Skip to contentSkip to conversation

ath10k: Fix un-initialized debug objects.

PATCH In review [email protected] 2 participants 7 messages Oct 24, 2013 – Oct 28, 2013 Qualcomm Atheros ATH10K wireless driver

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ath10k: Fix un-initialized debug objects.

From: Ben Greear <[email protected]>

If the 'ar' registration fails early, the debug objects were not initialized, and trouble ensued when the object was later destroyed and the un-initialized objects were accessed.

ath10k: Unable to wakeup target
ath10k: Failed to reset target, target did not wake up: -110
ath10k: early firmware event indicated
ath10k: could not start pci hif (-110)
ath10k: could not probe fw (-110)
ath10k: could not register driver core (-110)

------------[ cut here ]------------
WARNING: CPU: 0 PID: 11760 at /mnt/sda/home/greearb/git/ath/lib/debugobjects.c:260 debug_print_object+0x7c/)
ODEBUG: assert_init not available (active state 0) object type: timer_list hint: stub_timer+0x0/0x17
Modules linked in: ath10k_pci(+) ath10k_core ath5k ath9k ath9k_common ath9k_hw nfsv3 nfs_acl nfs fscache nf]
CPU: 0 PID: 11760 Comm: modprobe Tainted: G        WC   3.12.0-rc5-wl+ #2
Hardware name: To be filled by O.E.M. To be filled by O.E.M./HURONRIVER, BIOS 4.6.5 05/02/2012
 0000000000000009 ffff88020b2db8d8 ffffffff8158a1af 0000000000000006
 ffff88020b2db928 ffff88020b2db918 ffffffff8109f45e ffff88020b2db948
 ffffffff812c3274 ffff88020b2db9d0 ffffffff81a41030 ffffffff81822095
Call Trace:
 [<ffffffff8158a1af>] dump_stack+0x55/0x86
 [<ffffffff8109f45e>] warn_slowpath_common+0x77/0x91
 [<ffffffff812c3274>] ? debug_print_object+0x7c/0x8d
 [<ffffffff8109f50c>] warn_slowpath_fmt+0x41/0x43
 [<ffffffff812c3274>] debug_print_object+0x7c/0x8d
 [<ffffffff810a8324>] ? timer_debug_hint+0xa/0xa
 [<ffffffff812c3a4c>] debug_object_assert_init+0x95/0xb1
 [<ffffffff810a8fe2>] del_timer+0x1a/0x5f
 [<ffffffff810eca6d>] ? trace_hardirqs_off_caller+0x37/0xa6
 [<ffffffff810b48eb>] try_to_grab_pending+0x4c/0x141
 [<ffffffff810b4b12>] __cancel_work_timer+0x22/0xbf
 [<ffffffff810b4bbd>] cancel_delayed_work_sync+0xe/0x10
 [<ffffffffa095823e>] ath10k_debug_destroy+0x10/0x12 [ath10k_core]
 [<ffffffffa0958a25>] ath10k_core_destroy+0xf/0x33 [ath10k_core]
 [<ffffffffa0975b4d>] ath10k_pci_probe+0x2ef/0x326 [ath10k_pci]
 [<ffffffff812e2f6d>] local_pci_probe+0x39/0x61
 [<ffffffff813a0723>] ? driver_probe_device+0x2f9/0x2f9
 [<ffffffff812e3195>] pci_device_probe+0xba/0xe0
 [<ffffffff813a054f>] driver_probe_device+0x125/0x2f9
 [<ffffffff813a0723>] ? driver_probe_device+0x2f9/0x2f9
 [<ffffffff813a0771>] __driver_attach+0x4e/0x6f
 [<ffffffff8139e847>] bus_for_each_dev+0x5a/0x8c
 [<ffffffff8139fee8>] driver_attach+0x19/0x1b
 [<ffffffff8139fb1d>] bus_add_driver+0x113/0x263
 [<ffffffff813a0d61>] driver_register+0x87/0xbe
 [<ffffffff812e2605>] __pci_register_driver+0x5c/0x60
 [<ffffffffa097c000>] ? 0xffffffffa097bfff
 [<ffffffffa097c021>] ath10k_pci_init+0x21/0x3e [ath10k_pci]
 [<ffffffff81128bce>] ? jump_label_lock+0x12/0x14
 [<ffffffffa097c000>] ? 0xffffffffa097bfff
 [<ffffffff810020d9>] do_one_initcall+0x9f/0x132
 [<ffffffff81128bce>] ? jump_label_lock+0x12/0x14
 [<ffffffff810be655>] ? __blocking_notifier_call_chain+0x4c/0x5a
 [<ffffffff810f977d>] load_module+0x1beb/0x1efe
 [<ffffffff810f68b9>] ? mod_kobject_put+0x78/0x78
 [<ffffffff810f9b5c>] SyS_init_module+0xcc/0xdb
 [<ffffffff815967fd>] system_call_fastpath+0x1a/0x1f
---[ end trace 67d86d1d06d553fa ]---
---
 drivers/net/wireless/ath/ath10k/core.c  |  2 ++
 drivers/net/wireless/ath/ath10k/debug.c | 13 ++++++++-----
 drivers/net/wireless/ath/ath10k/debug.h |  6 ++++++
 3 files changed, 16 insertions(+), 5 deletions(-)
drivers/net/wireless/ath/ath10k/core.c +2 -0
index 2a360ca..282c187 100644
@@ -728,6 +728,8 @@ struct ath10k *ath10k_core_create(void *hif_priv, struct device *dev,
 
 	INIT_WORK(&ar->restart_work, ath10k_core_restart);
 
+	ath10k_debug_init(ar);
+
 	return ar;
 
 err_wq:
drivers/net/wireless/ath/ath10k/debug.c +8 -5
index e4ab546..aac5227 100644
@@ -639,6 +639,14 @@ void ath10k_debug_stop(struct ath10k *ar)
 		cancel_delayed_work(&ar->debug.htt_stats_dwork);
 }
 
+
+void ath10k_debug_init(struct ath10k *ar)
+{
+	INIT_DELAYED_WORK(&ar->debug.htt_stats_dwork,
+			  ath10k_debug_htt_stats_dwork);
+	init_completion(&ar->debug.event_stats_compl);
+}
+
 int ath10k_debug_create(struct ath10k *ar)
 {
 	ar->debug.debugfs_phy = debugfs_create_dir("ath10k",
@@ -647,11 +655,6 @@ int ath10k_debug_create(struct ath10k *ar)
 	if (!ar->debug.debugfs_phy)
 		return -ENOMEM;
 
-	INIT_DELAYED_WORK(&ar->debug.htt_stats_dwork,
-			  ath10k_debug_htt_stats_dwork);
-
-	init_completion(&ar->debug.event_stats_compl);
-
 	debugfs_create_file("fw_stats", S_IRUSR, ar->debug.debugfs_phy, ar,
 			    &fops_fw_stats);
 
drivers/net/wireless/ath/ath10k/debug.h +6 -0
index 807bf03..3c12d44 100644
@@ -47,6 +47,7 @@ extern __printf(1, 2) int ath10k_warn(const char *fmt, ...);
 int ath10k_debug_start(struct ath10k *ar);
 void ath10k_debug_stop(struct ath10k *ar);
 int ath10k_debug_create(struct ath10k *ar);
+void ath10k_debug_init(struct ath10k *ar);
 void ath10k_debug_destroy(struct ath10k *ar);
 void ath10k_debug_read_service_map(struct ath10k *ar,
 				   void *service_map,
@@ -69,6 +70,11 @@ static inline int ath10k_debug_create(struct ath10k *ar)
 	return 0;
 }
 
+static inline void ath10k_debug_init(struct ath10k *ar)
+{
+	return;
+}
+
 static inline void ath10k_debug_destroy(struct ath10k *ar)
 {
 }
Signature · 7 lines
1.7.11.7


_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
#1 lore raw
Quoted 74 lines · [email protected] writes:

From: Ben Greear <[email protected]>

If the 'ar' registration fails early, the debug objects were not initialized, and trouble ensued when the object was later destroyed and the un-initialized objects were accessed.

ath10k: Unable to wakeup target
ath10k: Failed to reset target, target did not wake up: -110
ath10k: early firmware event indicated
ath10k: could not start pci hif (-110)
ath10k: could not probe fw (-110)
ath10k: could not register driver core (-110)

------------[ cut here ]------------
WARNING: CPU: 0 PID: 11760 at /mnt/sda/home/greearb/git/ath/lib/debugobjects.c:260 debug_print_object+0x7c/)
ODEBUG: assert_init not available (active state 0) object type: timer_list hint: stub_timer+0x0/0x17
Modules linked in: ath10k_pci(+) ath10k_core ath5k ath9k ath9k_common ath9k_hw nfsv3 nfs_acl nfs fscache nf]
CPU: 0 PID: 11760 Comm: modprobe Tainted: G        WC   3.12.0-rc5-wl+ #2
Hardware name: To be filled by O.E.M. To be filled by O.E.M./HURONRIVER, BIOS 4.6.5 05/02/2012
 0000000000000009 ffff88020b2db8d8 ffffffff8158a1af 0000000000000006
 ffff88020b2db928 ffff88020b2db918 ffffffff8109f45e ffff88020b2db948
 ffffffff812c3274 ffff88020b2db9d0 ffffffff81a41030 ffffffff81822095
Call Trace:
 [<ffffffff8158a1af>] dump_stack+0x55/0x86
 [<ffffffff8109f45e>] warn_slowpath_common+0x77/0x91
 [<ffffffff812c3274>] ? debug_print_object+0x7c/0x8d
 [<ffffffff8109f50c>] warn_slowpath_fmt+0x41/0x43
 [<ffffffff812c3274>] debug_print_object+0x7c/0x8d
 [<ffffffff810a8324>] ? timer_debug_hint+0xa/0xa
 [<ffffffff812c3a4c>] debug_object_assert_init+0x95/0xb1
 [<ffffffff810a8fe2>] del_timer+0x1a/0x5f
 [<ffffffff810eca6d>] ? trace_hardirqs_off_caller+0x37/0xa6
 [<ffffffff810b48eb>] try_to_grab_pending+0x4c/0x141
 [<ffffffff810b4b12>] __cancel_work_timer+0x22/0xbf
 [<ffffffff810b4bbd>] cancel_delayed_work_sync+0xe/0x10
 [<ffffffffa095823e>] ath10k_debug_destroy+0x10/0x12 [ath10k_core]
 [<ffffffffa0958a25>] ath10k_core_destroy+0xf/0x33 [ath10k_core]
 [<ffffffffa0975b4d>] ath10k_pci_probe+0x2ef/0x326 [ath10k_pci]
 [<ffffffff812e2f6d>] local_pci_probe+0x39/0x61
 [<ffffffff813a0723>] ? driver_probe_device+0x2f9/0x2f9
 [<ffffffff812e3195>] pci_device_probe+0xba/0xe0
 [<ffffffff813a054f>] driver_probe_device+0x125/0x2f9
 [<ffffffff813a0723>] ? driver_probe_device+0x2f9/0x2f9
 [<ffffffff813a0771>] __driver_attach+0x4e/0x6f
 [<ffffffff8139e847>] bus_for_each_dev+0x5a/0x8c
 [<ffffffff8139fee8>] driver_attach+0x19/0x1b
 [<ffffffff8139fb1d>] bus_add_driver+0x113/0x263
 [<ffffffff813a0d61>] driver_register+0x87/0xbe
 [<ffffffff812e2605>] __pci_register_driver+0x5c/0x60
 [<ffffffffa097c000>] ? 0xffffffffa097bfff
 [<ffffffffa097c021>] ath10k_pci_init+0x21/0x3e [ath10k_pci]
 [<ffffffff81128bce>] ? jump_label_lock+0x12/0x14
 [<ffffffffa097c000>] ? 0xffffffffa097bfff
 [<ffffffff810020d9>] do_one_initcall+0x9f/0x132
 [<ffffffff81128bce>] ? jump_label_lock+0x12/0x14
 [<ffffffff810be655>] ? __blocking_notifier_call_chain+0x4c/0x5a
 [<ffffffff810f977d>] load_module+0x1beb/0x1efe
 [<ffffffff810f68b9>] ? mod_kobject_put+0x78/0x78
 [<ffffffff810f9b5c>] SyS_init_module+0xcc/0xdb
 [<ffffffff815967fd>] system_call_fastpath+0x1a/0x1f
---[ end trace 67d86d1d06d553fa ]---
---
 drivers/net/wireless/ath/ath10k/core.c  |  2 ++
 drivers/net/wireless/ath/ath10k/debug.c | 13 ++++++++-----
 drivers/net/wireless/ath/ath10k/debug.h |  6 ++++++
 3 files changed, 16 insertions(+), 5 deletions(-)
drivers/net/wireless/ath/ath10k/core.c +0 -0
index 2a360ca..282c187 100644

drivers/net/wireless/ath/ath10k/core.c struct ath10k *ath10k_core_create(void *hif_priv, struct device *dev, · line 731

Quoted 4 lines
drivers/net/wireless/ath/ath10k/core.c +1 -0
@@ -728,6 +728,8 @@ struct ath10k *ath10k_core_create(void *hif_priv, struct device *dev,
 
 	INIT_WORK(&ar->restart_work, ath10k_core_restart);
 
+	ath10k_debug_init(ar);

For symmetry would it make more sense to move ath10k_debug_destroy() to ath10k_core_unregister()? That way we could avoid adding a new function.

Signature · 6 lines
Kalle Valo

_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
author
#2 lore raw
Quoted 9 lines · Kalle Valo

drivers/net/wireless/ath/ath10k/core.c struct ath10k *ath10k_core_create(void *hif_priv, struct device *dev, · line 731

Quoted 3 lines · [email protected] writes:
  	INIT_WORK(&ar->restart_work, ath10k_core_restart);
drivers/net/wireless/ath/ath10k/core.c +1 -0
+	ath10k_debug_init(ar);

For symmetry would it make more sense to move ath10k_debug_destroy() to ath10k_core_unregister()? That way we could avoid adding a new function.

In my opinion, you should initialize such things as early as possible so you don't have to worry so much about the various error cases leaving things un-initialized. I believe my patch accomplished that.

What new function are you planning to avoid?

Thanks,
Ben

Signature · 8 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com


_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
#3 lore raw
Quoted 11 lines · Ben Greear
Quoted 9 lines · Kalle Valo

Review

Quoted 3 lines · [email protected] writes:
  	INIT_WORK(&ar->restart_work, ath10k_core_restart);
(quoted diff) +1 -0
+	ath10k_debug_init(ar);

For symmetry would it make more sense to move ath10k_debug_destroy() to ath10k_core_unregister()? That way we could avoid adding a new function.

Quoted 3 lines

In my opinion, you should initialize such things as early as possible so you don't have to worry so much about the various error cases leaving things un-initialized. I believe my patch accomplished that.

Well, I again like symmetry and simplicity.

Quoted 1 line

What new function are you planning to avoid?

I would prefer not to create ath10k_debug_init().

What I was trying to suggest is this, I think it should fix the bug you are seeing:

drivers/net/wireless/ath/ath10k/core.c +2 -2
@@ -738,8 +738,6 @@ EXPORT_SYMBOL(ath10k_core_create);
 
 void ath10k_core_destroy(struct ath10k *ar)
 {
-       ath10k_debug_destroy(ar);
-
        flush_workqueue(ar->workqueue);
        destroy_workqueue(ar->workqueue);
 
@@ -986,6 +984,8 @@ void ath10k_core_unregister(struct ath10k *ar)
        ath10k_mac_unregister(ar);
 
        ath10k_core_free_firmware_files(ar);
+
+       ath10k_debug_destroy(ar);
 }
 EXPORT_SYMBOL(ath10k_core_unregister);
 
Signature · 6 lines
Kalle Valo

_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
author
#4 lore raw
Quoted 17 lines · Kalle Valo
Quoted 11 lines · Ben Greear
Quoted 9 lines · Kalle Valo

Review

Quoted 3 lines · [email protected] writes:
   	INIT_WORK(&ar->restart_work, ath10k_core_restart);
(quoted diff) +1 -0
+	ath10k_debug_init(ar);

For symmetry would it make more sense to move ath10k_debug_destroy() to ath10k_core_unregister()? That way we could avoid adding a new function.

Quoted 3 lines

In my opinion, you should initialize such things as early as possible so you don't have to worry so much about the various error cases leaving things un-initialized. I believe my patch accomplished that.

Quoted 8 lines

Well, I again like symmetry and simplicity.

Quoted 1 line

What new function are you planning to avoid?

I would prefer not to create ath10k_debug_init().

What I was trying to suggest is this, I think it should fix the bug you are seeing:

I don't think so. The problem is that the init logic can fail before the debug logic is ever initialized, so any cleanup that attempts to access the debug logic can blow up.

All my patch does is ensure the debug related members are always initialized, and so can be cleaned up later without problem.

Maybe your patch fixes something else though, I haven't investigated that part of the code in detail yet.

Thanks,
Ben

Quoted 21 lines
drivers/net/wireless/ath/ath10k/core.c +2 -2
@@ -738,8 +738,6 @@ EXPORT_SYMBOL(ath10k_core_create);
 
  void ath10k_core_destroy(struct ath10k *ar)
  {
-       ath10k_debug_destroy(ar);
-
         flush_workqueue(ar->workqueue);
         destroy_workqueue(ar->workqueue);
 
@@ -986,6 +984,8 @@ void ath10k_core_unregister(struct ath10k *ar)
         ath10k_mac_unregister(ar);
 
         ath10k_core_free_firmware_files(ar);
+
+       ath10k_debug_destroy(ar);
  }
  EXPORT_SYMBOL(ath10k_core_unregister);
 
Signature · 8 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com


_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
#5 lore raw
Quoted 22 lines · Ben Greear
Quoted 21 lines · Kalle Valo
Quoted 11 lines · Ben Greear
Quoted 9 lines · Kalle Valo

Review

Quoted 3 lines · [email protected] writes:
   	INIT_WORK(&ar->restart_work, ath10k_core_restart);
(quoted diff) +1 -0
+	ath10k_debug_init(ar);

For symmetry would it make more sense to move ath10k_debug_destroy() to ath10k_core_unregister()? That way we could avoid adding a new function.

Quoted 3 lines

In my opinion, you should initialize such things as early as possible so you don't have to worry so much about the various error cases leaving things un-initialized. I believe my patch accomplished that.

Well, I again like symmetry and simplicity.

Quoted 1 line

What new function are you planning to avoid?

Quoted 8 lines
Quoted 4 lines

I would prefer not to create ath10k_debug_init().

What I was trying to suggest is this, I think it should fix the bug you are seeing:

I don't think so. The problem is that the init logic can fail before the debug logic is ever initialized, so any cleanup that attempts to access the debug logic can blow up.

Please look again. My idea was to fix the logic so that ath10k_debug_destroy() is called only if ath10k_debug_create() succeeds.

Quoted 2 lines

All my patch does is ensure the debug related members are always initialized, and so can be cleaned up later without problem.

It adds an unnecessary function which is not needed AFAICS.

Signature · 6 lines
Kalle Valo

_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
author
#6 lore raw
Quoted 29 lines · Kalle Valo
Quoted 27 lines · Ben Greear
Quoted 25 lines · Kalle Valo
Quoted 11 lines · Ben Greear
Quoted 9 lines · Kalle Valo

Review

Quoted 3 lines · [email protected] writes:
   	INIT_WORK(&ar->restart_work, ath10k_core_restart);
(quoted diff) +1 -0
+	ath10k_debug_init(ar);

For symmetry would it make more sense to move ath10k_debug_destroy() to ath10k_core_unregister()? That way we could avoid adding a new function.

Quoted 3 lines

In my opinion, you should initialize such things as early as possible so you don't have to worry so much about the various error cases leaving things un-initialized. I believe my patch accomplished that.

Well, I again like symmetry and simplicity.

Quoted 1 line

What new function are you planning to avoid?

I would prefer not to create ath10k_debug_init().

What I was trying to suggest is this, I think it should fix the bug you are seeing:

Quoted 6 lines
Quoted 3 lines

I don't think so. The problem is that the init logic can fail before the debug logic is ever initialized, so any cleanup that attempts to access the debug logic can blow up.

Please look again. My idea was to fix the logic so that ath10k_debug_destroy() is called only if ath10k_debug_create() succeeds.

Ok, I think your patch will work. I'll back mine out when you push yours and will let you know if I find any more problems in this area.

Thanks,
Ben

Signature · 8 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com


_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k