Skip to contentSkip to conversation

NTB: EPF: Fix error code in epf_ntb_bind()

PATCH In review Dan Carpenter 3 participants 8 messages Aug 1, 2022 – Aug 12, 2022

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 NTB: EPF: Fix error code in epf_ntb_bind()

Return an error code if pci_register_driver() fails. Don't return success.

Fixes da51fd247424 ("NTB: EPF: support NTB transfer between PCI RC and EP connection")
---
 drivers/pci/endpoint/functions/pci-epf-vntb.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/pci/endpoint/functions/pci-epf-vntb.c +2 -1
index 111568089d45..cf338f3cf348 100644
@@ -1312,7 +1312,8 @@ static int epf_ntb_bind(struct pci_epf *epf)
 
 	epf_set_drvdata(epf, ntb);
 
-	if (pci_register_driver(&vntb_pci_driver)) {
+	ret = pci_register_driver(&vntb_pci_driver);
+	if (ret) {
 		dev_err(dev, "failure register vntb pci driver\n");
 		goto err_bar_alloc;
 	}
Signature · 1 line
2.35.1
#1 lore raw

PATCH 2/2 NTB: EPF: Tidy up some bounds checks

This sscanf() is reading from the filename which was set by the kernel so it should be trust worthy. Although the data is likely trust worthy there is some bounds checking but unfortunately, it is not complete or consistent. Additionally, the Smatch static checker marks everything that comes from sscanf() as tainted and so Smatch complains that this code can lead to an out of bounds issue. Let's clean things up and make Smatch happy.

The first problem is that there is no bounds checking in the _show() functions. The _store() and _show() functions are very similar so make the bounds checking the same in both.

The second issue is that if "win_no" is zero it leads to an array underflow so add an if (win_no <= 0) check for that.

---
 drivers/pci/endpoint/functions/pci-epf-vntb.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
drivers/pci/endpoint/functions/pci-epf-vntb.c +9 -2
index cf338f3cf348..a7fe86f43739 100644
@@ -831,9 +831,16 @@ static ssize_t epf_ntb_##_name##_show(struct config_item *item,		\
 {									\
 	struct config_group *group = to_config_group(item);		\
 	struct epf_ntb *ntb = to_epf_ntb(group);			\
+	struct device *dev = &ntb->epf->dev;				\
 	int win_no;							\
 									\
-	sscanf(#_name, "mw%d", &win_no);				\
+	if (sscanf(#_name, "mw%d", &win_no) != 1)			\
+		return -EINVAL;						\
+									\
+	if (win_no <= 0 || win_no > ntb->num_mws) {			\
+		dev_err(dev, "Invalid num_nws: %d value\n", ntb->num_mws); \
+		return -EINVAL;						\
+	}								\
 									\
 	return sprintf(page, "%lld\n", ntb->mws_size[win_no - 1]);	\
 }
@@ -856,7 +863,7 @@ static ssize_t epf_ntb_##_name##_store(struct config_item *item,	\
 	if (sscanf(#_name, "mw%d", &win_no) != 1)			\
 		return -EINVAL;						\
 									\
-	if (ntb->num_mws < win_no) {					\
+	if (win_no <= 0 || win_no > ntb->num_mws) {			\
 		dev_err(dev, "Invalid num_nws: %d value\n", ntb->num_mws); \
 		return -EINVAL;						\
 	}								\
Signature · 1 line
2.35.1
#2 lore raw

Re: [PATCH 2/2] NTB: EPF: Tidy up some bounds checks

Quoted 44 lines · Dan Carpenter

This sscanf() is reading from the filename which was set by the kernel so it should be trust worthy. Although the data is likely trust worthy there is some bounds checking but unfortunately, it is not complete or consistent. Additionally, the Smatch static checker marks everything that comes from sscanf() as tainted and so Smatch complains that this code can lead to an out of bounds issue. Let's clean things up and make Smatch happy.

The first problem is that there is no bounds checking in the _show() functions. The _store() and _show() functions are very similar so make the bounds checking the same in both.

The second issue is that if "win_no" is zero it leads to an array underflow so add an if (win_no <= 0) check for that.

---
 drivers/pci/endpoint/functions/pci-epf-vntb.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
drivers/pci/endpoint/functions/pci-epf-vntb.c +8 -1
index cf338f3cf348..a7fe86f43739 100644
@@ -831,9 +831,16 @@ static ssize_t epf_ntb_##_name##_show(struct config_item *item,            \
 {                                                                      \
        struct config_group *group = to_config_group(item);             \
        struct epf_ntb *ntb = to_epf_ntb(group);                        \
+       struct device *dev = &ntb->epf->dev;                            \
        int win_no;                                                     \
                                                                        \
-       sscanf(#_name, "mw%d", &win_no);                                \
+       if (sscanf(#_name, "mw%d", &win_no) != 1)                       \
+               return -EINVAL;                                         \
+                                                                       \
+       if (win_no <= 0 || win_no > ntb->num_mws) {                     \
+               dev_err(dev, "Invalid num_nws: %d value\n", ntb->num_mws); \
+               return -EINVAL;                                         \
+       }                                                               \
                                                                        \
        return sprintf(page, "%lld\n", ntb->mws_size[win_no - 1]);      \
 }

drivers/pci/endpoint/functions/pci-epf-vntb.c static ssize_t epf_ntb_##_name##_store(struct config_item *item, \ · line 866

Quoted 5 lines
drivers/pci/endpoint/functions/pci-epf-vntb.c +1 -1
@@ -856,7 +863,7 @@ static ssize_t epf_ntb_##_name##_store(struct config_item *item,    \
        if (sscanf(#_name, "mw%d", &win_no) != 1)                       \
                return -EINVAL;                                         \
                                                                        \
-       if (ntb->num_mws < win_no) {                                    \
+       if (win_no <= 0 || win_no > ntb->num_mws) {                     \

This might change the existing logic. will it be ? if (win_no <= 0 || win_no >= ntb->num_mws) {

Quoted 6 lines
drivers/pci/endpoint/functions/pci-epf-vntb.c +0 -1
                dev_err(dev, "Invalid num_nws: %d value\n", ntb->num_mws); \
                return -EINVAL;                                         \
        }                                                               \
--

2.35.1

#3 lore raw

Re: [PATCH 2/2] NTB: EPF: Tidy up some bounds checks

Quoted 45 lines · Souptick Joarder
Quoted 44 lines · Dan Carpenter

This sscanf() is reading from the filename which was set by the kernel so it should be trust worthy. Although the data is likely trust worthy there is some bounds checking but unfortunately, it is not complete or consistent. Additionally, the Smatch static checker marks everything that comes from sscanf() as tainted and so Smatch complains that this code can lead to an out of bounds issue. Let's clean things up and make Smatch happy.

The first problem is that there is no bounds checking in the _show() functions. The _store() and _show() functions are very similar so make the bounds checking the same in both.

The second issue is that if "win_no" is zero it leads to an array underflow so add an if (win_no <= 0) check for that.

---
 drivers/pci/endpoint/functions/pci-epf-vntb.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
drivers/pci/endpoint/functions/pci-epf-vntb.c +8 -1
index cf338f3cf348..a7fe86f43739 100644
@@ -831,9 +831,16 @@ static ssize_t epf_ntb_##_name##_show(struct config_item *item,            \
 {                                                                      \
        struct config_group *group = to_config_group(item);             \
        struct epf_ntb *ntb = to_epf_ntb(group);                        \
+       struct device *dev = &ntb->epf->dev;                            \
        int win_no;                                                     \
                                                                        \
-       sscanf(#_name, "mw%d", &win_no);                                \
+       if (sscanf(#_name, "mw%d", &win_no) != 1)                       \
+               return -EINVAL;                                         \
+                                                                       \
+       if (win_no <= 0 || win_no > ntb->num_mws) {                     \
+               dev_err(dev, "Invalid num_nws: %d value\n", ntb->num_mws); \
+               return -EINVAL;                                         \
+       }                                                               \
                                                                        \
        return sprintf(page, "%lld\n", ntb->mws_size[win_no - 1]);      \
 }
@@ -856,7 +863,7 @@ static ssize_t epf_ntb_##_name##_store(struct config_item *item,    \
Quoted 8 lines

drivers/pci/endpoint/functions/pci-epf-vntb.c static ssize_t epf_ntb_##_name##_store(struct config_item *item, \ · line 866

Quoted 5 lines
drivers/pci/endpoint/functions/pci-epf-vntb.c +1 -1
        if (sscanf(#_name, "mw%d", &win_no) != 1)                       \
                return -EINVAL;                                         \
                                                                        \
-       if (ntb->num_mws < win_no) {                                    \
+       if (win_no <= 0 || win_no > ntb->num_mws) {                     \

This might change the existing logic. will it be ? if (win_no <= 0 || win_no >= ntb->num_mws) {

No, it doesn't change the exiting logic with regards to the upper bounds. if (foo > bar) is the same as if (bar < foo). It just adds a lower bounds check.

Normally, the variable part of the condition is on the right. Check patch enforces that rule where it can. It's the same in English, the variable comes first. "If twelve foot is greater than the height of the tree then blah blah" vs "if the tree is less than twelve foot blah blah".

regards,
dan carpenter

#4 lore raw

Re: [PATCH 2/2] NTB: EPF: Tidy up some bounds checks

Quoted 3 lines · Dan Carpenter

It's the same in English, the variable comes first. "If twelve foot is greater than the height of the tree then blah blah" vs "if the tree is less than twelve foot blah blah".

Hehe... I had a brief momement of panic where I worried that these statements might not be equivalent. The backwards if is so confusing.

regards,
dan carpenter

#5 lore raw

Re: [PATCH 2/2] NTB: EPF: Tidy up some bounds checks

Quoted 8 lines · Dan Carpenter
Quoted 3 lines · Dan Carpenter

It's the same in English, the variable comes first. "If twelve foot is greater than the height of the tree then blah blah" vs "if the tree is less than twelve foot blah blah".

Hehe... I had a brief momement of panic where I worried that these statements might not be equivalent. The backwards if is so confusing.

hehe.. Sorry I read the condition wrong with an example value at first.

Quoted 3 lines

regards,
dan carpenter

Acked-by on v1 1/2
#6 lore raw
In reply to Dan Carpenter
Quoted 6 lines · Dan Carpenter

Return an error code if pci_register_driver() fails. Don't return success.

Fixes da51fd247424 ("NTB: EPF: support NTB transfer between PCI RC and EP connection")
Quoted 21 lines
---
 drivers/pci/endpoint/functions/pci-epf-vntb.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/pci/endpoint/functions/pci-epf-vntb.c +2 -2
index 111568089d45..cf338f3cf348 100644
@@ -1312,7 +1312,8 @@ static int epf_ntb_bind(struct pci_epf *epf)
 
        epf_set_drvdata(epf, ntb);
 
-       if (pci_register_driver(&vntb_pci_driver)) {
+       ret = pci_register_driver(&vntb_pci_driver);
+       if (ret) {
                dev_err(dev, "failure register vntb pci driver\n");
                goto err_bar_alloc;
        }
--

2.35.1

maintainer
#7 lore raw
In reply to Dan Carpenter
Quoted 26 lines · Dan Carpenter

Return an error code if pci_register_driver() fails. Don't return success.

Fixes da51fd247424 ("NTB: EPF: support NTB transfer between PCI RC and EP connection")
---
 drivers/pci/endpoint/functions/pci-epf-vntb.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/pci/endpoint/functions/pci-epf-vntb.c +2 -1
index 111568089d45..cf338f3cf348 100644
@@ -1312,7 +1312,8 @@ static int epf_ntb_bind(struct pci_epf *epf)
 
 	epf_set_drvdata(epf, ntb);
 
-	if (pci_register_driver(&vntb_pci_driver)) {
+	ret = pci_register_driver(&vntb_pci_driver);
+	if (ret) {
 		dev_err(dev, "failure register vntb pci driver\n");
 		goto err_bar_alloc;
 	}
Signature · 1 line
2.35.1

Sorry for the delay in response. This series is in my ntb branch and will be in my pull request for v5.20 which should be going out later today.

Thanks,
Jon