[PATCH 13/26] ppc: Convert mmu context allocation to new IDA API

ida_alloc_range is the perfect fit for this use case.  Eliminates
a custom spinlock, a call to ida_pre_get and a local check for the
allocated ID exceeding a maximum.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 arch/powerpc/mm/mmu_context_book3s64.c | 44 +++-----------------------
 1 file changed, 4 insertions(+), 40 deletions(-)

diff --git a/arch/powerpc/mm/mmu_context_book3s64.c b/arch/powerpc/mm/mmu_context_book3s64.c
index f3d4b4a0e561..5a0cf2cc8ba0 100644
--- a/arch/powerpc/mm/mmu_context_book3s64.c
+++ b/arch/powerpc/mm/mmu_context_book3s64.c

@@ -26,48 +26,16 @@
 #include <asm/mmu_context.h>
 #include <asm/pgalloc.h>
 
-static DEFINE_SPINLOCK(mmu_context_lock);
 static DEFINE_IDA(mmu_context_ida);
 
 static int alloc_context_id(int min_id, int max_id)
 {
-	int index, err;
-
-again:
-	if (!ida_pre_get(&mmu_context_ida, GFP_KERNEL))
-		return -ENOMEM;
-
-	spin_lock(&mmu_context_lock);
-	err = ida_get_new_above(&mmu_context_ida, min_id, &index);
-	spin_unlock(&mmu_context_lock);
-
-	if (err == -EAGAIN)
-		goto again;
-	else if (err)
-		return err;
-
-	if (index > max_id) {
-		spin_lock(&mmu_context_lock);
-		ida_remove(&mmu_context_ida, index);
-		spin_unlock(&mmu_context_lock);
-		return -ENOMEM;
-	}
-
-	return index;
+	return ida_alloc_range(&mmu_context_ida, min_id, max_id, GFP_KERNEL);
 }
 
 void hash__reserve_context_id(int id)
 {
-	int rc, result = 0;
-
-	do {
-		if (!ida_pre_get(&mmu_context_ida, GFP_KERNEL))
-			break;
-
-		spin_lock(&mmu_context_lock);
-		rc = ida_get_new_above(&mmu_context_ida, id, &result);
-		spin_unlock(&mmu_context_lock);
-	} while (rc == -EAGAIN);
+	int result = ida_alloc_range(&mmu_context_ida, id, id, GFP_KERNEL);
 
 	WARN(result != id, "mmu: Failed to reserve context id %d (rc %d)\n", id, result);
 }

@@ -172,9 +140,7 @@ int init_new_context(struct task_struct *tsk, struct mm_struct *mm)
 
 void __destroy_context(int context_id)
 {
-	spin_lock(&mmu_context_lock);
-	ida_remove(&mmu_context_ida, context_id);
-	spin_unlock(&mmu_context_lock);
+	ida_free(&mmu_context_ida, context_id);
 }
 EXPORT_SYMBOL_GPL(__destroy_context);
 

@@ -182,13 +148,11 @@ static void destroy_contexts(mm_context_t *ctx)
 {
 	int index, context_id;
 
-	spin_lock(&mmu_context_lock);
 	for (index = 0; index < ARRAY_SIZE(ctx->extended_id); index++) {
 		context_id = ctx->extended_id[index];
 		if (context_id)
-			ida_remove(&mmu_context_ida, context_id);
+			ida_free(&mmu_context_ida, context_id);
 	}
-	spin_unlock(&mmu_context_lock);
 }
 
 static void pte_frag_destroy(void *pte_frag)

-- 
2.17.1

On Thu, 21 Jun 2018 14:28:22 -0700
Matthew Wilcox [off-list ref] wrote:

This hunk should be okay because the mmu_context_lock does not protect
the extended_id array, right Aneesh?

Reviewed-by: Nicholas Piggin <npiggin@gmail.com>

Thanks,
Nick

On Fri, Jun 22, 2018 at 12:15:11PM +1000, Nicholas Piggin wrote:

On Thu, 21 Jun 2018 14:28:22 -0700
Matthew Wilcox [off-list ref] wrote:

quoted

 static int alloc_context_id(int min_id, int max_id)

...

quoted

-	spin_lock(&mmu_context_lock);
-	err = ida_get_new_above(&mmu_context_ida, min_id, &index);
-	spin_unlock(&mmu_context_lock);

...

quoted

@@ -182,13 +148,11 @@ static void destroy_contexts(mm_context_t *ctx)
 {
 	int index, context_id;
 
-	spin_lock(&mmu_context_lock);
 	for (index = 0; index < ARRAY_SIZE(ctx->extended_id); index++) {
 		context_id = ctx->extended_id[index];
 		if (context_id)
-			ida_remove(&mmu_context_ida, context_id);
+			ida_free(&mmu_context_ida, context_id);
 	}
-	spin_unlock(&mmu_context_lock);
 }
 
 static void pte_frag_destroy(void *pte_frag)

This hunk should be okay because the mmu_context_lock does not protect
the extended_id array, right Aneesh?

That's my understanding.  The code today does this:

static inline int alloc_extended_context(struct mm_struct *mm,
                                         unsigned long ea)
{
        int context_id;

        int index = ea >> MAX_EA_BITS_PER_CONTEXT;

        context_id = hash__alloc_context_id();
        if (context_id < 0)
                return context_id;

        VM_WARN_ON(mm->context.extended_id[index]);
        mm->context.extended_id[index] = context_id;

so it's not currently protected by this lock.  I suppose we are currently
protected from destroy_contexts() being called twice simultaneously, but
you'll notice that we don't zero the array elements in destroy_contexts(),
so if we somehow had a code path which could call it concurrently, we'd
be seeing warnings when the second caller tried to remove the context
IDs from the IDA.  I deduced that something else must be preventing
this situation from occurring (like, oh i don't know, this function only
being called on process exit, so implicitly only called once per context).

Reviewed-by: Nicholas Piggin <npiggin@gmail.com>

Thanks.

On Thu, 21 Jun 2018 21:38:15 -0700
Matthew Wilcox [off-list ref] wrote:

On Fri, Jun 22, 2018 at 12:15:11PM +1000, Nicholas Piggin wrote:

quoted

On Thu, 21 Jun 2018 14:28:22 -0700
Matthew Wilcox [off-list ref] wrote:  

quoted

 static int alloc_context_id(int min_id, int max_id)  

...

quoted

quoted

-	spin_lock(&mmu_context_lock);
-	err = ida_get_new_above(&mmu_context_ida, min_id, &index);
-	spin_unlock(&mmu_context_lock);  

...

quoted

quoted

@@ -182,13 +148,11 @@ static void destroy_contexts(mm_context_t *ctx)
 {
 	int index, context_id;
 
-	spin_lock(&mmu_context_lock);
 	for (index = 0; index < ARRAY_SIZE(ctx->extended_id); index++) {
 		context_id = ctx->extended_id[index];
 		if (context_id)
-			ida_remove(&mmu_context_ida, context_id);
+			ida_free(&mmu_context_ida, context_id);
 	}
-	spin_unlock(&mmu_context_lock);
 }
 
 static void pte_frag_destroy(void *pte_frag)  

This hunk should be okay because the mmu_context_lock does not protect
the extended_id array, right Aneesh?  

That's my understanding.  The code today does this:

static inline int alloc_extended_context(struct mm_struct *mm,
                                         unsigned long ea)
{
        int context_id;

        int index = ea >> MAX_EA_BITS_PER_CONTEXT;

        context_id = hash__alloc_context_id();
        if (context_id < 0)
                return context_id;

        VM_WARN_ON(mm->context.extended_id[index]);
        mm->context.extended_id[index] = context_id;

so it's not currently protected by this lock.  I suppose we are currently
protected from destroy_contexts() being called twice simultaneously, but
you'll notice that we don't zero the array elements in destroy_contexts(),
so if we somehow had a code path which could call it concurrently, we'd
be seeing warnings when the second caller tried to remove the context

Yeah that'd be an existing bug.

IDs from the IDA.  I deduced that something else must be preventing
this situation from occurring (like, oh i don't know, this function only
being called on process exit, so implicitly only called once per context).

I think that's exactly right.

Thanks,
Nick

Nicholas Piggin [off-list ref] writes:

On Thu, 21 Jun 2018 14:28:22 -0700
Matthew Wilcox [off-list ref] wrote:

quoted

ida_alloc_range is the perfect fit for this use case.  Eliminates
a custom spinlock, a call to ida_pre_get and a local check for the
allocated ID exceeding a maximum.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 arch/powerpc/mm/mmu_context_book3s64.c | 44 +++-----------------------
 1 file changed, 4 insertions(+), 40 deletions(-)

diff --git a/arch/powerpc/mm/mmu_context_book3s64.c b/arch/powerpc/mm/mmu_context_book3s64.c
index f3d4b4a0e561..5a0cf2cc8ba0 100644
--- a/arch/powerpc/mm/mmu_context_book3s64.c
+++ b/arch/powerpc/mm/mmu_context_book3s64.c

@@ -26,48 +26,16 @@
 #include <asm/mmu_context.h>
 #include <asm/pgalloc.h>
 
-static DEFINE_SPINLOCK(mmu_context_lock);
 static DEFINE_IDA(mmu_context_ida);
 
 static int alloc_context_id(int min_id, int max_id)
 {
-	int index, err;
-
-again:
-	if (!ida_pre_get(&mmu_context_ida, GFP_KERNEL))
-		return -ENOMEM;
-
-	spin_lock(&mmu_context_lock);
-	err = ida_get_new_above(&mmu_context_ida, min_id, &index);
-	spin_unlock(&mmu_context_lock);
-
-	if (err == -EAGAIN)
-		goto again;
-	else if (err)
-		return err;
-
-	if (index > max_id) {
-		spin_lock(&mmu_context_lock);
-		ida_remove(&mmu_context_ida, index);
-		spin_unlock(&mmu_context_lock);
-		return -ENOMEM;
-	}
-
-	return index;
+	return ida_alloc_range(&mmu_context_ida, min_id, max_id, GFP_KERNEL);
 }
 
 void hash__reserve_context_id(int id)
 {
-	int rc, result = 0;
-
-	do {
-		if (!ida_pre_get(&mmu_context_ida, GFP_KERNEL))
-			break;
-
-		spin_lock(&mmu_context_lock);
-		rc = ida_get_new_above(&mmu_context_ida, id, &result);
-		spin_unlock(&mmu_context_lock);
-	} while (rc == -EAGAIN);
+	int result = ida_alloc_range(&mmu_context_ida, id, id, GFP_KERNEL);
 
 	WARN(result != id, "mmu: Failed to reserve context id %d (rc %d)\n", id, result);
 }

@@ -172,9 +140,7 @@ int init_new_context(struct task_struct *tsk, struct mm_struct *mm)
 
 void __destroy_context(int context_id)
 {
-	spin_lock(&mmu_context_lock);
-	ida_remove(&mmu_context_ida, context_id);
-	spin_unlock(&mmu_context_lock);
+	ida_free(&mmu_context_ida, context_id);
 }
 EXPORT_SYMBOL_GPL(__destroy_context);
 

@@ -182,13 +148,11 @@ static void destroy_contexts(mm_context_t *ctx)
 {
 	int index, context_id;
 
-	spin_lock(&mmu_context_lock);
 	for (index = 0; index < ARRAY_SIZE(ctx->extended_id); index++) {
 		context_id = ctx->extended_id[index];
 		if (context_id)
-			ida_remove(&mmu_context_ida, context_id);
+			ida_free(&mmu_context_ida, context_id);
 	}
-	spin_unlock(&mmu_context_lock);
 }
 
 static void pte_frag_destroy(void *pte_frag)

This hunk should be okay because the mmu_context_lock does not protect
the extended_id array, right Aneesh?

Yes. This is called at process exit, so we should not find parallel
calls. On the allocation side, we are protected by mmap_sem. We do
allocate extended_id when doing mmap.

-aneesh

Matthew Wilcox [off-list ref] writes:

this situation from occurring (like, oh i don't know, this function only
being called on process exit, so implicitly only called once per context).

That is correct.