Re: [PATCH 0/3] properly account for stack randomization and guard gap

2017-06-23 Thread Rik van Riel 
On Fri, 2017-06-23 at 10:35 +0200, Ingo Molnar wrote:
> * r...@redhat.com  wrote:
> 
> > When RLIMIT_STACK is larger than the minimum gap enforced by
> > mmap_base(), it is possible for the kernel to place the mmap
> > area where the stack wants to grow, resulting in the stack
> > not being able to use the space that should have been allocated
> > to it through RLIMIT_STACK.
> > 
> > This series ensures that x86, ARM64, and PPC have at least
> > RLIMIT_STACK + stack randomization + the stack guard gap
> > space available for the stack.
> > 
> > s390 seems to be ok. I have not checked other architectures.
> 
> x86 patch LGTM:
> 
>   Acked-by: Ingo Molnar 
> 
> ... but I suspect this wants to go via -mm or Linus directly?

I believe Andrew picked it up yesterday.

Re: [PATCH 0/3] properly account for stack randomization and guard gap

2017-06-23 Thread Rik van Riel 
On Fri, 2017-06-23 at 10:35 +0200, Ingo Molnar wrote:
> * r...@redhat.com  wrote:
> 
> > When RLIMIT_STACK is larger than the minimum gap enforced by
> > mmap_base(), it is possible for the kernel to place the mmap
> > area where the stack wants to grow, resulting in the stack
> > not being able to use the space that should have been allocated
> > to it through RLIMIT_STACK.
> > 
> > This series ensures that x86, ARM64, and PPC have at least
> > RLIMIT_STACK + stack randomization + the stack guard gap
> > space available for the stack.
> > 
> > s390 seems to be ok. I have not checked other architectures.
> 
> x86 patch LGTM:
> 
>   Acked-by: Ingo Molnar 
> 
> ... but I suspect this wants to go via -mm or Linus directly?

I believe Andrew picked it up yesterday.

Re: [PATCH 0/3] properly account for stack randomization and guard gap

2017-06-23 Thread Ingo Molnar 

* r...@redhat.com  wrote:

> When RLIMIT_STACK is larger than the minimum gap enforced by
> mmap_base(), it is possible for the kernel to place the mmap
> area where the stack wants to grow, resulting in the stack
> not being able to use the space that should have been allocated
> to it through RLIMIT_STACK.
> 
> This series ensures that x86, ARM64, and PPC have at least
> RLIMIT_STACK + stack randomization + the stack guard gap
> space available for the stack.
> 
> s390 seems to be ok. I have not checked other architectures.

x86 patch LGTM:

  Acked-by: Ingo Molnar 

... but I suspect this wants to go via -mm or Linus directly?

Thanks,

Ingo

Re: [PATCH 0/3] properly account for stack randomization and guard gap

2017-06-23 Thread Ingo Molnar 

* r...@redhat.com  wrote:

> When RLIMIT_STACK is larger than the minimum gap enforced by
> mmap_base(), it is possible for the kernel to place the mmap
> area where the stack wants to grow, resulting in the stack
> not being able to use the space that should have been allocated
> to it through RLIMIT_STACK.
> 
> This series ensures that x86, ARM64, and PPC have at least
> RLIMIT_STACK + stack randomization + the stack guard gap
> space available for the stack.
> 
> s390 seems to be ok. I have not checked other architectures.

x86 patch LGTM:

  Acked-by: Ingo Molnar 

... but I suspect this wants to go via -mm or Linus directly?

Thanks,

Ingo

[PATCH 0/3] properly account for stack randomization and guard gap

2017-06-22 Thread riel 
When RLIMIT_STACK is larger than the minimum gap enforced by
mmap_base(), it is possible for the kernel to place the mmap
area where the stack wants to grow, resulting in the stack
not being able to use the space that should have been allocated
to it through RLIMIT_STACK.

This series ensures that x86, ARM64, and PPC have at least
RLIMIT_STACK + stack randomization + the stack guard gap
space available for the stack.

s390 seems to be ok. I have not checked other architectures.

[PATCH 0/3] properly account for stack randomization and guard gap

2017-06-22 Thread riel 
When RLIMIT_STACK is larger than the minimum gap enforced by
mmap_base(), it is possible for the kernel to place the mmap
area where the stack wants to grow, resulting in the stack
not being able to use the space that should have been allocated
to it through RLIMIT_STACK.

This series ensures that x86, ARM64, and PPC have at least
RLIMIT_STACK + stack randomization + the stack guard gap
space available for the stack.

s390 seems to be ok. I have not checked other architectures.