Skip to contentSkip to conversation

fix memcpy_to_page and memzero_page

PATCH Applied Christoph Hellwig 3 participants 6 messages Jul 13, 2021 – Jul 14, 2021

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi all,

we need to call flush_dcache_page on any kmapped page, and on any page we write to that may be mapped into the page cache, which can happen quite frequently for anything we use raw pages for.

Patch 1 make sure memcpy_to_page and memzero_page to do the right thing for that, and patch 2 fixes up the lost switch to kmap_local_page for memzero_page.

#1 lore raw

PATCH 1/2 mm: call flush_dcache_page in memcpy_to_page and memzero_page

memcpy_to_page and memzero_page can write to arbitrary pages, which could be in the page cache or in high memory, so call flush_kernel_dcache_pages to flush the dcache.

Fixes bb90d4bc7b6a ("mm/highmem: Lift memcpy_[to|from]_page to core")
Fixes 28961998f858 ("iov_iter: lift memzero_page() to highmem.h")
---
 include/linux/highmem.h | 2 ++
 1 file changed, 2 insertions(+)
include/linux/highmem.h +2 -0
index 8c6e8e996c87..8e7e50a53a12 100644
@@ -318,6 +318,7 @@ static inline void memcpy_to_page(struct page *page, size_t offset,
 
 	VM_BUG_ON(offset + len > PAGE_SIZE);
 	memcpy(to + offset, from, len);
+	flush_dcache_page(page);
 	kunmap_local(to);
 }
 
@@ -325,6 +326,7 @@ static inline void memzero_page(struct page *page, size_t offset, size_t len)
 {
 	char *addr = kmap_atomic(page);
 	memset(addr + offset, 0, len);
+	flush_dcache_page(page);
 	kunmap_atomic(addr);
 }
 
Signature · 1 line
2.30.2
maintainer
#2 lore raw

Re: [PATCH 1/2] mm: call flush_dcache_page in memcpy_to_page and memzero_page

Quoted 3 lines · On Tue, 13 Jul 2021 07:52:30 +0200 Christoph Hellwig <[email protected]> wrote:

memcpy_to_page and memzero_page can write to arbitrary pages, which could be in the page cache or in high memory, so call flush_kernel_dcache_pages to flush the dcache.

I assume this presently is not known to cause any problems, but that some problems might be discovered in the future? In which case, should we cc:stable?

#3 lore raw

Re: [PATCH 1/2] mm: call flush_dcache_page in memcpy_to_page and memzero_page

Quoted 9 lines · Andrew Morton
Quoted 3 lines · On Tue, 13 Jul 2021 07:52:30 +0200 Christoph Hellwig <[email protected]> wrote:

memcpy_to_page and memzero_page can write to arbitrary pages, which could be in the page cache or in high memory, so call flush_kernel_dcache_pages to flush the dcache.

I assume this presently is not known to cause any problems, but that some problems might be discovered in the future? In which case, should we cc:stable?

This is a problem when using these helpers on dcache challeneged architectures. Right now there are just a few users, chances are no one used the PC floppy drіver, the aha1542 driver for an ISA SCSI HBA, and a few advanced and optional btrfs and ext4 features on those platforms yet since the conversion. So yes, stable would be good.

Reviewed-by on v1 1/2
#4 lore raw

Re: [PATCH 1/2] mm: call flush_dcache_page in memcpy_to_page and memzero_page

In reply to Christoph Hellwig
Quoted 7 lines · Christoph Hellwig

memcpy_to_page and memzero_page can write to arbitrary pages, which could be in the page cache or in high memory, so call flush_kernel_dcache_pages to flush the dcache.

Fixes bb90d4bc7b6a ("mm/highmem: Lift memcpy_[to|from]_page to core")
Fixes 28961998f858 ("iov_iter: lift memzero_page() to highmem.h")
Quoted 27 lines
---
 include/linux/highmem.h | 2 ++
 1 file changed, 2 insertions(+)
include/linux/highmem.h +2 -0
index 8c6e8e996c87..8e7e50a53a12 100644
@@ -318,6 +318,7 @@ static inline void memcpy_to_page(struct page *page, size_t offset,
 
 	VM_BUG_ON(offset + len > PAGE_SIZE);
 	memcpy(to + offset, from, len);
+	flush_dcache_page(page);
 	kunmap_local(to);
 }
 
@@ -325,6 +326,7 @@ static inline void memzero_page(struct page *page, size_t offset, size_t len)
 {
 	char *addr = kmap_atomic(page);
 	memset(addr + offset, 0, len);
+	flush_dcache_page(page);
 	kunmap_atomic(addr);
 }
 
Signature · 1 line
2.30.2
#5 lore raw

PATCH 2/2 mm: use kmap_local_page in memzero_page

In reply to Christoph Hellwig

The commit message introducing the global memzero_page explicitly mentions switching to kmap_local_page in the commit log but doesn't actually do that.

Fixes 28961998f858 ("iov_iter: lift memzero_page() to highmem.h")
---
 include/linux/highmem.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
include/linux/highmem.h +2 -2
index 8e7e50a53a12..d9a606a9fc64 100644
@@ -324,10 +324,10 @@ static inline void memcpy_to_page(struct page *page, size_t offset,
 
 static inline void memzero_page(struct page *page, size_t offset, size_t len)
 {
-	char *addr = kmap_atomic(page);
+	char *addr = kmap_local_page(page);
 	memset(addr + offset, 0, len);
 	flush_dcache_page(page);
-	kunmap_atomic(addr);
+	kunmap_local(addr);
 }
 
 #endif /* _LINUX_HIGHMEM_H */
Signature · 1 line
2.30.2