diff mbox

[v3,3/3] PCI: endpoint: Fix find_first_zero_bit() usage

Message ID 20171127154955.22214-4-niklas.cassel@axis.com (mailing list archive)
State New, archived
Delegated to: Bjorn Helgaas
Headers show

Commit Message

Niklas Cassel Nov. 27, 2017, 3:49 p.m. UTC
find_first_zero_bit()'s parameter 'size' is defined in bits,
not in bytes.

Calling find_first_zero_bit() with the wrong size unit
will lead to insidious bugs.

Fix this by calling find_first_zero_bit() with size
BITS_PER_LONG, rather than sizeof().

Also add proper error handling for find_first_zero_bit(),
since this was missing.

Fixes: d74679911610 ("PCI: endpoint: Introduce configfs entry for configuring EP functions")
Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
---
 drivers/pci/endpoint/pci-ep-cfs.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

Comments

Lorenzo Pieralisi Nov. 27, 2017, 5:44 p.m. UTC | #1
On Mon, Nov 27, 2017 at 04:49:55PM +0100, Niklas Cassel wrote:
> find_first_zero_bit()'s parameter 'size' is defined in bits,
> not in bytes.
> 
> Calling find_first_zero_bit() with the wrong size unit
> will lead to insidious bugs.
> 
> Fix this by calling find_first_zero_bit() with size BITS_PER_LONG,
> rather than sizeof().
> 
> Also add proper error handling for find_first_zero_bit(), since this
> was missing.

"Fix this by calling find_first_zero_bit() with size BITS_PER_LONG,
rather than sizeof() and add missing find_first_zero_bit() return
handling"

> Fixes: d74679911610 ("PCI: endpoint: Introduce configfs entry for configuring EP functions")
> Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
> ---
>  drivers/pci/endpoint/pci-ep-cfs.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/pci/endpoint/pci-ep-cfs.c b/drivers/pci/endpoint/pci-ep-cfs.c
> index e1f5adc9e113..0a22a7976580 100644
> --- a/drivers/pci/endpoint/pci-ep-cfs.c
> +++ b/drivers/pci/endpoint/pci-ep-cfs.c
> @@ -109,7 +109,12 @@ static int pci_epc_epf_link(struct config_item *epc_item,
>  		return ret;
>  
>  	func_no = find_first_zero_bit(&epc_group->function_num_map,
> -				      sizeof(epc_group->function_num_map));
> +				      BITS_PER_LONG);
> +	if (func_no >= BITS_PER_LONG) {
> +		ret = -EINVAL;
> +		dev_err(&epc->dev, "failed to link endpoint function device\n");

I would not log the error for the time being or make it more explicit:

dev_error(&epc->dev, "unsupported number of functions, failed to link
endpoint function device\n");

Side note: we _do_ know the max number of functions the epc supports
(epc->max_functions) when this code is called so we do know the
bitmap size - we can rework this code path based on that (@kishon,
thoughts ?).

Other than that:

Acked-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>

> +		goto err_func_no;
> +	}
>  	set_bit(func_no, &epc_group->function_num_map);
>  	epf->func_no = func_no;
>  
> @@ -121,6 +126,7 @@ static int pci_epc_epf_link(struct config_item *epc_item,
>  
>  err_epf_bind:
>  	clear_bit(func_no, &epc_group->function_num_map);
> +err_func_no:
>  	pci_epc_remove_epf(epc, epf);
>  
>  	return ret;
> -- 
> 2.14.2
>
David Laight Nov. 28, 2017, 9:53 a.m. UTC | #2
From: Niklas Cassel
> Sent: 27 November 2017 15:50
> find_first_zero_bit()'s parameter 'size' is defined in bits,
> not in bytes.
> 
> Calling find_first_zero_bit() with the wrong size unit
> will lead to insidious bugs.
> 
> Fix this by calling find_first_zero_bit() with size
> BITS_PER_LONG, rather than sizeof().
> 
> Also add proper error handling for find_first_zero_bit(),
> since this was missing.
> 
> Fixes: d74679911610 ("PCI: endpoint: Introduce configfs entry for configuring EP functions")
> Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
> ---
>  drivers/pci/endpoint/pci-ep-cfs.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/pci/endpoint/pci-ep-cfs.c b/drivers/pci/endpoint/pci-ep-cfs.c
> index e1f5adc9e113..0a22a7976580 100644
> --- a/drivers/pci/endpoint/pci-ep-cfs.c
> +++ b/drivers/pci/endpoint/pci-ep-cfs.c
> @@ -109,7 +109,12 @@ static int pci_epc_epf_link(struct config_item *epc_item,
>  		return ret;
> 
>  	func_no = find_first_zero_bit(&epc_group->function_num_map,
> -				      sizeof(epc_group->function_num_map));
> +				      BITS_PER_LONG);

Surely this should be either 8 * sizeof() or you should use ffz()
that takes a numeric argument rather than the function designed
for arbitrary size bitmaps.

	David
Lorenzo Pieralisi Nov. 28, 2017, 12:13 p.m. UTC | #3
On Tue, Nov 28, 2017 at 09:53:12AM +0000, David Laight wrote:
> From: Niklas Cassel
> > Sent: 27 November 2017 15:50
> > find_first_zero_bit()'s parameter 'size' is defined in bits,
> > not in bytes.
> > 
> > Calling find_first_zero_bit() with the wrong size unit
> > will lead to insidious bugs.
> > 
> > Fix this by calling find_first_zero_bit() with size
> > BITS_PER_LONG, rather than sizeof().
> > 
> > Also add proper error handling for find_first_zero_bit(),
> > since this was missing.
> > 
> > Fixes: d74679911610 ("PCI: endpoint: Introduce configfs entry for configuring EP functions")
> > Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
> > ---
> >  drivers/pci/endpoint/pci-ep-cfs.c | 8 +++++++-
> >  1 file changed, 7 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/pci/endpoint/pci-ep-cfs.c b/drivers/pci/endpoint/pci-ep-cfs.c
> > index e1f5adc9e113..0a22a7976580 100644
> > --- a/drivers/pci/endpoint/pci-ep-cfs.c
> > +++ b/drivers/pci/endpoint/pci-ep-cfs.c
> > @@ -109,7 +109,12 @@ static int pci_epc_epf_link(struct config_item *epc_item,
> >  		return ret;
> > 
> >  	func_no = find_first_zero_bit(&epc_group->function_num_map,
> > -				      sizeof(epc_group->function_num_map));
> > +				      BITS_PER_LONG);
> 
> Surely this should be either 8 * sizeof() or you should use ffz()
> that takes a numeric argument rather than the function designed
> for arbitrary size bitmaps.

Do you see a problem with this code ? It can be made a bitmap by
allocating it using epc->max_functions as bitmap number of bits
and that can be made on top of the code above that would go in
as a fix unless you strongly object to it, as Joe mentioned
both ffz+| and find_first_zero_bit()+set_bit() are ok with me as
long as we choose one and fix the issue.

Thanks,
Lorenzo
diff mbox

Patch

diff --git a/drivers/pci/endpoint/pci-ep-cfs.c b/drivers/pci/endpoint/pci-ep-cfs.c
index e1f5adc9e113..0a22a7976580 100644
--- a/drivers/pci/endpoint/pci-ep-cfs.c
+++ b/drivers/pci/endpoint/pci-ep-cfs.c
@@ -109,7 +109,12 @@  static int pci_epc_epf_link(struct config_item *epc_item,
 		return ret;
 
 	func_no = find_first_zero_bit(&epc_group->function_num_map,
-				      sizeof(epc_group->function_num_map));
+				      BITS_PER_LONG);
+	if (func_no >= BITS_PER_LONG) {
+		ret = -EINVAL;
+		dev_err(&epc->dev, "failed to link endpoint function device\n");
+		goto err_func_no;
+	}
 	set_bit(func_no, &epc_group->function_num_map);
 	epf->func_no = func_no;
 
@@ -121,6 +126,7 @@  static int pci_epc_epf_link(struct config_item *epc_item,
 
 err_epf_bind:
 	clear_bit(func_no, &epc_group->function_num_map);
+err_func_no:
 	pci_epc_remove_epf(epc, epf);
 
 	return ret;