Re: [PATCH 1/7] Add a new flags-accepting interface for anonymous inodes

Thanks for taking a look

On Mon, Oct 14, 2019 at 8:39 AM Jann Horn [off-list ref] wrote:

On Sat, Oct 12, 2019 at 9:16 PM Daniel Colascione [off-list ref] wrote:

quoted

Add functions forwarding from the old names to the new ones so we
don't need to change any callers.

This patch does more than the commit message says; it also refactors
the body of the function. (I would've moved that refactoring over into
patch 2, but I guess this works, too.)

[...]

quoted

-struct file *anon_inode_getfile(const char *name,
-                               const struct file_operations *fops,
-                               void *priv, int flags)
+struct file *anon_inode_getfile2(const char *name,
+                                const struct file_operations *fops,
+                                void *priv, int flags, int anon_inode_flags)

(AFAIK, normal kernel style is to slap a "__" prefix in front of the
function name instead of appending a digit, but I guess it doesn't
really matter.)

I thought prefixing "_" was for signaling "this is an implementation
detail and you probably don't want to call it unless you know what
you're doing", not "here's a new version that does a new thing".

quoted

 {
+       struct inode *inode;
        struct file *file;

-       if (IS_ERR(anon_inode_inode))
-               return ERR_PTR(-ENODEV);
-
-       if (fops->owner && !try_module_get(fops->owner))
-               return ERR_PTR(-ENOENT);
+       if (anon_inode_flags)
+               return ERR_PTR(-EINVAL);

+       inode = anon_inode_inode;
+       if (IS_ERR(inode))
+               return ERR_PTR(-ENODEV);
        /*
-        * We know the anon_inode inode count is always greater than zero,
-        * so ihold() is safe.
+        * We know the anon_inode inode count is always
+        * greater than zero, so ihold() is safe.
         */

This looks like maybe you started editing the comment, then un-did the
change, but left the modified line wrapping in your patch? Please
avoid that - code changes with no real reason make "git blame" output
more annoying and create trouble when porting patches between kernel
versions.

I'll fix it.

[...]

quoted

 EXPORT_SYMBOL_GPL(anon_inode_getfile);
+EXPORT_SYMBOL_GPL(anon_inode_getfile2);

[...]

quoted

 EXPORT_SYMBOL_GPL(anon_inode_getfd);
+EXPORT_SYMBOL_GPL(anon_inode_getfd2);

Since anon_inode_getfd() is now a static inline function in
include/linux/anon_inodes.h, exporting it doesn't make sense anymore.
Same for anon_inode_getfile().

I didn't want to break modules unnecessarily. Declaring the function
inline and also exporting it gives us an efficiency win while avoiding
an ABI break, right?

[...]

quoted

+#define ANON_INODE_SECURE 1

That #define belongs in a later patch, right?

Yep.