2021-10-29 19:26:39

by Evan Green

[permalink] [raw]
Subject: [PATCH] PM / hibernate: Fix snapshot partial write lengths

snapshot_write() is inappropriately limiting the amount of data that can
be written in cases where a partial page has already been written. For
example, one would expect to be able to write 1 byte, then 4095 bytes to
the snapshot device, and have both of those complete fully (since now
we're aligned to a page again). But what ends up happening is we write 1
byte, then 4094/4095 bytes complete successfully.

The reason is that simple_write_to_buffer()'s second argument is the
total size of the buffer, not the size of the buffer minus the offset.
Since simple_write_to_buffer() accounts for the offset in its
implementation, snapshot_write() can just pass the full page size
directly down.

Signed-off-by: Evan Green <[email protected]>
---

kernel/power/user.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/power/user.c b/kernel/power/user.c
index 740723bb388524..ad241b4ff64c58 100644
--- a/kernel/power/user.c
+++ b/kernel/power/user.c
@@ -177,7 +177,7 @@ static ssize_t snapshot_write(struct file *filp, const char __user *buf,
if (res <= 0)
goto unlock;
} else {
- res = PAGE_SIZE - pg_offp;
+ res = PAGE_SIZE;
}

if (!data_of(data->handle)) {
--
2.31.0


2021-11-15 17:44:04

by Evan Green

[permalink] [raw]
Subject: Re: [PATCH] PM / hibernate: Fix snapshot partial write lengths

Gentle bump.


On Fri, Oct 29, 2021 at 12:24 PM Evan Green <[email protected]> wrote:
>
> snapshot_write() is inappropriately limiting the amount of data that can
> be written in cases where a partial page has already been written. For
> example, one would expect to be able to write 1 byte, then 4095 bytes to
> the snapshot device, and have both of those complete fully (since now
> we're aligned to a page again). But what ends up happening is we write 1
> byte, then 4094/4095 bytes complete successfully.
>
> The reason is that simple_write_to_buffer()'s second argument is the
> total size of the buffer, not the size of the buffer minus the offset.
> Since simple_write_to_buffer() accounts for the offset in its
> implementation, snapshot_write() can just pass the full page size
> directly down.
>
> Signed-off-by: Evan Green <[email protected]>
> ---
>
> kernel/power/user.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/power/user.c b/kernel/power/user.c
> index 740723bb388524..ad241b4ff64c58 100644
> --- a/kernel/power/user.c
> +++ b/kernel/power/user.c
> @@ -177,7 +177,7 @@ static ssize_t snapshot_write(struct file *filp, const char __user *buf,
> if (res <= 0)
> goto unlock;
> } else {
> - res = PAGE_SIZE - pg_offp;
> + res = PAGE_SIZE;
> }
>
> if (!data_of(data->handle)) {
> --
> 2.31.0
>

2021-11-16 17:54:37

by Rafael J. Wysocki

[permalink] [raw]
Subject: Re: [PATCH] PM / hibernate: Fix snapshot partial write lengths

On Mon, Nov 15, 2021 at 6:13 PM Evan Green <[email protected]> wrote:
>
> Gentle bump.
>
>
> On Fri, Oct 29, 2021 at 12:24 PM Evan Green <[email protected]> wrote:
> >
> > snapshot_write() is inappropriately limiting the amount of data that can
> > be written in cases where a partial page has already been written. For
> > example, one would expect to be able to write 1 byte, then 4095 bytes to
> > the snapshot device, and have both of those complete fully (since now
> > we're aligned to a page again). But what ends up happening is we write 1
> > byte, then 4094/4095 bytes complete successfully.
> >
> > The reason is that simple_write_to_buffer()'s second argument is the
> > total size of the buffer, not the size of the buffer minus the offset.
> > Since simple_write_to_buffer() accounts for the offset in its
> > implementation, snapshot_write() can just pass the full page size
> > directly down.
> >
> > Signed-off-by: Evan Green <[email protected]>
> > ---
> >
> > kernel/power/user.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/kernel/power/user.c b/kernel/power/user.c
> > index 740723bb388524..ad241b4ff64c58 100644
> > --- a/kernel/power/user.c
> > +++ b/kernel/power/user.c
> > @@ -177,7 +177,7 @@ static ssize_t snapshot_write(struct file *filp, const char __user *buf,
> > if (res <= 0)
> > goto unlock;
> > } else {
> > - res = PAGE_SIZE - pg_offp;
> > + res = PAGE_SIZE;
> > }
> >
> > if (!data_of(data->handle)) {
> > --

Do you actually see this problem in practice?

2021-11-16 20:22:52

by Evan Green

[permalink] [raw]
Subject: Re: [PATCH] PM / hibernate: Fix snapshot partial write lengths

On Tue, Nov 16, 2021 at 9:54 AM Rafael J. Wysocki <[email protected]> wrote:
>
> On Mon, Nov 15, 2021 at 6:13 PM Evan Green <[email protected]> wrote:
> >
> > Gentle bump.
> >
> >
> > On Fri, Oct 29, 2021 at 12:24 PM Evan Green <[email protected]> wrote:
> > >
> > > snapshot_write() is inappropriately limiting the amount of data that can
> > > be written in cases where a partial page has already been written. For
> > > example, one would expect to be able to write 1 byte, then 4095 bytes to
> > > the snapshot device, and have both of those complete fully (since now
> > > we're aligned to a page again). But what ends up happening is we write 1
> > > byte, then 4094/4095 bytes complete successfully.
> > >
> > > The reason is that simple_write_to_buffer()'s second argument is the
> > > total size of the buffer, not the size of the buffer minus the offset.
> > > Since simple_write_to_buffer() accounts for the offset in its
> > > implementation, snapshot_write() can just pass the full page size
> > > directly down.
> > >
> > > Signed-off-by: Evan Green <[email protected]>
> > > ---
> > >
> > > kernel/power/user.c | 2 +-
> > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/kernel/power/user.c b/kernel/power/user.c
> > > index 740723bb388524..ad241b4ff64c58 100644
> > > --- a/kernel/power/user.c
> > > +++ b/kernel/power/user.c
> > > @@ -177,7 +177,7 @@ static ssize_t snapshot_write(struct file *filp, const char __user *buf,
> > > if (res <= 0)
> > > goto unlock;
> > > } else {
> > > - res = PAGE_SIZE - pg_offp;
> > > + res = PAGE_SIZE;
> > > }
> > >
> > > if (!data_of(data->handle)) {
> > > --
>
> Do you actually see this problem in practice?

Yes. I may fire up another thread to explain why I'm stuck doing a
partial page write, and how I might be able to stop doing that in the
future with some kernel help. But either way, this is a bug.
-Evan

2021-11-24 12:56:39

by Rafael J. Wysocki

[permalink] [raw]
Subject: Re: [PATCH] PM / hibernate: Fix snapshot partial write lengths

On Tue, Nov 16, 2021 at 9:22 PM Evan Green <[email protected]> wrote:
>
> On Tue, Nov 16, 2021 at 9:54 AM Rafael J. Wysocki <[email protected]> wrote:
> >
> > On Mon, Nov 15, 2021 at 6:13 PM Evan Green <[email protected]> wrote:
> > >
> > > Gentle bump.
> > >
> > >
> > > On Fri, Oct 29, 2021 at 12:24 PM Evan Green <[email protected]> wrote:
> > > >
> > > > snapshot_write() is inappropriately limiting the amount of data that can
> > > > be written in cases where a partial page has already been written. For
> > > > example, one would expect to be able to write 1 byte, then 4095 bytes to
> > > > the snapshot device, and have both of those complete fully (since now
> > > > we're aligned to a page again). But what ends up happening is we write 1
> > > > byte, then 4094/4095 bytes complete successfully.
> > > >
> > > > The reason is that simple_write_to_buffer()'s second argument is the
> > > > total size of the buffer, not the size of the buffer minus the offset.
> > > > Since simple_write_to_buffer() accounts for the offset in its
> > > > implementation, snapshot_write() can just pass the full page size
> > > > directly down.
> > > >
> > > > Signed-off-by: Evan Green <[email protected]>
> > > > ---
> > > >
> > > > kernel/power/user.c | 2 +-
> > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/kernel/power/user.c b/kernel/power/user.c
> > > > index 740723bb388524..ad241b4ff64c58 100644
> > > > --- a/kernel/power/user.c
> > > > +++ b/kernel/power/user.c
> > > > @@ -177,7 +177,7 @@ static ssize_t snapshot_write(struct file *filp, const char __user *buf,
> > > > if (res <= 0)
> > > > goto unlock;
> > > > } else {
> > > > - res = PAGE_SIZE - pg_offp;
> > > > + res = PAGE_SIZE;
> > > > }
> > > >
> > > > if (!data_of(data->handle)) {
> > > > --
> >
> > Do you actually see this problem in practice?
>
> Yes. I may fire up another thread to explain why I'm stuck doing a
> partial page write, and how I might be able to stop doing that in the
> future with some kernel help. But either way, this is a bug.

OK, patch applied as 5.16-rc material.

I guess it should go into -stable kernels too?

2021-11-29 20:17:10

by Evan Green

[permalink] [raw]
Subject: Re: [PATCH] PM / hibernate: Fix snapshot partial write lengths

On Wed, Nov 24, 2021 at 4:54 AM Rafael J. Wysocki <[email protected]> wrote:
>
> On Tue, Nov 16, 2021 at 9:22 PM Evan Green <[email protected]> wrote:
> >
> > On Tue, Nov 16, 2021 at 9:54 AM Rafael J. Wysocki <[email protected]> wrote:
> > >
> > > On Mon, Nov 15, 2021 at 6:13 PM Evan Green <[email protected]> wrote:
> > > >
> > > > Gentle bump.
> > > >
> > > >
> > > > On Fri, Oct 29, 2021 at 12:24 PM Evan Green <[email protected]> wrote:
> > > > >
> > > > > snapshot_write() is inappropriately limiting the amount of data that can
> > > > > be written in cases where a partial page has already been written. For
> > > > > example, one would expect to be able to write 1 byte, then 4095 bytes to
> > > > > the snapshot device, and have both of those complete fully (since now
> > > > > we're aligned to a page again). But what ends up happening is we write 1
> > > > > byte, then 4094/4095 bytes complete successfully.
> > > > >
> > > > > The reason is that simple_write_to_buffer()'s second argument is the
> > > > > total size of the buffer, not the size of the buffer minus the offset.
> > > > > Since simple_write_to_buffer() accounts for the offset in its
> > > > > implementation, snapshot_write() can just pass the full page size
> > > > > directly down.
> > > > >
> > > > > Signed-off-by: Evan Green <[email protected]>
> > > > > ---
> > > > >
> > > > > kernel/power/user.c | 2 +-
> > > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > > >
> > > > > diff --git a/kernel/power/user.c b/kernel/power/user.c
> > > > > index 740723bb388524..ad241b4ff64c58 100644
> > > > > --- a/kernel/power/user.c
> > > > > +++ b/kernel/power/user.c
> > > > > @@ -177,7 +177,7 @@ static ssize_t snapshot_write(struct file *filp, const char __user *buf,
> > > > > if (res <= 0)
> > > > > goto unlock;
> > > > > } else {
> > > > > - res = PAGE_SIZE - pg_offp;
> > > > > + res = PAGE_SIZE;
> > > > > }
> > > > >
> > > > > if (!data_of(data->handle)) {
> > > > > --
> > >
> > > Do you actually see this problem in practice?
> >
> > Yes. I may fire up another thread to explain why I'm stuck doing a
> > partial page write, and how I might be able to stop doing that in the
> > future with some kernel help. But either way, this is a bug.
>
> OK, patch applied as 5.16-rc material.
>
> I guess it should go into -stable kernels too?

Yes, putting it into -stable would make sense also. I should have CCed
them originally, doing that now.
-Evan

2021-11-30 06:51:52

by Greg Kroah-Hartman

[permalink] [raw]
Subject: Re: [PATCH] PM / hibernate: Fix snapshot partial write lengths

On Mon, Nov 29, 2021 at 08:50:06AM -0800, Evan Green wrote:
> On Wed, Nov 24, 2021 at 4:54 AM Rafael J. Wysocki <[email protected]> wrote:
> >
> > On Tue, Nov 16, 2021 at 9:22 PM Evan Green <[email protected]> wrote:
> > >
> > > On Tue, Nov 16, 2021 at 9:54 AM Rafael J. Wysocki <[email protected]> wrote:
> > > >
> > > > On Mon, Nov 15, 2021 at 6:13 PM Evan Green <[email protected]> wrote:
> > > > >
> > > > > Gentle bump.
> > > > >
> > > > >
> > > > > On Fri, Oct 29, 2021 at 12:24 PM Evan Green <[email protected]> wrote:
> > > > > >
> > > > > > snapshot_write() is inappropriately limiting the amount of data that can
> > > > > > be written in cases where a partial page has already been written. For
> > > > > > example, one would expect to be able to write 1 byte, then 4095 bytes to
> > > > > > the snapshot device, and have both of those complete fully (since now
> > > > > > we're aligned to a page again). But what ends up happening is we write 1
> > > > > > byte, then 4094/4095 bytes complete successfully.
> > > > > >
> > > > > > The reason is that simple_write_to_buffer()'s second argument is the
> > > > > > total size of the buffer, not the size of the buffer minus the offset.
> > > > > > Since simple_write_to_buffer() accounts for the offset in its
> > > > > > implementation, snapshot_write() can just pass the full page size
> > > > > > directly down.
> > > > > >
> > > > > > Signed-off-by: Evan Green <[email protected]>
> > > > > > ---
> > > > > >
> > > > > > kernel/power/user.c | 2 +-
> > > > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > > > >
> > > > > > diff --git a/kernel/power/user.c b/kernel/power/user.c
> > > > > > index 740723bb388524..ad241b4ff64c58 100644
> > > > > > --- a/kernel/power/user.c
> > > > > > +++ b/kernel/power/user.c
> > > > > > @@ -177,7 +177,7 @@ static ssize_t snapshot_write(struct file *filp, const char __user *buf,
> > > > > > if (res <= 0)
> > > > > > goto unlock;
> > > > > > } else {
> > > > > > - res = PAGE_SIZE - pg_offp;
> > > > > > + res = PAGE_SIZE;
> > > > > > }
> > > > > >
> > > > > > if (!data_of(data->handle)) {
> > > > > > --
> > > >
> > > > Do you actually see this problem in practice?
> > >
> > > Yes. I may fire up another thread to explain why I'm stuck doing a
> > > partial page write, and how I might be able to stop doing that in the
> > > future with some kernel help. But either way, this is a bug.
> >
> > OK, patch applied as 5.16-rc material.
> >
> > I guess it should go into -stable kernels too?
>
> Yes, putting it into -stable would make sense also. I should have CCed
> them originally, doing that now.


<formletter>

This is not the correct way to submit patches for inclusion in the
stable kernel tree. Please read:
https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
for how to do this properly.

</formletter>

2021-11-30 11:47:32

by Rafael J. Wysocki

[permalink] [raw]
Subject: Re: [PATCH] PM / hibernate: Fix snapshot partial write lengths

On Mon, Nov 29, 2021 at 5:50 PM Evan Green <[email protected]> wrote:
>
> On Wed, Nov 24, 2021 at 4:54 AM Rafael J. Wysocki <[email protected]> wrote:
> >
> > On Tue, Nov 16, 2021 at 9:22 PM Evan Green <[email protected]> wrote:
> > >
> > > On Tue, Nov 16, 2021 at 9:54 AM Rafael J. Wysocki <[email protected]> wrote:
> > > >
> > > > On Mon, Nov 15, 2021 at 6:13 PM Evan Green <[email protected]> wrote:
> > > > >
> > > > > Gentle bump.
> > > > >
> > > > >
> > > > > On Fri, Oct 29, 2021 at 12:24 PM Evan Green <[email protected]> wrote:
> > > > > >
> > > > > > snapshot_write() is inappropriately limiting the amount of data that can
> > > > > > be written in cases where a partial page has already been written. For
> > > > > > example, one would expect to be able to write 1 byte, then 4095 bytes to
> > > > > > the snapshot device, and have both of those complete fully (since now
> > > > > > we're aligned to a page again). But what ends up happening is we write 1
> > > > > > byte, then 4094/4095 bytes complete successfully.
> > > > > >
> > > > > > The reason is that simple_write_to_buffer()'s second argument is the
> > > > > > total size of the buffer, not the size of the buffer minus the offset.
> > > > > > Since simple_write_to_buffer() accounts for the offset in its
> > > > > > implementation, snapshot_write() can just pass the full page size
> > > > > > directly down.
> > > > > >
> > > > > > Signed-off-by: Evan Green <[email protected]>
> > > > > > ---
> > > > > >
> > > > > > kernel/power/user.c | 2 +-
> > > > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > > > >
> > > > > > diff --git a/kernel/power/user.c b/kernel/power/user.c
> > > > > > index 740723bb388524..ad241b4ff64c58 100644
> > > > > > --- a/kernel/power/user.c
> > > > > > +++ b/kernel/power/user.c
> > > > > > @@ -177,7 +177,7 @@ static ssize_t snapshot_write(struct file *filp, const char __user *buf,
> > > > > > if (res <= 0)
> > > > > > goto unlock;
> > > > > > } else {
> > > > > > - res = PAGE_SIZE - pg_offp;
> > > > > > + res = PAGE_SIZE;
> > > > > > }
> > > > > >
> > > > > > if (!data_of(data->handle)) {
> > > > > > --
> > > >
> > > > Do you actually see this problem in practice?
> > >
> > > Yes. I may fire up another thread to explain why I'm stuck doing a
> > > partial page write, and how I might be able to stop doing that in the
> > > future with some kernel help. But either way, this is a bug.
> >
> > OK, patch applied as 5.16-rc material.
> >
> > I guess it should go into -stable kernels too?
>
> Yes, putting it into -stable would make sense also. I should have CCed
> them originally, doing that now.

Well, you need to point them to the upstream commit to backport.

In this particular case it would be

commit 88a5045f176b78c33a269a30a7b146e99c550bd9 (pm-sleep)
Author: Evan Green <[email protected]>
Date: Fri Oct 29 12:24:22 2021 -0700

PM: hibernate: Fix snapshot partial write lengths

I'll send an inclusion request for this. I guess it should go into
all of the applicable -stable series, right?

2021-11-30 18:21:48

by Evan Green

[permalink] [raw]
Subject: Re: [PATCH] PM / hibernate: Fix snapshot partial write lengths

On Tue, Nov 30, 2021 at 3:46 AM Rafael J. Wysocki <[email protected]> wrote:
>
> On Mon, Nov 29, 2021 at 5:50 PM Evan Green <[email protected]> wrote:
> >
> > On Wed, Nov 24, 2021 at 4:54 AM Rafael J. Wysocki <[email protected]> wrote:
> > >
> > > On Tue, Nov 16, 2021 at 9:22 PM Evan Green <[email protected]> wrote:
> > > >
> > > > On Tue, Nov 16, 2021 at 9:54 AM Rafael J. Wysocki <[email protected]> wrote:
> > > > >
> > > > > On Mon, Nov 15, 2021 at 6:13 PM Evan Green <[email protected]> wrote:
> > > > > >
> > > > > > Gentle bump.
> > > > > >
> > > > > >
> > > > > > On Fri, Oct 29, 2021 at 12:24 PM Evan Green <[email protected]> wrote:
> > > > > > >
> > > > > > > snapshot_write() is inappropriately limiting the amount of data that can
> > > > > > > be written in cases where a partial page has already been written. For
> > > > > > > example, one would expect to be able to write 1 byte, then 4095 bytes to
> > > > > > > the snapshot device, and have both of those complete fully (since now
> > > > > > > we're aligned to a page again). But what ends up happening is we write 1
> > > > > > > byte, then 4094/4095 bytes complete successfully.
> > > > > > >
> > > > > > > The reason is that simple_write_to_buffer()'s second argument is the
> > > > > > > total size of the buffer, not the size of the buffer minus the offset.
> > > > > > > Since simple_write_to_buffer() accounts for the offset in its
> > > > > > > implementation, snapshot_write() can just pass the full page size
> > > > > > > directly down.
> > > > > > >
> > > > > > > Signed-off-by: Evan Green <[email protected]>
> > > > > > > ---
> > > > > > >
> > > > > > > kernel/power/user.c | 2 +-
> > > > > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > > > > >
> > > > > > > diff --git a/kernel/power/user.c b/kernel/power/user.c
> > > > > > > index 740723bb388524..ad241b4ff64c58 100644
> > > > > > > --- a/kernel/power/user.c
> > > > > > > +++ b/kernel/power/user.c
> > > > > > > @@ -177,7 +177,7 @@ static ssize_t snapshot_write(struct file *filp, const char __user *buf,
> > > > > > > if (res <= 0)
> > > > > > > goto unlock;
> > > > > > > } else {
> > > > > > > - res = PAGE_SIZE - pg_offp;
> > > > > > > + res = PAGE_SIZE;
> > > > > > > }
> > > > > > >
> > > > > > > if (!data_of(data->handle)) {
> > > > > > > --
> > > > >
> > > > > Do you actually see this problem in practice?
> > > >
> > > > Yes. I may fire up another thread to explain why I'm stuck doing a
> > > > partial page write, and how I might be able to stop doing that in the
> > > > future with some kernel help. But either way, this is a bug.
> > >
> > > OK, patch applied as 5.16-rc material.
> > >
> > > I guess it should go into -stable kernels too?
> >
> > Yes, putting it into -stable would make sense also. I should have CCed
> > them originally, doing that now.
>
> Well, you need to point them to the upstream commit to backport.
>
> In this particular case it would be
>
> commit 88a5045f176b78c33a269a30a7b146e99c550bd9 (pm-sleep)
> Author: Evan Green <[email protected]>
> Date: Fri Oct 29 12:24:22 2021 -0700
>
> PM: hibernate: Fix snapshot partial write lengths
>
> I'll send an inclusion request for this. I guess it should go into
> all of the applicable -stable series, right?

Correct. From what I can tell this code and simple_write_to_buffer()
haven't changed since its introduction in 2010, so my patch should
have had a fixes tag like:

Fixes: d3c1b24c50e8 ("PM / Hibernate: Snapshot cleanup")

Thanks for doing that Rafael, sorry for flubbing the stable aspect.

-Evan