diff options
author | Mika Westerberg <mika.westerberg@linux.intel.com> | 2018-04-20 15:22:02 +0300 |
---|---|---|
committer | Rafael J. Wysocki <rafael.j.wysocki@intel.com> | 2018-04-23 08:57:35 +0200 |
commit | ae860a19f37c686e7c5816e96640168b7174a096 (patch) | |
tree | 92abb123df7bb49a9b805fd044a52a31a963ff9a /drivers/pci/pci-driver.c | |
parent | 6d08b06e67cd117f6992c46611dfb4ce267cd71e (diff) | |
download | linux-stable-ae860a19f37c686e7c5816e96640168b7174a096.tar.gz linux-stable-ae860a19f37c686e7c5816e96640168b7174a096.tar.bz2 linux-stable-ae860a19f37c686e7c5816e96640168b7174a096.zip |
PCI / PM: Do not clear state_saved in pci_pm_freeze() when smart suspend is set
If a driver uses DPM_FLAG_SMART_SUSPEND and the device is already
runtime suspended when hibernate is started PCI core skips runtime
resuming the device but still clears pci_dev->state_saved. After the
hibernation image is written pci_pm_thaw_noirq() makes sure subsequent
thaw phases for the device are also skipped leaving it runtime suspended
with pci_dev->state_saved == false.
When the device is eventually runtime resumed pci_pm_runtime_resume()
restores config space by calling pci_restore_standard_config(), however
because pci_dev->state_saved == false pci_restore_state() never actually
restores the config space leaving the device in a state that is not what
the driver might expect.
For example here is what happens for intel-lpss I2C devices once the
hibernation snapshot is taken:
intel-lpss 0000:00:15.0: power state changed by ACPI to D0
intel-lpss 0000:00:1e.0: power state changed by ACPI to D3cold
video LNXVIDEO:00: Restoring backlight state
PM: hibernation exit
i2c_designware i2c_designware.1: Unknown Synopsys component type: 0xffffffff
i2c_designware i2c_designware.0: Unknown Synopsys component type: 0xffffffff
i2c_designware i2c_designware.1: timeout in disabling adapter
i2c_designware i2c_designware.0: timeout in disabling adapter
Since PCI config space is not restored the device is still in D3hot
making MMIO register reads return 0xffffffff.
Fix this by clearing pci_dev->state_saved only if we actually end up
runtime resuming the device.
Fixes: c4b65157aeef (PCI / PM: Take SMART_SUSPEND driver flag into account)
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Cc: 4.15+ <stable@vger.kernel.org> # 4.15+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Diffstat (limited to 'drivers/pci/pci-driver.c')
-rw-r--r-- | drivers/pci/pci-driver.c | 5 |
1 files changed, 3 insertions, 2 deletions
diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c index 6ace47099fc5..b9a131137e64 100644 --- a/drivers/pci/pci-driver.c +++ b/drivers/pci/pci-driver.c @@ -958,10 +958,11 @@ static int pci_pm_freeze(struct device *dev) * devices should not be touched during freeze/thaw transitions, * however. */ - if (!dev_pm_test_driver_flags(dev, DPM_FLAG_SMART_SUSPEND)) + if (!dev_pm_smart_suspend_and_suspended(dev)) { pm_runtime_resume(dev); + pci_dev->state_saved = false; + } - pci_dev->state_saved = false; if (pm->freeze) { int error; |