On Mon 05-11-18 22:32:07, Rasmus Villemoes wrote: > kstrdup_quotable_cmdline takes gfp flags and passes those on to > kstrdup_quotable, but before that it has done a kmalloc(PAGE_SIZE) with > a hard-coded GFP_KERNEL. There is one caller of kstrdup_quotable_cmdline > which passes GFP_ATOMIC, and the commit introducing that (65a3c2748e) > conveniently has this piece of history: > > v2: Use GFP_ATOMIC while holding the rcu lock per Chris Wilson > > So, should the GFP_KERNEL in kstrdup_quotable_cmdline simply be changed > to use the passed-in gfp, or is there some deeper reason for the > GFP_KERNEL (in which case it doesn't really make sense to take gfp at > all...)? I would just drop the gfp argument and move comm = kstrdup(task->comm, GFP_ATOMIC); before rcu read lock The code in its current form is buggy. -- Michal Hocko SUSE Labs
