diff mbox series

[7/9] xen/arm: take care of concurrency on static memory allocation

Message ID 20210607024318.3988467-8-penny.zheng@arm.com (mailing list archive)
State Superseded
Headers show
Series Domain on Static Allocation | expand

Commit Message

Penny Zheng June 7, 2021, 2:43 a.m. UTC 
In the future, user may want to allocate static memory at runtime,
and it is quite important to get the code protected from concurrent
access.

Re-use heap_lock to protect concurrent access in alloc_staticmem_pages.

Signed-off-by: Penny Zheng <penny.zheng@arm.com>
---
changes v2:
- new commit
---
 xen/common/page_alloc.c | 5 +++++
 1 file changed, 5 insertions(+)

Comments

Jan Beulich June 10, 2021, 10:53 a.m. UTC | #1 
On 07.06.2021 04:43, Penny Zheng wrote:
> --- a/xen/common/page_alloc.c
> +++ b/xen/common/page_alloc.c
> @@ -1087,6 +1087,9 @@ static struct page_info *alloc_staticmem_pages(unsigned long nr_mfns,
>                 nr_mfns, mfn_x(smfn));
>          return NULL;
>      }
> +
> +    spin_lock(&heap_lock);
> +
>      pg = mfn_to_page(smfn);
>  
>      for ( i = 0; i < nr_mfns; i++ )
> @@ -1127,6 +1130,8 @@ static struct page_info *alloc_staticmem_pages(unsigned long nr_mfns,
>                              !(memflags & MEMF_no_icache_flush));
>      }
>  
> +    spin_unlock(&heap_lock);
> +
>      if ( need_tlbflush )
>          filtered_flush_tlb_mask(tlbflush_timestamp);

Besides, as indicated there, the need to fold this into the previous
patch, you will also want to pay attention to alloc_heap_pages()
carefully avoiding to scrub or flush pages with the heap lock held.
You will want to follow this for your additions.

Jan
diff mbox series

Patch

diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
index a0eea5f1a4..c6ccfc3216 100644
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -1087,6 +1087,9 @@  static struct page_info *alloc_staticmem_pages(unsigned long nr_mfns,
                nr_mfns, mfn_x(smfn));
         return NULL;
     }
+
+    spin_lock(&heap_lock);
+
     pg = mfn_to_page(smfn);
 
     for ( i = 0; i < nr_mfns; i++ )
@@ -1127,6 +1130,8 @@  static struct page_info *alloc_staticmem_pages(unsigned long nr_mfns,
                             !(memflags & MEMF_no_icache_flush));
     }
 
+    spin_unlock(&heap_lock);
+
     if ( need_tlbflush )
         filtered_flush_tlb_mask(tlbflush_timestamp);