On Wed, 20 Nov 2019, Pete Zaitcev wrote: > On Wed, 20 Nov 2019 11:14:05 -0500 (EST) > Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote: > > > As it happens, I spent a little time investigating this bug report just > > yesterday. It seems to me that the easiest fix would be to disallow > > resizing the buffer while it is mapped by any users. (Besides, > > allowing that seems like a bad idea in any case.) > > > > Pete, does that seem reasonable to you? > > Actually, please hold on a little, I think to think more about this. > The deadlock is between mon_bin_read and mon_bin_vma_fault. > To disallow resizing isn't going to fix _that_. As I understand it (and my understanding is pretty limited, since I only started to look seriously at the code one day ago), the reason why mon_bin_vma_fault acquires fetch_lock is to prevent a resize from happening while the fault is being handled. Is there another reason? If you disallow resizing while the buffer is mapped then mon_bin_vma_fault won't need to hold fetch_lock at all. That would fix the deadlock, right? Alan Stern
