Skip to contentSkip to conversation

scsi: lsi: exit infinite loop while executing script (CVE-2019-12068)

PATCH Superseded P J P 4 participants 9 messages Aug 8, 2019

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

P J P

author
#0 lore raw

PATCH 1/1 scsi: lsi: exit infinite loop while executing script (CVE-2019-12068)

From: Prasad J Pandit <[email protected]>

When executing script in lsi_execute_script(), the LSI scsi adapter emulator advances 's->dsp' index to read next opcode. This can lead to an infinite loop if the next opcode is empty. Exit such loop after reading 10k empty opcodes.

---
 hw/scsi/lsi53c895a.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
hw/scsi/lsi53c895a.c +5 -1
index 10468c1ec1..c23a40525e 100644
@@ -1132,7 +1132,10 @@ static void lsi_execute_script(LSIState *s)
 
     s->istat1 |= LSI_ISTAT1_SRUN;
 again:
-    insn_processed++;
+    if (++insn_processed > 10000) {
+        s->waiting = LSI_NOWAIT;
+        goto exitloop;
+    }
     insn = read_dword(s, s->dsp);
     if (!insn) {
         /* If we receive an empty opcode increment the DSP by 4 bytes
@@ -1569,6 +1572,7 @@ again:
             }
         }
     }
+exitloop:
     if (insn_processed > 10000 && s->waiting == LSI_NOWAIT) {
         /* Some windows drivers make the device spin waiting for a memory
            location to change.  If we have been executed a lot of code then
Signature · 1 line
2.21.0
#1 lore raw
Quoted 18 lines · P J P

From: Prasad J Pandit <[email protected]>

When executing script in lsi_execute_script(), the LSI scsi adapter emulator advances 's->dsp' index to read next opcode. This can lead to an infinite loop if the next opcode is empty. Exit such loop after reading 10k empty opcodes.

---
 hw/scsi/lsi53c895a.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
hw/scsi/lsi53c895a.c +0 -0
index 10468c1ec1..c23a40525e 100644

hw/scsi/lsi53c895a.c static void lsi_execute_script(LSIState *s) · line 1135

Quoted 5 lines
hw/scsi/lsi53c895a.c +1 -1
@@ -1132,7 +1132,10 @@ static void lsi_execute_script(LSIState *s)
 
     s->istat1 |= LSI_ISTAT1_SRUN;
 again:
-    insn_processed++;
+    if (++insn_processed > 10000) {
                              ^

Since we are using this "magic" number in several lines, should we define a macro?

Thanks,
Stefano

Quoted 18 lines
hw/scsi/lsi53c895a.c +4 -0
+        s->waiting = LSI_NOWAIT;
+        goto exitloop;
+    }
     insn = read_dword(s, s->dsp);
     if (!insn) {
         /* If we receive an empty opcode increment the DSP by 4 bytes
@@ -1569,6 +1572,7 @@ again:
             }
         }
     }
+exitloop:
     if (insn_processed > 10000 && s->waiting == LSI_NOWAIT) {
         /* Some windows drivers make the device spin waiting for a memory
            location to change.  If we have been executed a lot of code then
Signature · 1 line
2.21.0
author
#2 lore raw
+-- On Thu, 8 Aug 2019, Stefano Garzarella wrote --+
| > +    if (++insn_processed > 10000) {
|                               ^
| Since we are using this "magic" number in several lines,
| should we define a macro?

Sent patch v2. Thank you.
--
Prasad J Pandit / Red Hat Product Security Team
47AF CE69 3A90 54AA 9045 1053 DD13 3D32 FE5B 041F

#3 lore raw
In reply to P J P
Quoted 31 lines · P J P

From: Prasad J Pandit <[email protected]>

When executing script in lsi_execute_script(), the LSI scsi adapter emulator advances 's->dsp' index to read next opcode. This can lead to an infinite loop if the next opcode is empty. Exit such loop after reading 10k empty opcodes.

---
 hw/scsi/lsi53c895a.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
hw/scsi/lsi53c895a.c +4 -1
index 10468c1ec1..c23a40525e 100644
@@ -1132,7 +1132,10 @@ static void lsi_execute_script(LSIState *s)
 
     s->istat1 |= LSI_ISTAT1_SRUN;
 again:
-    insn_processed++;
+    if (++insn_processed > 10000) {
+        s->waiting = LSI_NOWAIT;
+        goto exitloop;
+    }
     insn = read_dword(s, s->dsp);
     if (!insn) {
         /* If we receive an empty opcode increment the DSP by 4 bytes

hw/scsi/lsi53c895a.c again: · line 1578

Quoted 7 lines
hw/scsi/lsi53c895a.c +1 -0
@@ -1569,6 +1572,7 @@ again:
             }
         }
     }
+exitloop:
     if (insn_processed > 10000 && s->waiting == LSI_NOWAIT) {
         /* Some windows drivers make the device spin waiting for a memory
            location to change.  If we have been executed a lot of code then

I am not sure this is worth a CVE. The kernel can cause QEMU to break, but is there a practical case in which an unprivileged user can do that?

Paolo

author
#4 lore raw
+-- On Thu, 8 Aug 2019, Paolo Bonzini wrote --+
| I am not sure this is worth a CVE. 

True, it is a low one, as QEMU consumes cycles on the host.

| The kernel can cause QEMU to break, but is there a practical case in which | an unprivileged user can do that?

QEMU does not break, it keeps running in interruptible sleep 'S' state. They've a reproducer wherein guest does mmio calls to trigger the issue.

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

#5 lore raw
Quoted 10 lines · P J P
+-- On Thu, 8 Aug 2019, Paolo Bonzini wrote --+
| I am not sure this is worth a CVE. 

True, it is a low one, as QEMU consumes cycles on the host.

| The kernel can cause QEMU to break, but is there a practical case in which | an unprivileged user can do that?

QEMU does not break, it keeps running in interruptible sleep 'S' state. They've a reproducer wherein guest does mmio calls to trigger the issue.

From user-mode? As unprivileged user?

author
#6 lore raw
+-- On Thu, 8 Aug 2019, Philippe Mathieu-Daudé wrote --+
| >From user-mode? As unprivileged user?

No, needs privileges inside guest.

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

#7 lore raw
In reply to P J P
Quoted 4 lines · P J P
+-- On Thu, 8 Aug 2019, Paolo Bonzini wrote --+
| I am not sure this is worth a CVE. 

True, it is a low one, as QEMU consumes cycles on the host.

A guest that runs an infinite loop would be an easier way to do that. I suppose this one also blocks the monitor, but then "kill -9" is always your friend. :)

Paolo

Quoted 5 lines

| The kernel can cause QEMU to break, but is there a practical case in which | an unprivileged user can do that?

QEMU does not break, it keeps running in interruptible sleep 'S' state. They've a reproducer wherein guest does mmio calls to trigger the issue.

author
#8 lore raw
+-- On Thu, 8 Aug 2019, Paolo Bonzini wrote --+
| I suppose this one also blocks the monitor, but then "kill -9" is always 
| your friend. :)

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