lock_device_hotplug() serializes hotplug & online/offline operations.
The lock is held in common sysfs online/offline interfaces and ACPI
hotplug code paths.
try_offline_node() off-lines a node if all memory sections and cpus
are removed on the node. It is called from acpi_processor_remove()
and acpi_memory_remove_memory()->remove_memory() paths, both of which
are in the ACPI hotplug code.
try_offline_node() calls stop_machine() to stop all cpus while checking
all cpu status with the assumption that the caller is not protected from
CPU hotplug or CPU online/offline operations. However, the caller is
always serialized with lock_device_hotplug(). Also, the code needs to
be properly serialized with a lock, not by stopping all cpus at a random
place with stop_machine().
This patch removes the use of stop_machine() in try_offline_node() and
adds comments to try_offline_node() and remove_memory() that
lock_device_hotplug() is required.
Signed-off-by: Toshi Kani <[email protected]>
---
mm/memory_hotplug.c | 31 ++++++++++++++++++++++---------
1 file changed, 22 insertions(+), 9 deletions(-)
diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
index ca1dd3a..0b4b0f7 100644
--- a/mm/memory_hotplug.c
+++ b/mm/memory_hotplug.c
@@ -1674,9 +1674,8 @@ static int is_memblock_offlined_cb(struct memory_block *mem, void *arg)
return ret;
}
-static int check_cpu_on_node(void *data)
+static int check_cpu_on_node(pg_data_t *pgdat)
{
- struct pglist_data *pgdat = data;
int cpu;
for_each_present_cpu(cpu) {
@@ -1691,10 +1690,9 @@ static int check_cpu_on_node(void *data)
return 0;
}
-static void unmap_cpu_on_node(void *data)
+static void unmap_cpu_on_node(pg_data_t *pgdat)
{
#ifdef CONFIG_ACPI_NUMA
- struct pglist_data *pgdat = data;
int cpu;
for_each_possible_cpu(cpu)
@@ -1703,10 +1701,11 @@ static void unmap_cpu_on_node(void *data)
#endif
}
-static int check_and_unmap_cpu_on_node(void *data)
+static int check_and_unmap_cpu_on_node(pg_data_t *pgdat)
{
- int ret = check_cpu_on_node(data);
+ int ret;
+ ret = check_cpu_on_node(pgdat);
if (ret)
return ret;
@@ -1715,11 +1714,18 @@ static int check_and_unmap_cpu_on_node(void *data)
* the cpu_to_node() now.
*/
- unmap_cpu_on_node(data);
+ unmap_cpu_on_node(pgdat);
return 0;
}
-/* offline the node if all memory sections of this node are removed */
+/**
+ * try_offline_node
+ *
+ * Offline a node if all memory sections and cpus of the node are removed.
+ *
+ * NOTE: The caller must call lock_device_hotplug() to serialize hotplug
+ * and online/offline operations before this call.
+ */
void try_offline_node(int nid)
{
pg_data_t *pgdat = NODE_DATA(nid);
@@ -1745,7 +1751,7 @@ void try_offline_node(int nid)
return;
}
- if (stop_machine(check_and_unmap_cpu_on_node, pgdat, NULL))
+ if (check_and_unmap_cpu_on_node(pgdat))
return;
/*
@@ -1782,6 +1788,13 @@ void try_offline_node(int nid)
}
EXPORT_SYMBOL(try_offline_node);
+/**
+ * remove_memory
+ *
+ * NOTE: The caller must call lock_device_hotplug() to serialize hotplug
+ * and online/offline operations before this call, as required by
+ * try_offline_node().
+ */
void __ref remove_memory(int nid, u64 start, u64 size)
{
int ret;
On Monday, August 12, 2013 01:34:31 PM Toshi Kani wrote:
> lock_device_hotplug() serializes hotplug & online/offline operations.
> The lock is held in common sysfs online/offline interfaces and ACPI
> hotplug code paths.
>
> try_offline_node() off-lines a node if all memory sections and cpus
> are removed on the node. It is called from acpi_processor_remove()
> and acpi_memory_remove_memory()->remove_memory() paths, both of which
> are in the ACPI hotplug code.
>
> try_offline_node() calls stop_machine() to stop all cpus while checking
> all cpu status with the assumption that the caller is not protected from
> CPU hotplug or CPU online/offline operations. However, the caller is
> always serialized with lock_device_hotplug(). Also, the code needs to
> be properly serialized with a lock, not by stopping all cpus at a random
> place with stop_machine().
>
> This patch removes the use of stop_machine() in try_offline_node() and
> adds comments to try_offline_node() and remove_memory() that
> lock_device_hotplug() is required.
>
> Signed-off-by: Toshi Kani <[email protected]>
Acked-by: Rafael J. Wysocki <[email protected]>
> ---
> mm/memory_hotplug.c | 31 ++++++++++++++++++++++---------
> 1 file changed, 22 insertions(+), 9 deletions(-)
>
> diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
> index ca1dd3a..0b4b0f7 100644
> --- a/mm/memory_hotplug.c
> +++ b/mm/memory_hotplug.c
> @@ -1674,9 +1674,8 @@ static int is_memblock_offlined_cb(struct memory_block *mem, void *arg)
> return ret;
> }
>
> -static int check_cpu_on_node(void *data)
> +static int check_cpu_on_node(pg_data_t *pgdat)
> {
> - struct pglist_data *pgdat = data;
> int cpu;
>
> for_each_present_cpu(cpu) {
> @@ -1691,10 +1690,9 @@ static int check_cpu_on_node(void *data)
> return 0;
> }
>
> -static void unmap_cpu_on_node(void *data)
> +static void unmap_cpu_on_node(pg_data_t *pgdat)
> {
> #ifdef CONFIG_ACPI_NUMA
> - struct pglist_data *pgdat = data;
> int cpu;
>
> for_each_possible_cpu(cpu)
> @@ -1703,10 +1701,11 @@ static void unmap_cpu_on_node(void *data)
> #endif
> }
>
> -static int check_and_unmap_cpu_on_node(void *data)
> +static int check_and_unmap_cpu_on_node(pg_data_t *pgdat)
> {
> - int ret = check_cpu_on_node(data);
> + int ret;
>
> + ret = check_cpu_on_node(pgdat);
> if (ret)
> return ret;
>
> @@ -1715,11 +1714,18 @@ static int check_and_unmap_cpu_on_node(void *data)
> * the cpu_to_node() now.
> */
>
> - unmap_cpu_on_node(data);
> + unmap_cpu_on_node(pgdat);
> return 0;
> }
>
> -/* offline the node if all memory sections of this node are removed */
> +/**
> + * try_offline_node
> + *
> + * Offline a node if all memory sections and cpus of the node are removed.
> + *
> + * NOTE: The caller must call lock_device_hotplug() to serialize hotplug
> + * and online/offline operations before this call.
> + */
> void try_offline_node(int nid)
> {
> pg_data_t *pgdat = NODE_DATA(nid);
> @@ -1745,7 +1751,7 @@ void try_offline_node(int nid)
> return;
> }
>
> - if (stop_machine(check_and_unmap_cpu_on_node, pgdat, NULL))
> + if (check_and_unmap_cpu_on_node(pgdat))
> return;
>
> /*
> @@ -1782,6 +1788,13 @@ void try_offline_node(int nid)
> }
> EXPORT_SYMBOL(try_offline_node);
>
> +/**
> + * remove_memory
> + *
> + * NOTE: The caller must call lock_device_hotplug() to serialize hotplug
> + * and online/offline operations before this call, as required by
> + * try_offline_node().
> + */
> void __ref remove_memory(int nid, u64 start, u64 size)
> {
> int ret;
--
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
On Tue, 2013-08-13 at 13:41 +0200, Rafael J. Wysocki wrote:
> On Monday, August 12, 2013 01:34:31 PM Toshi Kani wrote:
> > lock_device_hotplug() serializes hotplug & online/offline operations.
> > The lock is held in common sysfs online/offline interfaces and ACPI
> > hotplug code paths.
> >
> > try_offline_node() off-lines a node if all memory sections and cpus
> > are removed on the node. It is called from acpi_processor_remove()
> > and acpi_memory_remove_memory()->remove_memory() paths, both of which
> > are in the ACPI hotplug code.
> >
> > try_offline_node() calls stop_machine() to stop all cpus while checking
> > all cpu status with the assumption that the caller is not protected from
> > CPU hotplug or CPU online/offline operations. However, the caller is
> > always serialized with lock_device_hotplug(). Also, the code needs to
> > be properly serialized with a lock, not by stopping all cpus at a random
> > place with stop_machine().
> >
> > This patch removes the use of stop_machine() in try_offline_node() and
> > adds comments to try_offline_node() and remove_memory() that
> > lock_device_hotplug() is required.
> >
> > Signed-off-by: Toshi Kani <[email protected]>
>
> Acked-by: Rafael J. Wysocki <[email protected]>
Thanks!
-Toshi
(8/12/13 3:34 PM), Toshi Kani wrote:
> lock_device_hotplug() serializes hotplug & online/offline operations.
> The lock is held in common sysfs online/offline interfaces and ACPI
> hotplug code paths.
>
> try_offline_node() off-lines a node if all memory sections and cpus
> are removed on the node. It is called from acpi_processor_remove()
> and acpi_memory_remove_memory()->remove_memory() paths, both of which
> are in the ACPI hotplug code.
>
> try_offline_node() calls stop_machine() to stop all cpus while checking
> all cpu status with the assumption that the caller is not protected from
> CPU hotplug or CPU online/offline operations. However, the caller is
> always serialized with lock_device_hotplug(). Also, the code needs to
> be properly serialized with a lock, not by stopping all cpus at a random
> place with stop_machine().
>
> This patch removes the use of stop_machine() in try_offline_node() and
> adds comments to try_offline_node() and remove_memory() that
> lock_device_hotplug() is required.
This patch need more verbose explanation. check_cpu_on_node() traverse cpus
and cpu hotplug seems to use cpu_hotplug_driver_lock() instead of lock_device_hotplug().
That said, the race is not happen against another memeory happen. It's likely happen
another cpu hotplug. So commenting remove_memory() doesn't make much sense.
On Wed, 2013-08-14 at 21:21 -0400, KOSAKI Motohiro wrote:
> (8/12/13 3:34 PM), Toshi Kani wrote:
> > lock_device_hotplug() serializes hotplug & online/offline operations.
> > The lock is held in common sysfs online/offline interfaces and ACPI
> > hotplug code paths.
> >
> > try_offline_node() off-lines a node if all memory sections and cpus
> > are removed on the node. It is called from acpi_processor_remove()
> > and acpi_memory_remove_memory()->remove_memory() paths, both of which
> > are in the ACPI hotplug code.
> >
> > try_offline_node() calls stop_machine() to stop all cpus while checking
> > all cpu status with the assumption that the caller is not protected from
> > CPU hotplug or CPU online/offline operations. However, the caller is
> > always serialized with lock_device_hotplug(). Also, the code needs to
> > be properly serialized with a lock, not by stopping all cpus at a random
> > place with stop_machine().
> >
> > This patch removes the use of stop_machine() in try_offline_node() and
> > adds comments to try_offline_node() and remove_memory() that
> > lock_device_hotplug() is required.
>
> This patch need more verbose explanation. check_cpu_on_node() traverse cpus
> and cpu hotplug seems to use cpu_hotplug_driver_lock() instead of lock_device_hotplug().
As described:
| lock_device_hotplug() serializes hotplug & online/offline operations.
| The lock is held in common sysfs online/offline interfaces and ACPI
| hotplug code paths.
And here are their code paths.
- CPU & Mem online/offline via sysfs online
store_online()->lock_device_hotplug()
- Mem online via sysfs state:
store_mem_state()->lock_device_hotplug()
- ACPI CPU & Mem hot-add:
acpi_scan_bus_device_check()->lock_device_hotplug()
- ACPI CPU & Mem hot-delete:
acpi_scan_hot_remove()->lock_device_hotplug()
> That said, the race is not happen against another memeory happen. It's likely happen
> another cpu hotplug. So commenting remove_memory() doesn't make much sense.
This lock scheme protects from both CPU and memory hotplug, and requires
the caller to call lock_device_hotplug().
Thanks,
-Toshi
>>> This patch removes the use of stop_machine() in try_offline_node() and
>>> adds comments to try_offline_node() and remove_memory() that
>>> lock_device_hotplug() is required.
>>
>> This patch need more verbose explanation. check_cpu_on_node() traverse cpus
>> and cpu hotplug seems to use cpu_hotplug_driver_lock() instead of lock_device_hotplug().
>
> As described:
>
> | lock_device_hotplug() serializes hotplug & online/offline operations.
> | The lock is held in common sysfs online/offline interfaces and ACPI
> | hotplug code paths.
>
> And here are their code paths.
>
> - CPU & Mem online/offline via sysfs online
> store_online()->lock_device_hotplug()
>
> - Mem online via sysfs state:
> store_mem_state()->lock_device_hotplug()
>
> - ACPI CPU & Mem hot-add:
> acpi_scan_bus_device_check()->lock_device_hotplug()
>
> - ACPI CPU & Mem hot-delete:
> acpi_scan_hot_remove()->lock_device_hotplug()
O.K.