Skip to contentSkip to conversation

PCI: hv: Add paravirtual PCI front-end for Microsoft Hyper-V VMs

bug report Dan Carpenter 2 participants 3 messages Feb 7, 2017 PCI

lore
Share & tools
lore
#0 lore raw

[ No idea why I haven never sent this email before. I was just going
through all the use after free warnings again today and noticed it. ]

Hello Jake Oshins,

The patch 4daace0d8ce8: "PCI: hv: Add paravirtual PCI front-end for Microsoft Hyper-V VMs" from Feb 16, 2016, leads to the following static checker warning:

drivers/pci/host/pci-hyperv.c:1441 pci_devices_present_work()
error: dereferencing freed memory 'dr'

drivers/pci/host/pci-hyperv.c

  1410          /* Pull this off the queue and process it if it was the last one. */
  1411          spin_lock_irqsave(&hbus->device_list_lock, flags);
  1412          while (!list_empty(&hbus->dr_list)) {
  1413                  dr = list_first_entry(&hbus->dr_list, struct hv_dr_state,
  1414                                        list_entry);
  1415                  list_del(&dr->list_entry);
  1416  
  1417                  /* Throw this away if the list still has stuff in it. */
  1418                  if (!list_empty(&hbus->dr_list)) {
  1419                          kfree(dr);
                                ^^^^^^^^^

We free "dr". Presumably we should set dr = NULL here?

1420 continue;
1421 }
1422 }
1423 spin_unlock_irqrestore(&hbus->device_list_lock, flags);
1424
1425 if (!dr) {

  1426                  up(&hbus->enum_sem);
  1427                  put_hvpcibus(hbus);
  1428                  return;
  1429          }
  1430  
  1431          /* First, mark all existing children as reported missing. */
  1432          spin_lock_irqsave(&hbus->device_list_lock, flags);
  1433          list_for_each(iter, &hbus->children) {
  1434                          hpdev = container_of(iter, struct hv_pci_dev,
  1435                                               list_entry);
  1436                          hpdev->reported_missing = true;
  1437          }
  1438          spin_unlock_irqrestore(&hbus->device_list_lock, flags);
  1439  
  1440          /* Next, add back any reported devices. */
  1441          for (child_no = 0; child_no < dr->device_count; child_no++) {
                                              ^^^^^^^^^^^^^^^^

Use after free.

  1442                  found = false;
  1443                  new_desc = &dr->func[child_no];
  1444  
  1445                  spin_lock_irqsave(&hbus->device_list_lock, flags);

regards,
dan carpenter

#1 lore raw
Quoted 6 lines

-----Original Message-----
From: Dan Carpenter [mailto:[email protected]]
Sent: Monday, February 6, 2017 11:12 PM
To: Jake Oshins <[email protected]>
Cc: [email protected]; [email protected]
Subject: [bug report] PCI: hv: Add paravirtual PCI front-end for Microsof=

t

Quoted 8 lines

Hyper-V VMs
=20
[ No idea why I haven never sent this email before. I was just going
through all the use after free warnings again today and noticed it. ]
=20
Hello Jake Oshins,
=20
The patch 4daace0d8ce8: "PCI: hv: Add paravirtual PCI front-end for

Quoted 8 lines

Microsoft Hyper-V VMs" from Feb 16, 2016, leads to the following static checker warning:
=20
drivers/pci/host/pci-hyperv.c:1441 pci_devices_present_work()
error: dereferencing freed memory 'dr'
=20
drivers/pci/host/pci-hyperv.c
1410 /* Pull this off the queue and process it if it was the l=

ast one. */

Quoted 3 lines
  1411          spin_lock_irqsave(&hbus->device_list_lock, flags);
  1412          while (!list_empty(&hbus->dr_list)) {
  1413                  dr =3D list_first_entry(&hbus->dr_list, struct hv=

_dr_state,

Quoted 4 lines

1414 list_entry);
1415 list_del(&dr->list_entry);
1416
1417 /* Throw this away if the list still has stuff in=

it. */

Quoted 11 lines

1418 if (!list_empty(&hbus->dr_list)) {

  1419                          kfree(dr);
                                ^^^^^^^^^

We free "dr". Presumably we should set dr =3D NULL here? =20
1420 continue;
1421 }
1422 }
1423 spin_unlock_irqrestore(&hbus->device_list_lock, flags);
1424
1425 if (!dr) {

Quoted 6 lines
  1426                  up(&hbus->enum_sem);
  1427                  put_hvpcibus(hbus);
  1428                  return;

1429 }
1430
1431 /* First, mark all existing children as reported missing.=

*/

Quoted 3 lines
  1432          spin_lock_irqsave(&hbus->device_list_lock, flags);
  1433          list_for_each(iter, &hbus->children) {
  1434                          hpdev =3D container_of(iter, struct hv_pc=

i_dev,

Quoted 7 lines
  1435                                               list_entry);
  1436                          hpdev->reported_missing =3D true;
  1437          }
  1438          spin_unlock_irqrestore(&hbus->device_list_lock, flags);
  1439
  1440          /* Next, add back any reported devices. */
  1441          for (child_no =3D 0; child_no < dr->device_count; child_n=

o++) {

Quoted 7 lines
                                              ^^^^^^^^^^^^^^^^

Use after free.
=20
1442 found =3D false;
1443 new_desc =3D &dr->func[child_no];
1444
1445 spin_lock_irqsave(&hbus->device_list_lock, flags)=

;

Quoted 4 lines

=20
=20
regards,
dan carpenter

I'm pretty sure that this is a false positive. It only frees the struct if=
there is another entry in the list, and then it immediately overwrites the= pointer with the next entry in the list.

What's the right move here? Should we send a patch that nulls the pointer = just to make a static analysis hit go away? It's not a hot path.

Thanks,
Jake Oshins

#2 lore raw
Quoted 68 lines · Jake Oshins
Quoted 59 lines

-----Original Message-----
From: Dan Carpenter [mailto:[email protected]]
Sent: Monday, February 6, 2017 11:12 PM
To: Jake Oshins <[email protected]>
Cc: [email protected]; [email protected]
Subject: [bug report] PCI: hv: Add paravirtual PCI front-end for Microsoft Hyper-V VMs

[ No idea why I haven never sent this email before. I was just going
through all the use after free warnings again today and noticed it. ]

Hello Jake Oshins,

The patch 4daace0d8ce8: "PCI: hv: Add paravirtual PCI front-end for Microsoft Hyper-V VMs" from Feb 16, 2016, leads to the following static checker warning:

drivers/pci/host/pci-hyperv.c:1441 pci_devices_present_work()
error: dereferencing freed memory 'dr'

drivers/pci/host/pci-hyperv.c

  1410          /* Pull this off the queue and process it if it was the last one. */
  1411          spin_lock_irqsave(&hbus->device_list_lock, flags);
  1412          while (!list_empty(&hbus->dr_list)) {
  1413                  dr = list_first_entry(&hbus->dr_list, struct hv_dr_state,
  1414                                        list_entry);
  1415                  list_del(&dr->list_entry);
  1416
  1417                  /* Throw this away if the list still has stuff in it. */
  1418                  if (!list_empty(&hbus->dr_list)) {
  1419                          kfree(dr);
                                ^^^^^^^^^

We free "dr". Presumably we should set dr = NULL here?

1420 continue;
1421 }
1422 }
1423 spin_unlock_irqrestore(&hbus->device_list_lock, flags);
1424
1425 if (!dr) {

  1426                  up(&hbus->enum_sem);
  1427                  put_hvpcibus(hbus);
  1428                  return;
  1429          }
  1430
  1431          /* First, mark all existing children as reported missing. */
  1432          spin_lock_irqsave(&hbus->device_list_lock, flags);
  1433          list_for_each(iter, &hbus->children) {
  1434                          hpdev = container_of(iter, struct hv_pci_dev,
  1435                                               list_entry);
  1436                          hpdev->reported_missing = true;
  1437          }
  1438          spin_unlock_irqrestore(&hbus->device_list_lock, flags);
  1439
  1440          /* Next, add back any reported devices. */
  1441          for (child_no = 0; child_no < dr->device_count; child_no++) {
                                              ^^^^^^^^^^^^^^^^

Use after free.

Quoted 8 lines
  1442                  found = false;
  1443                  new_desc = &dr->func[child_no];
  1444
  1445                  spin_lock_irqsave(&hbus->device_list_lock, flags);

regards,
dan carpenter

Quoted 7 lines

I'm pretty sure that this is a false positive. It only frees the struct if there is another entry in the list, and then it immediately overwrites the pointer with the next entry in the list.

What's the right move here? Should we send a patch that nulls the pointer just to make a static analysis hit go away? It's not a hot path.

Nah... Sorry. I remember now that I did send this warning before and K. Y. Srinivasan explained this to me like you did.

I'm close to being able to silence this false positive in Smatch. Don't worry about it. My bad.

regards,
dan carpenter