Skip to contentSkip to conversation

net: ne2000: fix bounds check in ioport operations

PATCH Superseded P J P 2 participants 5 messages Dec 31, 2015

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

P J P

author
#0 lore raw

PATCH v2 1/1 net: ne2000: fix bounds check in ioport operations

From: Prasad J Pandit <[email protected]>

While doing ioport r/w operations, ne2000 device emulation suffers from OOB r/w errors. Update respective array bounds check to avoid OOB access.

---
 hw/net/ne2000.c | 28 +++++++++++++++-------------
 1 file changed, 15 insertions(+), 13 deletions(-)

Updated as per review in
-> https://lists.gnu.org/archive/html/qemu-devel/2015-12/msg04856.html

hw/net/ne2000.c +15 -13
index 010f9ef..d1f764b 100644
@@ -447,8 +447,7 @@ static uint32_t ne2000_ioport_read(void *opaque, uint32_t addr)
 static inline void ne2000_mem_writeb(NE2000State *s, uint32_t addr,
                                      uint32_t val)
 {
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32 || (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
         s->mem[addr] = val;
     }
 }
@@ -457,9 +456,10 @@ static inline void ne2000_mem_writew(NE2000State *s, uint32_t addr,
                                      uint32_t val)
 {
     addr &= ~1; /* XXX: check exact behaviour if not even */
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
-        *(uint16_t *)(s->mem + addr) = cpu_to_le16(val);
+    if (addr < 32
+        || (addr >= NE2000_PMEM_START
+            && addr + sizeof(uint16_t) < NE2000_MEM_SIZE)) {
+        *(uint16_t *)(s->mem + addr) = cpu_to_le16(data);
     }
 }
 
@@ -467,16 +467,16 @@ static inline void ne2000_mem_writel(NE2000State *s, uint32_t addr,
                                      uint32_t val)
 {
     addr &= ~1; /* XXX: check exact behaviour if not even */
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32
+        || (addr >= NE2000_PMEM_START
+            && addr + sizeof(uint32_t) < NE2000_MEM_SIZE)) {
         stl_le_p(s->mem + addr, val);
     }
 }
 
 static inline uint32_t ne2000_mem_readb(NE2000State *s, uint32_t addr)
 {
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32 || (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
         return s->mem[addr];
     } else {
         return 0xff;
@@ -486,8 +486,9 @@ static inline uint32_t ne2000_mem_readb(NE2000State *s, uint32_t addr)
 static inline uint32_t ne2000_mem_readw(NE2000State *s, uint32_t addr)
 {
     addr &= ~1; /* XXX: check exact behaviour if not even */
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32
+        || (addr >= NE2000_PMEM_START
+            && addr + sizeof(uint16_t) < NE2000_MEM_SIZE)) {
         return le16_to_cpu(*(uint16_t *)(s->mem + addr));
     } else {
         return 0xffff;
@@ -497,8 +498,9 @@ static inline uint32_t ne2000_mem_readw(NE2000State *s, uint32_t addr)
 static inline uint32_t ne2000_mem_readl(NE2000State *s, uint32_t addr)
 {
     addr &= ~1; /* XXX: check exact behaviour if not even */
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32
+        || (addr >= NE2000_PMEM_START
+            && addr + sizeof(uint32_t) < NE2000_MEM_SIZE)) {
         return ldl_le_p(s->mem + addr);
     } else {
         return 0xffffffff;
Signature · 1 line
2.4.3
#1 lore raw
Quoted 20 lines · P J P

From: Prasad J Pandit <[email protected]>

While doing ioport r/w operations, ne2000 device emulation suffers from OOB r/w errors. Update respective array bounds check to avoid OOB access.

---
 hw/net/ne2000.c | 28 +++++++++++++++-------------
 1 file changed, 15 insertions(+), 13 deletions(-)

Updated as per review in
-> https://lists.gnu.org/archive/html/qemu-devel/2015-12/msg04856.html

hw/net/ne2000.c +0 -0
index 010f9ef..d1f764b 100644

hw/net/ne2000.c static uint32_t ne2000_ioport_read(void *opaque, uint32_t addr) · line 450

Quoted 6 lines
hw/net/ne2000.c +1 -2
@@ -447,8 +447,7 @@ static uint32_t ne2000_ioport_read(void *opaque, uint32_t addr)
 static inline void ne2000_mem_writeb(NE2000State *s, uint32_t addr,
                                      uint32_t val)
 {
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32 || (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {

The change is unnecessary.

hw/net/ne2000.c static inline void ne2000_mem_writew(NE2000State *s, uint32_t addr, · line 461

Quoted 13 lines
hw/net/ne2000.c +3 -3
         s->mem[addr] = val;
     }
 }
@@ -457,9 +456,10 @@ static inline void ne2000_mem_writew(NE2000State *s, uint32_t addr,
                                      uint32_t val)
 {
     addr &= ~1; /* XXX: check exact behaviour if not even */
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
-        *(uint16_t *)(s->mem + addr) = cpu_to_le16(val);
+    if (addr < 32
+        || (addr >= NE2000_PMEM_START
+            && addr + sizeof(uint16_t) < NE2000_MEM_SIZE)) {

I think you mean '<=' instead of '<' here? (And for the other checks below).

Quoted 49 lines
hw/net/ne2000.c +11 -8
+        *(uint16_t *)(s->mem + addr) = cpu_to_le16(data);
     }
 }
 
@@ -467,16 +467,16 @@ static inline void ne2000_mem_writel(NE2000State *s, uint32_t addr,
                                      uint32_t val)
 {
     addr &= ~1; /* XXX: check exact behaviour if not even */
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32
+        || (addr >= NE2000_PMEM_START
+            && addr + sizeof(uint32_t) < NE2000_MEM_SIZE)) {
         stl_le_p(s->mem + addr, val);
     }
 }
 
 static inline uint32_t ne2000_mem_readb(NE2000State *s, uint32_t addr)
 {
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32 || (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
         return s->mem[addr];
     } else {
         return 0xff;
@@ -486,8 +486,9 @@ static inline uint32_t ne2000_mem_readb(NE2000State *s, uint32_t addr)
 static inline uint32_t ne2000_mem_readw(NE2000State *s, uint32_t addr)
 {
     addr &= ~1; /* XXX: check exact behaviour if not even */
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32
+        || (addr >= NE2000_PMEM_START
+            && addr + sizeof(uint16_t) < NE2000_MEM_SIZE)) {
         return le16_to_cpu(*(uint16_t *)(s->mem + addr));
     } else {
         return 0xffff;
@@ -497,8 +498,9 @@ static inline uint32_t ne2000_mem_readw(NE2000State *s, uint32_t addr)
 static inline uint32_t ne2000_mem_readl(NE2000State *s, uint32_t addr)
 {
     addr &= ~1; /* XXX: check exact behaviour if not even */
-    if (addr < 32 ||
-        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
+    if (addr < 32
+        || (addr >= NE2000_PMEM_START
+            && addr + sizeof(uint32_t) < NE2000_MEM_SIZE)) {
         return ldl_le_p(s->mem + addr);
     } else {
         return 0xffffffff;
author
#2 lore raw
+-- On Thu, 31 Dec 2015, Jason Wang wrote --+
| > -        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
| > +    if (addr < 32 || (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
| 
| The change is unnecessary.

Okay.

| > +    if (addr < 32
| > +        || (addr >= NE2000_PMEM_START
| > +            && addr + sizeof(uint16_t) < NE2000_MEM_SIZE)) {
| 
| I think you mean '<=' instead of '<' here? (And for the other checks below).

I think <= would lead to an off-by-one, no? As the last array index would be
one less than the size; Same as ne2000_mem_readb() above.

Thank you.
--
Prasad J Pandit / Red Hat Product Security Team
47AF CE69 3A90 54AA 9045 1053 DD13 3D32 FE5B 041F

#3 lore raw
Quoted 8 lines · P J P
+-- On Thu, 31 Dec 2015, Jason Wang wrote --+
| > -        (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
| > +    if (addr < 32 || (addr >= NE2000_PMEM_START && addr < NE2000_MEM_SIZE)) {
| 
| The change is unnecessary.

Okay.

Quoted 7 lines
| > +    if (addr < 32
| > +        || (addr >= NE2000_PMEM_START
| > +            && addr + sizeof(uint16_t) < NE2000_MEM_SIZE)) {
| 
| I think you mean '<=' instead of '<' here? (And for the other checks below).

I think <= would lead to an off-by-one, no?

The real byte we could touch is in fact addr + sizeof(uint16_t) -1 here.

Consider we should allow double bytes access at NE2000_MEM_SIZE - 2, but this patch forbids this.

Btw, looking at ne2000_mem_writew(), it has:

addr &= ~1;

at the beginning, so looks like we are really safe, Need only to care about writel?

Quoted 7 lines

As the last array index would be
one less than the size; Same as ne2000_mem_readb() above.

Thank you.
--
Prasad J Pandit / Red Hat Product Security Team
47AF CE69 3A90 54AA 9045 1053 DD13 3D32 FE5B 041F

author
#4 lore raw
+-- On Thu, 31 Dec 2015, Jason Wang wrote --+
| Btw, looking at ne2000_mem_writew(), it has:
| addr &= ~1;

Yes, this seems to ensure that write starts at an even address.

| at the beginning, so looks like we are really safe, Need only to care | about writel?

Right, I've sent an updated patch for the same.

Thank you.
--
Prasad J Pandit / Red Hat Product Security Team
47AF CE69 3A90 54AA 9045 1053 DD13 3D32 FE5B 041F