Skip to contentSkip to conversation

powernv: eeh: fix possible buffer overrun in ioda_eeh_phb_diag()

PATCH In review Brian W Hart 2 participants 4 messages Dec 19, 2013 – Dec 20, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 powernv: eeh: fix possible buffer overrun in ioda_eeh_phb_diag()

PHB diagnostic buffer may be smaller than PAGE_SIZE, especially when PAGE_SIZE > 4KB.

---
 arch/powerpc/platforms/powernv/eeh-ioda.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
arch/powerpc/platforms/powernv/eeh-ioda.c +3 -2
index 02245ce..8184ef5 100644
@@ -820,14 +820,15 @@ static void ioda_eeh_phb_diag(struct pci_controller *hose)
 	struct OpalIoPhbErrorCommon *common;
 	long rc;
 
-	common = (struct OpalIoPhbErrorCommon *)phb->diag.blob;
-	rc = opal_pci_get_phb_diag_data2(phb->opal_id, common, PAGE_SIZE);
+	rc = opal_pci_get_phb_diag_data2(phb->opal_id, phb->diag.blob,
+					 PNV_PCI_DIAG_BUF_SIZE);
 	if (rc != OPAL_SUCCESS) {
 		pr_warning("%s: Failed to get diag-data for PHB#%x (%ld)\n",
 			    __func__, hose->global_number, rc);
 		return;
 	}
 
+	common = (struct OpalIoPhbErrorCommon *)phb->diag.blob;
 	switch (common->ioType) {
 	case OPAL_PHB_ERROR_DATA_TYPE_P7IOC:
 		ioda_eeh_p7ioc_phb_diag(hose, common);
Signature · 1 line
1.8.3.1
Acked-by on v1 1/1
#1 lore raw
Quoted 3 lines · Brian W Hart

PHB diagnostic buffer may be smaller than PAGE_SIZE, especially when PAGE_SIZE > 4KB.

I think you're talking about that PAGE_SIZE could be configured to have variable size (e.g. 4KB). So it's not safe to pass PAGE_SIZE to OPAL API opal_pci_get_phb_diag_data2(). Instead, we should pass PNV_PCI_DIAG_BUF_SIZE and it makes sense to me :-)

Also, it needs to be backported to stable kernel as well.

Quoted 1 line
Quoted 16 lines

---
arch/powerpc/platforms/powernv/eeh-ioda.c | 5 +++--

1 file changed, 3 insertions(+), 2 deletions(-)
arch/powerpc/platforms/powernv/eeh-ioda.c +2 -2
index 02245ce..8184ef5 100644
@@ -820,14 +820,15 @@ static void ioda_eeh_phb_diag(struct pci_controller *hose)
 	struct OpalIoPhbErrorCommon *common;
 	long rc;
 
-	common = (struct OpalIoPhbErrorCommon *)phb->diag.blob;
-	rc = opal_pci_get_phb_diag_data2(phb->opal_id, common, PAGE_SIZE);
+	rc = opal_pci_get_phb_diag_data2(phb->opal_id, phb->diag.blob,
+					 PNV_PCI_DIAG_BUF_SIZE);

arch/powerpc/platforms/powernv/eeh-ioda.c static void ioda_eeh_phb_diag(struct pci_controller *hose) · line 834

Quoted 10 lines
arch/powerpc/platforms/powernv/eeh-ioda.c +1 -0
 	if (rc != OPAL_SUCCESS) {
 		pr_warning("%s: Failed to get diag-data for PHB#%x (%ld)\n",
 			    __func__, hose->global_number, rc);
 		return;
 	}
 
+	common = (struct OpalIoPhbErrorCommon *)phb->diag.blob;
 	switch (common->ioType) {
 	case OPAL_PHB_ERROR_DATA_TYPE_P7IOC:
 		ioda_eeh_p7ioc_phb_diag(hose, common);

Thanks,
Gavin

#2 lore raw
Quoted 10 lines · Gavin Shan
Quoted 3 lines · Brian W Hart

PHB diagnostic buffer may be smaller than PAGE_SIZE, especially when PAGE_SIZE > 4KB.

I think you're talking about that PAGE_SIZE could be configured to have variable size (e.g. 4KB). So it's not safe to pass PAGE_SIZE to OPAL API opal_pci_get_phb_diag_data2(). Instead, we should pass PNV_PCI_DIAG_BUF_SIZE and it makes sense to me :-)

Quoted 6 lines

Also, it needs to be backported to stable kernel as well.

Quoted 1 line

Sorry, Brian. It has been fixed as part of the following commit, which has been put into Ben's powerpc-next branch :-)

commit 93aef2a789778e7ec787179fc9b34ca4885a5ef3

    161  static void ioda_eeh_phb_diag(struct pci_controller *hose)
    162  {
    163         struct pnv_phb *phb = hose->private_data;
    164 -       struct OpalIoPhbErrorCommon *common;
    165         long rc;
    166  
    167 -       common = (struct OpalIoPhbErrorCommon *)phb->diag.blob;
    168 -       rc = opal_pci_get_phb_diag_data2(phb->opal_id, common, PAGE_SIZE);
    169 +       rc = opal_pci_get_phb_diag_data2(phb->opal_id, phb->diag.blob,
    170 +                                        PNV_PCI_DIAG_BUF_SIZE);
Quoted 22 lines
Quoted 22 lines

---
arch/powerpc/platforms/powernv/eeh-ioda.c | 5 +++--

1 file changed, 3 insertions(+), 2 deletions(-)
arch/powerpc/platforms/powernv/eeh-ioda.c +2 -2
index 02245ce..8184ef5 100644
@@ -820,14 +820,15 @@ static void ioda_eeh_phb_diag(struct pci_controller *hose)
 	struct OpalIoPhbErrorCommon *common;
 	long rc;
 
-	common = (struct OpalIoPhbErrorCommon *)phb->diag.blob;
-	rc = opal_pci_get_phb_diag_data2(phb->opal_id, common, PAGE_SIZE);
+	rc = opal_pci_get_phb_diag_data2(phb->opal_id, phb->diag.blob,
+					 PNV_PCI_DIAG_BUF_SIZE);
 	if (rc != OPAL_SUCCESS) {
 		pr_warning("%s: Failed to get diag-data for PHB#%x (%ld)\n",
 			    __func__, hose->global_number, rc);
 		return;
 	}
 
Quoted 5 lines

arch/powerpc/platforms/powernv/eeh-ioda.c static void ioda_eeh_phb_diag(struct pci_controller *hose) · line 834

Quoted 4 lines
arch/powerpc/platforms/powernv/eeh-ioda.c +1 -0
+	common = (struct OpalIoPhbErrorCommon *)phb->diag.blob;
 	switch (common->ioType) {
 	case OPAL_PHB_ERROR_DATA_TYPE_P7IOC:
 		ioda_eeh_p7ioc_phb_diag(hose, common);

Thanks,
Gavin

#3 lore raw
Quoted 10 lines · Gavin Shan
Quoted 9 lines · Gavin Shan
Quoted 3 lines · Brian W Hart

PHB diagnostic buffer may be smaller than PAGE_SIZE, especially when PAGE_SIZE > 4KB.

I think you're talking about that PAGE_SIZE could be configured to have variable size (e.g. 4KB). So it's not safe to pass PAGE_SIZE to OPAL API opal_pci_get_phb_diag_data2(). Instead, we should pass PNV_PCI_DIAG_BUF_SIZE and it makes sense to me :-)

Yeah, I noticed the problem because our test machine has PAGE_SIZE of 64K with the buffer only being 8K.

[...]

Quoted 2 lines

Sorry, Brian. It has been fixed as part of the following commit, which has been put into Ben's powerpc-next branch :-)

Thank you!