Before invoke vmap, we need offer a pages pointer array which each page
need to map in vmalloc area.

But currently vmap_udmabuf only set each folio's head page into pages,
missed each offset pages when iter.

This patch set the correctly offset page in each folio into array.

Signed-off-by: Huan Yang <link@vivo.com>
Fixes: 5e72b2b41a21 ("udmabuf: convert udmabuf driver to use folios")
---
 drivers/dma-buf/udmabuf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/dma-buf/udmabuf.c b/drivers/dma-buf/udmabuf.c
index 79845565089d..af5200e360a6 100644
--- a/drivers/dma-buf/udmabuf.c
+++ b/drivers/dma-buf/udmabuf.c
@@ -120,7 +120,8 @@ static int vmap_udmabuf(struct dma_buf *buf, struct iosys_map *map)
 		return -ENOMEM;
 
 	for (pg = 0; pg < ubuf->pagecount; pg++)
-		pages[pg] = &ubuf->folios[pg]->page;
+		pages[pg] = folio_page(ubuf->folios[pg],
+				       ubuf->offsets[pg] >> PAGE_SHIFT);
 
 	vaddr = vm_map_ram(pages, ubuf->pagecount, -1);
 	kvfree(pages);
-- 
2.48.1

Hi Huan,

> Subject: [PATCH 2/2] udmabuf: fix vmap missed offset page
> 
> Before invoke vmap, we need offer a pages pointer array which each page
> need to map in vmalloc area.
> 
> But currently vmap_udmabuf only set each folio's head page into pages,
> missed each offset pages when iter.
> 
> This patch set the correctly offset page in each folio into array.
> 
> Signed-off-by: Huan Yang <link@vivo.com>
> Fixes: 5e72b2b41a21 ("udmabuf: convert udmabuf driver to use folios")
> ---
>  drivers/dma-buf/udmabuf.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/dma-buf/udmabuf.c b/drivers/dma-buf/udmabuf.c
> index 79845565089d..af5200e360a6 100644
> --- a/drivers/dma-buf/udmabuf.c
> +++ b/drivers/dma-buf/udmabuf.c
> @@ -120,7 +120,8 @@ static int vmap_udmabuf(struct dma_buf *buf, struct
> iosys_map *map)
>  		return -ENOMEM;
> 
>  	for (pg = 0; pg < ubuf->pagecount; pg++)
> -		pages[pg] = &ubuf->folios[pg]->page;
> +		pages[pg] = folio_page(ubuf->folios[pg],
> +				       ubuf->offsets[pg] >> PAGE_SHIFT);
IIUC, it does not look like vm_map_ram() or the other functions it calls would
write to these tail page pointers (struct page*), which should be safe even
when HVO is enabled (based on your conversations with Muchun). However,
I am wondering whether Bingbu can test this out with HVO enabled?

Regardless,
Acked-by: Vivek Kasireddy <vivek.kasireddy@intel.com>

Thanks,
Vivek

> 
>  	vaddr = vm_map_ram(pages, ubuf->pagecount, -1);
>  	kvfree(pages);
> --
> 2.48.1

On 4/22/25 1:22 PM, Kasireddy, Vivek wrote:
> Hi Huan,
> 
>> Subject: [PATCH 2/2] udmabuf: fix vmap missed offset page
>>
>> Before invoke vmap, we need offer a pages pointer array which each page
>> need to map in vmalloc area.
>>
>> But currently vmap_udmabuf only set each folio's head page into pages,
>> missed each offset pages when iter.
>>
>> This patch set the correctly offset page in each folio into array.
>>
>> Signed-off-by: Huan Yang <link@vivo.com>
>> Fixes: 5e72b2b41a21 ("udmabuf: convert udmabuf driver to use folios")
>> ---
>>  drivers/dma-buf/udmabuf.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/dma-buf/udmabuf.c b/drivers/dma-buf/udmabuf.c
>> index 79845565089d..af5200e360a6 100644
>> --- a/drivers/dma-buf/udmabuf.c
>> +++ b/drivers/dma-buf/udmabuf.c
>> @@ -120,7 +120,8 @@ static int vmap_udmabuf(struct dma_buf *buf, struct
>> iosys_map *map)
>>  		return -ENOMEM;
>>
>>  	for (pg = 0; pg < ubuf->pagecount; pg++)
>> -		pages[pg] = &ubuf->folios[pg]->page;
>> +		pages[pg] = folio_page(ubuf->folios[pg],
>> +				       ubuf->offsets[pg] >> PAGE_SHIFT);
> IIUC, it does not look like vm_map_ram() or the other functions it calls would
> write to these tail page pointers (struct page*), which should be safe even
> when HVO is enabled (based on your conversations with Muchun). However,
> I am wondering whether Bingbu can test this out with HVO enabled?

Sorry, I cannot test HVO enabled case. I was running my case with local
revert patch. :)

> 
> Regardless,
> Acked-by: Vivek Kasireddy <vivek.kasireddy@intel.com>
> 
> Thanks,
> Vivek
> 
>>
>>  	vaddr = vm_map_ram(pages, ubuf->pagecount, -1);
>>  	kvfree(pages);
>> --
>> 2.48.1
> 

-- 
Best regards,
Bingbu Cao

Hi Vivek

在 2025/4/22 13:22, Kasireddy, Vivek 写道:
> Hi Huan,
>
>> Subject: [PATCH 2/2] udmabuf: fix vmap missed offset page
>>
>> Before invoke vmap, we need offer a pages pointer array which each page
>> need to map in vmalloc area.
>>
>> But currently vmap_udmabuf only set each folio's head page into pages,
>> missed each offset pages when iter.
>>
>> This patch set the correctly offset page in each folio into array.
>>
>> Signed-off-by: Huan Yang <link@vivo.com>
>> Fixes: 5e72b2b41a21 ("udmabuf: convert udmabuf driver to use folios")
>> ---
>>   drivers/dma-buf/udmabuf.c | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/dma-buf/udmabuf.c b/drivers/dma-buf/udmabuf.c
>> index 79845565089d..af5200e360a6 100644
>> --- a/drivers/dma-buf/udmabuf.c
>> +++ b/drivers/dma-buf/udmabuf.c
>> @@ -120,7 +120,8 @@ static int vmap_udmabuf(struct dma_buf *buf, struct
>> iosys_map *map)
>>   		return -ENOMEM;
>>
>>   	for (pg = 0; pg < ubuf->pagecount; pg++)
>> -		pages[pg] = &ubuf->folios[pg]->page;
>> +		pages[pg] = folio_page(ubuf->folios[pg],
>> +				       ubuf->offsets[pg] >> PAGE_SHIFT);
> IIUC, it does not look like vm_map_ram() or the other functions it calls would
> write to these tail page pointers (struct page*), which should be safe even
> when HVO is enabled (based on your conversations with Muchun). However,

Yes, and need point, each write to tail page's va pointer is no permit.(HVO changed this va in vmemmap's prot

into RO), so, easy to observe. And  I see each vmap-api, only turn to read it, no write. :)

Thanks,

Huan

> I am wondering whether Bingbu can test this out with HVO enabled?
>
> Regardless,
> Acked-by: Vivek Kasireddy <vivek.kasireddy@intel.com>
>
> Thanks,
> Vivek
>
>>   	vaddr = vm_map_ram(pages, ubuf->pagecount, -1);
>>   	kvfree(pages);
>> --
>> 2.48.1