linux-kernel - Re: [RFC PATCH] extcon: Fix sleeping in atomic context after connecting USB cable

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-id: <368375031.68241411964736944.JavaMail.weblogic@epmlwas08a>
Date:	Mon, 29 Sep 2014 04:25:37 +0000 (GMT)
From:	함명주 <myungjoo.ham@...sung.com>
To:	Krzysztof Kozlowski <k.kozlowski@...sung.com>,
	최찬우 <cw00.choi@...sung.com>,
	Felipe Balbi <balbi@...com>,
	우상정 <sangjung.woo@...sung.com>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	한진구 <jg1.han@...sung.com>
Cc:	박경민 <kyungmin.park@...sung.com>,
	Marek Szyprowski <m.szyprowski@...sung.com>,
	Bartlomiej Zolnierkiewicz <b.zolnierkie@...sung.com>
Subject: Re: [RFC PATCH] extcon: Fix sleeping in atomic context after
 connecting USB cable

>   
>  Kernel built with extcon and charger-manager.
> 
> After connecting the USB cable sleeping function was called from atomic
> context:
> [   63.328648] BUG: sleeping function called from invalid context at kernel/locking/mutex.c:586
[]
> [   63.388743] Workqueue: events max14577_muic_irq_work
> [   63.393707] [<c00167a4>] (unwind_backtrace) from [<c0012c08>] (show_stack+0x10/0x14)
> [   63.401422] [<c0012c08>] (show_stack) from [<c06232ec>] (dump_stack+0x70/0xbc)
> [   63.408625] [<c06232ec>] (dump_stack) from [<c0629d0c>] (mutex_lock_nested+0x24/0x410)
> [   63.416525] [<c0629d0c>] (mutex_lock_nested) from [<c0354474>] (regmap_read+0x30/0x60)
> [   63.424424] [<c0354474>] (regmap_read) from [<c0406b84>] (max14577_charger_get_property+0x1e4/0x250)
> [   63.433535] [<c0406b84>] (max14577_charger_get_property) from [<c0403834>] (is_ext_pwr_online+0x30/0x6c)
> [   63.442994] [<c0403834>] (is_ext_pwr_online) from [<c0404a54>] (charger_extcon_notifier+0x40/0x70)
> [   63.451934] [<c0404a54>] (charger_extcon_notifier) from [<c044cbe4>] (_call_per_cable+0x40/0x4c)
> [   63.460704] [<c044cbe4>] (_call_per_cable) from [<c0051560>] (notifier_call_chain+0x64/0x128)
> [   63.469209] [<c0051560>] (notifier_call_chain) from [<c0051640>] (raw_notifier_call_chain+0x18/0x20)
> [   63.478321] [<c0051640>] (raw_notifier_call_chain) from [<c044d0fc>] (extcon_update_state+0xa8/0x204)
> [   63.487522] [<c044d0fc>] (extcon_update_state) from [<c044e5f8>] (max14577_muic_chg_handler+0xdc/0x180)
> [   63.496897] [<c044e5f8>] (max14577_muic_chg_handler) from [<c044e924>] (max14577_muic_irq_work+0x7c/0xd8)
> [   63.506445] [<c044e924>] (max14577_muic_irq_work) from [<c004480c>] (process_one_work+0x198/0x66c)
> [   63.515385] [<c004480c>] (process_one_work) from [<c0044d4c>] (worker_thread+0x38/0x564)
> [   63.523455] [<c0044d4c>] (worker_thread) from [<c004c4ec>] (kthread+0xcc/0xe8)
> [   63.530661] [<c004c4ec>] (kthread) from [<c000f1f8>] (ret_from_fork+0x14/0x3c)
> [   63.543926] charger_manager: Set current limit of CHARGER : 450000uA ~ 450000uA
> [   63.548870] extcon-port jack: jack event CHGDET=usb
> [   63.550592] extcon-port jack: jack event CHGDET=charger
> [   64.188607] charger-manager charger-manager@0: Failed to get battery temperature
> [   64.200684] charger-manager charger-manager@0: CHARGING
> 
> The first sleeping function is is_ext_pwr_online()
> (drivers/power/charger-manager.c). The atomic context initiating the
> flow is set up in extcon_update_state() (drivers/extcon/extcon-class.c).
> 
> The extcon_update_state() uses spin locks which are not necessary
> because the function is not called from interrupt service routines.
> Instead, the extcon_update_state() is called from:
> 1. Threaded interrupt handlers.
> 2. Work queues.
> 
> Replace the spin lock with mutex and update the documentation of this
> function.

No. You've done it in the opposite way.

update_state is often called in a interrupt handler that cannot sleep.

For the context you've mentioned, we'd need workqueue invoked by
_call_per_cable or notifier callback.


Cheers,
MyungJoo

> 
> Signed-off-by: Krzysztof Kozlowski <k.kozlowski@...sung.com>
> ---
>  drivers/extcon/extcon-class.c | 15 ++++++++-------
>  include/linux/extcon.h        |  3 ++-
>  2 files changed, 10 insertions(+), 8 deletions(-)
> 
[]
> -	spin_lock_irqsave(&edev->lock, flags);
> +	mutex_lock(&edev->lock);
[]