ksmbd: clean up a type in ksmbd_vfs_stream_write()
Subsystems (3)
PATCH 1/1 ksmbd: clean up a type in ksmbd_vfs_stream_write()
The existing code works fine because ksmbd_vfs_getcasexattr() isn't going to return values greater than INT_MAX. But it's ugly to do the casting and using a ssize_t makes everything cleaner.
--- fs/ksmbd/vfs.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)
fs/ksmbd/vfs.c +4 -3
@@ -411,7 +411,8 @@ static int ksmbd_vfs_stream_write(struct ksmbd_file *fp, char *buf, loff_t *pos, {char *stream_buf =NULL , *wbuf;struct user_namespace *user_ns =file_mnt_user_ns (fp->filp); -size_t size, v_len; +ssize_t v_len; +size_t size;int err =0 ; ksmbd_debug(VFS, "write stream data pos : %llu, count : %zd\n",
@@ -428,9 +429,9 @@ static int ksmbd_vfs_stream_write(struct ksmbd_file *fp, char *buf, loff_t *pos, fp->stream.name, fp->stream.size, &stream_buf); -if ((int )v_len <0 ) { +if (v_len <0 ) {pr_err ("not found stream in xattr : %zd\n" , v_len); - err = (int )v_len; + err = v_len; goto out; }
Signature · 1 line
2.35.1
Namjae Jeon
maintainer2022-06-02 19:51 GMT+09:00, Dan Carpenter <[email protected]>:
Quoted 25 lines
The existing code works fine because
ksmbd_vfs_getcasexattr()isn't going to return values greater thanINT_MAX. But it's ugly to do the casting and using assize_tmakes everything cleaner.Signed-off-by Dan Carpenter <[email protected]>--- fs/ksmbd/vfs.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)fs/ksmbd/vfs.c +2
-1@@ -411,7 +411,8 @@ static int ksmbd_vfs_stream_write(struct ksmbd_file *fp, char *buf, loff_t *pos, { char *stream_buf = NULL, *wbuf;struct user_namespace *user_ns =file_mnt_user_ns (fp->filp); -size_t size, v_len; +ssize_t v_len; +size_t size;int err =0 ;fs/ksmbd/vfs.c +0
-0ksmbd_debug(VFS, "write stream data pos : %llu, count : %zd\n",
fs/ksmbd/vfs.c static int ksmbd_vfs_stream_write(struct ksmbd_file *fp, · line 435
Quoted 9 lines
fs/ksmbd/vfs.c +2
-2@@ -428,9 +429,9 @@ static int ksmbd_vfs_stream_write(struct ksmbd_file *fp, char *buf, loff_t *pos, fp->stream.name, fp->stream.size, &stream_buf); -if ((int )v_len <0 ) { +if (v_len <0 ) {pr_err ("not found stream in xattr : %zd\n" , v_len); - err = (int )v_len; + err = v_len;
Data type of ssize_t is long. Wouldn't some static checker warn us that this is a problem?
Thanks!
Quoted 7 lines
fs/ksmbd/vfs.c +0
-1goto out; } --2.35.1
Quoted 13 lines · Namjae Jeon
2022-06-02 19:51 GMT+09:00, Dan Carpenter <[email protected]>:
fs/ksmbd/vfs.c static int ksmbd_vfs_stream_write(struct ksmbd_file *fp, · line 435
Quoted 10 lines
fs/ksmbd/vfs.c +2
-2@@ -428,9 +429,9 @@ static int ksmbd_vfs_stream_write(struct ksmbd_file *fp, char *buf, loff_t *pos, fp->stream.name, fp->stream.size, &stream_buf); -if ((int )v_len <0 ) { +if (v_len <0 ) {pr_err ("not found stream in xattr : %zd\n" , v_len); - err = (int )v_len; + err = v_len;Data type of
ssize_tis long. Wouldn't some static checker warn us that this is a problem?
None that I know of.
To a human reader, the cast isn't needed because when a function returns ssize_t the negatives can only be error codes in the (-4095)-(-1) range. No other negative sizes make sense.
On the other hand, the "if ((int)v_len < 0) {" line really should trigger a static checker because there are times when sizes could be over 2GB. I will write down that I need to create that checker.
regards,
dan carpenter
Namjae Jeon
maintainer2022-06-03 9:10 GMT+09:00, Dan Carpenter <[email protected]>:
Quoted 18 lines · Namjae Jeon
Quoted 14 lines
2022-06-02 19:51 GMT+09:00, Dan Carpenter <[email protected]>:
fs/ksmbd/vfs.c static int ksmbd_vfs_stream_write(struct ksmbd_file · line 436
Quoted 11 lines
fs/ksmbd/vfs.c +2
-2@@ -428,9 +429,9 @@ static int ksmbd_vfs_stream_write(struct ksmbd_file *fp, char *buf, loff_t *pos, fp->stream.name, fp->stream.size, &stream_buf); -if ((int )v_len <0 ) { +if (v_len <0 ) {pr_err ("not found stream in xattr : %zd\n" , v_len); - err = (int )v_len; + err = v_len;Data type of
ssize_tis long. Wouldn't some static checker warn us that this is a problem?None that I know of.
Quoted 7 lines
To a human reader, the cast isn't needed because when a function returns
ssize_tthe negatives can only be error codes in the (-4095)-(-1) range. No other negative sizes make sense.On the other hand, the "if ((int)v_len < 0) {" line really should trigger a static checker because there are times when sizes could be over 2GB. I will write down that I need to create that checker.
Okay, And there is a similar type casting in ksmbd_vfs_stream_read(). Is it not necessary to change together?
Quoted 5 lines
regards,
dan carpenter
Quoted 23 lines · Namjae Jeon
2022-06-03 9:10 GMT+09:00, Dan Carpenter <[email protected]>:
Quoted 22 lines · Namjae Jeon
Quoted 14 lines
2022-06-02 19:51 GMT+09:00, Dan Carpenter <[email protected]>:
fs/ksmbd/vfs.c static int ksmbd_vfs_stream_write(struct ksmbd_file · line 436
Quoted 11 lines
fs/ksmbd/vfs.c +2
-2@@ -428,9 +429,9 @@ static int ksmbd_vfs_stream_write(struct ksmbd_file *fp, char *buf, loff_t *pos, fp->stream.name, fp->stream.size, &stream_buf); -if ((int )v_len <0 ) { +if (v_len <0 ) {pr_err ("not found stream in xattr : %zd\n" , v_len); - err = (int )v_len; + err = v_len;Data type of
ssize_tis long. Wouldn't some static checker warn us that this is a problem?None that I know of.
To a human reader, the cast isn't needed because when a function returns
ssize_tthe negatives can only be error codes in the (-4095)-(-1) range. No other negative sizes make sense.
Quoted 5 lines
Quoted 3 lines
On the other hand, the "if ((int)v_len < 0) {" line really should trigger a static checker because there are times when sizes could be over 2GB. I will write down that I need to create that checker.
Okay, And there is a similar type casting in
ksmbd_vfs_stream_read(). Is it not necessary to change together?
The motivation for this change was that I was looking for places which save negative error codes in a positive value. Since the ksmbd_vfs_stream_read() is using ssize_t then it didn't trigger that warning. But ksmbd_vfs_stream_read() has some minor issues.
if ((int)v_len <= 0)
This cast is "buggy". Harmless though.
return (int)v_len;
This is fine, but unnecessary.
if (v_len <= *pos) {
count = -EINVAL;I feel like this should return 0 bytes instead of -EINVAL; It probably doesn't matter.
I am going to try write a static checker that complains specifically about casting size_t (not long) to int. I will make it complain about ssize_t as well, but that's kind of unusual because ssize_t can already store negatives.
regards,
dan carpenter