Certain drivers release resources (pinned pages, etc.) into system
memory during the prepare freeze PM op, making them swappable.
Currently, hibernate_preallocate_memory is called before prepare freeze,
so those drivers have no opportunity to release resources first. If a
driver is holding a large amount of unswappable system RAM, this can
cause hibernate_preallocate_memory to fail.

Move the call to hibernate_preallocate_memory after prepare freeze.
According to the documentation for the prepare callback, devices should
be left in a usable state, so storage drivers should still be able to
service I/O requests. This allows drivers to release unswappable
resources prior to preallocation, so they can be swapped out through
hibernate_preallocate_memory's reclaim path.

Signed-off-by: Matthew Leach <matthew.leach@collabora.com>
---
 kernel/power/hibernate.c | 14 ++++++--------
 1 file changed, 6 insertions(+), 8 deletions(-)

diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c
index af8d07bafe02..ff62e220ded1 100644
--- a/kernel/power/hibernate.c
+++ b/kernel/power/hibernate.c
@@ -425,14 +425,9 @@ int hibernation_snapshot(int platform_mode)
 	if (error)
 		goto Close;
 
-	/* Preallocate image memory before shutting down devices. */
-	error = hibernate_preallocate_memory();
-	if (error)
-		goto Close;
-
 	error = freeze_kernel_threads();
 	if (error)
-		goto Cleanup;
+		goto Close;
 
 	if (hibernation_test(TEST_FREEZER)) {
 
@@ -450,6 +445,11 @@ int hibernation_snapshot(int platform_mode)
 		goto Thaw;
 	}
 
+	/* Preallocate image memory before shutting down devices. */
+	error = hibernate_preallocate_memory();
+	if (error)
+		goto Thaw;
+
 	/*
 	 * Device drivers may move lots of data to shmem in dpm_prepare(). The shmem
 	 * pages will use lots of system memory, causing hibernation image creation
@@ -494,8 +494,6 @@ int hibernation_snapshot(int platform_mode)
 
  Thaw:
 	thaw_kernel_threads();
- Cleanup:
-	swsusp_free();
 	goto Close;
 }
 

---
base-commit: f338e77383789c0cae23ca3d48adcc5e9e137e3c
change-id: 20260321-hibernation-fixes-69bca2ee5b65

Best regards,
-- 
Matt

On Sat, Mar 21, 2026 at 08:51:31AM +0000, Matthew Leach wrote:

Hi Matthew,

> +     /* Preallocate image memory before shutting down devices. */
> +     error = hibernate_preallocate_memory();

The shrink_shmem_memory() below is also intended to run after
dpm_prepare(), and the motivation is the same. Since
hibernate_preallocate_memory() already does shrinking internally,
how about pulling out the shrink-related parts and consolidating
them together?

> +     if (error)
> +             goto Thaw;

Don't we need to call dpm_complete(PMSG_RECOVER) in this error path?

>  Thaw:
>       thaw_kernel_threads();
> - Cleanup:
> -     swsusp_free();
>       goto Close;

Thanks,
Youngjun Park

YoungJun Park <youngjun.park@lge.com> writes:

> On Sat, Mar 21, 2026 at 08:51:31AM +0000, Matthew Leach wrote:
>
> Hi Matthew,

Hi Youngjun,

Thanks for the review.

>> +     /* Preallocate image memory before shutting down devices. */
>> +     error = hibernate_preallocate_memory();
>
> The shrink_shmem_memory() below is also intended to run after
> dpm_prepare(), and the motivation is the same. Since
> hibernate_preallocate_memory() already does shrinking internally,
> how about pulling out the shrink-related parts and consolidating
> them together?

Agreed. I'll remove shrink_shmem_memory() completely since
hibernate_preallocate_memory() will have swapped out as many pages as is
required to construct the image in system RAM. There's no point doing
another reclaim.

>> +     if (error)
>> +             goto Thaw;
>
> Don't we need to call dpm_complete(PMSG_RECOVER) in this error path?

Yes, good catch.

Regards,
-- 
Matt