[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20080310222735.GA22619@ldl.fc.hp.com>
Date: Mon, 10 Mar 2008 16:27:35 -0600
From: Alex Chiang <achiang@...com>
To: tony.luck@...el.com, lenb@...nel.org
Cc: linux-ia64@...r.kernel.org, linux-acpi@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: [PATCH, RFC] Create 'slot' sysfs attribute in
/sys/devices/system/cpu/cpuN/topology/
Hi Len, Tony,
A while back, I submitted a patch to expose the value of the _SUN
method for CPUs. On ia64, the "physical id" field in
/proc/cpuinfo isn't sufficient for some legacy HP ia64 platforms,
as they do not have a modern SAL or PAL.
http://lkml.org/lkml/2007/10/30/529
That patch wasn't so great because as Len pointed out, /proc/acpi
is a deprecated interface.
Here is another try that attempts to present a generic "slot"
attribute in sysfs that leaves the various architectures free to
implement however they like.
Yes, I am aware of the physical_package_id attribute. However,
that doesn't necessarily map to a slot on the system board.
Additionally, it's broken on the same ia64 systems where
/proc/cpuinfo is broken, because it's trying to read the same
field. :-/
On ia64, I have a somewhat ugly implementation, where the ACPI
processor driver is filling in the answer. I chose this approach
because I couldn't figure out how to do it in the base CPU
driver, since calling the _SUN method requires the ACPI
interpreter to be initialized, and it didn't look like that was
the case when we did things like identify_siblings(), for
instance.
Finally, there is something that looks odd in struct
cpuinfo_ia64, where there are now two #ifdef CONFIG_SMP
statements. I did this because adding the new 'slot' member
created a 4-byte hole in the struct. It felt a little cleaner to
me to move the 'int cpu;' member upwards to fill in the hole, and
then my 'slot' could take the spot of 'cpu'.
I'd like to get feedback on both:
- the concept in general of creating a new sysfs
attribute
- the ia64 implementation, specifically
Thanks.
/ac
From: Alex Chiang <achiang@...com>
Physical slot information for CPUs is important for many reasons,
including managability. For instance, if a platform supports
physical CPU hotplug, it would be nice for the user to know which
physical CPU module to actually unplug after offlining it.
The "physical id" field in /proc/cpuinfo is broken/missing on
older ia64 platforms that do not have modern firmware.
/sys/devices/system/cpu/cpuN/topology/physical_package_id exists,
but is broken on those same platforms because it's simply reading
the same field that would have been presented in /proc/cpuinfo.
Create a 'slot' attribute and define an interface that may be
populated via alternate sources, without breaking existing
functionality.
Implement this interface on ia64 by invoking the _SUN method on a
per cpu basis.
Leave other archs alone for now.
Signed-off-by: Alex Chiang <achiang@...com>
---
arch/ia64/kernel/setup.c | 1 +
arch/ia64/kernel/topology.c | 6 ++++++
drivers/acpi/processor_core.c | 5 +++++
drivers/base/topology.c | 9 +++++++++
include/asm-ia64/cpu.h | 1 +
include/asm-ia64/processor.h | 7 ++++++-
include/asm-ia64/topology.h | 1 +
include/asm-x86/cpu.h | 1 +
8 files changed, 30 insertions(+), 1 deletions(-)
diff --git a/arch/ia64/kernel/setup.c b/arch/ia64/kernel/setup.c
index ebd1a09..34de68a 100644
--- a/arch/ia64/kernel/setup.c
+++ b/arch/ia64/kernel/setup.c
@@ -735,6 +735,7 @@ identify_cpu (struct cpuinfo_ia64 *c)
*/
c->threads_per_core = c->cores_per_socket = c->num_log = 1;
c->socket_id = -1;
+ c->slot = -1;
identify_siblings(c);
diff --git a/arch/ia64/kernel/topology.c b/arch/ia64/kernel/topology.c
index a2484fc..3810b48 100644
--- a/arch/ia64/kernel/topology.c
+++ b/arch/ia64/kernel/topology.c
@@ -27,6 +27,12 @@
static struct ia64_cpu *sysfs_cpus;
+void arch_set_topology_slot(int num, u32 slot)
+{
+ cpu_data(num)->slot = slot;
+}
+EXPORT_SYMBOL(arch_set_topology_slot);
+
int arch_register_cpu(int num)
{
#if defined (CONFIG_ACPI) && defined (CONFIG_HOTPLUG_CPU)
diff --git a/drivers/acpi/processor_core.c b/drivers/acpi/processor_core.c
index a3cc8a9..89a9ad5 100644
--- a/drivers/acpi/processor_core.c
+++ b/drivers/acpi/processor_core.c
@@ -590,6 +590,11 @@ static int acpi_processor_get_info(struct acpi_processor *pr, unsigned has_uid)
ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Processor [%d:%d]\n", pr->id,
pr->acpi_id));
+ status = acpi_evaluate_object(pr->handle, "_SUN", NULL, &buffer);
+ if (ACPI_FAILURE(status))
+ object.integer.value = -1;
+ arch_set_topology_slot(pr->id, object.integer.value);
+
if (!object.processor.pblk_address)
ACPI_DEBUG_PRINT((ACPI_DB_INFO, "No PBLK (NULL address)\n"));
else if (object.processor.pblk_length != 6)
diff --git a/drivers/base/topology.c b/drivers/base/topology.c
index e1d3ad4..4fcffca 100644
--- a/drivers/base/topology.c
+++ b/drivers/base/topology.c
@@ -57,6 +57,14 @@ define_one_ro(physical_package_id);
#define ref_physical_package_id_attr
#endif
+#ifdef topology_slot
+define_id_show_func(slot);
+define_one_ro(slot);
+#define ref_slot_attr &attr_slot.attr,
+#else
+#define ref_slot_attr
+#endif
+
#ifdef topology_core_id
define_id_show_func(core_id);
define_one_ro(core_id);
@@ -83,6 +91,7 @@ define_one_ro(core_siblings);
static struct attribute *default_attrs[] = {
ref_physical_package_id_attr
+ ref_slot_attr
ref_core_id_attr
ref_thread_siblings_attr
ref_core_siblings_attr
diff --git a/include/asm-ia64/cpu.h b/include/asm-ia64/cpu.h
index e87fa32..393e545 100644
--- a/include/asm-ia64/cpu.h
+++ b/include/asm-ia64/cpu.h
@@ -14,6 +14,7 @@ DECLARE_PER_CPU(struct ia64_cpu, cpu_devices);
DECLARE_PER_CPU(int, cpu_state);
+extern void arch_set_topology_slot(int num, u32 slot);
extern int arch_register_cpu(int num);
#ifdef CONFIG_HOTPLUG_CPU
extern void arch_unregister_cpu(int);
diff --git a/include/asm-ia64/processor.h b/include/asm-ia64/processor.h
index 741f7ec..066553c 100644
--- a/include/asm-ia64/processor.h
+++ b/include/asm-ia64/processor.h
@@ -125,6 +125,10 @@ struct ia64_psr {
*/
struct cpuinfo_ia64 {
__u32 softirq_pending;
+
+#ifdef CONFIG_SMP
+ int cpu;
+#endif
__u64 itm_delta; /* # of clock cycles between clock ticks */
__u64 itm_next; /* interval timer mask value to use for next clock tick */
__u64 nsec_per_cyc; /* (1000000000<<IA64_NSEC_PER_CYC_SHIFT)/itc_freq */
@@ -140,7 +144,6 @@ struct cpuinfo_ia64 {
#ifdef CONFIG_SMP
__u64 loops_per_jiffy;
- int cpu;
__u32 socket_id; /* physical processor socket id */
__u16 core_id; /* core id */
__u16 thread_id; /* thread id */
@@ -150,6 +153,8 @@ struct cpuinfo_ia64 {
__u8 threads_per_core; /* Threads per core */
#endif
+ __u32 slot; /* physical slot; can differ from socket_id */
+
/* CPUID-derived information: */
__u64 ppn;
__u64 features;
diff --git a/include/asm-ia64/topology.h b/include/asm-ia64/topology.h
index 2d67b72..3b0ab1c 100644
--- a/include/asm-ia64/topology.h
+++ b/include/asm-ia64/topology.h
@@ -110,6 +110,7 @@ void build_cpu_to_node_map(void);
#ifdef CONFIG_SMP
#define topology_physical_package_id(cpu) (cpu_data(cpu)->socket_id)
+#define topology_slot(cpu) (cpu_data(cpu)->slot)
#define topology_core_id(cpu) (cpu_data(cpu)->core_id)
#define topology_core_siblings(cpu) (cpu_core_map[cpu])
#define topology_thread_siblings(cpu) (per_cpu(cpu_sibling_map, cpu))
diff --git a/include/asm-x86/cpu.h b/include/asm-x86/cpu.h
index 73f2ea8..43132a2 100644
--- a/include/asm-x86/cpu.h
+++ b/include/asm-x86/cpu.h
@@ -11,6 +11,7 @@ struct x86_cpu {
struct cpu cpu;
};
+extern void arch_set_topology_slot(int num, u32 slot);
#ifdef CONFIG_HOTPLUG_CPU
extern int arch_register_cpu(int num);
extern void arch_unregister_cpu(int);
--
1.5.3.1.g1e61
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists