diff mbox series

i2c: xiic: Fix reference count leaks.

Message ID 20200613215923.2611-1-wu000273@umn.edu
State Changes Requested
Headers show
Series i2c: xiic: Fix reference count leaks. | expand

Commit Message

wu000273@umn.edu June 13, 2020, 9:59 p.m. UTC
From: Qiushi Wu <wu000273@umn.edu>

pm_runtime_get_sync() increments the runtime PM usage counter even
when it returns an error code. Thus call pm_runtime_put_noidle()
if pm_runtime_get_sync() fails.

Fixes: 36ecbcab84d0 ("i2c: xiic: Implement power management")
Signed-off-by: Qiushi Wu <wu000273@umn.edu>
---
 drivers/i2c/busses/i2c-xiic.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

Comments

Wolfram Sang June 14, 2020, 9:09 a.m. UTC | #1
On Sat, Jun 13, 2020 at 04:59:23PM -0500, wu000273@umn.edu wrote:
> From: Qiushi Wu <wu000273@umn.edu>
> 
> pm_runtime_get_sync() increments the runtime PM usage counter even
> when it returns an error code. Thus call pm_runtime_put_noidle()
> if pm_runtime_get_sync() fails.

Can you point me to a discussion where it was decided that this is a
proper fix? I'd think we rather should fix pm_runtime_get_sync() but
maybe there are technical reasons against it.
David Laight June 17, 2020, 2:30 p.m. UTC | #2
From: Wolfram Sang
> Sent: 14 June 2020 10:10
> 
> On Sat, Jun 13, 2020 at 04:59:23PM -0500, wu000273@umn.edu wrote:
> > From: Qiushi Wu <wu000273@umn.edu>
> >
> > pm_runtime_get_sync() increments the runtime PM usage counter even
> > when it returns an error code. Thus call pm_runtime_put_noidle()
> > if pm_runtime_get_sync() fails.
> 
> Can you point me to a discussion where it was decided that this is a
> proper fix? I'd think we rather should fix pm_runtime_get_sync() but
> maybe there are technical reasons against it.

Or, if there is one place that actually needs the reference split the
code so that unusual case keeps the reference.

In one of the patches I also spotted:
	ret = pm_runtime_get_sync();
	if (ret < 0 && ret != _EAGAIN)
		...

(I think it was EAGAIN.)
I can't help feeling that is just wrong somewhere.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
Michal Simek July 13, 2020, 7:54 a.m. UTC | #3
Hi Qiushi,

On 17. 06. 20 16:30, David Laight wrote:
> From: Wolfram Sang
>> Sent: 14 June 2020 10:10
>>
>> On Sat, Jun 13, 2020 at 04:59:23PM -0500, wu000273@umn.edu wrote:
>>> From: Qiushi Wu <wu000273@umn.edu>
>>>
>>> pm_runtime_get_sync() increments the runtime PM usage counter even
>>> when it returns an error code. Thus call pm_runtime_put_noidle()
>>> if pm_runtime_get_sync() fails.
>>
>> Can you point me to a discussion where it was decided that this is a
>> proper fix? I'd think we rather should fix pm_runtime_get_sync() but
>> maybe there are technical reasons against it.
> 
> Or, if there is one place that actually needs the reference split the
> code so that unusual case keeps the reference.
> 
> In one of the patches I also spotted:
> 	ret = pm_runtime_get_sync();
> 	if (ret < 0 && ret != _EAGAIN)
> 		...
> 
> (I think it was EAGAIN.)
> I can't help feeling that is just wrong somewhere.

Qiushi: Any update on this one?

Thanks,
Michal
Michal Simek July 21, 2020, 6:02 a.m. UTC | #4
Hi Qiushi,

On 13. 07. 20 21:41, Qiushi Wu wrote:
> Hi Michal,
> I think multiple previous patches also fixed similar problems, such as
> https://patchwork.kernel.org/patch/2404751/
> https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg2227849.html
> https://patchwork.ozlabs.org/project/linux-tegra/patch/20200520095148.10995-1-dinghao.liu@zju.edu.cn/
> 
> But recently, people are discussing reimplementing
> function pm_runtime_get_sync(), because it often misguides users and
> introducing bugs. 
> So at this time point, I'm not sure if we should fix this issue in the
> current version or wait for the new implementation of this API. : )

If we apply this patch then it needs to be reverted when API is fixed.
Is there any timeline mentioned?
If this will take some time I am fine with applying this patch.

Thanks,
Michal
Michal Simek July 27, 2020, 6:46 a.m. UTC | #5
Hi,

On 27. 07. 20 8:16, Qiushi Wu wrote:
> Hi Michal,
> 
> Thanks for your reply!
> I checked the mailing list for the related emails against this API, but
> I didn't find any specific timeline they mentioned.
> Maybe we can patch this bug to prevent potential issues at this time
> point if you are willing to.

ok. I am fine with this but please send v2 and add there some comments
about it and extend commit message with more information you provided.
I just want to make sure that in several months timeframe when someone
look at this it will be clear that it was workaround.

Thanks,
Michal
diff mbox series

Patch

diff --git a/drivers/i2c/busses/i2c-xiic.c b/drivers/i2c/busses/i2c-xiic.c
index 90c1c362394d..ffec41e6be72 100644
--- a/drivers/i2c/busses/i2c-xiic.c
+++ b/drivers/i2c/busses/i2c-xiic.c
@@ -696,8 +696,10 @@  static int xiic_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs, int num)
 		xiic_getreg8(i2c, XIIC_SR_REG_OFFSET));
 
 	err = pm_runtime_get_sync(i2c->dev);
-	if (err < 0)
+	if (err < 0) {
+		pm_runtime_put_noidle(i2c->dev);
 		return err;
+	}
 
 	err = xiic_busy(i2c);
 	if (err)
@@ -860,8 +862,10 @@  static int xiic_i2c_remove(struct platform_device *pdev)
 	i2c_del_adapter(&i2c->adap);
 
 	ret = pm_runtime_get_sync(i2c->dev);
-	if (ret < 0)
+	if (ret < 0) {
+		pm_runtime_put_noidle(i2c->dev);
 		return ret;
+	}
 
 	xiic_deinit(i2c);
 	pm_runtime_put_sync(i2c->dev);