[v5] cifs: Fix copy_to_iter return value check

[PATCH v5] cifs: Fix copy_to_iter return value check

Posted by Fushuai Wang 2 months, 1 week ago

The return value of copy_to_iter() function will never be negative,
it is the number of bytes copied, or zero if nothing was copied.
Update the check to treat 0 as an error, and return -1 in that case.

Fixes: d08089f649a0 ("cifs: Change the I/O paths to use an iterator rather than a page list")
Signed-off-by: Fushuai Wang <wangfushuai@baidu.com>
---
v5: no code changes, only improve commit format
v4: no code changes, only add version description
v3: use size_t type for (copied) and check for (copied == 0) as error.
v2: use (!length) check for error condition.
v1: use (length <= 0) check for error condition.

 fs/smb/client/smb2ops.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/fs/smb/client/smb2ops.c b/fs/smb/client/smb2ops.c
index 058050f744c0..ac8a5bd6aec4 100644
--- a/fs/smb/client/smb2ops.c
+++ b/fs/smb/client/smb2ops.c
@@ -4650,7 +4650,7 @@ handle_read_data(struct TCP_Server_Info *server, struct mid_q_entry *mid,
 	unsigned int pad_len;
 	struct cifs_io_subrequest *rdata = mid->callback_data;
 	struct smb2_hdr *shdr = (struct smb2_hdr *)buf;
-	int length;
+	size_t copied;
 	bool use_rdma_mr = false;
 
 	if (shdr->Command != SMB2_READ) {
@@ -4763,10 +4763,10 @@ handle_read_data(struct TCP_Server_Info *server, struct mid_q_entry *mid,
 	} else if (buf_len >= data_offset + data_len) {
 		/* read response payload is in buf */
 		WARN_ONCE(buffer, "read data can be either in buf or in buffer");
-		length = copy_to_iter(buf + data_offset, data_len, &rdata->subreq.io_iter);
-		if (length < 0)
-			return length;
-		rdata->got_bytes = data_len;
+		copied = copy_to_iter(buf + data_offset, data_len, &rdata->subreq.io_iter);
+		if (copied == 0)
+			return -1;
+		rdata->got_bytes = copied;
 	} else {
 		/* read response payload cannot be in both buf and pages */
 		WARN_ONCE(1, "buf can not contain only a part of read data");
-- 
2.36.1

Re: [PATCH v5] cifs: Fix copy_to_iter return value check

Posted by Steve French 2 months, 1 week ago

David suggested fixing this line:

 +                       return -1;

to

 +                       return -EIO;

which makes sense to me since returning -1 could be confusing.

On Tue, Oct 7, 2025 at 3:27 AM Fushuai Wang <wangfushuai@baidu.com> wrote:
>
> The return value of copy_to_iter() function will never be negative,
> it is the number of bytes copied, or zero if nothing was copied.
> Update the check to treat 0 as an error, and return -1 in that case.
>
> Fixes: d08089f649a0 ("cifs: Change the I/O paths to use an iterator rather than a page list")
> Signed-off-by: Fushuai Wang <wangfushuai@baidu.com>
> ---
> v5: no code changes, only improve commit format
> v4: no code changes, only add version description
> v3: use size_t type for (copied) and check for (copied == 0) as error.
> v2: use (!length) check for error condition.
> v1: use (length <= 0) check for error condition.
>
>  fs/smb/client/smb2ops.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/fs/smb/client/smb2ops.c b/fs/smb/client/smb2ops.c
> index 058050f744c0..ac8a5bd6aec4 100644
> --- a/fs/smb/client/smb2ops.c
> +++ b/fs/smb/client/smb2ops.c
> @@ -4650,7 +4650,7 @@ handle_read_data(struct TCP_Server_Info *server, struct mid_q_entry *mid,
>         unsigned int pad_len;
>         struct cifs_io_subrequest *rdata = mid->callback_data;
>         struct smb2_hdr *shdr = (struct smb2_hdr *)buf;
> -       int length;
> +       size_t copied;
>         bool use_rdma_mr = false;
>
>         if (shdr->Command != SMB2_READ) {
> @@ -4763,10 +4763,10 @@ handle_read_data(struct TCP_Server_Info *server, struct mid_q_entry *mid,
>         } else if (buf_len >= data_offset + data_len) {
>                 /* read response payload is in buf */
>                 WARN_ONCE(buffer, "read data can be either in buf or in buffer");
> -               length = copy_to_iter(buf + data_offset, data_len, &rdata->subreq.io_iter);
> -               if (length < 0)
> -                       return length;
> -               rdata->got_bytes = data_len;
> +               copied = copy_to_iter(buf + data_offset, data_len, &rdata->subreq.io_iter);
> +               if (copied == 0)
> +                       return -1;
> +               rdata->got_bytes = copied;
>         } else {
>                 /* read response payload cannot be in both buf and pages */
>                 WARN_ONCE(1, "buf can not contain only a part of read data");

> --
> 2.36.1
>
>


--
Thanks,

Steve

Re: [PATCH v5] cifs: Fix copy_to_iter return value check

Posted by Tom Talpey 2 months, 1 week ago

I like it. For v5 feel free to add my Reviewed-by and ack.

Tom.

Oct 7, 2025 2:45:33 PM Steve French <smfrench@gmail.com>:

> David suggested fixing this line:
> 
> +                       return -1;
> 
> to
> 
> +                       return -EIO;
> 
> which makes sense to me since returning -1 could be confusing.
> 
> On Tue, Oct 7, 2025 at 3:27 AM Fushuai Wang <wangfushuai@baidu.com> wrote:
>> 
>> The return value of copy_to_iter() function will never be negative,
>> it is the number of bytes copied, or zero if nothing was copied.
>> Update the check to treat 0 as an error, and return -1 in that case.
>> 
>> Fixes: d08089f649a0 ("cifs: Change the I/O paths to use an iterator rather than a page list")
>> Signed-off-by: Fushuai Wang <wangfushuai@baidu.com>
>> ---
>> v5: no code changes, only improve commit format
>> v4: no code changes, only add version description
>> v3: use size_t type for (copied) and check for (copied == 0) as error.
>> v2: use (!length) check for error condition.
>> v1: use (length <= 0) check for error condition.
>> 
>> fs/smb/client/smb2ops.c | 10 +++++-----
>> 1 file changed, 5 insertions(+), 5 deletions(-)
>> 
>> diff --git a/fs/smb/client/smb2ops.c b/fs/smb/client/smb2ops.c
>> index 058050f744c0..ac8a5bd6aec4 100644
>> --- a/fs/smb/client/smb2ops.c
>> +++ b/fs/smb/client/smb2ops.c
>> @@ -4650,7 +4650,7 @@ handle_read_data(struct TCP_Server_Info *server, struct mid_q_entry *mid,
>>         unsigned int pad_len;
>>         struct cifs_io_subrequest *rdata = mid->callback_data;
>>         struct smb2_hdr *shdr = (struct smb2_hdr *)buf;
>> -       int length;
>> +       size_t copied;
>>         bool use_rdma_mr = false;
>> 
>>         if (shdr->Command != SMB2_READ) {
>> @@ -4763,10 +4763,10 @@ handle_read_data(struct TCP_Server_Info *server, struct mid_q_entry *mid,
>>         } else if (buf_len >= data_offset + data_len) {
>>                 /* read response payload is in buf */
>>                 WARN_ONCE(buffer, "read data can be either in buf or in buffer");
>> -               length = copy_to_iter(buf + data_offset, data_len, &rdata->subreq.io_iter);
>> -               if (length < 0)
>> -                       return length;
>> -               rdata->got_bytes = data_len;
>> +               copied = copy_to_iter(buf + data_offset, data_len, &rdata->subreq.io_iter);
>> +               if (copied == 0)
>> +                       return -1;
>> +               rdata->got_bytes = copied;
>>         } else {
>>                 /* read response payload cannot be in both buf and pages */
>>                 WARN_ONCE(1, "buf can not contain only a part of read data");
> 
>> -- 
>> 2.36.1
>> 
>> 
> 
> 
> -- 
> Thanks,
> 
> Steve