On Sun, Oct 08, 2017 at 12:17:49PM -0700, Eric Biggers wrote: g> > This is an improvement, but it's still broken because there is no locking. If > two processes ->open() the same inode at the same time, one can observe > ext4_has_encryption_key() as true before the key has been fully initialized. It > may be a good idea to add a really simple > DEFINE_MUTEX(ext4_key_generation_mutex), then do the ext4_has_encryption_key() > check and key generation under the mutex... Thanks for the review. You're right, we need to add a mutex because otherwise this patch would cause a regression. Previously, a racing open would "only" leak kernel memory. Now, it could cause a kernel crash. I'll respin a patch with the mutex added. At this point, looking at 4.1, I can't really recommend that _anyone_ try to use ext4 encryption with the 4.1 LTS kernel. Fortunately, I don't think there are that many users of 4.1, and those that do probably aren't using ext4 encryption. (The Android ecosystem seems to have largely settled on 3.18 and 4.4, and there hasn't been much use of ext4 encryption outside of Android yet.) I just want to make sure keep the latest upstream LTS kernels mostly clean with respect to xfstests runs, so I more easily detect regressions. Speaking of which, over the weekend I've sent a dozen or so stable kernel backport/cherry pick requests to stable@xxxxxxxxxxxxxxx. Most of them I didn't bother to cc linux-ext4 since they were minor changes to allow the patches to apply and I decided not to spam the list with the backports. I cc'ed the patch in this thread because it was new code, and I'm glad I did, since your review was very helpful. If folks are interested the backports, I can either send them all to the list, or I can publish them on git.kernel.org. Hopefully they will all show up in the LTS stable review process and then into a LTS kernel in the next month or so. Also, there are a number of the backports which are needed if you want to be able to run xfstests without tests causing older kernels (especialy 3.18 and 4.1) to crash or hang in the middle of the test run. So if anyone else is interested in running xfstests against older kernels before these patches land in the LTS kernels, let me know. (I still have one known problem which still needs to be fixed, which is that generic/269 will cause 4.4 and earlier kernels to soft lockup.) Cheers, - Ted
