Matt DeVillier has uploaded this change for review. ( https://review.coreboot.org/25325
Change subject: soc/intel/broadwell: If L1 Sub state is disabled, do not set capability ......................................................................
soc/intel/broadwell: If L1 Sub state is disabled, do not set capability
I finally found out why disabling the L1 sub state option did not prevent some NVMe drives from locking up in L1 substate. I expect that the disabled L1 substate initialization that coreboot does is negated because Linux might itself configure it if it finds the capability enabled on the PCIe root port.
Removing the capability from the PCIe root port when L1 sub states are disabled in the configuration should fix the problem.
This was not tested because it's a difficult issue to reproduce and I do not have the problematic hardware that caused it anymore.
Change-Id: I293a650db307e77cee024a43fbfc81e1d8c86265 Signed-off-by: Youness Alaoui kakaroto@kakaroto.homelinux.net --- M src/soc/intel/broadwell/pcie.c 1 file changed, 4 insertions(+), 1 deletion(-)
git pull ssh://review.coreboot.org:29418/coreboot refs/changes/25/25325/1
diff --git a/src/soc/intel/broadwell/pcie.c b/src/soc/intel/broadwell/pcie.c index aff66a3..cf25749 100644 --- a/src/soc/intel/broadwell/pcie.c +++ b/src/soc/intel/broadwell/pcie.c @@ -558,7 +558,10 @@ (1 << 29) | 0x10001);
/* Set L1 Sub-State Cap ID to 1Eh and Next Cap Pointer to None. */ - pci_update_config32(dev, 0x200, ~0xffff, 0x001e); + if (IS_ENABLED(CONFIG_PCIEXP_L1_SUB_STATE)) + pci_update_config32(dev, 0x200, ~0xfffff, 0x001e); + else + pci_update_config32(dev, 0x200, ~0xfffff, 0);
pci_update_config32(dev, 0x320, ~(3 << 20) & ~(7 << 6), (1 << 20) | (3 << 6));