[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <5abb1e89-bf49-6c08-742a-a782111b41dc@samsung.com>
Date: Thu, 6 Dec 2018 16:07:14 +0100
From: Marek Szyprowski <m.szyprowski@...sung.com>
To: Dan Carpenter <dan.carpenter@...cle.com>
Cc: linux-kernel@...r.kernel.org, linux-usb@...r.kernel.org,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
Minas Harutyunyan <hminas@...opsys.com>,
Felipe Balbi <felipe.balbi@...ux.intel.com>,
Geert Uytterhoeven <geert@...ux-m68k.org>,
Bartlomiej Zolnierkiewicz <b.zolnierkie@...sung.com>
Subject: Re: [PATCH] usb: dwc2: Revert
"usb: dwc2: Disable all EP's on disconnect"
Hi Dan,
On 2018-12-06 15:52, Dan Carpenter wrote:
> Hi Marek,
>
> I'm surprised you don't get deadlocks when you apply this patch.
Why should I get it? It is just a revert to the state before applying
the mentioned incorrect patch.
> On Wed, Nov 21, 2018 at 04:45:04PM +0100, Marek Szyprowski wrote:
>
>> @@ -4020,9 +4008,7 @@ static int dwc2_hsotg_ep_disable(struct usb_ep *ep)
>>
>> epctrl_reg = dir_in ? DIEPCTL(index) : DOEPCTL(index);
>>
>> - locked = spin_is_locked(&hsotg->lock);
>> - if (!locked)
>> - spin_lock_irqsave(&hsotg->lock, flags);
>> + spin_lock_irqsave(&hsotg->lock, flags);
>>
> One of the callers is already holding the hsotg->log. The
> spin_is_locked() test would avoid the deadlock.
Before that broken patch, there was no call to any function which takes
spinlock again, so no deadlock.
Best regards
--
Marek Szyprowski, PhD
Samsung R&D Institute Poland
Powered by blists - more mailing lists