[PATCH 0/3] Fixups for hibernate issues reported by Askar Safin

[PATCH 0/3] Fixups for hibernate issues reported by Askar Safin

[Mario Limonciello (AMD)]

Askar Safin reported some usability issues with hibernate messageing in
some failure cases.  This series fixes those issues as well as some style
issues I noticed while touching the function.

Cc: Askar Safin <safinaskar@gmail.com>

Mario Limonciello (AMD) (3):
  PM: hibernate: Emit an error when image writing fails
  PM: hibernate: Use atomic64_t for compressed_size variable
  PM: hibernate: Fix style issues in save_compressed_image()

 kernel/power/swap.c | 21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)

-- 
2.43.0

[PATCH 1/3] PM: hibernate: Emit an error when image writing fails

[Mario Limonciello (AMD)]

If image writing fails a return code is passed up to the caller but
none of the callers log anything to the log and so the only record
of it is the return code that userspace gets.

Adjust the logging so that the image size and speed of writing is
only emitted on success and if there is an error it's saved the logs.

Reported-by: Askar Safin <safinaskar@gmail.com>
Closes: https://lore.kernel.org/linux-pm/20251105180506.137448-1-safinaskar@gmail.com/
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
---
 kernel/power/swap.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/kernel/power/swap.c b/kernel/power/swap.c
index 0beff7eeaabaf..f9bf01f355189 100644
--- a/kernel/power/swap.c
+++ b/kernel/power/swap.c
@@ -877,11 +877,13 @@ static int save_compressed_image(struct swap_map_handle *handle,
 	stop = ktime_get();
 	if (!ret)
 		ret = err2;
-	if (!ret)
+	if (!ret) {
+		swsusp_show_speed(start, stop, nr_to_write, "Wrote");
+		pr_info("Image size after compression: %d kbytes\n",
+			(atomic_read(&compressed_size) / 1024));
 		pr_info("Image saving done\n");
-	swsusp_show_speed(start, stop, nr_to_write, "Wrote");
-	pr_info("Image size after compression: %d kbytes\n",
-		(atomic_read(&compressed_size) / 1024));
+	} else
+		pr_err("Image saving failed: %d\n", ret);
 
 out_clean:
 	hib_finish_batch(&hb);
-- 
2.43.0

[PATCH 2/3] PM: hibernate: Use atomic64_t for compressed_size variable

[Mario Limonciello (AMD)]

`compressed_size` can overflow, showing nonsensical values.
Change from `atomic_t` to `atomic64_t` to prevent overflow.

Reported-by: Askar Safin <safinaskar@gmail.com>
Closes: https://lore.kernel.org/linux-pm/20251105180506.137448-1-safinaskar@gmail.com/
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
---
 kernel/power/swap.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/kernel/power/swap.c b/kernel/power/swap.c
index f9bf01f355189..37270fa5ea600 100644
--- a/kernel/power/swap.c
+++ b/kernel/power/swap.c
@@ -635,7 +635,7 @@ struct cmp_data {
 };
 
 /* Indicates the image size after compression */
-static atomic_t compressed_size = ATOMIC_INIT(0);
+static atomic64_t compressed_size = ATOMIC_INIT(0);
 
 /*
  * Compression function that runs in its own thread.
@@ -664,7 +664,7 @@ static int compress_threadfn(void *data)
 		d->ret = crypto_acomp_compress(d->cr);
 		d->cmp_len = d->cr->dlen;
 
-		atomic_set(&compressed_size, atomic_read(&compressed_size) + d->cmp_len);
+		atomic64_add(d->cmp_len, &compressed_size);
 		atomic_set_release(&d->stop, 1);
 		wake_up(&d->done);
 	}
@@ -696,7 +696,7 @@ static int save_compressed_image(struct swap_map_handle *handle,
 
 	hib_init_batch(&hb);
 
-	atomic_set(&compressed_size, 0);
+	atomic64_set(&compressed_size, 0);
 
 	/*
 	 * We'll limit the number of threads for compression to limit memory
@@ -879,8 +879,8 @@ static int save_compressed_image(struct swap_map_handle *handle,
 		ret = err2;
 	if (!ret) {
 		swsusp_show_speed(start, stop, nr_to_write, "Wrote");
-		pr_info("Image size after compression: %d kbytes\n",
-			(atomic_read(&compressed_size) / 1024));
+		pr_info("Image size after compression: %lld kbytes\n",
+			(atomic64_read(&compressed_size) / 1024));
 		pr_info("Image saving done\n");
 	} else
 		pr_err("Image saving failed: %d\n", ret);
-- 
2.43.0

[PATCH 3/3] PM: hibernate: Fix style issues in save_compressed_image()

[Mario Limonciello (AMD)]

Fix the two issues caught by checkpatch:
	Trailing statements should be on next line.
	Prefer 'unsigned int' to bare use of 'unsigned'

Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
---
 kernel/power/swap.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/kernel/power/swap.c b/kernel/power/swap.c
index 37270fa5ea600..488a74d2a56b8 100644
--- a/kernel/power/swap.c
+++ b/kernel/power/swap.c
@@ -689,7 +689,7 @@ static int save_compressed_image(struct swap_map_handle *handle,
 	ktime_t start;
 	ktime_t stop;
 	size_t off;
-	unsigned thr, run_threads, nr_threads;
+	unsigned int thr, run_threads, nr_threads;
 	unsigned char *page = NULL;
 	struct cmp_data *data = NULL;
 	struct crc_data *crc = NULL;
@@ -901,7 +901,8 @@ static int save_compressed_image(struct swap_map_handle *handle,
 		}
 		vfree(data);
 	}
-	if (page) free_page((unsigned long)page);
+	if (page)
+		free_page((unsigned long)page);
 
 	return ret;
 }
-- 
2.43.0

Re: [PATCH 1/3] PM: hibernate: Emit an error when image writing fails

[Askar Safin]

"Mario Limonciello (AMD)" <superm1@kernel.org>:
> If image writing fails a return code is passed up to the caller but
> none of the callers log anything to the log and so the only record
> of it is the return code that userspace gets.

Thank you! This is bug fix, so, please, add to this particular commit:

> Cc: <stable@vger.kernel.org>
> Fixes: a06c6f5d3cc9 ("PM: hibernate: Move to crypto APIs for LZO compression")

a06c6f5d3cc9 introduced that pr_info("Image size after compression...")

-- 
Askar Safin

Re: [PATCH 2/3] PM: hibernate: Use atomic64_t for compressed_size variable

[Askar Safin]

"Mario Limonciello (AMD)" <superm1@kernel.org>:
> `compressed_size` can overflow, showing nonsensical values.
> Change from `atomic_t` to `atomic64_t` to prevent overflow.

This is bug fix, too. Please, add:

> Cc: <stable@vger.kernel.org>
> Fixes: a06c6f5d3cc9 ("PM: hibernate: Move to crypto APIs for LZO compression")

a06c6f5d3cc9 introduced variable "compressed_size".

(Yes, this commit responsible for both bugs.)

> -		atomic_set(&compressed_size, atomic_read(&compressed_size) + d->cmp_len);

So, we read atomic variable, then add something to it, and then write?
This looks like textbook parallel programming error. Very strange.
Anyway, you fixed this, Mario! Thank you!

-- 
Askar Safin

Re: [PATCH 0/3] Fixups for hibernate issues reported by Askar Safin

[Askar Safin]

"Mario Limonciello (AMD)" <superm1@kernel.org>:
> Askar Safin reported some usability issues with hibernate messageing in
> some failure cases.  This series fixes those issues as well as some style
> issues I noticed while touching the function.

I tested in Qemu. Now everything works. Thank you!

Tested-By: <safinaskar@gmail.com>

-- 
Askar Safin

Re: [PATCH 0/3] Fixups for hibernate issues reported by Askar Safin

[Mario Limonciello]

On 11/6/25 2:03 PM, Askar Safin wrote:
> "Mario Limonciello (AMD)" <superm1@kernel.org>:
>> Askar Safin reported some usability issues with hibernate messageing in
>> some failure cases.  This series fixes those issues as well as some style
>> issues I noticed while touching the function.
> 
> I tested in Qemu. Now everything works. Thank you!
> 
> Tested-By: <safinaskar@gmail.com>
> 

Thanks!  I'll let Rafael pick up the extra tags you had for the patches 
if he agrees with them.

Re: [PATCH 2/3] PM: hibernate: Use atomic64_t for compressed_size variable

[Rafael J. Wysocki]

On Thu, Nov 6, 2025 at 5:52 AM Mario Limonciello (AMD)
<superm1@kernel.org> wrote:
>
> `compressed_size` can overflow, showing nonsensical values.
> Change from `atomic_t` to `atomic64_t` to prevent overflow.
>
> Reported-by: Askar Safin <safinaskar@gmail.com>
> Closes: https://lore.kernel.org/linux-pm/20251105180506.137448-1-safinaskar@gmail.com/
> Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
> ---
>  kernel/power/swap.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/kernel/power/swap.c b/kernel/power/swap.c
> index f9bf01f355189..37270fa5ea600 100644
> --- a/kernel/power/swap.c
> +++ b/kernel/power/swap.c
> @@ -635,7 +635,7 @@ struct cmp_data {
>  };
>
>  /* Indicates the image size after compression */
> -static atomic_t compressed_size = ATOMIC_INIT(0);
> +static atomic64_t compressed_size = ATOMIC_INIT(0);
>
>  /*
>   * Compression function that runs in its own thread.
> @@ -664,7 +664,7 @@ static int compress_threadfn(void *data)
>                 d->ret = crypto_acomp_compress(d->cr);
>                 d->cmp_len = d->cr->dlen;
>
> -               atomic_set(&compressed_size, atomic_read(&compressed_size) + d->cmp_len);

Hmm, how did I miss this?

> +               atomic64_add(d->cmp_len, &compressed_size);
>                 atomic_set_release(&d->stop, 1);
>                 wake_up(&d->done);
>         }
> @@ -696,7 +696,7 @@ static int save_compressed_image(struct swap_map_handle *handle,
>
>         hib_init_batch(&hb);
>
> -       atomic_set(&compressed_size, 0);
> +       atomic64_set(&compressed_size, 0);
>
>         /*
>          * We'll limit the number of threads for compression to limit memory
> @@ -879,8 +879,8 @@ static int save_compressed_image(struct swap_map_handle *handle,
>                 ret = err2;
>         if (!ret) {
>                 swsusp_show_speed(start, stop, nr_to_write, "Wrote");
> -               pr_info("Image size after compression: %d kbytes\n",
> -                       (atomic_read(&compressed_size) / 1024));
> +               pr_info("Image size after compression: %lld kbytes\n",
> +                       (atomic64_read(&compressed_size) / 1024));
>                 pr_info("Image saving done\n");
>         } else
>                 pr_err("Image saving failed: %d\n", ret);
> --

Applied as 6.18-rc material along with the other patches in the
series, with some changelog adjustments, added tags etc.

Thanks!