On Wed, 7 Feb 2024, Jian-Hong Pan wrote: > The original __pci_enable_link_state() configs the links directly without: > * Check the L1 substates features which are supported, or not > * Calculate & program related parameters for L1.2, such as T_POWER_ON, > Common_Mode_Restore_Time, and LTR_L1.2_THRESHOLD > > This leads some supported L1 PM substates of the link between VMD remapped > PCIe Root Port and NVMe get wrong configs when a caller tries to enabled > them. > > Here is a failed example on ASUS B1400CEAE with enabled VMD: > > Capabilities: [900 v1] L1 PM Substates > L1SubCap: PCI-PM_L1.2+ PCI-PM_L1.1- ASPM_L1.2+ ASPM_L1.1- L1_PM_Substates+ > PortCommonModeRestoreTime=32us PortTPowerOnTime=10us > L1SubCtl1: PCI-PM_L1.2- PCI-PM_L1.1- ASPM_L1.2+ ASPM_L1.1- > T_CommonMode=0us LTR1.2_Threshold=0ns > L1SubCtl2: T_PwrOn=10us > > This patch initializes the link's L1 PM substates to get the supported > features and programs relating paramters, if some of them are going to be > enabled in __pci_enable_link_state(). Then, enables the L1 PM substates if > the caller intends to enable them and they are supported. > > Link: https://bugzilla.kernel.org/show_bug.cgi?id=218394 > Signed-off-by: Jian-Hong Pan <jhp@xxxxxxxxxxxxx> > --- > v2: > - Prepare the PCIe LTR parameters before enable L1 Substates > > v3: > - Only enable supported features for the L1 Substates part > > drivers/pci/pcie/aspm.c | 12 +++++++----- > 1 file changed, 7 insertions(+), 5 deletions(-) > > diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c > index a39d2ee744cb..c866971cae70 100644 > --- a/drivers/pci/pcie/aspm.c > +++ b/drivers/pci/pcie/aspm.c > @@ -1389,14 +1389,16 @@ static int __pci_enable_link_state(struct pci_dev *pdev, int state, bool locked) > link->aspm_default |= ASPM_STATE_L0S; > if (state & PCIE_LINK_STATE_L1) > link->aspm_default |= ASPM_STATE_L1; > - /* L1 PM substates require L1 */ > - if (state & PCIE_LINK_STATE_L1_1) > + if (state & ASPM_STATE_L1_2_MASK) > + aspm_l1ss_init(link); Your commit message didn't explain why you need to add this call here, why isn't the existing call to aspm_l1ss_init() enough to initialize what is needed? > + /* L1 PM substates require L1 and should be in supported list */ > + if (state & link->aspm_support & PCIE_LINK_STATE_L1_1) > link->aspm_default |= ASPM_STATE_L1_1 | ASPM_STATE_L1; > - if (state & PCIE_LINK_STATE_L1_2) > + if (state & link->aspm_support & PCIE_LINK_STATE_L1_2) > link->aspm_default |= ASPM_STATE_L1_2 | ASPM_STATE_L1; > - if (state & PCIE_LINK_STATE_L1_1_PCIPM) > + if (state & link->aspm_support & PCIE_LINK_STATE_L1_1_PCIPM) > link->aspm_default |= ASPM_STATE_L1_1_PCIPM | ASPM_STATE_L1; > - if (state & PCIE_LINK_STATE_L1_2_PCIPM) > + if (state & link->aspm_support & PCIE_LINK_STATE_L1_2_PCIPM) > link->aspm_default |= ASPM_STATE_L1_2_PCIPM | ASPM_STATE_L1; It would be simpler and cleaner to just AND once with link->aspm_support to clear unsupported bits from state rather than doing it on every line like that. -- i.
