PCI驱动的suspend/resume方法，在何时调用？

zx_wing

这次好像猜错了

>I would say my friend is working on PM, I am on behalf of him to
>confirm before question.
>He said, in linus's opinion, STD should be called as
>"snapshot-and-restore" and should not be considered as suspend/resume.

People always have different views from different angles. But if you refer to ACPI spec, sleep sequence is always suggested similar.

>So you said "linux treats S1/S3 exactly same" should be for D-sates.
>STR in linux and S1 in linux are different, is he right?

Not only D-states, including cpu offline, etc. I mean the whole sleep flow is same except the last value written to platform is different.

zx_wing

更多补充

>Oh, sorry, I mis-understanding what he wants.
>His question is "for CPU, does linux treat S1/S3 same"?
>I guess the answer is no

The answer is yes for APs, and no for BSPs. All APs are put offline for both S1/S3. But for BSP, S1 doesn't need to save/restore BSP context since once resume, the instruction right after sleep point is resumed. However for S3, because all cpus are reset, one special real-mode wakeup code is required to bring BSP back to original context.

albcamus

原帖由 zx_wing 于 2008-3-21 17:36 发表
更多补充

多谢多谢

这么精彩的回答，我要加个精华(不过加到自己头上了，咳咳 )

albcamus

http://www.advogato.org/article/913.html

albcamus

事情好象改变了^_^
http://kerneltrap.org/Linux/Separating_Suspend_and_Hibernation

fishpxie

继续问一下,那么S3 resume之后，PCI空间的值是如何暂存的？
又是谁来做这个工作的？有没有一个标准的做法？

albcamus

原帖由 fishpxie 于 2009-6-25 15:15 发表
继续问一下,那么S3 resume之后，PCI空间的值是如何暂存的？
又是谁来做这个工作的？有没有一个标准的做法？

driver自己的suspend方法，调用pci_save_state，来把配置空间（包括PCI、PCI-X、PCI-E）存到pci_dev里头去； resueme方法调用pci_restore_state，从pci_dev里头保存的值来恢复config space的状态。

accessory

好贴顶下。。。

羡慕下WING老兄有内部专家可以问。。。