From: Si-Wei Liu <hidden> Date: 2021-02-10 21:51:48
This set attempts to fix a few independent issues recently found
in mlx5_vdpa driver. Please find details for each in the commit
message.
Patch 1 and patch 3 are already Ack'ed by Jason Wang. Patch 2 is
reworked to move virtio feature capability query to mlx5v_probe()
as suggested by Eli.
--
v1->v2: move feature capability query to probing (Eli)
Si-Wei Liu (3):
vdpa/mlx5: should exclude header length and fcs from mtu
vdpa/mlx5: fix feature negotiation across device reset
vdpa/mlx5: defer clear_virtqueues to until DRIVER_OK
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 4 ++++
drivers/vdpa/mlx5/net/mlx5_vnet.c | 42 +++++++++++++++++++++++++++-----------
2 files changed, 34 insertions(+), 12 deletions(-)
--
1.8.3.1
From: Si-Wei Liu <hidden> Date: 2021-02-10 21:50:42
The mlx_features denotes the capability for which
set of virtio features is supported by device. In
principle, this field needs not be cleared during
virtio device reset, as this capability is static
and does not change across reset.
In fact, the current code may have the assumption
that mlx_features can be reloaded from firmware
via the .get_features ops after device is reset
(via the .set_status ops), which is unfortunately
not true. The userspace VMM might save a copy
of backend capable features and won't call into
kernel again to get it on reset. This causes all
virtio features getting disabled on newly created
virtqs after device reset, while guest would hold
mismatched view of available features. For e.g.,
the guest may still assume tx checksum offload
is available after reset and feature negotiation,
causing frames with bogus (incomplete) checksum
transmitted on the wire.
Fixes: 1a86b377aa21 ("vdpa/mlx5: Add VDPA driver for supported mlx5 devices")
Signed-off-by: Si-Wei Liu <redacted>
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 25 +++++++++++++++----------
1 file changed, 15 insertions(+), 10 deletions(-)
From: Eli Cohen <hidden> Date: 2021-02-11 07:34:47
On Wed, Feb 10, 2021 at 01:47:59PM -0800, Si-Wei Liu wrote:
The mlx_features denotes the capability for which
set of virtio features is supported by device. In
principle, this field needs not be cleared during
virtio device reset, as this capability is static
and does not change across reset.
In fact, the current code may have the assumption
that mlx_features can be reloaded from firmware
via the .get_features ops after device is reset
(via the .set_status ops), which is unfortunately
not true. The userspace VMM might save a copy
of backend capable features and won't call into
kernel again to get it on reset. This causes all
virtio features getting disabled on newly created
virtqs after device reset, while guest would hold
mismatched view of available features. For e.g.,
the guest may still assume tx checksum offload
is available after reset and feature negotiation,
causing frames with bogus (incomplete) checksum
transmitted on the wire.
Fixes: 1a86b377aa21 ("vdpa/mlx5: Add VDPA driver for supported mlx5 devices")
Signed-off-by: Si-Wei Liu <redacted>
From: Jason Wang <hidden> Date: 2021-02-18 07:26:00
On 2021/2/11 上午5:47, Si-Wei Liu wrote:
The mlx_features denotes the capability for which
set of virtio features is supported by device. In
principle, this field needs not be cleared during
virtio device reset, as this capability is static
and does not change across reset.
In fact, the current code may have the assumption
that mlx_features can be reloaded from firmware
via the .get_features ops after device is reset
(via the .set_status ops), which is unfortunately
not true. The userspace VMM might save a copy
of backend capable features and won't call into
kernel again to get it on reset. This causes all
virtio features getting disabled on newly created
virtqs after device reset, while guest would hold
mismatched view of available features. For e.g.,
the guest may still assume tx checksum offload
is available after reset and feature negotiation,
causing frames with bogus (incomplete) checksum
transmitted on the wire.
Fixes: 1a86b377aa21 ("vdpa/mlx5: Add VDPA driver for supported mlx5 devices")
Signed-off-by: Si-Wei Liu <redacted>
From: Si-Wei Liu <hidden> Date: 2021-02-10 21:51:13
While virtq is stopped, get_vq_state() is supposed to
be called to get sync'ed with the latest internal
avail_index from device. The saved avail_index is used
to restate the virtq once device is started. Commit
b35ccebe3ef7 introduced the clear_virtqueues() routine
to reset the saved avail_index, however, the index
gets cleared a bit earlier before get_vq_state() tries
to read it. This would cause consistency problems when
virtq is restarted, e.g. through a series of link down
and link up events. We could defer the clearing of
avail_index to until the device is to be started,
i.e. until VIRTIO_CONFIG_S_DRIVER_OK is set again in
set_status().
Fixes: b35ccebe3ef7 ("vdpa/mlx5: Restore the hardware used index after change map")
Signed-off-by: Si-Wei Liu <redacted>
Acked-by: Jason Wang <redacted>
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
From: Eli Cohen <hidden> Date: 2021-02-11 07:35:14
On Wed, Feb 10, 2021 at 01:48:00PM -0800, Si-Wei Liu wrote:
While virtq is stopped, get_vq_state() is supposed to
be called to get sync'ed with the latest internal
avail_index from device. The saved avail_index is used
to restate the virtq once device is started. Commit
b35ccebe3ef7 introduced the clear_virtqueues() routine
to reset the saved avail_index, however, the index
gets cleared a bit earlier before get_vq_state() tries
to read it. This would cause consistency problems when
virtq is restarted, e.g. through a series of link down
and link up events. We could defer the clearing of
avail_index to until the device is to be started,
i.e. until VIRTIO_CONFIG_S_DRIVER_OK is set again in
set_status().
Fixes: b35ccebe3ef7 ("vdpa/mlx5: Restore the hardware used index after change map")
Signed-off-by: Si-Wei Liu <redacted>
Acked-by: Jason Wang <redacted>
From: Eli Cohen <hidden> Date: 2021-02-16 15:23:25
On Thu, Feb 11, 2021 at 09:33:14AM +0200, Eli Cohen wrote:
On Wed, Feb 10, 2021 at 01:48:00PM -0800, Si-Wei Liu wrote:
quoted
While virtq is stopped, get_vq_state() is supposed to
be called to get sync'ed with the latest internal
avail_index from device. The saved avail_index is used
to restate the virtq once device is started. Commit
b35ccebe3ef7 introduced the clear_virtqueues() routine
to reset the saved avail_index, however, the index
gets cleared a bit earlier before get_vq_state() tries
to read it. This would cause consistency problems when
virtq is restarted, e.g. through a series of link down
and link up events. We could defer the clearing of
avail_index to until the device is to be started,
i.e. until VIRTIO_CONFIG_S_DRIVER_OK is set again in
set_status().
Fixes: b35ccebe3ef7 ("vdpa/mlx5: Restore the hardware used index after change map")
Signed-off-by: Si-Wei Liu <redacted>
Acked-by: Jason Wang <redacted>
Acked-by: Eli Cohen <redacted>
I take it back. I think we don't need to clear the indexes at all. In
case we need to restore indexes we'll get the right values through
set_vq_state(). If we suspend the virtqueue due to VM being suspended,
qemu will query first and will provide the the queried value. In case of
VM reboot, it will provide 0 in set_vq_state().
I am sending a patch that addresses both reboot and suspend.
From: Si-Wei Liu <hidden> Date: 2021-02-17 21:56:40
On 2/16/2021 7:21 AM, Eli Cohen wrote:
On Thu, Feb 11, 2021 at 09:33:14AM +0200, Eli Cohen wrote:
quoted
On Wed, Feb 10, 2021 at 01:48:00PM -0800, Si-Wei Liu wrote:
quoted
While virtq is stopped, get_vq_state() is supposed to
be called to get sync'ed with the latest internal
avail_index from device. The saved avail_index is used
to restate the virtq once device is started. Commit
b35ccebe3ef7 introduced the clear_virtqueues() routine
to reset the saved avail_index, however, the index
gets cleared a bit earlier before get_vq_state() tries
to read it. This would cause consistency problems when
virtq is restarted, e.g. through a series of link down
and link up events. We could defer the clearing of
avail_index to until the device is to be started,
i.e. until VIRTIO_CONFIG_S_DRIVER_OK is set again in
set_status().
Fixes: b35ccebe3ef7 ("vdpa/mlx5: Restore the hardware used index after change map")
Signed-off-by: Si-Wei Liu <redacted>
Acked-by: Jason Wang <redacted>
Acked-by: Eli Cohen <redacted>
I take it back. I think we don't need to clear the indexes at all. In
case we need to restore indexes we'll get the right values through
set_vq_state(). If we suspend the virtqueue due to VM being suspended,
qemu will query first and will provide the the queried value. In case of
VM reboot, it will provide 0 in set_vq_state().
I am sending a patch that addresses both reboot and suspend.
With set_vq_state() repurposed to restoring used_index I'm fine with
this approach.
Do I have to repost a v3 of this series while dropping the 3rd patch?
-Siwei
From: Jason Wang <hidden> Date: 2021-02-18 07:27:28
On 2021/2/18 上午5:55, Si-Wei Liu wrote:
On 2/16/2021 7:21 AM, Eli Cohen wrote:
quoted
On Thu, Feb 11, 2021 at 09:33:14AM +0200, Eli Cohen wrote:
quoted
On Wed, Feb 10, 2021 at 01:48:00PM -0800, Si-Wei Liu wrote:
quoted
While virtq is stopped, get_vq_state() is supposed to
be called to get sync'ed with the latest internal
avail_index from device. The saved avail_index is used
to restate the virtq once device is started. Commit
b35ccebe3ef7 introduced the clear_virtqueues() routine
to reset the saved avail_index, however, the index
gets cleared a bit earlier before get_vq_state() tries
to read it. This would cause consistency problems when
virtq is restarted, e.g. through a series of link down
and link up events. We could defer the clearing of
avail_index to until the device is to be started,
i.e. until VIRTIO_CONFIG_S_DRIVER_OK is set again in
set_status().
Fixes: b35ccebe3ef7 ("vdpa/mlx5: Restore the hardware used index
after change map")
Signed-off-by: Si-Wei Liu <redacted>
Acked-by: Jason Wang <redacted>
Acked-by: Eli Cohen <redacted>
I take it back. I think we don't need to clear the indexes at all. In
case we need to restore indexes we'll get the right values through
set_vq_state(). If we suspend the virtqueue due to VM being suspended,
qemu will query first and will provide the the queried value. In case of
VM reboot, it will provide 0 in set_vq_state().
I am sending a patch that addresses both reboot and suspend.
With set_vq_state() repurposed to restoring used_index I'm fine with
this approach.
Do I have to repost a v3 of this series while dropping the 3rd patch?
-Siwei
From: Si-Wei Liu <hidden> Date: 2021-02-10 21:52:11
When feature VIRTIO_NET_F_MTU is negotiated on mlx5_vdpa,
22 extra bytes worth of MTU length is shown in guest.
This is because the mlx5_query_port_max_mtu API returns
the "hardware" MTU value, which does not just contain the
Ethernet payload, but includes extra lengths starting
from the Ethernet header up to the FCS altogether.
Fix the MTU so packets won't get dropped silently.
Fixes: 1a86b377aa21 ("vdpa/mlx5: Add VDPA driver for supported mlx5 devices")
Signed-off-by: Si-Wei Liu <redacted>
Acked-by: Jason Wang <redacted>
Acked-by: Eli Cohen <redacted>
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 4 ++++
drivers/vdpa/mlx5/net/mlx5_vnet.c | 15 ++++++++++++++-
2 files changed, 18 insertions(+), 1 deletion(-)