diff mbox series

PCI: fix a potential uninitentional integer overflow issue

Message ID 20201007114615.19966-1-colin.king@canonical.com (mailing list archive)
State Superseded, archived
Delegated to: Bjorn Helgaas
Headers show
Series PCI: fix a potential uninitentional integer overflow issue | expand

Commit Message

Colin King Oct. 7, 2020, 11:46 a.m. UTC
From: Colin Ian King <colin.king@canonical.com>

The shift of 1 by align_order is evaluated using 32 bit arithmetic
and the result is assigned to a resource_size_t type variable that
is a 64 bit unsigned integer on 64 bit platforms. Fix an overflow
before widening issue by using the BIT_ULL macro to perform the
shift.

Addresses-Coverity: ("Uninitentional integer overflow")
Fixes: 07d8d7e57c28 ("PCI: Make specifying PCI devices in kernel parameters reusable")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
---
 drivers/pci/pci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Dan Carpenter Oct. 7, 2020, 12:33 p.m. UTC | #1
On Wed, Oct 07, 2020 at 12:46:15PM +0100, Colin King wrote:
> From: Colin Ian King <colin.king@canonical.com>
> 
> The shift of 1 by align_order is evaluated using 32 bit arithmetic
> and the result is assigned to a resource_size_t type variable that
> is a 64 bit unsigned integer on 64 bit platforms. Fix an overflow
> before widening issue by using the BIT_ULL macro to perform the
> shift.
> 
> Addresses-Coverity: ("Uninitentional integer overflow")
> Fixes: 07d8d7e57c28 ("PCI: Make specifying PCI devices in kernel parameters reusable")
> Signed-off-by: Colin Ian King <colin.king@canonical.com>
> ---
>  drivers/pci/pci.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 6d4d5a2f923d..1a5844d7af35 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -6209,7 +6209,7 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
>  			if (align_order == -1)
>  				align = PAGE_SIZE;
>  			else
> -				align = 1 << align_order;
> +				align = BIT_ULL(align_order);

"align_order" comes from sscanf() so Smatch thinks it's not trusted.
Anything above 63 is undefined behavior.  There should be a bounds check
on this but I don't know what the valid values of "align" are.

regards,
dan carpenter
Bjorn Helgaas Nov. 5, 2020, 10:24 p.m. UTC | #2
On Wed, Oct 07, 2020 at 03:33:45PM +0300, Dan Carpenter wrote:
> On Wed, Oct 07, 2020 at 12:46:15PM +0100, Colin King wrote:
> > From: Colin Ian King <colin.king@canonical.com>
> > 
> > The shift of 1 by align_order is evaluated using 32 bit arithmetic
> > and the result is assigned to a resource_size_t type variable that
> > is a 64 bit unsigned integer on 64 bit platforms. Fix an overflow
> > before widening issue by using the BIT_ULL macro to perform the
> > shift.
> > 
> > Addresses-Coverity: ("Uninitentional integer overflow")

s/Uninitentional/Unintentional/
Also in subject (please also capitalize subject)

Doesn't Coverity also assign an ID number for this specific issue?
Can you include that as well, e.g.,

  Addresses-Coverity-ID: 1226899 ("Unintentional integer overflow")

> > Fixes: 07d8d7e57c28 ("PCI: Make specifying PCI devices in kernel parameters reusable")
> > Signed-off-by: Colin Ian King <colin.king@canonical.com>
> > ---
> >  drivers/pci/pci.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > index 6d4d5a2f923d..1a5844d7af35 100644
> > --- a/drivers/pci/pci.c
> > +++ b/drivers/pci/pci.c
> > @@ -6209,7 +6209,7 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
> >  			if (align_order == -1)
> >  				align = PAGE_SIZE;
> >  			else
> > -				align = 1 << align_order;
> > +				align = BIT_ULL(align_order);
> 
> "align_order" comes from sscanf() so Smatch thinks it's not trusted.
> Anything above 63 is undefined behavior.  There should be a bounds check
> on this but I don't know what the valid values of "align" are.

The spec doesn't explicitly say what the size limit for 64-bit BARs
is, but it does say 32-bit BARs can support up to 2GB (2^31).  So I
infer that 2^63 would be the limit for 64-bit BARs.

What about something like the following?  To me, BIT_ULL doesn't seem
like an advantage over "1ULL << ", but maybe there's a reason to use
it.

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 8b9bea8ba751..6e17d0a6828a 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -6197,19 +6197,21 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
 	while (*p) {
 		count = 0;
 		if (sscanf(p, "%d%n", &align_order, &count) == 1 &&
-							p[count] == '@') {
+		    p[count] == '@') {
 			p += count + 1;
+			if (align_order > 63) {
+				pr_err("PCI: Invalid requested alignment (order %d)\n",
+				       align_order);
+				align_order = PAGE_SHIFT;
+			}
 		} else {
-			align_order = -1;
+			align_order = PAGE_SHIFT;
 		}
 
 		ret = pci_dev_str_match(dev, p, &p);
 		if (ret == 1) {
 			*resize = true;
-			if (align_order == -1)
-				align = PAGE_SIZE;
-			else
-				align = 1 << align_order;
+			align = 1ULL << align_order;
 			break;
 		} else if (ret < 0) {
 			pr_err("PCI: Can't parse resource_alignment parameter: %s\n",
Dan Carpenter Nov. 6, 2020, 8:04 a.m. UTC | #3
On Thu, Nov 05, 2020 at 04:24:30PM -0600, Bjorn Helgaas wrote:
> On Wed, Oct 07, 2020 at 03:33:45PM +0300, Dan Carpenter wrote:
> > On Wed, Oct 07, 2020 at 12:46:15PM +0100, Colin King wrote:
> > > From: Colin Ian King <colin.king@canonical.com>
> > > 
> > > The shift of 1 by align_order is evaluated using 32 bit arithmetic
> > > and the result is assigned to a resource_size_t type variable that
> > > is a 64 bit unsigned integer on 64 bit platforms. Fix an overflow
> > > before widening issue by using the BIT_ULL macro to perform the
> > > shift.
> > > 
> > > Addresses-Coverity: ("Uninitentional integer overflow")
> 
> s/Uninitentional/Unintentional/
> Also in subject (please also capitalize subject)
> 
> Doesn't Coverity also assign an ID number for this specific issue?
> Can you include that as well, e.g.,
> 
>   Addresses-Coverity-ID: 1226899 ("Unintentional integer overflow")
> 
> > > Fixes: 07d8d7e57c28 ("PCI: Make specifying PCI devices in kernel parameters reusable")
> > > Signed-off-by: Colin Ian King <colin.king@canonical.com>
> > > ---
> > >  drivers/pci/pci.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > > index 6d4d5a2f923d..1a5844d7af35 100644
> > > --- a/drivers/pci/pci.c
> > > +++ b/drivers/pci/pci.c
> > > @@ -6209,7 +6209,7 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
> > >  			if (align_order == -1)
> > >  				align = PAGE_SIZE;
> > >  			else
> > > -				align = 1 << align_order;
> > > +				align = BIT_ULL(align_order);
> > 
> > "align_order" comes from sscanf() so Smatch thinks it's not trusted.
> > Anything above 63 is undefined behavior.  There should be a bounds check
> > on this but I don't know what the valid values of "align" are.
> 
> The spec doesn't explicitly say what the size limit for 64-bit BARs
> is, but it does say 32-bit BARs can support up to 2GB (2^31).  So I
> infer that 2^63 would be the limit for 64-bit BARs.
> 
> What about something like the following?  To me, BIT_ULL doesn't seem
> like an advantage over "1ULL << ", but maybe there's a reason to use
> it.

The advantage of BIT_ULL() is that checkpatch and I think Coccinelle
will suggest using it.  It's only recently where a few people have
complained (actually you're probably the second person) that BIT() is
sort of a weird thing to use for size variables.

regards,
dan carpenter
Bjorn Helgaas Nov. 10, 2020, 8:54 p.m. UTC | #4
On Fri, Nov 06, 2020 at 11:04:19AM +0300, Dan Carpenter wrote:
> On Thu, Nov 05, 2020 at 04:24:30PM -0600, Bjorn Helgaas wrote:
> > On Wed, Oct 07, 2020 at 03:33:45PM +0300, Dan Carpenter wrote:
> > > On Wed, Oct 07, 2020 at 12:46:15PM +0100, Colin King wrote:
> > > > From: Colin Ian King <colin.king@canonical.com>
> > > > 
> > > > The shift of 1 by align_order is evaluated using 32 bit arithmetic
> > > > and the result is assigned to a resource_size_t type variable that
> > > > is a 64 bit unsigned integer on 64 bit platforms. Fix an overflow
> > > > before widening issue by using the BIT_ULL macro to perform the
> > > > shift.
> > > > 
> > > > Addresses-Coverity: ("Uninitentional integer overflow")
> > 
> > s/Uninitentional/Unintentional/
> > Also in subject (please also capitalize subject)
> > 
> > Doesn't Coverity also assign an ID number for this specific issue?
> > Can you include that as well, e.g.,
> > 
> >   Addresses-Coverity-ID: 1226899 ("Unintentional integer overflow")
> > 
> > > > Fixes: 07d8d7e57c28 ("PCI: Make specifying PCI devices in kernel parameters reusable")
> > > > Signed-off-by: Colin Ian King <colin.king@canonical.com>
> > > > ---
> > > >  drivers/pci/pci.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > > > index 6d4d5a2f923d..1a5844d7af35 100644
> > > > --- a/drivers/pci/pci.c
> > > > +++ b/drivers/pci/pci.c
> > > > @@ -6209,7 +6209,7 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
> > > >  			if (align_order == -1)
> > > >  				align = PAGE_SIZE;
> > > >  			else
> > > > -				align = 1 << align_order;
> > > > +				align = BIT_ULL(align_order);
> > > 
> > > "align_order" comes from sscanf() so Smatch thinks it's not trusted.
> > > Anything above 63 is undefined behavior.  There should be a bounds check
> > > on this but I don't know what the valid values of "align" are.
> > 
> > The spec doesn't explicitly say what the size limit for 64-bit BARs
> > is, but it does say 32-bit BARs can support up to 2GB (2^31).  So I
> > infer that 2^63 would be the limit for 64-bit BARs.
> > 
> > What about something like the following?  To me, BIT_ULL doesn't seem
> > like an advantage over "1ULL << ", but maybe there's a reason to use
> > it.
> 
> The advantage of BIT_ULL() is that checkpatch and I think Coccinelle
> will suggest using it.  It's only recently where a few people have
> complained (actually you're probably the second person) that BIT() is
> sort of a weird thing to use for size variables.

If that's the only reason, I definitely prefer "1ULL << align_order".

BIT_ULL is just a pointless abstraction in this case.
Colin King Nov. 10, 2020, 10 p.m. UTC | #5
On 10/11/2020 20:54, Bjorn Helgaas wrote:
> On Fri, Nov 06, 2020 at 11:04:19AM +0300, Dan Carpenter wrote:
>> On Thu, Nov 05, 2020 at 04:24:30PM -0600, Bjorn Helgaas wrote:
>>> On Wed, Oct 07, 2020 at 03:33:45PM +0300, Dan Carpenter wrote:
>>>> On Wed, Oct 07, 2020 at 12:46:15PM +0100, Colin King wrote:
>>>>> From: Colin Ian King <colin.king@canonical.com>
>>>>>
>>>>> The shift of 1 by align_order is evaluated using 32 bit arithmetic
>>>>> and the result is assigned to a resource_size_t type variable that
>>>>> is a 64 bit unsigned integer on 64 bit platforms. Fix an overflow
>>>>> before widening issue by using the BIT_ULL macro to perform the
>>>>> shift.
>>>>>
>>>>> Addresses-Coverity: ("Uninitentional integer overflow")
>>>
>>> s/Uninitentional/Unintentional/
>>> Also in subject (please also capitalize subject)

OK

>>>
>>> Doesn't Coverity also assign an ID number for this specific issue?
>>> Can you include that as well, e.g.,

I'm running this from an internal coverity scan, so the ID is not public.

>>>
>>>   Addresses-Coverity-ID: 1226899 ("Unintentional integer overflow")
>>>
>>>>> Fixes: 07d8d7e57c28 ("PCI: Make specifying PCI devices in kernel parameters reusable")
>>>>> Signed-off-by: Colin Ian King <colin.king@canonical.com>
>>>>> ---
>>>>>  drivers/pci/pci.c | 2 +-
>>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>>>>> index 6d4d5a2f923d..1a5844d7af35 100644
>>>>> --- a/drivers/pci/pci.c
>>>>> +++ b/drivers/pci/pci.c
>>>>> @@ -6209,7 +6209,7 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
>>>>>  			if (align_order == -1)
>>>>>  				align = PAGE_SIZE;
>>>>>  			else
>>>>> -				align = 1 << align_order;
>>>>> +				align = BIT_ULL(align_order);
>>>>
>>>> "align_order" comes from sscanf() so Smatch thinks it's not trusted.
>>>> Anything above 63 is undefined behavior.  There should be a bounds check
>>>> on this but I don't know what the valid values of "align" are.
>>>
>>> The spec doesn't explicitly say what the size limit for 64-bit BARs
>>> is, but it does say 32-bit BARs can support up to 2GB (2^31).  So I
>>> infer that 2^63 would be the limit for 64-bit BARs.
>>>
>>> What about something like the following?  To me, BIT_ULL doesn't seem
>>> like an advantage over "1ULL << ", but maybe there's a reason to use
>>> it.
>>
>> The advantage of BIT_ULL() is that checkpatch and I think Coccinelle
>> will suggest using it.  It's only recently where a few people have
>> complained (actually you're probably the second person) that BIT() is
>> sort of a weird thing to use for size variables.
> 
> If that's the only reason, I definitely prefer "1ULL << align_order".
> 
> BIT_ULL is just a pointless abstraction in this case.
> 
OK. V2 Arriving later today

Colin
diff mbox series

Patch

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 6d4d5a2f923d..1a5844d7af35 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -6209,7 +6209,7 @@  static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
 			if (align_order == -1)
 				align = PAGE_SIZE;
 			else
-				align = 1 << align_order;
+				align = BIT_ULL(align_order);
 			break;
 		} else if (ret < 0) {
 			pr_err("PCI: Can't parse resource_alignment parameter: %s\n",