Skip to contentSkip to conversation

assertion failure in net_tx_pkt_add_raw_fragment() in hw/net/net_tx_pkt.c

PATCH Superseded Mauro Matteo Cascella 4 participants 11 messages Jul 27, 2020 – Jul 31, 2020

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover assertion failure in net_tx_pkt_add_raw_fragment() in hw/net/net_tx_pkt.c

An assertion failure issue was reported by Mr. Ziming Zhang (CC'd). It occurs in the code that processes network packets while adding data fragments into packet context. This flaw could potentially be abused by a malicious guest to abort the QEMU process on the host. This two patch series does a couple of things:

- introduces a new function in net_tx_pkt.{c,h} to check the maximum number of data fragments
- adds a check in both e1000e and vmxnet3 devices to skip the packet if the current data fragment exceeds max_raw_frags, preventing net_tx_pkt_add_raw_fragment() to be called with an invalid raw_frags

Mauro Matteo Cascella (2):
hw/net/net_tx_pkt: add function to check pkt->max_raw_frags hw/net: check max_raw_frags in e1000e and vmxnet3 devices

 hw/net/e1000e_core.c | 3 ++-
 hw/net/net_tx_pkt.c  | 5 +++++
 hw/net/net_tx_pkt.h  | 8 ++++++++
 hw/net/vmxnet3.c     | 3 ++-
 4 files changed, 17 insertions(+), 2 deletions(-)
Signature · 1 line
2.26.2
#1 lore raw

PATCH 1/2 hw/net/net_tx_pkt: add function to check pkt->max_raw_frags

This patch introduces a new function in hw/net/net_tx_pkt.{c,h} to check the current data fragment against the maximum number of data fragments.

---
 hw/net/net_tx_pkt.c | 5 +++++
 hw/net/net_tx_pkt.h | 8 ++++++++
 2 files changed, 13 insertions(+)
hw/net/net_tx_pkt.c +5 -0
index 9560e4a49e..d035618f2c 100644
@@ -400,6 +400,11 @@ bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa,
     }
 }
 
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt)
+{
+    return pkt->raw_frags >= pkt->max_raw_frags;
+}
+
 bool net_tx_pkt_has_fragments(struct NetTxPkt *pkt)
 {
     return pkt->raw_frags > 0;
hw/net/net_tx_pkt.h +8 -0
index 4ec8bbe9bd..e2ee46ae03 100644
@@ -179,6 +179,14 @@ bool net_tx_pkt_send_loopback(struct NetTxPkt *pkt, NetClientState *nc);
  */
 bool net_tx_pkt_parse(struct NetTxPkt *pkt);
 
+/**
+* indicates if the current data fragment exceeds max_raw_frags
+*
+* @pkt:            packet
+*
+*/
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt);
+
 /**
 * indicates if there are data fragments held by this packet object.
 *
Signature · 1 line
2.26.2
#2 lore raw

Re: [PATCH 1/2] hw/net/net_tx_pkt: add function to check pkt->max_raw_frags

Quoted 2 lines · Mauro Matteo Cascella

This patch introduces a new function in hw/net/net_tx_pkt.{c,h} to check the current data fragment against the maximum number of data fragments.

I wonder whether it's better to do the check in net_tx_pkt_add_raw_fragment() and fail there.

Btw, I find net_tx_pkt_add_raw_fragment() does not unmap dma when returning to true, is this a bug?

Thanks

Quoted 43 lines
---
  hw/net/net_tx_pkt.c | 5 +++++
  hw/net/net_tx_pkt.h | 8 ++++++++
  2 files changed, 13 insertions(+)
hw/net/net_tx_pkt.c +5 -0
index 9560e4a49e..d035618f2c 100644
@@ -400,6 +400,11 @@ bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa,
      }
  }
 
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt)
+{
+    return pkt->raw_frags >= pkt->max_raw_frags;
+}
+
  bool net_tx_pkt_has_fragments(struct NetTxPkt *pkt)
  {
      return pkt->raw_frags > 0;
hw/net/net_tx_pkt.h +8 -0
index 4ec8bbe9bd..e2ee46ae03 100644
@@ -179,6 +179,14 @@ bool net_tx_pkt_send_loopback(struct NetTxPkt *pkt, NetClientState *nc);
   */
  bool net_tx_pkt_parse(struct NetTxPkt *pkt);
 
+/**
+* indicates if the current data fragment exceeds max_raw_frags
+*
+* @pkt:            packet
+*
+*/
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt);
+
  /**
  * indicates if there are data fragments held by this packet object.
  *
#3 lore raw

Re: [PATCH 1/2] hw/net/net_tx_pkt: add function to check pkt->max_raw_frags

Quoted 9 lines · Jason Wang
Quoted 2 lines · Mauro Matteo Cascella

This patch introduces a new function in hw/net/net_tx_pkt.{c,h} to check the current data fragment against the maximum number of data fragments.

I wonder whether it's better to do the check in
net_tx_pkt_add_raw_fragment() and fail there.

Given the assertion, I assumed the caller is responsible for the check, but moving the check in net_tx_pkt_add_raw_fragment() totally makes sense to me.

Quoted 2 lines

Btw, I find net_tx_pkt_add_raw_fragment() does not unmap dma when returning to true, is this a bug?

Isn't it unmapped in net_tx_pkt_reset()?

Quoted 47 lines

Thanks

Quoted 43 lines
---
  hw/net/net_tx_pkt.c | 5 +++++
  hw/net/net_tx_pkt.h | 8 ++++++++
  2 files changed, 13 insertions(+)
hw/net/net_tx_pkt.c +5 -0
index 9560e4a49e..d035618f2c 100644
@@ -400,6 +400,11 @@ bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa,
      }
  }
 
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt)
+{
+    return pkt->raw_frags >= pkt->max_raw_frags;
+}
+
  bool net_tx_pkt_has_fragments(struct NetTxPkt *pkt)
  {
      return pkt->raw_frags > 0;
hw/net/net_tx_pkt.h +8 -0
index 4ec8bbe9bd..e2ee46ae03 100644
@@ -179,6 +179,14 @@ bool net_tx_pkt_send_loopback(struct NetTxPkt *pkt, NetClientState *nc);
   */
  bool net_tx_pkt_parse(struct NetTxPkt *pkt);
 
+/**
+* indicates if the current data fragment exceeds max_raw_frags
+*
+* @pkt:            packet
+*
+*/
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt);
+
  /**
  * indicates if there are data fragments held by this packet object.
  *
Signature · 2 lines
Mauro Matteo Cascella, Red Hat Product Security
6F78 E20B 5935 928C F0A8  1A9D 4E55 23B8 BB34 10B0
#4 lore raw

Re: [PATCH 1/2] hw/net/net_tx_pkt: add function to check pkt->max_raw_frags

Quoted 11 lines · Mauro Matteo Cascella
Quoted 7 lines · Jason Wang
Quoted 2 lines · Mauro Matteo Cascella

This patch introduces a new function in hw/net/net_tx_pkt.{c,h} to check the current data fragment against the maximum number of data fragments.

I wonder whether it's better to do the check in
net_tx_pkt_add_raw_fragment() and fail there.

Given the assertion, I assumed the caller is responsible for the check, but moving the check in net_tx_pkt_add_raw_fragment() totally makes sense to me.

Want to send a new version for this?

Quoted 4 lines
Quoted 3 lines

Btw, I find net_tx_pkt_add_raw_fragment() does not unmap dma when returning to true, is this a bug?

Isn't it unmapped in net_tx_pkt_reset()?

Probably but see how it was used in e1000e, the net_tx_pkt_reset() is only called when eop is set. Is this a bug?

Thanks

Quoted 46 lines
Quoted 46 lines

Thanks

Quoted 42 lines
---
   hw/net/net_tx_pkt.c | 5 +++++
   hw/net/net_tx_pkt.h | 8 ++++++++
   2 files changed, 13 insertions(+)
hw/net/net_tx_pkt.c +5 -0
index 9560e4a49e..d035618f2c 100644
@@ -400,6 +400,11 @@ bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa,
       }
   }
 
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt)
+{
+    return pkt->raw_frags >= pkt->max_raw_frags;
+}
+
   bool net_tx_pkt_has_fragments(struct NetTxPkt *pkt)
   {
       return pkt->raw_frags > 0;
hw/net/net_tx_pkt.h +8 -0
index 4ec8bbe9bd..e2ee46ae03 100644
@@ -179,6 +179,14 @@ bool net_tx_pkt_send_loopback(struct NetTxPkt *pkt, NetClientState *nc);
    */
   bool net_tx_pkt_parse(struct NetTxPkt *pkt);
 
+/**
+* indicates if the current data fragment exceeds max_raw_frags
+*
+* @pkt:            packet
+*
+*/
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt);
+
   /**
   * indicates if there are data fragments held by this packet object.
   *
#5 lore raw

Re: [PATCH 1/2] hw/net/net_tx_pkt: add function to check pkt->max_raw_frags

Quoted 9 lines · Jason Wang
Quoted 6 lines · Mauro Matteo Cascella
Quoted 5 lines · Jason Wang
Quoted 2 lines · Mauro Matteo Cascella

This patch introduces a new function in hw/net/net_tx_pkt.{c,h} to check the current data fragment against the maximum number of data fragments.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

I wonder whether it's better to do the check in
net_tx_pkt_add_raw_fragment() and fail there.

Given the assertion, I assumed the caller is responsible for the check, but moving the check in net_tx_pkt_add_raw_fragment() totally makes sense to me.

Want to send a new version for this?

Sure, I'll send a new version. Thank you.

Quoted 9 lines
Quoted 5 lines
Quoted 4 lines

Btw, I find net_tx_pkt_add_raw_fragment() does not unmap dma when returning to true, is this a bug?

Isn't it unmapped in net_tx_pkt_reset()?

Probably but see how it was used in e1000e, the net_tx_pkt_reset() is only called when eop is set. Is this a bug?

Yeah it all depends on E1000_TXD_CMD_EOP. Besides, if not set, e1000e_tx_pkt_send() would never be called. Honestly, I don't know if this is a reasonable scenario or not.

Quoted 37 lines

Thanks

Quoted 35 lines
Quoted 35 lines

Thanks

Quoted 31 lines
---
   hw/net/net_tx_pkt.c | 5 +++++
   hw/net/net_tx_pkt.h | 8 ++++++++
   2 files changed, 13 insertions(+)
hw/net/net_tx_pkt.c +5 -0
index 9560e4a49e..d035618f2c 100644
@@ -400,6 +400,11 @@ bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa,
       }
   }
 
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt)
+{
+    return pkt->raw_frags >= pkt->max_raw_frags;
+}
+
   bool net_tx_pkt_has_fragments(struct NetTxPkt *pkt)
   {
       return pkt->raw_frags > 0;
hw/net/net_tx_pkt.h +0 -0
index 4ec8bbe9bd..e2ee46ae03 100644
@@ -179,6 +179,14 @@ bool net_tx_pkt_send_loopback(struct NetTxPkt *pkt, NetClientState *nc);
    */
   bool net_tx_pkt_parse(struct NetTxPkt *pkt);
 
Quoted 12 lines
Quoted 11 lines
Quoted 11 lines

hw/net/net_tx_pkt.h bool net_tx_pkt_send_loopback(struct NetTxPkt *pkt, NetClientState *nc); · line 192

Quoted 11 lines
hw/net/net_tx_pkt.h +8 -0
+/**
+* indicates if the current data fragment exceeds max_raw_frags
+*
+* @pkt:            packet
+*
+*/
+bool net_tx_pkt_exceed_max_fragments(struct NetTxPkt *pkt);
+
   /**
   * indicates if there are data fragments held by this packet object.
   *

--
Mauro Matteo Cascella, Red Hat Product Security 6F78 E20B 5935 928C F0A8 1A9D 4E55 23B8 BB34 10B0

#6 lore raw

Re: [PATCH 1/2] hw/net/net_tx_pkt: add function to check pkt->max_raw_frags

Quoted 14 lines · Mauro Matteo Cascella
Quoted 11 lines · Jason Wang
Quoted 9 lines · Mauro Matteo Cascella
Quoted 5 lines · Jason Wang
Quoted 2 lines · Mauro Matteo Cascella

This patch introduces a new function in hw/net/net_tx_pkt.{c,h} to check the current data fragment against the maximum number of data fragments.

I wonder whether it's better to do the check in
net_tx_pkt_add_raw_fragment() and fail there.

Given the assertion, I assumed the caller is responsible for the check, but moving the check in net_tx_pkt_add_raw_fragment() totally makes sense to me.

Want to send a new version for this?

Sure, I'll send a new version. Thank you.

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Quoted 2 lines

Btw, I find net_tx_pkt_add_raw_fragment() does not unmap dma when returning to true, is this a bug?

Isn't it unmapped in net_tx_pkt_reset()?

Probably but see how it was used in e1000e, the net_tx_pkt_reset() is only called when eop is set. Is this a bug?

Yeah it all depends on E1000_TXD_CMD_EOP. Besides, if not set, e1000e_tx_pkt_send() would never be called. Honestly, I don't know if this is a reasonable scenario or not.

It's probably fine since anyway e1000e_core_reset() will call net_tx_pkt_reset().

Thanks

Quoted 3 lines
Quoted 3 lines

Thanks

#7 lore raw

PATCH 2/2 hw/net: check max_raw_frags in e1000e and vmxnet3 devices

In reply to Mauro Matteo Cascella

This patch adds a check in both e1000e and vmxnet3 devices to skip the packet if the current data fragment exceeds max_raw_frags, preventing net_tx_pkt_add_raw_fragment() to be called with an invalid raw_frags.

---
 hw/net/e1000e_core.c | 3 ++-
 hw/net/vmxnet3.c     | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)
hw/net/e1000e_core.c +2 -1
index bcd186cac5..c573a30d63 100644
@@ -728,7 +728,8 @@ e1000e_process_tx_desc(E1000ECore *core,
     addr = le64_to_cpu(dp->buffer_addr);
 
     if (!tx->skip_cp) {
-        if (!net_tx_pkt_add_raw_fragment(tx->tx_pkt, addr, split_size)) {
+        if (net_tx_pkt_exceed_max_fragments(tx->tx_pkt) ||
+            !net_tx_pkt_add_raw_fragment(tx->tx_pkt, addr, split_size)) {
             tx->skip_cp = true;
         }
     }
hw/net/vmxnet3.c +2 -1
index 7a6ca4ec35..f482806037 100644
@@ -650,7 +650,8 @@ static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx)
             data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE;
             data_pa = txd.addr;
 
-            if (!net_tx_pkt_add_raw_fragment(s->tx_pkt,
+            if (net_tx_pkt_exceed_max_fragments(s->tx_pkt) ||
+                !net_tx_pkt_add_raw_fragment(s->tx_pkt,
                                                 data_pa,
                                                 data_len)) {
                 s->skip_current_tx_pkt = true;
Signature · 1 line
2.26.2
#8 lore raw
In reply to Mauro Matteo Cascella

I sent a reproducer for the to the list some time ago, but never created a Launchpad bug...
https://www.mail-archive.com/[email protected]/msg701930.html

Anyways.. I can confirm that I can't reproduce the issue with these patches.

Minimized Reproducer:
cat << EOF | ./i386-softmmu/qemu-system-i386 -M pc-q35-5.0 -nographic \
-display none -serial none -monitor none -qtest stdio
outl 0xcf8 0x80001010
outl 0xcfc 0xe1020000
outl 0xcf8 0x80001004
outw 0xcfc 0x7
write 0xe10207e8 0x4 0x25ff13ff
write 0xe10200b8 0x7 0xe3055e411b0202
write 0xe1020100 0x5 0x5e411b0202
write 0xe1020110 0x4 0x1b0202e1
write 0xe1020118 0x4 0x06fff105
write 0xe1020128 0x7 0xf3055e411b0202
write 0xe1020402 0x2 0x5e41
write 0xe1020420 0x4 0x1b0202e1
write 0xe1020428 0x4 0x06ff6105
write 0xe1020438 0x1 0x63
write 0xe1020439 0x1 0x05
EOF

-Alex

Quoted 26 lines · Mauro Matteo Cascella

An assertion failure issue was reported by Mr. Ziming Zhang (CC'd). It occurs in the code that processes network packets while adding data fragments into packet context. This flaw could potentially be abused by a malicious guest to abort the QEMU process on the host. This two patch series does a couple of things:

- introduces a new function in net_tx_pkt.{c,h} to check the maximum number of data fragments
- adds a check in both e1000e and vmxnet3 devices to skip the packet if the current data fragment exceeds max_raw_frags, preventing net_tx_pkt_add_raw_fragment() to be called with an invalid raw_frags

Mauro Matteo Cascella (2):
hw/net/net_tx_pkt: add function to check pkt->max_raw_frags hw/net: check max_raw_frags in e1000e and vmxnet3 devices

 hw/net/e1000e_core.c | 3 ++-
 hw/net/net_tx_pkt.c  | 5 +++++
 hw/net/net_tx_pkt.h  | 8 ++++++++
 hw/net/vmxnet3.c     | 3 ++-
 4 files changed, 17 insertions(+), 2 deletions(-)
Signature · 1 line
2.26.2
#9 lore raw

Thank you Alexander for testing the patch and providing the reproducer. I think you should be credited, along with Ziming, for independently reporting the same issue.

Quoted 58 lines · Alexander Bulekov

I sent a reproducer for the to the list some time ago, but never created a Launchpad bug...
https://www.mail-archive.com/[email protected]/msg701930.html

Anyways.. I can confirm that I can't reproduce the issue with these patches.

Minimized Reproducer:
cat << EOF | ./i386-softmmu/qemu-system-i386 -M pc-q35-5.0 -nographic \
-display none -serial none -monitor none -qtest stdio
outl 0xcf8 0x80001010
outl 0xcfc 0xe1020000
outl 0xcf8 0x80001004
outw 0xcfc 0x7
write 0xe10207e8 0x4 0x25ff13ff
write 0xe10200b8 0x7 0xe3055e411b0202
write 0xe1020100 0x5 0x5e411b0202
write 0xe1020110 0x4 0x1b0202e1
write 0xe1020118 0x4 0x06fff105
write 0xe1020128 0x7 0xf3055e411b0202
write 0xe1020402 0x2 0x5e41
write 0xe1020420 0x4 0x1b0202e1
write 0xe1020428 0x4 0x06ff6105
write 0xe1020438 0x1 0x63
write 0xe1020439 0x1 0x05
EOF

-Alex

Quoted 26 lines · Mauro Matteo Cascella

An assertion failure issue was reported by Mr. Ziming Zhang (CC'd). It occurs in the code that processes network packets while adding data fragments into packet context. This flaw could potentially be abused by a malicious guest to abort the QEMU process on the host. This two patch series does a couple of things:

- introduces a new function in net_tx_pkt.{c,h} to check the maximum number of data fragments
- adds a check in both e1000e and vmxnet3 devices to skip the packet if the current data fragment exceeds max_raw_frags, preventing net_tx_pkt_add_raw_fragment() to be called with an invalid raw_frags

Mauro Matteo Cascella (2):
hw/net/net_tx_pkt: add function to check pkt->max_raw_frags hw/net: check max_raw_frags in e1000e and vmxnet3 devices

 hw/net/e1000e_core.c | 3 ++-
 hw/net/net_tx_pkt.c  | 5 +++++
 hw/net/net_tx_pkt.h  | 8 ++++++++
 hw/net/vmxnet3.c     | 3 ++-
 4 files changed, 17 insertions(+), 2 deletions(-)

--
2.26.2

Signature · 2 lines
Mauro Matteo Cascella, Red Hat Product Security
6F78 E20B 5935 928C F0A8  1A9D 4E55 23B8 BB34 10B0
Reviewed-by
#10 lore raw
In reply to Alexander Bulekov

The idea looks good to me. I believe it makes sense to do the check in net_tx_pkt_add_raw_fragment() as suggested by Jason.

Quoted 58 lines · Alexander Bulekov

I sent a reproducer for the to the list some time ago, but never created a Launchpad bug...
https://www.mail-archive.com/[email protected]/msg701930.html

Anyways.. I can confirm that I can't reproduce the issue with these patches.

Minimized Reproducer:
cat << EOF | ./i386-softmmu/qemu-system-i386 -M pc-q35-5.0 -nographic \
-display none -serial none -monitor none -qtest stdio
outl 0xcf8 0x80001010
outl 0xcfc 0xe1020000
outl 0xcf8 0x80001004
outw 0xcfc 0x7
write 0xe10207e8 0x4 0x25ff13ff
write 0xe10200b8 0x7 0xe3055e411b0202
write 0xe1020100 0x5 0x5e411b0202
write 0xe1020110 0x4 0x1b0202e1
write 0xe1020118 0x4 0x06fff105
write 0xe1020128 0x7 0xf3055e411b0202
write 0xe1020402 0x2 0x5e41
write 0xe1020420 0x4 0x1b0202e1
write 0xe1020428 0x4 0x06ff6105
write 0xe1020438 0x1 0x63
write 0xe1020439 0x1 0x05
EOF

-Alex

Quoted 26 lines · Mauro Matteo Cascella

An assertion failure issue was reported by Mr. Ziming Zhang (CC'd). It occurs in the code that processes network packets while adding data fragments into packet context. This flaw could potentially be abused by a malicious guest to abort the QEMU process on the host. This two patch series does a couple of things:

- introduces a new function in net_tx_pkt.{c,h} to check the maximum number of data fragments
- adds a check in both e1000e and vmxnet3 devices to skip the packet if the current data fragment exceeds max_raw_frags, preventing net_tx_pkt_add_raw_fragment() to be called with an invalid raw_frags

Mauro Matteo Cascella (2):
hw/net/net_tx_pkt: add function to check pkt->max_raw_frags hw/net: check max_raw_frags in e1000e and vmxnet3 devices

hw/net/e1000e_core.c | 3 ++-
hw/net/net_tx_pkt.c  | 5 +++++
hw/net/net_tx_pkt.h  | 8 ++++++++
hw/net/vmxnet3.c     | 3 ++-
4 files changed, 17 insertions(+), 2 deletions(-)
Signature · 1 line
2.26.2