>>> On 25.04.19 at 15:33, <ta...@tklengyel.com> wrote: > On Thu, Apr 25, 2019 at 5:14 AM Jan Beulich <jbeul...@suse.com> wrote: >> >> >>> On 15.04.19 at 19:51, <ta...@tklengyel.com> wrote: >> > Simply changing to get_page with dom_cow is not enough, I ran into >> > issues where the system still deadlocks due to the ordering of >> > put_page_and_type being called before mem_sharing_page_unlock. So the >> > ordering still has to be changed. >> >> Dead locks because of the ordering of releases (or locks or alike), >> not acquires? That looks suspicious to me, i.e. suggests yet >> another problem elsewhere. > > FWIW page_lock has this warning: > > * We must never call _put_page_type() while holding a page_lock() for > * that page; doing so may cause a deadlock under the right > * conditions.
Oh, of course. I once again forgot that mem_sharing_page_unlock() calls page_unlock(). > That said, I'm working on getting rid of using page_lock altogether > because it just over-complicates things for no good reason. Should be > able to send the new patches today. Oh, that's good news. Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xenproject.org https://lists.xenproject.org/mailman/listinfo/xen-devel
