On Thu, 2023-09-28 at 20:59 +0300, Leon Romanovsky wrote:
> On Thu, Sep 28, 2023 at 03:55:47PM +0200, Niklas Schnelle wrote:
> > Since commit 06cd555f73ca ("net/mlx5: split mlx5_cmd_init() to probe and
> > reload routines") mlx5_cmd_init() is called in mlx5_mdev_init() which is
> > called in probe_one() before mlx5_pci_init(). This is a problem because
> > mlx5_pci_init() is where the DMA and coherent mask is set but
> > mlx5_cmd_init() already does a dma_alloc_coherent(). Thus a DMA
> > allocation is done during probe before the correct mask is set. This
> > causes probe to fail initialization of the cmdif SW structs on s390x
> > after that is converted to the common dma-iommu code. This is because on
> > s390x DMA addresses below 4 GiB are reserved on current machines and
> > unlike the old s390x specific DMA API implementation common code
> > enforces DMA masks. Fix this by switching the order of the
> > mlx5_mdev_init() and mlx5_pci_init() in probe_one().
> >
> > Link: https://lore.kernel.org/linux-iommu/cfc9e9128ed5571d2e36421e347301057662a09e.camel@xxxxxxxxxxxxx/
> > Fixes: 06cd555f73ca ("net/mlx5: split mlx5_cmd_init() to probe and reload routines")
> > Signed-off-by: Niklas Schnelle <schnelle@xxxxxxxxxxxxx>
> > ---
> > Note: I ran into this while testing the linked series for converting
> > s390x to use dma-iommu. The existing s390x specific DMA API
> > implementation doesn't respect DMA masks and is thus not affected
> > despite of course also only supporting DMA addresses above 4 GiB.
> > That said ConnectX VFs are the primary users of native PCI on s390x and
> > we'd really like to get the DMA API conversion into v6.7 so this has
> > high priority for us.
> > ---
> > drivers/net/ethernet/mellanox/mlx5/core/main.c | 12 ++++++------
> > 1 file changed, 6 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
> > index 15561965d2af..06744dedd928 100644
> > --- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
> > +++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
> > @@ -1908,10 +1908,6 @@ static int probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
> > goto adev_init_err;
> > }
> >
> > - err = mlx5_mdev_init(dev, prof_sel);
> > - if (err)
> > - goto mdev_init_err;
> > -
> > err = mlx5_pci_init(dev, pdev, id);
> > if (err) {
> > mlx5_core_err(dev, "mlx5_pci_init failed with error code %d\n",
> > @@ -1919,6 +1915,10 @@ static int probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
> > goto pci_init_err;
> > }
> >
> > + err = mlx5_mdev_init(dev, prof_sel);
> > + if (err)
> > + goto mdev_init_err;
> > +
>
> I had something different in mind as I'm worry that call to pci_enable_device()
> in mlx5_pci_init() before we finished FW command interface initialization is a bit
> premature.
>
> What about the following patch?
>
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
> index 15561965d2af..31f1d701116a 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
> @@ -905,12 +905,6 @@ static int mlx5_pci_init(struct mlx5_core_dev *dev, struct pci_dev *pdev,
>
> pci_set_master(pdev);
>
> - err = set_dma_caps(pdev);
> - if (err) {
> - mlx5_core_err(dev, "Failed setting DMA capabilities mask, aborting\n");
> - goto err_clr_master;
> - }
> -
> if (pci_enable_atomic_ops_to_root(pdev, PCI_EXP_DEVCAP2_ATOMIC_COMP32) &&
> pci_enable_atomic_ops_to_root(pdev, PCI_EXP_DEVCAP2_ATOMIC_COMP64) &&
> pci_enable_atomic_ops_to_root(pdev, PCI_EXP_DEVCAP2_ATOMIC_COMP128))
> @@ -1908,9 +1902,15 @@ static int probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
> goto adev_init_err;
> }
>
> + err = set_dma_caps(pdev);
> + if (err) {
> + mlx5_core_err(dev, "Failed setting DMA capabilities mask, aborting\n");
> + goto dma_cap_err;
> + }
> +
> err = mlx5_mdev_init(dev, prof_sel);
> if (err)
> - goto mdev_init_err;
> + goto dma_cap_err;
>
> err = mlx5_pci_init(dev, pdev, id);
> if (err) {
> @@ -1942,7 +1942,7 @@ static int probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
> mlx5_pci_close(dev);
> pci_init_err:
> mlx5_mdev_uninit(dev);
> -mdev_init_err:
> +dma_cap_err:
> mlx5_adev_idx_free(dev->priv.adev_idx);
> adev_init_err:
> mlx5_devlink_free(devlink);
>
> Thanks
The above works too. Maybe for consistency within probe_one() it would
then make sense to also rename set_dma_caps() to mlx5_dma_init()?
Thanks,
Niklas
