
+		return -ENOMEM; \
+\
+	spin_lock(&type##_idr_lock); \
+	err = idr_get_new(&type##_idr, NULL, &i); \
+	spin_unlock(&type##_idr_lock); \
+\
+	if (unlikely(err = -EAGAIN)) \
+		goto again; \
+	else if (unlikely(err)) \
+		return err; \
+\
+	*id = i & MAX_ID_MASK; \
+	return 0; \

drivers/hwmon/Kconfig

Quoted 10 lines
drivers/hwmon/Kconfig +10 -0
+}
+
+/* Release an object ID */
+#define AEM_IDR_PUT(type) \
+static void type##_idr_put(int id) \
+{ \
+	spin_lock(&type##_idr_lock); \
+	idr_remove(&type##_idr, id); \
+	spin_unlock(&type##_idr_lock); \
+}

ick.

Quoted 6 lines

...

drivers/hwmon/Kconfig +3 -0
+/* Find and initialize all AEM1 instances */
+static void aem_init_aem1(struct aem_ipmi_data *probe)
+{

drivers/hwmon/Kconfig

Quoted 10 lines
drivers/hwmon/Kconfig +10 -0
+	int num, i, err;
+
+	num = aem_find_aem1_count(probe);
+	for (i = 0; i < num; i++) {
+		err = aem_init_aem1_inst(probe, i);
+		if (err) {
+			dev_err(probe->bmc_device,
+				"Error %d initializing AEM1 0x%X\n",
+				err, i);
+			return;

Should the error really be dropped on the floor?

Quoted 22 lines
drivers/hwmon/Kconfig +19 -0
+		}
+	}
+}
+
 
...
 
+/* Find and initialize one AEM2 instance */
+static int aem_init_aem2_inst(struct aem_ipmi_data *probe,
+			      struct aem_find_instance_resp *fi_resp)
+{
+	struct aem2_data *data;
+	int i;
+	int res = -ENOMEM;
+
+	data = kzalloc(sizeof(*data), GFP_KERNEL);
+	if (!data)
+		return res;
+	mutex_init(&data->fw.lock);
+
+	/* Copy instance data */
+	data->fw.ver_major = fi_resp->major;

drivers/hwmon/Kconfig

Quoted 10 lines
drivers/hwmon/Kconfig +10 -0
+	data->fw.ver_minor = fi_resp->minor;
+	data->fw.module_handle = fi_resp->module_handle;
+	for (i = 0; i < AEM2_NUM_ENERGY_REGS; i++)
+		data->power_period[i] = AEM_DEFAULT_POWER_INTERVAL;
+
+	/* Create sub-device for this fw instance */
+	if (aem2_idr_get(&data->fw.id))
+		goto id_err;
+
+	data->fw.pdev = platform_device_alloc(DRVNAME "2", data->fw.id);

platform_device_alloc() can fail.

drivers/hwmon/Kconfig

Quoted 1 line
drivers/hwmon/Kconfig +1 -0
+	data->fw.pdev->dev.driver = &aem_driver;

which will result in an oops here.

drivers/hwmon/Kconfig

Quoted 1 line
drivers/hwmon/Kconfig +1 -0
+	if (IS_ERR(data->fw.pdev))

platform_device_alloc() returns NULL on failure, not ERR_PTR.

Quoted 291 lines
drivers/hwmon/Kconfig +291 -0
+		goto dev_err;
+
+	res = platform_device_add(data->fw.pdev);
+	if (res)
+		goto ipmi_err;
+
+	dev_set_drvdata(&data->fw.pdev->dev, &data->fw);
+
+	/* Set up IPMI interface */
+	if (aem_init_ipmi_data(&data->fw.ipmi, probe->interface,
+			       probe->bmc_device))
+		goto ipmi_err;
+
+	/* Register with hwmon */
+	data->fw.hwmon_dev = hwmon_device_register(&data->fw.pdev->dev);
+
+	if (IS_ERR(data->fw.hwmon_dev)) {
+		dev_err(&data->fw.pdev->dev, "Unable to register hwmon "
+			"device for IPMI interface %d\n",
+			probe->interface);
+		goto hwmon_reg_err;
+	}
+
+	/* Find sensors */
+	if (aem2_find_sensors(data))
+		goto sensor_err;
+
+	/* Add to our list of AEM2 devices */
+	list_add_tail(&data->list, &driver_data.aem2_devices);
+	dev_info(data->fw.ipmi.bmc_device, "Found AEM v%d.%d at 0x%X\n",
+		 data->fw.ver_major, data->fw.ver_minor,
+		 data->fw.module_handle);
+	return 0;
+
+sensor_err:
+	hwmon_device_unregister(data->fw.hwmon_dev);
+hwmon_reg_err:
+	ipmi_destroy_user(data->fw.ipmi.user);
+ipmi_err:
+	dev_set_drvdata(&data->fw.pdev->dev, NULL);
+	platform_device_unregister(data->fw.pdev);
+dev_err:
+	aem2_idr_put(data->fw.id);
+id_err:
+	kfree(data);
+
+	return res;
+}
+
+/* Find and initialize all AEM2 instances */
+static void aem_init_aem2(struct aem_ipmi_data *probe)
+{
+	struct aem_find_instance_resp fi_resp;
+	int err;
+	int i = 0;
+
+	while (!aem_find_aem2(probe, &fi_resp, i)) {
+		if (fi_resp.major != 2) {
+			dev_err(probe->bmc_device, "Unknown AEM v%d; please "
+				"report this to the maintainer.\n",
+				fi_resp.major);
+			i++;
+			continue;
+		}
+		err = aem_init_aem2_inst(probe, &fi_resp);
+		if (err) {
+			dev_err(probe->bmc_device,
+				"Error %d initializing AEM2 0x%X\n",
+				err, fi_resp.module_handle);
+			return;
+		}
+		i++;
+	}
+}
+
+/* Delete an AEM2 instance */
+AEM_DELETE(aem2)
+
+/* Sensor support functions */
+
+/* Read a sensor value */
+static int aem_read_sensor(struct aem_fw_data *fwdata, u8 elt, u8 reg,
+			   void *data, size_t size)
+{
+	int rs_size;
+	struct aem_read_sensor_req	rs_req;
+	struct aem_read_sensor_resp	*rs_resp;
+	struct aem_ipmi_data *ipmi = &fwdata->ipmi;
+
+	/* AEM registers are 1, 2, 4 or 8 bytes */
+	switch (size) {
+	case 1:
+	case 2:
+	case 4:
+	case 8:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	rs_req.id = system_x_id;
+	rs_req.module_handle = fwdata->module_handle;
+	rs_req.element = elt;
+	rs_req.subcommand = AEM_READ_REGISTER;
+	rs_req.reg = reg;
+	rs_req.rx_buf_size = size;
+
+	ipmi->tx_message.cmd = AEM_ELEMENT_CMD;
+	ipmi->tx_message.data = (char *)&rs_req;
+	ipmi->tx_message.data_len = sizeof(rs_req);
+
+	rs_size = sizeof(*rs_resp) + size;
+	rs_resp = kzalloc(rs_size, GFP_KERNEL);
+	if (!rs_resp)
+		return -ENOMEM;
+
+	ipmi->rx_msg_data = rs_resp;
+	ipmi->rx_msg_len = rs_size;
+
+	aem_send_message(ipmi);
+
+	wait_for_completion(&ipmi->read_complete);
+
+	if (ipmi->rx_result || ipmi->rx_msg_len != rs_size ||
+	    memcmp(&rs_resp->id, &system_x_id, sizeof(system_x_id))) {
+		kfree(rs_resp);
+		return -ENOENT;
+	}
+
+	switch (size) {
+	case 1: {
+		u8 *x = data;
+		*x = rs_resp->bytes[0];
+		break;
+	}
+	case 2: {
+		u16 *x = data;
+		*x = be16_to_cpup((u16 *)rs_resp->bytes);
+		break;
+	}
+	case 4: {
+		u32 *x = data;
+		*x = be32_to_cpup((u32 *)rs_resp->bytes);
+		break;
+	}
+	case 8: {
+		u64 *x = data;
+		*x = be64_to_cpup((u64 *)rs_resp->bytes);
+		break;
+	}
+	}
+
+	return 0;
+}
+
+/* Update AEM1 energy sensors */
+static void update_aem1_energy_sensors(struct aem1_data *data)
+{
+	aem_read_sensor(&data->fw, AEM_ENERGY_ELEMENT, 0, &data->energy, 8);
+}
+
+/* Update all AEM1 sensors */
+static void update_aem1_sensors(struct aem1_data *data)
+{
+	mutex_lock(&data->fw.lock);
+	if (time_before(jiffies, data->fw.last_updated + REFRESH_INTERVAL) &&
+	    data->fw.valid)
+		goto out;
+
+	update_aem1_energy_sensors(data);
+out:
+	mutex_unlock(&data->fw.lock);
+}
+
+/* Update AEM2 energy sensors */
+static void update_aem2_energy_sensors(struct aem2_data *data)
+{
+	aem_read_sensor(&data->fw, AEM_ENERGY_ELEMENT,	0, &data->energy[0], 8);
+	aem_read_sensor(&data->fw, AEM_ENERGY_ELEMENT,	1, &data->energy[1], 8);
+}
+
+/* Update all AEM2 sensors */
+static void update_aem2_sensors(struct aem2_data *data)
+{
+	int i;
+
+	mutex_lock(&data->fw.lock);
+	if (time_before(jiffies, data->fw.last_updated + REFRESH_INTERVAL) &&
+	    data->fw.valid)
+		goto out;
+
+	update_aem2_energy_sensors(data);
+	aem_read_sensor(&data->fw, AEM_EXHAUST_ELEMENT,	0, &data->temp[0], 1);
+	aem_read_sensor(&data->fw, AEM_EXHAUST_ELEMENT,	1, &data->temp[1], 1);
+
+	for (i = POWER_CAP; i <= POWER_AUX; i++)
+		aem_read_sensor(&data->fw, AEM_POWER_CAP_ELEMENT, i,
+				&data->pcap[i], 2);
+out:
+	mutex_unlock(&data->fw.lock);
+}
+
+/* sysfs support functions */
+
+#define AEM_SHOW_POWER(type) \
+static ssize_t type##_show_power(struct device *dev, \
+				 struct device_attribute *devattr, \
+				 char *buf) \
+{ \
+	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr); \
+	struct aem_fw_data *fwdata = dev_get_drvdata(dev); \
+	struct type##_data * a = container_of(fwdata, \
+						 struct type##_data, fw); \
+	u64 before, after; \
+	signed long leftover; \
+\
+	mutex_lock(&fwdata->lock); \
+	update_##type##_energy_sensors(a); \
+	before = a->energy[attr->index]; \
+\
+	leftover = schedule_timeout_interruptible( \
+			msecs_to_jiffies(a->power_period[attr->index]) \
+		   ); \
+	if (leftover) { \
+		mutex_unlock(&fwdata->lock); \
+		return 0; \
+	} \
+\
+	update_##type##_energy_sensors(a); \
+	after = a->energy[attr->index]; \
+	mutex_unlock(&fwdata->lock); \
+\
+	return sprintf(buf, "%llu\n", (after - before) * 1000000 / \
+		       a->power_period[attr->index]); \
+}
+
+/* Display energy use */
+#define AEM_SHOW_ENERGY(type) \
+static ssize_t type##_show_energy(struct device *dev, \
+				  struct device_attribute *devattr, \
+				  char *buf) \
+{ \
+	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr); \
+	struct aem_fw_data *fwdata = dev_get_drvdata(dev); \
+	struct type##_data * a = container_of(fwdata, struct type##_data, fw); \
+	update_##type##_sensors(a); \
+\
+	return sprintf(buf, "%llu\n", a->energy[attr->index] * 1000); \
+}
+
+/* Display power interval registers */
+#define AEM_SHOW_POWER_PERIOD(type) \
+static ssize_t type##_show_power_period(struct device *dev, \
+					struct device_attribute *devattr, \
+					char *buf) \
+{ \
+	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr); \
+	struct aem_fw_data *fwdata = dev_get_drvdata(dev); \
+	struct type##_data * a = container_of(fwdata, struct type##_data, fw); \
+	update_##type##_sensors(a); \
+\
+	return sprintf(buf, "%d\n", a->power_period[attr->index]); \
+}
+
+/* Set power interval registers */
+#define AEM_SET_POWER_PERIOD(type) \
+static ssize_t type##_set_power_period(struct device *dev, \
+				       struct device_attribute *devattr, \
+				       const char *buf, size_t count) \
+{ \
+	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr); \
+	struct aem_fw_data *fwdata = dev_get_drvdata(dev); \
+	struct type##_data * a = container_of(fwdata, struct type##_data, fw); \
+	int temp = simple_strtol(buf, NULL, 10); \
+\
+	if (temp < AEM_MIN_POWER_INTERVAL) \
+		return -EINVAL; \
+\
+	mutex_lock(&a->fw.lock); \
+	a->power_period[attr->index] = temp; \
+	mutex_unlock(&a->fw.lock); \
+\
+	return count; \
+}
+
+/* Remove sensors attached to an AEM device */
+#define AEM_REMOVE_SENSORS(type, num_sensors) \
+static void type##_remove_sensors(struct type##_data *data) \
+{ \
+	int i; \
+\

drivers/hwmon/Kconfig

Quoted 10 lines
drivers/hwmon/Kconfig +10 -0
+	for (i = 0; i < num_sensors; i++) { \
+		if (!data->sensors[i].dev_attr.attr.name) \
+			continue; \
+		device_remove_file(&data->fw.pdev->dev, \
+				   &data->sensors[i].dev_attr); \
+	} \
+\
+	device_remove_file(&data->fw.pdev->dev, \
+			   &sensor_dev_attr_name.dev_attr); \
+}

ho hum.

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#8 lore raw

Re: [resend] [PATCH v2] ibmaem: New driver for power/energy meters in IBM System X hardware

In reply to Darrick J. Wong

drivers/hwmon/ibmaem.c line 918

Quoted 10 lines · On Tue, 6 May 2008 15:38:13 -0700 "Darrick J. Wong" <[email protected]> wrote:

[resend due to truncation]
Refactor the registration function to shrink the macros. Let me know if more aggressive de-macroing is desirable.
---
New driver for power meters in IBM System X hardware, with a few cleanups suggested by Anthony Liguori.

...

drivers/hwmon/ibmaem.c new file +1 -0
+	int temp = simple_strtol(buf, NULL, 10); \

A checkpatch version which you might not have says

WARNING: consider using strict_strtol in preference to simple_strtol #1015: FILE: drivers/hwmon/ibmaem.c:918:
+ int temp = simple_strtol(buf, NULL, 10); \

total: 0 errors, 1 warnings, 1221 lines checked

because the code as you have it will accept input of the form "12foo". strict_strtol() will error on that.

afacit the calling code would perhaps need a little changing to handle such errors appropriately.

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#9 lore raw

Re: [resend] [PATCH v2] ibmaem: New driver for power/energy meters in IBM System X hardware

Quoted 10 lines · Andrew Morton

A checkpatch version which you might not have says

WARNING: consider using strict_strtol in preference to simple_strtol #1015: FILE: drivers/hwmon/ibmaem.c:918:
+ int temp = simple_strtol(buf, NULL, 10); \

total: 0 errors, 1 warnings, 1221 lines checked

because the code as you have it will accept input of the form "12foo". strict_strtol() will error on that.

As a side note, I really wonder why this new function wasn't named just strtol(), especially if the general idea is to get rid of the "simple" variant after some time.

Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
#10 lore raw

Re: [resend] [PATCH v2] ibmaem: New driver for power/energy meters in IBM System X hardware

Quoted 7 lines · On Wed, 7 May 2008 08:01:05 +0200 Jean Delvare <[email protected]> wrote:
Quoted 6 lines · Andrew Morton

A checkpatch version which you might not have says

WARNING: consider using strict_strtol in preference to simple_strtol #1015: FILE: drivers/hwmon/ibmaem.c:918:
+ int temp = simple_strtol(buf, NULL, 10); \

Quoted 8 lines
Quoted 4 lines

total: 0 errors, 1 warnings, 1221 lines checked

because the code as you have it will accept input of the form "12foo". strict_strtol() will error on that.

As a side note, I really wonder why this new function wasn't named just strtol(), especially if the general idea is to get rid of the "simple" variant after some time.

didn't think of doing that.

But:

y:/home/akpm> cat t.c
#include <stdlib.h>
#include <stdio.h>

main()

{
        unsigned long u = strtoul("12foo", 0, 10);
        printf("%lu\n", u);
}

y:/home/akpm> gcc t.c
y:/home/akpm> ./a.out
12

strtoul() has different behaviour.

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors