[PATCH v4 0/2] Fix issues with vmalloc flush flag

[PATCH v4 0/2] Fix issues with vmalloc flush flag

[Rick Edgecombe]

These two patches address issues with the recently added
VM_FLUSH_RESET_PERMS vmalloc flag.

Patch 1 is the most critical and addresses an issue that could cause a
crash on x86.

Patch 2 addresses an issue where in a rare case strange arguments
could be provided to flush_tlb_kernel_range(). Most arch's should be
fine with it, but some may not like them, so we should avoid doing
this.

v3->v4:
 - Drop patch that switched vm_unmap_alias() calls to regular flush
 - Add patch to address correctness previously fixed in dropped patch

v2->v3:
 - Split into two patches

v1->v2:
 - Update commit message with more detail
 - Fix flush end range on !CONFIG_ARCH_HAS_SET_DIRECT_MAP case

Rick Edgecombe (2):
  vmalloc: Fix calculation of direct map addr range
  vmalloc: Avoid rare case of flushing tlb with weird arguements

 mm/vmalloc.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

-- 
2.20.1

[PATCH v4 1/2] vmalloc: Fix calculation of direct map addr range

[Rick Edgecombe]

The calculation of the direct map address range to flush was wrong.
This could cause problems on x86 if a RO direct map alias ever got loaded
into the TLB. This shouldn't normally happen, but it could cause the
permissions to remain RO on the direct map alias, and then the page
would return from the page allocator to some other component as RO and
cause a crash.

So fix fix the address range calculation so the flush will include the
direct map range.

Fixes: 868b104d7379 ("mm/vmalloc: Add flag for freeing of special permsissions")
Cc: Meelis Roos <mroos@linux.ee>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Nadav Amit <namit@vmware.com>
Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
---
 mm/vmalloc.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index c42872ed82ac..836888ae01f6 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -2159,9 +2159,10 @@ static void vm_remove_mappings(struct vm_struct *area, int deallocate_pages)
 	 * the vm_unmap_aliases() flush includes the direct map.
 	 */
 	for (i = 0; i < area->nr_pages; i++) {
-		if (page_address(area->pages[i])) {
+		addr = (unsigned long)page_address(area->pages[i]);
+		if (addr) {
 			start = min(addr, start);
-			end = max(addr, end);
+			end = max(addr + PAGE_SIZE, end);
 		}
 	}
 
-- 
2.20.1

[PATCH v4 2/2] vmalloc: Avoid rare case of flushing tlb with weird arguements

[Rick Edgecombe]

In a rare case, flush_tlb_kernel_range() could be called with a start
higher than the end. Most architectures should be fine with with this, but
some may not like it, so avoid doing this.

In vm_remove_mappings(), in case page_address() returns 0 for all pages,
_vm_unmap_aliases() will be called with start = ULONG_MAX, end = 0 and
flush = 1.

If at the same time, the vmalloc purge operation is triggered by something
else while the current operation is between remove_vm_area() and
_vm_unmap_aliases(), then the vm mapping just removed will be already
purged. In this case the call of vm_unmap_aliases() may not find any other
mappings to flush and so end up flushing start = ULONG_MAX, end = 0. So
only set flush = true if we find something in the direct mapping that we
need to flush, and this way this can't happen.

Fixes: 868b104d7379 ("mm/vmalloc: Add flag for freeing of special permsissions")
Cc: Meelis Roos <mroos@linux.ee>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Nadav Amit <namit@vmware.com>
Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
---
 mm/vmalloc.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 836888ae01f6..537d1134b40e 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -2125,6 +2125,7 @@ static void vm_remove_mappings(struct vm_struct *area, int deallocate_pages)
 	unsigned long addr = (unsigned long)area->addr;
 	unsigned long start = ULONG_MAX, end = 0;
 	int flush_reset = area->flags & VM_FLUSH_RESET_PERMS;
+	int flush_dmap = 0;
 	int i;
 
 	/*
@@ -2163,6 +2164,7 @@ static void vm_remove_mappings(struct vm_struct *area, int deallocate_pages)
 		if (addr) {
 			start = min(addr, start);
 			end = max(addr + PAGE_SIZE, end);
+			flush_dmap = 1;
 		}
 	}
 
@@ -2172,7 +2174,7 @@ static void vm_remove_mappings(struct vm_struct *area, int deallocate_pages)
 	 * reset the direct map permissions to the default.
 	 */
 	set_area_direct_map(area, set_direct_map_invalid_noflush);
-	_vm_unmap_aliases(start, end, 1);
+	_vm_unmap_aliases(start, end, flush_dmap);
 	set_area_direct_map(area, set_direct_map_default_noflush);
 }
 
-- 
2.20.1

Re: [PATCH v4 1/2] vmalloc: Fix calculation of direct map addr range

[Peter Zijlstra]

On Tue, May 21, 2019 at 01:51:36PM -0700, Rick Edgecombe wrote:
> The calculation of the direct map address range to flush was wrong.
> This could cause problems on x86 if a RO direct map alias ever got loaded
> into the TLB. This shouldn't normally happen, but it could cause the
> permissions to remain RO on the direct map alias, and then the page
> would return from the page allocator to some other component as RO and
> cause a crash.
> 
> So fix fix the address range calculation so the flush will include the
> direct map range.
> 
> Fixes: 868b104d7379 ("mm/vmalloc: Add flag for freeing of special permsissions")
> Cc: Meelis Roos <mroos@linux.ee>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Dave Hansen <dave.hansen@intel.com>
> Cc: Borislav Petkov <bp@alien8.de>
> Cc: Andy Lutomirski <luto@kernel.org>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Nadav Amit <namit@vmware.com>
> Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
> ---
>  mm/vmalloc.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index c42872ed82ac..836888ae01f6 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2159,9 +2159,10 @@ static void vm_remove_mappings(struct vm_struct *area, int deallocate_pages)
>  	 * the vm_unmap_aliases() flush includes the direct map.
>  	 */
>  	for (i = 0; i < area->nr_pages; i++) {
> -		if (page_address(area->pages[i])) {
> +		addr = (unsigned long)page_address(area->pages[i]);
> +		if (addr) {
>  			start = min(addr, start);
> -			end = max(addr, end);
> +			end = max(addr + PAGE_SIZE, end);
>  		}
>  	}
>  

Indeed; howevr I'm thinking this bug was caused to exist by the dual use
of @addr in this function, so should we not, perhaps, do something like
the below instead?

Also; having looked at this, it makes me question the use of
flush_tlb_kernel_range() in _vm_unmap_aliases() and
__purge_vmap_area_lazy(), it's potentially combining multiple ranges,
which never really works well.

Arguably, we should just do flush_tlb_all() here, but that's for another
patch I'm thinking.

---
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -2123,7 +2123,6 @@ static inline void set_area_direct_map(c
 /* Handle removing and resetting vm mappings related to the vm_struct. */
 static void vm_remove_mappings(struct vm_struct *area, int deallocate_pages)
 {
-	unsigned long addr = (unsigned long)area->addr;
 	unsigned long start = ULONG_MAX, end = 0;
 	int flush_reset = area->flags & VM_FLUSH_RESET_PERMS;
 	int i;
@@ -2135,8 +2134,8 @@ static void vm_remove_mappings(struct vm
 	 * execute permissions, without leaving a RW+X window.
 	 */
 	if (flush_reset && !IS_ENABLED(CONFIG_ARCH_HAS_SET_DIRECT_MAP)) {
-		set_memory_nx(addr, area->nr_pages);
-		set_memory_rw(addr, area->nr_pages);
+		set_memory_nx((unsigned long)area->addr, area->nr_pages);
+		set_memory_rw((unsigned long)area->addr, area->nr_pages);
 	}
 
 	remove_vm_area(area->addr);
@@ -2160,9 +2159,10 @@ static void vm_remove_mappings(struct vm
 	 * the vm_unmap_aliases() flush includes the direct map.
 	 */
 	for (i = 0; i < area->nr_pages; i++) {
-		if (page_address(area->pages[i])) {
+		unsigned long addr = (unsigned long)page_address(area->pages[i]);
+		if (addr) {
 			start = min(addr, start);
-			end = max(addr, end);
+			end = max(addr + PAGE_SIZE, end);
 		}
 	}
 

Re: [PATCH v4 2/2] vmalloc: Avoid rare case of flushing tlb with weird arguements

[Peter Zijlstra]

On Tue, May 21, 2019 at 01:51:37PM -0700, Rick Edgecombe wrote:
> In a rare case, flush_tlb_kernel_range() could be called with a start
> higher than the end. Most architectures should be fine with with this, but
> some may not like it, so avoid doing this.
> 
> In vm_remove_mappings(), in case page_address() returns 0 for all pages,
> _vm_unmap_aliases() will be called with start = ULONG_MAX, end = 0 and
> flush = 1.
> 
> If at the same time, the vmalloc purge operation is triggered by something
> else while the current operation is between remove_vm_area() and
> _vm_unmap_aliases(), then the vm mapping just removed will be already
> purged. In this case the call of vm_unmap_aliases() may not find any other
> mappings to flush and so end up flushing start = ULONG_MAX, end = 0. So
> only set flush = true if we find something in the direct mapping that we
> need to flush, and this way this can't happen.
> 
> Fixes: 868b104d7379 ("mm/vmalloc: Add flag for freeing of special permsissions")
> Cc: Meelis Roos <mroos@linux.ee>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Dave Hansen <dave.hansen@intel.com>
> Cc: Borislav Petkov <bp@alien8.de>
> Cc: Andy Lutomirski <luto@kernel.org>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Nadav Amit <namit@vmware.com>
> Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
> ---
>  mm/vmalloc.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index 836888ae01f6..537d1134b40e 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2125,6 +2125,7 @@ static void vm_remove_mappings(struct vm_struct *area, int deallocate_pages)
>  	unsigned long addr = (unsigned long)area->addr;
>  	unsigned long start = ULONG_MAX, end = 0;
>  	int flush_reset = area->flags & VM_FLUSH_RESET_PERMS;
> +	int flush_dmap = 0;
>  	int i;
>  
>  	/*
> @@ -2163,6 +2164,7 @@ static void vm_remove_mappings(struct vm_struct *area, int deallocate_pages)
>  		if (addr) {
>  			start = min(addr, start);
>  			end = max(addr + PAGE_SIZE, end);
> +			flush_dmap = 1;
>  		}
>  	}
>  
> @@ -2172,7 +2174,7 @@ static void vm_remove_mappings(struct vm_struct *area, int deallocate_pages)
>  	 * reset the direct map permissions to the default.
>  	 */
>  	set_area_direct_map(area, set_direct_map_invalid_noflush);
> -	_vm_unmap_aliases(start, end, 1);
> +	_vm_unmap_aliases(start, end, flush_dmap);
>  	set_area_direct_map(area, set_direct_map_default_noflush);
>  }

Hurmph.. another clue that this range flushing is crap I feel. The phys
addrs of the page array can be scattered all over the place, a range
doesn't properly represent things.

But yes, this seems like a minimal fix in spirit with the existing code.

Re: [PATCH v4 1/2] vmalloc: Fix calculation of direct map addr range

[Edgecombe, Rick P]

On Mon, 2019-05-27 at 14:20 +0200, Peter Zijlstra wrote:
> On Tue, May 21, 2019 at 01:51:36PM -0700, Rick Edgecombe wrote:
> > The calculation of the direct map address range to flush was wrong.
> > This could cause problems on x86 if a RO direct map alias ever got
> > loaded
> > into the TLB. This shouldn't normally happen, but it could cause
> > the
> > permissions to remain RO on the direct map alias, and then the page
> > would return from the page allocator to some other component as RO
> > and
> > cause a crash.
> > 
> > So fix fix the address range calculation so the flush will include
> > the
> > direct map range.
> > 
> > Fixes: 868b104d7379 ("mm/vmalloc: Add flag for freeing of special
> > permsissions")
> > Cc: Meelis Roos <mroos@linux.ee>
> > Cc: Peter Zijlstra <peterz@infradead.org>
> > Cc: "David S. Miller" <davem@davemloft.net>
> > Cc: Dave Hansen <dave.hansen@intel.com>
> > Cc: Borislav Petkov <bp@alien8.de>
> > Cc: Andy Lutomirski <luto@kernel.org>
> > Cc: Ingo Molnar <mingo@redhat.com>
> > Cc: Nadav Amit <namit@vmware.com>
> > Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
> > ---
> >  mm/vmalloc.c | 5 +++--
> >  1 file changed, 3 insertions(+), 2 deletions(-)
> > 
> > diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> > index c42872ed82ac..836888ae01f6 100644
> > --- a/mm/vmalloc.c
> > +++ b/mm/vmalloc.c
> > @@ -2159,9 +2159,10 @@ static void vm_remove_mappings(struct
> > vm_struct *area, int deallocate_pages)
> >  	 * the vm_unmap_aliases() flush includes the direct map.
> >  	 */
> >  	for (i = 0; i < area->nr_pages; i++) {
> > -		if (page_address(area->pages[i])) {
> > +		addr = (unsigned long)page_address(area->pages[i]);
> > +		if (addr) {
> >  			start = min(addr, start);
> > -			end = max(addr, end);
> > +			end = max(addr + PAGE_SIZE, end);
> >  		}
> >  	}
> >  
> 
> Indeed; howevr I'm thinking this bug was caused to exist by the dual
> use
> of @addr in this function, so should we not, perhaps, do something
> like
> the below instead?
> 
> Also; having looked at this, it makes me question the use of
> flush_tlb_kernel_range() in _vm_unmap_aliases() and
> __purge_vmap_area_lazy(), it's potentially combining multiple ranges,
> which never really works well.
> 
> Arguably, we should just do flush_tlb_all() here, but that's for
> another
> patch I'm thinking.

Thanks. It mostly got broken implementing a style suggestion late in
the series. I'll change the addr variable around like you suggest to
make it more resistant.

The flush_tlb_all() suggestion makes sense to me, but I'll leave it for
now.

> ---
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2123,7 +2123,6 @@ static inline void set_area_direct_map(c
>  /* Handle removing and resetting vm mappings related to the
> vm_struct. */
>  static void vm_remove_mappings(struct vm_struct *area, int
> deallocate_pages)
>  {
> -	unsigned long addr = (unsigned long)area->addr;
>  	unsigned long start = ULONG_MAX, end = 0;
>  	int flush_reset = area->flags & VM_FLUSH_RESET_PERMS;
>  	int i;
> @@ -2135,8 +2134,8 @@ static void vm_remove_mappings(struct vm
>  	 * execute permissions, without leaving a RW+X window.
>  	 */
>  	if (flush_reset && !IS_ENABLED(CONFIG_ARCH_HAS_SET_DIRECT_MAP))
> {
> -		set_memory_nx(addr, area->nr_pages);
> -		set_memory_rw(addr, area->nr_pages);
> +		set_memory_nx((unsigned long)area->addr, area-
> >nr_pages);
> +		set_memory_rw((unsigned long)area->addr, area-
> >nr_pages);
>  	}
>  
>  	remove_vm_area(area->addr);
> @@ -2160,9 +2159,10 @@ static void vm_remove_mappings(struct vm
>  	 * the vm_unmap_aliases() flush includes the direct map.
>  	 */
>  	for (i = 0; i < area->nr_pages; i++) {
> -		if (page_address(area->pages[i])) {
> +		unsigned long addr = (unsigned long)page_address(area-
> >pages[i]);
> +		if (addr) {
>  			start = min(addr, start);
> -			end = max(addr, end);
> +			end = max(addr + PAGE_SIZE, end);
>  		}
>  	}
>