wireguard-openbsd/sys/uvm, branch master WireGuard implementation for the OpenBSD kernel https://git.zx2c4.com/wireguard-openbsd/atom/sys/uvm?h=master 2021-03-31T08:53:39Z Introduce UAO_USES_SWHASH() and use tabs instead of spaces in #defines. 2021-03-31T08:53:39Z mpi mpi@openbsd.org 2021-03-31T08:53:39Z urn:sha1:39c73ac77e4aa3b464eba41d5bf747e695f91b02 No functionnal change, reduce the difference with NetBSD. ok jmatthew@ Remove parenthesis around return value to reduce the diff with NetBSD. 2021-03-26T13:40:05Z mpi mpi@openbsd.org 2021-03-26T13:40:05Z urn:sha1:b9df1565ef82604eba98afb01045f088eb20cf68 No functional change. ok mlarkin@ Sync some comments in order to reduce the difference with NetBSD. 2021-03-20T10:24:21Z mpi mpi@openbsd.org 2021-03-20T10:24:21Z urn:sha1:52887a3857412054fd51c1af58c9c8fe944e10f7 No functionnal change. ok kettenis@ spelling 2021-03-12T14:15:49Z jsg jsg@openbsd.org 2021-03-12T14:15:49Z urn:sha1:4af3577ff281f32bc2b40d899fa25046271de1d3 ok mpi@ ansi 2021-03-05T05:30:44Z jsg jsg@openbsd.org 2021-03-05T05:30:44Z urn:sha1:fbaef5a95e6253a38e8c205cadf0514f8c268842 Modify `uvmexp.swpgonly' atomically, required for uvm_fault() w/o KERNEL_LOCK() 2021-03-04T09:00:03Z mpi mpi@openbsd.org 2021-03-04T09:00:03Z urn:sha1:c4a864ba38a189859dcf0af4cf9fb01aabd6a92c ok kettenis@ Bring back previous fix for UVM vnode deadlock. 2021-03-04T08:38:48Z mpi mpi@openbsd.org 2021-03-04T08:38:48Z urn:sha1:00e7696ef29d639c47992c239d3900a5b4cbb811 tb@ reports that refaulting when there's contention on the vnode makes firefox start very slowly on his machine. To revisit when the fault handler will be unlocked. ok anton@ Original commit message: Fix a deadlock between uvn_io() and uvn_flush(). While faulting on a page backed by a vnode, uvn_io() will end up being called in order to populate newly allocated pages using I/O on the backing vnode. Before performing the I/O, newly allocated pages are flagged as busy by uvn_get(), that is before uvn_io() tries to lock the vnode. Such pages could then end up being flushed by uvn_flush() which already has acquired the vnode lock. Since such pages are flagged as busy, uvn_flush() will wait for them to be flagged as not busy. This will never happens as uvn_io() cannot make progress until the vnode lock is released. Instead, grab the vnode lock before allocating and flagging pages as busy in uvn_get(). This does extend the scope in uvn_get() in which the vnode is locked but resolves the deadlock. ok mpi@ Reported-by: syzbot+e63407b35dff08dbee02@syzkaller.appspotmail.com Fix the deadlock between uvn_io() and uvn_flush() by restarting the fault. 2021-03-02T10:12:37Z mpi mpi@openbsd.org 2021-03-02T10:12:37Z urn:sha1:3426345d211be9d90bb8991ff6daa953ea3b752a Do not allow a faulting thread to sleep on a contended vnode lock to prevent lock ordering issues with upcoming per-uobj lock. ok anton@ Reported-by: syzbot+e63407b35dff08dbee02@syzkaller.appspotmail.com Revert the fix for the deadlock between uvn_io() and uvn_flush(). 2021-03-02T10:09:20Z mpi mpi@openbsd.org 2021-03-02T10:09:20Z urn:sha1:4deeff89dc578f0b4221cd4a79a8e3693afe5c5b This fix (ab)use the vnode lock to serialize access to some fields of the corresponding pages associated with UVM vnode object and this will create new deadlocks with the introduction of a per-uobj lock. ok anton@ If an anon is associated with a page, acquire its lock before any modification. 2021-03-01T09:13:33Z mpi mpi@openbsd.org 2021-03-01T09:13:33Z urn:sha1:1c92e3af03b7e2f4288e77e83c65213588485773 This change should have been part of the previous anon-locking diff and is necessary to run the top part of uvm_fault() unlocked. ok jmatthew@