Re: [Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
At 11/15/2012 07:34 AM, Rafael J. Wysocki Wrote: > On Tuesday, November 13, 2012 10:04:53 AM Wen Congyang wrote: >> At 11/13/2012 05:00 AM, Toshi Kani Wrote: >>> On Thu, 2012-11-08 at 19:04 +0800, Wen Congyang wrote: The memory device can be removed by 2 ways: 1. send eject request by SCI 2. echo 1 >/sys/bus/pci/devices/PNP0C80:XX/eject This 2 events may happen at the same time, so we may touch acpi_memory_device.res_list at the same time. This patch introduce a lock to protect this list. >>> >>> Hi Wen, >>> >>> This race condition is not unique in memory hot-remove as the sysfs >>> eject interface is created for all objects with _EJ0. For CPU >>> hot-remove, I addressed this race condition by making the notify handler >>> to run the hot-remove operation on kacpi_hotplug_wq by calling >>> acpi_os_hotplug_execute(). This serializes the hot-remove operations >>> among the two events since the sysfs eject also runs on >>> kacpi_hotplug_wq. This way is much simpler and is easy to maintain, >>> although it does not allow both operations to run simultaneously (which >>> I do not think we need). Can it be used for memory hot-remove as well? >> >> Good idea. I will update it. > > Still waiting. :-) > > But if you want that in v3.8, please repost ASAP. I think I will send it today. It is in test now. Thanks Wen Congyang > > Thanks, > Rafael > > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
On Tuesday, November 13, 2012 10:04:53 AM Wen Congyang wrote: > At 11/13/2012 05:00 AM, Toshi Kani Wrote: > > On Thu, 2012-11-08 at 19:04 +0800, Wen Congyang wrote: > >> The memory device can be removed by 2 ways: > >> 1. send eject request by SCI > >> 2. echo 1 >/sys/bus/pci/devices/PNP0C80:XX/eject > >> > >> This 2 events may happen at the same time, so we may touch > >> acpi_memory_device.res_list at the same time. This patch > >> introduce a lock to protect this list. > > > > Hi Wen, > > > > This race condition is not unique in memory hot-remove as the sysfs > > eject interface is created for all objects with _EJ0. For CPU > > hot-remove, I addressed this race condition by making the notify handler > > to run the hot-remove operation on kacpi_hotplug_wq by calling > > acpi_os_hotplug_execute(). This serializes the hot-remove operations > > among the two events since the sysfs eject also runs on > > kacpi_hotplug_wq. This way is much simpler and is easy to maintain, > > although it does not allow both operations to run simultaneously (which > > I do not think we need). Can it be used for memory hot-remove as well? > > Good idea. I will update it. Still waiting. :-) But if you want that in v3.8, please repost ASAP. Thanks, Rafael -- I speak only for myself. Rafael J. Wysocki, Intel Open Source Technology Center. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
On Tuesday, November 13, 2012 10:04:53 AM Wen Congyang wrote: At 11/13/2012 05:00 AM, Toshi Kani Wrote: On Thu, 2012-11-08 at 19:04 +0800, Wen Congyang wrote: The memory device can be removed by 2 ways: 1. send eject request by SCI 2. echo 1 /sys/bus/pci/devices/PNP0C80:XX/eject This 2 events may happen at the same time, so we may touch acpi_memory_device.res_list at the same time. This patch introduce a lock to protect this list. Hi Wen, This race condition is not unique in memory hot-remove as the sysfs eject interface is created for all objects with _EJ0. For CPU hot-remove, I addressed this race condition by making the notify handler to run the hot-remove operation on kacpi_hotplug_wq by calling acpi_os_hotplug_execute(). This serializes the hot-remove operations among the two events since the sysfs eject also runs on kacpi_hotplug_wq. This way is much simpler and is easy to maintain, although it does not allow both operations to run simultaneously (which I do not think we need). Can it be used for memory hot-remove as well? Good idea. I will update it. Still waiting. :-) But if you want that in v3.8, please repost ASAP. Thanks, Rafael -- I speak only for myself. Rafael J. Wysocki, Intel Open Source Technology Center. -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
At 11/15/2012 07:34 AM, Rafael J. Wysocki Wrote: On Tuesday, November 13, 2012 10:04:53 AM Wen Congyang wrote: At 11/13/2012 05:00 AM, Toshi Kani Wrote: On Thu, 2012-11-08 at 19:04 +0800, Wen Congyang wrote: The memory device can be removed by 2 ways: 1. send eject request by SCI 2. echo 1 /sys/bus/pci/devices/PNP0C80:XX/eject This 2 events may happen at the same time, so we may touch acpi_memory_device.res_list at the same time. This patch introduce a lock to protect this list. Hi Wen, This race condition is not unique in memory hot-remove as the sysfs eject interface is created for all objects with _EJ0. For CPU hot-remove, I addressed this race condition by making the notify handler to run the hot-remove operation on kacpi_hotplug_wq by calling acpi_os_hotplug_execute(). This serializes the hot-remove operations among the two events since the sysfs eject also runs on kacpi_hotplug_wq. This way is much simpler and is easy to maintain, although it does not allow both operations to run simultaneously (which I do not think we need). Can it be used for memory hot-remove as well? Good idea. I will update it. Still waiting. :-) But if you want that in v3.8, please repost ASAP. I think I will send it today. It is in test now. Thanks Wen Congyang Thanks, Rafael -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
At 11/13/2012 05:00 AM, Toshi Kani Wrote: > On Thu, 2012-11-08 at 19:04 +0800, Wen Congyang wrote: >> The memory device can be removed by 2 ways: >> 1. send eject request by SCI >> 2. echo 1 >/sys/bus/pci/devices/PNP0C80:XX/eject >> >> This 2 events may happen at the same time, so we may touch >> acpi_memory_device.res_list at the same time. This patch >> introduce a lock to protect this list. > > Hi Wen, > > This race condition is not unique in memory hot-remove as the sysfs > eject interface is created for all objects with _EJ0. For CPU > hot-remove, I addressed this race condition by making the notify handler > to run the hot-remove operation on kacpi_hotplug_wq by calling > acpi_os_hotplug_execute(). This serializes the hot-remove operations > among the two events since the sysfs eject also runs on > kacpi_hotplug_wq. This way is much simpler and is easy to maintain, > although it does not allow both operations to run simultaneously (which > I do not think we need). Can it be used for memory hot-remove as well? Good idea. I will update it. Thanks Wen Congyang > > Thanks, > -Toshi > > -- > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > the body of a message to majord...@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > Please read the FAQ at http://www.tux.org/lkml/ > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
On Thu, 2012-11-08 at 19:04 +0800, Wen Congyang wrote: > The memory device can be removed by 2 ways: > 1. send eject request by SCI > 2. echo 1 >/sys/bus/pci/devices/PNP0C80:XX/eject > > This 2 events may happen at the same time, so we may touch > acpi_memory_device.res_list at the same time. This patch > introduce a lock to protect this list. Hi Wen, This race condition is not unique in memory hot-remove as the sysfs eject interface is created for all objects with _EJ0. For CPU hot-remove, I addressed this race condition by making the notify handler to run the hot-remove operation on kacpi_hotplug_wq by calling acpi_os_hotplug_execute(). This serializes the hot-remove operations among the two events since the sysfs eject also runs on kacpi_hotplug_wq. This way is much simpler and is easy to maintain, although it does not allow both operations to run simultaneously (which I do not think we need). Can it be used for memory hot-remove as well? Thanks, -Toshi -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
On Thu, 2012-11-08 at 19:04 +0800, Wen Congyang wrote: The memory device can be removed by 2 ways: 1. send eject request by SCI 2. echo 1 /sys/bus/pci/devices/PNP0C80:XX/eject This 2 events may happen at the same time, so we may touch acpi_memory_device.res_list at the same time. This patch introduce a lock to protect this list. Hi Wen, This race condition is not unique in memory hot-remove as the sysfs eject interface is created for all objects with _EJ0. For CPU hot-remove, I addressed this race condition by making the notify handler to run the hot-remove operation on kacpi_hotplug_wq by calling acpi_os_hotplug_execute(). This serializes the hot-remove operations among the two events since the sysfs eject also runs on kacpi_hotplug_wq. This way is much simpler and is easy to maintain, although it does not allow both operations to run simultaneously (which I do not think we need). Can it be used for memory hot-remove as well? Thanks, -Toshi -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
At 11/13/2012 05:00 AM, Toshi Kani Wrote: On Thu, 2012-11-08 at 19:04 +0800, Wen Congyang wrote: The memory device can be removed by 2 ways: 1. send eject request by SCI 2. echo 1 /sys/bus/pci/devices/PNP0C80:XX/eject This 2 events may happen at the same time, so we may touch acpi_memory_device.res_list at the same time. This patch introduce a lock to protect this list. Hi Wen, This race condition is not unique in memory hot-remove as the sysfs eject interface is created for all objects with _EJ0. For CPU hot-remove, I addressed this race condition by making the notify handler to run the hot-remove operation on kacpi_hotplug_wq by calling acpi_os_hotplug_execute(). This serializes the hot-remove operations among the two events since the sysfs eject also runs on kacpi_hotplug_wq. This way is much simpler and is easy to maintain, although it does not allow both operations to run simultaneously (which I do not think we need). Can it be used for memory hot-remove as well? Good idea. I will update it. Thanks Wen Congyang Thanks, -Toshi -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
The memory device can be removed by 2 ways: 1. send eject request by SCI 2. echo 1 >/sys/bus/pci/devices/PNP0C80:XX/eject This 2 events may happen at the same time, so we may touch acpi_memory_device.res_list at the same time. This patch introduce a lock to protect this list. CC: David Rientjes CC: Jiang Liu CC: Len Brown CC: Benjamin Herrenschmidt CC: Paul Mackerras CC: Christoph Lameter Cc: Minchan Kim CC: Andrew Morton CC: KOSAKI Motohiro CC: Yasuaki Ishimatsu CC: Rafael J. Wysocki CC: Konrad Rzeszutek Wilk Signed-off-by: Wen Congyang --- The commit in pm tree is 85fcb375 drivers/acpi/acpi_memhotplug.c | 21 ++--- 1 file changed, 18 insertions(+), 3 deletions(-) diff --git a/drivers/acpi/acpi_memhotplug.c b/drivers/acpi/acpi_memhotplug.c index 1e90e8f..4c18ee3 100644 --- a/drivers/acpi/acpi_memhotplug.c +++ b/drivers/acpi/acpi_memhotplug.c @@ -83,7 +83,8 @@ struct acpi_memory_info { struct acpi_memory_device { struct acpi_device * device; unsigned int state; /* State of the memory device */ - struct list_head res_list; + struct mutex list_lock; + struct list_head res_list; /* protected by list_lock */ }; static int acpi_hotmem_initialized; @@ -101,19 +102,23 @@ acpi_memory_get_resource(struct acpi_resource *resource, void *context) (address64.resource_type != ACPI_MEMORY_RANGE)) return AE_OK; + mutex_lock(_device->list_lock); list_for_each_entry(info, _device->res_list, list) { /* Can we combine the resource range information? */ if ((info->caching == address64.info.mem.caching) && (info->write_protect == address64.info.mem.write_protect) && (info->start_addr + info->length == address64.minimum)) { info->length += address64.address_length; + mutex_unlock(_device->list_lock); return AE_OK; } } new = kzalloc(sizeof(struct acpi_memory_info), GFP_KERNEL); - if (!new) + if (!new) { + mutex_unlock(_device->list_lock); return AE_ERROR; + } INIT_LIST_HEAD(>list); new->caching = address64.info.mem.caching; @@ -121,6 +126,7 @@ acpi_memory_get_resource(struct acpi_resource *resource, void *context) new->start_addr = address64.minimum; new->length = address64.address_length; list_add_tail(>list, _device->res_list); + mutex_unlock(_device->list_lock); return AE_OK; } @@ -138,9 +144,11 @@ acpi_memory_get_device_resources(struct acpi_memory_device *mem_device) status = acpi_walk_resources(mem_device->device->handle, METHOD_NAME__CRS, acpi_memory_get_resource, mem_device); if (ACPI_FAILURE(status)) { + mutex_lock(_device->list_lock); list_for_each_entry_safe(info, n, _device->res_list, list) kfree(info); INIT_LIST_HEAD(_device->res_list); + mutex_unlock(_device->list_lock); return -EINVAL; } @@ -236,6 +244,7 @@ static int acpi_memory_enable_device(struct acpi_memory_device *mem_device) * We don't have memory-hot-add rollback function,now. * (i.e. memory-hot-remove function) */ + mutex_lock(_device->list_lock); list_for_each_entry(info, _device->res_list, list) { if (info->enabled) { /* just sanity check...*/ num_enabled++; @@ -256,6 +265,7 @@ static int acpi_memory_enable_device(struct acpi_memory_device *mem_device) info->enabled = 1; num_enabled++; } + mutex_unlock(_device->list_lock); if (!num_enabled) { printk(KERN_ERR PREFIX "add_memory failed\n"); mem_device->state = MEMORY_INVALID_STATE; @@ -316,14 +326,18 @@ static int acpi_memory_disable_device(struct acpi_memory_device *mem_device) * Ask the VM to offline this memory range. * Note: Assume that this function returns zero on success */ + mutex_lock(_device->list_lock); list_for_each_entry_safe(info, n, _device->res_list, list) { if (info->enabled) { result = remove_memory(info->start_addr, info->length); - if (result) + if (result) { + mutex_unlock(_device->list_lock); return result; + } } kfree(info); } + mutex_unlock(_device->list_lock); /* Power-off and eject the device */ result = acpi_memory_powerdown_device(mem_device); @@ -438,6 +452,7 @@ static int acpi_memory_device_add(struct acpi_device *device) mem_device->device = device;
[Patch v4 1/7] acpi,memory-hotplug: introduce a mutex lock to protect the list in acpi_memory_device
The memory device can be removed by 2 ways: 1. send eject request by SCI 2. echo 1 /sys/bus/pci/devices/PNP0C80:XX/eject This 2 events may happen at the same time, so we may touch acpi_memory_device.res_list at the same time. This patch introduce a lock to protect this list. CC: David Rientjes rient...@google.com CC: Jiang Liu liu...@gmail.com CC: Len Brown len.br...@intel.com CC: Benjamin Herrenschmidt b...@kernel.crashing.org CC: Paul Mackerras pau...@samba.org CC: Christoph Lameter c...@linux.com Cc: Minchan Kim minchan@gmail.com CC: Andrew Morton a...@linux-foundation.org CC: KOSAKI Motohiro kosaki.motoh...@jp.fujitsu.com CC: Yasuaki Ishimatsu isimatu.yasu...@jp.fujitsu.com CC: Rafael J. Wysocki r...@sisk.pl CC: Konrad Rzeszutek Wilk konrad.w...@oracle.com Signed-off-by: Wen Congyang we...@cn.fujitsu.com --- The commit in pm tree is 85fcb375 drivers/acpi/acpi_memhotplug.c | 21 ++--- 1 file changed, 18 insertions(+), 3 deletions(-) diff --git a/drivers/acpi/acpi_memhotplug.c b/drivers/acpi/acpi_memhotplug.c index 1e90e8f..4c18ee3 100644 --- a/drivers/acpi/acpi_memhotplug.c +++ b/drivers/acpi/acpi_memhotplug.c @@ -83,7 +83,8 @@ struct acpi_memory_info { struct acpi_memory_device { struct acpi_device * device; unsigned int state; /* State of the memory device */ - struct list_head res_list; + struct mutex list_lock; + struct list_head res_list; /* protected by list_lock */ }; static int acpi_hotmem_initialized; @@ -101,19 +102,23 @@ acpi_memory_get_resource(struct acpi_resource *resource, void *context) (address64.resource_type != ACPI_MEMORY_RANGE)) return AE_OK; + mutex_lock(mem_device-list_lock); list_for_each_entry(info, mem_device-res_list, list) { /* Can we combine the resource range information? */ if ((info-caching == address64.info.mem.caching) (info-write_protect == address64.info.mem.write_protect) (info-start_addr + info-length == address64.minimum)) { info-length += address64.address_length; + mutex_unlock(mem_device-list_lock); return AE_OK; } } new = kzalloc(sizeof(struct acpi_memory_info), GFP_KERNEL); - if (!new) + if (!new) { + mutex_unlock(mem_device-list_lock); return AE_ERROR; + } INIT_LIST_HEAD(new-list); new-caching = address64.info.mem.caching; @@ -121,6 +126,7 @@ acpi_memory_get_resource(struct acpi_resource *resource, void *context) new-start_addr = address64.minimum; new-length = address64.address_length; list_add_tail(new-list, mem_device-res_list); + mutex_unlock(mem_device-list_lock); return AE_OK; } @@ -138,9 +144,11 @@ acpi_memory_get_device_resources(struct acpi_memory_device *mem_device) status = acpi_walk_resources(mem_device-device-handle, METHOD_NAME__CRS, acpi_memory_get_resource, mem_device); if (ACPI_FAILURE(status)) { + mutex_lock(mem_device-list_lock); list_for_each_entry_safe(info, n, mem_device-res_list, list) kfree(info); INIT_LIST_HEAD(mem_device-res_list); + mutex_unlock(mem_device-list_lock); return -EINVAL; } @@ -236,6 +244,7 @@ static int acpi_memory_enable_device(struct acpi_memory_device *mem_device) * We don't have memory-hot-add rollback function,now. * (i.e. memory-hot-remove function) */ + mutex_lock(mem_device-list_lock); list_for_each_entry(info, mem_device-res_list, list) { if (info-enabled) { /* just sanity check...*/ num_enabled++; @@ -256,6 +265,7 @@ static int acpi_memory_enable_device(struct acpi_memory_device *mem_device) info-enabled = 1; num_enabled++; } + mutex_unlock(mem_device-list_lock); if (!num_enabled) { printk(KERN_ERR PREFIX add_memory failed\n); mem_device-state = MEMORY_INVALID_STATE; @@ -316,14 +326,18 @@ static int acpi_memory_disable_device(struct acpi_memory_device *mem_device) * Ask the VM to offline this memory range. * Note: Assume that this function returns zero on success */ + mutex_lock(mem_device-list_lock); list_for_each_entry_safe(info, n, mem_device-res_list, list) { if (info-enabled) { result = remove_memory(info-start_addr, info-length); - if (result) + if (result) { + mutex_unlock(mem_device-list_lock); return result; + } } kfree(info); } +