On Wed, Nov 20, 2024 at 11:53:29AM -0500, Frank Li wrote:
> On Wed, Nov 20, 2024 at 04:57:33PM +0100, Niklas Cassel wrote:
> > If running pci_endpoint_test.c (host side) against a version of
> > pci-epf-test.c (EP side), we will not see any capabilities being set.
> >
> > For now, only add the CAP_HAS_ALIGN_ADDR capability.
> >
> > If the CAP_HAS_ALIGN_ADDR is set, that means that the EP side supports
> > reading/writing to an address without any alignment requirements.
> >
> > Thus, if CAP_HAS_ALIGN_ADDR is set, make sure that we do not add any
> > specific padding to the buffers that we allocate (which was only made
> > in order to get the buffers to satisfy certain alignment requirements).
> >
> > Signed-off-by: Niklas Cassel <cassel@xxxxxxxxxx>
> > ---
> > drivers/misc/pci_endpoint_test.c | 21 +++++++++++++++++++++
> > 1 file changed, 21 insertions(+)
> >
> > diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c
> > index 3aaaf47fa4ee..ab2b322410fb 100644
> > --- a/drivers/misc/pci_endpoint_test.c
> > +++ b/drivers/misc/pci_endpoint_test.c
> > @@ -69,6 +69,9 @@
> > #define PCI_ENDPOINT_TEST_FLAGS 0x2c
> > #define FLAG_USE_DMA BIT(0)
> >
> > +#define PCI_ENDPOINT_TEST_CAPS 0x30
> > +#define CAP_HAS_ALIGN_ADDR BIT(0)
> > +
> > #define PCI_DEVICE_ID_TI_AM654 0xb00c
> > #define PCI_DEVICE_ID_TI_J7200 0xb00f
> > #define PCI_DEVICE_ID_TI_AM64 0xb010
> > @@ -805,6 +808,22 @@ static const struct file_operations pci_endpoint_test_fops = {
> > .unlocked_ioctl = pci_endpoint_test_ioctl,
> > };
> >
> > +static void pci_endpoint_test_get_capabilities(struct pci_endpoint_test *test)
> > +{
> > + struct pci_dev *pdev = test->pdev;
> > + struct device *dev = &pdev->dev;
> > + u32 caps;
> > + bool has_align_addr;
> > +
> > + caps = pci_endpoint_test_readl(test, PCI_ENDPOINT_TEST_CAPS);
>
> I worry about if there are problem if EP have not such register. for
> example, if reg's space size is 64, but host try to read pos 68. I think it
> is original design problem, it should have one 'version' or 'size' in
> register list.
Hello Frank,
The test BAR is allocated using pci_epf_alloc_space(), which allocates the
backing memory using dma_alloc_coherent(), which will return zeroed memory
regardless of __GFP_ZERO was set or not.
This means that running a new version of pci-endpoint-test.c (host side)
with and old version of pci-epf-test.c (EP side) will not see any
capabilities being set (as intended), so this is backwards compatible.
And as you probably know, pci-epf-test will always allocate at least 128
bytes for the test BAR:
https://github.com/torvalds/linux/blob/v6.12/drivers/pci/endpoint/functions/pci-epf-test.c#L833
This patch uses:
#define PCI_ENDPOINT_TEST_CAPS 0x30
0x30 == 48, so we are currently only using 49 bytes.
So we should be good.
Kind regards,
Niklas
