Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1681828 > unrolled thread

[RFC v5 34/38] procfs: display the protection-key number associated with a vma

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [RFC v5 34/38] procfs: display the protection-key number associated with a vma Ram Pai <linuxram@us.ibm.com> - 2017-07-05 23:30 +0200
    Re: [RFC v5 34/38] procfs: display the protection-key number  associated with a vma Anshuman Khandual <khandual@linux.vnet.ibm.com> - 2017-07-10 05:10 +0200
      Re: [RFC v5 34/38] procfs: display the protection-key number  associated with a vma Ram Pai <linuxram@us.ibm.com> - 2017-07-10 08:10 +0200
    Re: [RFC v5 34/38] procfs: display the protection-key number  associated with a vma Dave Hansen <dave.hansen@intel.com> - 2017-07-11 20:20 +0200
      Re: [RFC v5 34/38] procfs: display the protection-key number  associated with a vma Ram Pai <linuxram@us.ibm.com> - 2017-07-13 10:10 +0200
        Re: [RFC v5 34/38] procfs: display the protection-key number  associated with a vma Dave Hansen <dave.hansen@intel.com> - 2017-07-13 16:10 +0200
          Re: [RFC v5 34/38] procfs: display the protection-key number  associated with a vma Ram Pai <linuxram@us.ibm.com> - 2017-07-13 19:10 +0200

#1681828 — [RFC v5 34/38] procfs: display the protection-key number associated with a vma

Display the pkey number associated with the vma in smaps of a task.
The key will be seen as below:

ProtectionKey: 0

Signed-off-by: Ram Pai <linuxram@us.ibm.com>
---
 arch/powerpc/kernel/setup_64.c |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
index f35ff9d..ebc82b3 100644
--- a/arch/powerpc/kernel/setup_64.c
+++ b/arch/powerpc/kernel/setup_64.c
@@ -37,6 +37,7 @@
 #include <linux/memblock.h>
 #include <linux/memory.h>
 #include <linux/nmi.h>
+#include <linux/pkeys.h>
 
 #include <asm/io.h>
 #include <asm/kdump.h>
@@ -745,3 +746,10 @@ static int __init disable_hardlockup_detector(void)
 }
 early_initcall(disable_hardlockup_detector);
 #endif
+
+#ifdef CONFIG_PPC64_MEMORY_PROTECTION_KEYS
+void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
+{
+	seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
+}
+#endif /* CONFIG_PPC64_MEMORY_PROTECTION_KEYS */
-- 
1.7.1

[toc] | [next] | [standalone]

#1683922 — Re: [RFC v5 34/38] procfs: display the protection-key number associated with a vma

On 07/06/2017 02:52 AM, Ram Pai wrote:
> Display the pkey number associated with the vma in smaps of a task.
> The key will be seen as below:
> 
> ProtectionKey: 0
> 
> Signed-off-by: Ram Pai <linuxram@us.ibm.com>
> ---
>  arch/powerpc/kernel/setup_64.c |    8 ++++++++
>  1 files changed, 8 insertions(+), 0 deletions(-)
> 
> diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
> index f35ff9d..ebc82b3 100644
> --- a/arch/powerpc/kernel/setup_64.c
> +++ b/arch/powerpc/kernel/setup_64.c
> @@ -37,6 +37,7 @@
>  #include <linux/memblock.h>
>  #include <linux/memory.h>
>  #include <linux/nmi.h>
> +#include <linux/pkeys.h>
>  
>  #include <asm/io.h>
>  #include <asm/kdump.h>
> @@ -745,3 +746,10 @@ static int __init disable_hardlockup_detector(void)
>  }
>  early_initcall(disable_hardlockup_detector);
>  #endif
> +
> +#ifdef CONFIG_PPC64_MEMORY_PROTECTION_KEYS

Why not for X86 protection keys ?

[toc] | [prev] | [next] | [standalone]

#1683975 — Re: [RFC v5 34/38] procfs: display the protection-key number associated with a vma

On Mon, Jul 10, 2017 at 08:37:28AM +0530, Anshuman Khandual wrote:
> On 07/06/2017 02:52 AM, Ram Pai wrote:
> > Display the pkey number associated with the vma in smaps of a task.
> > The key will be seen as below:
> > 
> > ProtectionKey: 0
> > 
> > Signed-off-by: Ram Pai <linuxram@us.ibm.com>
> > ---
> >  arch/powerpc/kernel/setup_64.c |    8 ++++++++
> >  1 files changed, 8 insertions(+), 0 deletions(-)
> > 
> > diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
> > index f35ff9d..ebc82b3 100644
> > --- a/arch/powerpc/kernel/setup_64.c
> > +++ b/arch/powerpc/kernel/setup_64.c
> > @@ -37,6 +37,7 @@
> >  #include <linux/memblock.h>
> >  #include <linux/memory.h>
> >  #include <linux/nmi.h>
> > +#include <linux/pkeys.h>
> >  
> >  #include <asm/io.h>
> >  #include <asm/kdump.h>
> > @@ -745,3 +746,10 @@ static int __init disable_hardlockup_detector(void)
> >  }
> >  early_initcall(disable_hardlockup_detector);
> >  #endif
> > +
> > +#ifdef CONFIG_PPC64_MEMORY_PROTECTION_KEYS
> 
> Why not for X86 protection keys ?

hmm.. I dont understand the comment.


-- 
Ram Pai

[toc] | [prev] | [next] | [standalone]

#1685276 — Re: [RFC v5 34/38] procfs: display the protection-key number associated with a vma

On 07/05/2017 02:22 PM, Ram Pai wrote:
> +#ifdef CONFIG_PPC64_MEMORY_PROTECTION_KEYS
> +void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
> +{
> +	seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
> +}
> +#endif /* CONFIG_PPC64_MEMORY_PROTECTION_KEYS */

This seems like kinda silly unnecessary duplication.  Could we just put
this in the fs/proc/ code and #ifdef it on ARCH_HAS_PKEYS?

[toc] | [prev] | [next] | [standalone]

#1686356 — Re: [RFC v5 34/38] procfs: display the protection-key number associated with a vma

On Tue, Jul 11, 2017 at 11:13:56AM -0700, Dave Hansen wrote:
> On 07/05/2017 02:22 PM, Ram Pai wrote:
> > +#ifdef CONFIG_PPC64_MEMORY_PROTECTION_KEYS
> > +void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
> > +{
> > +	seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
> > +}
> > +#endif /* CONFIG_PPC64_MEMORY_PROTECTION_KEYS */
> 
> This seems like kinda silly unnecessary duplication.  Could we just put
> this in the fs/proc/ code and #ifdef it on ARCH_HAS_PKEYS?

Well x86 predicates it based on availability of X86_FEATURE_OSPKE.

powerpc doesn't need that check or any similar check. So trying to
generalize the code does not save much IMHO.

maybe have a seperate inline function that does
seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
and is called from x86 and powerpc's arch_show_smap()?
At least will keep the string format captured in 
one single place.

thoughts?
RP

[toc] | [prev] | [next] | [standalone]

#1686544 — Re: [RFC v5 34/38] procfs: display the protection-key number associated with a vma

On 07/13/2017 01:03 AM, Ram Pai wrote:
> On Tue, Jul 11, 2017 at 11:13:56AM -0700, Dave Hansen wrote:
>> On 07/05/2017 02:22 PM, Ram Pai wrote:
>>> +#ifdef CONFIG_PPC64_MEMORY_PROTECTION_KEYS
>>> +void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
>>> +{
>>> +	seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
>>> +}
>>> +#endif /* CONFIG_PPC64_MEMORY_PROTECTION_KEYS */
>>
>> This seems like kinda silly unnecessary duplication.  Could we just put
>> this in the fs/proc/ code and #ifdef it on ARCH_HAS_PKEYS?
> 
> Well x86 predicates it based on availability of X86_FEATURE_OSPKE.
> 
> powerpc doesn't need that check or any similar check. So trying to
> generalize the code does not save much IMHO.

I know all your hardware doesn't support it. :)

So, for instance, if you are running on a new POWER9 with radix page
tables, you will just always output "ProtectionKey: 0" in every VMA,
regardless?

> maybe have a seperate inline function that does
> seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
> and is called from x86 and powerpc's arch_show_smap()?
> At least will keep the string format captured in 
> one single place.

Now that we have two architectures, is there a strong reason we can't
just have an arch_pkeys_enabled(), and stick the seq_printf() back in
generic code?

[toc] | [prev] | [next] | [standalone]

#1686783 — Re: [RFC v5 34/38] procfs: display the protection-key number associated with a vma

On Thu, Jul 13, 2017 at 07:07:48AM -0700, Dave Hansen wrote:
> On 07/13/2017 01:03 AM, Ram Pai wrote:
> > On Tue, Jul 11, 2017 at 11:13:56AM -0700, Dave Hansen wrote:
> >> On 07/05/2017 02:22 PM, Ram Pai wrote:
> >>> +#ifdef CONFIG_PPC64_MEMORY_PROTECTION_KEYS
> >>> +void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
> >>> +{
> >>> +	seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
> >>> +}
> >>> +#endif /* CONFIG_PPC64_MEMORY_PROTECTION_KEYS */
> >>
> >> This seems like kinda silly unnecessary duplication.  Could we just put
> >> this in the fs/proc/ code and #ifdef it on ARCH_HAS_PKEYS?
> > 
> > Well x86 predicates it based on availability of X86_FEATURE_OSPKE.
> > 
> > powerpc doesn't need that check or any similar check. So trying to
> > generalize the code does not save much IMHO.
> 
> I know all your hardware doesn't support it. :)

Wow! you bring a good point which I had not considered yet. I need some
runtime checks for RPT.

But regardless, my above statement is still partially true. x86
predicates it based on availability of X86_FEATURE_OSPKE, and powerpc
should predicate it based on HPT. So we have our own
customized checks. Hence a unified function won't suffice.

> 
> So, for instance, if you are running on a new POWER9 with radix page
> tables, you will just always output "ProtectionKey: 0" in every VMA,
> regardless?
> 
> > maybe have a seperate inline function that does
> > seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
> > and is called from x86 and powerpc's arch_show_smap()?
> > At least will keep the string format captured in 
> > one single place.
> 
> Now that we have two architectures, is there a strong reason we can't
> just have an arch_pkeys_enabled(), and stick the seq_printf() back in
> generic code?

correct. that looks like the correct approach. Was trying to avoid
touching arch neutral code. But this approach will force me
do so. Will do.

-- 
Ram Pai

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web