On Wed, May 27, 2020 at 06:15:10PM -0700, Bart Van Assche wrote: > On 2020-05-26 20:12, Luis Chamberlain wrote: > > + /* > > + * Blktrace needs a debugsfs name even for queues that don't register > > + * a gendisk, so it lazily registers the debugfs directory. But that > > + * can get us into a situation where a SCSI device is found, with no > > + * driver for it (yet). Then blktrace is used on the device, creating > > + * the debugfs directory, and only after that a drivers is loaded. In > ^^^^^^^ > driver? Fixed. > > @@ -494,6 +490,38 @@ static int do_blk_trace_setup(struct request_queue *q, char *name, dev_t dev, > > */ > > strreplace(buts->name, '/', '_'); > > > > + /* > > + * We also have to use a partition directory if a partition is > > + * being worked on, even though the same request_queue is shared. > > + */ > > + if (bdev && bdev != bdev->bd_contains) > > + dir = bdev->bd_part->debugfs_dir; > > Please balance braces in if-statements as required by the kernel coding style. Sure thing. > > + else { > > + /* > > + * For queues that do not have a gendisk attached to them, the > > + * debugfs directory will not have been created at setup time. > > + * Create it here lazily, it will only be removed when the > > + * queue is torn down. > > + */ > > Is the above comment perhaps a reference to blk_register_queue()? If so, please > mention the name of that function explicitly. No, it actually is in reference to *add_disk()* helpers, so I'll add that there. scsi-generic is the ugly child we have which we don't talk too much about, not sure if we have a proper name for *non* add_disk() related use of the request_queue... oh and mmc I think? I've changed this to (ignore spaces, I'll adjust): * For queues that do not have a gendisk attached to them, that is those * which do not use *add_disk*() or similar, the debugfs directory will * not have been created at setup time. This is the case for * scsi-generic drivers. Create it here lazily, it will only be removed * when the queue is torn down. > > + if (!q->debugfs_dir) { > > + q->debugfs_dir = > > + debugfs_create_dir(buts->name, > > + blk_debugfs_root); > > + } > > + dir = q->debugfs_dir; > > + } > > + > > + /* > > + * As blktrace relies on debugfs for its interface the debugfs directory > > + * is required, contrary to the usual mantra of not checking for debugfs > > + * files or directories. > > + */ > > + if (IS_ERR_OR_NULL(q->debugfs_dir)) { > > + pr_warn("debugfs_dir not present for %s so skipping\n", > > + buts->name); > > + return -ENOENT; > > + } > > How are do_blk_trace_setup() calls serialized against the debugfs directory > creation code in blk_register_queue()? Perhaps via q->blk_trace_mutex? Yes, hence the mutex lock that Christoph added as an alternative to the whole symlink stuff for scsi-generic and addressing this on the class interface driver. > Are > mutex lock and unlock calls for that mutex perhaps missing from > compat_blk_trace_setup()? No, because that is called from blk_trace_ioctl(), and that holds the mutex. > How about adding a lockdep_assert_held(&q->blk_trace_mutex) statement in > do_blk_trace_setup()? Sure, however that doesn't seem part of the fix. How about adding that as a separat patch? Luis