> > After the 'restore_registers' it returns and we end up called
> > restore_processor_state() - where we reload the GDT. The reload of
> > the GDT is not needed as bootup kernel has already loaded the GDT which
> > is at the same physical location as the the restored kernel.
>
> I'm not sure if this particular statement is actually correct. It is correct
> on 32-bit, but here it is not necessary for the bootup kernel to be the same
> as the image one. Different kernel version may be used for that even (at
> least theoretically). So the question is, and I'm quite unsure about the
> answer, if the GDT of from the bootup kernel is really *guaranteed* to be
> at the same location (given that those kernels may be really different).
A bit of testing with different bootup kernel provided me with this error:
PM: Image mismatch: version
which after a bit of digging pointed me to 'check_image_kernel'. The criteria
there imply that the different kernel versions or releases cannot be used with
hibernation.
For "fun" I subverted the checks and tried resuming a differently compiled
kernel (CONFIG_KALLSYMS_ALL=y) vs the one that was booting
(# CONFIG_KALLSYMS_ALL is not set) and found out that it does not work on a vanilla
v3.9 kernel.
diff --git a/kernel/power/snapshot.c b/kernel/power/snapshot.c
index 0de2857..8589e05 100644
--- a/kernel/power/snapshot.c
+++ b/kernel/power/snapshot.c
@@ -1633,10 +1633,16 @@ static char *check_image_kernel(struct swsusp_info *info)
return "kernel version";
if (strcmp(info->uts.sysname,init_utsname()->sysname))
return "system type";
- if (strcmp(info->uts.release,init_utsname()->release))
- return "kernel release";
- if (strcmp(info->uts.version,init_utsname()->version))
- return "version";
+ if (strcmp(info->uts.release,init_utsname()->release)) {
+ printk(KERN_ERR "%s != %s .. continuing on.\n", info->uts.release,
+ init_utsname()->release);
+ return NULL;
+ }
+ if (strcmp(info->uts.version,init_utsname()->version)) {
+ printk(KERN_ERR "%s != %s .. continuing on.\n", info->uts.version,
+ init_utsname()->version);
+ return NULL;
+ }
if (strcmp(info->uts.machine,init_utsname()->machine))
return "machine";
return NULL;
Meaning without these patches we do crash when trying to load a different kernel.
Here is the log:
# echo "8:1" > /sys/power/resume
[ 18.881169] PM: Starting manual resume from disk
[ 18.884164] PM: Hibernation image partition 8:1 present
[ 18.885418] PM: Looking for hibernation image.
[ 18.888344] PM: Image signature found, resuming
[ 18.895387] PM: Marking nosave pages: [mem 0x0009f000-0x000fffff]
[ 18.896892] PM: Basic memory bitmaps created
[ 18.897937] PM: Preparing processes for restore.
[ 18.900102] Freezing user space processes ... (elapsed 0.01 seconds) done.
[ 18.914402] PM: Loading hibernation image.
[ 19.201178] PM: Using 2 thread(s) for decompression.
[ 19.201178] PM: Loading and decompressing image data (53823 pages)...
[ 19.210490] 3.9.0upstream != 3.9.0upstream-dirty .. continuing on.
[ 19.874986] PM: Image loading progress: 0%
[ 20.027158] PM: Image loading progress: 10%
[ 20.177723] PM: Image loading progress: 20%
[ 20.335267] PM: Image loading progress: 30%
[ 20.491458] PM: Image loading progress: 40%
[ 20.578486] PM: Image loading progress: 50%
[ 20.663751] PM: Image loading progress: 60%
[ 20.985474] PM: Image loading progress: 70%
[ 21.112780] PM: Image loading progress: 80%
[ 21.219476] PM: Image loading progress: 90%
[ 21.336482] PM: Image loading progress: 100%
[ 21.340302] PM: Image loading done.
[ 21.342830] PM: Read 215292 kbytes in 2.12 seconds (101.55 MB/s)
[ 21.350612] PM: Image successfully loaded
[ 21.353913] Suspending console(s) (use no_console_suspend to debug)
.. and the machine either reboots or just hangs
So I think the assumptions I made are OK?
--
To unsubscribe from this list: send the line "unsubscribe linux-tip-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
|