x104/0x140
entry_SYSCALL_64_after_hwframe+0x76/0x7e
RIP: 0033:0x7f0c25b0feca
[...]
---[ end trace ]---
Signed-off-by: Dragos Tatulea
Fixes: 83e445e64f48 ("vdpa/mlx5: Fix error path during device add")
Reported-by: Wenli Quan
Closes:
https://lore.kerne
evious value of the needs_teardown
flag when re-calculating it during VQ size configuration.
Fixes: 0fe963d6fc16 ("vdpa/mlx5: Re-create HW VQs under certain conditions")
Signed-off-by: Dragos Tatulea
Reviewed-by: Shahar Shitrit
Reviewed-by: Si-Wei Liu
Tested-by: Si-Wei Liu
---
drive
are correctly populated.
Fixes: 94abbccdf291 ("vdpa/mlx5: Add shared memory registration code")
Cc: sta...@vger.kernel.org
Reported-by: Cong Meng
Signed-off-by: Si-Wei Liu
Signed-off-by: Dragos Tatulea
Acked-by: Eugenio Pérez
---
Changes in v2:
- Added Reported-by tag.
- Fixed typo fo
On 02/18, Eugenio Perez Martin wrote:
> On Mon, Feb 17, 2025 at 8:45 PM Dragos Tatulea wrote:
> >
> > From: Si-Wei Liu
> >
> > create_user_mr() has correct code to count the number of null keys
> > used to fill in a hole for the memory map. However, fill_indir(
correctly populated.
Fixes: 94abbccdf291 ("vdpa/mlx5: Add shared memory registration code")
Cc: sta...@vger.kernel.org
Signed-off-by: Si-Wei Liu
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/core/mr.c | 7 +--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/dr
gt;actual_features =
> + (device_features & BIT_ULL(VIRTIO_F_VERSION_1));
>
> ndev->vqs = kcalloc(max_vqs, sizeof(*ndev->vqs), GFP_KERNEL);
> ndev->event_cbs = kcalloc(max_vqs + 1, sizeof(*ndev->event_cbs),
> GFP_KERNEL);
> --
> 2.34.1
>
The patch looks good now. Thanks!
Reviewed-by: Dragos Tatulea
On 01/30, Konstantin Shkolnyy wrote:
> mlx5_vdpa_dev_add() doesn’t initialize mvdev->actual_features. It’s
> initialized later by mlx5_vdpa_set_driver_features(). However,
> mlx5_vdpa_get_config() depends on the VIRTIO_F_VERSION_1 flag in
> actual_features to return data with correct endianness. Wh
On 13.11.24 15:49, Michael S. Tsirkin wrote:
> On Wed, Nov 13, 2024 at 03:33:35PM +0100, Dragos Tatulea wrote:
>>
>>
>> On 13.11.24 07:32, Michael S. Tsirkin wrote:
>>> On Mon, Oct 21, 2024 at 04:40:40PM +0300, Dragos Tatulea wrote:
>>>> From: Si-Wei L
On 13.11.24 07:32, Michael S. Tsirkin wrote:
> On Mon, Oct 21, 2024 at 04:40:40PM +0300, Dragos Tatulea wrote:
>> From: Si-Wei Liu
>>
>> The starting iova address to iterate iotlb map entry within a range
>> was set to an irrelevant value when passing to the itree_n
On 21.10.24 15:40, Dragos Tatulea wrote:
> Here are 2 fixes from Si-Wei:
> - The first one is an important fix that has to be applied as far
> back as possible (hence CC'ing linux-stable).
> - The second is more of an improvement. That's why it doesn't have the
>
t endianness of speed.
>>
>> Signed-off-by: Carlos Bilbao
>> Reviewed-by: Dragos Tatulea
>> ---
>> drivers/vdpa/mlx5/net/mlx5_vnet.c | 12
>> 1 file changed, 12 insertions(+)
>>
>> diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c
>>
ffeace233f3
Modules linked in: ...
CR2: 00b8
Fixes: 62111654481d ("vdpa/mlx5: Postpone MR deletion")
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 21 +
1 file changed, 5 insertions(+), 16 deletions(-)
diff --git a/drivers/vdpa/ml
Here are 2 fixes from Si-Wei:
- The first one is an important fix that has to be applied as far
back as possible (hence CC'ing linux-stable).
- The second is more of an improvement. That's why it doesn't have the
Fixes tag.
I'd like to thank Si-Wei for the effort of finding and fixing these
is
ke
it consistent with the following for-loop.
Fixes: 94abbccdf291 ("vdpa/mlx5: Add shared memory registration code")
Signed-off-by: Si-Wei Liu
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/core/mr.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/drivers/vdpa/m
code is already correct:
the range in [mr->start, map->start) was handled by a different
iteration.
Fixes: 94abbccdf291 ("vdpa/mlx5: Add shared memory registration code")
Cc: sta...@vger.kernel.org
Signed-off-by: Si-Wei Liu
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5
On 11.09.24 10:02, Eugenio Perez Martin wrote:
> On Mon, Sep 9, 2024 at 11:30 AM Dragos Tatulea wrote:
>>
>>
>>
>> On 30.08.24 12:58, Dragos Tatulea wrote:
>>> This series improves the time of .set_map() operations by parallelizing
>>> the MKEY cr
On 11.09.24 10:00, Eugenio Perez Martin wrote:
> On Mon, Sep 9, 2024 at 11:33 AM Dragos Tatulea wrote:
>>
>>
>>
>> On 16.08.24 11:01, Dragos Tatulea wrote:
>>> Currently, commands that qualify as throttled can't be used via the
>>> async API.
On 16.08.24 11:01, Dragos Tatulea wrote:
> Currently, commands that qualify as throttled can't be used via the
> async API. That's due to the fact that the throttle semaphore can sleep
> but the async API can't.
>
> This patch allows throttling in the async API by
On 30.08.24 12:58, Dragos Tatulea wrote:
> This series improves the time of .set_map() operations by parallelizing
> the MKEY creation and deletion for direct MKEYs. Looking at the top
> level MKEY creation/deletion functions, the following improvement can
On 04.09.24 08:34, Jason Wang wrote:
> On Wed, Sep 4, 2024 at 1:59 PM Dragos Tatulea wrote:
>>
>>
>>
>> On 04.09.24 05:38, Jason Wang wrote:
>>> On Wed, Sep 4, 2024 at 1:15 AM Carlos Bilbao
>>> wrote:
>>>>
>>>> From: Carlos
On 04.09.24 05:38, Jason Wang wrote:
> On Wed, Sep 4, 2024 at 1:15 AM Carlos Bilbao
> wrote:
>>
>> From: Carlos Bilbao
>>
>> Remove invalid ioctl VHOST_VDPA_SET_CONFIG and all its implementations
>> with vdpa_config_ops->set_config(). This is needed per virtio spec
>> requirements; virtio-spec
On 03.09.24 10:10, Eugenio Perez Martin wrote:
> On Tue, Sep 3, 2024 at 9:48 AM Dragos Tatulea wrote:
>>
>>
>>
>> On 03.09.24 09:40, Lei Yang wrote:
>>> On Mon, Sep 2, 2024 at 7:05 PM Dragos Tatulea wrote:
>>>>
>>>> Hi Lei,
On 03.09.24 09:40, Lei Yang wrote:
> On Mon, Sep 2, 2024 at 7:05 PM Dragos Tatulea wrote:
>>
>> Hi Lei,
>>
>> On 02.09.24 12:03, Lei Yang wrote:
>>> Hi Dragos
>>>
>>> QE tested this series with mellanox nic, it failed with [1] when
>>
Hi Cindy,
On 30.08.24 15:52, Dragos Tatulea wrote:
>
>
> On 30.08.24 11:12, Cindy Lu wrote:
>> On Thu, 29 Aug 2024 at 18:00, Dragos Tatulea wrote:
>>>
>>>
>>>
>>> On 29.08.24 11:05, Cindy Lu wrote:
>>>> On Wed, 28 Aug 2024 at 17:37
On 30.08.24 11:12, Cindy Lu wrote:
> On Thu, 29 Aug 2024 at 18:00, Dragos Tatulea wrote:
>>
>>
>>
>> On 29.08.24 11:05, Cindy Lu wrote:
>>> On Wed, 28 Aug 2024 at 17:37, Dragos Tatulea wrote:
>>>>
>>>>
>>>>
>>>&g
On 29.08.24 17:15, Eugenio Perez Martin wrote:
> On Thu, Aug 29, 2024 at 3:54 PM Dragos Tatulea wrote:
>>
>>
>>
>> On 29.08.24 15:10, Eugenio Perez Martin wrote:
>>> On Wed, Aug 21, 2024 at 1:41 PM Dragos Tatulea wrote:
>>>>
>>>> Us
Now that the mr resources have their own namespace in the
struct, give the lock a clearer name.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
Acked-by: Eugenio Pérez
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 2 +-
drivers/vdpa/mlx5/core/mr.c| 20 ++--
drivers
lock taken, this operation it will have to wait for the handler
to finish.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 10 ++
drivers/vdpa/mlx5/core/mr.c| 55 --
drivers/vdpa/mlx5/net/mlx5_vnet.c | 4
A followup patch will use this name for something else.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 2 +-
drivers/vdpa/mlx5/core/mr.c| 2 +-
drivers/vdpa/mlx5/net/mlx5_vnet.c | 8
3 files changed, 6 insertions(+), 6
urces() / mlx5_vdpa_free_resources()
into these new functions. However, the lifetime at the
outer scope remains the same:
mlx5_vdpa_dev_add() / mlx5_vdpa_dev_free()
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 2 ++
drivers/vdpa/mlx5/core/mr.c
Group all mapping related resources into their own structure.
Upcoming patches will add more members in this new structure.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
Acked-by: Eugenio Pérez
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 13 ++-
drivers/vdpa/mlx5/core/mr.c
Use the async interface to issue MTT MKEY deletion.
This makes destroy_user_mr() on average 8x times faster.
This number is also dependent on the size of the MR being
deleted.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
Acked-by: Eugenio Pérez
---
drivers/vdpa/mlx5/core/mr.c | 64
create_user_mr() 3-5x faster, depending on
the size of the MR.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mr.c | 120 +---
1 file changed, 98 insertions(+), 22 deletions(-)
diff --git a/drivers/vdpa/mlx5/core/mr.c b/drivers/vdpa/mlx5
where appropriate
(only second patch).
- Added macro define for MTT alignment in first patch.
- Improved commit messages/comments based on review comments.
- Removed extra newlines.
---
Dragos Tatulea (7):
vdpa/mlx5: Create direct MKEYs in parallel
vdpa/mlx5: Delete direct MKEYs in parallel
Hi Cindy,
On 30.08.24 11:29, Cindy Lu wrote:
> On Fri, 30 Aug 2024 at 03:03, Dragos Tatulea wrote:
>>
>>
>>
>> On 29.08.24 12:00, Dragos Tatulea wrote:
>>>
>>>
>>> On 29.08.24 11:05, Cindy Lu wrote:
>>>> On Wed, 28 Aug 2024 at
(resending as I accidentally replied only to Carlos)
On 29.08.24 18:16, Carlos Bilbao wrote:
> From: Carlos Bilbao
>
> Include support to update the vDPA configuration fields of speed and
> duplex (as needed by VHOST_VDPA_SET_CONFIG). This includes function
> mlx5_vdpa_set_config() as well as ch
On 29.08.24 12:00, Dragos Tatulea wrote:
>
>
> On 29.08.24 11:05, Cindy Lu wrote:
>> On Wed, 28 Aug 2024 at 17:37, Dragos Tatulea wrote:
>>>
>>>
>>>
>>> On 28.08.24 11:00, Cindy Lu wrote:
>>>> On Wed, 28 Aug 2024 at 09:51,
and duplex. Add
> needed helper cpu_to_mlx5vdpa32() to convert endianness of speed.
>
> Signed-off-by: Carlos Bilbao
Nit: prefix is vdpa/mlx5. Once that is fixed, for this patch:
Reviewed-by: Dragos Tatulea
> ---
> drivers/vdpa/mlx5/net/mlx5_vnet.c | 12
> 1 file ch
On 29.08.24 16:37, Eugenio Perez Martin wrote:
> On Wed, Aug 21, 2024 at 1:42 PM Dragos Tatulea wrote:
>>
>> There's currently not a lot of action happening during
>> the init/destroy of MR resources. But more will be added
>> in the upcoming patches.
>
On 29.08.24 17:07, Eugenio Perez Martin wrote:
> On Wed, Aug 21, 2024 at 1:42 PM Dragos Tatulea wrote:
>>
>> Currently, when a new MR is set up, the old MR is deleted. MR deletion
>> is about 30-40% the time of MR creation. As deleting the old MR is not
>> important
On 28.08.24 20:16, Carlos Bilbao wrote:
> From: Carlos Bilbao
>
> mlx5_vdpa vDPA devices currently don't support reporting or setting the
> speed and duplex and hence should be UNKNOWN instead of zero.
>
> Signed-off-by: Carlos Bilbao
> ---
> drivers/vdpa/mlx5/net/mlx5_vnet.c | 7 +++
>
On 29.08.24 11:05, Cindy Lu wrote:
> On Wed, 28 Aug 2024 at 17:37, Dragos Tatulea wrote:
>>
>>
>>
>> On 28.08.24 11:00, Cindy Lu wrote:
>>> On Wed, 28 Aug 2024 at 09:51, Jason Wang wrote:
>>>>
>>>> On Wed, Aug 28, 2024 at 12:03 AM D
On 28.08.24 11:00, Cindy Lu wrote:
> On Wed, 28 Aug 2024 at 09:51, Jason Wang wrote:
>>
>> On Wed, Aug 28, 2024 at 12:03 AM Dragos Tatulea wrote:
>>>
>>> When the vdpa device is configured without a specific MAC
>>> address, the vport MAC address i
On 28.08.24 08:22, Si-Wei Liu wrote:
>
>
> On 8/27/2024 9:08 AM, Dragos Tatulea wrote:
>> Certain error paths from mlx5_vdpa_dev_add() can end up releasing mr
>> resources which never got initialized in the first place.
>>
>> This patch adds the missing check
On 28.08.24 07:54, Si-Wei Liu wrote:
>
>
> On 8/27/2024 9:02 AM, Dragos Tatulea wrote:
>> When the vdpa device is configured without a specific MAC
>> address, the vport MAC address is used. However, this
>> address can be 0 which prevents the driver from properly
On 27.08.24 04:03, Jason Wang wrote:
> On Tue, Aug 27, 2024 at 12:11 AM Dragos Tatulea wrote:
>>
>>
>> On 26.08.24 16:24, Andrew Lunn wrote:
>>> On Mon, Aug 26, 2024 at 11:06:09AM +0200, Dragos Tatulea wrote:
>>>>
>>>>
>>>> On
? clear_bhb_loop+0x25/0x80
? clear_bhb_loop+0x25/0x80
? clear_bhb_loop+0x25/0x80
? clear_bhb_loop+0x25/0x80
entry_SYSCALL_64_after_hwframe+0x78/0x80
Fixes: 512c0cdd80c1 ("vdpa/mlx5: Decouple cvq iotlb handling from hw mapping
code")
Signed-off-by: Dragos Tatulea
Reviewed-by: Co
vport.
Now it's possible to create a vdpa device without a
MAC address and run qemu with this device without needing
to configure an explicit MAC address.
Signed-off-by: Dragos Tatulea
Reviewed-by: Jiri Pirko
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 3 +++
1 file changed, 3 insertions(+)
On 26.08.24 16:24, Andrew Lunn wrote:
> On Mon, Aug 26, 2024 at 11:06:09AM +0200, Dragos Tatulea wrote:
>>
>>
>> On 23.08.24 18:54, Carlos Bilbao wrote:
>>> Hello,
>>>
>>> I'm debugging my vDPA setup, and when using ioctl to retrieve the
&
On 26.08.24 16:26, Carlos Bilbao wrote:
> Hello Dragos,
>
> On 8/26/24 4:06 AM, Dragos Tatulea wrote:
>>
>> On 23.08.24 18:54, Carlos Bilbao wrote:
>>> Hello,
>>>
>>> I'm debugging my vDPA setup, and when using ioctl to retrieve the
>>
On 23.08.24 18:54, Carlos Bilbao wrote:
> Hello,
>
> I'm debugging my vDPA setup, and when using ioctl to retrieve the
> configuration, I noticed that it's running in half duplex mode:
>
> Configuration data (24 bytes):
> MAC address: (Mac address)
> Status: 0x0001
> Max virtqueue pairs:
lock taken, this operation it will have to wait for the handler
to finish.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 10 ++
drivers/vdpa/mlx5/core/mr.c| 51 --
drivers/vdpa/mlx5/net/mlx5_vnet.c | 3
Now that the mr resources have their own namespace in the
struct, give the lock a clearer name.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 2 +-
drivers/vdpa/mlx5/core/mr.c| 20 ++--
drivers/vdpa/mlx5/core
VQPs)
This series must be applied on top of the parallel VQ suspend/resume
series [0].
[0] https://lore.kernel.org/all/20240816090159.1967650-1-dtatu...@nvidia.com/
Dragos Tatulea (7):
vdpa/mlx5: Create direct MKEYs in parallel
vdpa/mlx5: Delete direct MKEYs in parallel
vdpa/mlx5: Rename
There's currently not a lot of action happening during
the init/destroy of MR resources. But more will be added
in the upcoming patches.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 2 ++
drivers/vdpa/mlx5/core/mr.c
Group all mapping related resources into their own structure.
Upcoming patches will add more members in this new structure.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 13 ++-
drivers/vdpa/mlx5/core/mr.c| 30
Use the async interface to issue MTT MKEY deletion.
This makes destroy_user_mr() on average 8x times faster.
This number is also dependent on the size of the MR being
deleted.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mr.c | 66
A followup patch will use this name for something else.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 2 +-
drivers/vdpa/mlx5/core/mr.c| 2 +-
drivers/vdpa/mlx5/net/mlx5_vnet.c | 8
3 files changed, 6 insertions(+), 6
create_user_mr() 3-5x faster, depending on
the size of the MR.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin Ratiu
---
drivers/vdpa/mlx5/core/mr.c | 118 +---
1 file changed, 96 insertions(+), 22 deletions(-)
diff --git a/drivers/vdpa/mlx5/core/mr.c b/drivers/vdpa/mlx5
On 02.08.24 15:14, Michael S. Tsirkin wrote:
> On Fri, Aug 02, 2024 at 10:20:17AM +0300, Dragos Tatulea wrote:
>> This series parallelizes the mlx5_vdpa device suspend and resume
>> operations through the firmware async API. The purpose is to reduce live
>> migration downt
() will exit early. If
the VQ is not initialized, setup_vq() will create it and resume_vqs()
will resume it.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 22 --
1 file changed, 12 insertions(+), 10 deletions(-)
diff --git a
change_num_qps() has a lot of multiplications by 2 to convert
the number of VQ pairs to number of VQs. This patch simplifies
the code by doing the VQP -> VQ count conversion at the beginning
in a variable.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/
changes.
For 1 vDPA device x 32 VQs (16 VQPs) attached to a large VM (256 GB RAM,
32 CPUs x 2 threads per core), the device suspend time is reduced from
~13 ms to ~2.5 ms.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
Acked-by: Eugenio Pérez
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 6
is reduced from
~16 ms to ~4.5 ms.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
Acked-by: Eugenio Pérez
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 40 +++
1 file changed, 14 insertions(+), 26 deletions(-)
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c
b
large VM (256 GB RAM,
32 CPUs x 2 threads per core), the device suspend time is reduced from
~37 ms to ~13 ms.
A later patch will remove the link unregister operation which will make
it even faster.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
Acked-by: Eugenio Pérez
---
drivers/vdpa
modifying multiple vqs in parallel.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 154 --
1 file changed, 106 insertions(+), 48 deletions(-)
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c
b/drivers/vdpa/mlx5/net/mlx5_vnet.c
issuing more commands at a time.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 2 +
drivers/vdpa/mlx5/net/mlx5_vnet.c | 101 ++---
2 files changed, 78 insertions(+), 25 deletions(-)
diff --git a/drivers/vdpa/mlx5
-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 15 ++
drivers/vdpa/mlx5/core/resources.c | 73 ++
2 files changed, 88 insertions(+)
diff --git a/drivers/vdpa/mlx5/core/mlx5_vdpa.h
b/drivers/vdpa/mlx5/core/mlx5_vdpa.h
(semaphore at 0) returns EBUSY
to signal to the caller that they need to wait for the completion of
previously issued commands.
Furthermore, make sure that the semaphore is released in the callback.
Signed-off-by: Dragos Tatulea
Cc: Leon Romanovsky
Reviewed-by: Tariq Toukan
---
drivers/net/etherne
mlx5_vdpa_err() was missing. This patch adds it and uses it in the
necessary places.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
Acked-by: Eugenio Pérez
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 5 +
drivers/vdpa/mlx5/net/mlx5_vnet.c | 24
2 files
in completely in
the vhost tree [0]. The mlx5_core patch was reviewed by Tariq who is
also a maintainer for mlx5_core.
[0] -
https://lore.kernel.org/virtualization/6582792d-8db2-4bc0-bf3a-248fe5c8f...@nvidia.com/T/#maefabb2fde5adfb322d16ca16ae64d540f75b7d2
Dragos Tatulea (10):
net/mlx5: Supp
On 14.08.24 07:29, Jason Wang wrote:
> On Tue, Aug 13, 2024 at 8:53 PM Dragos Tatulea wrote:
>>
>>
>>
>> On 13.08.24 08:26, Jason Wang wrote:
>>> On Mon, Aug 12, 2024 at 7:22 PM Dragos Tatulea wrote:
>>>>
>>>>
>>>>
&
On 13.08.24 08:26, Jason Wang wrote:
> On Mon, Aug 12, 2024 at 7:22 PM Dragos Tatulea wrote:
>>
>>
>>
>> On 12.08.24 08:49, Jason Wang wrote:
>>> On Mon, Aug 12, 2024 at 1:47 PM Jason Wang wrote:
>>>>
>>>> On Fri, Aug 9, 2024
On 12.08.24 08:49, Jason Wang wrote:
> On Mon, Aug 12, 2024 at 1:47 PM Jason Wang wrote:
>>
>> On Fri, Aug 9, 2024 at 2:04 AM Dragos Tatulea wrote:
>>>
>>>
>>>
>>> On 08.08.24 10:20, Jason Wang wrote:
>>>> We used to call i
On 08.08.24 10:20, Jason Wang wrote:
> We used to call irq_bypass_unregister_producer() in
> vhost_vdpa_setup_vq_irq() which is problematic as we don't know if the
> token pointer is still valid or not.
>
> Actually, we use the eventfd_ctx as the token so the life cycle of the
> token should be
On 07.08.24 15:25, Eugenio Perez Martin wrote:
> On Fri, Aug 2, 2024 at 9:24 AM Dragos Tatulea wrote:
>>
>> This series parallelizes the mlx5_vdpa device suspend and resume
>> operations through the firmware async API. The purpose is to reduce live
>> migration
On 06.08.24 10:18, Dragos Tatulea wrote:
> (Re-sending. I messed up the previous message, sorry about that.)
>
> On 06.08.24 04:57, Jason Wang wrote:
>> On Mon, Aug 5, 2024 at 11:59 PM Dragos Tatulea wrote:
>>>
>>> On 05.08.24 05:17, Jason Wang wrote:
&g
(Re-sending. I messed up the previous message, sorry about that.)
On 06.08.24 04:57, Jason Wang wrote:
> On Mon, Aug 5, 2024 at 11:59 PM Dragos Tatulea wrote:
>>
>> On 05.08.24 05:17, Jason Wang wrote:
>>> On Fri, Aug 2, 2024 at 2:51 PM Dragos Tatulea wrote:
>>&g
On 05.08.24 05:17, Jason Wang wrote:
> On Fri, Aug 2, 2024 at 2:51 PM Dragos Tatulea wrote:
>>
>> On Fri, 2024-08-02 at 11:29 +0800, Jason Wang wrote:
>>> On Thu, Aug 1, 2024 at 11:38 PM Dragos Tatulea wrote:
>>>>
>>>> The following workflow
changes.
For 1 vDPA device x 32 VQs (16 VQPs) attached to a large VM (256 GB RAM,
32 CPUs x 2 threads per core), the device suspend time is reduced from
~13 ms to ~2.5 ms.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 6 --
1 file changed, 4
is reduced from
~16 ms to ~4.5 ms.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 40 +++
1 file changed, 14 insertions(+), 26 deletions(-)
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c
b/drivers/vdpa/mlx5/net
large VM (256 GB RAM,
32 CPUs x 2 threads per core), the device suspend time is reduced from
~37 ms to ~13 ms.
A later patch will remove the link unregister operation which will make
it even faster.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 56
modifying multiple vqs in parallel.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 150 --
1 file changed, 103 insertions(+), 47 deletions(-)
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c
b/drivers/vdpa/mlx5/net/mlx5_vnet.c
but the infrastructure is there
to issue commands in parallel, including ratelimiting the number
of issued async commands to firmware.
A later patch will switch to issuing more commands at a time.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h
mlx5_vdpa_err() was missing. This patch adds it and uses it in the
necessary places.
Signed-off-by: Dragos Tatulea
Reviewed-by: Tariq Toukan
---
drivers/vdpa/mlx5/core/mlx5_vdpa.h | 5 +
drivers/vdpa/mlx5/net/mlx5_vnet.c | 24
2 files changed, 17 insertions
the vhost
tree and only then the remaining patches can be applied.
[0]
https://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux.git/log/?h=mlx5-vhost
Dragos Tatulea (7):
net/mlx5: Support throttled commands from async API
vdpa/mlx5: Introduce error logging function
vdpa/mlx5: Use async
On Fri, 2024-08-02 at 11:29 +0800, Jason Wang wrote:
> On Thu, Aug 1, 2024 at 11:38 PM Dragos Tatulea wrote:
> >
> > The following workflow triggers the crash referenced below:
> >
> > 1) vhost_vdpa_unsetup_vq_irq() unregisters the irq bypass producer
> >
0
R10: R11: 0246 R12: 562dfe134360
R13: 562dfe134d20 R14: R15: 7f9df801e190
Signed-off-by: Dragos Tatulea
---
drivers/vhost/vdpa.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/vhost/vdpa.c b/drivers/vhost/vdpa.c
index
> + if (add_config->mask & (1 << VDPA_ATTR_DEV_NET_CFG_MACADDR)) {
> + pfmdev = pci_get_drvdata(pci_physfn(mdev->pdev));
> + err = mlx5_mpfs_add_mac(pfmdev, config->mac);
> + if (0 == err)
if (!err) would be nicer. Not a deal bre
On Mon, 2024-07-22 at 15:48 +0800, Jason Wang wrote:
> On Mon, Jul 22, 2024 at 9:06 AM Cindy Lu wrote:
> >
> > Add the function to support setting the MAC address.
> > For vdpa/mlx5, the function will use mlx5_mpfs_add_mac
> > to set the mac address
> >
> > Tested in ConnectX-6 Dx device
> >
>
Hi Zhu Yanjun,
On Mon, 2024-07-08 at 18:22 +0200, Zhu Yanjun wrote:
> 在 2024/6/17 17:07, Dragos Tatulea 写道:
> > Currently, hardware VQs are created right when the vdpa device gets into
> > DRIVER_OK state. That is easier because most of the VQ state is known by
> > then
resources.
To solve this problem, avoid resetting hardware VQs if the VQs are still
in a blank state.
Reviewed-by: Cosmin Ratiu
Acked-by: Eugenio Pérez
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 30 +++---
1 file changed, 27 insertions(+), 3
a VQ instead of tearing it down when the number of
VQ pairs decreases. But only if the resume operation is supported.
Reviewed-by: Cosmin Ratiu
Acked-by: Eugenio Pérez
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 14 +++---
1 file changed, 11 insertions(+), 3
VQ indices in the range [cur_num_qps, max_vqs) represent queues that
have not yet been activated. .set_vq_ready should not activate these
VQs.
Reviewed-by: Cosmin Ratiu
Acked-by: Eugenio Pérez
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 3 +++
1 file changed, 3
Start using the suspend/resume_vq() error return codes previously added.
Reviewed-by: Cosmin Ratiu
Reviewed-by: Zhu Yanjun
Reviewed-by: Eugenio Pérez
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 12
1 file changed, 8 insertions(+), 4 deletions(-)
diff
these cases, the TIR and RQT need to be
re-created.
Add a needs_teardown configuration variable and set it when detecting
the above scenarios. On next DRIVER_OK, the resources will be torn down
first.
Reviewed-by: Cosmin Ratiu
Acked-by: Eugenio Pérez
Signed-off-by: Dragos Tatulea
---
drivers
er in setup_vq() so that
resume_vq() doesn't return early.
Reviewed-by: Cosmin Ratiu
Acked-by: Eugenio Pérez
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 18 ++
1 file changed, 6 insertions(+), 12 deletions(-)
diff --git a/drivers/vdpa/mlx5/net/mlx5
DRIVER_OK.
On a 64 CPU, 256 GB VM with 1 vDPA device of 16 VQps, the full VQ
resource creation + resume time was ~370ms. Now it's down to 60 ms
(only VQ config and resume). The measurements were done on a ConnectX6DX
based vDPA device.
Signed-off-by: Dragos Tatulea
Reviewed-by: Cosmin
Until now resume_vq() was used only for the suspend/resume scenario.
This change also allows calling resume_vq() to bring it from Init to
Ready state (VQ initialization).
Reviewed-by: Cosmin Ratiu
Acked-by: Eugenio Pérez
Signed-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 24
-off-by: Dragos Tatulea
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 77 +++
1 file changed, 54 insertions(+), 23 deletions(-)
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c
b/drivers/vdpa/mlx5/net/mlx5_vnet.c
index adcc4d63cf83..8ab5cf1bbc43 100644
--- a/drivers
1 - 100 of 318 matches
Mail list logo
