David, I am finally working on cachefs and have hit my first bug: About a quarter of the way into "cachefs_fill_super", a cachefs object is allocated (using "kmem_cache_alloc(cachefs_object_jar,...)") that will correspond to the root of the cachefs filesystem. When "cachefs_object_jar" was created, its constructor was set to "cache_object_init_once()". That routine calls "fscache_object_init()" to initialize the fscache object part of the cachefs object being constructed. Here is the problem: The first thing "fscache_object_init()" does is an atomic inc of the object_count of the fscache_cache struct passed (by pointer) as the third argument of "fscache_object_init()". :::BUT::: when "fscache_object_init()" is called from "cache_object_init_once()", the third argument is NULL, causing an "unable to handle kernel NULL pointer dereference" Oops. I don't see that there is any valid fscache_cache struct here because the fscache_object is part of a cachefs_object and is not contained in a fscache_cache. So my inclination is to modify "fscache_object_init()" to handle the case where it's fscache_cache argument is NULL. Unless there is something in the fscache-api to handle a case like this. I will look into that, but I thought I would check with you to make sure I was headed in the right direction. What do you think? -=# Paul Gilliam #=- -- Linux-cachefs mailing list Linux-cachefs@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cachefs
