> -----Original Message-----
> From: Wu, Hao <hao.wu@xxxxxxxxx>
> Sent: Monday, February 28, 2022 7:10 PM
> To: Zhang, Tianfei <tianfei.zhang@xxxxxxxxx>; trix@xxxxxxxxxx;
> mdf@xxxxxxxxxx; Xu, Yilun <yilun.xu@xxxxxxxxx>; linux-fpga@xxxxxxxxxxxxxxx;
> linux-doc@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> Cc: corbet@xxxxxxx
> Subject: RE: [PATCH v1] fpga: dfl: check feature type before parse irq info
>
> > Subject: [PATCH v1] fpga: dfl: check feature type before parse irq
> > info
> >
> > From: Tianfei Zhang <tianfei.zhang@xxxxxxxxx>
> >
> > The feature ID of "Port User Interrupt" and the "PMCI Subsystem" are
> > identical, 0x12, but one is for FME, other is for Port. It should
> > check the feature type While parsing the irq info in
> > parse_feature_irqs().
> >
> > Fixes: 8d021039cbb5 ("fpga: dfl: parse interrupt info for feature
> > devices on
> > enumeration")
>
> Actually this is not a real bug, as in original design, there is no overlap for FME
> and Port features. This is why you see features for Port doesn't start from 0. But
> anyway I am good with such extension.
I think this is a bug, I add some printk debug log in DFL driver, we observed that it was mistaken for
the "Port User Interrput" feature device while we are parsing the FME feature devices, so actually it
read out the PMCI MMIO space for "Port User Interrput".
Here is the log:
[90273.482859] parse_feature: DFH_TYPE_FIU
[90273.482861] parse_feature_fiu: id:0 fme
[90273.482864] create_feature_instance: feature id: 0xfe
[90273.482868] parse_feature: DFH_TYPE_PRIVATE
[90273.482870] create_feature_instance: feature id: 0x1
[90273.482872] parse_feature: DFH_TYPE_PRIVATE
[90273.482874] create_feature_instance: feature id: 0x7
...
[90273.482895] create_feature_instance: feature id: 0x12
[90273.482898] parse_feature_irqs, PORT_UINT_DFH: 0x3000000200001012
[90273.482898] parse_feature_irqs, PORT_UINT_CAP: 0xbaadbeefdeadbeef
[90273.482899] dfl-pci 0000:b1:00.0: feature: 0x12, irq_base: 2779, nr_irqs: 3823
[90273.482901] dfl-pci 0000:b1:00.0: Ignoring nvalid interrupt number in feature 0x12 6602 > 7
>
> > Link: https://lore.kernel.org/linux-
> >
> fpga/BN9PR11MB54833D7636348D62F931526CE33A9@xxxxxxxxxxxxxxxxx
> > prd11.prod.outlook.com/
> >
> > Signed-off-by: Tianfei Zhang <tianfei.zhang@xxxxxxxxx>
> > ---
> > Documentation/fpga/dfl.rst | 5 +++++
> > drivers/fpga/dfl.c | 38 ++++++++++++++++++++++----------------
> > 2 files changed, 27 insertions(+), 16 deletions(-)
> >
> > diff --git a/Documentation/fpga/dfl.rst b/Documentation/fpga/dfl.rst
> > index ef9eec71f6f3..9ce418da1876 100644
> > --- a/Documentation/fpga/dfl.rst
> > +++ b/Documentation/fpga/dfl.rst
> > @@ -502,6 +502,11 @@ Developer only needs to provide a sub feature
> > driver with matched feature id.
> > FME Partial Reconfiguration Sub Feature driver (see
> > drivers/fpga/dfl-fme-pr.c) could be a reference.
> >
> > +Individual DFL drivers are bound DFL devices based on Feature Type
> > +and
> > Feature ID.
> > +The definition of Feature Type and Feature ID can be found:
> > +
> > +https://github.com/OPAE/linux-dfl-feature-id/blob/master/dfl-feature-
> > +ids.rst
>
> Thanks for tracking ID allocations. could we also add some description that if
> user want to implement a new private feature, then they need to submit new ID
> application to https://github.com/OPAE/linux-dfl-feature-id, and add some
> README file to guide people for the application process?
That is a good point, I will add this for next version patch.
>
> Thanks
> Hao
>
> > +
> > Location of DFLs on a PCI Device
> > ================================
> > The original method for finding a DFL on a PCI device assumed the
> > start of the diff --git a/drivers/fpga/dfl.c b/drivers/fpga/dfl.c
> > index 599bb21d86af..6bff39ff21a0 100644
> > --- a/drivers/fpga/dfl.c
> > +++ b/drivers/fpga/dfl.c
> > @@ -940,9 +940,12 @@ static int parse_feature_irqs(struct
> > build_feature_devs_info *binfo, {
> > void __iomem *base = binfo->ioaddr + ofst;
> > unsigned int i, ibase, inr = 0;
> > + enum dfl_id_type type;
> > int virq;
> > u64 v;
> >
> > + type = feature_dev_id_type(binfo->feature_dev);
> > +
> > /*
> > * Ideally DFL framework should only read info from DFL header, but
> > * current version DFL only provides mmio resources information for
> > @@ -957,22 +960,25 @@ static int parse_feature_irqs(struct
> > build_feature_devs_info *binfo,
> > * code will be added. But in order to be compatible to old version
> > * DFL, the driver may still fall back to these quirks.
> > */
> > - switch (fid) {
> > - case PORT_FEATURE_ID_UINT:
> > - v = readq(base + PORT_UINT_CAP);
> > - ibase = FIELD_GET(PORT_UINT_CAP_FST_VECT, v);
> > - inr = FIELD_GET(PORT_UINT_CAP_INT_NUM, v);
> > - break;
> > - case PORT_FEATURE_ID_ERROR:
> > - v = readq(base + PORT_ERROR_CAP);
> > - ibase = FIELD_GET(PORT_ERROR_CAP_INT_VECT, v);
> > - inr = FIELD_GET(PORT_ERROR_CAP_SUPP_INT, v);
> > - break;
> > - case FME_FEATURE_ID_GLOBAL_ERR:
> > - v = readq(base + FME_ERROR_CAP);
> > - ibase = FIELD_GET(FME_ERROR_CAP_INT_VECT, v);
> > - inr = FIELD_GET(FME_ERROR_CAP_SUPP_INT, v);
> > - break;
> > + if (type == PORT_ID) {
> > + switch (fid) {
> > + case PORT_FEATURE_ID_UINT:
> > + v = readq(base + PORT_UINT_CAP);
> > + ibase = FIELD_GET(PORT_UINT_CAP_FST_VECT, v);
> > + inr = FIELD_GET(PORT_UINT_CAP_INT_NUM, v);
> > + break;
> > + case PORT_FEATURE_ID_ERROR:
> > + v = readq(base + PORT_ERROR_CAP);
> > + ibase = FIELD_GET(PORT_ERROR_CAP_INT_VECT, v);
> > + inr = FIELD_GET(PORT_ERROR_CAP_SUPP_INT, v);
> > + break;
> > + }
> > + } else if (type == FME_ID) {
> > + if (fid == FME_FEATURE_ID_GLOBAL_ERR) {
> > + v = readq(base + FME_ERROR_CAP);
> > + ibase = FIELD_GET(FME_ERROR_CAP_INT_VECT, v);
> > + inr = FIELD_GET(FME_ERROR_CAP_SUPP_INT, v);
> > + }
> > }
> >
> > if (!inr) {
> > --
> > 2.26.2
