Hi Ishimatsu-san:

By the way, if you want to reproduce this problem, just
modify my patch1 to call __container_notify_cb() directly
in container_notify_cb(). And apply my patch2.

Then, you add a container, and remove it.
The deadlock will be triggered.

And this patch is based on Lu Yinghai's tree:
git://git.kernel.org/pub/scm/linux/kernel/git/yinghai/linux-yinghai.git
for-pci-split-pci-root-hp-2


On 10/24/2012 02:54 PM, Yasuaki Ishimatsu wrote:

Hi Tang,

2012/10/24 15:05, Tang Chen wrote:

quoted

As the comments in __acpi_os_execute() said:

	We can't run hotplug code in keventd_wq/kacpid_wq/kacpid_notify_wq
	because the hotplug code may call driver .remove() functions,
	which invoke flush_scheduled_work/acpi_os_wait_events_complete
	to flush these workqueues.

we should keep the hotplug code in kacpi_hotplug_wq.

But we have the following call series in kernel now:
	acpi_ev_queue_notify_request()
	|-->  acpi_os_execute()
	     |-->  __acpi_os_execute(type, function, context, 0)

The last parameter 0 makes the container_notify_cb() executed in
kacpi_notify_wq or kacpid_wq. So, we need to put the real hotplug code
into kacpi_hotplug_wq.

I cannot understand the purpose of the patch.
Is the patch a bug fix patch? If yes, what problem happens?

Thanks,
Yasuaki Ishimatsu

quoted

Signed-off-by: Tang Chen<redacted>
---
   drivers/acpi/container.c |   17 ++++++++++++++++-
   1 files changed, 16 insertions(+), 1 deletions(-)

diff --git a/drivers/acpi/container.c b/drivers/acpi/container.c
index 69e2d6b..d300e03 100644
--- a/drivers/acpi/container.c
+++ b/drivers/acpi/container.c

@@ -35,6 +35,7 @@
   #include<acpi/acpi_bus.h>
   #include<acpi/acpi_drivers.h>
   #include<acpi/container.h>
+#include<acpi/acpiosxf.h>

   #define PREFIX "ACPI: "

@@ -165,14 +166,21 @@ static int container_device_add(struct acpi_device **device, acpi_handle handle)
   	return result;
   }

-static void container_notify_cb(acpi_handle handle, u32 type, void *context)
+static void __container_notify_cb(struct work_struct *work)
   {
   	struct acpi_device *device = NULL;
   	int result;
   	int present;
   	acpi_status status;
+	struct acpi_hp_work *hp_work;
+	acpi_handle handle;
+	u32 type;
   	u32 ost_code = ACPI_OST_SC_NON_SPECIFIC_FAILURE; /* default */

+	hp_work = container_of(work, struct acpi_hp_work, work);
+	handle = hp_work->handle;
+	type = hp_work->type;
+
   	switch (type) {
   	case ACPI_NOTIFY_BUS_CHECK:
   		/* Fall through */

@@ -224,6 +232,13 @@ static void container_notify_cb(acpi_handle handle, u32 type, void *context)
   	return;
   }

+static void container_notify_cb(acpi_handle handle, u32 type,
+				void *context)
+{
+	alloc_acpi_hp_work(handle, type, context,
+			   __container_notify_cb);
+}
+
   static acpi_status
   container_walk_namespace_cb(acpi_handle handle,
   			    u32 lvl, void *context, void **rv)

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