[PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion

Ye Liu posted 1 patch 2 months, 2 weeks ago
include/linux/mm.h    | 9 +++++++++
kernel/rcu/rcuscale.c | 2 +-
kernel/sched/fair.c   | 5 ++---
mm/backing-dev.c      | 2 +-
mm/huge_memory.c      | 2 +-
mm/swap.c             | 2 +-
6 files changed, 15 insertions(+), 7 deletions(-)
[PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by Ye Liu 2 months, 2 weeks ago
From: Ye Liu <liuye@kylinos.cn>

Replace repeated (20 - PAGE_SHIFT) calculations with standard macros:
- MB_TO_PAGES(mb)    converts MB to page count
- PAGES_TO_MB(pages) converts pages to MB

No functional change.

Signed-off-by: Ye Liu <liuye@kylinos.cn>
---
 include/linux/mm.h    | 9 +++++++++
 kernel/rcu/rcuscale.c | 2 +-
 kernel/sched/fair.c   | 5 ++---
 mm/backing-dev.c      | 2 +-
 mm/huge_memory.c      | 2 +-
 mm/swap.c             | 2 +-
 6 files changed, 15 insertions(+), 7 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 957acde6ae62..0c1b2c074142 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -69,6 +69,15 @@ static inline void totalram_pages_add(long count)
 
 extern void * high_memory;
 
+/*
+ * Convert between pages and MB
+ * 20 is the shift for 1MB (2^20 = 1MB)
+ * PAGE_SHIFT is the shift for page size (e.g., 12 for 4KB pages)
+ * So (20 - PAGE_SHIFT) converts between pages and MB
+ */
+#define PAGES_TO_MB(pages) ((pages) >> (20 - PAGE_SHIFT))
+#define MB_TO_PAGES(mb)    ((mb) << (20 - PAGE_SHIFT))
+
 #ifdef CONFIG_SYSCTL
 extern int sysctl_legacy_va_layout;
 #else
diff --git a/kernel/rcu/rcuscale.c b/kernel/rcu/rcuscale.c
index b521d0455992..7484d8ad5767 100644
--- a/kernel/rcu/rcuscale.c
+++ b/kernel/rcu/rcuscale.c
@@ -796,7 +796,7 @@ kfree_scale_thread(void *arg)
 		pr_alert("Total time taken by all kfree'ers: %llu ns, loops: %d, batches: %ld, memory footprint: %lldMB\n",
 		       (unsigned long long)(end_time - start_time), kfree_loops,
 		       rcuscale_seq_diff(b_rcu_gp_test_finished, b_rcu_gp_test_started),
-		       (mem_begin - mem_during) >> (20 - PAGE_SHIFT));
+		       PAGES_TO_MB(mem_begin - mem_during));
 
 		if (shutdown) {
 			smp_mb(); /* Assign before wake. */
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index b9b4bbbf0af6..ae1d9a7ef202 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -1489,7 +1489,7 @@ static unsigned int task_nr_scan_windows(struct task_struct *p)
 	 * by the PTE scanner and NUMA hinting faults should be trapped based
 	 * on resident pages
 	 */
-	nr_scan_pages = sysctl_numa_balancing_scan_size << (20 - PAGE_SHIFT);
+	nr_scan_pages = MB_TO_PAGES(sysctl_numa_balancing_scan_size);
 	rss = get_mm_rss(p->mm);
 	if (!rss)
 		rss = nr_scan_pages;
@@ -1926,8 +1926,7 @@ bool should_numa_migrate_memory(struct task_struct *p, struct folio *folio,
 		}
 
 		def_th = sysctl_numa_balancing_hot_threshold;
-		rate_limit = sysctl_numa_balancing_promote_rate_limit << \
-			(20 - PAGE_SHIFT);
+		rate_limit = MB_TO_PAGES(sysctl_numa_balancing_promote_rate_limit);
 		numa_promotion_adjust_threshold(pgdat, rate_limit, def_th);
 
 		th = pgdat->nbp_threshold ? : def_th;
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index 783904d8c5ef..e4d578e6121c 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -510,7 +510,7 @@ static void wb_update_bandwidth_workfn(struct work_struct *work)
 /*
  * Initial write bandwidth: 100 MB/s
  */
-#define INIT_BW		(100 << (20 - PAGE_SHIFT))
+#define INIT_BW		MB_TO_PAGES(100)
 
 static int wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi,
 		   gfp_t gfp)
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 389620c65a5f..dcc33d9c300f 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -911,7 +911,7 @@ static int __init hugepage_init(void)
 	 * where the extra memory used could hurt more than TLB overhead
 	 * is likely to save.  The admin can still enable it through /sys.
 	 */
-	if (totalram_pages() < (512 << (20 - PAGE_SHIFT))) {
+	if (totalram_pages() < MB_TO_PAGES(512)) {
 		transparent_hugepage_flags = 0;
 		return 0;
 	}
diff --git a/mm/swap.c b/mm/swap.c
index 3632dd061beb..cb164f9ef9e3 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -1096,7 +1096,7 @@ static const struct ctl_table swap_sysctl_table[] = {
  */
 void __init swap_setup(void)
 {
-	unsigned long megs = totalram_pages() >> (20 - PAGE_SHIFT);
+	unsigned long megs = PAGES_TO_MB(totalram_pages());
 
 	/* Use a smaller cluster for small-memory machines */
 	if (megs < 16)
-- 
2.43.0
Re: [PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by Chris Li 2 months, 2 weeks ago
On Thu, Jul 17, 2025 at 7:42 PM Ye Liu <ye.liu@linux.dev> wrote:
>
> From: Ye Liu <liuye@kylinos.cn>
>
> Replace repeated (20 - PAGE_SHIFT) calculations with standard macros:
> - MB_TO_PAGES(mb)    converts MB to page count
> - PAGES_TO_MB(pages) converts pages to MB
>
> No functional change.

Thanks for doing this.

Acked-by: Chris Li <chrisl@kernel.org>

Chris
Re: [PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by Zi Yan 2 months, 2 weeks ago
On 17 Jul 2025, at 22:41, Ye Liu wrote:

> From: Ye Liu <liuye@kylinos.cn>
>
> Replace repeated (20 - PAGE_SHIFT) calculations with standard macros:
> - MB_TO_PAGES(mb)    converts MB to page count
> - PAGES_TO_MB(pages) converts pages to MB
>
> No functional change.
>
> Signed-off-by: Ye Liu <liuye@kylinos.cn>
> ---
>  include/linux/mm.h    | 9 +++++++++
>  kernel/rcu/rcuscale.c | 2 +-
>  kernel/sched/fair.c   | 5 ++---
>  mm/backing-dev.c      | 2 +-
>  mm/huge_memory.c      | 2 +-
>  mm/swap.c             | 2 +-
>  6 files changed, 15 insertions(+), 7 deletions(-)
>

Thanks.

Acked-by: Zi Yan <ziy@nvidia.com>

--
Best Regards,
Yan, Zi
Re: [PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by Lorenzo Stoakes 2 months, 2 weeks ago
On Fri, Jul 18, 2025 at 10:41:32AM +0800, Ye Liu wrote:
> From: Ye Liu <liuye@kylinos.cn>
>
> Replace repeated (20 - PAGE_SHIFT) calculations with standard macros:
> - MB_TO_PAGES(mb)    converts MB to page count
> - PAGES_TO_MB(pages) converts pages to MB
>
> No functional change.
>
> Signed-off-by: Ye Liu <liuye@kylinos.cn>

Nice idea :)

NOte I see arch/x86/include/asm/pgtable.h has a pages_to_mb() static inline
declaration, but probably being an asm include can't ref mm.h so meh not a big
deal.

LGTM, so:

Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>

> ---
>  include/linux/mm.h    | 9 +++++++++
>  kernel/rcu/rcuscale.c | 2 +-
>  kernel/sched/fair.c   | 5 ++---
>  mm/backing-dev.c      | 2 +-
>  mm/huge_memory.c      | 2 +-
>  mm/swap.c             | 2 +-
>  6 files changed, 15 insertions(+), 7 deletions(-)
>
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 957acde6ae62..0c1b2c074142 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -69,6 +69,15 @@ static inline void totalram_pages_add(long count)
>
>  extern void * high_memory;
>
> +/*
> + * Convert between pages and MB
> + * 20 is the shift for 1MB (2^20 = 1MB)
> + * PAGE_SHIFT is the shift for page size (e.g., 12 for 4KB pages)
> + * So (20 - PAGE_SHIFT) converts between pages and MB
> + */
> +#define PAGES_TO_MB(pages) ((pages) >> (20 - PAGE_SHIFT))
> +#define MB_TO_PAGES(mb)    ((mb) << (20 - PAGE_SHIFT))
> +
>  #ifdef CONFIG_SYSCTL
>  extern int sysctl_legacy_va_layout;
>  #else
> diff --git a/kernel/rcu/rcuscale.c b/kernel/rcu/rcuscale.c
> index b521d0455992..7484d8ad5767 100644
> --- a/kernel/rcu/rcuscale.c
> +++ b/kernel/rcu/rcuscale.c
> @@ -796,7 +796,7 @@ kfree_scale_thread(void *arg)
>  		pr_alert("Total time taken by all kfree'ers: %llu ns, loops: %d, batches: %ld, memory footprint: %lldMB\n",
>  		       (unsigned long long)(end_time - start_time), kfree_loops,
>  		       rcuscale_seq_diff(b_rcu_gp_test_finished, b_rcu_gp_test_started),
> -		       (mem_begin - mem_during) >> (20 - PAGE_SHIFT));
> +		       PAGES_TO_MB(mem_begin - mem_during));
>
>  		if (shutdown) {
>  			smp_mb(); /* Assign before wake. */
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index b9b4bbbf0af6..ae1d9a7ef202 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -1489,7 +1489,7 @@ static unsigned int task_nr_scan_windows(struct task_struct *p)
>  	 * by the PTE scanner and NUMA hinting faults should be trapped based
>  	 * on resident pages
>  	 */
> -	nr_scan_pages = sysctl_numa_balancing_scan_size << (20 - PAGE_SHIFT);
> +	nr_scan_pages = MB_TO_PAGES(sysctl_numa_balancing_scan_size);
>  	rss = get_mm_rss(p->mm);
>  	if (!rss)
>  		rss = nr_scan_pages;
> @@ -1926,8 +1926,7 @@ bool should_numa_migrate_memory(struct task_struct *p, struct folio *folio,
>  		}
>
>  		def_th = sysctl_numa_balancing_hot_threshold;
> -		rate_limit = sysctl_numa_balancing_promote_rate_limit << \
> -			(20 - PAGE_SHIFT);
> +		rate_limit = MB_TO_PAGES(sysctl_numa_balancing_promote_rate_limit);
>  		numa_promotion_adjust_threshold(pgdat, rate_limit, def_th);
>
>  		th = pgdat->nbp_threshold ? : def_th;
> diff --git a/mm/backing-dev.c b/mm/backing-dev.c
> index 783904d8c5ef..e4d578e6121c 100644
> --- a/mm/backing-dev.c
> +++ b/mm/backing-dev.c
> @@ -510,7 +510,7 @@ static void wb_update_bandwidth_workfn(struct work_struct *work)
>  /*
>   * Initial write bandwidth: 100 MB/s
>   */
> -#define INIT_BW		(100 << (20 - PAGE_SHIFT))
> +#define INIT_BW		MB_TO_PAGES(100)
>
>  static int wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi,
>  		   gfp_t gfp)
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 389620c65a5f..dcc33d9c300f 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -911,7 +911,7 @@ static int __init hugepage_init(void)
>  	 * where the extra memory used could hurt more than TLB overhead
>  	 * is likely to save.  The admin can still enable it through /sys.
>  	 */
> -	if (totalram_pages() < (512 << (20 - PAGE_SHIFT))) {
> +	if (totalram_pages() < MB_TO_PAGES(512)) {
>  		transparent_hugepage_flags = 0;
>  		return 0;
>  	}
> diff --git a/mm/swap.c b/mm/swap.c
> index 3632dd061beb..cb164f9ef9e3 100644
> --- a/mm/swap.c
> +++ b/mm/swap.c
> @@ -1096,7 +1096,7 @@ static const struct ctl_table swap_sysctl_table[] = {
>   */
>  void __init swap_setup(void)
>  {
> -	unsigned long megs = totalram_pages() >> (20 - PAGE_SHIFT);
> +	unsigned long megs = PAGES_TO_MB(totalram_pages());
>
>  	/* Use a smaller cluster for small-memory machines */
>  	if (megs < 16)
> --
> 2.43.0
>
Re: [PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by Matthew Wilcox 2 months, 2 weeks ago
On Fri, Jul 18, 2025 at 10:57:36AM +0100, Lorenzo Stoakes wrote:
> NOte I see arch/x86/include/asm/pgtable.h has a pages_to_mb() static inline
> declaration, but probably being an asm include can't ref mm.h so meh not a big
> deal.

Should probably go to linux/sizes.h, except that it uses PAGE_SIZE which
isn't available there.  But asm-generic/getorder.h might be a good place
for it.

(hm, is including getorder.h safe by itself?  looks like it relies on
something else to bring in the definition of PAGE_SHIFT)
Re: [PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by Dev Jain 2 months, 2 weeks ago
On 18/07/25 8:11 am, Ye Liu wrote:
> From: Ye Liu <liuye@kylinos.cn>
>
> Replace repeated (20 - PAGE_SHIFT) calculations with standard macros:
> - MB_TO_PAGES(mb)    converts MB to page count
> - PAGES_TO_MB(pages) converts pages to MB
>
> No functional change.
>
> Signed-off-by: Ye Liu <liuye@kylinos.cn>
> ---
>   

sh and x86 have their own pages_to_mb, drivers/target/target_core_user.c too.
I guess no one likes to clean the kernel :)
Re: [PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by Lorenzo Stoakes 2 months, 2 weeks ago
On Fri, Jul 18, 2025 at 03:16:55PM +0530, Dev Jain wrote:
>
> On 18/07/25 8:11 am, Ye Liu wrote:
> > From: Ye Liu <liuye@kylinos.cn>
> >
> > Replace repeated (20 - PAGE_SHIFT) calculations with standard macros:
> > - MB_TO_PAGES(mb)    converts MB to page count
> > - PAGES_TO_MB(pages) converts pages to MB
> >
> > No functional change.
> >
> > Signed-off-by: Ye Liu <liuye@kylinos.cn>
> > ---
>
> sh and x86 have their own pages_to_mb, drivers/target/target_core_user.c too.
> I guess no one likes to clean the kernel :)
>

*A wild Lorenzo appaers*
Re: [PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by Dev Jain 2 months, 2 weeks ago
On 18/07/25 8:11 am, Ye Liu wrote:
> From: Ye Liu <liuye@kylinos.cn>
>
> Replace repeated (20 - PAGE_SHIFT) calculations with standard macros:
> - MB_TO_PAGES(mb)    converts MB to page count
> - PAGES_TO_MB(pages) converts pages to MB
>
> No functional change.
>
> Signed-off-by: Ye Liu <liuye@kylinos.cn>
> ---
>   

Reviewed-by: Dev Jain <dev.jain@arm.com>
Re: [PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by David Hildenbrand 2 months, 2 weeks ago
On 18.07.25 04:41, Ye Liu wrote:
> From: Ye Liu <liuye@kylinos.cn>
> 
> Replace repeated (20 - PAGE_SHIFT) calculations with standard macros:
> - MB_TO_PAGES(mb)    converts MB to page count
> - PAGES_TO_MB(pages) converts pages to MB
> 
> No functional change.
> 
> Signed-off-by: Ye Liu <liuye@kylinos.cn>
> ---

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers,

David / dhildenb
Re: [PATCH] mm: Replace (20 - PAGE_SHIFT) with common macros for pages<->MB conversion
Posted by Andrew Morton 2 months, 2 weeks ago
On Fri, 18 Jul 2025 10:41:32 +0800 Ye Liu <ye.liu@linux.dev> wrote:

> From: Ye Liu <liuye@kylinos.cn>
> 
> Replace repeated (20 - PAGE_SHIFT) calculations with standard macros:
> - MB_TO_PAGES(mb)    converts MB to page count
> - PAGES_TO_MB(pages) converts pages to MB
> 
> No functional change.
> 
> ...
>
> +/*
> + * Convert between pages and MB
> + * 20 is the shift for 1MB (2^20 = 1MB)
> + * PAGE_SHIFT is the shift for page size (e.g., 12 for 4KB pages)
> + * So (20 - PAGE_SHIFT) converts between pages and MB
> + */
> +#define PAGES_TO_MB(pages) ((pages) >> (20 - PAGE_SHIFT))
> +#define MB_TO_PAGES(mb)    ((mb) << (20 - PAGE_SHIFT))
> +
>  #ifdef CONFIG_SYSCTL
>  extern int sysctl_legacy_va_layout;
>  #else
>
> ...
>
> @@ -796,7 +796,7 @@ kfree_scale_thread(void *arg)
>  		pr_alert("Total time taken by all kfree'ers: %llu ns, loops: %d, batches: %ld, memory footprint: %lldMB\n",
>  		       (unsigned long long)(end_time - start_time), kfree_loops,
>  		       rcuscale_seq_diff(b_rcu_gp_test_finished, b_rcu_gp_test_started),
> -		       (mem_begin - mem_during) >> (20 - PAGE_SHIFT));
> +		       PAGES_TO_MB(mem_begin - mem_during));
>  
>  		if (shutdown) {
>  			smp_mb(); /* Assign before wake. */

But, but, but, obscure hard-coded magic numbers are there for our job
security!

Oh well, we got caught.  Applied, thanks.