[3.2-stable] Suspend regression after "x86, mm: Set NX across entire PMD at boot"

[3.2-stable] Suspend regression after "x86, mm: Set NX across entire PMD at boot"

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 1363 bytes --]

Commit 45e2a9d4701d8c624d4a4bcdd1084eae31e92f58 ("x86, mm: Set NX across
entire PMD at boot") has been backported to many stable branches:

v3.10.62: 0b15c16c4f4a x86, mm: Set NX across entire PMD at boot
v3.12.35: 868b19cb0314 x86, mm: Set NX across entire PMD at boot
v3.14.26: 1a829c59816e x86, mm: Set NX across entire PMD at boot
v3.17.5: 6231385167e0 x86, mm: Set NX across entire PMD at boot
v3.2.65: a5c187d92d2c x86, mm: Set NX across entire PMD at boot
v3.13.11-ckt12: 3657489f7009 x86, mm: Set NX across entire PMD at boot
v3.16.7-ckt3: de32229b546e x86, mm: Set NX across entire PMD at boot

Unfortunately, in 3.2.65 suspend/resume stopped working on at least some
systems - specifically, the system would reboot when woken:

https://bugs.debian.org/774436 - Lenovo ThinkPad T530
https://bugs.debian.org/774461 - Lenovo ThinkPad X230
https://bugs.debian.org/774526 - ECS P4M900T-M2
https://bugs.debian.org/775059 - ASUS X201EP

I could reproduce this on my own ASUS P8Z68-V LX and was able to bisect
to the backport of this commit.

Assuming that this regression has not occurred on mainline or the other
branches, I will revert this patch for 3.2.y unless someone can point
out a missing dependency that I should apply.

Ben.

-- 
Ben Hutchings
One of the nice things about standards is that there are so many of them.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 811 bytes --]

[3.2-stable] Suspend regression after "x86, 64bit, mm: Mark data/bss/brk to nx"

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 1799 bytes --]

On Sun, 2015-01-11 at 00:23 +0000, Ben Hutchings wrote:
> Commit 45e2a9d4701d8c624d4a4bcdd1084eae31e92f58 ("x86, mm: Set NX across
> entire PMD at boot") has been backported to many stable branches:
> 
> v3.10.62: 0b15c16c4f4a x86, mm: Set NX across entire PMD at boot
> v3.12.35: 868b19cb0314 x86, mm: Set NX across entire PMD at boot
> v3.14.26: 1a829c59816e x86, mm: Set NX across entire PMD at boot
> v3.17.5: 6231385167e0 x86, mm: Set NX across entire PMD at boot
> v3.2.65: a5c187d92d2c x86, mm: Set NX across entire PMD at boot
> v3.13.11-ckt12: 3657489f7009 x86, mm: Set NX across entire PMD at boot
> v3.16.7-ckt3: de32229b546e x86, mm: Set NX across entire PMD at boot

Sorry, I think I booted the wrong version at the last step of bisection.
It was actually the previous commit on the 3.2.y branch, which was
commit 72212675d1c96f5db8ec6fb35701879911193158 upstream ("x86, 64bit,
mm: Mark data/bss/brk to nx") that introduced the regression.  I intend
to revert both of these.

Ben.

> Unfortunately, in 3.2.65 suspend/resume stopped working on at least some
> systems - specifically, the system would reboot when woken:
> 
> https://bugs.debian.org/774436 - Lenovo ThinkPad T530
> https://bugs.debian.org/774461 - Lenovo ThinkPad X230
> https://bugs.debian.org/774526 - ECS P4M900T-M2
> https://bugs.debian.org/775059 - ASUS X201EP
> 
> I could reproduce this on my own ASUS P8Z68-V LX and was able to bisect
> to the backport of this commit.
> 
> Assuming that this regression has not occurred on mainline or the other
> branches, I will revert this patch for 3.2.y unless someone can point
> out a missing dependency that I should apply.
> 
> Ben.
> 

-- 
Ben Hutchings
One of the nice things about standards is that there are so many of them.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 811 bytes --]

Re: [3.2-stable] Suspend regression after "x86, 64bit, mm: Mark data/bss/brk to nx"

[Luis Henriques]

On Sun, Jan 11, 2015 at 09:16:23PM +0000, Ben Hutchings wrote:
> On Sun, 2015-01-11 at 00:23 +0000, Ben Hutchings wrote:
> > Commit 45e2a9d4701d8c624d4a4bcdd1084eae31e92f58 ("x86, mm: Set NX across
> > entire PMD at boot") has been backported to many stable branches:
> > 
> > v3.10.62: 0b15c16c4f4a x86, mm: Set NX across entire PMD at boot
> > v3.12.35: 868b19cb0314 x86, mm: Set NX across entire PMD at boot
> > v3.14.26: 1a829c59816e x86, mm: Set NX across entire PMD at boot
> > v3.17.5: 6231385167e0 x86, mm: Set NX across entire PMD at boot
> > v3.2.65: a5c187d92d2c x86, mm: Set NX across entire PMD at boot
> > v3.13.11-ckt12: 3657489f7009 x86, mm: Set NX across entire PMD at boot
> > v3.16.7-ckt3: de32229b546e x86, mm: Set NX across entire PMD at boot
> 
> Sorry, I think I booted the wrong version at the last step of bisection.
> It was actually the previous commit on the 3.2.y branch, which was
> commit 72212675d1c96f5db8ec6fb35701879911193158 upstream ("x86, 64bit,
> mm: Mark data/bss/brk to nx") that introduced the regression.  I intend
> to revert both of these.
> 
> Ben.
> 

This is an old (3.8) commit that seems to have been backported to 3.2
as a dependency of 45e2a9d4701d8c624d4a4bcdd1084eae31e92f58 and all
the other stable kernels already included it.  Since I am not aware of
any new regression, I believe this is could be a 3.2 specific issue.

Anyway, thanks a lot for sharing this, as we have an Ubuntu kernel
that is based on 3.2, to which we regularly apply 3.2 releases.

Cheers,
--
Lu�s

> > Unfortunately, in 3.2.65 suspend/resume stopped working on at least some
> > systems - specifically, the system would reboot when woken:
> > 
> > https://bugs.debian.org/774436 - Lenovo ThinkPad T530
> > https://bugs.debian.org/774461 - Lenovo ThinkPad X230
> > https://bugs.debian.org/774526 - ECS P4M900T-M2
> > https://bugs.debian.org/775059 - ASUS X201EP
> > 
> > I could reproduce this on my own ASUS P8Z68-V LX and was able to bisect
> > to the backport of this commit.
> > 
> > Assuming that this regression has not occurred on mainline or the other
> > branches, I will revert this patch for 3.2.y unless someone can point
> > out a missing dependency that I should apply.
> > 
> > Ben.
> > 
> 
> -- 
> Ben Hutchings
> One of the nice things about standards is that there are so many of them.