[PATCH v2] mm/mprotect.c: don't imply PROT_EXEC on non-exec fs

[PATCH v2] mm/mprotect.c: don't imply PROT_EXEC on non-exec fs

[Piotr Kwapulinski]

The mprotect(PROT_READ) fails when called by the READ_IMPLIES_EXEC binary
on a memory mapped file located on non-exec fs. The mprotect does not
check whether fs is _executable_ or not. The PROT_EXEC flag is set
automatically even if a memory mapped file is located on non-exec fs.
Fix it by checking whether a memory mapped file is located on a non-exec
fs. If so the PROT_EXEC is not implied by the PROT_READ.
The implementation uses the VM_MAYEXEC flag set properly in mmap.
Now it is consistent with mmap.

I did the isolated tests (PT_GNU_STACK X/NX, multiple VMAs, X/NX fs).
I also patched the official 3.19.0-47-generic Ubuntu 14.04 kernel
and it seems to work.

Signed-off-by: Piotr Kwapulinski <kwapulinski.piotr@gmail.com>
---
The difference between v1 is that the prot variable is reset to
 reqprot for each loop iteration (thanks to Konstantin Khlebnikov for
pointing this out).
rier means "(current->personality & [R]EAD_[I]MPLIES_[E]XEC) &&
(prot & PROT_[R]EAD)".

 mm/mprotect.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/mm/mprotect.c b/mm/mprotect.c
index 8eb7bb4..1b9597f 100644
--- a/mm/mprotect.c
+++ b/mm/mprotect.c
@@ -352,10 +352,12 @@ fail:
 SYSCALL_DEFINE3(mprotect, unsigned long, start, size_t, len,
 		unsigned long, prot)
 {
-	unsigned long vm_flags, nstart, end, tmp, reqprot;
+	unsigned long nstart, end, tmp, reqprot;
 	struct vm_area_struct *vma, *prev;
 	int error = -EINVAL;
 	const int grows = prot & (PROT_GROWSDOWN|PROT_GROWSUP);
+	const bool rier = (current->personality & READ_IMPLIES_EXEC) &&
+				(prot & PROT_READ);
 	prot &= ~(PROT_GROWSDOWN|PROT_GROWSUP);
 	if (grows == (PROT_GROWSDOWN|PROT_GROWSUP)) /* can't be both */
 		return -EINVAL;
@@ -372,13 +374,6 @@ SYSCALL_DEFINE3(mprotect, unsigned long, start, size_t, len,
 		return -EINVAL;
 
 	reqprot = prot;
-	/*
-	 * Does the application expect PROT_READ to imply PROT_EXEC:
-	 */
-	if ((prot & PROT_READ) && (current->personality & READ_IMPLIES_EXEC))
-		prot |= PROT_EXEC;
-
-	vm_flags = calc_vm_prot_bits(prot);
 
 	down_write(&current->mm->mmap_sem);
 
@@ -412,7 +407,11 @@ SYSCALL_DEFINE3(mprotect, unsigned long, start, size_t, len,
 
 		/* Here we know that vma->vm_start <= nstart < vma->vm_end. */
 
-		newflags = vm_flags;
+		/* Does the application expect PROT_READ to imply PROT_EXEC */
+		if (rier && (vma->vm_flags & VM_MAYEXEC))
+			prot |= PROT_EXEC;
+
+		newflags = calc_vm_prot_bits(prot);
 		newflags |= (vma->vm_flags & ~(VM_READ | VM_WRITE | VM_EXEC));
 
 		/* newflags >> 4 shift VM_MAY% in place of VM_% */
@@ -443,6 +442,7 @@ SYSCALL_DEFINE3(mprotect, unsigned long, start, size_t, len,
 			error = -ENOMEM;
 			goto out;
 		}
+		prot = reqprot;
 	}
 out:
 	up_write(&current->mm->mmap_sem);
-- 
2.7.0

Re: [PATCH v2] mm/mprotect.c: don't imply PROT_EXEC on non-exec fs

[Andrew Morton]

On Wed, 27 Jan 2016 17:29:37 +0100 Piotr Kwapulinski <kwapulinski.piotr@gmail.com> wrote:

> The mprotect(PROT_READ) fails when called by the READ_IMPLIES_EXEC binary
> on a memory mapped file located on non-exec fs. The mprotect does not
> check whether fs is _executable_ or not. The PROT_EXEC flag is set
> automatically even if a memory mapped file is located on non-exec fs.
> Fix it by checking whether a memory mapped file is located on a non-exec
> fs. If so the PROT_EXEC is not implied by the PROT_READ.
> The implementation uses the VM_MAYEXEC flag set properly in mmap.
> Now it is consistent with mmap.
> 
> I did the isolated tests (PT_GNU_STACK X/NX, multiple VMAs, X/NX fs).
> I also patched the official 3.19.0-47-generic Ubuntu 14.04 kernel
> and it seems to work.

sys_mprotect() just took a mangling in linux-next due to 

commit 62b5f7d013fc455b8db26cf01e421f4c0d264b92
Author:     Dave Hansen <dave.hansen@linux.intel.com>
AuthorDate: Fri Feb 12 13:02:40 2016 -0800
Commit:     Ingo Molnar <mingo@kernel.org>
CommitDate: Thu Feb 18 19:46:33 2016 +0100

    mm/core, x86/mm/pkeys: Add execute-only protection keys support


Here is my rework of your "mm/mprotect.c: don't imply PROT_EXEC on
non-exec fs" to handle this.  Please check very carefully.


From: Piotr Kwapulinski <kwapulinski.piotr@gmail.com>
Subject: mm/mprotect.c: don't imply PROT_EXEC on non-exec fs

The mprotect(PROT_READ) fails when called by the READ_IMPLIES_EXEC binary
on a memory mapped file located on non-exec fs.  The mprotect does not
check whether fs is _executable_ or not.  The PROT_EXEC flag is set
automatically even if a memory mapped file is located on non-exec fs.  Fix
it by checking whether a memory mapped file is located on a non-exec fs. 
If so the PROT_EXEC is not implied by the PROT_READ.  The implementation
uses the VM_MAYEXEC flag set properly in mmap.  Now it is consistent with
mmap.

I did the isolated tests (PT_GNU_STACK X/NX, multiple VMAs, X/NX fs).  I
also patched the official 3.19.0-47-generic Ubuntu 14.04 kernel and it
seems to work.

Signed-off-by: Piotr Kwapulinski <kwapulinski.piotr@gmail.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 mm/mprotect.c |   13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff -puN mm/mprotect.c~mm-mprotectc-dont-imply-prot_exec-on-non-exec-fs mm/mprotect.c
--- a/mm/mprotect.c~mm-mprotectc-dont-imply-prot_exec-on-non-exec-fs
+++ a/mm/mprotect.c
@@ -359,6 +359,9 @@ SYSCALL_DEFINE3(mprotect, unsigned long,
 	struct vm_area_struct *vma, *prev;
 	int error = -EINVAL;
 	const int grows = prot & (PROT_GROWSDOWN|PROT_GROWSUP);
+	const bool rier = (current->personality & READ_IMPLIES_EXEC) &&
+				(prot & PROT_READ);
+
 	prot &= ~(PROT_GROWSDOWN|PROT_GROWSUP);
 	if (grows == (PROT_GROWSDOWN|PROT_GROWSUP)) /* can't be both */
 		return -EINVAL;
@@ -375,11 +378,6 @@ SYSCALL_DEFINE3(mprotect, unsigned long,
 		return -EINVAL;
 
 	reqprot = prot;
-	/*
-	 * Does the application expect PROT_READ to imply PROT_EXEC:
-	 */
-	if ((prot & PROT_READ) && (current->personality & READ_IMPLIES_EXEC))
-		prot |= PROT_EXEC;
 
 	down_write(&current->mm->mmap_sem);
 
@@ -414,6 +412,10 @@ SYSCALL_DEFINE3(mprotect, unsigned long,
 
 		/* Here we know that vma->vm_start <= nstart < vma->vm_end. */
 
+		/* Does the application expect PROT_READ to imply PROT_EXEC */
+		if (rier && (vma->vm_flags & VM_MAYEXEC))
+			prot |= PROT_EXEC;
+
 		newflags = calc_vm_prot_bits(prot, pkey);
 		newflags |= (vma->vm_flags & ~(VM_READ | VM_WRITE | VM_EXEC));
 
@@ -445,6 +447,7 @@ SYSCALL_DEFINE3(mprotect, unsigned long,
 			error = -ENOMEM;
 			goto out;
 		}
+		prot = reqprot;
 	}
 out:
 	up_write(&current->mm->mmap_sem);
_

Re: [PATCH v2] mm/mprotect.c: don't imply PROT_EXEC on non-exec fs

[Piotr Kwapulinski]

On Fri, Feb 26, 2016 at 12:20:32PM -0800, Andrew Morton wrote:
> On Wed, 27 Jan 2016 17:29:37 +0100 Piotr Kwapulinski <kwapulinski.piotr@gmail.com> wrote:
> 
> > The mprotect(PROT_READ) fails when called by the READ_IMPLIES_EXEC binary
> > on a memory mapped file located on non-exec fs. The mprotect does not
> > check whether fs is _executable_ or not. The PROT_EXEC flag is set
> > automatically even if a memory mapped file is located on non-exec fs.
> > Fix it by checking whether a memory mapped file is located on a non-exec
> > fs. If so the PROT_EXEC is not implied by the PROT_READ.
> > The implementation uses the VM_MAYEXEC flag set properly in mmap.
> > Now it is consistent with mmap.
> > 
> > I did the isolated tests (PT_GNU_STACK X/NX, multiple VMAs, X/NX fs).
> > I also patched the official 3.19.0-47-generic Ubuntu 14.04 kernel
> > and it seems to work.
> 
> sys_mprotect() just took a mangling in linux-next due to 
> 
> commit 62b5f7d013fc455b8db26cf01e421f4c0d264b92
> Author:     Dave Hansen <dave.hansen@linux.intel.com>
> AuthorDate: Fri Feb 12 13:02:40 2016 -0800
> Commit:     Ingo Molnar <mingo@kernel.org>
> CommitDate: Thu Feb 18 19:46:33 2016 +0100
> 
>     mm/core, x86/mm/pkeys: Add execute-only protection keys support
> 
> 
> Here is my rework of your "mm/mprotect.c: don't imply PROT_EXEC on
> non-exec fs" to handle this.  Please check very carefully.
> 
> 
> From: Piotr Kwapulinski <kwapulinski.piotr@gmail.com>
> Subject: mm/mprotect.c: don't imply PROT_EXEC on non-exec fs
> 
> The mprotect(PROT_READ) fails when called by the READ_IMPLIES_EXEC binary
> on a memory mapped file located on non-exec fs.  The mprotect does not
> check whether fs is _executable_ or not.  The PROT_EXEC flag is set
> automatically even if a memory mapped file is located on non-exec fs.  Fix
> it by checking whether a memory mapped file is located on a non-exec fs. 
> If so the PROT_EXEC is not implied by the PROT_READ.  The implementation
> uses the VM_MAYEXEC flag set properly in mmap.  Now it is consistent with
> mmap.
> 
> I did the isolated tests (PT_GNU_STACK X/NX, multiple VMAs, X/NX fs).  I
> also patched the official 3.19.0-47-generic Ubuntu 14.04 kernel and it
> seems to work.
> 
> Signed-off-by: Piotr Kwapulinski <kwapulinski.piotr@gmail.com>
> Cc: Mel Gorman <mgorman@suse.de>
> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> Cc: Konstantin Khlebnikov <koct9i@gmail.com>
> Cc: Dan Williams <dan.j.williams@intel.com>
> Cc: Dave Hansen <dave.hansen@linux.intel.com>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> ---
> 
>  mm/mprotect.c |   13 ++++++++-----
>  1 file changed, 8 insertions(+), 5 deletions(-)
> 
> diff -puN mm/mprotect.c~mm-mprotectc-dont-imply-prot_exec-on-non-exec-fs mm/mprotect.c
> --- a/mm/mprotect.c~mm-mprotectc-dont-imply-prot_exec-on-non-exec-fs
> +++ a/mm/mprotect.c
> @@ -359,6 +359,9 @@ SYSCALL_DEFINE3(mprotect, unsigned long,
>  	struct vm_area_struct *vma, *prev;
>  	int error = -EINVAL;
>  	const int grows = prot & (PROT_GROWSDOWN|PROT_GROWSUP);
> +	const bool rier = (current->personality & READ_IMPLIES_EXEC) &&
> +				(prot & PROT_READ);
> +
>  	prot &= ~(PROT_GROWSDOWN|PROT_GROWSUP);
>  	if (grows == (PROT_GROWSDOWN|PROT_GROWSUP)) /* can't be both */
>  		return -EINVAL;
> @@ -375,11 +378,6 @@ SYSCALL_DEFINE3(mprotect, unsigned long,
>  		return -EINVAL;
>  
>  	reqprot = prot;
> -	/*
> -	 * Does the application expect PROT_READ to imply PROT_EXEC:
> -	 */
> -	if ((prot & PROT_READ) && (current->personality & READ_IMPLIES_EXEC))
> -		prot |= PROT_EXEC;
>  
>  	down_write(&current->mm->mmap_sem);
>  
> @@ -414,6 +412,10 @@ SYSCALL_DEFINE3(mprotect, unsigned long,
>  
>  		/* Here we know that vma->vm_start <= nstart < vma->vm_end. */
>  
> +		/* Does the application expect PROT_READ to imply PROT_EXEC */
> +		if (rier && (vma->vm_flags & VM_MAYEXEC))
> +			prot |= PROT_EXEC;
> +
>  		newflags = calc_vm_prot_bits(prot, pkey);
>  		newflags |= (vma->vm_flags & ~(VM_READ | VM_WRITE | VM_EXEC));
>  
> @@ -445,6 +447,7 @@ SYSCALL_DEFINE3(mprotect, unsigned long,
>  			error = -ENOMEM;
>  			goto out;
>  		}
> +		prot = reqprot;
>  	}
>  out:
>  	up_write(&current->mm->mmap_sem);
> _
> 
It looks good. I also did some tests (non-MPK CPU) - passed.
Thank you.