2022-10-21 14:00:06

by Yang Yingliang

[permalink] [raw]
Subject: [PATCH -next] rfkill: replace BUG_ON() with WARN_ON() in core.c

Replace BUG_ON() with WARN_ON() to handle fault more gracefully.

Suggested-by: Johannes Berg <[email protected]>
Signed-off-by: Yang Yingliang <[email protected]>
---
net/rfkill/core.c | 25 ++++++++++++++++---------
1 file changed, 16 insertions(+), 9 deletions(-)

diff --git a/net/rfkill/core.c b/net/rfkill/core.c
index dac4fdc7488a..4f493e76520d 100644
--- a/net/rfkill/core.c
+++ b/net/rfkill/core.c
@@ -150,7 +150,8 @@ EXPORT_SYMBOL(rfkill_get_led_trigger_name);

void rfkill_set_led_trigger_name(struct rfkill *rfkill, const char *name)
{
- BUG_ON(!rfkill);
+ if (WARN_ON(!rfkill))
+ return;

rfkill->ledtrigname = name;
}
@@ -532,7 +533,8 @@ bool rfkill_set_hw_state_reason(struct rfkill *rfkill,
unsigned long flags;
bool ret, prev;

- BUG_ON(!rfkill);
+ if (WARN_ON(!rfkill))
+ return blocked;

if (WARN(reason &
~(RFKILL_HARD_BLOCK_SIGNAL | RFKILL_HARD_BLOCK_NOT_OWNER),
@@ -581,7 +583,8 @@ bool rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
unsigned long flags;
bool prev, hwblock;

- BUG_ON(!rfkill);
+ if (WARN_ON(!rfkill))
+ return blocked;

spin_lock_irqsave(&rfkill->lock, flags);
prev = !!(rfkill->state & RFKILL_BLOCK_SW);
@@ -607,8 +610,8 @@ void rfkill_init_sw_state(struct rfkill *rfkill, bool blocked)
{
unsigned long flags;

- BUG_ON(!rfkill);
- BUG_ON(rfkill->registered);
+ if (WARN_ON(!rfkill) || WARN_ON(rfkill->registered))
+ return;

spin_lock_irqsave(&rfkill->lock, flags);
__rfkill_set_sw_state(rfkill, blocked);
@@ -622,7 +625,8 @@ void rfkill_set_states(struct rfkill *rfkill, bool sw, bool hw)
unsigned long flags;
bool swprev, hwprev;

- BUG_ON(!rfkill);
+ if (WARN_ON(!rfkill))
+ return;

spin_lock_irqsave(&rfkill->lock, flags);

@@ -860,7 +864,8 @@ static int rfkill_dev_uevent(struct device *dev, struct kobj_uevent_env *env)

void rfkill_pause_polling(struct rfkill *rfkill)
{
- BUG_ON(!rfkill);
+ if (WARN_ON(!rfkill))
+ return;

if (!rfkill->ops->poll)
return;
@@ -872,7 +877,8 @@ EXPORT_SYMBOL(rfkill_pause_polling);

void rfkill_resume_polling(struct rfkill *rfkill)
{
- BUG_ON(!rfkill);
+ if (WARN_ON(!rfkill))
+ return;

if (!rfkill->ops->poll)
return;
@@ -1115,7 +1121,8 @@ EXPORT_SYMBOL(rfkill_register);

void rfkill_unregister(struct rfkill *rfkill)
{
- BUG_ON(!rfkill);
+ if (WARN_ON(!rfkill))
+ return;

if (rfkill->ops->poll)
cancel_delayed_work_sync(&rfkill->poll_work);
--
2.25.1


2022-10-23 09:13:38

by Leon Romanovsky

[permalink] [raw]
Subject: Re: [PATCH -next] rfkill: replace BUG_ON() with WARN_ON() in core.c

On Fri, Oct 21, 2022 at 09:57:38PM +0800, Yang Yingliang wrote:
> Replace BUG_ON() with WARN_ON() to handle fault more gracefully.
>
> Suggested-by: Johannes Berg <[email protected]>
> Signed-off-by: Yang Yingliang <[email protected]>
> ---
> net/rfkill/core.c | 25 ++++++++++++++++---------
> 1 file changed, 16 insertions(+), 9 deletions(-)

Please add changelog and version numbers when you set your series.

The same comment as https://lore.kernel.org/all/Y1T3a1y/pWdbt2ow@unreal
or you should delete BUG_ONs completely or simply replace them with WARN_ONs.

There is no need in all these if (...).

Thanks

2022-10-24 02:05:42

by Yang Yingliang

[permalink] [raw]
Subject: Re: [PATCH -next] rfkill: replace BUG_ON() with WARN_ON() in core.c


On 2022/10/23 16:16, Leon Romanovsky wrote:
> On Fri, Oct 21, 2022 at 09:57:38PM +0800, Yang Yingliang wrote:
>> Replace BUG_ON() with WARN_ON() to handle fault more gracefully.
>>
>> Suggested-by: Johannes Berg <[email protected]>
>> Signed-off-by: Yang Yingliang <[email protected]>
>> ---
>> net/rfkill/core.c | 25 ++++++++++++++++---------
>> 1 file changed, 16 insertions(+), 9 deletions(-)
> Please add changelog and version numbers when you set your series.
>
> The same comment as https://lore.kernel.org/all/Y1T3a1y/pWdbt2ow@unreal
The link is unreachable.
> or you should delete BUG_ONs completely or simply replace them with WARN_ONs.
>
> There is no need in all these if (...).
If remove BUG_ONs or not use if (...), it may lead null-ptr-deref, it's
same as
using BUG_ON(), which leads system crash.

Thanks,
Yang
>
> Thanks
> .

2022-10-24 08:44:20

by Leon Romanovsky

[permalink] [raw]
Subject: Re: [PATCH -next] rfkill: replace BUG_ON() with WARN_ON() in core.c

On Mon, Oct 24, 2022 at 09:58:59AM +0800, Yang Yingliang wrote:
>
> On 2022/10/23 16:16, Leon Romanovsky wrote:
> > On Fri, Oct 21, 2022 at 09:57:38PM +0800, Yang Yingliang wrote:
> > > Replace BUG_ON() with WARN_ON() to handle fault more gracefully.
> > >
> > > Suggested-by: Johannes Berg <[email protected]>
> > > Signed-off-by: Yang Yingliang <[email protected]>
> > > ---
> > > net/rfkill/core.c | 25 ++++++++++++++++---------
> > > 1 file changed, 16 insertions(+), 9 deletions(-)
> > Please add changelog and version numbers when you set your series.
> >
> > The same comment as https://lore.kernel.org/all/Y1T3a1y/pWdbt2ow@unreal
> The link is unreachable.

Try this https://lore.kernel.org/netdev/Y1T3a1y%2FpWdbt2ow@unreal/

> > or you should delete BUG_ONs completely or simply replace them with WARN_ONs.
> >
> > There is no need in all these if (...).
> If remove BUG_ONs or not use if (...), it may lead null-ptr-deref, it's same
> as
> using BUG_ON(), which leads system crash.

May or will? Do you have crash report in hand?

This is rfkill API and an expectation is to have valid struct rfkill *rfkill.
Callers shouldn't call to these API functions if they know what rfkill is NULL.

Thanks

>
> Thanks,
> Yang
> >
> > Thanks
> > .