[PATCH] xen/p2m: check MFN is in range before using the m2p table

[PATCH] xen/p2m: check MFN is in range before using the m2p table

[David Vrabel]

From: David Vrabel <david.vrabel@citrix.com>

On hosts with more than 168 GB of memory, a 32-bit guest may attempt
to grant map an MFN that is not cannot lookup in its mapping of the
m2p table.  There is an m2p lookup as part of m2p_add_override() and
m2p_remove_override().  The lookup falls off the end of the mapped
portion of the m2p and (because the mapping is at the highest virtual
address) wraps around and the lookup causes a fault on what appears to
be a user space address.

do_page_fault() (thinking it's a fault to a userspace address), tries
to lock mm->mmap_sem.  If the gntdev device is used for the grant map,
m2p_add_override() is called from from gnttab_mmap() with mm->mmap_sem
already locked.  do_page_fault() then deadlocks.

The deadlock would most commonly occur when a 64-bit guest is started
and xenconsoled attempts to grant map its console ring.

Introduce mfn_to_pfn_no_overrides() which checks the MFN is within the
mapped portion of the m2p table before accessing the table and use
this in m2p_add_override(), m2p_remove_override(), and mfn_to_pfn()
(which already had the correct range check).

All faults caused by accessing the non-existant parts of the m2p are
thus within the kernel address space and exception_fixup() is called
without trying to lock mm->mmap_sem.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
---
 arch/x86/include/asm/xen/page.h |   32 +++++++++++++++++++-------------
 arch/x86/xen/p2m.c              |   10 ++++------
 2 files changed, 23 insertions(+), 19 deletions(-)

diff --git a/arch/x86/include/asm/xen/page.h b/arch/x86/include/asm/xen/page.h
index 6aef9fb..17c7b6b 100644
--- a/arch/x86/include/asm/xen/page.h
+++ b/arch/x86/include/asm/xen/page.h
@@ -79,30 +79,35 @@ static inline int phys_to_machine_mapping_valid(unsigned long pfn)
 	return get_phys_to_machine(pfn) != INVALID_P2M_ENTRY;
 }
 
-static inline unsigned long mfn_to_pfn(unsigned long mfn)
+static inline unsigned long mfn_to_pfn_no_overrides(unsigned long mfn)
 {
 	unsigned long pfn;
-	int ret = 0;
+	int ret;
 
-	if (xen_feature(XENFEAT_auto_translated_physmap))
-		return mfn;
+	if (unlikely(mfn >= machine_to_phys_nr))
+		return ~0;
 
-	if (unlikely(mfn >= machine_to_phys_nr)) {
-		pfn = ~0;
-		goto try_override;
-	}
-	pfn = 0;
 	/*
 	 * The array access can fail (e.g., device space beyond end of RAM).
 	 * In such cases it doesn't matter what we return (we return garbage),
 	 * but we must handle the fault without crashing!
 	 */
 	ret = __get_user(pfn, &machine_to_phys_mapping[mfn]);
-try_override:
-	/* ret might be < 0 if there are no entries in the m2p for mfn */
 	if (ret < 0)
-		pfn = ~0;
-	else if (get_phys_to_machine(pfn) != mfn)
+		return ~0;
+
+	return pfn;
+}
+
+static inline unsigned long mfn_to_pfn(unsigned long mfn)
+{
+	unsigned long pfn;
+
+	if (xen_feature(XENFEAT_auto_translated_physmap))
+		return mfn;
+
+	pfn = mfn_to_pfn_no_overrides(mfn);
+	if (pfn != ~0 && get_phys_to_machine(pfn) != mfn) {
 		/*
 		 * If this appears to be a foreign mfn (because the pfn
 		 * doesn't map back to the mfn), then check the local override
@@ -111,6 +116,7 @@ try_override:
 		 * m2p_find_override_pfn returns ~0 if it doesn't find anything.
 		 */
 		pfn = m2p_find_override_pfn(mfn, ~0);
+	}
 
 	/* 
 	 * pfn is ~0 if there are no entries in the m2p for mfn or if the
diff --git a/arch/x86/xen/p2m.c b/arch/x86/xen/p2m.c
index 95fb2aa..1563444 100644
--- a/arch/x86/xen/p2m.c
+++ b/arch/x86/xen/p2m.c
@@ -878,7 +878,6 @@ int m2p_add_override(unsigned long mfn, struct page *page,
 	unsigned long uninitialized_var(address);
 	unsigned level;
 	pte_t *ptep = NULL;
-	int ret = 0;
 
 	pfn = page_to_pfn(page);
 	if (!PageHighMem(page)) {
@@ -925,8 +924,8 @@ int m2p_add_override(unsigned long mfn, struct page *page,
 	 * frontend pages while they are being shared with the backend,
 	 * because mfn_to_pfn (that ends up being called by GUPF) will
 	 * return the backend pfn rather than the frontend pfn. */
-	ret = __get_user(pfn, &machine_to_phys_mapping[mfn]);
-	if (ret == 0 && get_phys_to_machine(pfn) == mfn)
+	pfn = mfn_to_pfn_no_overrides(mfn);
+	if (get_phys_to_machine(pfn) == mfn)
 		set_phys_to_machine(pfn, FOREIGN_FRAME(mfn));
 
 	return 0;
@@ -941,7 +940,6 @@ int m2p_remove_override(struct page *page,
 	unsigned long uninitialized_var(address);
 	unsigned level;
 	pte_t *ptep = NULL;
-	int ret = 0;
 
 	pfn = page_to_pfn(page);
 	mfn = get_phys_to_machine(pfn);
@@ -1019,8 +1017,8 @@ int m2p_remove_override(struct page *page,
 	 * the original pfn causes mfn_to_pfn(mfn) to return the frontend
 	 * pfn again. */
 	mfn &= ~FOREIGN_FRAME_BIT;
-	ret = __get_user(pfn, &machine_to_phys_mapping[mfn]);
-	if (ret == 0 && get_phys_to_machine(pfn) == FOREIGN_FRAME(mfn) &&
+	pfn = mfn_to_pfn_no_overrides(mfn);
+	if (get_phys_to_machine(pfn) == FOREIGN_FRAME(mfn) &&
 			m2p_find_override(mfn) == NULL)
 		set_phys_to_machine(pfn, mfn);
 
-- 
1.7.2.5

Re: [PATCH] xen/p2m: check MFN is in range before using the m2p table

[Jan Beulich]

>>> On 28.08.13 at 19:30, David Vrabel <david.vrabel@citrix.com> wrote:
> From: David Vrabel <david.vrabel@citrix.com>
> 
> On hosts with more than 168 GB of memory, a 32-bit guest may attempt
> to grant map an MFN that is not cannot lookup in its mapping of the
> m2p table.  There is an m2p lookup as part of m2p_add_override() and
> m2p_remove_override().  The lookup falls off the end of the mapped
> portion of the m2p and (because the mapping is at the highest virtual
> address) wraps around and the lookup causes a fault on what appears to
> be a user space address.
> 
> do_page_fault() (thinking it's a fault to a userspace address), tries
> to lock mm->mmap_sem.  If the gntdev device is used for the grant map,
> m2p_add_override() is called from from gnttab_mmap() with mm->mmap_sem
> already locked.  do_page_fault() then deadlocks.
> 
> The deadlock would most commonly occur when a 64-bit guest is started
> and xenconsoled attempts to grant map its console ring.
> 
> Introduce mfn_to_pfn_no_overrides() which checks the MFN is within the
> mapped portion of the m2p table before accessing the table and use
> this in m2p_add_override(), m2p_remove_override(), and mfn_to_pfn()
> (which already had the correct range check).
> 
> All faults caused by accessing the non-existant parts of the m2p are
> thus within the kernel address space and exception_fixup() is called
> without trying to lock mm->mmap_sem.
> 
> Signed-off-by: David Vrabel <david.vrabel@citrix.com>

This all looks quite fine to me, but iiuc it only removes the deadlock,
it doesn't make things work. In order to make it work I think we'd
need a hypercall (albeit even then it would work only up to 1Tb).
Otoh the question of course is whether driving this big a system
with a 32-bit Dom0 kernel is a reasonable thing in the first place.

Jan

Re: [PATCH] xen/p2m: check MFN is in range before using the m2p table

[David Vrabel]

On 29/08/13 08:07, Jan Beulich wrote:
>>>> On 28.08.13 at 19:30, David Vrabel <david.vrabel@citrix.com> wrote:
>> From: David Vrabel <david.vrabel@citrix.com>
>>
>> On hosts with more than 168 GB of memory, a 32-bit guest may attempt
>> to grant map an MFN that is not cannot lookup in its mapping of the
>> m2p table.  There is an m2p lookup as part of m2p_add_override() and
>> m2p_remove_override().  The lookup falls off the end of the mapped
>> portion of the m2p and (because the mapping is at the highest virtual
>> address) wraps around and the lookup causes a fault on what appears to
>> be a user space address.
>>
>> do_page_fault() (thinking it's a fault to a userspace address), tries
>> to lock mm->mmap_sem.  If the gntdev device is used for the grant map,
>> m2p_add_override() is called from from gnttab_mmap() with mm->mmap_sem
>> already locked.  do_page_fault() then deadlocks.
>>
>> The deadlock would most commonly occur when a 64-bit guest is started
>> and xenconsoled attempts to grant map its console ring.
>>
>> Introduce mfn_to_pfn_no_overrides() which checks the MFN is within the
>> mapped portion of the m2p table before accessing the table and use
>> this in m2p_add_override(), m2p_remove_override(), and mfn_to_pfn()
>> (which already had the correct range check).
>>
>> All faults caused by accessing the non-existant parts of the m2p are
>> thus within the kernel address space and exception_fixup() is called
>> without trying to lock mm->mmap_sem.
>>
>> Signed-off-by: David Vrabel <david.vrabel@citrix.com>
> 
> This all looks quite fine to me, but iiuc it only removes the deadlock,
> it doesn't make things work. In order to make it work I think we'd
> need a hypercall (albeit even then it would work only up to 1Tb).
> Otoh the question of course is whether driving this big a system
> with a 32-bit Dom0 kernel is a reasonable thing in the first place.

It should be sufficient as either an MFN is owned by this domain and
thus it is within the m2p mapping or it's a foreign MFN and it will be
looked up via the m2p overrides.

However, I note I got one of the tests wrong in mfn_to_pfn() and it's
broken.  It should be:

	pfn = mfn_to_pfn_no_overrides(mfn);
	if (pfn == ~0 || get_phys_to_machine(pfn) != mfn)
		pfn = m2p_find_override_pfn(mfn, ~0);

Or the equivalent (because get_phys_to_machine(~0) returns
INVALID_P2M_ENTRY which is != mfn:

	pfn = mfn_to_pfn_no_overrides(mfn);
	if (get_phys_to_machine(pfn) != mfn)
		pfn = m2p_find_override_pfn(mfn, ~0);

David