kernel.git/drivers/xen/events, branch linux-4.1.y

xen/events: Don't move disabled irqs

2016-06-06T23:12:34Z

[ Upstream commit f0f393877c71ad227d36705d61d1e4062bc29cf5 ] Commit ff1e22e7a638 ("xen/events: Mask a moving irq") open-coded irq_move_irq() but left out checking if the IRQ is disabled. This broke resuming from suspend since it tries to move a (disabled) irq without holding the IRQ's desc->lock. Fix it by adding in a check for disabled IRQs. The resulting stacktrace was: kernel BUG at /build/linux-UbQGH5/linux-4.4.0/kernel/irq/migration.c:31! invalid opcode: 0000 [#1] SMP Modules linked in: xenfs xen_privcmd ... CPU: 0 PID: 9 Comm: migration/0 Not tainted 4.4.0-22-generic #39-Ubuntu Hardware name: Xen HVM domU, BIOS 4.6.1-xs125180 05/04/2016 task: ffff88003d75ee00 ti: ffff88003d7bc000 task.ti: ffff88003d7bc000 RIP: 0010:[] [] irq_move_masked_irq+0xd2/0xe0 RSP: 0018:ffff88003d7bfc50 EFLAGS: 00010046 RAX: 0000000000000000 RBX: ffff88003d40ba00 RCX: 0000000000000001 RDX: 0000000000000001 RSI: 0000000000000100 RDI: ffff88003d40bad8 RBP: ffff88003d7bfc68 R08: 0000000000000000 R09: ffff88003d000000 R10: 0000000000000000 R11: 000000000000023c R12: ffff88003d40bad0 R13: ffffffff81f3a4a0 R14: 0000000000000010 R15: 00000000ffffffff FS: 0000000000000000(0000) GS:ffff88003da00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007fd4264de624 CR3: 0000000037922000 CR4: 00000000003406f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Stack: ffff88003d40ba38 0000000000000024 0000000000000000 ffff88003d7bfca0 ffffffff814c8d92 00000010813ef89d 00000000805ea732 0000000000000009 0000000000000024 ffff88003cc39b80 ffff88003d7bfce0 ffffffff814c8f66 Call Trace: [] eoi_pirq+0xb2/0xf0 [] __startup_pirq+0xe6/0x150 [] xen_irq_resume+0x319/0x360 [] xen_suspend+0xb5/0x180 [] multi_cpu_stop+0xb5/0xe0 [] ? cpu_stop_queue_work+0x80/0x80 [] cpu_stopper_thread+0xb0/0x140 [] ? finish_task_switch+0x76/0x220 [] ? __raw_callee_save___pv_queued_spin_unlock+0x11/0x20 [] smpboot_thread_fn+0x105/0x160 [] ? sort_range+0x30/0x30 [] kthread+0xd8/0xf0 [] ? kthread_create_on_node+0x1e0/0x1e0 [] ret_from_fork+0x3f/0x70 [] ? kthread_create_on_node+0x1e0/0x1e0 Signed-off-by: Ross Lagerwall Reviewed-by: Boris Ostrovsky Cc: Signed-off-by: David Vrabel Signed-off-by: Sasha Levin

xen/events: Mask a moving irq

2016-04-20T05:03:44Z

[ Upstream commit ff1e22e7a638a0782f54f81a6c9cb139aca2da35 ] Moving an unmasked irq may result in irq handler being invoked on both source and target CPUs. With 2-level this can happen as follows: On source CPU: evtchn_2l_handle_events() -> generic_handle_irq() -> handle_edge_irq() -> eoi_pirq(): irq_move_irq(data); /***** WE ARE HERE *****/ if (VALID_EVTCHN(evtchn)) clear_evtchn(evtchn); If at this moment target processor is handling an unrelated event in evtchn_2l_handle_events()'s loop it may pick up our event since target's cpu_evtchn_mask claims that this event belongs to it *and* the event is unmasked and still pending. At the same time, source CPU will continue executing its own handle_edge_irq(). With FIFO interrupt the scenario is similar: irq_move_irq() may result in a EVTCHNOP_unmask hypercall which, in turn, may make the event pending on the target CPU. We can avoid this situation by moving and clearing the event while keeping event masked. Signed-off-by: Boris Ostrovsky Cc: stable@vger.kernel.org Signed-off-by: David Vrabel Signed-off-by: Sasha Levin

xen/events: don't bind non-percpu VIRQs with percpu chip

2015-05-19T18:55:36Z

A non-percpu VIRQ (e.g., VIRQ_CONSOLE) may be freed on a different VCPU than it is bound to. This can result in a race between handle_percpu_irq() and removing the action in __free_irq() because handle_percpu_irq() does not take desc->lock. The interrupt handler sees a NULL action and oopses. Only use the percpu chip/handler for per-CPU VIRQs (like VIRQ_TIMER). # cat /proc/interrupts | grep virq 40: 87246 0 xen-percpu-virq timer0 44: 0 0 xen-percpu-virq debug0 47: 0 20995 xen-percpu-virq timer1 51: 0 0 xen-percpu-virq debug1 69: 0 0 xen-dyn-virq xen-pcpu 74: 0 0 xen-dyn-virq mce 75: 29 0 xen-dyn-virq hvc_console Signed-off-by: David Vrabel Cc:

xen/events: Set irq_info->evtchn before binding the channel to CPU in __startup_pirq()

2015-05-05T17:27:14Z

.. because bind_evtchn_to_cpu(evtchn, cpu) will map evtchn to 'info' and pass 'info' down to xen_evtchn_port_bind_to_cpu(). Signed-off-by: Boris Ostrovsky Tested-by: Annie Li Cc: Signed-off-by: David Vrabel

xen/events: Clear cpu_evtchn_mask before resuming

2015-05-05T17:27:11Z

When a guest is resumed, the hypervisor may change event channel assignments. If this happens and the guest uses 2-level events it is possible for the interrupt to be claimed by wrong VCPU since cpu_evtchn_mask bits may be stale. This can happen even though evtchn_2l_bind_to_cpu() attempts to clear old bits: irq_info that is passed in is not necessarily the original one (from pre-migration times) but instead is freshly allocated during resume and so any information about which CPU the channel was bound to is lost. Thus we should clear the mask during resume. We also need to make sure that bits for xenstore and console channels are set when these two subsystems are resumed. While rebind_evtchn_irq() (which is invoked for both of them on a resume) calls irq_set_affinity(), the latter will in fact postpone setting affinity until handling the interrupt. But because cpu_evtchn_mask will have bits for these two cleared we won't be able to take the interrupt. With that in mind, we need to bind those two channels explicitly in rebind_evtchn_irq(). We will keep irq_set_affinity() so that we have a pass through generic irq affinity code later, in case something needs to be updated there as well. (Also replace cpumask_of(0) with cpumask_of(info->cpu) in rebind_evtchn_irq(): it should be set to zero in preceding xen_irq_info_evtchn_setup().) Signed-off-by: Boris Ostrovsky Reported-by: Annie Li Cc: # 3.14+ Signed-off-by: David Vrabel

xen/events: avoid NULL pointer dereference in dom0 on large machines

2015-03-06T13:16:32Z

Using the pvops kernel a NULL pointer dereference was detected on a large machine (144 processors) when booting as dom0 in evtchn_fifo_unmask() during assignment of a pirq. The event channel in question was the first to need a new entry in event_array[] in events_fifo.c. Unfortunately xen_irq_info_pirq_setup() is called with evtchn being 0 for a new pirq and the real event channel number is assigned to the pirq only during __startup_pirq(). It is mandatory to call xen_evtchn_port_setup() after assigning the event channel number to the pirq to make sure all memory needed for the event channel is allocated. Signed-off-by: Juergen Gross Cc: # 3.14+ Signed-off-by: David Vrabel

xen/events: support threaded irqs for interdomain event channels

2014-09-23T13:36:19Z

Export bind_interdomain_evtchn_to_irq() so drivers can use threaded interrupt handlers with: irq = bind_interdomain_evtchn_to_irq(remote_dom, remote_port); if (irq < 0) /* error */ ret = request_threaded_irq(...); Signed-off-by: Juergen Gross Acked-by: Konrad Rzeszutek Wilk Signed-off-by: David Vrabel

xen/events: drop negativity check of unsigned parameter

2014-08-01T14:44:45Z

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=80531 Reported-by: David Binderman Signed-off-by: Andrey Utkin Signed-off-by: David Vrabel

xen/events/fifo: remove a unecessary use of BM()

2014-07-31T16:59:18Z

Since 05a812ac474d0d6aef6d54b66bb08b81abde79c6 (xen/events/fifo: correctly align bitops), ready is an unsigned long instead of uint32_t and the BM() macro is no longer required. Signed-off-by: Frediano Ziglio Signed-off-by: David Vrabel Reviewed-by: Boris Ostrovsky

xen/events/fifo: ensure all bitops are properly aligned even on x86

2014-07-31T16:58:38Z

When using the FIFO-based ABI on x86_64, if the last port is at the end of an event array page then sync_test_bit() on this port's event word will read beyond the end of the page and in certain circumstances this may fault. The fault requires the following page in the kernel's direct mapping to be not present, which would mean: a) the array page is the last page of RAM; or b) the following page is ballooned out /and/ it has been used for a foreign mapping by a kernel driver (such as netback or blkback) /and/ the grant has been unmapped. Use the infrastructure added for arm64 to ensure that all bitops operating on event words are unsigned long aligned. Signed-off-by: David Vrabel Reviewed-by: Boris Ostrovsky Cc: stable@vger.kernel.org