[PATCH] tcmu: Fix possible to/from address overflow when doing the memcpy

[PATCH] tcmu: Fix possible to/from address overflow when doing the memcpy

[lixiubo]

From: Xiubo Li <lixiubo@cmss.chinamobile.com>

For most case the sg->length equals to PAGE_SIZE, so this bug won't
be triggered. Otherwise this will crash the kernel, for example when
all segments' sg->length equal to 1K.

Signed-off-by: Xiubo Li <lixiubo@cmss.chinamobile.com>
---
 drivers/target/target_core_user.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/drivers/target/target_core_user.c b/drivers/target/target_core_user.c
index 8bf0823..9030c2a 100644
--- a/drivers/target/target_core_user.c
+++ b/drivers/target/target_core_user.c
@@ -590,8 +590,6 @@ static int scatter_data_area(struct tcmu_dev *udev,
 					block_remaining);
 			to_offset = get_block_offset_user(udev, dbi,
 					block_remaining);
-			offset = DATA_BLOCK_SIZE - block_remaining;
-			to += offset;
 
 			if (*iov_cnt != 0 &&
 			    to_offset == iov_tail(*iov)) {
@@ -602,8 +600,10 @@ static int scatter_data_area(struct tcmu_dev *udev,
 				(*iov)->iov_len = copy_bytes;
 			}
 			if (copy_data) {
-				memcpy(to, from + sg->length - sg_remaining,
-					copy_bytes);
+				offset = DATA_BLOCK_SIZE - block_remaining;
+				memcpy(to + offset,
+				       from + sg->length - sg_remaining,
+				       copy_bytes);
 				tcmu_flush_dcache_range(to, copy_bytes);
 			}
 			sg_remaining -= copy_bytes;
@@ -664,9 +664,8 @@ static void gather_data_area(struct tcmu_dev *udev, struct tcmu_cmd *cmd,
 			copy_bytes = min_t(size_t, sg_remaining,
 					block_remaining);
 			offset = DATA_BLOCK_SIZE - block_remaining;
-			from += offset;
 			tcmu_flush_dcache_range(from, copy_bytes);
-			memcpy(to + sg->length - sg_remaining, from,
+			memcpy(to + sg->length - sg_remaining, from + offset,
 					copy_bytes);
 
 			sg_remaining -= copy_bytes;
-- 
1.8.3.1

Re: [PATCH] tcmu: Fix possible to/from address overflow when doing the memcpy

[Mike Christie]

On 07/12/2017 02:51 AM, lixiubo@cmss.chinamobile.com wrote:
> From: Xiubo Li <lixiubo@cmss.chinamobile.com>
> 
> For most case the sg->length equals to PAGE_SIZE, so this bug won't
> be triggered. Otherwise this will crash the kernel, for example when
> all segments' sg->length equal to 1K.
> 
> Signed-off-by: Xiubo Li <lixiubo@cmss.chinamobile.com>
> ---
>  drivers/target/target_core_user.c | 11 +++++------
>  1 file changed, 5 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/target/target_core_user.c b/drivers/target/target_core_user.c
> index 8bf0823..9030c2a 100644
> --- a/drivers/target/target_core_user.c
> +++ b/drivers/target/target_core_user.c
> @@ -590,8 +590,6 @@ static int scatter_data_area(struct tcmu_dev *udev,
>  					block_remaining);
>  			to_offset = get_block_offset_user(udev, dbi,
>  					block_remaining);
> -			offset = DATA_BLOCK_SIZE - block_remaining;
> -			to += offset;
>  
>  			if (*iov_cnt != 0 &&
>  			    to_offset == iov_tail(*iov)) {
> @@ -602,8 +600,10 @@ static int scatter_data_area(struct tcmu_dev *udev,
>  				(*iov)->iov_len = copy_bytes;
>  			}
>  			if (copy_data) {
> -				memcpy(to, from + sg->length - sg_remaining,
> -					copy_bytes);
> +				offset = DATA_BLOCK_SIZE - block_remaining;
> +				memcpy(to + offset,
> +				       from + sg->length - sg_remaining,
> +				       copy_bytes);
>  				tcmu_flush_dcache_range(to, copy_bytes);
>  			}
>  			sg_remaining -= copy_bytes;
> @@ -664,9 +664,8 @@ static void gather_data_area(struct tcmu_dev *udev, struct tcmu_cmd *cmd,
>  			copy_bytes = min_t(size_t, sg_remaining,
>  					block_remaining);
>  			offset = DATA_BLOCK_SIZE - block_remaining;
> -			from += offset;
>  			tcmu_flush_dcache_range(from, copy_bytes);
> -			memcpy(to + sg->length - sg_remaining, from,
> +			memcpy(to + sg->length - sg_remaining, from + offset,
>  					copy_bytes);
>  
>  			sg_remaining -= copy_bytes;
> 

Nice.

Reviewed-by: Mike Christie <mchristi@redhat.com>

Re: [PATCH] tcmu: Fix possible to/from address overflow when doing the memcpy

[Nicholas A. Bellinger]

On Wed, 2017-07-12 at 15:51 +0800, lixiubo@cmss.chinamobile.com wrote:
> From: Xiubo Li <lixiubo@cmss.chinamobile.com>
> 
> For most case the sg->length equals to PAGE_SIZE, so this bug won't
> be triggered. Otherwise this will crash the kernel, for example when
> all segments' sg->length equal to 1K.
> 
> Signed-off-by: Xiubo Li <lixiubo@cmss.chinamobile.com>
> ---
>  drivers/target/target_core_user.c | 11 +++++------
>  1 file changed, 5 insertions(+), 6 deletions(-)
> 

Applied to target-pending/master.

Thankx Xiubo + MNC.