On 09-03-22, 15:48, Greg Kroah-Hartman wrote: > On Wed, Mar 09, 2022 at 02:57:08PM +0100, Pavel Machek wrote: > > On Wed 2022-03-09 14:40:56, Greg Kroah-Hartman wrote: > > > On Wed, Mar 09, 2022 at 01:35:09PM +0100, Pavel Machek wrote: > > > > On Wed 2022-03-09 12:58:46, Greg Kroah-Hartman wrote: > > > > > On Wed, Mar 09, 2022 at 11:54:20AM +0100, Pavel Machek wrote: > > > > > > Hi! > > > > > > > > > > > > > From: Yongzhi Liu <lyz_cs@xxxxxxxxxx> > > > > > > > > > > > > > > [ Upstream commit 455896c53d5b803733ddd84e1bf8a430644439b6 ] > > > > > > > > > > > > > > pm_runtime_get_() increments the runtime PM usage counter even > > > > > > > when it returns an error code, thus a matching decrement is needed on > > > > > > > the error handling path to keep the counter balanced. > > > > > > > > > > > > This patch will break things. > > > > > > > > > > > > Notice that -ret is ignored (checked 4.4 and 5.10), so we don't > > > > > > actually abort/return error; we just printk. We'll do two > > > > > > pm_runtime_put's after the "fix". > > > > > > > > > > > > Please drop from -stable. > > > > > > > > > > > > It was discussed during AUTOSEL review: > > > > > > > > > > > > Date: Fri, 25 Feb 2022 14:25:10 +0800 (GMT+08:00) > > > > > > From: 刘永志 <lyz_cs@xxxxxxxxxx> > > > > > > To: pavel machek <pavel@xxxxxxx> > > > > > > Cc: sasha levin <sashal@xxxxxxxxxx>, linux-kernel@xxxxxxxxxxxxxxx, > > > > > > Subject: Re: [PATCH AUTOSEL 5.16 24/30] dmaengine: shdma: Fix runtime PM > > > > > > imbalance on error > > > > > > > > > > So 5.15 and 5.16 is ok, but older is not? > > > > > > > > I believe commit is wrong for mainline and all stable releases, and > > > > author seems to agree. Drop from everywhere. > > > > > > Is it reverted in Linus's tree yet? > > > > It will take you a minute to check. > > > > Take a look at the patch. There's no return in error path, thus doing > > runtime_put is clearly bogus. Should take you less than minute to > > verify. > > > > Please drop the patch. > > I want to have it reverted in Linus's tree as well, otherwise that's a > regression that people will hit. I have reverted now, it will be in -next tomorrow and in mainline during upcoming merge window -- ~Vinod
