hibernation fails from kexec'ed kernel

hibernation fails from kexec'ed kernel

[Jiri Slaby]

Hi,

when I try to hibernate and resume from a kexec'ed kernel, the resume
fails with:
PM: Image not found (code -22)

I dumped sig and orig_sig from swsusp_check and it looks like this:
53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2
53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2

So it looks like the image (or only the signature) did not land to the disk.

I use only the in-kernel hibernation, I do not even have the suspend
package installed. This is with 3.7.2 and happens this way for some time
already. When I boot the kernel normally, hib+resume works just fine.

Any ideas?

thanks,
-- 
js

Re: hibernation fails from kexec'ed kernel

[Jiri Slaby]

On 01/12/2013 04:47 PM, Jiri Slaby wrote:
> Hi,
> 
> when I try to hibernate and resume from a kexec'ed kernel, the resume
> fails with:
> PM: Image not found (code -22)
> 
> I dumped sig and orig_sig from swsusp_check and it looks like this:
> 53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2
> 53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2
> 
> So it looks like the image (or only the signature) did not land to the disk.

Note that I see "PM: S|" and the error returned from mark_swapfiles is 0.

> I use only the in-kernel hibernation, I do not even have the suspend
> package installed. This is with 3.7.2 and happens this way for some time
> already. When I boot the kernel normally, hib+resume works just fine.
> 
> Any ideas?
> 
> thanks,
-- 
js

Re: hibernation fails from kexec'ed kernel

[Jiri Slaby]

On 01/12/2013 09:06 PM, Rafael J. Wysocki wrote:
> On Saturday, January 12, 2013 06:26:27 PM Jiri Slaby wrote:
>> On 01/12/2013 04:47 PM, Jiri Slaby wrote:
>>> Hi,
>>>
>>> when I try to hibernate and resume from a kexec'ed kernel, the resume
>>> fails with:
>>> PM: Image not found (code -22)
>>>
>>> I dumped sig and orig_sig from swsusp_check and it looks like this:
>>> 53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2
>>> 53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2
>>>
>>> So it looks like the image (or only the signature) did not land to the disk.
>>
>> Note that I see "PM: S|" and the error returned from mark_swapfiles is 0.
>>
>>> I use only the in-kernel hibernation, I do not even have the suspend
>>> package installed. This is with 3.7.2 and happens this way for some time
>>> already. When I boot the kernel normally, hib+resume works just fine.
>>>
>>> Any ideas?
> 
> Quite frankly, I'm not really surprised, because this is the first time I know
> of anybody has tested hibernate and resume from a kexec'ed kernel.

It's quite usual and natural, that people kexec to a new kernel, I
think. I used to do that for a long time, but it stopped workign at some
point.

> I have no idea why exactly it doesn't work, though.

I'm trying to debug that ATM...

> Did you use s2disk or the internal hibernation?

As I write above, I don't have suspend package installed. This is a pure
in-kernel implementation.

-- 
js

Re: hibernation fails from kexec'ed kernel

[Rafael J. Wysocki]

On Saturday, January 12, 2013 06:26:27 PM Jiri Slaby wrote:
> On 01/12/2013 04:47 PM, Jiri Slaby wrote:
> > Hi,
> > 
> > when I try to hibernate and resume from a kexec'ed kernel, the resume
> > fails with:
> > PM: Image not found (code -22)
> > 
> > I dumped sig and orig_sig from swsusp_check and it looks like this:
> > 53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2
> > 53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2
> > 
> > So it looks like the image (or only the signature) did not land to the disk.
> 
> Note that I see "PM: S|" and the error returned from mark_swapfiles is 0.
> 
> > I use only the in-kernel hibernation, I do not even have the suspend
> > package installed. This is with 3.7.2 and happens this way for some time
> > already. When I boot the kernel normally, hib+resume works just fine.
> > 
> > Any ideas?

Quite frankly, I'm not really surprised, because this is the first time I know
of anybody has tested hibernate and resume from a kexec'ed kernel.

I have no idea why exactly it doesn't work, though.

Did you use s2disk or the internal hibernation?

Rafael


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.

Re: hibernation fails from kexec'ed kernel

[Rafael J. Wysocki]

On Saturday, January 12, 2013 09:06:28 PM Jiri Slaby wrote:
> On 01/12/2013 09:06 PM, Rafael J. Wysocki wrote:
> > On Saturday, January 12, 2013 06:26:27 PM Jiri Slaby wrote:
> >> On 01/12/2013 04:47 PM, Jiri Slaby wrote:
> >>> Hi,
> >>>
> >>> when I try to hibernate and resume from a kexec'ed kernel, the resume
> >>> fails with:
> >>> PM: Image not found (code -22)
> >>>
> >>> I dumped sig and orig_sig from swsusp_check and it looks like this:
> >>> 53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2
> >>> 53 57 41 50 53 50 41 43 45 32                    SWAPSPACE2
> >>>
> >>> So it looks like the image (or only the signature) did not land to the disk.
> >>
> >> Note that I see "PM: S|" and the error returned from mark_swapfiles is 0.
> >>
> >>> I use only the in-kernel hibernation, I do not even have the suspend
> >>> package installed. This is with 3.7.2 and happens this way for some time
> >>> already. When I boot the kernel normally, hib+resume works just fine.
> >>>
> >>> Any ideas?
> > 
> > Quite frankly, I'm not really surprised, because this is the first time I know
> > of anybody has tested hibernate and resume from a kexec'ed kernel.
> 
> It's quite usual and natural, that people kexec to a new kernel, I
> think. I used to do that for a long time, but it stopped workign at some
> point.
> 
> > I have no idea why exactly it doesn't work, though.
> 
> I'm trying to debug that ATM...
> 
> > Did you use s2disk or the internal hibernation?
> 
> As I write above, I don't have suspend package installed. This is a pure
> in-kernel implementation.

I see.

Well, if it used to work and it doesn't any more, then definitely something
broke.

The symptom seems to indicate that the image header is in a wrong place.

Thanks,
Rafael


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.

Re: hibernation fails from kexec'ed kernel

[Jiri Slaby]

On 01/12/2013 10:32 PM, Rafael J. Wysocki wrote:
> The symptom seems to indicate that the image header is in a wrong place.

Or a memory corruption which I saw today (without hibernation) -- data
which should have gone to the video memory appeared in struct hid_device
:/. I suggest to suspend this thread unless the corruption gets
resolved. (I suppose the corruption won't affect this issue, because I
haven't ever seen a problem with hibernation unless kexec was used. But
who knows...)

-- 
js