All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v8] xen/grant-table: Avoid m2p_override during mapping
@ 2014-02-06 12:56 Zoltan Kiss
  2014-02-07 18:42 ` David Vrabel
                   ` (3 more replies)
  0 siblings, 4 replies; 6+ messages in thread
From: Zoltan Kiss @ 2014-02-06 12:56 UTC (permalink / raw)
  To: Konrad Rzeszutek Wilk, Boris Ostrovsky, David Vrabel,
	Thomas Gleixner, Ingo Molnar, H. Peter Anvin, x86, xen-devel,
	linux-kernel, Roger Pau Monné,
	Jan Beulich, Ian Campbell, stefano.stabellini
  Cc: Zoltan Kiss

The grant mapping API does m2p_override unnecessarily: only gntdev needs it,
for blkback and future netback patches it just cause a lock contention, as
those pages never go to userspace. Therefore this series does the following:
- the original functions were renamed to __gnttab_[un]map_refs, with a new
  parameter m2p_override
- based on m2p_override either they follow the original behaviour, or just set
  the private flag and call set_phys_to_machine
- gnttab_[un]map_refs are now a wrapper to call __gnttab_[un]map_refs with
  m2p_override false
- a new function gnttab_[un]map_refs_userspace provides the old behaviour
- it cuts out common parts from m2p_*_override functions to
  *_foreign_p2m_mapping functions

It also removes a stray space from page.h and change ret to 0 if
XENFEAT_auto_translated_physmap, as that is the only possible return value
there.

v2:
- move the storing of the old mfn in page->index to gnttab_map_refs
- move the function header update to a separate patch

v3:
- a new approach to retain old behaviour where it needed
- squash the patches into one

v4:
- move out the common bits from m2p* functions, and pass pfn/mfn as parameter
- clear page->private before doing anything with the page, so m2p_find_override
  won't race with this

v5:
- change return value handling in __gnttab_[un]map_refs
- remove a stray space in page.h
- add detail why ret = 0 now at some places

v6:
- don't pass pfn to m2p* functions, just get it locally

v7:
- the previous version broke build on ARM, as there is no need for those p2m
  changes. I've put them into arch specific functions, which are stubs on arm

v8:
- give credit to Anthony Liguori who submitted a very similar patch originally:
http://marc.info/?i=1384307336-5328-1-git-send-email-anthony%40codemonkey.ws
- create ARM stub for get_phys_to_machine
- move definition of mfn in __gnttab_unmap_refs to the right place

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Original-by: Anthony Liguori <aliguori@amazon.com>
---
 arch/arm/include/asm/xen/page.h     |   20 +++++++++-
 arch/x86/include/asm/xen/page.h     |    7 +++-
 arch/x86/xen/p2m.c                  |   49 ++++++++++++++++-------
 drivers/block/xen-blkback/blkback.c |   15 +++----
 drivers/xen/gntdev.c                |   13 +++---
 drivers/xen/grant-table.c           |   74 +++++++++++++++++++++++++++++------
 include/xen/grant_table.h           |    8 +++-
 7 files changed, 140 insertions(+), 46 deletions(-)

diff --git a/arch/arm/include/asm/xen/page.h b/arch/arm/include/asm/xen/page.h
index e0965ab..d26c3d7 100644
--- a/arch/arm/include/asm/xen/page.h
+++ b/arch/arm/include/asm/xen/page.h
@@ -97,13 +97,31 @@ static inline pte_t *lookup_address(unsigned long address, unsigned int *level)
 	return NULL;
 }
 
+static inline unsigned long get_phys_to_machine(unsigned long pfn)
+{
+	return 0;
+}
+
+static inline int set_foreign_p2m_mapping(struct page *page, unsigned long mfn)
+{
+	return 0;
+}
+
 static inline int m2p_add_override(unsigned long mfn, struct page *page,
 		struct gnttab_map_grant_ref *kmap_op)
 {
 	return 0;
 }
 
-static inline int m2p_remove_override(struct page *page, bool clear_pte)
+static inline int restore_foreign_p2m_mapping(struct page *page,
+					      unsigned long mfn)
+{
+	return 0;
+}
+
+static inline int m2p_remove_override(struct page *page,
+				      struct gnttab_map_grant_ref *kmap_op,
+				      unsigned long mfn)
 {
 	return 0;
 }
diff --git a/arch/x86/include/asm/xen/page.h b/arch/x86/include/asm/xen/page.h
index 3e276eb..0340954 100644
--- a/arch/x86/include/asm/xen/page.h
+++ b/arch/x86/include/asm/xen/page.h
@@ -49,10 +49,13 @@ extern bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn);
 extern unsigned long set_phys_range_identity(unsigned long pfn_s,
 					     unsigned long pfn_e);
 
+extern int set_foreign_p2m_mapping(struct page *page, unsigned long mfn);
 extern int m2p_add_override(unsigned long mfn, struct page *page,
 			    struct gnttab_map_grant_ref *kmap_op);
+extern int restore_foreign_p2m_mapping(struct page *page, unsigned long mfn);
 extern int m2p_remove_override(struct page *page,
-				struct gnttab_map_grant_ref *kmap_op);
+			       struct gnttab_map_grant_ref *kmap_op,
+			       unsigned long mfn);
 extern struct page *m2p_find_override(unsigned long mfn);
 extern unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn);
 
@@ -121,7 +124,7 @@ static inline unsigned long mfn_to_pfn(unsigned long mfn)
 		pfn = m2p_find_override_pfn(mfn, ~0);
 	}
 
-	/* 
+	/*
 	 * pfn is ~0 if there are no entries in the m2p for mfn or if the
 	 * entry doesn't map back to the mfn and m2p_override doesn't have a
 	 * valid entry for it.
diff --git a/arch/x86/xen/p2m.c b/arch/x86/xen/p2m.c
index 696c694..3e7cfa9 100644
--- a/arch/x86/xen/p2m.c
+++ b/arch/x86/xen/p2m.c
@@ -881,6 +881,22 @@ static unsigned long mfn_hash(unsigned long mfn)
 	return hash_long(mfn, M2P_OVERRIDE_HASH_SHIFT);
 }
 
+int set_foreign_p2m_mapping(struct page *page, unsigned long mfn)
+{
+	unsigned long pfn = page_to_pfn(page);
+
+	WARN_ON(PagePrivate(page));
+	SetPagePrivate(page);
+	set_page_private(page, mfn);
+
+	page->index = pfn_to_mfn(pfn);
+	if (unlikely(!set_phys_to_machine(pfn, FOREIGN_FRAME(mfn))))
+		return -ENOMEM;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(set_foreign_p2m_mapping);
+
 /* Add an MFN override for a particular page */
 int m2p_add_override(unsigned long mfn, struct page *page,
 		struct gnttab_map_grant_ref *kmap_op)
@@ -899,13 +915,6 @@ int m2p_add_override(unsigned long mfn, struct page *page,
 					"m2p_add_override: pfn %lx not mapped", pfn))
 			return -EINVAL;
 	}
-	WARN_ON(PagePrivate(page));
-	SetPagePrivate(page);
-	set_page_private(page, mfn);
-	page->index = pfn_to_mfn(pfn);
-
-	if (unlikely(!set_phys_to_machine(pfn, FOREIGN_FRAME(mfn))))
-		return -ENOMEM;
 
 	if (kmap_op != NULL) {
 		if (!PageHighMem(page)) {
@@ -943,20 +952,33 @@ int m2p_add_override(unsigned long mfn, struct page *page,
 	return 0;
 }
 EXPORT_SYMBOL_GPL(m2p_add_override);
+
+int restore_foreign_p2m_mapping(struct page *page, unsigned long mfn)
+{
+	unsigned long pfn = page_to_pfn(page);
+
+	if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT))
+		return -EINVAL;
+
+	set_page_private(page, INVALID_P2M_ENTRY);
+	WARN_ON(!PagePrivate(page));
+	ClearPagePrivate(page);
+	set_phys_to_machine(pfn, page->index);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(restore_foreign_p2m_mapping);
+
 int m2p_remove_override(struct page *page,
-		struct gnttab_map_grant_ref *kmap_op)
+			struct gnttab_map_grant_ref *kmap_op,
+			unsigned long mfn)
 {
 	unsigned long flags;
-	unsigned long mfn;
 	unsigned long pfn;
 	unsigned long uninitialized_var(address);
 	unsigned level;
 	pte_t *ptep = NULL;
 
 	pfn = page_to_pfn(page);
-	mfn = get_phys_to_machine(pfn);
-	if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT))
-		return -EINVAL;
 
 	if (!PageHighMem(page)) {
 		address = (unsigned long)__va(pfn << PAGE_SHIFT);
@@ -970,10 +992,7 @@ int m2p_remove_override(struct page *page,
 	spin_lock_irqsave(&m2p_override_lock, flags);
 	list_del(&page->lru);
 	spin_unlock_irqrestore(&m2p_override_lock, flags);
-	WARN_ON(!PagePrivate(page));
-	ClearPagePrivate(page);
 
-	set_phys_to_machine(pfn, page->index);
 	if (kmap_op != NULL) {
 		if (!PageHighMem(page)) {
 			struct multicall_space mcs;
diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c
index 4b97b86..da18046 100644
--- a/drivers/block/xen-blkback/blkback.c
+++ b/drivers/block/xen-blkback/blkback.c
@@ -285,8 +285,7 @@ static void free_persistent_gnts(struct xen_blkif *blkif, struct rb_root *root,
 
 		if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST ||
 			!rb_next(&persistent_gnt->node)) {
-			ret = gnttab_unmap_refs(unmap, NULL, pages,
-				segs_to_unmap);
+			ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
 			BUG_ON(ret);
 			put_free_pages(blkif, pages, segs_to_unmap);
 			segs_to_unmap = 0;
@@ -321,8 +320,7 @@ static void unmap_purged_grants(struct work_struct *work)
 		pages[segs_to_unmap] = persistent_gnt->page;
 
 		if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
-			ret = gnttab_unmap_refs(unmap, NULL, pages,
-				segs_to_unmap);
+			ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
 			BUG_ON(ret);
 			put_free_pages(blkif, pages, segs_to_unmap);
 			segs_to_unmap = 0;
@@ -330,7 +328,7 @@ static void unmap_purged_grants(struct work_struct *work)
 		kfree(persistent_gnt);
 	}
 	if (segs_to_unmap > 0) {
-		ret = gnttab_unmap_refs(unmap, NULL, pages, segs_to_unmap);
+		ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
 		BUG_ON(ret);
 		put_free_pages(blkif, pages, segs_to_unmap);
 	}
@@ -670,15 +668,14 @@ static void xen_blkbk_unmap(struct xen_blkif *blkif,
 				    GNTMAP_host_map, pages[i]->handle);
 		pages[i]->handle = BLKBACK_INVALID_HANDLE;
 		if (++invcount == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
-			ret = gnttab_unmap_refs(unmap, NULL, unmap_pages,
-			                        invcount);
+			ret = gnttab_unmap_refs(unmap, unmap_pages, invcount);
 			BUG_ON(ret);
 			put_free_pages(blkif, unmap_pages, invcount);
 			invcount = 0;
 		}
 	}
 	if (invcount) {
-		ret = gnttab_unmap_refs(unmap, NULL, unmap_pages, invcount);
+		ret = gnttab_unmap_refs(unmap, unmap_pages, invcount);
 		BUG_ON(ret);
 		put_free_pages(blkif, unmap_pages, invcount);
 	}
@@ -740,7 +737,7 @@ again:
 	}
 
 	if (segs_to_map) {
-		ret = gnttab_map_refs(map, NULL, pages_to_gnt, segs_to_map);
+		ret = gnttab_map_refs(map, pages_to_gnt, segs_to_map);
 		BUG_ON(ret);
 	}
 
diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
index 073b4a1..34a2704 100644
--- a/drivers/xen/gntdev.c
+++ b/drivers/xen/gntdev.c
@@ -284,8 +284,10 @@ static int map_grant_pages(struct grant_map *map)
 	}
 
 	pr_debug("map %d+%d\n", map->index, map->count);
-	err = gnttab_map_refs(map->map_ops, use_ptemod ? map->kmap_ops : NULL,
-			map->pages, map->count);
+	err = gnttab_map_refs_userspace(map->map_ops,
+					use_ptemod ? map->kmap_ops : NULL,
+					map->pages,
+					map->count);
 	if (err)
 		return err;
 
@@ -315,9 +317,10 @@ static int __unmap_grant_pages(struct grant_map *map, int offset, int pages)
 		}
 	}
 
-	err = gnttab_unmap_refs(map->unmap_ops + offset,
-			use_ptemod ? map->kmap_ops + offset : NULL, map->pages + offset,
-			pages);
+	err = gnttab_unmap_refs_userspace(map->unmap_ops + offset,
+					  use_ptemod ? map->kmap_ops + offset : NULL,
+					  map->pages + offset,
+					  pages);
 	if (err)
 		return err;
 
diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c
index b84e3ab..c719bc8 100644
--- a/drivers/xen/grant-table.c
+++ b/drivers/xen/grant-table.c
@@ -928,15 +928,17 @@ void gnttab_batch_copy(struct gnttab_copy *batch, unsigned count)
 }
 EXPORT_SYMBOL_GPL(gnttab_batch_copy);
 
-int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
+int __gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
 		    struct gnttab_map_grant_ref *kmap_ops,
-		    struct page **pages, unsigned int count)
+		    struct page **pages, unsigned int count,
+		    bool m2p_override)
 {
 	int i, ret;
 	bool lazy = false;
 	pte_t *pte;
 	unsigned long mfn;
 
+	BUG_ON(kmap_ops && !m2p_override);
 	ret = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, map_ops, count);
 	if (ret)
 		return ret;
@@ -955,10 +957,12 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
 			set_phys_to_machine(map_ops[i].host_addr >> PAGE_SHIFT,
 					map_ops[i].dev_bus_addr >> PAGE_SHIFT);
 		}
-		return ret;
+		return 0;
 	}
 
-	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
+	if (m2p_override &&
+	    !in_interrupt() &&
+	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
 		arch_enter_lazy_mmu_mode();
 		lazy = true;
 	}
@@ -975,8 +979,14 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
 		} else {
 			mfn = PFN_DOWN(map_ops[i].dev_bus_addr);
 		}
-		ret = m2p_add_override(mfn, pages[i], kmap_ops ?
-				       &kmap_ops[i] : NULL);
+
+		ret = set_foreign_p2m_mapping(pages[i], mfn);
+		if (ret)
+			goto out;
+
+		if (m2p_override)
+			ret = m2p_add_override(mfn, pages[i], kmap_ops ?
+					       &kmap_ops[i] : NULL);
 		if (ret)
 			goto out;
 	}
@@ -987,15 +997,31 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
 
 	return ret;
 }
+
+int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
+		    struct page **pages, unsigned int count)
+{
+	return __gnttab_map_refs(map_ops, NULL, pages, count, false);
+}
 EXPORT_SYMBOL_GPL(gnttab_map_refs);
 
-int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
+int gnttab_map_refs_userspace(struct gnttab_map_grant_ref *map_ops,
+			      struct gnttab_map_grant_ref *kmap_ops,
+			      struct page **pages, unsigned int count)
+{
+	return __gnttab_map_refs(map_ops, kmap_ops, pages, count, true);
+}
+EXPORT_SYMBOL_GPL(gnttab_map_refs_userspace);
+
+int __gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
 		      struct gnttab_map_grant_ref *kmap_ops,
-		      struct page **pages, unsigned int count)
+		      struct page **pages, unsigned int count,
+		      bool m2p_override)
 {
 	int i, ret;
 	bool lazy = false;
 
+	BUG_ON(kmap_ops && !m2p_override);
 	ret = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, unmap_ops, count);
 	if (ret)
 		return ret;
@@ -1006,17 +1032,27 @@ int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
 			set_phys_to_machine(unmap_ops[i].host_addr >> PAGE_SHIFT,
 					INVALID_P2M_ENTRY);
 		}
-		return ret;
+		return 0;
 	}
 
-	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
+	if (m2p_override &&
+	    !in_interrupt() &&
+	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
 		arch_enter_lazy_mmu_mode();
 		lazy = true;
 	}
 
 	for (i = 0; i < count; i++) {
-		ret = m2p_remove_override(pages[i], kmap_ops ?
-				       &kmap_ops[i] : NULL);
+		unsigned long mfn = get_phys_to_machine(page_to_pfn(pages[i]));
+		ret = restore_foreign_p2m_mapping(pages[i], mfn);
+		if (ret)
+			goto out;
+
+		if (m2p_override)
+			ret = m2p_remove_override(pages[i],
+						  kmap_ops ?
+						   &kmap_ops[i] : NULL,
+						  mfn);
 		if (ret)
 			goto out;
 	}
@@ -1027,8 +1063,22 @@ int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
 
 	return ret;
 }
+
+int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *map_ops,
+		    struct page **pages, unsigned int count)
+{
+	return __gnttab_unmap_refs(map_ops, NULL, pages, count, false);
+}
 EXPORT_SYMBOL_GPL(gnttab_unmap_refs);
 
+int gnttab_unmap_refs_userspace(struct gnttab_unmap_grant_ref *map_ops,
+				struct gnttab_map_grant_ref *kmap_ops,
+				struct page **pages, unsigned int count)
+{
+	return __gnttab_unmap_refs(map_ops, kmap_ops, pages, count, true);
+}
+EXPORT_SYMBOL_GPL(gnttab_unmap_refs_userspace);
+
 static unsigned nr_status_frames(unsigned nr_grant_frames)
 {
 	BUG_ON(grefs_per_grant_frame == 0);
diff --git a/include/xen/grant_table.h b/include/xen/grant_table.h
index a5af2a2..7ad033d 100644
--- a/include/xen/grant_table.h
+++ b/include/xen/grant_table.h
@@ -191,11 +191,15 @@ void gnttab_free_auto_xlat_frames(void);
 #define gnttab_map_vaddr(map) ((void *)(map.host_virt_addr))
 
 int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
-		    struct gnttab_map_grant_ref *kmap_ops,
 		    struct page **pages, unsigned int count);
+int gnttab_map_refs_userspace(struct gnttab_map_grant_ref *map_ops,
+			      struct gnttab_map_grant_ref *kmap_ops,
+			      struct page **pages, unsigned int count);
 int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
-		      struct gnttab_map_grant_ref *kunmap_ops,
 		      struct page **pages, unsigned int count);
+int gnttab_unmap_refs_userspace(struct gnttab_unmap_grant_ref *unmap_ops,
+				struct gnttab_map_grant_ref *kunmap_ops,
+				struct page **pages, unsigned int count);
 
 /* Perform a batch of grant map/copy operations. Retry every batch slot
  * for which the hypervisor returns GNTST_eagain. This is typically due

^ permalink raw reply related	[flat|nested] 6+ messages in thread

* Re: [Xen-devel] [PATCH v8] xen/grant-table: Avoid m2p_override during mapping
  2014-02-06 12:56 [PATCH v8] xen/grant-table: Avoid m2p_override during mapping Zoltan Kiss
  2014-02-07 18:42 ` David Vrabel
@ 2014-02-07 18:42 ` David Vrabel
  2014-02-11 18:55 ` Stefano Stabellini
  2014-02-11 18:55 ` Stefano Stabellini
  3 siblings, 0 replies; 6+ messages in thread
From: David Vrabel @ 2014-02-07 18:42 UTC (permalink / raw)
  To: Zoltan Kiss
  Cc: Konrad Rzeszutek Wilk, Boris Ostrovsky, David Vrabel,
	Thomas Gleixner, Ingo Molnar, H. Peter Anvin, x86, xen-devel,
	linux-kernel, Roger Pau Monné,
	Jan Beulich, Ian Campbell, stefano.stabellini

On 06/02/14 12:56, Zoltan Kiss wrote:
> The grant mapping API does m2p_override unnecessarily: only gntdev needs it,
> for blkback and future netback patches it just cause a lock contention, as
> those pages never go to userspace. Therefore this series does the following:
> - the original functions were renamed to __gnttab_[un]map_refs, with a new
>   parameter m2p_override
> - based on m2p_override either they follow the original behaviour, or just set
>   the private flag and call set_phys_to_machine
> - gnttab_[un]map_refs are now a wrapper to call __gnttab_[un]map_refs with
>   m2p_override false
> - a new function gnttab_[un]map_refs_userspace provides the old behaviour
> - it cuts out common parts from m2p_*_override functions to
>   *_foreign_p2m_mapping functions
> 
> It also removes a stray space from page.h and change ret to 0 if
> XENFEAT_auto_translated_physmap, as that is the only possible return value
> there.
> 
> v2:

Please put this version information after the '---' marker.  It doesn't
need to end up in the commit message.

> @@ -955,10 +957,12 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  			set_phys_to_machine(map_ops[i].host_addr >> PAGE_SHIFT,
>  					map_ops[i].dev_bus_addr >> PAGE_SHIFT);
>  		}
> -		return ret;
> +		return 0;
>  	}
>  
> -	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
> +	if (m2p_override &&
> +	    !in_interrupt() &&
> +	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
>  		arch_enter_lazy_mmu_mode();
>  		lazy = true;
>  	}

I think this block and the loop should be in an arch-specific function
(e.g., set_foreign_p2m_mappings(), but I would like to hear Stefano's
opinion.

Similarly for clear_foreign_p2m_mappings().

David

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v8] xen/grant-table: Avoid m2p_override during mapping
  2014-02-06 12:56 [PATCH v8] xen/grant-table: Avoid m2p_override during mapping Zoltan Kiss
@ 2014-02-07 18:42 ` David Vrabel
  2014-02-07 18:42 ` [Xen-devel] " David Vrabel
                   ` (2 subsequent siblings)
  3 siblings, 0 replies; 6+ messages in thread
From: David Vrabel @ 2014-02-07 18:42 UTC (permalink / raw)
  To: Zoltan Kiss
  Cc: Ian Campbell, x86, linux-kernel, Ingo Molnar, David Vrabel,
	Jan Beulich, H. Peter Anvin, xen-devel, Boris Ostrovsky,
	Thomas Gleixner, stefano.stabellini, Roger Pau Monné

On 06/02/14 12:56, Zoltan Kiss wrote:
> The grant mapping API does m2p_override unnecessarily: only gntdev needs it,
> for blkback and future netback patches it just cause a lock contention, as
> those pages never go to userspace. Therefore this series does the following:
> - the original functions were renamed to __gnttab_[un]map_refs, with a new
>   parameter m2p_override
> - based on m2p_override either they follow the original behaviour, or just set
>   the private flag and call set_phys_to_machine
> - gnttab_[un]map_refs are now a wrapper to call __gnttab_[un]map_refs with
>   m2p_override false
> - a new function gnttab_[un]map_refs_userspace provides the old behaviour
> - it cuts out common parts from m2p_*_override functions to
>   *_foreign_p2m_mapping functions
> 
> It also removes a stray space from page.h and change ret to 0 if
> XENFEAT_auto_translated_physmap, as that is the only possible return value
> there.
> 
> v2:

Please put this version information after the '---' marker.  It doesn't
need to end up in the commit message.

> @@ -955,10 +957,12 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  			set_phys_to_machine(map_ops[i].host_addr >> PAGE_SHIFT,
>  					map_ops[i].dev_bus_addr >> PAGE_SHIFT);
>  		}
> -		return ret;
> +		return 0;
>  	}
>  
> -	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
> +	if (m2p_override &&
> +	    !in_interrupt() &&
> +	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
>  		arch_enter_lazy_mmu_mode();
>  		lazy = true;
>  	}

I think this block and the loop should be in an arch-specific function
(e.g., set_foreign_p2m_mappings(), but I would like to hear Stefano's
opinion.

Similarly for clear_foreign_p2m_mappings().

David

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v8] xen/grant-table: Avoid m2p_override during mapping
  2014-02-06 12:56 [PATCH v8] xen/grant-table: Avoid m2p_override during mapping Zoltan Kiss
                   ` (2 preceding siblings ...)
  2014-02-11 18:55 ` Stefano Stabellini
@ 2014-02-11 18:55 ` Stefano Stabellini
  3 siblings, 0 replies; 6+ messages in thread
From: Stefano Stabellini @ 2014-02-11 18:55 UTC (permalink / raw)
  To: Zoltan Kiss
  Cc: Konrad Rzeszutek Wilk, Boris Ostrovsky, David Vrabel,
	Thomas Gleixner, Ingo Molnar, H. Peter Anvin, x86, xen-devel,
	linux-kernel, Roger Pau Monné,
	Jan Beulich, Ian Campbell, stefano.stabellini

On Thu, 6 Feb 2014, Zoltan Kiss wrote:
> The grant mapping API does m2p_override unnecessarily: only gntdev needs it,
> for blkback and future netback patches it just cause a lock contention, as
> those pages never go to userspace. Therefore this series does the following:
> - the original functions were renamed to __gnttab_[un]map_refs, with a new
>   parameter m2p_override
> - based on m2p_override either they follow the original behaviour, or just set
>   the private flag and call set_phys_to_machine
> - gnttab_[un]map_refs are now a wrapper to call __gnttab_[un]map_refs with
>   m2p_override false
> - a new function gnttab_[un]map_refs_userspace provides the old behaviour
> - it cuts out common parts from m2p_*_override functions to
>   *_foreign_p2m_mapping functions
> 
> It also removes a stray space from page.h and change ret to 0 if
> XENFEAT_auto_translated_physmap, as that is the only possible return value
> there.
> 
> v2:
> - move the storing of the old mfn in page->index to gnttab_map_refs
> - move the function header update to a separate patch
> 
> v3:
> - a new approach to retain old behaviour where it needed
> - squash the patches into one
> 
> v4:
> - move out the common bits from m2p* functions, and pass pfn/mfn as parameter
> - clear page->private before doing anything with the page, so m2p_find_override
>   won't race with this
> 
> v5:
> - change return value handling in __gnttab_[un]map_refs
> - remove a stray space in page.h
> - add detail why ret = 0 now at some places
> 
> v6:
> - don't pass pfn to m2p* functions, just get it locally
> 
> v7:
> - the previous version broke build on ARM, as there is no need for those p2m
>   changes. I've put them into arch specific functions, which are stubs on arm
> 
> v8:
> - give credit to Anthony Liguori who submitted a very similar patch originally:
> http://marc.info/?i=1384307336-5328-1-git-send-email-anthony%40codemonkey.ws
> - create ARM stub for get_phys_to_machine
> - move definition of mfn in __gnttab_unmap_refs to the right place
> 
> Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
> Original-by: Anthony Liguori <aliguori@amazon.com>
> ---
>  arch/arm/include/asm/xen/page.h     |   20 +++++++++-
>  arch/x86/include/asm/xen/page.h     |    7 +++-
>  arch/x86/xen/p2m.c                  |   49 ++++++++++++++++-------
>  drivers/block/xen-blkback/blkback.c |   15 +++----
>  drivers/xen/gntdev.c                |   13 +++---
>  drivers/xen/grant-table.c           |   74 +++++++++++++++++++++++++++++------
>  include/xen/grant_table.h           |    8 +++-
>  7 files changed, 140 insertions(+), 46 deletions(-)
> 
> diff --git a/arch/arm/include/asm/xen/page.h b/arch/arm/include/asm/xen/page.h
> index e0965ab..d26c3d7 100644
> --- a/arch/arm/include/asm/xen/page.h
> +++ b/arch/arm/include/asm/xen/page.h
> @@ -97,13 +97,31 @@ static inline pte_t *lookup_address(unsigned long address, unsigned int *level)
>  	return NULL;
>  }
>  
> +static inline unsigned long get_phys_to_machine(unsigned long pfn)
> +{
> +	return 0;
> +}

We can probably just rename phys_to_machine to get_phys_to_machine.
Be careful about the other callers in page.h.


> +static inline int set_foreign_p2m_mapping(struct page *page, unsigned long mfn)
> +{
> +	return 0;
> +}
> +
>  static inline int m2p_add_override(unsigned long mfn, struct page *page,
>  		struct gnttab_map_grant_ref *kmap_op)
>  {
>  	return 0;
>  }
>  
> -static inline int m2p_remove_override(struct page *page, bool clear_pte)
> +static inline int restore_foreign_p2m_mapping(struct page *page,
> +					      unsigned long mfn)
> +{
> +	return 0;
> +}
> +
> +static inline int m2p_remove_override(struct page *page,
> +				      struct gnttab_map_grant_ref *kmap_op,
> +				      unsigned long mfn)
>  {
>  	return 0;
>  }

I like set_foreign_p2m_mapping and restore_foreign_p2m_mapping but we
need to use them coherently.
I think that on ARM it makes sense to rename the current implementation of
set_phys_to_machine to set_foreign_p2m_mapping.
I also makes sense to change the call to set_phys_to_machine with
INVALID_P2M_ENTRY as a parameter to restore_foreign_p2m_mapping.

Then you can implement set_phys_to_machine as

BUG_ON(pfn != mfn):
    
and get_phys_to_machine as

return pfn;


> diff --git a/arch/x86/include/asm/xen/page.h b/arch/x86/include/asm/xen/page.h
> index 3e276eb..0340954 100644
> --- a/arch/x86/include/asm/xen/page.h
> +++ b/arch/x86/include/asm/xen/page.h
> @@ -49,10 +49,13 @@ extern bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn);
>  extern unsigned long set_phys_range_identity(unsigned long pfn_s,
>  					     unsigned long pfn_e);
>  
> +extern int set_foreign_p2m_mapping(struct page *page, unsigned long mfn);
>  extern int m2p_add_override(unsigned long mfn, struct page *page,
>  			    struct gnttab_map_grant_ref *kmap_op);
> +extern int restore_foreign_p2m_mapping(struct page *page, unsigned long mfn);
>  extern int m2p_remove_override(struct page *page,
> -				struct gnttab_map_grant_ref *kmap_op);
> +			       struct gnttab_map_grant_ref *kmap_op,
> +			       unsigned long mfn);
>  extern struct page *m2p_find_override(unsigned long mfn);
>  extern unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn);
>  
> @@ -121,7 +124,7 @@ static inline unsigned long mfn_to_pfn(unsigned long mfn)
>  		pfn = m2p_find_override_pfn(mfn, ~0);
>  	}
>  
> -	/* 
> +	/*
>  	 * pfn is ~0 if there are no entries in the m2p for mfn or if the
>  	 * entry doesn't map back to the mfn and m2p_override doesn't have a
>  	 * valid entry for it.
> diff --git a/arch/x86/xen/p2m.c b/arch/x86/xen/p2m.c
> index 696c694..3e7cfa9 100644
> --- a/arch/x86/xen/p2m.c
> +++ b/arch/x86/xen/p2m.c
> @@ -881,6 +881,22 @@ static unsigned long mfn_hash(unsigned long mfn)
>  	return hash_long(mfn, M2P_OVERRIDE_HASH_SHIFT);
>  }
>  
> +int set_foreign_p2m_mapping(struct page *page, unsigned long mfn)
> +{
> +	unsigned long pfn = page_to_pfn(page);
> +
> +	WARN_ON(PagePrivate(page));
> +	SetPagePrivate(page);
> +	set_page_private(page, mfn);
> +
> +	page->index = pfn_to_mfn(pfn);
> +	if (unlikely(!set_phys_to_machine(pfn, FOREIGN_FRAME(mfn))))
> +		return -ENOMEM;
> +
> +	return 0;
> +}
> +EXPORT_SYMBOL_GPL(set_foreign_p2m_mapping);
> +
>  /* Add an MFN override for a particular page */
>  int m2p_add_override(unsigned long mfn, struct page *page,
>  		struct gnttab_map_grant_ref *kmap_op)
> @@ -899,13 +915,6 @@ int m2p_add_override(unsigned long mfn, struct page *page,
>  					"m2p_add_override: pfn %lx not mapped", pfn))
>  			return -EINVAL;
>  	}
> -	WARN_ON(PagePrivate(page));
> -	SetPagePrivate(page);
> -	set_page_private(page, mfn);
> -	page->index = pfn_to_mfn(pfn);
> -
> -	if (unlikely(!set_phys_to_machine(pfn, FOREIGN_FRAME(mfn))))
> -		return -ENOMEM;
>  
>  	if (kmap_op != NULL) {
>  		if (!PageHighMem(page)) {
> @@ -943,20 +952,33 @@ int m2p_add_override(unsigned long mfn, struct page *page,
>  	return 0;
>  }
>  EXPORT_SYMBOL_GPL(m2p_add_override);
> +
> +int restore_foreign_p2m_mapping(struct page *page, unsigned long mfn)
> +{
> +	unsigned long pfn = page_to_pfn(page);
> +
> +	if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT))
> +		return -EINVAL;
> +
> +	set_page_private(page, INVALID_P2M_ENTRY);
> +	WARN_ON(!PagePrivate(page));
> +	ClearPagePrivate(page);
> +	set_phys_to_machine(pfn, page->index);
> +	return 0;
> +}
> +EXPORT_SYMBOL_GPL(restore_foreign_p2m_mapping);
> +
>  int m2p_remove_override(struct page *page,
> -		struct gnttab_map_grant_ref *kmap_op)
> +			struct gnttab_map_grant_ref *kmap_op,
> +			unsigned long mfn)
>  {
>  	unsigned long flags;
> -	unsigned long mfn;
>  	unsigned long pfn;
>  	unsigned long uninitialized_var(address);
>  	unsigned level;
>  	pte_t *ptep = NULL;
>  
>  	pfn = page_to_pfn(page);
> -	mfn = get_phys_to_machine(pfn);
> -	if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT))
> -		return -EINVAL;
>  
>  	if (!PageHighMem(page)) {
>  		address = (unsigned long)__va(pfn << PAGE_SHIFT);
> @@ -970,10 +992,7 @@ int m2p_remove_override(struct page *page,
>  	spin_lock_irqsave(&m2p_override_lock, flags);
>  	list_del(&page->lru);
>  	spin_unlock_irqrestore(&m2p_override_lock, flags);
> -	WARN_ON(!PagePrivate(page));
> -	ClearPagePrivate(page);
>  
> -	set_phys_to_machine(pfn, page->index);
>  	if (kmap_op != NULL) {
>  		if (!PageHighMem(page)) {
>  			struct multicall_space mcs;
> diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c
> index 4b97b86..da18046 100644
> --- a/drivers/block/xen-blkback/blkback.c
> +++ b/drivers/block/xen-blkback/blkback.c
> @@ -285,8 +285,7 @@ static void free_persistent_gnts(struct xen_blkif *blkif, struct rb_root *root,
>  
>  		if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST ||
>  			!rb_next(&persistent_gnt->node)) {
> -			ret = gnttab_unmap_refs(unmap, NULL, pages,
> -				segs_to_unmap);
> +			ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
>  			BUG_ON(ret);
>  			put_free_pages(blkif, pages, segs_to_unmap);
>  			segs_to_unmap = 0;
> @@ -321,8 +320,7 @@ static void unmap_purged_grants(struct work_struct *work)
>  		pages[segs_to_unmap] = persistent_gnt->page;
>  
>  		if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
> -			ret = gnttab_unmap_refs(unmap, NULL, pages,
> -				segs_to_unmap);
> +			ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
>  			BUG_ON(ret);
>  			put_free_pages(blkif, pages, segs_to_unmap);
>  			segs_to_unmap = 0;
> @@ -330,7 +328,7 @@ static void unmap_purged_grants(struct work_struct *work)
>  		kfree(persistent_gnt);
>  	}
>  	if (segs_to_unmap > 0) {
> -		ret = gnttab_unmap_refs(unmap, NULL, pages, segs_to_unmap);
> +		ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
>  		BUG_ON(ret);
>  		put_free_pages(blkif, pages, segs_to_unmap);
>  	}
> @@ -670,15 +668,14 @@ static void xen_blkbk_unmap(struct xen_blkif *blkif,
>  				    GNTMAP_host_map, pages[i]->handle);
>  		pages[i]->handle = BLKBACK_INVALID_HANDLE;
>  		if (++invcount == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
> -			ret = gnttab_unmap_refs(unmap, NULL, unmap_pages,
> -			                        invcount);
> +			ret = gnttab_unmap_refs(unmap, unmap_pages, invcount);
>  			BUG_ON(ret);
>  			put_free_pages(blkif, unmap_pages, invcount);
>  			invcount = 0;
>  		}
>  	}
>  	if (invcount) {
> -		ret = gnttab_unmap_refs(unmap, NULL, unmap_pages, invcount);
> +		ret = gnttab_unmap_refs(unmap, unmap_pages, invcount);
>  		BUG_ON(ret);
>  		put_free_pages(blkif, unmap_pages, invcount);
>  	}
> @@ -740,7 +737,7 @@ again:
>  	}
>  
>  	if (segs_to_map) {
> -		ret = gnttab_map_refs(map, NULL, pages_to_gnt, segs_to_map);
> +		ret = gnttab_map_refs(map, pages_to_gnt, segs_to_map);
>  		BUG_ON(ret);
>  	}
>  
> diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
> index 073b4a1..34a2704 100644
> --- a/drivers/xen/gntdev.c
> +++ b/drivers/xen/gntdev.c
> @@ -284,8 +284,10 @@ static int map_grant_pages(struct grant_map *map)
>  	}
>  
>  	pr_debug("map %d+%d\n", map->index, map->count);
> -	err = gnttab_map_refs(map->map_ops, use_ptemod ? map->kmap_ops : NULL,
> -			map->pages, map->count);
> +	err = gnttab_map_refs_userspace(map->map_ops,
> +					use_ptemod ? map->kmap_ops : NULL,
> +					map->pages,
> +					map->count);
>  	if (err)
>  		return err;
>  
> @@ -315,9 +317,10 @@ static int __unmap_grant_pages(struct grant_map *map, int offset, int pages)
>  		}
>  	}
>  
> -	err = gnttab_unmap_refs(map->unmap_ops + offset,
> -			use_ptemod ? map->kmap_ops + offset : NULL, map->pages + offset,
> -			pages);
> +	err = gnttab_unmap_refs_userspace(map->unmap_ops + offset,
> +					  use_ptemod ? map->kmap_ops + offset : NULL,
> +					  map->pages + offset,
> +					  pages);
>  	if (err)
>  		return err;
>  
> diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c
> index b84e3ab..c719bc8 100644
> --- a/drivers/xen/grant-table.c
> +++ b/drivers/xen/grant-table.c
> @@ -928,15 +928,17 @@ void gnttab_batch_copy(struct gnttab_copy *batch, unsigned count)
>  }
>  EXPORT_SYMBOL_GPL(gnttab_batch_copy);
>  
> -int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
> +int __gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  		    struct gnttab_map_grant_ref *kmap_ops,
> -		    struct page **pages, unsigned int count)
> +		    struct page **pages, unsigned int count,
> +		    bool m2p_override)
>  {
>  	int i, ret;
>  	bool lazy = false;
>  	pte_t *pte;
>  	unsigned long mfn;
>  
> +	BUG_ON(kmap_ops && !m2p_override);
>  	ret = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, map_ops, count);
>  	if (ret)
>  		return ret;
> @@ -955,10 +957,12 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  			set_phys_to_machine(map_ops[i].host_addr >> PAGE_SHIFT,
>  					map_ops[i].dev_bus_addr >> PAGE_SHIFT);

I think that this should become a call to set_foreign_p2m_mapping


>  		}
> -		return ret;
> +		return 0;
>  	}
>  
> -	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
> +	if (m2p_override &&
> +	    !in_interrupt() &&
> +	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
>  		arch_enter_lazy_mmu_mode();
>  		lazy = true;
>  	}
> @@ -975,8 +979,14 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  		} else {
>  			mfn = PFN_DOWN(map_ops[i].dev_bus_addr);
>  		}
> -		ret = m2p_add_override(mfn, pages[i], kmap_ops ?
> -				       &kmap_ops[i] : NULL);
> +
> +		ret = set_foreign_p2m_mapping(pages[i], mfn);
> +		if (ret)
> +			goto out;
> +
> +		if (m2p_override)
> +			ret = m2p_add_override(mfn, pages[i], kmap_ops ?
> +					       &kmap_ops[i] : NULL);
>  		if (ret)
>  			goto out;
>  	}
> @@ -987,15 +997,31 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  
>  	return ret;
>  }
> +
> +int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
> +		    struct page **pages, unsigned int count)
> +{
> +	return __gnttab_map_refs(map_ops, NULL, pages, count, false);
> +}
>  EXPORT_SYMBOL_GPL(gnttab_map_refs);
>  
> -int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
> +int gnttab_map_refs_userspace(struct gnttab_map_grant_ref *map_ops,
> +			      struct gnttab_map_grant_ref *kmap_ops,
> +			      struct page **pages, unsigned int count)
> +{
> +	return __gnttab_map_refs(map_ops, kmap_ops, pages, count, true);
> +}
> +EXPORT_SYMBOL_GPL(gnttab_map_refs_userspace);
> +
> +int __gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
>  		      struct gnttab_map_grant_ref *kmap_ops,
> -		      struct page **pages, unsigned int count)
> +		      struct page **pages, unsigned int count,
> +		      bool m2p_override)
>  {
>  	int i, ret;
>  	bool lazy = false;
>  
> +	BUG_ON(kmap_ops && !m2p_override);
>  	ret = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, unmap_ops, count);
>  	if (ret)
>  		return ret;
> @@ -1006,17 +1032,27 @@ int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
>  			set_phys_to_machine(unmap_ops[i].host_addr >> PAGE_SHIFT,
>  					INVALID_P2M_ENTRY);

this should become a call to restore_foreign_p2m_mapping


>  		}
> -		return ret;
> +		return 0;
>  	}
>  
> -	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
> +	if (m2p_override &&
> +	    !in_interrupt() &&
> +	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
>  		arch_enter_lazy_mmu_mode();
>  		lazy = true;
>  	}
>  
>  	for (i = 0; i < count; i++) {
> -		ret = m2p_remove_override(pages[i], kmap_ops ?
> -				       &kmap_ops[i] : NULL);
> +		unsigned long mfn = get_phys_to_machine(page_to_pfn(pages[i]));
> +		ret = restore_foreign_p2m_mapping(pages[i], mfn);
> +		if (ret)
> +			goto out;
> +
> +		if (m2p_override)
> +			ret = m2p_remove_override(pages[i],
> +						  kmap_ops ?
> +						   &kmap_ops[i] : NULL,
> +						  mfn);
>  		if (ret)
>  			goto out;
>  	}
> @@ -1027,8 +1063,22 @@ int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
>  
>  	return ret;
>  }
> +
> +int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *map_ops,
> +		    struct page **pages, unsigned int count)
> +{
> +	return __gnttab_unmap_refs(map_ops, NULL, pages, count, false);
> +}
>  EXPORT_SYMBOL_GPL(gnttab_unmap_refs);
>  
> +int gnttab_unmap_refs_userspace(struct gnttab_unmap_grant_ref *map_ops,
> +				struct gnttab_map_grant_ref *kmap_ops,
> +				struct page **pages, unsigned int count)
> +{
> +	return __gnttab_unmap_refs(map_ops, kmap_ops, pages, count, true);
> +}
> +EXPORT_SYMBOL_GPL(gnttab_unmap_refs_userspace);
> +
>  static unsigned nr_status_frames(unsigned nr_grant_frames)
>  {
>  	BUG_ON(grefs_per_grant_frame == 0);
> diff --git a/include/xen/grant_table.h b/include/xen/grant_table.h
> index a5af2a2..7ad033d 100644
> --- a/include/xen/grant_table.h
> +++ b/include/xen/grant_table.h
> @@ -191,11 +191,15 @@ void gnttab_free_auto_xlat_frames(void);
>  #define gnttab_map_vaddr(map) ((void *)(map.host_virt_addr))
>  
>  int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
> -		    struct gnttab_map_grant_ref *kmap_ops,
>  		    struct page **pages, unsigned int count);
> +int gnttab_map_refs_userspace(struct gnttab_map_grant_ref *map_ops,
> +			      struct gnttab_map_grant_ref *kmap_ops,
> +			      struct page **pages, unsigned int count);
>  int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
> -		      struct gnttab_map_grant_ref *kunmap_ops,
>  		      struct page **pages, unsigned int count);
> +int gnttab_unmap_refs_userspace(struct gnttab_unmap_grant_ref *unmap_ops,
> +				struct gnttab_map_grant_ref *kunmap_ops,
> +				struct page **pages, unsigned int count);
>  
>  /* Perform a batch of grant map/copy operations. Retry every batch slot
>   * for which the hypervisor returns GNTST_eagain. This is typically due
> 

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v8] xen/grant-table: Avoid m2p_override during mapping
  2014-02-06 12:56 [PATCH v8] xen/grant-table: Avoid m2p_override during mapping Zoltan Kiss
  2014-02-07 18:42 ` David Vrabel
  2014-02-07 18:42 ` [Xen-devel] " David Vrabel
@ 2014-02-11 18:55 ` Stefano Stabellini
  2014-02-11 18:55 ` Stefano Stabellini
  3 siblings, 0 replies; 6+ messages in thread
From: Stefano Stabellini @ 2014-02-11 18:55 UTC (permalink / raw)
  To: Zoltan Kiss
  Cc: Ian Campbell, x86, linux-kernel, Ingo Molnar, David Vrabel,
	Jan Beulich, H. Peter Anvin, xen-devel, Boris Ostrovsky,
	Thomas Gleixner, stefano.stabellini, Roger Pau Monné

On Thu, 6 Feb 2014, Zoltan Kiss wrote:
> The grant mapping API does m2p_override unnecessarily: only gntdev needs it,
> for blkback and future netback patches it just cause a lock contention, as
> those pages never go to userspace. Therefore this series does the following:
> - the original functions were renamed to __gnttab_[un]map_refs, with a new
>   parameter m2p_override
> - based on m2p_override either they follow the original behaviour, or just set
>   the private flag and call set_phys_to_machine
> - gnttab_[un]map_refs are now a wrapper to call __gnttab_[un]map_refs with
>   m2p_override false
> - a new function gnttab_[un]map_refs_userspace provides the old behaviour
> - it cuts out common parts from m2p_*_override functions to
>   *_foreign_p2m_mapping functions
> 
> It also removes a stray space from page.h and change ret to 0 if
> XENFEAT_auto_translated_physmap, as that is the only possible return value
> there.
> 
> v2:
> - move the storing of the old mfn in page->index to gnttab_map_refs
> - move the function header update to a separate patch
> 
> v3:
> - a new approach to retain old behaviour where it needed
> - squash the patches into one
> 
> v4:
> - move out the common bits from m2p* functions, and pass pfn/mfn as parameter
> - clear page->private before doing anything with the page, so m2p_find_override
>   won't race with this
> 
> v5:
> - change return value handling in __gnttab_[un]map_refs
> - remove a stray space in page.h
> - add detail why ret = 0 now at some places
> 
> v6:
> - don't pass pfn to m2p* functions, just get it locally
> 
> v7:
> - the previous version broke build on ARM, as there is no need for those p2m
>   changes. I've put them into arch specific functions, which are stubs on arm
> 
> v8:
> - give credit to Anthony Liguori who submitted a very similar patch originally:
> http://marc.info/?i=1384307336-5328-1-git-send-email-anthony%40codemonkey.ws
> - create ARM stub for get_phys_to_machine
> - move definition of mfn in __gnttab_unmap_refs to the right place
> 
> Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
> Original-by: Anthony Liguori <aliguori@amazon.com>
> ---
>  arch/arm/include/asm/xen/page.h     |   20 +++++++++-
>  arch/x86/include/asm/xen/page.h     |    7 +++-
>  arch/x86/xen/p2m.c                  |   49 ++++++++++++++++-------
>  drivers/block/xen-blkback/blkback.c |   15 +++----
>  drivers/xen/gntdev.c                |   13 +++---
>  drivers/xen/grant-table.c           |   74 +++++++++++++++++++++++++++++------
>  include/xen/grant_table.h           |    8 +++-
>  7 files changed, 140 insertions(+), 46 deletions(-)
> 
> diff --git a/arch/arm/include/asm/xen/page.h b/arch/arm/include/asm/xen/page.h
> index e0965ab..d26c3d7 100644
> --- a/arch/arm/include/asm/xen/page.h
> +++ b/arch/arm/include/asm/xen/page.h
> @@ -97,13 +97,31 @@ static inline pte_t *lookup_address(unsigned long address, unsigned int *level)
>  	return NULL;
>  }
>  
> +static inline unsigned long get_phys_to_machine(unsigned long pfn)
> +{
> +	return 0;
> +}

We can probably just rename phys_to_machine to get_phys_to_machine.
Be careful about the other callers in page.h.


> +static inline int set_foreign_p2m_mapping(struct page *page, unsigned long mfn)
> +{
> +	return 0;
> +}
> +
>  static inline int m2p_add_override(unsigned long mfn, struct page *page,
>  		struct gnttab_map_grant_ref *kmap_op)
>  {
>  	return 0;
>  }
>  
> -static inline int m2p_remove_override(struct page *page, bool clear_pte)
> +static inline int restore_foreign_p2m_mapping(struct page *page,
> +					      unsigned long mfn)
> +{
> +	return 0;
> +}
> +
> +static inline int m2p_remove_override(struct page *page,
> +				      struct gnttab_map_grant_ref *kmap_op,
> +				      unsigned long mfn)
>  {
>  	return 0;
>  }

I like set_foreign_p2m_mapping and restore_foreign_p2m_mapping but we
need to use them coherently.
I think that on ARM it makes sense to rename the current implementation of
set_phys_to_machine to set_foreign_p2m_mapping.
I also makes sense to change the call to set_phys_to_machine with
INVALID_P2M_ENTRY as a parameter to restore_foreign_p2m_mapping.

Then you can implement set_phys_to_machine as

BUG_ON(pfn != mfn):
    
and get_phys_to_machine as

return pfn;


> diff --git a/arch/x86/include/asm/xen/page.h b/arch/x86/include/asm/xen/page.h
> index 3e276eb..0340954 100644
> --- a/arch/x86/include/asm/xen/page.h
> +++ b/arch/x86/include/asm/xen/page.h
> @@ -49,10 +49,13 @@ extern bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn);
>  extern unsigned long set_phys_range_identity(unsigned long pfn_s,
>  					     unsigned long pfn_e);
>  
> +extern int set_foreign_p2m_mapping(struct page *page, unsigned long mfn);
>  extern int m2p_add_override(unsigned long mfn, struct page *page,
>  			    struct gnttab_map_grant_ref *kmap_op);
> +extern int restore_foreign_p2m_mapping(struct page *page, unsigned long mfn);
>  extern int m2p_remove_override(struct page *page,
> -				struct gnttab_map_grant_ref *kmap_op);
> +			       struct gnttab_map_grant_ref *kmap_op,
> +			       unsigned long mfn);
>  extern struct page *m2p_find_override(unsigned long mfn);
>  extern unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn);
>  
> @@ -121,7 +124,7 @@ static inline unsigned long mfn_to_pfn(unsigned long mfn)
>  		pfn = m2p_find_override_pfn(mfn, ~0);
>  	}
>  
> -	/* 
> +	/*
>  	 * pfn is ~0 if there are no entries in the m2p for mfn or if the
>  	 * entry doesn't map back to the mfn and m2p_override doesn't have a
>  	 * valid entry for it.
> diff --git a/arch/x86/xen/p2m.c b/arch/x86/xen/p2m.c
> index 696c694..3e7cfa9 100644
> --- a/arch/x86/xen/p2m.c
> +++ b/arch/x86/xen/p2m.c
> @@ -881,6 +881,22 @@ static unsigned long mfn_hash(unsigned long mfn)
>  	return hash_long(mfn, M2P_OVERRIDE_HASH_SHIFT);
>  }
>  
> +int set_foreign_p2m_mapping(struct page *page, unsigned long mfn)
> +{
> +	unsigned long pfn = page_to_pfn(page);
> +
> +	WARN_ON(PagePrivate(page));
> +	SetPagePrivate(page);
> +	set_page_private(page, mfn);
> +
> +	page->index = pfn_to_mfn(pfn);
> +	if (unlikely(!set_phys_to_machine(pfn, FOREIGN_FRAME(mfn))))
> +		return -ENOMEM;
> +
> +	return 0;
> +}
> +EXPORT_SYMBOL_GPL(set_foreign_p2m_mapping);
> +
>  /* Add an MFN override for a particular page */
>  int m2p_add_override(unsigned long mfn, struct page *page,
>  		struct gnttab_map_grant_ref *kmap_op)
> @@ -899,13 +915,6 @@ int m2p_add_override(unsigned long mfn, struct page *page,
>  					"m2p_add_override: pfn %lx not mapped", pfn))
>  			return -EINVAL;
>  	}
> -	WARN_ON(PagePrivate(page));
> -	SetPagePrivate(page);
> -	set_page_private(page, mfn);
> -	page->index = pfn_to_mfn(pfn);
> -
> -	if (unlikely(!set_phys_to_machine(pfn, FOREIGN_FRAME(mfn))))
> -		return -ENOMEM;
>  
>  	if (kmap_op != NULL) {
>  		if (!PageHighMem(page)) {
> @@ -943,20 +952,33 @@ int m2p_add_override(unsigned long mfn, struct page *page,
>  	return 0;
>  }
>  EXPORT_SYMBOL_GPL(m2p_add_override);
> +
> +int restore_foreign_p2m_mapping(struct page *page, unsigned long mfn)
> +{
> +	unsigned long pfn = page_to_pfn(page);
> +
> +	if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT))
> +		return -EINVAL;
> +
> +	set_page_private(page, INVALID_P2M_ENTRY);
> +	WARN_ON(!PagePrivate(page));
> +	ClearPagePrivate(page);
> +	set_phys_to_machine(pfn, page->index);
> +	return 0;
> +}
> +EXPORT_SYMBOL_GPL(restore_foreign_p2m_mapping);
> +
>  int m2p_remove_override(struct page *page,
> -		struct gnttab_map_grant_ref *kmap_op)
> +			struct gnttab_map_grant_ref *kmap_op,
> +			unsigned long mfn)
>  {
>  	unsigned long flags;
> -	unsigned long mfn;
>  	unsigned long pfn;
>  	unsigned long uninitialized_var(address);
>  	unsigned level;
>  	pte_t *ptep = NULL;
>  
>  	pfn = page_to_pfn(page);
> -	mfn = get_phys_to_machine(pfn);
> -	if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT))
> -		return -EINVAL;
>  
>  	if (!PageHighMem(page)) {
>  		address = (unsigned long)__va(pfn << PAGE_SHIFT);
> @@ -970,10 +992,7 @@ int m2p_remove_override(struct page *page,
>  	spin_lock_irqsave(&m2p_override_lock, flags);
>  	list_del(&page->lru);
>  	spin_unlock_irqrestore(&m2p_override_lock, flags);
> -	WARN_ON(!PagePrivate(page));
> -	ClearPagePrivate(page);
>  
> -	set_phys_to_machine(pfn, page->index);
>  	if (kmap_op != NULL) {
>  		if (!PageHighMem(page)) {
>  			struct multicall_space mcs;
> diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c
> index 4b97b86..da18046 100644
> --- a/drivers/block/xen-blkback/blkback.c
> +++ b/drivers/block/xen-blkback/blkback.c
> @@ -285,8 +285,7 @@ static void free_persistent_gnts(struct xen_blkif *blkif, struct rb_root *root,
>  
>  		if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST ||
>  			!rb_next(&persistent_gnt->node)) {
> -			ret = gnttab_unmap_refs(unmap, NULL, pages,
> -				segs_to_unmap);
> +			ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
>  			BUG_ON(ret);
>  			put_free_pages(blkif, pages, segs_to_unmap);
>  			segs_to_unmap = 0;
> @@ -321,8 +320,7 @@ static void unmap_purged_grants(struct work_struct *work)
>  		pages[segs_to_unmap] = persistent_gnt->page;
>  
>  		if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
> -			ret = gnttab_unmap_refs(unmap, NULL, pages,
> -				segs_to_unmap);
> +			ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
>  			BUG_ON(ret);
>  			put_free_pages(blkif, pages, segs_to_unmap);
>  			segs_to_unmap = 0;
> @@ -330,7 +328,7 @@ static void unmap_purged_grants(struct work_struct *work)
>  		kfree(persistent_gnt);
>  	}
>  	if (segs_to_unmap > 0) {
> -		ret = gnttab_unmap_refs(unmap, NULL, pages, segs_to_unmap);
> +		ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
>  		BUG_ON(ret);
>  		put_free_pages(blkif, pages, segs_to_unmap);
>  	}
> @@ -670,15 +668,14 @@ static void xen_blkbk_unmap(struct xen_blkif *blkif,
>  				    GNTMAP_host_map, pages[i]->handle);
>  		pages[i]->handle = BLKBACK_INVALID_HANDLE;
>  		if (++invcount == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
> -			ret = gnttab_unmap_refs(unmap, NULL, unmap_pages,
> -			                        invcount);
> +			ret = gnttab_unmap_refs(unmap, unmap_pages, invcount);
>  			BUG_ON(ret);
>  			put_free_pages(blkif, unmap_pages, invcount);
>  			invcount = 0;
>  		}
>  	}
>  	if (invcount) {
> -		ret = gnttab_unmap_refs(unmap, NULL, unmap_pages, invcount);
> +		ret = gnttab_unmap_refs(unmap, unmap_pages, invcount);
>  		BUG_ON(ret);
>  		put_free_pages(blkif, unmap_pages, invcount);
>  	}
> @@ -740,7 +737,7 @@ again:
>  	}
>  
>  	if (segs_to_map) {
> -		ret = gnttab_map_refs(map, NULL, pages_to_gnt, segs_to_map);
> +		ret = gnttab_map_refs(map, pages_to_gnt, segs_to_map);
>  		BUG_ON(ret);
>  	}
>  
> diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
> index 073b4a1..34a2704 100644
> --- a/drivers/xen/gntdev.c
> +++ b/drivers/xen/gntdev.c
> @@ -284,8 +284,10 @@ static int map_grant_pages(struct grant_map *map)
>  	}
>  
>  	pr_debug("map %d+%d\n", map->index, map->count);
> -	err = gnttab_map_refs(map->map_ops, use_ptemod ? map->kmap_ops : NULL,
> -			map->pages, map->count);
> +	err = gnttab_map_refs_userspace(map->map_ops,
> +					use_ptemod ? map->kmap_ops : NULL,
> +					map->pages,
> +					map->count);
>  	if (err)
>  		return err;
>  
> @@ -315,9 +317,10 @@ static int __unmap_grant_pages(struct grant_map *map, int offset, int pages)
>  		}
>  	}
>  
> -	err = gnttab_unmap_refs(map->unmap_ops + offset,
> -			use_ptemod ? map->kmap_ops + offset : NULL, map->pages + offset,
> -			pages);
> +	err = gnttab_unmap_refs_userspace(map->unmap_ops + offset,
> +					  use_ptemod ? map->kmap_ops + offset : NULL,
> +					  map->pages + offset,
> +					  pages);
>  	if (err)
>  		return err;
>  
> diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c
> index b84e3ab..c719bc8 100644
> --- a/drivers/xen/grant-table.c
> +++ b/drivers/xen/grant-table.c
> @@ -928,15 +928,17 @@ void gnttab_batch_copy(struct gnttab_copy *batch, unsigned count)
>  }
>  EXPORT_SYMBOL_GPL(gnttab_batch_copy);
>  
> -int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
> +int __gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  		    struct gnttab_map_grant_ref *kmap_ops,
> -		    struct page **pages, unsigned int count)
> +		    struct page **pages, unsigned int count,
> +		    bool m2p_override)
>  {
>  	int i, ret;
>  	bool lazy = false;
>  	pte_t *pte;
>  	unsigned long mfn;
>  
> +	BUG_ON(kmap_ops && !m2p_override);
>  	ret = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, map_ops, count);
>  	if (ret)
>  		return ret;
> @@ -955,10 +957,12 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  			set_phys_to_machine(map_ops[i].host_addr >> PAGE_SHIFT,
>  					map_ops[i].dev_bus_addr >> PAGE_SHIFT);

I think that this should become a call to set_foreign_p2m_mapping


>  		}
> -		return ret;
> +		return 0;
>  	}
>  
> -	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
> +	if (m2p_override &&
> +	    !in_interrupt() &&
> +	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
>  		arch_enter_lazy_mmu_mode();
>  		lazy = true;
>  	}
> @@ -975,8 +979,14 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  		} else {
>  			mfn = PFN_DOWN(map_ops[i].dev_bus_addr);
>  		}
> -		ret = m2p_add_override(mfn, pages[i], kmap_ops ?
> -				       &kmap_ops[i] : NULL);
> +
> +		ret = set_foreign_p2m_mapping(pages[i], mfn);
> +		if (ret)
> +			goto out;
> +
> +		if (m2p_override)
> +			ret = m2p_add_override(mfn, pages[i], kmap_ops ?
> +					       &kmap_ops[i] : NULL);
>  		if (ret)
>  			goto out;
>  	}
> @@ -987,15 +997,31 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
>  
>  	return ret;
>  }
> +
> +int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
> +		    struct page **pages, unsigned int count)
> +{
> +	return __gnttab_map_refs(map_ops, NULL, pages, count, false);
> +}
>  EXPORT_SYMBOL_GPL(gnttab_map_refs);
>  
> -int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
> +int gnttab_map_refs_userspace(struct gnttab_map_grant_ref *map_ops,
> +			      struct gnttab_map_grant_ref *kmap_ops,
> +			      struct page **pages, unsigned int count)
> +{
> +	return __gnttab_map_refs(map_ops, kmap_ops, pages, count, true);
> +}
> +EXPORT_SYMBOL_GPL(gnttab_map_refs_userspace);
> +
> +int __gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
>  		      struct gnttab_map_grant_ref *kmap_ops,
> -		      struct page **pages, unsigned int count)
> +		      struct page **pages, unsigned int count,
> +		      bool m2p_override)
>  {
>  	int i, ret;
>  	bool lazy = false;
>  
> +	BUG_ON(kmap_ops && !m2p_override);
>  	ret = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, unmap_ops, count);
>  	if (ret)
>  		return ret;
> @@ -1006,17 +1032,27 @@ int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
>  			set_phys_to_machine(unmap_ops[i].host_addr >> PAGE_SHIFT,
>  					INVALID_P2M_ENTRY);

this should become a call to restore_foreign_p2m_mapping


>  		}
> -		return ret;
> +		return 0;
>  	}
>  
> -	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
> +	if (m2p_override &&
> +	    !in_interrupt() &&
> +	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
>  		arch_enter_lazy_mmu_mode();
>  		lazy = true;
>  	}
>  
>  	for (i = 0; i < count; i++) {
> -		ret = m2p_remove_override(pages[i], kmap_ops ?
> -				       &kmap_ops[i] : NULL);
> +		unsigned long mfn = get_phys_to_machine(page_to_pfn(pages[i]));
> +		ret = restore_foreign_p2m_mapping(pages[i], mfn);
> +		if (ret)
> +			goto out;
> +
> +		if (m2p_override)
> +			ret = m2p_remove_override(pages[i],
> +						  kmap_ops ?
> +						   &kmap_ops[i] : NULL,
> +						  mfn);
>  		if (ret)
>  			goto out;
>  	}
> @@ -1027,8 +1063,22 @@ int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
>  
>  	return ret;
>  }
> +
> +int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *map_ops,
> +		    struct page **pages, unsigned int count)
> +{
> +	return __gnttab_unmap_refs(map_ops, NULL, pages, count, false);
> +}
>  EXPORT_SYMBOL_GPL(gnttab_unmap_refs);
>  
> +int gnttab_unmap_refs_userspace(struct gnttab_unmap_grant_ref *map_ops,
> +				struct gnttab_map_grant_ref *kmap_ops,
> +				struct page **pages, unsigned int count)
> +{
> +	return __gnttab_unmap_refs(map_ops, kmap_ops, pages, count, true);
> +}
> +EXPORT_SYMBOL_GPL(gnttab_unmap_refs_userspace);
> +
>  static unsigned nr_status_frames(unsigned nr_grant_frames)
>  {
>  	BUG_ON(grefs_per_grant_frame == 0);
> diff --git a/include/xen/grant_table.h b/include/xen/grant_table.h
> index a5af2a2..7ad033d 100644
> --- a/include/xen/grant_table.h
> +++ b/include/xen/grant_table.h
> @@ -191,11 +191,15 @@ void gnttab_free_auto_xlat_frames(void);
>  #define gnttab_map_vaddr(map) ((void *)(map.host_virt_addr))
>  
>  int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
> -		    struct gnttab_map_grant_ref *kmap_ops,
>  		    struct page **pages, unsigned int count);
> +int gnttab_map_refs_userspace(struct gnttab_map_grant_ref *map_ops,
> +			      struct gnttab_map_grant_ref *kmap_ops,
> +			      struct page **pages, unsigned int count);
>  int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
> -		      struct gnttab_map_grant_ref *kunmap_ops,
>  		      struct page **pages, unsigned int count);
> +int gnttab_unmap_refs_userspace(struct gnttab_unmap_grant_ref *unmap_ops,
> +				struct gnttab_map_grant_ref *kunmap_ops,
> +				struct page **pages, unsigned int count);
>  
>  /* Perform a batch of grant map/copy operations. Retry every batch slot
>   * for which the hypervisor returns GNTST_eagain. This is typically due
> 

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [PATCH v8] xen/grant-table: Avoid m2p_override during mapping
@ 2014-02-06 12:56 Zoltan Kiss
  0 siblings, 0 replies; 6+ messages in thread
From: Zoltan Kiss @ 2014-02-06 12:56 UTC (permalink / raw)
  To: Konrad Rzeszutek Wilk, Boris Ostrovsky, David Vrabel,
	Thomas Gleixner, Ingo Molnar, H. Peter Anvin, x86, xen-devel,
	linux-kernel, Roger Pau Monné,
	Jan Beulich, Ian Campbell, stefano.stabellini
  Cc: Zoltan Kiss

The grant mapping API does m2p_override unnecessarily: only gntdev needs it,
for blkback and future netback patches it just cause a lock contention, as
those pages never go to userspace. Therefore this series does the following:
- the original functions were renamed to __gnttab_[un]map_refs, with a new
  parameter m2p_override
- based on m2p_override either they follow the original behaviour, or just set
  the private flag and call set_phys_to_machine
- gnttab_[un]map_refs are now a wrapper to call __gnttab_[un]map_refs with
  m2p_override false
- a new function gnttab_[un]map_refs_userspace provides the old behaviour
- it cuts out common parts from m2p_*_override functions to
  *_foreign_p2m_mapping functions

It also removes a stray space from page.h and change ret to 0 if
XENFEAT_auto_translated_physmap, as that is the only possible return value
there.

v2:
- move the storing of the old mfn in page->index to gnttab_map_refs
- move the function header update to a separate patch

v3:
- a new approach to retain old behaviour where it needed
- squash the patches into one

v4:
- move out the common bits from m2p* functions, and pass pfn/mfn as parameter
- clear page->private before doing anything with the page, so m2p_find_override
  won't race with this

v5:
- change return value handling in __gnttab_[un]map_refs
- remove a stray space in page.h
- add detail why ret = 0 now at some places

v6:
- don't pass pfn to m2p* functions, just get it locally

v7:
- the previous version broke build on ARM, as there is no need for those p2m
  changes. I've put them into arch specific functions, which are stubs on arm

v8:
- give credit to Anthony Liguori who submitted a very similar patch originally:
http://marc.info/?i=1384307336-5328-1-git-send-email-anthony%40codemonkey.ws
- create ARM stub for get_phys_to_machine
- move definition of mfn in __gnttab_unmap_refs to the right place

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Original-by: Anthony Liguori <aliguori@amazon.com>
---
 arch/arm/include/asm/xen/page.h     |   20 +++++++++-
 arch/x86/include/asm/xen/page.h     |    7 +++-
 arch/x86/xen/p2m.c                  |   49 ++++++++++++++++-------
 drivers/block/xen-blkback/blkback.c |   15 +++----
 drivers/xen/gntdev.c                |   13 +++---
 drivers/xen/grant-table.c           |   74 +++++++++++++++++++++++++++++------
 include/xen/grant_table.h           |    8 +++-
 7 files changed, 140 insertions(+), 46 deletions(-)

diff --git a/arch/arm/include/asm/xen/page.h b/arch/arm/include/asm/xen/page.h
index e0965ab..d26c3d7 100644
--- a/arch/arm/include/asm/xen/page.h
+++ b/arch/arm/include/asm/xen/page.h
@@ -97,13 +97,31 @@ static inline pte_t *lookup_address(unsigned long address, unsigned int *level)
 	return NULL;
 }
 
+static inline unsigned long get_phys_to_machine(unsigned long pfn)
+{
+	return 0;
+}
+
+static inline int set_foreign_p2m_mapping(struct page *page, unsigned long mfn)
+{
+	return 0;
+}
+
 static inline int m2p_add_override(unsigned long mfn, struct page *page,
 		struct gnttab_map_grant_ref *kmap_op)
 {
 	return 0;
 }
 
-static inline int m2p_remove_override(struct page *page, bool clear_pte)
+static inline int restore_foreign_p2m_mapping(struct page *page,
+					      unsigned long mfn)
+{
+	return 0;
+}
+
+static inline int m2p_remove_override(struct page *page,
+				      struct gnttab_map_grant_ref *kmap_op,
+				      unsigned long mfn)
 {
 	return 0;
 }
diff --git a/arch/x86/include/asm/xen/page.h b/arch/x86/include/asm/xen/page.h
index 3e276eb..0340954 100644
--- a/arch/x86/include/asm/xen/page.h
+++ b/arch/x86/include/asm/xen/page.h
@@ -49,10 +49,13 @@ extern bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn);
 extern unsigned long set_phys_range_identity(unsigned long pfn_s,
 					     unsigned long pfn_e);
 
+extern int set_foreign_p2m_mapping(struct page *page, unsigned long mfn);
 extern int m2p_add_override(unsigned long mfn, struct page *page,
 			    struct gnttab_map_grant_ref *kmap_op);
+extern int restore_foreign_p2m_mapping(struct page *page, unsigned long mfn);
 extern int m2p_remove_override(struct page *page,
-				struct gnttab_map_grant_ref *kmap_op);
+			       struct gnttab_map_grant_ref *kmap_op,
+			       unsigned long mfn);
 extern struct page *m2p_find_override(unsigned long mfn);
 extern unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn);
 
@@ -121,7 +124,7 @@ static inline unsigned long mfn_to_pfn(unsigned long mfn)
 		pfn = m2p_find_override_pfn(mfn, ~0);
 	}
 
-	/* 
+	/*
 	 * pfn is ~0 if there are no entries in the m2p for mfn or if the
 	 * entry doesn't map back to the mfn and m2p_override doesn't have a
 	 * valid entry for it.
diff --git a/arch/x86/xen/p2m.c b/arch/x86/xen/p2m.c
index 696c694..3e7cfa9 100644
--- a/arch/x86/xen/p2m.c
+++ b/arch/x86/xen/p2m.c
@@ -881,6 +881,22 @@ static unsigned long mfn_hash(unsigned long mfn)
 	return hash_long(mfn, M2P_OVERRIDE_HASH_SHIFT);
 }
 
+int set_foreign_p2m_mapping(struct page *page, unsigned long mfn)
+{
+	unsigned long pfn = page_to_pfn(page);
+
+	WARN_ON(PagePrivate(page));
+	SetPagePrivate(page);
+	set_page_private(page, mfn);
+
+	page->index = pfn_to_mfn(pfn);
+	if (unlikely(!set_phys_to_machine(pfn, FOREIGN_FRAME(mfn))))
+		return -ENOMEM;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(set_foreign_p2m_mapping);
+
 /* Add an MFN override for a particular page */
 int m2p_add_override(unsigned long mfn, struct page *page,
 		struct gnttab_map_grant_ref *kmap_op)
@@ -899,13 +915,6 @@ int m2p_add_override(unsigned long mfn, struct page *page,
 					"m2p_add_override: pfn %lx not mapped", pfn))
 			return -EINVAL;
 	}
-	WARN_ON(PagePrivate(page));
-	SetPagePrivate(page);
-	set_page_private(page, mfn);
-	page->index = pfn_to_mfn(pfn);
-
-	if (unlikely(!set_phys_to_machine(pfn, FOREIGN_FRAME(mfn))))
-		return -ENOMEM;
 
 	if (kmap_op != NULL) {
 		if (!PageHighMem(page)) {
@@ -943,20 +952,33 @@ int m2p_add_override(unsigned long mfn, struct page *page,
 	return 0;
 }
 EXPORT_SYMBOL_GPL(m2p_add_override);
+
+int restore_foreign_p2m_mapping(struct page *page, unsigned long mfn)
+{
+	unsigned long pfn = page_to_pfn(page);
+
+	if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT))
+		return -EINVAL;
+
+	set_page_private(page, INVALID_P2M_ENTRY);
+	WARN_ON(!PagePrivate(page));
+	ClearPagePrivate(page);
+	set_phys_to_machine(pfn, page->index);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(restore_foreign_p2m_mapping);
+
 int m2p_remove_override(struct page *page,
-		struct gnttab_map_grant_ref *kmap_op)
+			struct gnttab_map_grant_ref *kmap_op,
+			unsigned long mfn)
 {
 	unsigned long flags;
-	unsigned long mfn;
 	unsigned long pfn;
 	unsigned long uninitialized_var(address);
 	unsigned level;
 	pte_t *ptep = NULL;
 
 	pfn = page_to_pfn(page);
-	mfn = get_phys_to_machine(pfn);
-	if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT))
-		return -EINVAL;
 
 	if (!PageHighMem(page)) {
 		address = (unsigned long)__va(pfn << PAGE_SHIFT);
@@ -970,10 +992,7 @@ int m2p_remove_override(struct page *page,
 	spin_lock_irqsave(&m2p_override_lock, flags);
 	list_del(&page->lru);
 	spin_unlock_irqrestore(&m2p_override_lock, flags);
-	WARN_ON(!PagePrivate(page));
-	ClearPagePrivate(page);
 
-	set_phys_to_machine(pfn, page->index);
 	if (kmap_op != NULL) {
 		if (!PageHighMem(page)) {
 			struct multicall_space mcs;
diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c
index 4b97b86..da18046 100644
--- a/drivers/block/xen-blkback/blkback.c
+++ b/drivers/block/xen-blkback/blkback.c
@@ -285,8 +285,7 @@ static void free_persistent_gnts(struct xen_blkif *blkif, struct rb_root *root,
 
 		if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST ||
 			!rb_next(&persistent_gnt->node)) {
-			ret = gnttab_unmap_refs(unmap, NULL, pages,
-				segs_to_unmap);
+			ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
 			BUG_ON(ret);
 			put_free_pages(blkif, pages, segs_to_unmap);
 			segs_to_unmap = 0;
@@ -321,8 +320,7 @@ static void unmap_purged_grants(struct work_struct *work)
 		pages[segs_to_unmap] = persistent_gnt->page;
 
 		if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
-			ret = gnttab_unmap_refs(unmap, NULL, pages,
-				segs_to_unmap);
+			ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
 			BUG_ON(ret);
 			put_free_pages(blkif, pages, segs_to_unmap);
 			segs_to_unmap = 0;
@@ -330,7 +328,7 @@ static void unmap_purged_grants(struct work_struct *work)
 		kfree(persistent_gnt);
 	}
 	if (segs_to_unmap > 0) {
-		ret = gnttab_unmap_refs(unmap, NULL, pages, segs_to_unmap);
+		ret = gnttab_unmap_refs(unmap, pages, segs_to_unmap);
 		BUG_ON(ret);
 		put_free_pages(blkif, pages, segs_to_unmap);
 	}
@@ -670,15 +668,14 @@ static void xen_blkbk_unmap(struct xen_blkif *blkif,
 				    GNTMAP_host_map, pages[i]->handle);
 		pages[i]->handle = BLKBACK_INVALID_HANDLE;
 		if (++invcount == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
-			ret = gnttab_unmap_refs(unmap, NULL, unmap_pages,
-			                        invcount);
+			ret = gnttab_unmap_refs(unmap, unmap_pages, invcount);
 			BUG_ON(ret);
 			put_free_pages(blkif, unmap_pages, invcount);
 			invcount = 0;
 		}
 	}
 	if (invcount) {
-		ret = gnttab_unmap_refs(unmap, NULL, unmap_pages, invcount);
+		ret = gnttab_unmap_refs(unmap, unmap_pages, invcount);
 		BUG_ON(ret);
 		put_free_pages(blkif, unmap_pages, invcount);
 	}
@@ -740,7 +737,7 @@ again:
 	}
 
 	if (segs_to_map) {
-		ret = gnttab_map_refs(map, NULL, pages_to_gnt, segs_to_map);
+		ret = gnttab_map_refs(map, pages_to_gnt, segs_to_map);
 		BUG_ON(ret);
 	}
 
diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
index 073b4a1..34a2704 100644
--- a/drivers/xen/gntdev.c
+++ b/drivers/xen/gntdev.c
@@ -284,8 +284,10 @@ static int map_grant_pages(struct grant_map *map)
 	}
 
 	pr_debug("map %d+%d\n", map->index, map->count);
-	err = gnttab_map_refs(map->map_ops, use_ptemod ? map->kmap_ops : NULL,
-			map->pages, map->count);
+	err = gnttab_map_refs_userspace(map->map_ops,
+					use_ptemod ? map->kmap_ops : NULL,
+					map->pages,
+					map->count);
 	if (err)
 		return err;
 
@@ -315,9 +317,10 @@ static int __unmap_grant_pages(struct grant_map *map, int offset, int pages)
 		}
 	}
 
-	err = gnttab_unmap_refs(map->unmap_ops + offset,
-			use_ptemod ? map->kmap_ops + offset : NULL, map->pages + offset,
-			pages);
+	err = gnttab_unmap_refs_userspace(map->unmap_ops + offset,
+					  use_ptemod ? map->kmap_ops + offset : NULL,
+					  map->pages + offset,
+					  pages);
 	if (err)
 		return err;
 
diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c
index b84e3ab..c719bc8 100644
--- a/drivers/xen/grant-table.c
+++ b/drivers/xen/grant-table.c
@@ -928,15 +928,17 @@ void gnttab_batch_copy(struct gnttab_copy *batch, unsigned count)
 }
 EXPORT_SYMBOL_GPL(gnttab_batch_copy);
 
-int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
+int __gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
 		    struct gnttab_map_grant_ref *kmap_ops,
-		    struct page **pages, unsigned int count)
+		    struct page **pages, unsigned int count,
+		    bool m2p_override)
 {
 	int i, ret;
 	bool lazy = false;
 	pte_t *pte;
 	unsigned long mfn;
 
+	BUG_ON(kmap_ops && !m2p_override);
 	ret = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, map_ops, count);
 	if (ret)
 		return ret;
@@ -955,10 +957,12 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
 			set_phys_to_machine(map_ops[i].host_addr >> PAGE_SHIFT,
 					map_ops[i].dev_bus_addr >> PAGE_SHIFT);
 		}
-		return ret;
+		return 0;
 	}
 
-	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
+	if (m2p_override &&
+	    !in_interrupt() &&
+	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
 		arch_enter_lazy_mmu_mode();
 		lazy = true;
 	}
@@ -975,8 +979,14 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
 		} else {
 			mfn = PFN_DOWN(map_ops[i].dev_bus_addr);
 		}
-		ret = m2p_add_override(mfn, pages[i], kmap_ops ?
-				       &kmap_ops[i] : NULL);
+
+		ret = set_foreign_p2m_mapping(pages[i], mfn);
+		if (ret)
+			goto out;
+
+		if (m2p_override)
+			ret = m2p_add_override(mfn, pages[i], kmap_ops ?
+					       &kmap_ops[i] : NULL);
 		if (ret)
 			goto out;
 	}
@@ -987,15 +997,31 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
 
 	return ret;
 }
+
+int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
+		    struct page **pages, unsigned int count)
+{
+	return __gnttab_map_refs(map_ops, NULL, pages, count, false);
+}
 EXPORT_SYMBOL_GPL(gnttab_map_refs);
 
-int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
+int gnttab_map_refs_userspace(struct gnttab_map_grant_ref *map_ops,
+			      struct gnttab_map_grant_ref *kmap_ops,
+			      struct page **pages, unsigned int count)
+{
+	return __gnttab_map_refs(map_ops, kmap_ops, pages, count, true);
+}
+EXPORT_SYMBOL_GPL(gnttab_map_refs_userspace);
+
+int __gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
 		      struct gnttab_map_grant_ref *kmap_ops,
-		      struct page **pages, unsigned int count)
+		      struct page **pages, unsigned int count,
+		      bool m2p_override)
 {
 	int i, ret;
 	bool lazy = false;
 
+	BUG_ON(kmap_ops && !m2p_override);
 	ret = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, unmap_ops, count);
 	if (ret)
 		return ret;
@@ -1006,17 +1032,27 @@ int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
 			set_phys_to_machine(unmap_ops[i].host_addr >> PAGE_SHIFT,
 					INVALID_P2M_ENTRY);
 		}
-		return ret;
+		return 0;
 	}
 
-	if (!in_interrupt() && paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
+	if (m2p_override &&
+	    !in_interrupt() &&
+	    paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE) {
 		arch_enter_lazy_mmu_mode();
 		lazy = true;
 	}
 
 	for (i = 0; i < count; i++) {
-		ret = m2p_remove_override(pages[i], kmap_ops ?
-				       &kmap_ops[i] : NULL);
+		unsigned long mfn = get_phys_to_machine(page_to_pfn(pages[i]));
+		ret = restore_foreign_p2m_mapping(pages[i], mfn);
+		if (ret)
+			goto out;
+
+		if (m2p_override)
+			ret = m2p_remove_override(pages[i],
+						  kmap_ops ?
+						   &kmap_ops[i] : NULL,
+						  mfn);
 		if (ret)
 			goto out;
 	}
@@ -1027,8 +1063,22 @@ int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
 
 	return ret;
 }
+
+int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *map_ops,
+		    struct page **pages, unsigned int count)
+{
+	return __gnttab_unmap_refs(map_ops, NULL, pages, count, false);
+}
 EXPORT_SYMBOL_GPL(gnttab_unmap_refs);
 
+int gnttab_unmap_refs_userspace(struct gnttab_unmap_grant_ref *map_ops,
+				struct gnttab_map_grant_ref *kmap_ops,
+				struct page **pages, unsigned int count)
+{
+	return __gnttab_unmap_refs(map_ops, kmap_ops, pages, count, true);
+}
+EXPORT_SYMBOL_GPL(gnttab_unmap_refs_userspace);
+
 static unsigned nr_status_frames(unsigned nr_grant_frames)
 {
 	BUG_ON(grefs_per_grant_frame == 0);
diff --git a/include/xen/grant_table.h b/include/xen/grant_table.h
index a5af2a2..7ad033d 100644
--- a/include/xen/grant_table.h
+++ b/include/xen/grant_table.h
@@ -191,11 +191,15 @@ void gnttab_free_auto_xlat_frames(void);
 #define gnttab_map_vaddr(map) ((void *)(map.host_virt_addr))
 
 int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops,
-		    struct gnttab_map_grant_ref *kmap_ops,
 		    struct page **pages, unsigned int count);
+int gnttab_map_refs_userspace(struct gnttab_map_grant_ref *map_ops,
+			      struct gnttab_map_grant_ref *kmap_ops,
+			      struct page **pages, unsigned int count);
 int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops,
-		      struct gnttab_map_grant_ref *kunmap_ops,
 		      struct page **pages, unsigned int count);
+int gnttab_unmap_refs_userspace(struct gnttab_unmap_grant_ref *unmap_ops,
+				struct gnttab_map_grant_ref *kunmap_ops,
+				struct page **pages, unsigned int count);
 
 /* Perform a batch of grant map/copy operations. Retry every batch slot
  * for which the hypervisor returns GNTST_eagain. This is typically due

^ permalink raw reply related	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2014-02-11 18:55 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-02-06 12:56 [PATCH v8] xen/grant-table: Avoid m2p_override during mapping Zoltan Kiss
2014-02-07 18:42 ` David Vrabel
2014-02-07 18:42 ` [Xen-devel] " David Vrabel
2014-02-11 18:55 ` Stefano Stabellini
2014-02-11 18:55 ` Stefano Stabellini
  -- strict thread matches above, loose matches on Subject: below --
2014-02-06 12:56 Zoltan Kiss

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.