public inbox for linux-acpi@vger.kernel.org
 help / color / mirror / Atom feed
From: "Lee, Chun-Yi" <joeyli.kernel@gmail.com>
To: Robert Moore <robert.moore@intel.com>, Lv Zheng <lv.zheng@intel.com>
Cc: linux-acpi@vger.kernel.org, "Lee, Chun-Yi" <jlee@suse.com>
Subject: [PATCH] ACPICA: Add mutex to avoid race condition of reference count of notify object
Date: Mon,  9 Jun 2014 12:14:06 +0800	[thread overview]
Message-ID: <1402287246-10669-1-git-send-email-jlee@suse.com> (raw)

This issue found on v3.0 kernel, unfortunately there was no chance
to test latest kernel on issue mchine. This patch tested on v3.0 kernel
then sent to linux-acpi for review and note, maybe latest kernel also need.

The problem happened when acpi thermal driver evaluate _PSL, but acpi
processor driver install notify handler at the same time. In the
code path of evaluate _PSL, it updates reference count of processor
and its notify objects. When the notify handler installation done
after the reference count of processor updated, it causes the
ref_count of processor doesn't sync with its notify object's
ref_count value.

Here is an debugging log when issue reproduced:

[    3.481773] ACPI_TYPE_PROCESSOR set ACPI_DEVICE_NOTIFY, object_desc->common.reference_count: 3, notify_obj->common.reference_count: 1
[    3.481958] PROCESSOR device_hid: LNXCPU
...
[    3.487427] ACPI_TYPE_PROCESSOR, action = 1
[    3.487428] Update device_notify ref_count
[    3.487429] REF_DECREMENT ACPI_TYPE_LOCAL_NOTIFY original_count: 0
[    3.487431] ACPI Warning: Obj ffff8800b0f40b28, Reference Count is already zero, cannot decrement
[    3.487433]  (20110413/utdelete-431)
[    3.487434] REF_DECREMENT ACPI_TYPE_PROCESSOR original_count: 2

Accroding log, found the reference_count of parent object
(it's processor in this case) is 3, it doesn't match with notify_object's
reference_count, value is 1. It triggered "Reference Count is already zero"
warning, then happen object double free issue later.

To avoid rece condition, this patch introded ACPI_MTX_NOTIFY_REF_COUNT
mutex to keep the ref_count of notify object sync with its parent
object. And, it also set the reference_count value of new notify object
equals to its parent object's reference_count.

Signed-off-by: Lee, Chun-Yi <jlee@suse.com>
---
 drivers/acpi/acpica/aclocal.h  | 3 ++-
 drivers/acpi/acpica/evxface.c  | 7 ++++++-
 drivers/acpi/acpica/utdelete.c | 6 ++++++
 3 files changed, 14 insertions(+), 2 deletions(-)

diff --git a/drivers/acpi/acpica/aclocal.h b/drivers/acpi/acpica/aclocal.h
index c7f743c..e25a4af 100644
--- a/drivers/acpi/acpica/aclocal.h
+++ b/drivers/acpi/acpica/aclocal.h
@@ -85,8 +85,9 @@ union acpi_parse_object;
 #define ACPI_MTX_MEMORY                 5	/* Debug memory tracking lists */
 #define ACPI_MTX_DEBUG_CMD_COMPLETE     6	/* AML debugger */
 #define ACPI_MTX_DEBUG_CMD_READY        7	/* AML debugger */
+#define ACPI_MTX_NOTIFY_REF_COUNT	8	/* Reference count of notify object */
 
-#define ACPI_MAX_MUTEX                  7
+#define ACPI_MAX_MUTEX                  8
 #define ACPI_NUM_MUTEX                  ACPI_MAX_MUTEX+1
 
 /* Lock structure for reader/writer interfaces */
diff --git a/drivers/acpi/acpica/evxface.c b/drivers/acpi/acpica/evxface.c
index e114140..213fe1a 100644
--- a/drivers/acpi/acpica/evxface.c
+++ b/drivers/acpi/acpica/evxface.c
@@ -495,6 +495,10 @@ acpi_install_notify_handler(acpi_handle device,
 						handler, context,
 						NULL);
 
+		acpi_ut_acquire_mutex(ACPI_MTX_NOTIFY_REF_COUNT);
+
+		notify_obj->common.reference_count = obj_desc->common.reference_count;
+
 		if (handler_type & ACPI_SYSTEM_NOTIFY) {
 			obj_desc->common_notify.system_notify = notify_obj;
 		}
@@ -503,8 +507,9 @@ acpi_install_notify_handler(acpi_handle device,
 			obj_desc->common_notify.device_notify = notify_obj;
 		}
 
-		if (handler_type == ACPI_ALL_NOTIFY) {
+		acpi_ut_release_mutex(ACPI_MTX_NOTIFY_REF_COUNT);
 
+		if (handler_type == ACPI_ALL_NOTIFY) {
 			/* Extra ref if installed in both */
 
 			acpi_ut_add_reference(notify_obj);
diff --git a/drivers/acpi/acpica/utdelete.c b/drivers/acpi/acpica/utdelete.c
index 31f5a78..7559813 100644
--- a/drivers/acpi/acpica/utdelete.c
+++ b/drivers/acpi/acpica/utdelete.c
@@ -504,6 +504,7 @@ acpi_ut_update_object_reference(union acpi_operand_object *object, u16 action)
 
 			/* Update the notify objects for these types (if present) */
 
+			acpi_ut_acquire_mutex(ACPI_MTX_NOTIFY_REF_COUNT);
 			acpi_ut_update_ref_count(object->common_notify.
 						 system_notify, action);
 			acpi_ut_update_ref_count(object->common_notify.
@@ -592,6 +593,11 @@ acpi_ut_update_object_reference(union acpi_operand_object *object, u16 action)
 		 * main object to be deleted.
 		 */
 		acpi_ut_update_ref_count(object, action);
+		if (object->common.type == ACPI_TYPE_PROCESSOR ||
+		    object->common.type == ACPI_TYPE_DEVICE ||
+		    object->common.type == ACPI_TYPE_POWER ||
+		    object->common.type == ACPI_TYPE_THERMAL)
+			acpi_ut_release_mutex(ACPI_MTX_NOTIFY_REF_COUNT);
 		object = NULL;
 
 		/* Move on to the next object to be updated */
-- 
1.8.4.5


             reply	other threads:[~2014-06-09  4:18 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-06-09  4:14 Lee, Chun-Yi [this message]
  -- strict thread matches above, loose matches on Subject: below --
2014-06-16  2:11 [PATCH] ACPICA: Add mutex to avoid race condition of reference count of notify object Lee, Chun-Yi
2014-06-16  5:10 ` Zheng, Lv
2014-06-16  9:14   ` joeyli
2014-06-17  1:03     ` Zheng, Lv
2014-06-17  9:19       ` joeyli
2014-06-18  4:42         ` Zheng, Lv
2015-06-03  6:53           ` Hanjun Guo
2015-06-03  7:38             ` Zheng, Lv
2015-06-04  1:17               ` Hanjun Guo

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1402287246-10669-1-git-send-email-jlee@suse.com \
    --to=joeyli.kernel@gmail.com \
    --cc=jlee@suse.com \
    --cc=linux-acpi@vger.kernel.org \
    --cc=lv.zheng@intel.com \
    --cc=robert.moore@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox