On Wed, Mar 26, 2025 at 05:22:50PM +0200, Ilpo Järvinen wrote:
> On Thu, 20 Mar 2025, Michał Winiarski wrote:
>
> > Drivers could leverage the fact that the VF BAR MMIO reservation is
> > created for total number of VFs supported by the device by resizing the
> > BAR to larger size when smaller number of VFs is enabled.
> >
> > Add a pci_iov_vf_bar_set_size() function to control the size and a
> > pci_iov_vf_bar_get_sizes() helper to get the VF BAR sizes that will
> > allow up to num_vfs to be successfully enabled with the current
> > underlying reservation size.
> >
> > Signed-off-by: Michał Winiarski <michal.winiar...@intel.com>
> > ---
> > drivers/pci/iov.c | 78 +++++++++++++++++++++++++++++++++++++++++++++
> > include/linux/pci.h | 6 ++++
> > 2 files changed, 84 insertions(+)
> >
> > diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
> > index 861273ad9a580..751eef232685c 100644
> > --- a/drivers/pci/iov.c
> > +++ b/drivers/pci/iov.c
> > @@ -1291,3 +1291,81 @@ int pci_sriov_configure_simple(struct pci_dev *dev,
> > int nr_virtfn)
> > return nr_virtfn;
> > }
> > EXPORT_SYMBOL_GPL(pci_sriov_configure_simple);
> > +
> > +/**
> > + * pci_iov_vf_bar_set_size - set a new size for a VF BAR
> > + * @dev: the PCI device
> > + * @resno: the resource number
> > + * @size: new size as defined in the spec (0=1MB, 31=128TB)
> > + *
> > + * Set the new size of a VF BAR that supports VF resizable BAR capability.
> > + * Unlike pci_resize_resource(), this does not cause the resource that
> > + * reserves the MMIO space (originally up to total_VFs) to be resized,
> > which
> > + * means that following calls to pci_enable_sriov() can fail if the
> > resources
> > + * no longer fit.
> > + *
> > + * Returns 0 on success, or negative on failure.
>
> Return: is the correct kernel doc style.
Yeah, I just blindly followed the style from surrounding docs. I'll
change it here.
>
> > + */
> > +int pci_iov_vf_bar_set_size(struct pci_dev *dev, int resno, int size)
> > +{
> > + int ret;
> > + u32 sizes;
> > +
> > + if (!pci_resource_is_iov(resno))
> > + return -EINVAL;
> > +
> > + if (pci_iov_is_memory_decoding_enabled(dev))
> > + return -EBUSY;
> > +
> > + sizes = pci_rebar_get_possible_sizes(dev, resno);
> > + if (!sizes)
> > + return -ENOTSUPP;
> > +
> > + if (!(sizes & BIT(size)))
> > + return -EINVAL;
> > +
> > + ret = pci_rebar_set_size(dev, resno, size);
> > + if (ret)
> > + return ret;
> > +
> > + pci_iov_resource_set_size(dev, resno, pci_rebar_size_to_bytes(size));
> > +
> > + return 0;
> > +}
> > +EXPORT_SYMBOL_GPL(pci_iov_vf_bar_set_size);
> > +
> > +/**
> > + * pci_iov_vf_bar_get_sizes - get VF BAR sizes allowing to create up to
> > num_vfs
> > + * @dev: the PCI device
> > + * @resno: the resource number
> > + * @num_vfs: number of VFs
> > + *
> > + * Get the sizes of a VF resizable BAR that can be accommodated within the
> > + * resource that reserves the MMIO space if num_vfs are enabled.
>
> I'd rephrase to:
>
> Get the sizes of a VF resizable BAR that can be accommodate @num_vfs
> within the currently assigned size of the resource @resno.
Ok.
>
> > + *
> > + * Returns 0 if BAR isn't resizable, otherwise returns a bitmask in format
>
> Return:
>
> a bitmask of sizes
Ok.
>
> > + * defined in the spec (bit 0=1MB, bit 31=128TB).
> > + */
> > +u32 pci_iov_vf_bar_get_sizes(struct pci_dev *dev, int resno, int num_vfs)
> > +{
> > + resource_size_t size;
> > + u32 sizes;
> > + int i;
> > +
> > + sizes = pci_rebar_get_possible_sizes(dev, resno);
> > + if (!sizes)
> > + return 0;
> > +
> > + while (sizes > 0) {
> > + i = __fls(sizes);
> > + size = pci_rebar_size_to_bytes(i);
> > +
> > + if (size * num_vfs <= pci_resource_len(dev, resno))
> > + break;
> > +
> > + sizes &= ~BIT(i);
> > + }
>
> Couldn't this be handled without a loop:
>
> bar_sizes = (round_up(pci_resource_len(dev, resno) / num_vfs) - 1) >>
> ilog2(SZ_1M);
>
> sizes &= bar_sizes;
>
> (Just to given an idea, I wrote this into the email so it might contain
> some off-by-one errors or like).
I think the division will need to be wrapped with something like do_div
(because IIUC, we have 32bit architectures where resource_size_t is
u64).
But yeah, we can drop the loop, turning it into something like this:
vf_len = pci_resource_len(dev, resno);
do_div(vf_len, num_vfs);
sizes = (roundup_pow_of_two(vf_len + 1) - 1) >> ilog2(SZ_1M);
Thanks,
-Michał
>
> > +
> > + return sizes;
> > +}
> > +EXPORT_SYMBOL_GPL(pci_iov_vf_bar_get_sizes);
> > diff --git a/include/linux/pci.h b/include/linux/pci.h
> > index 0e8e3fd77e967..c8708f3749757 100644
> > --- a/include/linux/pci.h
> > +++ b/include/linux/pci.h
> > @@ -2389,6 +2389,8 @@ int pci_sriov_set_totalvfs(struct pci_dev *dev, u16
> > numvfs);
> > int pci_sriov_get_totalvfs(struct pci_dev *dev);
> > int pci_sriov_configure_simple(struct pci_dev *dev, int nr_virtfn);
> > resource_size_t pci_iov_resource_size(struct pci_dev *dev, int resno);
> > +int pci_iov_vf_bar_set_size(struct pci_dev *dev, int resno, int size);
> > +u32 pci_iov_vf_bar_get_sizes(struct pci_dev *dev, int resno, int num_vfs);
> > void pci_vf_drivers_autoprobe(struct pci_dev *dev, bool probe);
> >
> > /* Arch may override these (weak) */
> > @@ -2441,6 +2443,10 @@ static inline int pci_sriov_get_totalvfs(struct
> > pci_dev *dev)
> > #define pci_sriov_configure_simple NULL
> > static inline resource_size_t pci_iov_resource_size(struct pci_dev *dev,
> > int resno)
> > { return 0; }
> > +static inline int pci_iov_vf_bar_set_size(struct pci_dev *dev, int resno,
> > int size)
> > +{ return -ENODEV; }
> > +static inline u32 pci_iov_vf_bar_get_sizes(struct pci_dev *dev, int resno,
> > int num_vfs)
> > +{ return 0; }
> > static inline void pci_vf_drivers_autoprobe(struct pci_dev *dev, bool
> > probe) { }
> > #endif
> >
> >
>
> --
> i.
