RE: [ACPI] pci device resume too early?

[Li, S. <sha...@in...>]

How about add one line in pci_default_resume just like this:

static void pci_default_resume(struct pci_dev *pci_dev)
{
+	pci_set_power_state(pci_dev, 0);
	/* restore the PCI config space */
	pci_restore_state(pci_dev, pci_dev->saved_config_space);
	/* if the device was enabled before suspend, reenable */


Thanks,
Shaohua
>-----Original Message-----
>From: acp...@li... [mailto:acpi-devel-
>ad...@li...] On Behalf Of ste...@gm...
>Sent: Thursday, July 29, 2004 1:54 AM
>To: acp...@li...
>Subject: [ACPI] pci device resume too early?
>
>Hello,
>I have problems with a few pci devices after resume. Hardware
environment
>is a
>acer Travelmate 803(lspci attached), I use an unpatched linux-2.6.8-rc2
>kernel.
>
>The problem is that directly after S3 resume, any attempt to write to
pci
>configuration registers of three of my devices(00:1d.7, 00:1f.5,
00:1f.6)
>locks up my system. After some testing, I found out that
pci_restore_state
>works nice if I place it into a function which is scheduled with
>schedule_delayed_work. I use my dummydriver module for testing.
>
>Using mdelay(something) instead of scheduling a work doesn't help, so
it's
>not
>the hardware which needs some time.
>Is it possible that the driver's resume functions are called too early?
Can
>you point me to the place in the acpi code which calls the resume
methods
>that I can do further testing? I couldn't find it.
>
>Thanks in advance,
>Stefan