On Sat, Mar 01, 2025 at 11:37:14AM +0000, Marc Zyngier wrote:
> On Tue, 11 Feb 2025 19:21:57 +0000,
> Frank Li <Frank.Li@xxxxxxx> wrote:
> >
> > Some platform devices create child devices dynamically and require the
> > parent device's msi-map to map device IDs to actual sideband information.
> >
> > A typical use case is using ITS as a PCIe Endpoint Controller(EPC)'s
> > doorbell function, where PCI hosts send TLP memory writes to the EP
> > controller. The EP controller converts these writes to AXI transactions
> > and appends platform-specific sideband information. See below figure.
> >
> > ┌────────────────────────────────┐
> > │ │
> > │ PCI Endpoint Controller │
> > │ │
> > │ ┌─────┐ ┌─────┐ ┌─────┐ │
> > PCI Bus │ │ │ │ │ │ │ │
> > ─────────► │ │Func1│ │Func2│ ... │Func │ │
> > TLP Memory │ │ │ │ │ │<n> │ │
> > Write Push │ │ │ │ │ │ │ │
> > DoorBell │ └─┬─┬─┘ └──┬──┘ └──┬──┘ │
> > │ │ │ │ │ │
> > └────┼─┼────────┼───────────┼────┘
> > sideband │ │ Address│ │
> > information ▼ ▼ /Data ▼ ▼
> > ┌────────────────────────┐
> > │ MSI Controller │
> > └────────────────────────┘
> >
>
> Please stop using these figures in commit messages. I don't think they
> help, and they are not in consistent with the way the commit messages
> are managed.
Okay
>
> > EPC's DTS will provide such information by msi-map and msi-mask. A
> > simplified dts as
> >
> > pcie-ep@10000000 {
> > ...
> > msi-map = <0 &its 0xc 8>;
> > ^^^ 0xc is implement defined sideband information,
> > which append to AXI write transaction.
> > ^ 0 is function index.
>
> What does this sideband field represent?
ARM ITS use term "streamid" for this sideband field, which indicate which
MSI consumer write to address/data pair on bus. Such as PCI1 or PCI2.
> How is the ITS driver
> supposed to use that data?
ITS use a as devid, or info->scratchpad[0].ul = dev_id;
msi-map actually given a start dev_id (it is 0xc in example) for fuction0.
function1 will use dev_id + 1 ...
> Is it the full devid as presented to the
> ITS?
Yes,
> Something combined with the "function index"? Is the "function
> index" a full RID, as defined in the documentation?
Not a full RID. RID is related with host PCIe's topology. The EP function's
RID may 1:00:01 at PC1, 3:00:01 at the another PC.
So Endpoint driver can't use RID directly. It should related function and
virtual function number only.
PCI define 8 physicla funciton, and 64000 virutal function. Define device
id as vfunc[31:3], pfunc[2:0] as msi-map's input. DTS provide information
how map it to real streamid.
>
> Also, msi-map is so far reserved to a PCIe RC, not this sort of wonky
> contraption. This needs to be documented.
Okay, I can update document.
Frank
>
> >
> > msi-mask = <0x7>
> > }
> >
> > Check msi-map if msi-parent missed to keep compatility with existed system.
> >
> > Signed-off-by: Frank Li <Frank.Li@xxxxxxx>
> > ---
> > change from v14 to v15
> > - none
> >
> > change from v13 to v14
> > new patch
> > ---
> > drivers/irqchip/irq-gic-v3-its-msi-parent.c | 8 ++++++++
> > 1 file changed, 8 insertions(+)
> >
> > diff --git a/drivers/irqchip/irq-gic-v3-its-msi-parent.c b/drivers/irqchip/irq-gic-v3-its-msi-parent.c
> > index e150365fbe892..6c7389bb84a4a 100644
> > --- a/drivers/irqchip/irq-gic-v3-its-msi-parent.c
> > +++ b/drivers/irqchip/irq-gic-v3-its-msi-parent.c
> > @@ -118,6 +118,14 @@ static int of_pmsi_get_dev_id(struct irq_domain *domain, struct device *dev,
> > index++;
> > } while (!ret);
> >
> > + if (ret) {
> > + struct device_node *np = NULL;
> > +
> > + ret = of_map_id(dev->of_node, dev->id, "msi-map", "msi-map-mask", &np, dev_id);
> > + if (np)
> > + of_node_put(np);
> > + }
> > +
> > return ret;
> > }
> >
> >
>
> Thanks,
>
> M.
>
> --
> Without deviation from the norm, progress is not possible.
