diff mbox series

[RFC,v2,2/3] mm/memblock: Add reserved memory release function

Message ID 173920224789.826592.13627374781649157364.stgit@devnote2 (mailing list archive)
State New
Headers show
Series tracing: Improve persistent ring buffer | expand

Commit Message

Masami Hiramatsu (Google) Feb. 10, 2025, 3:44 p.m. UTC 
From: Masami Hiramatsu (Google) <mhiramat@kernel.org>

Add reserved_mem_release_by_name() to release a reserved memory region
with a given name. This allows us to release reserved memory which is
defined by kernel cmdline, after boot.

Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
---
 include/linux/mm.h |    1 +
 mm/memblock.c      |   72 +++++++++++++++++++++++++++++++++++++++++++---------
 2 files changed, 61 insertions(+), 12 deletions(-)

Comments

Steven Rostedt Feb. 11, 2025, 12:03 a.m. UTC | #1 
On Tue, 11 Feb 2025 00:44:08 +0900
"Masami Hiramatsu (Google)" <mhiramat@kernel.org> wrote:

> From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> 
> Add reserved_mem_release_by_name() to release a reserved memory region
> with a given name. This allows us to release reserved memory which is
> defined by kernel cmdline, after boot.

Since Mike wrote the reserve_mem code, we should Cc him on changes.

Also, I would make patches 2 and 3 a separate patchset, as it is unrelated
to the other code.

> 
> Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> ---
>  include/linux/mm.h |    1 +
>  mm/memblock.c      |   72 +++++++++++++++++++++++++++++++++++++++++++---------
>  2 files changed, 61 insertions(+), 12 deletions(-)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index f02925447e59..a7201824c1fc 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -4197,6 +4197,7 @@ void vma_pgtable_walk_begin(struct vm_area_struct *vma);
>  void vma_pgtable_walk_end(struct vm_area_struct *vma);
>  
>  int reserve_mem_find_by_name(const char *name, phys_addr_t *start, phys_addr_t *size);
> +int reserved_mem_release_by_name(const char *name);

Should be "reserve_mem_release_by_name()" and not "reserved_mem" to stay consistent.

>  
>  #ifdef CONFIG_64BIT
>  int do_mseal(unsigned long start, size_t len_in, unsigned long flags);
> diff --git a/mm/memblock.c b/mm/memblock.c
> index 095c18b5c430..9962fbb90597 100644
> --- a/mm/memblock.c
> +++ b/mm/memblock.c
> @@ -16,6 +16,7 @@
>  #include <linux/kmemleak.h>
>  #include <linux/seq_file.h>
>  #include <linux/memblock.h>
> +#include <linux/mutex.h>
>  
>  #include <asm/sections.h>
>  #include <linux/io.h>
> @@ -2263,6 +2264,7 @@ struct reserve_mem_table {
>  };
>  static struct reserve_mem_table reserved_mem_table[RESERVE_MEM_MAX_ENTRIES];
>  static int reserved_mem_count;
> +static DEFINE_MUTEX(reserved_mem_lock);

"reserve_mem_lock".

>  
>  /* Add wildcard region with a lookup name */
>  static void __init reserved_mem_add(phys_addr_t start, phys_addr_t size,
> @@ -2276,6 +2278,21 @@ static void __init reserved_mem_add(phys_addr_t start, phys_addr_t size,
>  	strscpy(map->name, name);
>  }
>  
> +static struct reserve_mem_table *reserve_mem_find_by_name_nolock(const char *name)
> +{
> +	struct reserve_mem_table *map;
> +	int i;
> +
> +	for (i = 0; i < reserved_mem_count; i++) {
> +		map = &reserved_mem_table[i];
> +		if (!map->size)
> +			continue;
> +		if (strcmp(name, map->name) == 0)
> +			return map;
> +	}
> +	return NULL;
> +}
> +
>  /**
>   * reserve_mem_find_by_name - Find reserved memory region with a given name
>   * @name: The name that is attached to a reserved memory region
> @@ -2289,22 +2306,53 @@ static void __init reserved_mem_add(phys_addr_t start, phys_addr_t size,
>  int reserve_mem_find_by_name(const char *name, phys_addr_t *start, phys_addr_t *size)
>  {
>  	struct reserve_mem_table *map;
> -	int i;
>  
> -	for (i = 0; i < reserved_mem_count; i++) {
> -		map = &reserved_mem_table[i];
> -		if (!map->size)
> -			continue;
> -		if (strcmp(name, map->name) == 0) {
> -			*start = map->start;
> -			*size = map->size;
> -			return 1;
> -		}
> -	}
> -	return 0;
> +	guard(mutex)(&reserved_mem_lock);
> +	map = reserve_mem_find_by_name_nolock(name);
> +	if (!map)
> +		return 0;
> +
> +	*start = map->start;
> +	*size = map->size;
> +	return 1;
>  }
>  EXPORT_SYMBOL_GPL(reserve_mem_find_by_name);
>  
> +/**
> + * reserved_mem_release_by_name - Release reserved memory region with a given name
> + * @name: The name that is attatched to a reserved memory region
> + *
> + * Forcibly release the pages in the reserved memory region so that those memory
> + * can be used as free memory. After released the reserved region size becomes 0.
> + *
> + * Returns: 1 if released or 0 if not found.
> + */
> +int reserved_mem_release_by_name(const char *name)
> +{
> +	struct reserve_mem_table *map;
> +	unsigned int page_count;
> +	phys_addr_t start;
> +
> +	guard(mutex)(&reserved_mem_lock);
> +	map = reserve_mem_find_by_name_nolock(name);
> +	if (!map)
> +		return 0;
> +
> +	start = map->start;
> +	page_count = DIV_ROUND_UP(map->size, PAGE_SIZE);
> +
> +	for (int i = 0; i < page_count; i++) {
> +		phys_addr_t addr = start + i * PAGE_SIZE;
> +		struct page *page = pfn_to_page(addr >> PAGE_SHIFT);
> +
> +		page->flags &= ~BIT(PG_reserved);
> +		__free_page(page);
> +	}
> +	map->size = 0;
> +
> +	return 1;
> +}
> +
>  /*
>   * Parse reserve_mem=nn:align:name
>   */


Thanks,

-- Steve
Masami Hiramatsu (Google) Feb. 11, 2025, 1:43 a.m. UTC | #2 
On Mon, 10 Feb 2025 19:03:06 -0500
Steven Rostedt <rostedt@goodmis.org> wrote:

> On Tue, 11 Feb 2025 00:44:08 +0900
> "Masami Hiramatsu (Google)" <mhiramat@kernel.org> wrote:
> 
> > From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> > 
> > Add reserved_mem_release_by_name() to release a reserved memory region
> > with a given name. This allows us to release reserved memory which is
> > defined by kernel cmdline, after boot.
> 
> Since Mike wrote the reserve_mem code, we should Cc him on changes.
> 
> Also, I would make patches 2 and 3 a separate patchset, as it is unrelated
> to the other code.

OK, let me make a separate series and sending Cc mm maintainers too.


> 
> > 
> > Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> > ---
> >  include/linux/mm.h |    1 +
> >  mm/memblock.c      |   72 +++++++++++++++++++++++++++++++++++++++++++---------
> >  2 files changed, 61 insertions(+), 12 deletions(-)
> > 
> > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > index f02925447e59..a7201824c1fc 100644
> > --- a/include/linux/mm.h
> > +++ b/include/linux/mm.h
> > @@ -4197,6 +4197,7 @@ void vma_pgtable_walk_begin(struct vm_area_struct *vma);
> >  void vma_pgtable_walk_end(struct vm_area_struct *vma);
> >  
> >  int reserve_mem_find_by_name(const char *name, phys_addr_t *start, phys_addr_t *size);
> > +int reserved_mem_release_by_name(const char *name);
> 
> Should be "reserve_mem_release_by_name()" and not "reserved_mem" to stay consistent.

Ahh, good catch. I confused reserved_mem and reserve_mem.

> 
> >  
> >  #ifdef CONFIG_64BIT
> >  int do_mseal(unsigned long start, size_t len_in, unsigned long flags);
> > diff --git a/mm/memblock.c b/mm/memblock.c
> > index 095c18b5c430..9962fbb90597 100644
> > --- a/mm/memblock.c
> > +++ b/mm/memblock.c
> > @@ -16,6 +16,7 @@
> >  #include <linux/kmemleak.h>
> >  #include <linux/seq_file.h>
> >  #include <linux/memblock.h>
> > +#include <linux/mutex.h>
> >  
> >  #include <asm/sections.h>
> >  #include <linux/io.h>
> > @@ -2263,6 +2264,7 @@ struct reserve_mem_table {
> >  };
> >  static struct reserve_mem_table reserved_mem_table[RESERVE_MEM_MAX_ENTRIES];
> >  static int reserved_mem_count;
> > +static DEFINE_MUTEX(reserved_mem_lock);
> 
> "reserve_mem_lock".

OK.

Thanks!

> 
> >  
> >  /* Add wildcard region with a lookup name */
> >  static void __init reserved_mem_add(phys_addr_t start, phys_addr_t size,
> > @@ -2276,6 +2278,21 @@ static void __init reserved_mem_add(phys_addr_t start, phys_addr_t size,
> >  	strscpy(map->name, name);
> >  }
> >  
> > +static struct reserve_mem_table *reserve_mem_find_by_name_nolock(const char *name)
> > +{
> > +	struct reserve_mem_table *map;
> > +	int i;
> > +
> > +	for (i = 0; i < reserved_mem_count; i++) {
> > +		map = &reserved_mem_table[i];
> > +		if (!map->size)
> > +			continue;
> > +		if (strcmp(name, map->name) == 0)
> > +			return map;
> > +	}
> > +	return NULL;
> > +}
> > +
> >  /**
> >   * reserve_mem_find_by_name - Find reserved memory region with a given name
> >   * @name: The name that is attached to a reserved memory region
> > @@ -2289,22 +2306,53 @@ static void __init reserved_mem_add(phys_addr_t start, phys_addr_t size,
> >  int reserve_mem_find_by_name(const char *name, phys_addr_t *start, phys_addr_t *size)
> >  {
> >  	struct reserve_mem_table *map;
> > -	int i;
> >  
> > -	for (i = 0; i < reserved_mem_count; i++) {
> > -		map = &reserved_mem_table[i];
> > -		if (!map->size)
> > -			continue;
> > -		if (strcmp(name, map->name) == 0) {
> > -			*start = map->start;
> > -			*size = map->size;
> > -			return 1;
> > -		}
> > -	}
> > -	return 0;
> > +	guard(mutex)(&reserved_mem_lock);
> > +	map = reserve_mem_find_by_name_nolock(name);
> > +	if (!map)
> > +		return 0;
> > +
> > +	*start = map->start;
> > +	*size = map->size;
> > +	return 1;
> >  }
> >  EXPORT_SYMBOL_GPL(reserve_mem_find_by_name);
> >  
> > +/**
> > + * reserved_mem_release_by_name - Release reserved memory region with a given name
> > + * @name: The name that is attatched to a reserved memory region
> > + *
> > + * Forcibly release the pages in the reserved memory region so that those memory
> > + * can be used as free memory. After released the reserved region size becomes 0.
> > + *
> > + * Returns: 1 if released or 0 if not found.
> > + */
> > +int reserved_mem_release_by_name(const char *name)
> > +{
> > +	struct reserve_mem_table *map;
> > +	unsigned int page_count;
> > +	phys_addr_t start;
> > +
> > +	guard(mutex)(&reserved_mem_lock);
> > +	map = reserve_mem_find_by_name_nolock(name);
> > +	if (!map)
> > +		return 0;
> > +
> > +	start = map->start;
> > +	page_count = DIV_ROUND_UP(map->size, PAGE_SIZE);
> > +
> > +	for (int i = 0; i < page_count; i++) {
> > +		phys_addr_t addr = start + i * PAGE_SIZE;
> > +		struct page *page = pfn_to_page(addr >> PAGE_SHIFT);
> > +
> > +		page->flags &= ~BIT(PG_reserved);
> > +		__free_page(page);
> > +	}
> > +	map->size = 0;
> > +
> > +	return 1;
> > +}
> > +
> >  /*
> >   * Parse reserve_mem=nn:align:name
> >   */
> 
> 
> Thanks,
> 
> -- Steve
diff mbox series

Patch

diff --git a/include/linux/mm.h b/include/linux/mm.h
index f02925447e59..a7201824c1fc 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -4197,6 +4197,7 @@  void vma_pgtable_walk_begin(struct vm_area_struct *vma);
 void vma_pgtable_walk_end(struct vm_area_struct *vma);
 
 int reserve_mem_find_by_name(const char *name, phys_addr_t *start, phys_addr_t *size);
+int reserved_mem_release_by_name(const char *name);
 
 #ifdef CONFIG_64BIT
 int do_mseal(unsigned long start, size_t len_in, unsigned long flags);
diff --git a/mm/memblock.c b/mm/memblock.c
index 095c18b5c430..9962fbb90597 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -16,6 +16,7 @@ 
 #include <linux/kmemleak.h>
 #include <linux/seq_file.h>
 #include <linux/memblock.h>
+#include <linux/mutex.h>
 
 #include <asm/sections.h>
 #include <linux/io.h>
@@ -2263,6 +2264,7 @@  struct reserve_mem_table {
 };
 static struct reserve_mem_table reserved_mem_table[RESERVE_MEM_MAX_ENTRIES];
 static int reserved_mem_count;
+static DEFINE_MUTEX(reserved_mem_lock);
 
 /* Add wildcard region with a lookup name */
 static void __init reserved_mem_add(phys_addr_t start, phys_addr_t size,
@@ -2276,6 +2278,21 @@  static void __init reserved_mem_add(phys_addr_t start, phys_addr_t size,
 	strscpy(map->name, name);
 }
 
+static struct reserve_mem_table *reserve_mem_find_by_name_nolock(const char *name)
+{
+	struct reserve_mem_table *map;
+	int i;
+
+	for (i = 0; i < reserved_mem_count; i++) {
+		map = &reserved_mem_table[i];
+		if (!map->size)
+			continue;
+		if (strcmp(name, map->name) == 0)
+			return map;
+	}
+	return NULL;
+}
+
 /**
  * reserve_mem_find_by_name - Find reserved memory region with a given name
  * @name: The name that is attached to a reserved memory region
@@ -2289,22 +2306,53 @@  static void __init reserved_mem_add(phys_addr_t start, phys_addr_t size,
 int reserve_mem_find_by_name(const char *name, phys_addr_t *start, phys_addr_t *size)
 {
 	struct reserve_mem_table *map;
-	int i;
 
-	for (i = 0; i < reserved_mem_count; i++) {
-		map = &reserved_mem_table[i];
-		if (!map->size)
-			continue;
-		if (strcmp(name, map->name) == 0) {
-			*start = map->start;
-			*size = map->size;
-			return 1;
-		}
-	}
-	return 0;
+	guard(mutex)(&reserved_mem_lock);
+	map = reserve_mem_find_by_name_nolock(name);
+	if (!map)
+		return 0;
+
+	*start = map->start;
+	*size = map->size;
+	return 1;
 }
 EXPORT_SYMBOL_GPL(reserve_mem_find_by_name);
 
+/**
+ * reserved_mem_release_by_name - Release reserved memory region with a given name
+ * @name: The name that is attatched to a reserved memory region
+ *
+ * Forcibly release the pages in the reserved memory region so that those memory
+ * can be used as free memory. After released the reserved region size becomes 0.
+ *
+ * Returns: 1 if released or 0 if not found.
+ */
+int reserved_mem_release_by_name(const char *name)
+{
+	struct reserve_mem_table *map;
+	unsigned int page_count;
+	phys_addr_t start;
+
+	guard(mutex)(&reserved_mem_lock);
+	map = reserve_mem_find_by_name_nolock(name);
+	if (!map)
+		return 0;
+
+	start = map->start;
+	page_count = DIV_ROUND_UP(map->size, PAGE_SIZE);
+
+	for (int i = 0; i < page_count; i++) {
+		phys_addr_t addr = start + i * PAGE_SIZE;
+		struct page *page = pfn_to_page(addr >> PAGE_SHIFT);
+
+		page->flags &= ~BIT(PG_reserved);
+		__free_page(page);
+	}
+	map->size = 0;
+
+	return 1;
+}
+
 /*
  * Parse reserve_mem=nn:align:name
  */