Skip to contentSkip to conversation

p2m/ept: pass correct level to atomic_write_ept_entry in ept_invalidate_emt

PATCH In review Roger Pau Monne 3 participants 8 messages Aug 20, 2019 – Sep 4, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 p2m/ept: pass correct level to atomic_write_ept_entry in ept_invalidate_emt

The level passed to ept_invalidate_emt corresponds to the EPT entry passed as the mfn parameter, which is a pointer to an EPT page table, hence the entries in that page table will have one level less than the parent.

Fix the call to atomic_write_ept_entry to pass the correct level, ie:
one level less than the parent.

Fixes 50fe6e73705 ('pvh dom0: add and remove foreign pages')

---
Cc: Jun Nakajima <[email protected]>
Cc: Kevin Tian <[email protected]>
Cc: George Dunlap <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
Cc: Sergey Dyasli <[email protected]>
Cc: Paul Durrant <[email protected]>

---
 xen/arch/x86/mm/p2m-ept.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
xen/arch/x86/mm/p2m-ept.c +1 -1
index 6b8468c793..23ae6e9da3 100644
@@ -367,7 +367,7 @@ static bool_t ept_invalidate_emt(struct p2m_domain *p2m, mfn_t mfn,
         e.emt = MTRR_NUM_TYPES;
         if ( recalc )
             e.recalc = 1;
-        rc = atomic_write_ept_entry(p2m, &epte[i], e, level);
+        rc = atomic_write_ept_entry(p2m, &epte[i], e, level - 1);
         ASSERT(rc == 0);
         changed = 1;
     }
Signature · 7 lines
2.22.0


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Reviewed-by on v1 1/1
#1 lore raw
Quoted 39 lines

From: Roger Pau Monne [mailto:[email protected]]
Sent: Tuesday, August 20, 2019 11:38 PM

The level passed to ept_invalidate_emt corresponds to the EPT entry passed as the mfn parameter, which is a pointer to an EPT page table, hence the entries in that page table will have one level less than the parent.

Fix the call to atomic_write_ept_entry to pass the correct level, ie:
one level less than the parent.

Fixes 50fe6e73705 ('pvh dom0: add and remove foreign pages')

---
Cc: Jun Nakajima <[email protected]>
Cc: Kevin Tian <[email protected]>
Cc: George Dunlap <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
Cc: Sergey Dyasli <[email protected]>
Cc: Paul Durrant <[email protected]>

---
 xen/arch/x86/mm/p2m-ept.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
xen/arch/x86/mm/p2m-ept.c +1 -1
index 6b8468c793..23ae6e9da3 100644
@@ -367,7 +367,7 @@ static bool_t ept_invalidate_emt(struct p2m_domain
 *p2m, mfn_t mfn,
         e.emt = MTRR_NUM_TYPES;
         if ( recalc )
             e.recalc = 1;
-        rc = atomic_write_ept_entry(p2m, &epte[i], e, level);
+        rc = atomic_write_ept_entry(p2m, &epte[i], e, level - 1);
         ASSERT(rc == 0);
         changed = 1;
Quoted 1 line
     }

One small comment about readability. What about renaming 'level' to 'parent_level' in this function?

Thanks
Kevin
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#2 lore raw
Quoted 43 lines · Kevin
Quoted 39 lines

From: Roger Pau Monne [mailto:[email protected]]
Sent: Tuesday, August 20, 2019 11:38 PM

The level passed to ept_invalidate_emt corresponds to the EPT entry passed as the mfn parameter, which is a pointer to an EPT page table, hence the entries in that page table will have one level less than the parent.

Fix the call to atomic_write_ept_entry to pass the correct level, ie:
one level less than the parent.

Fixes 50fe6e73705 ('pvh dom0: add and remove foreign pages')

---
Cc: Jun Nakajima <[email protected]>
Cc: Kevin Tian <[email protected]>
Cc: George Dunlap <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
Cc: Sergey Dyasli <[email protected]>
Cc: Paul Durrant <[email protected]>

---
 xen/arch/x86/mm/p2m-ept.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
xen/arch/x86/mm/p2m-ept.c +1 -1
index 6b8468c793..23ae6e9da3 100644
@@ -367,7 +367,7 @@ static bool_t ept_invalidate_emt(struct p2m_domain
 *p2m, mfn_t mfn,
         e.emt = MTRR_NUM_TYPES;
         if ( recalc )
             e.recalc = 1;
-        rc = atomic_write_ept_entry(p2m, &epte[i], e, level);
+        rc = atomic_write_ept_entry(p2m, &epte[i], e, level - 1);
         ASSERT(rc == 0);
         changed = 1;
Quoted 1 line
     }
Quoted 2 lines

One small comment about readability. What about renaming 'level' to 'parent_level' in this function?

Sure, I guess this can be done while committing it, or else I can send a follow up.

Thanks, Roger.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#3 lore raw
In reply to Tian, Kevin
Quoted 43 lines · Kevin
Quoted 39 lines

From: Roger Pau Monne [mailto:[email protected]]
Sent: Tuesday, August 20, 2019 11:38 PM

The level passed to ept_invalidate_emt corresponds to the EPT entry passed as the mfn parameter, which is a pointer to an EPT page table, hence the entries in that page table will have one level less than the parent.

Fix the call to atomic_write_ept_entry to pass the correct level, ie:
one level less than the parent.

Fixes 50fe6e73705 ('pvh dom0: add and remove foreign pages')

---
Cc: Jun Nakajima <[email protected]>
Cc: Kevin Tian <[email protected]>
Cc: George Dunlap <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
Cc: Sergey Dyasli <[email protected]>
Cc: Paul Durrant <[email protected]>

---
  xen/arch/x86/mm/p2m-ept.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
xen/arch/x86/mm/p2m-ept.c +1 -1
index 6b8468c793..23ae6e9da3 100644
@@ -367,7 +367,7 @@ static bool_t ept_invalidate_emt(struct p2m_domain
 *p2m, mfn_t mfn,
          e.emt = MTRR_NUM_TYPES;
          if ( recalc )
              e.recalc = 1;
-        rc = atomic_write_ept_entry(p2m, &epte[i], e, level);
+        rc = atomic_write_ept_entry(p2m, &epte[i], e, level - 1);
          ASSERT(rc == 0);
          changed = 1;
Quoted 1 line
      }
Quoted 2 lines

One small comment about readability. What about renaming 'level' to 'parent_level' in this function?

And also taking the opportunity and making it unsigned int?

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#4 lore raw
Quoted 44 lines · Jan Beulich
Quoted 43 lines · Kevin
Quoted 39 lines

From: Roger Pau Monne [mailto:[email protected]]
Sent: Tuesday, August 20, 2019 11:38 PM

The level passed to ept_invalidate_emt corresponds to the EPT entry passed as the mfn parameter, which is a pointer to an EPT page table, hence the entries in that page table will have one level less than the parent.

Fix the call to atomic_write_ept_entry to pass the correct level, ie:
one level less than the parent.

Fixes 50fe6e73705 ('pvh dom0: add and remove foreign pages')

---
Cc: Jun Nakajima <[email protected]>
Cc: Kevin Tian <[email protected]>
Cc: George Dunlap <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
Cc: Sergey Dyasli <[email protected]>
Cc: Paul Durrant <[email protected]>

---
  xen/arch/x86/mm/p2m-ept.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
xen/arch/x86/mm/p2m-ept.c +1 -1
index 6b8468c793..23ae6e9da3 100644
@@ -367,7 +367,7 @@ static bool_t ept_invalidate_emt(struct p2m_domain
 *p2m, mfn_t mfn,
          e.emt = MTRR_NUM_TYPES;
          if ( recalc )
              e.recalc = 1;
-        rc = atomic_write_ept_entry(p2m, &epte[i], e, level);
+        rc = atomic_write_ept_entry(p2m, &epte[i], e, level - 1);
          ASSERT(rc == 0);
          changed = 1;
Quoted 1 line
      }
Quoted 4 lines
Quoted 2 lines

One small comment about readability. What about renaming 'level' to 'parent_level' in this function?

And also taking the opportunity and making it unsigned int?

I've been thinking about this, and I'm not sure renaming level to parent_level is correct, level actually matches the level of the mfn also passed as a parameter, and hence it's correct. It's the function itself that actually iterates over the page table entries, and hence descends one level into the page tables.

ie: I could add something like ASSERT(level) to make sure no level 0 entries are passed to the function, but renaming doesn't seem correct to me.

Thanks, Roger.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#5 lore raw
Quoted 56 lines · Roger Pau Monné
Quoted 44 lines · Jan Beulich
Quoted 43 lines · Kevin
Quoted 39 lines

From: Roger Pau Monne [mailto:[email protected]]
Sent: Tuesday, August 20, 2019 11:38 PM

The level passed to ept_invalidate_emt corresponds to the EPT entry passed as the mfn parameter, which is a pointer to an EPT page table, hence the entries in that page table will have one level less than the parent.

Fix the call to atomic_write_ept_entry to pass the correct level, ie:
one level less than the parent.

Fixes 50fe6e73705 ('pvh dom0: add and remove foreign pages')

---
Cc: Jun Nakajima <[email protected]>
Cc: Kevin Tian <[email protected]>
Cc: George Dunlap <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
Cc: Sergey Dyasli <[email protected]>
Cc: Paul Durrant <[email protected]>

---
  xen/arch/x86/mm/p2m-ept.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
xen/arch/x86/mm/p2m-ept.c +1 -1
index 6b8468c793..23ae6e9da3 100644
@@ -367,7 +367,7 @@ static bool_t ept_invalidate_emt(struct p2m_domain
 *p2m, mfn_t mfn,
          e.emt = MTRR_NUM_TYPES;
          if ( recalc )
              e.recalc = 1;
-        rc = atomic_write_ept_entry(p2m, &epte[i], e, level);
+        rc = atomic_write_ept_entry(p2m, &epte[i], e, level - 1);
          ASSERT(rc == 0);
          changed = 1;
Quoted 1 line
      }
Quoted 4 lines
Quoted 2 lines

One small comment about readability. What about renaming 'level' to 'parent_level' in this function?

And also taking the opportunity and making it unsigned int?

I've been thinking about this, and I'm not sure renaming level to parent_level is correct, level actually matches the level of the mfn also passed as a parameter, and hence it's correct. It's the function itself that actually iterates over the page table entries, and hence descends one level into the page tables.

Quoted 3 lines

ie: I could add something like ASSERT(level) to make sure no level 0 entries are passed to the function, but renaming doesn't seem correct to me.

Hmm, I'm afraid I've made the change as requested by Kevin while committing. Personally I think either name is fine, but if Kevin agrees with your response, then maybe we should undo that adjustment.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#6 lore raw
In reply to Roger Pau Monné
Quoted 60 lines

From: Roger Pau Monné [mailto:[email protected]]
Sent: Thursday, August 29, 2019 6:26 PM

Quoted 27 lines · Jan Beulich
Quoted 26 lines · Kevin
Quoted 26 lines

From: Roger Pau Monne [mailto:[email protected]]
Sent: Tuesday, August 20, 2019 11:38 PM

The level passed to ept_invalidate_emt corresponds to the EPT entry passed as the mfn parameter, which is a pointer to an EPT page table, hence the entries in that page table will have one level less than the parent.

Fix the call to atomic_write_ept_entry to pass the correct level, ie:
one level less than the parent.

Fixes 50fe6e73705 ('pvh dom0: add and remove foreign pages')

---
Cc: Jun Nakajima <[email protected]>
Cc: Kevin Tian <[email protected]>
Cc: George Dunlap <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
Cc: Sergey Dyasli <[email protected]>
Cc: Paul Durrant <[email protected]>

---
  xen/arch/x86/mm/p2m-ept.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 5 lines
Quoted 5 lines

xen/arch/x86/mm/p2m-ept.c static bool_t ept_invalidate_emt(struct

Quoted 5 lines
xen/arch/x86/mm/p2m-ept.c +0 -0
index 6b8468c793..23ae6e9da3 100644
@@ -367,7 +367,7 @@ static bool_t ept_invalidate_emt(struct

p2m_domain

Quoted 10 lines
Quoted 10 lines
Quoted 9 lines
xen/arch/x86/mm/p2m-ept.c +1 -1
 *p2m, mfn_t mfn,
          e.emt = MTRR_NUM_TYPES;
          if ( recalc )
              e.recalc = 1;
-        rc = atomic_write_ept_entry(p2m, &epte[i], e, level);
+        rc = atomic_write_ept_entry(p2m, &epte[i], e, level - 1);
          ASSERT(rc == 0);
          changed = 1;
      }
Quoted 6 lines
Quoted 4 lines

One small comment about readability. What about renaming 'level' to 'parent_level' in this function?

And also taking the opportunity and making it unsigned int?

I've been thinking about this, and I'm not sure renaming level to parent_level is correct, level actually matches the level of the mfn also passed as a parameter, and hence it's correct. It's the function itself that actually iterates over the page table entries, and hence descends one level into the page tables.

Quoted 4 lines

ie: I could add something like ASSERT(level) to make sure no level 0 entries are passed to the function, but renaming doesn't seem correct to me.

The problem to me is that it is very obscure about what a 'level' actually means. Although I figured it out last time when reviewing this patch, now I'm getting confused again when looking at related code. e.g., you minus level by one in this patch when invoking atomic_write_ept_entry, while the latter increments the level by one when invoking p2m_entry_ modify. They are all about entry update according to the function name, but clearly the level means different thing. :/

specifically for this patch, maybe call ept_invalidate_emt_subtree can also improve readability a bit, along with ASSERT(level) thing?

Thanks
Kevin

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#7 lore raw
Quoted 65 lines · Kevin
Quoted 60 lines

From: Roger Pau Monné [mailto:[email protected]]
Sent: Thursday, August 29, 2019 6:26 PM

Quoted 27 lines · Jan Beulich
Quoted 26 lines · Kevin
Quoted 26 lines

From: Roger Pau Monne [mailto:[email protected]]
Sent: Tuesday, August 20, 2019 11:38 PM

The level passed to ept_invalidate_emt corresponds to the EPT entry passed as the mfn parameter, which is a pointer to an EPT page table, hence the entries in that page table will have one level less than the parent.

Fix the call to atomic_write_ept_entry to pass the correct level, ie:
one level less than the parent.

Fixes 50fe6e73705 ('pvh dom0: add and remove foreign pages')

---
Cc: Jun Nakajima <[email protected]>
Cc: Kevin Tian <[email protected]>
Cc: George Dunlap <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
Cc: Sergey Dyasli <[email protected]>
Cc: Paul Durrant <[email protected]>

---
  xen/arch/x86/mm/p2m-ept.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 5 lines
Quoted 5 lines

xen/arch/x86/mm/p2m-ept.c static bool_t ept_invalidate_emt(struct

Quoted 5 lines
xen/arch/x86/mm/p2m-ept.c +0 -0
index 6b8468c793..23ae6e9da3 100644
@@ -367,7 +367,7 @@ static bool_t ept_invalidate_emt(struct

p2m_domain

Quoted 10 lines
Quoted 10 lines
Quoted 9 lines
xen/arch/x86/mm/p2m-ept.c +1 -1
 *p2m, mfn_t mfn,
          e.emt = MTRR_NUM_TYPES;
          if ( recalc )
              e.recalc = 1;
-        rc = atomic_write_ept_entry(p2m, &epte[i], e, level);
+        rc = atomic_write_ept_entry(p2m, &epte[i], e, level - 1);
          ASSERT(rc == 0);
          changed = 1;
      }
Quoted 6 lines
Quoted 4 lines

One small comment about readability. What about renaming 'level' to 'parent_level' in this function?

And also taking the opportunity and making it unsigned int?

I've been thinking about this, and I'm not sure renaming level to parent_level is correct, level actually matches the level of the mfn also passed as a parameter, and hence it's correct. It's the function itself that actually iterates over the page table entries, and hence descends one level into the page tables.

Quoted 4 lines

ie: I could add something like ASSERT(level) to make sure no level 0 entries are passed to the function, but renaming doesn't seem correct to me.

Quoted 4 lines

The problem to me is that it is very obscure about what a 'level' actually means. Although I figured it out last time when reviewing this patch, now I'm getting confused again when looking at related code. e.g., you minus level by one in this patch when invoking atomic_write_ept_entry,

That minus one is because the level of EPT entry in the atomic_write_ept_entry call is one level less than the parent, which is the parameter of the function.

Quoted 3 lines

while the latter increments the level by one when invoking p2m_entry_ modify. They are all about entry update according to the function name, but clearly the level means different thing. :/

That's unfortunate, but NPT/shadow considers that level starts at 1 (ie: 4K page-table entries are level 1), while EPT consider that level starts at 0 (ie: 4K page-table entries are level 0). That's IMO a very bad choice, I guess no one realized before of this difference because level was always internal to NPT or EPT code, but no generic functions using such level parameter where being called from both implementations.

Quoted 2 lines

specifically for this patch, maybe call ept_invalidate_emt_subtree can also improve readability a bit, along with ASSERT(level) thing?

I agree, let me try to prepare a patch.

Thanks, Roger.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel