| Message ID | 20240814-coccinelle-followup-v2-2-88b4e4a9af56@chromium.org |
|---|---|
| State | Superseded |
| Headers | show |
| Series | media: Follow-up for coccinelle lock fixes patchset | expand |
Hi Mauro On Fri, 16 Aug 2024 at 10:17, Mauro Carvalho Chehab <mchehab+huawei@kernel.org> wrote: > > Em Wed, 14 Aug 2024 14:10:23 +0000 > Ricardo Ribalda <ribalda@chromium.org> escreveu: > > > Split out the wait function, and introduce some new toys: guard and > > lockdep. > > > > This fixes the following cocci warnings: > > drivers/media/dvb-core/dvb_frontend.c:2897:1-7: preceding lock on line 2776 > > drivers/media/dvb-core/dvb_frontend.c:2897:1-7: preceding lock on line 2786 > > drivers/media/dvb-core/dvb_frontend.c:2897:1-7: preceding lock on line 2809 > > Hi Ricardo, Hi Mauro > > Every time someone tries to fix this lock, we end having regression reports, > because of the diversity of devices, and the way they registers there. > > That's specially true for devices with multiple frontends and custom > zigzag methods. > > On what devices have you tested this patch? I do not have access to any device, it is just "compiled tested". I think that the patch is mainly a refactor, it does not really change how the lock is handled. Regards! > > Regards, > Mauro > > > > > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org> > > --- > > drivers/media/dvb-core/dvb_frontend.c | 58 ++++++++++++++++++++++------------- > > 1 file changed, 37 insertions(+), 21 deletions(-) > > > > diff --git a/drivers/media/dvb-core/dvb_frontend.c b/drivers/media/dvb-core/dvb_frontend.c > > index e81b9996530e..7f5d0c297464 100644 > > --- a/drivers/media/dvb-core/dvb_frontend.c > > +++ b/drivers/media/dvb-core/dvb_frontend.c > > @@ -30,6 +30,7 @@ > > #include <linux/kthread.h> > > #include <linux/ktime.h> > > #include <linux/compat.h> > > +#include <linux/lockdep.h> > > #include <asm/processor.h> > > > > #include <media/dvb_frontend.h> > > @@ -2826,6 +2827,34 @@ static int __dvb_frontend_open(struct inode *inode, struct file *file) > > return ret; > > } > > > > +static int wait_dvb_frontend(struct dvb_adapter *adapter, > > + struct dvb_device *mfedev) > > +{ > > + struct dvb_frontend *mfe = mfedev->priv; > > + struct dvb_frontend_private *mfepriv = mfe->frontend_priv; > > + int mferetry = (dvb_mfe_wait_time << 1); > > + int ret = 0; > > + > > + lockdep_assert_held(&adapter->mfe_lock); > > + > > + if (mfedev->users == -1 && !mfepriv->thread) > > + return 0; > > + > > + mutex_unlock(&adapter->mfe_lock); > > + > > + while (mferetry-- && (mfedev->users != -1 || mfepriv->thread)) { > > + if (msleep_interruptible(500)) > > + if (signal_pending(current)) { > > + ret = -EINTR; > > + break; > > + } > > + } > > + > > + mutex_lock(&adapter->mfe_lock); > > + > > + return ret; > > +} > > + > > static int dvb_frontend_open(struct inode *inode, struct file *file) > > { > > struct dvb_device *dvbdev = file->private_data; > > @@ -2840,19 +2869,16 @@ static int dvb_frontend_open(struct inode *inode, struct file *file) > > if (!adapter->mfe_shared) > > return __dvb_frontend_open(inode, file); > > > > + guard(mutex)(&adapter->mfe_lock); > > + > > if (adapter->mfe_shared == 2) { > > - mutex_lock(&adapter->mfe_lock); > > if ((file->f_flags & O_ACCMODE) != O_RDONLY) { > > if (adapter->mfe_dvbdev && > > - !adapter->mfe_dvbdev->writers) { > > - mutex_unlock(&adapter->mfe_lock); > > + !adapter->mfe_dvbdev->writers) > > return -EBUSY; > > - } > > adapter->mfe_dvbdev = dvbdev; > > } > > } else { > > - mutex_lock(&adapter->mfe_lock); > > - > > if (!adapter->mfe_dvbdev) { > > adapter->mfe_dvbdev = dvbdev; > > } else if (adapter->mfe_dvbdev != dvbdev) { > > @@ -2862,34 +2888,24 @@ static int dvb_frontend_open(struct inode *inode, struct file *file) > > *mfe = mfedev->priv; > > struct dvb_frontend_private > > *mfepriv = mfe->frontend_priv; > > - int mferetry = (dvb_mfe_wait_time << 1); > > - > > - mutex_unlock(&adapter->mfe_lock); > > - while (mferetry-- && (mfedev->users != -1 || > > - mfepriv->thread)) { > > - if (msleep_interruptible(500)) { > > - if (signal_pending(current)) > > - return -EINTR; > > - } > > - } > > > > - mutex_lock(&adapter->mfe_lock); > > + ret = wait_dvb_frontend(adapter, mfedev); > > + if (ret) > > + return ret; > > + > > if (adapter->mfe_dvbdev != dvbdev) { > > mfedev = adapter->mfe_dvbdev; > > mfe = mfedev->priv; > > mfepriv = mfe->frontend_priv; > > if (mfedev->users != -1 || > > - mfepriv->thread) { > > - mutex_unlock(&adapter->mfe_lock); > > + mfepriv->thread) > > return -EBUSY; > > - } > > adapter->mfe_dvbdev = dvbdev; > > } > > } > > } > > > > ret = __dvb_frontend_open(inode, file); > > - mutex_unlock(&adapter->mfe_lock); > > > > return ret; > > } > > > > > > Thanks, > Mauro
Em Fri, 16 Aug 2024 10:20:47 +0200 Ricardo Ribalda <ribalda@chromium.org> escreveu: > Hi Mauro > > On Fri, 16 Aug 2024 at 10:17, Mauro Carvalho Chehab > <mchehab+huawei@kernel.org> wrote: > > > > Em Wed, 14 Aug 2024 14:10:23 +0000 > > Ricardo Ribalda <ribalda@chromium.org> escreveu: > > > > > Split out the wait function, and introduce some new toys: guard and > > > lockdep. > > > > > > This fixes the following cocci warnings: > > > drivers/media/dvb-core/dvb_frontend.c:2897:1-7: preceding lock on line 2776 > > > drivers/media/dvb-core/dvb_frontend.c:2897:1-7: preceding lock on line 2786 > > > drivers/media/dvb-core/dvb_frontend.c:2897:1-7: preceding lock on line 2809 > > > > Hi Ricardo, > > Hi Mauro > > > > > Every time someone tries to fix this lock, we end having regression reports, > > because of the diversity of devices, and the way they registers there. > > > > That's specially true for devices with multiple frontends and custom > > zigzag methods. > > > > On what devices have you tested this patch? > > I do not have access to any device, it is just "compiled tested". > > I think that the patch is mainly a refactor, it does not really change > how the lock is handled. While I liked your approach, in the specific case of this lock, I have to disagree: there were at least 2 or 3 previous attempts to fix issues on it, with patches made by someone and dully tested on some hardware, ended being reverted due to corner cases with some boards. So, I'm not willing to take the risk of accept patches touching dvb frontend locking schema without tests with real hardware covering common and corner cases (multi-frontend, custom zigzag, ...) and/or a formal code verification to proof that the lock works on all circumstances. Regards, Mauro
diff --git a/drivers/media/dvb-core/dvb_frontend.c b/drivers/media/dvb-core/dvb_frontend.c index e81b9996530e..7f5d0c297464 100644 --- a/drivers/media/dvb-core/dvb_frontend.c +++ b/drivers/media/dvb-core/dvb_frontend.c @@ -30,6 +30,7 @@ #include <linux/kthread.h> #include <linux/ktime.h> #include <linux/compat.h> +#include <linux/lockdep.h> #include <asm/processor.h> #include <media/dvb_frontend.h> @@ -2826,6 +2827,34 @@ static int __dvb_frontend_open(struct inode *inode, struct file *file) return ret; } +static int wait_dvb_frontend(struct dvb_adapter *adapter, + struct dvb_device *mfedev) +{ + struct dvb_frontend *mfe = mfedev->priv; + struct dvb_frontend_private *mfepriv = mfe->frontend_priv; + int mferetry = (dvb_mfe_wait_time << 1); + int ret = 0; + + lockdep_assert_held(&adapter->mfe_lock); + + if (mfedev->users == -1 && !mfepriv->thread) + return 0; + + mutex_unlock(&adapter->mfe_lock); + + while (mferetry-- && (mfedev->users != -1 || mfepriv->thread)) { + if (msleep_interruptible(500)) + if (signal_pending(current)) { + ret = -EINTR; + break; + } + } + + mutex_lock(&adapter->mfe_lock); + + return ret; +} + static int dvb_frontend_open(struct inode *inode, struct file *file) { struct dvb_device *dvbdev = file->private_data; @@ -2840,19 +2869,16 @@ static int dvb_frontend_open(struct inode *inode, struct file *file) if (!adapter->mfe_shared) return __dvb_frontend_open(inode, file); + guard(mutex)(&adapter->mfe_lock); + if (adapter->mfe_shared == 2) { - mutex_lock(&adapter->mfe_lock); if ((file->f_flags & O_ACCMODE) != O_RDONLY) { if (adapter->mfe_dvbdev && - !adapter->mfe_dvbdev->writers) { - mutex_unlock(&adapter->mfe_lock); + !adapter->mfe_dvbdev->writers) return -EBUSY; - } adapter->mfe_dvbdev = dvbdev; } } else { - mutex_lock(&adapter->mfe_lock); - if (!adapter->mfe_dvbdev) { adapter->mfe_dvbdev = dvbdev; } else if (adapter->mfe_dvbdev != dvbdev) { @@ -2862,34 +2888,24 @@ static int dvb_frontend_open(struct inode *inode, struct file *file) *mfe = mfedev->priv; struct dvb_frontend_private *mfepriv = mfe->frontend_priv; - int mferetry = (dvb_mfe_wait_time << 1); - - mutex_unlock(&adapter->mfe_lock); - while (mferetry-- && (mfedev->users != -1 || - mfepriv->thread)) { - if (msleep_interruptible(500)) { - if (signal_pending(current)) - return -EINTR; - } - } - mutex_lock(&adapter->mfe_lock); + ret = wait_dvb_frontend(adapter, mfedev); + if (ret) + return ret; + if (adapter->mfe_dvbdev != dvbdev) { mfedev = adapter->mfe_dvbdev; mfe = mfedev->priv; mfepriv = mfe->frontend_priv; if (mfedev->users != -1 || - mfepriv->thread) { - mutex_unlock(&adapter->mfe_lock); + mfepriv->thread) return -EBUSY; - } adapter->mfe_dvbdev = dvbdev; } } } ret = __dvb_frontend_open(inode, file); - mutex_unlock(&adapter->mfe_lock); return ret; }
Split out the wait function, and introduce some new toys: guard and lockdep. This fixes the following cocci warnings: drivers/media/dvb-core/dvb_frontend.c:2897:1-7: preceding lock on line 2776 drivers/media/dvb-core/dvb_frontend.c:2897:1-7: preceding lock on line 2786 drivers/media/dvb-core/dvb_frontend.c:2897:1-7: preceding lock on line 2809 Signed-off-by: Ricardo Ribalda <ribalda@chromium.org> --- drivers/media/dvb-core/dvb_frontend.c | 58 ++++++++++++++++++++++------------- 1 file changed, 37 insertions(+), 21 deletions(-)