[ add Lukas ] Linus Torvalds wrote: > On Mon, 12 Feb 2024 at 14:04, Dan Williams <dan.j.williams@xxxxxxxxx> wrote: > > > > This works because the internals of virtio_fs_cleanup_dax(), "kill_dax() > > and put_dax()", know how to handle a NULL @dax_dev. It is still early > > days with the "cleanup" helpers, but I wonder if anyone else cares that > > the DEFINE_FREE() above does not check for NULL? > > Well, the main reason for DEFINE_FREE() to check for NULL is not > correctness, but code generation. See the comment about kfree() in > <linux/cleanup.h>: > > * NOTE: the DEFINE_FREE()'s @free expression includes a NULL test even though > * kfree() is fine to be called with a NULL value. This is on purpose. This way > * the compiler sees the end of our alloc_obj() function as [...] > > with the full explanation there. > > Now, whether the code wants to actually use the cleanup() helpers for > a single use-case is debatable. > > But yes, if it does, I suspect it should use !IS_ERR_OR_NULL(ptr).o I am trying to arrive at a common recommendation given Lukas found that IS_ERR_OR_NULL() resulted in unwanted NULL checks emitted in the assembly [1]. He is doing something similar: http://lore.kernel.org/r/4143b15418c4ecf87ddeceb36813943c3ede17aa.1707734526.git.lukas@xxxxxxxxx ...and introduced an assume() helper. However, Lukas, I think Linus is right, your DEFINE_FREE() should use IS_ERR_OR_NULL(). I.e. the problem is trying to use __free(x509_free_certificate) in x509_cert_parse(). > --- a/crypto/asymmetric_keys/x509_cert_parser.c > +++ b/crypto/asymmetric_keys/x509_cert_parser.c > @@ -60,24 +60,24 @@ void x509_free_certificate(struct x509_certificate *cert) > */ > struct x509_certificate *x509_cert_parse(const void *data, size_t datalen) > { > - struct x509_certificate *cert; > - struct x509_parse_context *ctx; > + struct x509_certificate *cert __free(x509_free_certificate); ...make this: struct x509_certificate *cert __free(kfree); ...and Mathieu, this should be IS_ERR_OR_NULL() to skip an unnecessary call to virtio_fs_cleanup_dax() at function exit that the compiler should elide.
