[net,0/4] Fix various issue of vhost

Message

Jason Wang Dec. 10, 2018, 9:44 a.m. UTC

Hi:

This series tries to fix various issues of vhost:

- Patch 1 adds a missing write barrier between used idx updating and
  logging.
- Patch 2-3 brings back the protection of device IOTLB through vq
  mutex, this fixes possible use after free in device IOTLB entries.
- Patch 4 fixes the diry page logging when device IOTLB is
  enabled. We should done through GPA instead of GIOVA, this was done
  through logging through iovec and traversing GPA->HPA list for the
  GPA.

Please consider them for -stable.

Thanks

Jason Wang (4):
  vhost: make sure used idx is seen before log in vhost_add_used_n()
  vhost_net: rework on the lock ordering for busy polling
  Revert "net: vhost: lock the vqs one by one"
  vhost: log dirty page correctly

 drivers/vhost/net.c   |  21 +++++++--
 drivers/vhost/vhost.c | 101 ++++++++++++++++++++++++++++++++++--------
 drivers/vhost/vhost.h |   3 +-
 3 files changed, 102 insertions(+), 23 deletions(-)

Comments

David Miller Dec. 10, 2018, 7:47 p.m. UTC | #1

From: Jason Wang <[e-mail]>
Date: Mon, 10 Dec 2018 17:44:50 +0800

> This series tries to fix various issues of vhost:
> 
> - Patch 1 adds a missing write barrier between used idx updating and
>   logging.
> - Patch 2-3 brings back the protection of device IOTLB through vq
>   mutex, this fixes possible use after free in device IOTLB entries.
> - Patch 4 fixes the diry page logging when device IOTLB is
>   enabled. We should done through GPA instead of GIOVA, this was done
>   through logging through iovec and traversing GPA->HPA list for the
>   GPA.
> 
> Please consider them for -stable.

Looks like the kbuild robot found some problems.

->used is a pointer (which might be 32-bit) and you're casting it to
a u64 in the translate_desc() calls of patch #4.

Please make sure that you don't actually require the full domain of
a u64 in these values, as obviously if vq->used is a pointer you will
only get a 32-bit domain on 32-bit architectures.

Jason Wang Dec. 11, 2018, 3:01 a.m. UTC | #2

On 2018/12/11 上午3:47, David Miller wrote:
> From: Jason Wang <[e-mail]>
> Date: Mon, 10 Dec 2018 17:44:50 +0800
>
>> This series tries to fix various issues of vhost:
>>
>> - Patch 1 adds a missing write barrier between used idx updating and
>>    logging.
>> - Patch 2-3 brings back the protection of device IOTLB through vq
>>    mutex, this fixes possible use after free in device IOTLB entries.
>> - Patch 4 fixes the diry page logging when device IOTLB is
>>    enabled. We should done through GPA instead of GIOVA, this was done
>>    through logging through iovec and traversing GPA->HPA list for the
>>    GPA.
>>
>> Please consider them for -stable.
> Looks like the kbuild robot found some problems.
>
> ->used is a pointer (which might be 32-bit) and you're casting it to
> a u64 in the translate_desc() calls of patch #4.
>
> Please make sure that you don't actually require the full domain of
> a u64 in these values, as obviously if vq->used is a pointer you will
> only get a 32-bit domain on 32-bit architectures.


It seems the reason is that I cast from plain void pointer directly. Let 
me cast it to uintptr_t first.

Thanks


>