[v2] PCI: Restore the original INTX_DISABLE bit by pcim_intx()

[PATCH v2] PCI: Restore the original INTX_DISABLE bit by pcim_intx()

Posted by Takashi Iwai 1 year, 3 months ago

pcim_intx() tries to restore the INTx bit at removal via devres, but
there is a chance that it restores a wrong value.
Because the value to be restored is blindly assumed to be the negative
of the enable argument, when a driver calls pcim_intx() unnecessarily
for the already enabled state, it'll restore to the disabled state in
turn.  That is, the function assumes the case like:

  // INTx == 1
  pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> correct

but it might be like the following, too:

  // INTx == 0
  pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> wrong

Also, when a driver calls pcim_intx() multiple times with different
enable argument values, the last one will win no matter what value it
is.  This can lead to inconsistency, e.g.

  // INTx == 1
  pcim_intx(pdev, 0); // OK
  ...
  pcim_intx(pdev, 1); // now old INTx wrongly assumed to be 0

This patch addresses those inconsistencies by saving the original
INTx state at the first pcim_intx() call.  For that,
get_or_create_intx_devres() is folded into pcim_intx() caller side;
it allows us to simply check the already allocated devres and record
the original INTx along with the devres_alloc() call.

Fixes: 25216afc9db5 ("PCI: Add managed pcim_intx()")
Cc: stable@vger.kernel.org # 6.11+
Link: https://lore.kernel.org/87v7xk2ps5.wl-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
---
v1->v2: refactoring, fold get_or_create_intx_devres() into the caller
instead of retrieving the original INTx there.
Also add comments and improve the patch description.

 drivers/pci/devres.c | 34 +++++++++++++++++++---------------
 1 file changed, 19 insertions(+), 15 deletions(-)

diff --git a/drivers/pci/devres.c b/drivers/pci/devres.c
index b133967faef8..c93d4d4499a0 100644
--- a/drivers/pci/devres.c
+++ b/drivers/pci/devres.c
@@ -438,19 +438,12 @@ static void pcim_intx_restore(struct device *dev, void *data)
 	__pcim_intx(pdev, res->orig_intx);
 }
 
-static struct pcim_intx_devres *get_or_create_intx_devres(struct device *dev)
+static void save_orig_intx(struct pci_dev *pdev, struct pcim_intx_devres *res)
 {
-	struct pcim_intx_devres *res;
+	u16 pci_command;
 
-	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
-	if (res)
-		return res;
-
-	res = devres_alloc(pcim_intx_restore, sizeof(*res), GFP_KERNEL);
-	if (res)
-		devres_add(dev, res);
-
-	return res;
+	pci_read_config_word(pdev, PCI_COMMAND, &pci_command);
+	res->orig_intx = !(pci_command & PCI_COMMAND_INTX_DISABLE);
 }
 
 /**
@@ -466,12 +459,23 @@ static struct pcim_intx_devres *get_or_create_intx_devres(struct device *dev)
 int pcim_intx(struct pci_dev *pdev, int enable)
 {
 	struct pcim_intx_devres *res;
+	struct device *dev = &pdev->dev;
 
-	res = get_or_create_intx_devres(&pdev->dev);
-	if (!res)
-		return -ENOMEM;
+	/*
+	 * pcim_intx() must only restore the INTx value that existed before the
+	 * driver was loaded, i.e., before it called pcim_intx() for the
+	 * first time.
+	 */
+	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
+	if (!res) {
+		res = devres_alloc(pcim_intx_restore, sizeof(*res), GFP_KERNEL);
+		if (!res)
+			return -ENOMEM;
+
+		save_orig_intx(pdev, res);
+		devres_add(dev, res);
+	}
 
-	res->orig_intx = !enable;
 	__pcim_intx(pdev, enable);
 
 	return 0;
-- 
2.43.0

Re: [PATCH v2] PCI: Restore the original INTX_DISABLE bit by pcim_intx()

Posted by Philipp Stanner 1 year, 3 months ago

On Thu, 2024-10-31 at 14:42 +0100, Takashi Iwai wrote:
> pcim_intx() tries to restore the INTx bit at removal via devres, but
> there is a chance that it restores a wrong value.
> Because the value to be restored is blindly assumed to be the
> negative
> of the enable argument, when a driver calls pcim_intx() unnecessarily
> for the already enabled state, it'll restore to the disabled state in
> turn.  That is, the function assumes the case like:
> 
>   // INTx == 1
>   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> correct
> 
> but it might be like the following, too:
> 
>   // INTx == 0
>   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> wrong
> 
> Also, when a driver calls pcim_intx() multiple times with different
> enable argument values, the last one will win no matter what value it
> is.  This can lead to inconsistency, e.g.
> 
>   // INTx == 1
>   pcim_intx(pdev, 0); // OK
>   ...
>   pcim_intx(pdev, 1); // now old INTx wrongly assumed to be 0
> 
> This patch addresses those inconsistencies by saving the original
> INTx state at the first pcim_intx() call.  For that,
> get_or_create_intx_devres() is folded into pcim_intx() caller side;
> it allows us to simply check the already allocated devres and record
> the original INTx along with the devres_alloc() call.
> 
> Fixes: 25216afc9db5 ("PCI: Add managed pcim_intx()")
> Cc: stable@vger.kernel.org # 6.11+
> Link: https://lore.kernel.org/87v7xk2ps5.wl-tiwai@suse.de
> Signed-off-by: Takashi Iwai <tiwai@suse.de>

Reviewed-by: Philipp Stanner <pstanner@redhat.com>

Nice!

> ---
> v1->v2: refactoring, fold get_or_create_intx_devres() into the caller
> instead of retrieving the original INTx there.
> Also add comments and improve the patch description.
> 
>  drivers/pci/devres.c | 34 +++++++++++++++++++---------------
>  1 file changed, 19 insertions(+), 15 deletions(-)
> 
> diff --git a/drivers/pci/devres.c b/drivers/pci/devres.c
> index b133967faef8..c93d4d4499a0 100644
> --- a/drivers/pci/devres.c
> +++ b/drivers/pci/devres.c
> @@ -438,19 +438,12 @@ static void pcim_intx_restore(struct device
> *dev, void *data)
>  	__pcim_intx(pdev, res->orig_intx);
>  }
>  
> -static struct pcim_intx_devres *get_or_create_intx_devres(struct
> device *dev)
> +static void save_orig_intx(struct pci_dev *pdev, struct
> pcim_intx_devres *res)
>  {
> -	struct pcim_intx_devres *res;
> +	u16 pci_command;
>  
> -	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
> -	if (res)
> -		return res;
> -
> -	res = devres_alloc(pcim_intx_restore, sizeof(*res),
> GFP_KERNEL);
> -	if (res)
> -		devres_add(dev, res);
> -
> -	return res;
> +	pci_read_config_word(pdev, PCI_COMMAND, &pci_command);
> +	res->orig_intx = !(pci_command & PCI_COMMAND_INTX_DISABLE);
>  }
>  
>  /**
> @@ -466,12 +459,23 @@ static struct pcim_intx_devres
> *get_or_create_intx_devres(struct device *dev)
>  int pcim_intx(struct pci_dev *pdev, int enable)
>  {
>  	struct pcim_intx_devres *res;
> +	struct device *dev = &pdev->dev;
>  
> -	res = get_or_create_intx_devres(&pdev->dev);
> -	if (!res)
> -		return -ENOMEM;
> +	/*
> +	 * pcim_intx() must only restore the INTx value that existed
> before the
> +	 * driver was loaded, i.e., before it called pcim_intx() for
> the
> +	 * first time.
> +	 */
> +	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
> +	if (!res) {
> +		res = devres_alloc(pcim_intx_restore, sizeof(*res),
> GFP_KERNEL);
> +		if (!res)
> +			return -ENOMEM;
> +
> +		save_orig_intx(pdev, res);
> +		devres_add(dev, res);
> +	}
>  
> -	res->orig_intx = !enable;
>  	__pcim_intx(pdev, enable);
>  
>  	return 0;

Re: [PATCH v2] PCI: Restore the original INTX_DISABLE bit by pcim_intx()

Posted by Philipp Stanner 1 year, 2 months ago

On Mon, 2024-11-04 at 10:14 +0100, Philipp Stanner wrote:
> On Thu, 2024-10-31 at 14:42 +0100, Takashi Iwai wrote:
> > pcim_intx() tries to restore the INTx bit at removal via devres,
> > but
> > there is a chance that it restores a wrong value.
> > Because the value to be restored is blindly assumed to be the
> > negative
> > of the enable argument, when a driver calls pcim_intx()
> > unnecessarily
> > for the already enabled state, it'll restore to the disabled state
> > in
> > turn.  That is, the function assumes the case like:
> > 
> >   // INTx == 1
> >   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> correct
> > 
> > but it might be like the following, too:
> > 
> >   // INTx == 0
> >   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> wrong
> > 
> > Also, when a driver calls pcim_intx() multiple times with different
> > enable argument values, the last one will win no matter what value
> > it
> > is.  This can lead to inconsistency, e.g.
> > 
> >   // INTx == 1
> >   pcim_intx(pdev, 0); // OK
> >   ...
> >   pcim_intx(pdev, 1); // now old INTx wrongly assumed to be 0
> > 
> > This patch addresses those inconsistencies by saving the original
> > INTx state at the first pcim_intx() call.  For that,
> > get_or_create_intx_devres() is folded into pcim_intx() caller side;
> > it allows us to simply check the already allocated devres and
> > record
> > the original INTx along with the devres_alloc() call.
> > 
> > Fixes: 25216afc9db5 ("PCI: Add managed pcim_intx()")
> > Cc: stable@vger.kernel.org # 6.11+
> > Link: https://lore.kernel.org/87v7xk2ps5.wl-tiwai@suse.de
> > Signed-off-by: Takashi Iwai <tiwai@suse.de>
> 
> Reviewed-by: Philipp Stanner <pstanner@redhat.com>

Hello,

it seems we forgot about this patch.

Regards,
P.


> 
> Nice!
> 
> > ---
> > v1->v2: refactoring, fold get_or_create_intx_devres() into the
> > caller
> > instead of retrieving the original INTx there.
> > Also add comments and improve the patch description.
> > 
> >  drivers/pci/devres.c | 34 +++++++++++++++++++---------------
> >  1 file changed, 19 insertions(+), 15 deletions(-)
> > 
> > diff --git a/drivers/pci/devres.c b/drivers/pci/devres.c
> > index b133967faef8..c93d4d4499a0 100644
> > --- a/drivers/pci/devres.c
> > +++ b/drivers/pci/devres.c
> > @@ -438,19 +438,12 @@ static void pcim_intx_restore(struct device
> > *dev, void *data)
> >  	__pcim_intx(pdev, res->orig_intx);
> >  }
> >  
> > -static struct pcim_intx_devres *get_or_create_intx_devres(struct
> > device *dev)
> > +static void save_orig_intx(struct pci_dev *pdev, struct
> > pcim_intx_devres *res)
> >  {
> > -	struct pcim_intx_devres *res;
> > +	u16 pci_command;
> >  
> > -	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
> > -	if (res)
> > -		return res;
> > -
> > -	res = devres_alloc(pcim_intx_restore, sizeof(*res),
> > GFP_KERNEL);
> > -	if (res)
> > -		devres_add(dev, res);
> > -
> > -	return res;
> > +	pci_read_config_word(pdev, PCI_COMMAND, &pci_command);
> > +	res->orig_intx = !(pci_command &
> > PCI_COMMAND_INTX_DISABLE);
> >  }
> >  
> >  /**
> > @@ -466,12 +459,23 @@ static struct pcim_intx_devres
> > *get_or_create_intx_devres(struct device *dev)
> >  int pcim_intx(struct pci_dev *pdev, int enable)
> >  {
> >  	struct pcim_intx_devres *res;
> > +	struct device *dev = &pdev->dev;
> >  
> > -	res = get_or_create_intx_devres(&pdev->dev);
> > -	if (!res)
> > -		return -ENOMEM;
> > +	/*
> > +	 * pcim_intx() must only restore the INTx value that
> > existed
> > before the
> > +	 * driver was loaded, i.e., before it called pcim_intx()
> > for
> > the
> > +	 * first time.
> > +	 */
> > +	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
> > +	if (!res) {
> > +		res = devres_alloc(pcim_intx_restore,
> > sizeof(*res),
> > GFP_KERNEL);
> > +		if (!res)
> > +			return -ENOMEM;
> > +
> > +		save_orig_intx(pdev, res);
> > +		devres_add(dev, res);
> > +	}
> >  
> > -	res->orig_intx = !enable;
> >  	__pcim_intx(pdev, enable);
> >  
> >  	return 0;
>

Re: [PATCH v2] PCI: Restore the original INTX_DISABLE bit by pcim_intx()

Posted by Takashi Iwai 1 year ago

On Mon, 09 Dec 2024 14:15:19 +0100,
Philipp Stanner wrote:
> 
> On Mon, 2024-11-04 at 10:14 +0100, Philipp Stanner wrote:
> > On Thu, 2024-10-31 at 14:42 +0100, Takashi Iwai wrote:
> > > pcim_intx() tries to restore the INTx bit at removal via devres,
> > > but
> > > there is a chance that it restores a wrong value.
> > > Because the value to be restored is blindly assumed to be the
> > > negative
> > > of the enable argument, when a driver calls pcim_intx()
> > > unnecessarily
> > > for the already enabled state, it'll restore to the disabled state
> > > in
> > > turn.  That is, the function assumes the case like:
> > > 
> > >   // INTx == 1
> > >   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> correct
> > > 
> > > but it might be like the following, too:
> > > 
> > >   // INTx == 0
> > >   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> wrong
> > > 
> > > Also, when a driver calls pcim_intx() multiple times with different
> > > enable argument values, the last one will win no matter what value
> > > it
> > > is.  This can lead to inconsistency, e.g.
> > > 
> > >   // INTx == 1
> > >   pcim_intx(pdev, 0); // OK
> > >   ...
> > >   pcim_intx(pdev, 1); // now old INTx wrongly assumed to be 0
> > > 
> > > This patch addresses those inconsistencies by saving the original
> > > INTx state at the first pcim_intx() call.  For that,
> > > get_or_create_intx_devres() is folded into pcim_intx() caller side;
> > > it allows us to simply check the already allocated devres and
> > > record
> > > the original INTx along with the devres_alloc() call.
> > > 
> > > Fixes: 25216afc9db5 ("PCI: Add managed pcim_intx()")
> > > Cc: stable@vger.kernel.org # 6.11+
> > > Link: https://lore.kernel.org/87v7xk2ps5.wl-tiwai@suse.de
> > > Signed-off-by: Takashi Iwai <tiwai@suse.de>
> > 
> > Reviewed-by: Philipp Stanner <pstanner@redhat.com>
> 
> Hello,
> 
> it seems we forgot about this patch.
> 
> Regards,
> P.

This has fallen through the cracks.
Do I need to resubmit?


thanks,

Takashi


> 
> 
> > 
> > Nice!
> > 
> > > ---
> > > v1->v2: refactoring, fold get_or_create_intx_devres() into the
> > > caller
> > > instead of retrieving the original INTx there.
> > > Also add comments and improve the patch description.
> > > 
> > >  drivers/pci/devres.c | 34 +++++++++++++++++++---------------
> > >  1 file changed, 19 insertions(+), 15 deletions(-)
> > > 
> > > diff --git a/drivers/pci/devres.c b/drivers/pci/devres.c
> > > index b133967faef8..c93d4d4499a0 100644
> > > --- a/drivers/pci/devres.c
> > > +++ b/drivers/pci/devres.c
> > > @@ -438,19 +438,12 @@ static void pcim_intx_restore(struct device
> > > *dev, void *data)
> > >  	__pcim_intx(pdev, res->orig_intx);
> > >  }
> > >  
> > > -static struct pcim_intx_devres *get_or_create_intx_devres(struct
> > > device *dev)
> > > +static void save_orig_intx(struct pci_dev *pdev, struct
> > > pcim_intx_devres *res)
> > >  {
> > > -	struct pcim_intx_devres *res;
> > > +	u16 pci_command;
> > >  
> > > -	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
> > > -	if (res)
> > > -		return res;
> > > -
> > > -	res = devres_alloc(pcim_intx_restore, sizeof(*res),
> > > GFP_KERNEL);
> > > -	if (res)
> > > -		devres_add(dev, res);
> > > -
> > > -	return res;
> > > +	pci_read_config_word(pdev, PCI_COMMAND, &pci_command);
> > > +	res->orig_intx = !(pci_command &
> > > PCI_COMMAND_INTX_DISABLE);
> > >  }
> > >  
> > >  /**
> > > @@ -466,12 +459,23 @@ static struct pcim_intx_devres
> > > *get_or_create_intx_devres(struct device *dev)
> > >  int pcim_intx(struct pci_dev *pdev, int enable)
> > >  {
> > >  	struct pcim_intx_devres *res;
> > > +	struct device *dev = &pdev->dev;
> > >  
> > > -	res = get_or_create_intx_devres(&pdev->dev);
> > > -	if (!res)
> > > -		return -ENOMEM;
> > > +	/*
> > > +	 * pcim_intx() must only restore the INTx value that
> > > existed
> > > before the
> > > +	 * driver was loaded, i.e., before it called pcim_intx()
> > > for
> > > the
> > > +	 * first time.
> > > +	 */
> > > +	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
> > > +	if (!res) {
> > > +		res = devres_alloc(pcim_intx_restore,
> > > sizeof(*res),
> > > GFP_KERNEL);
> > > +		if (!res)
> > > +			return -ENOMEM;
> > > +
> > > +		save_orig_intx(pdev, res);
> > > +		devres_add(dev, res);
> > > +	}
> > >  
> > > -	res->orig_intx = !enable;
> > >  	__pcim_intx(pdev, enable);
> > >  
> > >  	return 0;
> > 
>

Re: [PATCH v2] PCI: Restore the original INTX_DISABLE bit by pcim_intx()

Posted by Bjorn Helgaas 1 year ago

On Mon, Jan 27, 2025 at 05:00:25PM +0100, Takashi Iwai wrote:
> On Mon, 09 Dec 2024 14:15:19 +0100,
> Philipp Stanner wrote:
> > 
> > On Mon, 2024-11-04 at 10:14 +0100, Philipp Stanner wrote:
> > > On Thu, 2024-10-31 at 14:42 +0100, Takashi Iwai wrote:
> > > > pcim_intx() tries to restore the INTx bit at removal via devres,
> > > > but
> > > > there is a chance that it restores a wrong value.
> > > > Because the value to be restored is blindly assumed to be the
> > > > negative
> > > > of the enable argument, when a driver calls pcim_intx()
> > > > unnecessarily
> > > > for the already enabled state, it'll restore to the disabled state
> > > > in
> > > > turn.  That is, the function assumes the case like:
> > > > 
> > > >   // INTx == 1
> > > >   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> correct
> > > > 
> > > > but it might be like the following, too:
> > > > 
> > > >   // INTx == 0
> > > >   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> wrong
> > > > 
> > > > Also, when a driver calls pcim_intx() multiple times with different
> > > > enable argument values, the last one will win no matter what value
> > > > it
> > > > is.  This can lead to inconsistency, e.g.
> > > > 
> > > >   // INTx == 1
> > > >   pcim_intx(pdev, 0); // OK
> > > >   ...
> > > >   pcim_intx(pdev, 1); // now old INTx wrongly assumed to be 0
> > > > 
> > > > This patch addresses those inconsistencies by saving the original
> > > > INTx state at the first pcim_intx() call.  For that,
> > > > get_or_create_intx_devres() is folded into pcim_intx() caller side;
> > > > it allows us to simply check the already allocated devres and
> > > > record
> > > > the original INTx along with the devres_alloc() call.
> > > > 
> > > > Fixes: 25216afc9db5 ("PCI: Add managed pcim_intx()")
> > > > Cc: stable@vger.kernel.org # 6.11+
> > > > Link: https://lore.kernel.org/87v7xk2ps5.wl-tiwai@suse.de
> > > > Signed-off-by: Takashi Iwai <tiwai@suse.de>
> > > 
> > > Reviewed-by: Philipp Stanner <pstanner@redhat.com>
> > 
> > Hello,
> > 
> > it seems we forgot about this patch.
> > 
> > Regards,
> > P.
> 
> This has fallen through the cracks.
> Do I need to resubmit?

Oops, sorry, I did miss this.  I added to pci/for-linus for v6.14.  It
didn't apply quite cleanly, so take a look and make sure I resolved it
correctly:

https://git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=for-linus&id=d555ed45a5a10a813528c7685f432369d536ae3d

> > > Nice!
> > > 
> > > > ---
> > > > v1->v2: refactoring, fold get_or_create_intx_devres() into the
> > > > caller
> > > > instead of retrieving the original INTx there.
> > > > Also add comments and improve the patch description.
> > > > 
> > > >  drivers/pci/devres.c | 34 +++++++++++++++++++---------------
> > > >  1 file changed, 19 insertions(+), 15 deletions(-)
> > > > 
> > > > diff --git a/drivers/pci/devres.c b/drivers/pci/devres.c
> > > > index b133967faef8..c93d4d4499a0 100644
> > > > --- a/drivers/pci/devres.c
> > > > +++ b/drivers/pci/devres.c
> > > > @@ -438,19 +438,12 @@ static void pcim_intx_restore(struct device
> > > > *dev, void *data)
> > > >  	__pcim_intx(pdev, res->orig_intx);
> > > >  }
> > > >  
> > > > -static struct pcim_intx_devres *get_or_create_intx_devres(struct
> > > > device *dev)
> > > > +static void save_orig_intx(struct pci_dev *pdev, struct
> > > > pcim_intx_devres *res)
> > > >  {
> > > > -	struct pcim_intx_devres *res;
> > > > +	u16 pci_command;
> > > >  
> > > > -	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
> > > > -	if (res)
> > > > -		return res;
> > > > -
> > > > -	res = devres_alloc(pcim_intx_restore, sizeof(*res),
> > > > GFP_KERNEL);
> > > > -	if (res)
> > > > -		devres_add(dev, res);
> > > > -
> > > > -	return res;
> > > > +	pci_read_config_word(pdev, PCI_COMMAND, &pci_command);
> > > > +	res->orig_intx = !(pci_command &
> > > > PCI_COMMAND_INTX_DISABLE);
> > > >  }
> > > >  
> > > >  /**
> > > > @@ -466,12 +459,23 @@ static struct pcim_intx_devres
> > > > *get_or_create_intx_devres(struct device *dev)
> > > >  int pcim_intx(struct pci_dev *pdev, int enable)
> > > >  {
> > > >  	struct pcim_intx_devres *res;
> > > > +	struct device *dev = &pdev->dev;
> > > >  
> > > > -	res = get_or_create_intx_devres(&pdev->dev);
> > > > -	if (!res)
> > > > -		return -ENOMEM;
> > > > +	/*
> > > > +	 * pcim_intx() must only restore the INTx value that
> > > > existed
> > > > before the
> > > > +	 * driver was loaded, i.e., before it called pcim_intx()
> > > > for
> > > > the
> > > > +	 * first time.
> > > > +	 */
> > > > +	res = devres_find(dev, pcim_intx_restore, NULL, NULL);
> > > > +	if (!res) {
> > > > +		res = devres_alloc(pcim_intx_restore,
> > > > sizeof(*res),
> > > > GFP_KERNEL);
> > > > +		if (!res)
> > > > +			return -ENOMEM;
> > > > +
> > > > +		save_orig_intx(pdev, res);
> > > > +		devres_add(dev, res);
> > > > +	}
> > > >  
> > > > -	res->orig_intx = !enable;
> > > >  	__pcim_intx(pdev, enable);
> > > >  
> > > >  	return 0;
> > > 
> >

Re: [PATCH v2] PCI: Restore the original INTX_DISABLE bit by pcim_intx()

Posted by Takashi Iwai 1 year ago

On Mon, 27 Jan 2025 19:58:53 +0100,
Bjorn Helgaas wrote:
> 
> On Mon, Jan 27, 2025 at 05:00:25PM +0100, Takashi Iwai wrote:
> > On Mon, 09 Dec 2024 14:15:19 +0100,
> > Philipp Stanner wrote:
> > > 
> > > On Mon, 2024-11-04 at 10:14 +0100, Philipp Stanner wrote:
> > > > On Thu, 2024-10-31 at 14:42 +0100, Takashi Iwai wrote:
> > > > > pcim_intx() tries to restore the INTx bit at removal via devres,
> > > > > but
> > > > > there is a chance that it restores a wrong value.
> > > > > Because the value to be restored is blindly assumed to be the
> > > > > negative
> > > > > of the enable argument, when a driver calls pcim_intx()
> > > > > unnecessarily
> > > > > for the already enabled state, it'll restore to the disabled state
> > > > > in
> > > > > turn.  That is, the function assumes the case like:
> > > > > 
> > > > >   // INTx == 1
> > > > >   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> correct
> > > > > 
> > > > > but it might be like the following, too:
> > > > > 
> > > > >   // INTx == 0
> > > > >   pcim_intx(pdev, 0); // old INTx value assumed to be 1 -> wrong
> > > > > 
> > > > > Also, when a driver calls pcim_intx() multiple times with different
> > > > > enable argument values, the last one will win no matter what value
> > > > > it
> > > > > is.  This can lead to inconsistency, e.g.
> > > > > 
> > > > >   // INTx == 1
> > > > >   pcim_intx(pdev, 0); // OK
> > > > >   ...
> > > > >   pcim_intx(pdev, 1); // now old INTx wrongly assumed to be 0
> > > > > 
> > > > > This patch addresses those inconsistencies by saving the original
> > > > > INTx state at the first pcim_intx() call.  For that,
> > > > > get_or_create_intx_devres() is folded into pcim_intx() caller side;
> > > > > it allows us to simply check the already allocated devres and
> > > > > record
> > > > > the original INTx along with the devres_alloc() call.
> > > > > 
> > > > > Fixes: 25216afc9db5 ("PCI: Add managed pcim_intx()")
> > > > > Cc: stable@vger.kernel.org # 6.11+
> > > > > Link: https://lore.kernel.org/87v7xk2ps5.wl-tiwai@suse.de
> > > > > Signed-off-by: Takashi Iwai <tiwai@suse.de>
> > > > 
> > > > Reviewed-by: Philipp Stanner <pstanner@redhat.com>
> > > 
> > > Hello,
> > > 
> > > it seems we forgot about this patch.
> > > 
> > > Regards,
> > > P.
> > 
> > This has fallen through the cracks.
> > Do I need to resubmit?
> 
> Oops, sorry, I did miss this.  I added to pci/for-linus for v6.14.  It
> didn't apply quite cleanly, so take a look and make sure I resolved it
> correctly:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=for-linus&id=d555ed45a5a10a813528c7685f432369d536ae3d

Looks good to me.  Thanks!


Takashi