| Message ID | 20220902145835.344302-3-helgaas@kernel.org (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
| Series | PCI/PM: Always disable PTM for all devices during | expand |
On 9/2/22 7:58 AM, Bjorn Helgaas wrote: > From: Bjorn Helgaas <bhelgaas@google.com> > > The suspend path may disable PTM before saving config state, which means > the PCI_PTM_CTRL_ENABLE bit in the saved state may be cleared even though > we want PTM to be enabled when resuming. If suspend is disabling PTM separately, why not enable it during the resume operation? Why club it with PTM state restoration? > > If "dev->ptm_enabled" is set, it means PTM should be enabled, so make sure > PCI_PTM_CTRL_ENABLE is set when restoring the PTM state. > > Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> > --- > drivers/pci/pcie/ptm.c | 8 ++++++++ > 1 file changed, 8 insertions(+) > > diff --git a/drivers/pci/pcie/ptm.c b/drivers/pci/pcie/ptm.c > index b6a417247ce3..3115601a85ef 100644 > --- a/drivers/pci/pcie/ptm.c > +++ b/drivers/pci/pcie/ptm.c > @@ -82,6 +82,14 @@ void pci_restore_ptm_state(struct pci_dev *dev) > return; > > cap = (u16 *)&save_state->cap.data[0]; > + > + /* > + * The suspend path may disable PTM before saving config state. > + * Make sure PCI_PTM_CTRL_ENABLE is set if PTM should be enabled. > + */ > + if (dev->ptm_enabled) > + *cap |= PCI_PTM_CTRL_ENABLE; > + > pci_write_config_word(dev, ptm + PCI_PTM_CTRL, *cap); > } >
On Fri, Sep 02, 2022 at 10:25:54AM -0700, Sathyanarayanan Kuppuswamy wrote: > On 9/2/22 7:58 AM, Bjorn Helgaas wrote: > > From: Bjorn Helgaas <bhelgaas@google.com> > > > > The suspend path may disable PTM before saving config state, which means > > the PCI_PTM_CTRL_ENABLE bit in the saved state may be cleared even though > > we want PTM to be enabled when resuming. > > If suspend is disabling PTM separately, why not enable it during the resume > operation? Why club it with PTM state restoration? The long answer is in my previous reply [1]. The short answer is that pci_enable_ptm() only works with Endpoints, so if we enable PTM in the resume path, we need to rework it to handle Root Ports and Switch Ports as well. [1] https://lore.kernel.org/r/20220902203848.GA370638@bhelgaas > > If "dev->ptm_enabled" is set, it means PTM should be enabled, so make sure > > PCI_PTM_CTRL_ENABLE is set when restoring the PTM state. > > > > Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> > > --- > > drivers/pci/pcie/ptm.c | 8 ++++++++ > > 1 file changed, 8 insertions(+) > > > > diff --git a/drivers/pci/pcie/ptm.c b/drivers/pci/pcie/ptm.c > > index b6a417247ce3..3115601a85ef 100644 > > --- a/drivers/pci/pcie/ptm.c > > +++ b/drivers/pci/pcie/ptm.c > > @@ -82,6 +82,14 @@ void pci_restore_ptm_state(struct pci_dev *dev) > > return; > > > > cap = (u16 *)&save_state->cap.data[0]; > > + > > + /* > > + * The suspend path may disable PTM before saving config state. > > + * Make sure PCI_PTM_CTRL_ENABLE is set if PTM should be enabled. > > + */ > > + if (dev->ptm_enabled) > > + *cap |= PCI_PTM_CTRL_ENABLE; > > + > > pci_write_config_word(dev, ptm + PCI_PTM_CTRL, *cap); > > } > > > > -- > Sathyanarayanan Kuppuswamy > Linux Kernel Developer
diff --git a/drivers/pci/pcie/ptm.c b/drivers/pci/pcie/ptm.c index b6a417247ce3..3115601a85ef 100644 --- a/drivers/pci/pcie/ptm.c +++ b/drivers/pci/pcie/ptm.c @@ -82,6 +82,14 @@ void pci_restore_ptm_state(struct pci_dev *dev) return; cap = (u16 *)&save_state->cap.data[0]; + + /* + * The suspend path may disable PTM before saving config state. + * Make sure PCI_PTM_CTRL_ENABLE is set if PTM should be enabled. + */ + if (dev->ptm_enabled) + *cap |= PCI_PTM_CTRL_ENABLE; + pci_write_config_word(dev, ptm + PCI_PTM_CTRL, *cap); }