From: António Silva
I am trying to activate a PCIe link between two MPC8544 processor's on a custom board. One processor is configured as Root Complex (cfg_host_agt[0:2] = '111') and the other processor as endpoint (cfg_host_agt[0:2] = '101').
I've done this with many different configurations.
Only PCIe1 is active in both processors (cfg_IO_ports[0:2] = '010')
In u-boot I set the following flags in both processors:
#define CONFIG_PCI /* Enable PCI/PCIE */ #define CONFIG_PCIE1 /* PCIE controler 1 (slot 1) */ //MJ #define CONFIG_FSL_PCI_INIT /* Use common FSL init code */
In the first processor (configured as RC) I get the following output:
- pci_init_board: devdisr= 700 0008, io_sel=2, host_agent=7
Fine.
PCIE1 connected to Slot2 as Root Complex (base address e000a000)....PCIE link error. Skipping scan.LTSSM=0x00
As you note below, LTSSM zero is the issue.
PCIE1 on bus 00 - 00*In the second processor (configured as EP) I get the following output:
- pci_init_board: devdisr=7000008, io_sel=2, host_agent=5
** PCIE1 connected to Slot2 as End Point (base address e000a000) Scanning PCI bus 00 PCI Scan: Found Bus 0, Device 1, Function 0 00 01 1957 0033 0b20 00
The End Point should not scan the bus. Is this old u-boot code?
...
From the output of the RC processor:
....PCIE link error. Skipping scan.LTSSM=0x00
from table Table 18-109, the controller doesn't detect any activity on the bus.
It is worse than that, a value of zero indicates the interface is either not wired, in reset, or has no clocks. A value of 8 means the end point's receiver's are detected, but has not trained yet. And 16 is good.
Out of reset with no software support, it should go to 8 and then 16 when the end point responds.
I proceed trying to check in the HW for any activity in the PCIe bus. I don't see any activity in the bus from any of the processors. Both controller are getting the clock, and apart the PCIe detection both processor are working an proceed the booting stage.
Is there any configuration missing in u-boot for the PCIe configuration?
There is a controller reset for an errata workaround (see the setbits 0x08000000 to pdb_stat). But if that was set, I'd expect different symptoms.
Does anyone has any clue/hint to aid in the debugging process of the PCIe interface either for the SW and HW side?
Thanks in advance,
Antonio